2005-04-17 06:20:36 +08:00
|
|
|
/* Kernel module help for PPC64.
|
|
|
|
Copyright (C) 2001, 2003 Rusty Russell IBM Corporation.
|
|
|
|
|
|
|
|
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, write to the Free Software
|
|
|
|
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
|
|
*/
|
2014-09-17 12:39:35 +08:00
|
|
|
|
|
|
|
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/elf.h>
|
|
|
|
#include <linux/moduleloader.h>
|
|
|
|
#include <linux/err.h>
|
|
|
|
#include <linux/vmalloc.h>
|
2008-11-15 12:47:03 +08:00
|
|
|
#include <linux/ftrace.h>
|
2006-12-08 19:30:41 +08:00
|
|
|
#include <linux/bug.h>
|
2014-04-03 17:00:43 +08:00
|
|
|
#include <linux/uaccess.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <asm/module.h>
|
2006-10-20 09:47:19 +08:00
|
|
|
#include <asm/firmware.h>
|
2008-06-24 09:32:35 +08:00
|
|
|
#include <asm/code-patching.h>
|
2007-11-14 00:24:04 +08:00
|
|
|
#include <linux/sort.h>
|
2013-10-28 22:20:51 +08:00
|
|
|
#include <asm/setup.h>
|
2016-03-03 12:26:56 +08:00
|
|
|
#include <asm/sections.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* FIXME: We don't do .init separately. To do this, we'd need to have
|
|
|
|
a separate r2 value in the init and core section, and stub between
|
|
|
|
them, too.
|
|
|
|
|
|
|
|
Using a magic allocator which places modules within 32MB solves
|
|
|
|
this, and makes other things simpler. Anton?
|
|
|
|
--RR. */
|
|
|
|
|
2016-06-07 00:56:10 +08:00
|
|
|
#ifdef PPC64_ELF_ABI_v2
|
2014-03-19 08:12:22 +08:00
|
|
|
|
|
|
|
/* An address is simply the address of the function. */
|
|
|
|
typedef unsigned long func_desc_t;
|
|
|
|
|
|
|
|
static func_desc_t func_desc(unsigned long addr)
|
|
|
|
{
|
|
|
|
return addr;
|
|
|
|
}
|
|
|
|
static unsigned long func_addr(unsigned long addr)
|
|
|
|
{
|
|
|
|
return addr;
|
|
|
|
}
|
|
|
|
static unsigned long stub_func_addr(func_desc_t func)
|
|
|
|
{
|
|
|
|
return func;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* PowerPC64 specific values for the Elf64_Sym st_other field. */
|
|
|
|
#define STO_PPC64_LOCAL_BIT 5
|
|
|
|
#define STO_PPC64_LOCAL_MASK (7 << STO_PPC64_LOCAL_BIT)
|
|
|
|
#define PPC64_LOCAL_ENTRY_OFFSET(other) \
|
|
|
|
(((1 << (((other) & STO_PPC64_LOCAL_MASK) >> STO_PPC64_LOCAL_BIT)) >> 2) << 2)
|
|
|
|
|
|
|
|
static unsigned int local_entry_offset(const Elf64_Sym *sym)
|
|
|
|
{
|
|
|
|
/* sym->st_other indicates offset to local entry point
|
|
|
|
* (otherwise it will assume r12 is the address of the start
|
|
|
|
* of function and try to derive r2 from it). */
|
|
|
|
return PPC64_LOCAL_ENTRY_OFFSET(sym->st_other);
|
|
|
|
}
|
2014-03-18 17:42:44 +08:00
|
|
|
#else
|
2014-03-19 08:12:22 +08:00
|
|
|
|
|
|
|
/* An address is address of the OPD entry, which contains address of fn. */
|
|
|
|
typedef struct ppc64_opd_entry func_desc_t;
|
|
|
|
|
|
|
|
static func_desc_t func_desc(unsigned long addr)
|
|
|
|
{
|
|
|
|
return *(struct ppc64_opd_entry *)addr;
|
|
|
|
}
|
|
|
|
static unsigned long func_addr(unsigned long addr)
|
|
|
|
{
|
|
|
|
return func_desc(addr).funcaddr;
|
|
|
|
}
|
|
|
|
static unsigned long stub_func_addr(func_desc_t func)
|
|
|
|
{
|
|
|
|
return func.funcaddr;
|
|
|
|
}
|
|
|
|
static unsigned int local_entry_offset(const Elf64_Sym *sym)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
2017-11-10 07:48:27 +08:00
|
|
|
|
|
|
|
void *dereference_module_function_descriptor(struct module *mod, void *ptr)
|
|
|
|
{
|
|
|
|
if (ptr < (void *)mod->arch.start_opd ||
|
|
|
|
ptr >= (void *)mod->arch.end_opd)
|
|
|
|
return ptr;
|
|
|
|
|
|
|
|
return dereference_function_descriptor(ptr);
|
|
|
|
}
|
2014-03-18 17:42:44 +08:00
|
|
|
#endif
|
|
|
|
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
#define STUB_MAGIC 0x73747562 /* stub */
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* Like PPC32, we need little trampolines to do > 24-bit jumps (into
|
|
|
|
the kernel itself). But on PPC64, these need to be used for every
|
|
|
|
jump, actually, to reset r2 (TOC+0x8000). */
|
|
|
|
struct ppc64_stub_entry
|
|
|
|
{
|
2014-04-03 17:00:43 +08:00
|
|
|
/* 28 byte jump instruction sequence (7 instructions). We only
|
|
|
|
* need 6 instructions on ABIv2 but we always allocate 7 so
|
|
|
|
* so we don't have to modify the trampoline load instruction. */
|
2014-03-18 15:05:28 +08:00
|
|
|
u32 jump[7];
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
/* Used by ftrace to identify stubs */
|
|
|
|
u32 magic;
|
2005-04-17 06:20:36 +08:00
|
|
|
/* Data for the above code */
|
2014-03-19 08:12:22 +08:00
|
|
|
func_desc_t funcdata;
|
2005-04-17 06:20:36 +08:00
|
|
|
};
|
|
|
|
|
2014-03-18 17:43:03 +08:00
|
|
|
/*
|
|
|
|
* PPC64 uses 24 bit jumps, but we need to jump into other modules or
|
|
|
|
* the kernel which may be further. So we jump to a stub.
