OpenCloudOS-Kernel/tools/objtool/elf.c

1354 lines
29 KiB
C
Raw Normal View History

treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 13 Based on 2 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not see http www gnu org licenses this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details [based] [from] [clk] [highbank] [c] you should have received a copy of the gnu general public license along with this program if not see http www gnu org licenses extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 355 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Reviewed-by: Jilayne Lovejoy <opensource@jilayne.com> Reviewed-by: Steve Winslow <swinslow@gmail.com> Reviewed-by: Allison Randal <allison@lohutok.net> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190519154041.837383322@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2019-05-19 21:51:43 +08:00
// SPDX-License-Identifier: GPL-2.0-or-later
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
/*
* elf.c - ELF access library
*
* Adapted from kpatch (https://github.com/dynup/kpatch):
* Copyright (C) 2013-2015 Josh Poimboeuf <jpoimboe@redhat.com>
* Copyright (C) 2014 Seth Jennings <sjenning@redhat.com>
*/
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/mman.h>
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <errno.h>
#include <linux/interval_tree_generic.h>
objtool: Rework header include paths Currently objtool headers are being included either by their base name or included via ../ from a parent directory. In case of a base name usage: #include "warn.h" #include "arch_elf.h" it does not make it apparent from which directory the file comes from. To make it slightly better, and actually to avoid name clashes some arch specific files have "arch_" suffix. And files from an arch folder have to revert to including via ../ e.g: #include "../../elf.h" With additional architectures support and the code base growth there is a need for clearer headers naming scheme for multiple reasons: 1. to make it instantly obvious where these files come from (objtool itself / objtool arch|generic folders / some other external files), 2. to avoid name clashes of objtool arch specific headers, potential obtool arch generic headers and the system header files (there is /usr/include/elf.h already), 3. to avoid ../ includes and improve code readability. 4. to give a warm fuzzy feeling to developers who are mostly kernel developers and are accustomed to linux kernel headers arranging scheme. Doesn't this make it instantly obvious where are these files come from? #include <objtool/warn.h> #include <arch/elf.h> And doesn't it look nicer to avoid ugly ../ includes? Which also guarantees this is elf.h from the objtool and not /usr/include/elf.h. #include <objtool/elf.h> This patch defines and implements new objtool headers arranging scheme. Which is: - all generic headers go to include/objtool (similar to include/linux) - all arch headers go to arch/$(SRCARCH)/include/arch (to get arch prefix). This is similar to linux arch specific "asm/*" headers but we are not abusing "asm" name and calling it what it is. This also helps to prevent name clashes (arch is not used in system headers or kernel exports). To bring objtool to this state the following things are done: 1. current top level tools/objtool/ headers are moved into include/objtool/ subdirectory, 2. arch specific headers, currently only arch/x86/include/ are moved into arch/x86/include/arch/ and were stripped of "arch_" suffix, 3. new -I$(srctree)/tools/objtool/include include path to make includes like <objtool/warn.h> possible, 4. rewriting file includes, 5. make git not to ignore include/objtool/ subdirectory. Signed-off-by: Vasily Gorbik <gor@linux.ibm.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Masami Hiramatsu <mhiramat@kernel.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-11-13 07:03:32 +08:00
#include <objtool/builtin.h>
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
objtool: Rework header include paths Currently objtool headers are being included either by their base name or included via ../ from a parent directory. In case of a base name usage: #include "warn.h" #include "arch_elf.h" it does not make it apparent from which directory the file comes from. To make it slightly better, and actually to avoid name clashes some arch specific files have "arch_" suffix. And files from an arch folder have to revert to including via ../ e.g: #include "../../elf.h" With additional architectures support and the code base growth there is a need for clearer headers naming scheme for multiple reasons: 1. to make it instantly obvious where these files come from (objtool itself / objtool arch|generic folders / some other external files), 2. to avoid name clashes of objtool arch specific headers, potential obtool arch generic headers and the system header files (there is /usr/include/elf.h already), 3. to avoid ../ includes and improve code readability. 4. to give a warm fuzzy feeling to developers who are mostly kernel developers and are accustomed to linux kernel headers arranging scheme. Doesn't this make it instantly obvious where are these files come from? #include <objtool/warn.h> #include <arch/elf.h> And doesn't it look nicer to avoid ugly ../ includes? Which also guarantees this is elf.h from the objtool and not /usr/include/elf.h. #include <objtool/elf.h> This patch defines and implements new objtool headers arranging scheme. Which is: - all generic headers go to include/objtool (similar to include/linux) - all arch headers go to arch/$(SRCARCH)/include/arch (to get arch prefix). This is similar to linux arch specific "asm/*" headers but we are not abusing "asm" name and calling it what it is. This also helps to prevent name clashes (arch is not used in system headers or kernel exports). To bring objtool to this state the following things are done: 1. current top level tools/objtool/ headers are moved into include/objtool/ subdirectory, 2. arch specific headers, currently only arch/x86/include/ are moved into arch/x86/include/arch/ and were stripped of "arch_" suffix, 3. new -I$(srctree)/tools/objtool/include include path to make includes like <objtool/warn.h> possible, 4. rewriting file includes, 5. make git not to ignore include/objtool/ subdirectory. Signed-off-by: Vasily Gorbik <gor@linux.ibm.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Masami Hiramatsu <mhiramat@kernel.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-11-13 07:03:32 +08:00
#include <objtool/elf.h>
#include <objtool/warn.h>
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
#define MAX_NAME_LEN 128
static inline u32 str_hash(const char *str)
{
return jhash(str, strlen(str), 0);
}
#define __elf_table(name) (elf->name##_hash)
#define __elf_bits(name) (elf->name##_bits)
#define __elf_table_entry(name, key) \
__elf_table(name)[hash_min(key, __elf_bits(name))]
#define elf_hash_add(name, node, key) \
({ \
struct elf_hash_node *__node = node; \
__node->next = __elf_table_entry(name, key); \
__elf_table_entry(name, key) = __node; \
})
static inline void __elf_hash_del(struct elf_hash_node *node,
struct elf_hash_node **head)
{
struct elf_hash_node *cur, *prev;
if (node == *head) {
*head = node->next;
return;
}
for (prev = NULL, cur = *head; cur; prev = cur, cur = cur->next) {
if (cur == node) {
prev->next = cur->next;
break;
}
}
}
#define elf_hash_del(name, node, key) \
__elf_hash_del(node, &__elf_table_entry(name, key))
#define elf_list_entry(ptr, type, member) \
({ \
typeof(ptr) __ptr = (ptr); \
__ptr ? container_of(__ptr, type, member) : NULL; \
})
#define elf_hash_for_each_possible(name, obj, member, key) \
for (obj = elf_list_entry(__elf_table_entry(name, key), typeof(*obj), member); \
obj; \
obj = elf_list_entry(obj->member.next, typeof(*(obj)), member))
#define elf_alloc_hash(name, size) \
({ \
__elf_bits(name) = max(10, ilog2(size)); \
__elf_table(name) = mmap(NULL, sizeof(struct elf_hash_node *) << __elf_bits(name), \
PROT_READ|PROT_WRITE, \
MAP_PRIVATE|MAP_ANON, -1, 0); \
if (__elf_table(name) == (void *)-1L) { \
WARN("mmap fail " #name); \
__elf_table(name) = NULL; \
} \
__elf_table(name); \
})
static inline unsigned long __sym_start(struct symbol *s)
{
return s->offset;
}
static inline unsigned long __sym_last(struct symbol *s)
{
return s->offset + s->len - 1;
}
INTERVAL_TREE_DEFINE(struct symbol, node, unsigned long, __subtree_last,
__sym_start, __sym_last, static, __sym)
#define __sym_for_each(_iter, _tree, _start, _end) \
for (_iter = __sym_iter_first((_tree), (_start), (_end)); \
_iter; _iter = __sym_iter_next(_iter, (_start), (_end)))
struct symbol_hole {
unsigned long key;
const struct symbol *sym;
};
/*
* Find !section symbol where @offset is after it.