|
|
|
|
*
|
|
|
|
* For ELFv1 we need to use this to set up the new r2 value (aka TOC
|
|
|
|
* pointer). For ELFv2 it's the callee's responsibility to set up the
|
|
|
|
* new r2, but for both we need to save the old r2.
|
|
|
|
*
|
|
|
|
* We could simply patch the new r2 value and function pointer into
|
|
|
|
* the stub, but it's significantly shorter to put these values at the
|
|
|
|
* end of the stub code, and patch the stub address (32-bits relative
|
|
|
|
* to the TOC ptr, r2) into the stub.
|
|
|
|
*/
|
2014-04-03 17:00:43 +08:00
|
|
|
|
|
|
|
static u32 ppc64_stub_insns[] = {
|
2014-03-18 17:42:59 +08:00
|
|
|
0x3d620000, /* addis r11,r2, <high> */
|
|
|
|
0x396b0000, /* addi r11,r11, <low> */
|
2013-09-21 02:42:20 +08:00
|
|
|
/* Save current r2 value in magic place on the stack. */
|
2014-03-18 17:42:44 +08:00
|
|
|
0xf8410000|R2_STACK_OFFSET, /* std r2,R2_STACK_OFFSET(r1) */
|
2014-03-18 17:42:59 +08:00
|
|
|
0xe98b0020, /* ld r12,32(r11) */
|
2016-06-07 00:56:10 +08:00
|
|
|
#ifdef PPC64_ELF_ABI_v1
|
2014-03-18 17:43:03 +08:00
|
|
|
/* Set up new r2 from function descriptor */
|
2014-05-12 15:45:02 +08:00
|
|
|
0xe84b0028, /* ld r2,40(r11) */
|
2014-03-18 17:43:03 +08:00
|
|
|
#endif
|
2014-03-18 17:42:59 +08:00
|
|
|
0x7d8903a6, /* mtctr r12 */
|
2014-03-18 17:42:44 +08:00
|
|
|
0x4e800420 /* bctr */
|
2014-04-03 17:00:43 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
#ifdef CONFIG_DYNAMIC_FTRACE
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
int module_trampoline_target(struct module *mod, unsigned long addr,
|
|
|
|
unsigned long *target)
|
2014-04-03 17:00:43 +08:00
|
|
|
{
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
struct ppc64_stub_entry *stub;
|
|
|
|
func_desc_t funcdata;
|
|
|
|
u32 magic;
|
2014-04-03 17:00:43 +08:00
|
|
|
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
if (!within_module_core(addr, mod)) {
|
|
|
|
pr_err("%s: stub %lx not in module %s\n", __func__, addr, mod->name);
|
2014-04-03 17:00:43 +08:00
|
|
|
return -EFAULT;
|
|
|
|
}
|
|
|
|
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
stub = (struct ppc64_stub_entry *)addr;
|
2014-04-04 12:58:42 +08:00
|
|
|
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
if (probe_kernel_read(&magic, &stub->magic, sizeof(magic))) {
|
|
|
|
pr_err("%s: fault reading magic for stub %lx for %s\n", __func__, addr, mod->name);
|
2014-04-04 12:58:42 +08:00
|
|
|
return -EFAULT;
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
}
|
2014-04-04 12:58:42 +08:00
|
|
|
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
if (magic != STUB_MAGIC) {
|
|
|
|
pr_err("%s: bad magic for stub %lx for %s\n", __func__, addr, mod->name);
|
|
|
|
return -EFAULT;
|
|
|
|
}
|
2014-04-04 12:58:42 +08:00
|
|
|
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
if (probe_kernel_read(&funcdata, &stub->funcdata, sizeof(funcdata))) {
|
|
|
|
pr_err("%s: fault reading funcdata for stub %lx for %s\n", __func__, addr, mod->name);
|
|
|
|
return -EFAULT;
|
|
|
|
}
|
2014-04-04 12:58:42 +08:00
|
|
|
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
*target = stub_func_addr(funcdata);
|
2014-04-04 12:58:42 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2014-04-03 17:00:43 +08:00
|
|
|
#endif
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* Count how many different 24-bit relocations (different symbol,
|
|
|
|
different addend) */
|
|
|
|
static unsigned int count_relocs(const Elf64_Rela *rela, unsigned int num)
|
|
|
|
{
|
2007-11-14 00:24:04 +08:00
|
|
|
unsigned int i, r_info, r_addend, _count_relocs;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* FIXME: Only count external ones --RR */
|
2007-11-14 00:24:04 +08:00
|
|
|
_count_relocs = 0;
|
|
|
|
r_info = 0;
|
|
|
|
r_addend = 0;
|
|
|
|
for (i = 0; i < num; i++)
|
2005-04-17 06:20:36 +08:00
|
|
|
/* Only count 24-bit relocs, others don't need stubs */
|
2007-11-14 00:24:04 +08:00
|
|
|
if (ELF64_R_TYPE(rela[i].r_info) == R_PPC_REL24 &&
|
|
|
|
(r_info != ELF64_R_SYM(rela[i].r_info) ||
|
|
|
|
r_addend != rela[i].r_addend)) {
|
|
|
|
_count_relocs++;
|
|
|
|
r_info = ELF64_R_SYM(rela[i].r_info);
|
|
|
|
r_addend = rela[i].r_addend;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2007-11-14 00:24:04 +08:00
|
|
|
|
|
|
|
return _count_relocs;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2007-11-14 00:24:04 +08:00
|
|
|
static int relacmp(const void *_x, const void *_y)
|
|
|
|
{
|
|
|
|
const Elf64_Rela *x, *y;
|
|
|
|
|
|
|
|
y = (Elf64_Rela *)_x;
|
|
|
|
x = (Elf64_Rela *)_y;
|
|
|
|
|
|
|
|
/* Compare the entire r_info (as opposed to ELF64_R_SYM(r_info) only) to
|
|
|
|
* make the comparison cheaper/faster. It won't affect the sorting or
|
|
|
|
* the counting algorithms' performance
|
|
|
|
*/
|
|
|
|
if (x->r_info < y->r_info)
|
|
|
|
return -1;
|
|
|
|
else if (x->r_info > y->r_info)
|
|
|
|
return 1;
|
|
|
|
else if (x->r_addend < y->r_addend)
|
|
|
|
return -1;
|
|
|
|
else if (x->r_addend > y->r_addend)
|
|
|
|
return 1;
|
|
|
|
else
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void relaswap(void *_x, void *_y, int size)
|
|
|
|
{
|
|
|
|
uint64_t *x, *y, tmp;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
y = (uint64_t *)_x;
|
|
|
|
x = (uint64_t *)_y;
|
|
|
|
|
|
|
|
for (i = 0; i < sizeof(Elf64_Rela) / sizeof(uint64_t); i++) {
|
|
|
|
tmp = x[i];
|
|
|
|
x[i] = y[i];
|
|
|
|
y[i] = tmp;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* Get size of potential trampolines required. */
|
|
|
|
static unsigned long get_stubs_size(const Elf64_Ehdr *hdr,
|
|
|
|
const Elf64_Shdr *sechdrs)
|
|
|
|
{
|
|
|
|
/* One extra reloc so it's always 0-funcaddr terminated */
|
|
|
|
unsigned long relocs = 1;
|
|
|
|
unsigned i;
|
|
|
|
|
|
|
|
/* Every relocated section... */
|
|
|
|
for (i = 1; i < hdr->e_shnum; i++) {
|
|
|
|
if (sechdrs[i].sh_type == SHT_RELA) {
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_debug("Found relocations in section %u\n", i);
|
|
|
|
pr_debug("Ptr: %p. Number: %Lu\n",
|
2005-04-17 06:20:36 +08:00
|
|
|
(void *)sechdrs[i].sh_addr,
|
|
|
|
sechdrs[i].sh_size / sizeof(Elf64_Rela));
|
2007-11-14 00:24:04 +08:00
|
|
|
|
|
|
|
/* Sort the relocation information based on a symbol and
|
|
|
|
* addend key. This is a stable O(n*log n) complexity
|
|
|
|
* alogrithm but it will reduce the complexity of
|
|
|
|
* count_relocs() to linear complexity O(n)
|
|
|
|
*/
|
|
|
|
sort((void *)sechdrs[i].sh_addr,
|
|
|
|
sechdrs[i].sh_size / sizeof(Elf64_Rela),
|
|
|
|
sizeof(Elf64_Rela), relacmp, relaswap);
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
relocs += count_relocs((void *)sechdrs[i].sh_addr,
|
|
|
|
sechdrs[i].sh_size
|
|
|
|
/ sizeof(Elf64_Rela));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-11-15 12:47:03 +08:00
|
|
|
#ifdef CONFIG_DYNAMIC_FTRACE
|
|
|
|
/* make the trampoline to the ftrace_caller */
|
|
|
|
relocs++;
|
|
|
|
#endif
|
|
|
|
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_debug("Looks like a total of %lu stubs, max\n", relocs);
|
2005-04-17 06:20:36 +08:00
|
|
|
return relocs * sizeof(struct ppc64_stub_entry);
|
|
|
|
}
|
|
|
|
|
2014-03-18 17:41:28 +08:00
|
|
|
/* Still needed for ELFv2, for .TOC. */
|
2005-04-17 06:20:36 +08:00
|
|
|
static void dedotify_versions(struct modversion_info *vers,
|
|
|
|
unsigned long size)
|
|
|
|
{
|
|
|
|
struct modversion_info *end;
|
|
|
|
|
|
|
|
for (end = (void *)vers + size; vers < end; vers++)
|
2014-06-24 16:53:59 +08:00
|
|
|
if (vers->name[0] == '.') {
|
2005-04-17 06:20:36 +08:00
|
|
|
memmove(vers->name, vers->name+1, strlen(vers->name));
|
2014-06-24 16:53:59 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2016-01-15 17:52:22 +08:00
|
|
|
/*
|
|
|
|
* Undefined symbols which refer to .funcname, hack to funcname. Make .TOC.
|
|
|
|
* seem to be defined (value set later).