*/
static int symbol_hole_by_offset(const void *key, const struct rb_node *node)
{
const struct symbol *s = rb_entry(node, struct symbol, node);
struct symbol_hole *sh = (void *)key;
if (sh->key < s->offset)
return -1;
if (sh->key >= s->offset + s->len) {
if (s->type != STT_SECTION)
sh->sym = s;
return 1;
}
return 0;
}
struct section *find_section_by_name(const struct elf *elf, const char *name)
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
{
struct section *sec;
elf_hash_for_each_possible(section_name, sec, name_hash, str_hash(name)) {
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
if (!strcmp(sec->name, name))
return sec;
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return NULL;
}
static struct section *find_section_by_index(struct elf *elf,
unsigned int idx)
{
struct section *sec;
elf_hash_for_each_possible(section, sec, hash, idx) {
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
if (sec->idx == idx)
return sec;
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return NULL;
}
static struct symbol *find_symbol_by_index(struct elf *elf, unsigned int idx)
{
struct symbol *sym;
elf_hash_for_each_possible(symbol, sym, hash, idx) {
if (sym->idx == idx)
return sym;
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return NULL;
}
struct symbol *find_symbol_by_offset(struct section *sec, unsigned long offset)
{
struct rb_root_cached *tree = (struct rb_root_cached *)&sec->symbol_tree;
struct symbol *iter;
__sym_for_each(iter, tree, offset, offset) {
if (iter->offset == offset && iter->type != STT_SECTION)
return iter;
}
return NULL;
}
struct symbol *find_func_by_offset(struct section *sec, unsigned long offset)
{
struct rb_root_cached *tree = (struct rb_root_cached *)&sec->symbol_tree;
struct symbol *iter;
__sym_for_each(iter, tree, offset, offset) {
if (iter->offset == offset && iter->type == STT_FUNC)
return iter;
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return NULL;
}
struct symbol *find_symbol_containing(const struct section *sec, unsigned long offset)
{
struct rb_root_cached *tree = (struct rb_root_cached *)&sec->symbol_tree;
struct symbol *iter;
__sym_for_each(iter, tree, offset, offset) {
if (iter->type != STT_SECTION)
return iter;
}
return NULL;
}
/*
* Returns size of hole starting at @offset.
*/
int find_symbol_hole_containing(const struct section *sec, unsigned long offset)
{
struct symbol_hole hole = {
.key = offset,
.sym = NULL,
};
struct rb_node *n;
struct symbol *s;
/*
* Find the rightmost symbol for which @offset is after it.
*/
n = rb_find(&hole, &sec->symbol_tree.rb_root, symbol_hole_by_offset);
/* found a symbol that contains @offset */
if (n)
return 0; /* not a hole */
/* didn't find a symbol for which @offset is after it */
if (!hole.sym)
return 0; /* not a hole */
/* @offset >= sym->offset + sym->len, find symbol after it */
n = rb_next(&hole.sym->node);
if (!n)
return -1; /* until end of address space */
/* hole until start of next symbol */
s = rb_entry(n, struct symbol, node);
return s->offset - offset;
}
struct symbol *find_func_containing(struct section *sec, unsigned long offset)
{
struct rb_root_cached *tree = (struct rb_root_cached *)&sec->symbol_tree;
struct symbol *iter;
__sym_for_each(iter, tree, offset, offset) {
if (iter->type == STT_FUNC)
return iter;
}
return NULL;
}
struct symbol *find_symbol_by_name(const struct elf *elf, const char *name)
{
struct symbol *sym;
elf_hash_for_each_possible(symbol_name, sym, name_hash, str_hash(name)) {
if (!strcmp(sym->name, name))
return sym;
}
return NULL;
}
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
struct reloc *find_reloc_by_dest_range(const struct elf *elf, struct section *sec,
unsigned long offset, unsigned int len)
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
{
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
struct reloc *reloc, *r = NULL;
struct section *rsec;
unsigned long o;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
rsec = sec->rsec;
if (!rsec)
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return NULL;
for_offset_range(o, offset, offset + len) {
elf_hash_for_each_possible(reloc, reloc, hash,
sec_offset_hash(rsec, o)) {
if (reloc->sec != rsec)
continue;
if (reloc_offset(reloc) >= offset &&
reloc_offset(reloc) < offset + len) {
if (!r || reloc_offset(reloc) < reloc_offset(r))
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
r = reloc;
}
}
if (r)
return r;
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return NULL;
}
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
struct reloc *find_reloc_by_dest(const struct elf *elf, struct section *sec, unsigned long offset)
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
{
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
return find_reloc_by_dest_range(elf, sec, offset, 1);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
}
static bool is_dwarf_section(struct section *sec)
{
return !strncmp(sec->name, ".debug_", 7);
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
static int read_sections(struct elf *elf)
{
Elf_Scn *s = NULL;
struct section *sec;
size_t shstrndx, sections_nr;
int i;
if (elf_getshdrnum(elf->elf, &sections_nr)) {
WARN_ELF("elf_getshdrnum");
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return -1;
}
if (elf_getshdrstrndx(elf->elf, &shstrndx)) {
WARN_ELF("elf_getshdrstrndx");
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return -1;
}
if (!elf_alloc_hash(section, sections_nr) ||
!elf_alloc_hash(section_name, sections_nr))
return -1;
elf->section_data = calloc(sections_nr, sizeof(*sec));
if (!elf->section_data) {
perror("calloc");
return -1;
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
for (i = 0; i < sections_nr; i++) {
sec = &elf->section_data[i];
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
INIT_LIST_HEAD(&sec->symbol_list);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
s = elf_getscn(elf->elf, i);
if (!s) {
WARN_ELF("elf_getscn");
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return -1;
}
sec->idx = elf_ndxscn(s);
if (!gelf_getshdr(s, &sec->sh)) {
WARN_ELF("gelf_getshdr");
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return -1;
}
sec->name = elf_strptr(elf->elf, shstrndx, sec->sh.sh_name);
if (!sec->name) {
WARN_ELF("elf_strptr");
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return -1;
}
if (sec->sh.sh_size != 0 && !is_dwarf_section(sec)) {
sec->data = elf_getdata(s, NULL);
if (!sec->data) {
WARN_ELF("elf_getdata");
return -1;
}
if (sec->data->d_off != 0 ||
sec->data->d_size != sec->sh.sh_size) {
WARN("unexpected data attributes for %s",
sec->name);
return -1;
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
}
list_add_tail(&sec->list, &elf->sections);
elf_hash_add(section, &sec->hash, sec->idx);
elf_hash_add(section_name, &sec->name_hash, str_hash(sec->name));
if (is_reloc_sec(sec))
elf->num_relocs += sec_num_entries(sec);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
}
if (opts.stats) {
printf("nr_sections: %lu\n", (unsigned long)sections_nr);
printf("section_bits: %d\n", elf->section_bits);
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
/* sanity check, one more call to elf_nextscn() should return NULL */
if (elf_nextscn(elf->elf, s)) {
WARN("section entry mismatch");
return -1;
}
return 0;
}
static void elf_add_symbol(struct elf *elf, struct symbol *sym)
{
struct list_head *entry;
struct rb_node *pnode;
struct symbol *iter;
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
INIT_LIST_HEAD(&sym->pv_target);
sym->alias = sym;
sym->type = GELF_ST_TYPE(sym->sym.st_info);
sym->bind = GELF_ST_BIND(sym->sym.st_info);
if (sym->type == STT_FILE)
elf->num_files++;
sym->offset = sym->sym.st_value;
sym->len = sym->sym.st_size;
__sym_for_each(iter, &sym->sec->symbol_tree, sym->offset, sym->offset) {
if (iter->offset == sym->offset && iter->type == sym->type)
iter->alias = sym;
}
__sym_insert(sym, &sym->sec->symbol_tree);
pnode = rb_prev(&sym->node);
if (pnode)
entry = &rb_entry(pnode, struct symbol, node)->list;
else
entry = &sym->sec->symbol_list;
list_add(&sym->list, entry);
elf_hash_add(symbol, &sym->hash, sym->idx);
elf_hash_add(symbol_name, &sym->name_hash, str_hash(sym->name));
/*
* Don't store empty STT_NOTYPE symbols in the rbtree. They
* can exist within a function, confusing the sorting.
*/
if (!sym->len)
__sym_remove(sym, &sym->sec->symbol_tree);
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
static int read_symbols(struct elf *elf)
{
struct section *symtab, *symtab_shndx, *sec;
struct symbol *sym, *pfunc;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
int symbols_nr, i;
char *coldstr;
Elf_Data *shndx_data = NULL;
Elf32_Word shndx;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
symtab = find_section_by_name(elf, ".symtab");
if (symtab) {
symtab_shndx = find_section_by_name(elf, ".symtab_shndx");
if (symtab_shndx)
shndx_data = symtab_shndx->data;
symbols_nr = sec_num_entries(symtab);
} else {
/*
* A missing symbol table is actually possible if it's an empty
* .o file. This can happen for thunk_64.o. Make sure to at
* least allocate the symbol hash tables so we can do symbol
* lookups without crashing.
*/
symbols_nr = 0;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
}
if (!elf_alloc_hash(symbol, symbols_nr) ||
!elf_alloc_hash(symbol_name, symbols_nr))
return -1;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
elf->symbol_data = calloc(symbols_nr, sizeof(*sym));
if (!elf->symbol_data) {
perror("calloc");
return -1;
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
for (i = 0; i < symbols_nr; i++) {
sym = &elf->symbol_data[i];
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
sym->idx = i;
if (!gelf_getsymshndx(symtab->data, shndx_data, i, &sym->sym,
&shndx)) {
WARN_ELF("gelf_getsymshndx");
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
goto err;
}
sym->name = elf_strptr(elf->elf, symtab->sh.sh_link,
sym->sym.st_name);
if (!sym->name) {
WARN_ELF("elf_strptr");
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
goto err;
}
if ((sym->sym.st_shndx > SHN_UNDEF &&
sym->sym.st_shndx < SHN_LORESERVE) ||
(shndx_data && sym->sym.st_shndx == SHN_XINDEX)) {
if (sym->sym.st_shndx != SHN_XINDEX)
shndx = sym->sym.st_shndx;
sym->sec = find_section_by_index(elf, shndx);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
if (!sym->sec) {
WARN("couldn't find section for symbol %s",
sym->name);
goto err;
}
if (GELF_ST_TYPE(sym->sym.st_info) == STT_SECTION) {
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
sym->name = sym->sec->name;
sym->sec->sym = sym;
}
} else
sym->sec = find_section_by_index(elf, 0);
elf_add_symbol(elf, sym);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
}
if (opts.stats) {
printf("nr_symbols: %lu\n", (unsigned long)symbols_nr);
printf("symbol_bits: %d\n", elf->symbol_bits);
}
/* Create parent/child links for any cold subfunctions */
list_for_each_entry(sec, &elf->sections, list) {
sec_for_each_sym(sec, sym) {
char pname[MAX_NAME_LEN + 1];
size_t pnamelen;
if (sym->type != STT_FUNC)
continue;
if (sym->pfunc == NULL)
sym->pfunc = sym;
if (sym->cfunc == NULL)
sym->cfunc = sym;
coldstr = strstr(sym->name, ".cold");
objtool: Support GCC 8 '-fnoreorder-functions' Since the following commit: cd77849a69cf ("objtool: Fix GCC 8 cold subfunction detection for aliased functions") ... if the kernel is built with EXTRA_CFLAGS='-fno-reorder-functions', objtool can get stuck in an infinite loop. That flag causes the new GCC 8 cold subfunctions to be placed in .text instead of .text.unlikely. But it also has an unfortunate quirk: in the symbol table, the subfunction (e.g., nmi_panic.cold.7) is nested inside the parent (nmi_panic). That function overlap confuses objtool, and causes it to get into an infinite loop in next_insn_same_func(). Here's Allan's description of the loop: "Objtool iterates through the instructions in nmi_panic using next_insn_same_func. Once it reaches the end of nmi_panic at 0x534 it jumps to 0x528 as that's the start of nmi_panic.cold.7. However, since the instructions starting at 0x528 are still associated with nmi_panic objtool will get stuck in a loop, continually jumping back to 0x528 after reaching 0x534." Fix it by shortening the length of the parent function so that the functions no longer overlap. Reported-and-analyzed-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Allan Xavier <allan.x.xavier@oracle.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/9e704c52bee651129b036be14feda317ae5606ae.1530136978.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-06-28 06:03:45 +08:00
if (!coldstr)
continue;
pnamelen = coldstr - sym->name;
if (pnamelen > MAX_NAME_LEN) {
WARN("%s(): parent function name exceeds maximum length of %d characters",
sym->name, MAX_NAME_LEN);
return -1;
}
strncpy(pname, sym->name, pnamelen);
pname[pnamelen] = '\0';
pfunc = find_symbol_by_name(elf, pname);
objtool: Support GCC 8 '-fnoreorder-functions' Since the following commit: cd77849a69cf ("objtool: Fix GCC 8 cold subfunction detection for aliased functions") ... if the kernel is built with EXTRA_CFLAGS='-fno-reorder-functions', objtool can get stuck in an infinite loop. That flag causes the new GCC 8 cold subfunctions to be placed in .text instead of .text.unlikely. But it also has an unfortunate quirk: in the symbol table, the subfunction (e.g., nmi_panic.cold.7) is nested inside the parent (nmi_panic). That function overlap confuses objtool, and causes it to get into an infinite loop in next_insn_same_func(). Here's Allan's description of the loop: "Objtool iterates through the instructions in nmi_panic using next_insn_same_func. Once it reaches the end of nmi_panic at 0x534 it jumps to 0x528 as that's the start of nmi_panic.cold.7. However, since the instructions starting at 0x528 are still associated with nmi_panic objtool will get stuck in a loop, continually jumping back to 0x528 after reaching 0x534." Fix it by shortening the length of the parent function so that the functions no longer overlap. Reported-and-analyzed-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Allan Xavier <allan.x.xavier@oracle.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/9e704c52bee651129b036be14feda317ae5606ae.1530136978.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-06-28 06:03:45 +08:00
if (!pfunc) {
WARN("%s(): can't find parent function",
sym->name);
return -1;
objtool: Support GCC 8 '-fnoreorder-functions' Since the following commit: cd77849a69cf ("objtool: Fix GCC 8 cold subfunction detection for aliased functions") ... if the kernel is built with EXTRA_CFLAGS='-fno-reorder-functions', objtool can get stuck in an infinite loop. That flag causes the new GCC 8 cold subfunctions to be placed in .text instead of .text.unlikely. But it also has an unfortunate quirk: in the symbol table, the subfunction (e.g., nmi_panic.cold.7) is nested inside the parent (nmi_panic). That function overlap confuses objtool, and causes it to get into an infinite loop in next_insn_same_func(). Here's Allan's description of the loop: "Objtool iterates through the instructions in nmi_panic using next_insn_same_func. Once it reaches the end of nmi_panic at 0x534 it jumps to 0x528 as that's the start of nmi_panic.cold.7. However, since the instructions starting at 0x528 are still associated with nmi_panic objtool will get stuck in a loop, continually jumping back to 0x528 after reaching 0x534." Fix it by shortening the length of the parent function so that the functions no longer overlap. Reported-and-analyzed-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Allan Xavier <allan.x.xavier@oracle.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/9e704c52bee651129b036be14feda317ae5606ae.1530136978.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-06-28 06:03:45 +08:00