|
|
|
|
*/
|
2005-04-17 06:20:36 +08:00
|
|
|
static void dedotify(Elf64_Sym *syms, unsigned int numsyms, char *strtab)
|
|
|
|
{
|
|
|
|
unsigned int i;
|
|
|
|
|
|
|
|
for (i = 1; i < numsyms; i++) {
|
|
|
|
if (syms[i].st_shndx == SHN_UNDEF) {
|
|
|
|
char *name = strtab + syms[i].st_name;
|
2016-01-15 17:52:22 +08:00
|
|
|
if (name[0] == '.') {
|
|
|
|
if (strcmp(name+1, "TOC.") == 0)
|
|
|
|
syms[i].st_shndx = SHN_ABS;
|
2016-02-06 02:50:03 +08:00
|
|
|
syms[i].st_name++;
|
2016-01-15 17:52:22 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-03-18 17:29:26 +08:00
|
|
|
static Elf64_Sym *find_dot_toc(Elf64_Shdr *sechdrs,
|
|
|
|
const char *strtab,
|
|
|
|
unsigned int symindex)
|
|
|
|
{
|
|
|
|
unsigned int i, numsyms;
|
|
|
|
Elf64_Sym *syms;
|
|
|
|
|
|
|
|
syms = (Elf64_Sym *)sechdrs[symindex].sh_addr;
|
|
|
|
numsyms = sechdrs[symindex].sh_size / sizeof(Elf64_Sym);
|
|
|
|
|
|
|
|
for (i = 1; i < numsyms; i++) {
|
2016-01-15 17:52:22 +08:00
|
|
|
if (syms[i].st_shndx == SHN_ABS
|
2014-03-19 08:12:22 +08:00
|
|
|
&& strcmp(strtab + syms[i].st_name, "TOC.") == 0)
|
2014-03-18 17:29:26 +08:00
|
|
|
return &syms[i];
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
int module_frob_arch_sections(Elf64_Ehdr *hdr,
|
|
|
|
Elf64_Shdr *sechdrs,
|
|
|
|
char *secstrings,
|
|
|
|
struct module *me)
|
|
|
|
{
|
|
|
|
unsigned int i;
|
|
|
|
|
|
|
|
/* Find .toc and .stubs sections, symtab and strtab */
|
|
|
|
for (i = 1; i < hdr->e_shnum; i++) {
|
|
|
|
char *p;
|
|
|
|
if (strcmp(secstrings + sechdrs[i].sh_name, ".stubs") == 0)
|
|
|
|
me->arch.stubs_section = i;
|
powerpc/modules: Fix alignment of .toc section in kernel modules
powerpc64 gcc can generate code that offsets an address, to access
part of an object in memory. If the address is a -mcmodel=medium toc
pointer relative address then code like the following is possible.
addis r9,r2,var@toc@ha
ld r3,var@toc@l(r9)
ld r4,(var+8)@toc@l(r9)
This works fine so long as var is naturally aligned, *and* r2 is
sufficiently aligned. If not, there is a possibility that the offset
added to access var+8 wraps over a n*64k+32k boundary. Modules don't
have any guarantee that r2 is sufficiently aligned. Moreover, code
generated by older compilers generates a .toc section with 2**0
alignment, which can result in relocation failures at module load time
even without the wrap problem.
Thus, this patch links modules with an aligned .toc section (Makefile
and module.lds changes), and forces alignment for out of tree modules
or those without a .toc section (module_64.c changes).
Signed-off-by: Alan Modra <amodra@gmail.com>
[desnesn: updated patch to apply to powerpc-next kernel v4.15 ]
Signed-off-by: Desnes A. Nunes do Rosario <desnesn@linux.vnet.ibm.com>
[mpe: Fix out-of-tree build, swap -256 for ~0xff, reflow comment]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-12-07 03:12:28 +08:00
|
|
|
else if (strcmp(secstrings + sechdrs[i].sh_name, ".toc") == 0) {
|
2005-04-17 06:20:36 +08:00
|
|
|
me->arch.toc_section = i;
|
powerpc/modules: Fix alignment of .toc section in kernel modules
powerpc64 gcc can generate code that offsets an address, to access
part of an object in memory. If the address is a -mcmodel=medium toc
pointer relative address then code like the following is possible.
addis r9,r2,var@toc@ha
ld r3,var@toc@l(r9)
ld r4,(var+8)@toc@l(r9)
This works fine so long as var is naturally aligned, *and* r2 is
sufficiently aligned. If not, there is a possibility that the offset
added to access var+8 wraps over a n*64k+32k boundary. Modules don't
have any guarantee that r2 is sufficiently aligned. Moreover, code
generated by older compilers generates a .toc section with 2**0
alignment, which can result in relocation failures at module load time
even without the wrap problem.
Thus, this patch links modules with an aligned .toc section (Makefile
and module.lds changes), and forces alignment for out of tree modules
or those without a .toc section (module_64.c changes).