}
sym->pfunc = pfunc;
pfunc->cfunc = sym;
/*
* Unfortunately, -fnoreorder-functions puts the child
* inside the parent. Remove the overlap so we can
* have sane assumptions.
*
* Note that pfunc->len now no longer matches
* pfunc->sym.st_size.
*/
if (sym->sec == pfunc->sec &&
sym->offset >= pfunc->offset &&
sym->offset + sym->len == pfunc->offset + pfunc->len) {
pfunc->len -= sym->len;
}
}
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return 0;
err:
free(sym);
return -1;
}
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
/*
* @sym's idx has changed. Update the relocs which reference it.
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
*/
static int elf_update_sym_relocs(struct elf *elf, struct symbol *sym)
{
struct reloc *reloc;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
for (reloc = sym->relocs; reloc; reloc = reloc->sym_next_reloc)
set_reloc_sym(elf, reloc, reloc->sym->idx);
return 0;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
}
/*
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
* The libelf API is terrible; gelf_update_sym*() takes a data block relative
* index value, *NOT* the symbol index. As such, iterate the data blocks and
* adjust index until it fits.
*
* If no data block is found, allow adding a new data block provided the index
* is only one past the end.
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
*/
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
static int elf_update_symbol(struct elf *elf, struct section *symtab,
struct section *symtab_shndx, struct symbol *sym)
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
{
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
Elf32_Word shndx = sym->sec ? sym->sec->idx : SHN_UNDEF;
Elf_Data *symtab_data = NULL, *shndx_data = NULL;
Elf64_Xword entsize = symtab->sh.sh_entsize;
int max_idx, idx = sym->idx;
Elf_Scn *s, *t = NULL;
bool is_special_shndx = sym->sym.st_shndx >= SHN_LORESERVE &&
sym->sym.st_shndx != SHN_XINDEX;
if (is_special_shndx)
shndx = sym->sym.st_shndx;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
s = elf_getscn(elf->elf, symtab->idx);
if (!s) {
WARN_ELF("elf_getscn");
return -1;
}
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
if (symtab_shndx) {
t = elf_getscn(elf->elf, symtab_shndx->idx);
if (!t) {
WARN_ELF("elf_getscn");
return -1;
}
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
}
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
for (;;) {
/* get next data descriptor for the relevant sections */
symtab_data = elf_getdata(s, symtab_data);
if (t)
shndx_data = elf_getdata(t, shndx_data);
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
/* end-of-list */
if (!symtab_data) {
/*
* Over-allocate to avoid O(n^2) symbol creation
* behaviour. The down side is that libelf doesn't
* like this; see elf_truncate_section() for the fixup.
*/
int num = max(1U, sym->idx/3);
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
void *buf;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
if (idx) {
/* we don't do holes in symbol tables */
WARN("index out of range");
return -1;
}
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
/* if @idx == 0, it's the next contiguous entry, create it */
symtab_data = elf_newdata(s);
if (t)
shndx_data = elf_newdata(t);
buf = calloc(num, entsize);
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
if (!buf) {
WARN("malloc");
return -1;
}
symtab_data->d_buf = buf;
symtab_data->d_size = num * entsize;
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
symtab_data->d_align = 1;
symtab_data->d_type = ELF_T_SYM;
mark_sec_changed(elf, symtab, true);
symtab->truncate = true;
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
if (t) {
buf = calloc(num, sizeof(Elf32_Word));
if (!buf) {
WARN("malloc");
return -1;
}
shndx_data->d_buf = buf;
shndx_data->d_size = num * sizeof(Elf32_Word);
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
shndx_data->d_align = sizeof(Elf32_Word);
shndx_data->d_type = ELF_T_WORD;
mark_sec_changed(elf, symtab_shndx, true);
symtab_shndx->truncate = true;
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
}
break;
}
/* empty blocks should not happen */
if (!symtab_data->d_size) {
WARN("zero size data");
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
return -1;
}
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
/* is this the right block? */
max_idx = symtab_data->d_size / entsize;
if (idx < max_idx)
break;
/* adjust index and try again */
idx -= max_idx;
}
/* something went side-ways */
if (idx < 0) {
WARN("negative index");
return -1;
}
/* setup extended section index magic and write the symbol */
if ((shndx >= SHN_UNDEF && shndx < SHN_LORESERVE) || is_special_shndx) {
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
sym->sym.st_shndx = shndx;
if (!shndx_data)
shndx = 0;
} else {
sym->sym.st_shndx = SHN_XINDEX;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
if (!shndx_data) {
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
WARN("no .symtab_shndx");
return -1;
}
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
}
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
if (!gelf_update_symshndx(symtab_data, shndx_data, idx, &sym->sym, shndx)) {
WARN_ELF("gelf_update_symshndx");
return -1;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
}
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
return 0;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
}
static struct symbol *
__elf_create_symbol(struct elf *elf, struct symbol *sym)
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
{
struct section *symtab, *symtab_shndx;
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
Elf32_Word first_non_local, new_idx;
struct symbol *old;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
symtab = find_section_by_name(elf, ".symtab");
if (symtab) {
symtab_shndx = find_section_by_name(elf, ".symtab_shndx");
} else {
WARN("no .symtab");
return NULL;
}
new_idx = sec_num_entries(symtab);
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
if (GELF_ST_BIND(sym->sym.st_info) != STB_LOCAL)
goto non_local;
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
/*
* Move the first global symbol, as per sh_info, into a new, higher
* symbol index. This fees up a spot for a new local symbol.