Signed-off-by: Alan Modra <amodra@gmail.com>
[desnesn: updated patch to apply to powerpc-next kernel v4.15 ]
Signed-off-by: Desnes A. Nunes do Rosario <desnesn@linux.vnet.ibm.com>
[mpe: Fix out-of-tree build, swap -256 for ~0xff, reflow comment]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-12-07 03:12:28 +08:00
|
|
|
if (sechdrs[i].sh_addralign < 8)
|
|
|
|
sechdrs[i].sh_addralign = 8;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
else if (strcmp(secstrings+sechdrs[i].sh_name,"__versions")==0)
|
|
|
|
dedotify_versions((void *)hdr + sechdrs[i].sh_offset,
|
|
|
|
sechdrs[i].sh_size);
|
2017-11-10 07:48:27 +08:00
|
|
|
else if (!strcmp(secstrings + sechdrs[i].sh_name, ".opd")) {
|
|
|
|
me->arch.start_opd = sechdrs[i].sh_addr;
|
|
|
|
me->arch.end_opd = sechdrs[i].sh_addr +
|
|
|
|
sechdrs[i].sh_size;
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* We don't handle .init for the moment: rename to _init */
|
|
|
|
while ((p = strstr(secstrings + sechdrs[i].sh_name, ".init")))
|
|
|
|
p[0] = '_';
|
|
|
|
|
|
|
|
if (sechdrs[i].sh_type == SHT_SYMTAB)
|
|
|
|
dedotify((void *)hdr + sechdrs[i].sh_offset,
|
|
|
|
sechdrs[i].sh_size / sizeof(Elf64_Sym),
|
|
|
|
(void *)hdr
|
|
|
|
+ sechdrs[sechdrs[i].sh_link].sh_offset);
|
|
|
|
}
|
2006-04-24 20:41:51 +08:00
|
|
|
|
|
|
|
if (!me->arch.stubs_section) {
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_err("%s: doesn't contain .stubs.\n", me->name);
|
2005-04-17 06:20:36 +08:00
|
|
|
return -ENOEXEC;
|
|
|
|
}
|
|
|
|
|
2006-04-24 20:41:51 +08:00
|
|
|
/* If we don't have a .toc, just use .stubs. We need to set r2
|
|
|
|
to some reasonable value in case the module calls out to
|
|
|
|
other functions via a stub, or if a function pointer escapes
|
|
|
|
the module by some means. */
|
|
|
|
if (!me->arch.toc_section)
|
|
|
|
me->arch.toc_section = me->arch.stubs_section;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* Override the stubs size */
|
|
|
|
sechdrs[me->arch.stubs_section].sh_size = get_stubs_size(hdr, sechdrs);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
powerpc/modules: Fix alignment of .toc section in kernel modules
powerpc64 gcc can generate code that offsets an address, to access
part of an object in memory. If the address is a -mcmodel=medium toc
pointer relative address then code like the following is possible.
addis r9,r2,var@toc@ha
ld r3,var@toc@l(r9)
ld r4,(var+8)@toc@l(r9)
This works fine so long as var is naturally aligned, *and* r2 is
sufficiently aligned. If not, there is a possibility that the offset
added to access var+8 wraps over a n*64k+32k boundary. Modules don't
have any guarantee that r2 is sufficiently aligned. Moreover, code
generated by older compilers generates a .toc section with 2**0
alignment, which can result in relocation failures at module load time
even without the wrap problem.
Thus, this patch links modules with an aligned .toc section (Makefile
and module.lds changes), and forces alignment for out of tree modules
or those without a .toc section (module_64.c changes).
Signed-off-by: Alan Modra <amodra@gmail.com>
[desnesn: updated patch to apply to powerpc-next kernel v4.15 ]
Signed-off-by: Desnes A. Nunes do Rosario <desnesn@linux.vnet.ibm.com>
[mpe: Fix out-of-tree build, swap -256 for ~0xff, reflow comment]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-12-07 03:12:28 +08:00
|
|
|
/*
|
|
|
|
* r2 is the TOC pointer: it actually points 0x8000 into the TOC (this gives the
|
|
|
|
* value maximum span in an instruction which uses a signed offset). Round down
|
|
|
|
* to a 256 byte boundary for the odd case where we are setting up r2 without a
|
|
|
|
* .toc section.
|
|
|
|
*/
|
2016-03-03 12:26:54 +08:00
|
|
|
static inline unsigned long my_r2(const Elf64_Shdr *sechdrs, struct module *me)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
powerpc/modules: Fix alignment of .toc section in kernel modules
powerpc64 gcc can generate code that offsets an address, to access
part of an object in memory. If the address is a -mcmodel=medium toc
pointer relative address then code like the following is possible.
addis r9,r2,var@toc@ha
ld r3,var@toc@l(r9)
ld r4,(var+8)@toc@l(r9)
This works fine so long as var is naturally aligned, *and* r2 is
sufficiently aligned. If not, there is a possibility that the offset
added to access var+8 wraps over a n*64k+32k boundary. Modules don't
have any guarantee that r2 is sufficiently aligned. Moreover, code
generated by older compilers generates a .toc section with 2**0
alignment, which can result in relocation failures at module load time
even without the wrap problem.
Thus, this patch links modules with an aligned .toc section (Makefile
and module.lds changes), and forces alignment for out of tree modules
or those without a .toc section (module_64.c changes).
Signed-off-by: Alan Modra <amodra@gmail.com>
[desnesn: updated patch to apply to powerpc-next kernel v4.15 ]
Signed-off-by: Desnes A. Nunes do Rosario <desnesn@linux.vnet.ibm.com>
[mpe: Fix out-of-tree build, swap -256 for ~0xff, reflow comment]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-12-07 03:12:28 +08:00
|
|
|
return (sechdrs[me->arch.toc_section].sh_addr & ~0xfful) + 0x8000;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Both low and high 16 bits are added as SIGNED additions, so if low
|
|
|
|
16 bits has high bit set, high 16 bits must be adjusted. These
|
|
|
|
macros do that (stolen from binutils). */
|
|
|
|
#define PPC_LO(v) ((v) & 0xffff)
|
|
|
|
#define PPC_HI(v) (((v) >> 16) & 0xffff)
|
|
|
|
#define PPC_HA(v) PPC_HI ((v) + 0x8000)
|
|
|
|
|
|
|
|
/* Patch stub to reference function and correct r2 value. */
|
2016-03-03 12:26:54 +08:00
|
|
|
static inline int create_stub(const Elf64_Shdr *sechdrs,
|
2005-04-17 06:20:36 +08:00
|
|
|
struct ppc64_stub_entry *entry,
|
2014-03-19 08:12:22 +08:00
|
|
|
unsigned long addr,
|
2005-04-17 06:20:36 +08:00
|
|
|
struct module *me)
|
|
|
|
{
|
|
|
|
long reladdr;
|
|
|
|
|
2014-04-03 17:00:43 +08:00
|
|
|
memcpy(entry->jump, ppc64_stub_insns, sizeof(ppc64_stub_insns));
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* Stub uses address relative to r2. */
|
|
|
|
reladdr = (unsigned long)entry - my_r2(sechdrs, me);
|
|
|
|
if (reladdr > 0x7FFFFFFF || reladdr < -(0x80000000L)) {
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_err("%s: Address %p of stub out of range of %p.\n",
|
2005-04-17 06:20:36 +08:00
|
|
|
me->name, (void *)reladdr, (void *)my_r2);
|
|
|
|
return 0;
|
|
|
|
}
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_debug("Stub %p get data from reladdr %li\n", entry, reladdr);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2014-03-18 15:05:28 +08:00
|
|
|
entry->jump[0] |= PPC_HA(reladdr);
|
|
|
|
entry->jump[1] |= PPC_LO(reladdr);
|
2014-03-19 08:12:22 +08:00
|
|
|
entry->funcdata = func_desc(addr);
|
powerpc/module: Mark module stubs with a magic value
When a module is loaded, calls out to the kernel go via a stub which is
generated at runtime. One of these stubs is used to call _mcount(),
which is the default target of tracing calls generated by the compiler
with -pg.