*/
first_non_local = symtab->sh.sh_info;
old = find_symbol_by_index(elf, first_non_local);
if (old) {
elf_hash_del(symbol, &old->hash, old->idx);
elf_hash_add(symbol, &old->hash, new_idx);
old->idx = new_idx;
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
if (elf_update_symbol(elf, symtab, symtab_shndx, old)) {
WARN("elf_update_symbol move");
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
return NULL;
}
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
if (elf_update_sym_relocs(elf, old))
return NULL;
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
new_idx = first_non_local;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
}
/*
* Either way, we will add a LOCAL symbol.
*/
symtab->sh.sh_info += 1;
non_local:
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
sym->idx = new_idx;
if (elf_update_symbol(elf, symtab, symtab_shndx, sym)) {
WARN("elf_update_symbol");
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
return NULL;
}
symtab->sh.sh_size += symtab->sh.sh_entsize;
mark_sec_changed(elf, symtab, true);
if (symtab_shndx) {
symtab_shndx->sh.sh_size += sizeof(Elf32_Word);
mark_sec_changed(elf, symtab_shndx, true);
}
return sym;
}
static struct symbol *
elf_create_section_symbol(struct elf *elf, struct section *sec)
{
struct symbol *sym = calloc(1, sizeof(*sym));
if (!sym) {
perror("malloc");
return NULL;
}
objtool: Fix symbol creation Nathan reported objtool failing with the following messages: warning: objtool: no non-local symbols !? warning: objtool: gelf_update_symshndx: invalid section index The problem is due to commit 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") failing to consider the case where an object would have no non-local symbols. The problem that commit tries to address is adding a STB_LOCAL symbol to the symbol table in light of the ELF spec's requirement that: In each symbol table, all symbols with STB_LOCAL binding preced the weak and global symbols. As ``Sections'' above describes, a symbol table section's sh_info section header member holds the symbol table index for the first non-local symbol. The approach taken is to find this first non-local symbol, move that to the end and then re-use the freed spot to insert a new local symbol and increment sh_info. Except it never considered the case of object files without global symbols and got a whole bunch of details wrong -- so many in fact that it is a wonder it ever worked :/ Specifically: - It failed to re-hash the symbol on the new index, so a subsequent find_symbol_by_index() would not find it at the new location and a query for the old location would now return a non-deterministic choice between the old and new symbol. - It failed to appreciate that the GElf wrappers are not a valid disk format (it works because GElf is basically Elf64 and we only support x86_64 atm.) - It failed to fully appreciate how horrible the libelf API really is and got the gelf_update_symshndx() call pretty much completely wrong; with the direct consequence that if inserting a second STB_LOCAL symbol would require moving the same STB_GLOBAL symbol again it would completely come unstuck. Write a new elf_update_symbol() function that wraps all the magic required to update or create a new symbol at a given index. Specifically, gelf_update_sym*() require an @ndx argument that is relative to the @data argument; this means you have to manually iterate the section data descriptor list and update @ndx. Fixes: 4abff6d48dbc ("objtool: Fix code relocs vs weak symbols") Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/YoPCTEYjoPqE4ZxB@hirez.programming.kicks-ass.net
2022-05-17 23:42:04 +08:00
sym->name = sec->name;
sym->sec = sec;
// st_name 0
sym->sym.st_info = GELF_ST_INFO(STB_LOCAL, STT_SECTION);
// st_other 0
// st_value 0
// st_size 0
sym = __elf_create_symbol(elf, sym);
if (sym)
elf_add_symbol(elf, sym);
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
return sym;
}
static int elf_add_string(struct elf *elf, struct section *strtab, char *str);
struct symbol *
elf_create_prefix_symbol(struct elf *elf, struct symbol *orig, long size)
{
struct symbol *sym = calloc(1, sizeof(*sym));
size_t namelen = strlen(orig->name) + sizeof("__pfx_");
char *name = malloc(namelen);
if (!sym || !name) {
perror("malloc");
return NULL;
}
snprintf(name, namelen, "__pfx_%s", orig->name);
sym->name = name;
sym->sec = orig->sec;
sym->sym.st_name = elf_add_string(elf, NULL, name);
sym->sym.st_info = orig->sym.st_info;
sym->sym.st_value = orig->sym.st_value - size;
sym->sym.st_size = size;
sym = __elf_create_symbol(elf, sym);
if (sym)
elf_add_symbol(elf, sym);
return sym;
}
static struct reloc *elf_init_reloc(struct elf *elf, struct section *rsec,
unsigned int reloc_idx,
unsigned long offset, struct symbol *sym,
s64 addend, unsigned int type)
{
struct reloc *reloc, empty = { 0 };
if (reloc_idx >= sec_num_entries(rsec)) {
WARN("%s: bad reloc_idx %u for %s with %d relocs",
__func__, reloc_idx, rsec->name, sec_num_entries(rsec));
return NULL;
}
reloc = &rsec->relocs[reloc_idx];
if (memcmp(reloc, &empty, sizeof(empty))) {
WARN("%s: %s: reloc %d already initialized!",
__func__, rsec->name, reloc_idx);
return NULL;
}
reloc->sec = rsec;
reloc->sym = sym;
set_reloc_offset(elf, reloc, offset);
set_reloc_sym(elf, reloc, sym->idx);
set_reloc_type(elf, reloc, type);
set_reloc_addend(elf, reloc, addend);
elf_hash_add(reloc, &reloc->hash, reloc_hash(reloc));
reloc->sym_next_reloc = sym->relocs;
sym->relocs = reloc;
return reloc;
}
struct reloc *elf_init_reloc_text_sym(struct elf *elf, struct section *sec,
unsigned long offset,
unsigned int reloc_idx,
struct section *insn_sec,
unsigned long insn_off)
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
{
struct symbol *sym = insn_sec->sym;
int addend = insn_off;
if (!(insn_sec->sh.sh_flags & SHF_EXECINSTR)) {
WARN("bad call to %s() for data symbol %s",
__func__, sym->name);
return NULL;
}
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
if (!sym) {
/*
* Due to how weak functions work, we must use section based
* relocations. Symbol based relocations would result in the
* weak and non-weak function annotations being overlaid on the
* non-weak function after linking.
*/
sym = elf_create_section_symbol(elf, insn_sec);
if (!sym)
return NULL;
objtool: Fix code relocs vs weak symbols Occasionally objtool driven code patching (think .static_call_sites .retpoline_sites etc..) goes sideways and it tries to patch an instruction that doesn't match. Much head-scatching and cursing later the problem is as outlined below and affects every section that objtool generates for us, very much including the ORC data. The below uses .static_call_sites because it's convenient for demonstration purposes, but as mentioned the ORC sections, .retpoline_sites and __mount_loc are all similarly affected. Consider: foo-weak.c: extern void __SCT__foo(void); __attribute__((weak)) void foo(void) { return __SCT__foo(); } foo.c: extern void __SCT__foo(void); extern void my_foo(void); void foo(void) { my_foo(); return __SCT__foo(); } These generate the obvious code (gcc -O2 -fcf-protection=none -fno-asynchronous-unwind-tables -c foo*.c): foo-weak.o: 0000000000000000 <foo>: 0: e9 00 00 00 00 jmpq 5 <foo+0x5> 1: R_X86_64_PLT32 __SCT__foo-0x4 foo.o: 0000000000000000 <foo>: 0: 48 83 ec 08 sub $0x8,%rsp 4: e8 00 00 00 00 callq 9 <foo+0x9> 5: R_X86_64_PLT32 my_foo-0x4 9: 48 83 c4 08 add $0x8,%rsp d: e9 00 00 00 00 jmpq 12 <foo+0x12> e: R_X86_64_PLT32 __SCT__foo-0x4 Now, when we link these two files together, you get something like (ld -r -o foos.o foo-weak.o foo.o): foos.o: 0000000000000000 <foo-0x10>: 0: e9 00 00 00 00 jmpq 5 <foo-0xb> 1: R_X86_64_PLT32 __SCT__foo-0x4 5: 66 2e 0f 1f 84 00 00 00 00 00 nopw %cs:0x0(%rax,%rax,1) f: 90 nop 0000000000000010 <foo>: 10: 48 83 ec 08 sub $0x8,%rsp 14: e8 00 00 00 00 callq 19 <foo+0x9> 15: R_X86_64_PLT32 my_foo-0x4 19: 48 83 c4 08 add $0x8,%rsp 1d: e9 00 00 00 00 jmpq 22 <foo+0x12> 1e: R_X86_64_PLT32 __SCT__foo-0x4 Noting that ld preserves the weak function text, but strips the symbol off of it (hence objdump doing that funny negative offset thing). This does lead to 'interesting' unused code issues with objtool when ran on linked objects, but that seems to be working (fingers crossed). So far so good.. Now lets consider the objtool static_call output section (readelf output, old binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 .text + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 .text + 1d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 So we have two patch sites, one in the dead code of the weak foo and one in the real foo. All is well. *HOWEVER*, when the toolchain strips unused section symbols it generates things like this (using new enough binutils): foo-weak.o: Relocation section '.rela.static_call_sites' at offset 0x2c8 contains 1 entry: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foo.o: Relocation section '.rela.static_call_sites' at offset 0x310 contains 2 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000200000002 R_X86_64_PC32 0000000000000000 foo + d 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 foos.o: Relocation section '.rela.static_call_sites' at offset 0x430 contains 4 entries: Offset Info Type Symbol's Value Symbol's Name + Addend 0000000000000000 0000000100000002 R_X86_64_PC32 0000000000000000 foo + 0 0000000000000004 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 0000000000000008 0000000100000002 R_X86_64_PC32 0000000000000000 foo + d 000000000000000c 0000000d00000002 R_X86_64_PC32 0000000000000000 __SCT__foo + 1 And now we can see how that foos.o .static_call_sites goes side-ways, we now have _two_ patch sites in foo. One for the weak symbol at foo+0 (which is no longer a static_call site!) and one at foo+d which is in fact the right location. This seems to happen when objtool cannot find a section symbol, in which case it falls back to any other symbol to key off of, however in this case that goes terribly wrong! As such, teach objtool to create a section symbol when there isn't one. Fixes: 44f6a7c0755d ("objtool: Fix seg fault with Clang non-section symbols") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Link: https://lkml.kernel.org/r/20220419203807.655552918@infradead.org
2022-04-17 23:03:36 +08:00
insn_sec->sym = sym;
}
return elf_init_reloc(elf, sec->rsec, reloc_idx, offset, sym, addend,
elf_text_rela_type(elf));
}
struct reloc *elf_init_reloc_data_sym(struct elf *elf, struct section *sec,
unsigned long offset,
unsigned int reloc_idx,
struct symbol *sym,
s64 addend)
{
if (sym->sec && (sec->sh.sh_flags & SHF_EXECINSTR)) {
WARN("bad call to %s() for text symbol %s",
__func__, sym->name);
return NULL;
}
return elf_init_reloc(elf, sec->rsec, reloc_idx, offset, sym, addend,
elf_data_rela_type(elf));
}
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
static int read_relocs(struct elf *elf)
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
{
unsigned long nr_reloc, max_reloc = 0;
struct section *rsec;
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
struct reloc *reloc;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
unsigned int symndx;
struct symbol *sym;
int i;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
if (!elf_alloc_hash(reloc, elf->num_relocs))
return -1;
list_for_each_entry(rsec, &elf->sections, list) {
if (!is_reloc_sec(rsec))
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
continue;
rsec->base = find_section_by_index(elf, rsec->sh.sh_info);
if (!rsec->base) {
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
WARN("can't find base section for reloc section %s",
rsec->name);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return -1;
}
rsec->base->rsec = rsec;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
nr_reloc = 0;
rsec->relocs = calloc(sec_num_entries(rsec), sizeof(*reloc));
if (!rsec->relocs) {
perror("calloc");
return -1;
}
for (i = 0; i < sec_num_entries(rsec); i++) {
reloc = &rsec->relocs[i];
reloc->sec = rsec;
symndx = reloc_sym(reloc);
reloc->sym = sym = find_symbol_by_index(elf, symndx);
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
if (!reloc->sym) {
WARN("can't find reloc entry symbol %d for %s",
symndx, rsec->name);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
return -1;
}
elf_hash_add(reloc, &reloc->hash, reloc_hash(reloc));
reloc->sym_next_reloc = sym->relocs;
sym->relocs = reloc;
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
nr_reloc++;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
}
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
max_reloc = max(max_reloc, nr_reloc);
}
if (opts.stats) {
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
printf("max_reloc: %lu\n", max_reloc);
printf("num_relocs: %lu\n", elf->num_relocs);
printf("reloc_bits: %d\n", elf->reloc_bits);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
}
return 0;
}
struct elf *elf_open_read(const char *name, int flags)
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