If dynamic ftrace is enabled (which it typically is), another stub is
used to call ftrace_caller(), which is the target of tracing calls when
ftrace is actually active.
ftrace then wants to disable the calls to _mcount() at module startup,
and enable/disable the calls to ftrace_caller() when enabling/disabling
tracing - all of these it does by patching the code.
As part of that code patching, the ftrace code wants to confirm that the
branch it is about to modify, is in fact a call to a module stub which
calls _mcount() or ftrace_caller().
Currently it does that by inspecting the instructions and confirming
they are what it expects. Although that works, the code to do it is
pretty intricate because it requires lots of knowledge about the exact
format of the stub.
We can make that process easier by marking the generated stubs with a
magic value, and then looking for that magic value. Altough this is not
as rigorous as the current method, I believe it is sufficient in
practice.
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Reviewed-by: Torsten Duwe <duwe@suse.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-03-03 12:26:55 +08:00
|
|
|
entry->magic = STUB_MAGIC;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2014-03-19 08:12:22 +08:00
|
|
|
/* Create stub to jump to function described in this OPD/ptr: we need the
|
2005-04-17 06:20:36 +08:00
|
|
|
stub to set up the TOC ptr (r2) for the function. */
|
2016-03-03 12:26:54 +08:00
|
|
|
static unsigned long stub_for_addr(const Elf64_Shdr *sechdrs,
|
2014-03-19 08:12:22 +08:00
|
|
|
unsigned long addr,
|
2005-04-17 06:20:36 +08:00
|
|
|
struct module *me)
|
|
|
|
{
|
|
|
|
struct ppc64_stub_entry *stubs;
|
|
|
|
unsigned int i, num_stubs;
|
|
|
|
|
|
|
|
num_stubs = sechdrs[me->arch.stubs_section].sh_size / sizeof(*stubs);
|
|
|
|
|
|
|
|
/* Find this stub, or if that fails, the next avail. entry */
|
|
|
|
stubs = (void *)sechdrs[me->arch.stubs_section].sh_addr;
|
2014-03-19 08:12:22 +08:00
|
|
|
for (i = 0; stub_func_addr(stubs[i].funcdata); i++) {
|
2017-10-10 22:47:32 +08:00
|
|
|
if (WARN_ON(i >= num_stubs))
|
|
|
|
return 0;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2014-03-19 08:12:22 +08:00
|
|
|
if (stub_func_addr(stubs[i].funcdata) == func_addr(addr))
|
2005-04-17 06:20:36 +08:00
|
|
|
return (unsigned long)&stubs[i];
|
|
|
|
}
|
|
|
|
|
2014-03-19 08:12:22 +08:00
|
|
|
if (!create_stub(sechdrs, &stubs[i], addr, me))
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
return (unsigned long)&stubs[i];
|
|
|
|
}
|
|
|
|
|
2016-03-03 12:26:59 +08:00
|
|
|
#ifdef CC_USING_MPROFILE_KERNEL
|
|
|
|
static bool is_early_mcount_callsite(u32 *instruction)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Check if this is one of the -mprofile-kernel sequences.
|
|
|
|
*/
|
|
|
|
if (instruction[-1] == PPC_INST_STD_LR &&
|
|
|
|
instruction[-2] == PPC_INST_MFLR)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
if (instruction[-1] == PPC_INST_MFLR)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* In case of _mcount calls, do not save the current callee's TOC (in r2) into
|
|
|
|
* the original caller's stack frame. If we did we would clobber the saved TOC
|
|
|
|
* value of the original caller.
|
|
|
|
*/
|
|
|
|
static void squash_toc_save_inst(const char *name, unsigned long addr)
|
|
|
|
{
|
|
|
|
struct ppc64_stub_entry *stub = (struct ppc64_stub_entry *)addr;
|
|
|
|
|
|
|
|
/* Only for calls to _mcount */
|
|
|
|
if (strcmp("_mcount", name) != 0)
|
|
|
|
return;
|
|
|
|
|
|
|
|
stub->jump[2] = PPC_INST_NOP;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static void squash_toc_save_inst(const char *name, unsigned long addr) { }
|
|
|
|
|
|
|
|
/* without -mprofile-kernel, mcount calls are never early */
|
|
|
|
static bool is_early_mcount_callsite(u32 *instruction)
|
|
|
|
{
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* We expect a noop next: if it is, replace it with instruction to
|
|
|
|
restore r2. */
|
|
|
|
static int restore_r2(u32 *instruction, struct module *me)
|
|
|
|
{
|
2017-11-17 01:45:37 +08:00
|
|
|
u32 *prev_insn = instruction - 1;
|
|
|
|
|
|
|
|
if (is_early_mcount_callsite(prev_insn))
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure the branch isn't a sibling call. Sibling calls aren't
|
|
|
|
* "link" branches and they don't return, so they don't need the r2
|
|
|
|
* restore afterwards.
|
|
|
|
*/
|
|
|
|
if (!instr_is_relative_link_branch(*prev_insn))
|
powerpc/modules: Never restore r2 for a mprofile-kernel style mcount() call
In the module loader we process relocations, and for long jumps we
generate trampolines (aka stubs). At the call site for one of these
trampolines we usually need to generate a load instruction to restore
the TOC pointer into r2.