{
struct elf *elf;
Elf_Cmd cmd;
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
elf_version(EV_CURRENT);
elf = malloc(sizeof(*elf));
if (!elf) {
perror("malloc");
return NULL;
}
memset(elf, 0, sizeof(*elf));
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
INIT_LIST_HEAD(&elf->sections);
elf->fd = open(name, flags);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
if (elf->fd == -1) {
fprintf(stderr, "objtool: Can't open '%s': %s\n",
name, strerror(errno));
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
goto err;
}
if ((flags & O_ACCMODE) == O_RDONLY)
cmd = ELF_C_READ_MMAP;
else if ((flags & O_ACCMODE) == O_RDWR)
cmd = ELF_C_RDWR;
else /* O_WRONLY */
cmd = ELF_C_WRITE;
elf->elf = elf_begin(elf->fd, cmd, NULL);
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
if (!elf->elf) {
WARN_ELF("elf_begin");
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
goto err;
}
if (!gelf_getehdr(elf->elf, &elf->ehdr)) {
WARN_ELF("gelf_getehdr");
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
goto err;
}
if (read_sections(elf))
goto err;
if (read_symbols(elf))
goto err;
objtool: Rename rela to reloc Before supporting additional relocation types rename the relevant types and functions from "rela" to "reloc". This work be done with the following regex: sed -e 's/struct rela/struct reloc/g' \ -e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \ -e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \ -e 's/relasec/relocsec/g' \ -e 's/rela_list/reloc_list/g' \ -e 's/rela_hash/reloc_hash/g' \ -e 's/add_rela/add_reloc/g' \ -e 's/rela->/reloc->/g' \ -e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \ -e 's/rela =/reloc =/g' \ -e 's/relas =/relocs =/g' \ -e 's/relas\[/relocs[/g' \ -e 's/relaname =/relocname =/g' \ -e 's/= rela\;/= reloc\;/g' \ -e 's/= relas\;/= relocs\;/g' \ -e 's/= relaname\;/= relocname\;/g' \ -e 's/, rela)/, reloc)/g' \ -e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \ -e 's/ rela$/ reloc/g' \ -e 's/, relaname/, relocname/g' \ -e 's/sec->rela/sec->reloc/g' \ -e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \ -i \ arch.h \ arch/x86/decode.c \ check.c \ check.h \ elf.c \ elf.h \ orc_gen.c \ special.c Notable exceptions which complicate the regex include gelf_* library calls and standard/expected section names which still use "rela" because they encode the type of relocation expected. Also, keep "rela" in the struct because it encodes a specific type of relocation we currently expect. It will eventually turn into a member of an anonymous union when a susequent patch adds implicit addend, or "rel", relocation support. Signed-off-by: Matt Helsley <mhelsley@vmware.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
2020-05-30 05:01:13 +08:00
if (read_relocs(elf))
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
goto err;
return elf;
err:
elf_close(elf);
return NULL;
}
static int elf_add_string(struct elf *elf, struct section *strtab, char *str)
{
Elf_Data *data;
Elf_Scn *s;
int len;
if (!strtab)
strtab = find_section_by_name(elf, ".strtab");
if (!strtab) {
WARN("can't find .strtab section");
return -1;
}
s = elf_getscn(elf->elf, strtab->idx);
if (!s) {
WARN_ELF("elf_getscn");
return -1;
}
data = elf_newdata(s);
if (!data) {
WARN_ELF("elf_newdata");
return -1;
}
data->d_buf = str;
data->d_size = strlen(str) + 1;
data->d_align = 1;
len = strtab->sh.sh_size;
strtab->sh.sh_size += data->d_size;
mark_sec_changed(elf, strtab, true);
return len;
}
struct section *elf_create_section(struct elf *elf, const char *name,
size_t entsize, unsigned int nr)
{
struct section *sec, *shstrtab;
size_t size = entsize * nr;
Elf_Scn *s;
sec = malloc(sizeof(*sec));
if (!sec) {
perror("malloc");
return NULL;
}
memset(sec, 0, sizeof(*sec));
INIT_LIST_HEAD(&sec->symbol_list);
s = elf_newscn(elf->elf);
if (!s) {
WARN_ELF("elf_newscn");
return NULL;
}
sec->name = strdup(name);
if (!sec->name) {
perror("strdup");
return NULL;
}
sec->idx = elf_ndxscn(s);
sec->data = elf_newdata(s);
if (!sec->data) {
WARN_ELF("elf_newdata");
return NULL;
}
sec->data->d_size = size;
sec->data->d_align = 1;
if (size) {
sec->data->d_buf = malloc(size);
if (!sec->data->d_buf) {
perror("malloc");
return NULL;
}
memset(sec->data->d_buf, 0, size);
}
if (!gelf_getshdr(s, &sec->sh)) {
WARN_ELF("gelf_getshdr");
return NULL;
}
sec->sh.sh_size = size;
sec->sh.sh_entsize = entsize;
sec->sh.sh_type = SHT_PROGBITS;
sec->sh.sh_addralign = 1;
sec->sh.sh_flags = SHF_ALLOC;
/* Add section name to .shstrtab (or .strtab for Clang) */
shstrtab = find_section_by_name(elf, ".shstrtab");
if (!shstrtab)
shstrtab = find_section_by_name(elf, ".strtab");
if (!shstrtab) {
WARN("can't find .shstrtab or .strtab section");
return NULL;
}
sec->sh.sh_name = elf_add_string(elf, shstrtab, sec->name);
if (sec->sh.sh_name == -1)
return NULL;
list_add_tail(&sec->list, &elf->sections);
elf_hash_add(section, &sec->hash, sec->idx);
elf_hash_add(section_name, &sec->name_hash, str_hash(sec->name));
mark_sec_changed(elf, sec, true);
return sec;
}
static struct section *elf_create_rela_section(struct elf *elf,
struct section *sec,
unsigned int reloc_nr)
{
struct section *rsec;
char *rsec_name;
rsec_name = malloc(strlen(sec->name) + strlen(".rela") + 1);
if (!rsec_name) {
perror("malloc");
return NULL;
}
strcpy(rsec_name, ".rela");
strcat(rsec_name, sec->name);
rsec = elf_create_section(elf, rsec_name, elf_rela_size(elf), reloc_nr);
free(rsec_name);
if (!rsec)
return NULL;
rsec->data->d_type = ELF_T_RELA;
rsec->sh.sh_type = SHT_RELA;
rsec->sh.sh_addralign = elf_addr_size(elf);
rsec->sh.sh_link = find_section_by_name(elf, ".symtab")->idx;
rsec->sh.sh_info = sec->idx;
rsec->sh.sh_flags = SHF_INFO_LINK;
rsec->relocs = calloc(sec_num_entries(rsec), sizeof(struct reloc));
if (!rsec->relocs) {
perror("calloc");
return NULL;
}
sec->rsec = rsec;
rsec->base = sec;
return rsec;
}
struct section *elf_create_section_pair(struct elf *elf, const char *name,
size_t entsize, unsigned int nr,
unsigned int reloc_nr)
{
struct section *sec;
sec = elf_create_section(elf, name, entsize, nr);
if (!sec)
return NULL;
if (!elf_create_rela_section(elf, sec, reloc_nr))
return NULL;
return sec;
}
int elf_write_insn(struct elf *elf, struct section *sec,
unsigned long offset, unsigned int len,
const char *insn)
{
Elf_Data *data = sec->data;
if (data->d_type != ELF_T_BYTE || data->d_off) {
WARN("write to unexpected data for section: %s", sec->name);
return -1;
}
memcpy(data->d_buf + offset, insn, len);
mark_sec_changed(elf, sec, true);
return 0;
}
/*
* When Elf_Scn::sh_size is smaller than the combined Elf_Data::d_size
* do you:
*
* A) adhere to the section header and truncate the data, or
* B) ignore the section header and write out all the data you've got?
*
* Yes, libelf sucks and we need to manually truncate if we over-allocate data.
*/
static int elf_truncate_section(struct elf *elf, struct section *sec)
{
u64 size = sec->sh.sh_size;
bool truncated = false;
Elf_Data *data = NULL;
Elf_Scn *s;
s = elf_getscn(elf->elf, sec->idx);
if (!s) {
WARN_ELF("elf_getscn");
return -1;
}
for (;;) {
/* get next data descriptor for the relevant section */
data = elf_getdata(s, data);
if (!data) {
if (size) {
WARN("end of section data but non-zero size left\n");
return -1;
}
return 0;
}
if (truncated) {
/* when we remove symbols */
WARN("truncated; but more data\n");
return -1;
}
if (!data->d_size) {
WARN("zero size data");
return -1;
}
if (data->d_size > size) {
truncated = true;
data->d_size = size;
}
size -= data->d_size;
}
}
int elf_write(struct elf *elf)
{
struct section *sec;
Elf_Scn *s;
if (opts.dryrun)
return 0;
/* Update changed relocation sections and section headers: */
list_for_each_entry(sec, &elf->sections, list) {
if (sec->truncate)
elf_truncate_section(elf, sec);
if (sec_changed(sec)) {
s = elf_getscn(elf->elf, sec->idx);
if (!s) {
WARN_ELF("elf_getscn");
return -1;
}
/* Note this also flags the section dirty */
if (!gelf_update_shdr(s, &sec->sh)) {
WARN_ELF("gelf_update_shdr");
return -1;
}
mark_sec_changed(elf, sec, false);
}
}
/* Make sure the new section header entries get updated properly. */
elf_flagelf(elf->elf, ELF_C_SET, ELF_F_DIRTY);
/* Write all changes to the file. */
if (elf_update(elf->elf, ELF_C_WRITE) < 0) {
WARN_ELF("elf_update");
return -1;
}
elf->changed = false;
return 0;
}
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
void elf_close(struct elf *elf)
{
if (elf->elf)
elf_end(elf->elf);
if (elf->fd > 0)
close(elf->fd);
/*
* NOTE: All remaining allocations are leaked on purpose. Objtool is
* about to exit anyway.
*/
objtool: Add tool to perform compile-time stack metadata validation This adds a host tool named objtool which has a "check" subcommand which analyzes .o files to ensure the validity of stack metadata. It enforces a set of rules on asm code and C inline assembly code so that stack traces can be reliable. For each function, it recursively follows all possible code paths and validates the correct frame pointer state at each instruction. It also follows code paths involving kernel special sections, like .altinstructions, __jump_table, and __ex_table, which can add alternative execution paths to a given instruction (or set of instructions). Similarly, it knows how to follow switch statements, for which gcc sometimes uses jump tables. Here are some of the benefits of validating stack metadata: a) More reliable stack traces for frame pointer enabled kernels Frame pointers are used for debugging purposes. They allow runtime code and debug tools to be able to walk the stack to determine the chain of function call sites that led to the currently executing code. For some architectures, frame pointers are enabled by CONFIG_FRAME_POINTER. For some other architectures they may be required by the ABI (sometimes referred to as "backchain pointers"). For C code, gcc automatically generates instructions for setting up frame pointers when the -fno-omit-frame-pointer option is used. But for asm code, the frame setup instructions have to be written by hand, which most people don't do. So the end result is that CONFIG_FRAME_POINTER is honored for C code but not for most asm code. For stack traces based on frame pointers to be reliable, all functions which call other functions must first create a stack frame and update the frame pointer. If a first function doesn't properly create a stack frame before calling a second function, the *caller* of the first function will be skipped on the stack trace. For example, consider the following example backtrace with frame pointers enabled: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff8127f568>] seq_read+0x108/0x3e0 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 It correctly shows that the caller of cmdline_proc_show() is seq_read(). If we remove the frame pointer logic from cmdline_proc_show() by replacing the frame pointer related instructions with nops, here's what it looks like instead: [<ffffffff81812584>] dump_stack+0x4b/0x63 [<ffffffff812d6dc2>] cmdline_proc_show+0x12/0x30 [<ffffffff812cce62>] proc_reg_read+0x42/0x70 [<ffffffff81256197>] __vfs_read+0x37/0x100 [<ffffffff81256b16>] vfs_read+0x86/0x130 [<ffffffff81257898>] SyS_read+0x58/0xd0 [<ffffffff8181c1f2>] entry_SYSCALL_64_fastpath+0x12/0x76 Notice that cmdline_proc_show()'s caller, seq_read(), has been skipped. Instead the stack trace seems to show that cmdline_proc_show() was called by proc_reg_read(). The benefit of "objtool check" here is that because it ensures that *all* functions honor CONFIG_FRAME_POINTER, no functions will ever[*] be skipped on a stack trace. [*] unless an interrupt or exception has occurred at the very beginning of a function before the stack frame has been created, or at the very end of the function after the stack frame has been destroyed. This is an inherent limitation of frame pointers. b) 100% reliable stack traces for DWARF enabled kernels This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. c) Higher live patching compatibility rate This is not yet implemented. For more details about what is planned, see tools/objtool/Documentation/stack-validation.txt. To achieve the validation, "objtool check" enforces the following rules: 1. Each callable function must be annotated as such with the ELF function type. In asm code, this is typically done using the ENTRY/ENDPROC macros. If objtool finds a return instruction outside of a function, it flags an error since that usually indicates callable code which should be annotated accordingly. This rule is needed so that objtool can properly identify each callable function in order to analyze its stack metadata. 2. Conversely, each section of code which is *not* callable should *not* be annotated as an ELF function. The ENDPROC macro shouldn't be used in this case. This rule is needed so that objtool can ignore non-callable code. Such code doesn't have to follow any of the other rules. 3. Each callable function which calls another function must have the correct frame pointer logic, if required by CONFIG_FRAME_POINTER or the architecture's back chain rules. This can by done in asm code with the FRAME_BEGIN/FRAME_END macros. This rule ensures that frame pointer based stack traces will work as designed. If function A doesn't create a stack frame before calling function B, the _caller_ of function A will be skipped on the stack trace. 4. Dynamic jumps and jumps to undefined symbols are only allowed if: a) the jump is part of a switch statement; or b) the jump matches sibling call semantics and the frame pointer has the same value it had on function entry. This rule is needed so that objtool can reliably analyze all of a function's code paths. If a function jumps to code in another file, and it's not a sibling call, objtool has no way to follow the jump because it only analyzes a single file at a time. 5. A callable function may not execute kernel entry/exit instructions. The only code which needs such instructions is kernel entry code, which shouldn't be be in callable functions anyway. This rule is just a sanity check to ensure that callable functions return normally. It currently only supports x86_64. I tried to make the code generic so that support for other architectures can hopefully be plugged in relatively easily. On my Lenovo laptop with a i7-4810MQ 4-core/8-thread CPU, building the kernel with objtool checking every .o file adds about three seconds of total build time. It hasn't been optimized for performance yet, so there are probably some opportunities for better build performance. Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Bernd Petrovitsch <bernd@petrovitsch.priv.at> Cc: Borislav Petkov <bp@alien8.de> Cc: Chris J Arges <chris.j.arges@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Marek <mmarek@suse.cz> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Pedro Alves <palves@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: live-patching@vger.kernel.org Link: http://lkml.kernel.org/r/f3efb173de43bd067b060de73f856567c0fa1174.1456719558.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-02-29 12:22:41 +08:00
}