There is one exception however, which is calls to mcount() using the
mprofile-kernel ABI, they handle the TOC inside the stub, and so for
them we do not generate a TOC load.
The bug is in how the code in restore_r2() decides if it needs to
generate the TOC load. It does so by looking for a nop following the
branch, and if it sees a nop, it replaces it with the load. In general
the compiler has no reason to generate a nop following the mcount()
call and so that check works OK.
However if we combine a jump label at the start of a function, with an
early return, such that GCC applies the shrink-wrapping optimisation, we
can then end up with an mcount call followed immediately by a nop.
However the nop is not there for a TOC load, it is for the jump label.
That confuses restore_r2() into replacing the jump label nop with a TOC
load, which in turn confuses ftrace into replacing the mcount call with
a b +8 (fixed in the previous commit). The end result is we jump over
the jump label, which if it was supposed to return means we incorrectly
run the body of the function.
We have seen this in practice with some yet-to-be-merged patches that
use jump labels more extensively.
The fix is relatively simple, in restore_r2() we check for an
mprofile-kernel style mcount() call first, before looking for the
presence of a nop.
Fixes: 153086644fd1 ("powerpc/ftrace: Add support for -mprofile-kernel ftrace ABI")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-07-19 12:48:31 +08:00
|
|
|
return 1;
|
|
|
|
|
2009-02-11 04:10:44 +08:00
|
|
|
if (*instruction != PPC_INST_NOP) {
|
2017-11-14 17:29:10 +08:00
|
|
|
pr_err("%s: Expected nop after call, got %08x at %pS\n",
|
|
|
|
me->name, *instruction, instruction);
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
}
|
2014-03-18 17:42:44 +08:00
|
|
|
/* ld r2,R2_STACK_OFFSET(r1) */
|
2016-03-03 12:26:59 +08:00
|
|
|
*instruction = PPC_INST_LD_TOC;
|
2005-04-17 06:20:36 +08:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
int apply_relocate_add(Elf64_Shdr *sechdrs,
|
|
|
|
const char *strtab,
|
|
|
|
unsigned int symindex,
|
|
|
|
unsigned int relsec,
|
|
|
|
struct module *me)
|
|
|
|
{
|
|
|
|
unsigned int i;
|
|
|
|
Elf64_Rela *rela = (void *)sechdrs[relsec].sh_addr;
|
|
|
|
Elf64_Sym *sym;
|
|
|
|
unsigned long *location;
|
|
|
|
unsigned long value;
|
|
|
|
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_debug("Applying ADD relocate section %u to %u\n", relsec,
|
2005-04-17 06:20:36 +08:00
|
|
|
sechdrs[relsec].sh_info);
|
2014-03-18 17:29:26 +08:00
|
|
|
|
|
|
|
/* First time we're called, we can fix up .TOC. */
|
|
|
|
if (!me->arch.toc_fixed) {
|
|
|
|
sym = find_dot_toc(sechdrs, strtab, symindex);
|
|
|
|
/* It's theoretically possible that a module doesn't want a
|
|
|
|
* .TOC. so don't fail it just for that. */
|
|
|
|
if (sym)
|
|
|
|
sym->st_value = my_r2(sechdrs, me);
|
|
|
|
me->arch.toc_fixed = true;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
for (i = 0; i < sechdrs[relsec].sh_size / sizeof(*rela); i++) {
|
|
|
|
/* This is where to make the change */
|
|
|
|
location = (void *)sechdrs[sechdrs[relsec].sh_info].sh_addr
|
|
|
|
+ rela[i].r_offset;
|
|
|
|
/* This is the symbol it is referring to */
|
|
|
|
sym = (Elf64_Sym *)sechdrs[symindex].sh_addr
|
|
|
|
+ ELF64_R_SYM(rela[i].r_info);
|
|
|
|
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_debug("RELOC at %p: %li-type as %s (0x%lx) + %li\n",
|
2005-04-17 06:20:36 +08:00
|
|
|
location, (long)ELF64_R_TYPE(rela[i].r_info),
|
|
|
|
strtab + sym->st_name, (unsigned long)sym->st_value,
|
|
|
|
(long)rela[i].r_addend);
|
|
|
|
|
|
|
|
/* `Everything is relative'. */
|
|
|
|
value = sym->st_value + rela[i].r_addend;
|
|
|
|
|
|
|
|
switch (ELF64_R_TYPE(rela[i].r_info)) {
|
|
|
|
case R_PPC64_ADDR32:
|
|
|
|
/* Simply set it */
|
|
|
|
*(u32 *)location = value;
|
|
|
|
break;
|
2007-11-14 00:24:04 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
case R_PPC64_ADDR64:
|
|
|
|
/* Simply set it */
|
|
|
|
*(unsigned long *)location = value;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case R_PPC64_TOC:
|
|
|
|
*(unsigned long *)location = my_r2(sechdrs, me);
|
|
|
|
break;
|
|
|
|
|
2005-10-12 00:28:24 +08:00
|
|
|
case R_PPC64_TOC16:
|
2006-04-24 20:41:51 +08:00
|
|
|
/* Subtract TOC pointer */
|
2005-10-12 00:28:24 +08:00
|
|
|
value -= my_r2(sechdrs, me);
|
|
|
|
if (value + 0x8000 > 0xffff) {
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_err("%s: bad TOC16 relocation (0x%lx)\n",
|
2005-10-12 00:28:24 +08:00
|
|
|
me->name, value);
|
|
|
|
return -ENOEXEC;
|
|
|
|
}
|
|
|
|
*((uint16_t *) location)
|
|
|
|
= (*((uint16_t *) location) & ~0xffff)
|
|
|
|
| (value & 0xffff);
|
|
|
|
break;
|
|
|
|
|
2012-11-27 01:41:08 +08:00
|
|
|
case R_PPC64_TOC16_LO:
|
|
|
|
/* Subtract TOC pointer */
|
|
|
|
value -= my_r2(sechdrs, me);
|
|
|
|
*((uint16_t *) location)
|
|
|
|
= (*((uint16_t *) location) & ~0xffff)
|
|
|
|
| (value & 0xffff);
|
|
|
|
break;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
case R_PPC64_TOC16_DS:
|
2006-04-24 20:41:51 +08:00
|
|
|
/* Subtract TOC pointer */
|
2005-04-17 06:20:36 +08:00
|
|
|
value -= my_r2(sechdrs, me);
|
|
|
|
if ((value & 3) != 0 || value + 0x8000 > 0xffff) {
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_err("%s: bad TOC16_DS relocation (0x%lx)\n",
|
2005-04-17 06:20:36 +08:00
|
|
|
me->name, value);
|
|
|
|
return -ENOEXEC;
|
|
|
|
}
|
|
|
|
*((uint16_t *) location)
|
|
|
|
= (*((uint16_t *) location) & ~0xfffc)
|
|
|
|
| (value & 0xfffc);
|
|
|
|
break;
|
|
|
|
|
2012-11-27 01:41:08 +08:00
|
|
|
case R_PPC64_TOC16_LO_DS:
|
|
|
|
/* Subtract TOC pointer */
|
|
|
|
value -= my_r2(sechdrs, me);
|
|
|
|
if ((value & 3) != 0) {
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_err("%s: bad TOC16_LO_DS relocation (0x%lx)\n",
|
2012-11-27 01:41:08 +08:00
|
|
|
me->name, value);
|
|
|
|
return -ENOEXEC;
|
|
|
|
}
|
|
|
|
*((uint16_t *) location)
|
|
|
|
= (*((uint16_t *) location) & ~0xfffc)
|
|
|
|
| (value & 0xfffc);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case R_PPC64_TOC16_HA:
|
|
|
|
/* Subtract TOC pointer */
|
|
|
|
value -= my_r2(sechdrs, me);
|
|
|
|
value = ((value + 0x8000) >> 16);
|
|
|
|
*((uint16_t *) location)
|
|
|
|
= (*((uint16_t *) location) & ~0xffff)
|
|
|
|
| (value & 0xffff);
|
|
|
|
break;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
case R_PPC_REL24:
|
|
|
|
/* FIXME: Handle weak symbols here --RR */
|
2017-11-14 17:29:08 +08:00
|
|
|
if (sym->st_shndx == SHN_UNDEF ||
|
|
|
|
sym->st_shndx == SHN_LIVEPATCH) {
|
2005-04-17 06:20:36 +08:00
|
|
|
/* External: go via stub */
|
|
|
|
value = stub_for_addr(sechdrs, value, me);
|
|
|
|
if (!value)
|
|
|
|
return -ENOENT;
|
|
|
|
if (!restore_r2((u32 *)location + 1, me))
|
|
|
|
return -ENOEXEC;
|
2016-03-03 12:26:59 +08:00
|
|
|
|
|
|
|
squash_toc_save_inst(strtab + sym->st_name, value);
|
2014-03-19 08:12:22 +08:00
|
|
|
} else
|
|
|
|
value += local_entry_offset(sym);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* Convert value to relative */
|
|
|
|
value -= (unsigned long)location;
|
|
|
|
if (value + 0x2000000 > 0x3ffffff || (value & 3) != 0){
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_err("%s: REL24 %li out of range!\n",
|
2005-04-17 06:20:36 +08:00
|
|
|
me->name, (long int)value);
|
|
|
|
return -ENOEXEC;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Only replace bits 2 through 26 */
|
2007-11-14 00:24:04 +08:00
|
|
|
*(uint32_t *)location
|
2005-04-17 06:20:36 +08:00
|
|
|
= (*(uint32_t *)location & ~0x03fffffc)
|
|
|
|
| (value & 0x03fffffc);
|
|
|
|
break;
|
|
|
|
|
2006-10-20 09:47:19 +08:00
|
|
|
case R_PPC64_REL64:
|
|
|
|
/* 64 bits relative (used by features fixups) */
|
|
|
|
*location = value - (unsigned long)location;
|
|
|
|
break;
|
|
|
|
|
2016-10-26 11:51:12 +08:00
|
|
|
case R_PPC64_REL32:
|
|
|
|
/* 32 bits relative (used by relative exception tables) */
|
|
|
|
*(u32 *)location = value - (unsigned long)location;
|
|
|
|
break;
|
|
|
|
|
2014-03-18 15:06:28 +08:00
|
|
|
case R_PPC64_TOCSAVE:
|
|
|
|
/*
|
|
|
|
* Marker reloc indicates we don't have to save r2.
|
|
|
|
* That would only save us one instruction, so ignore
|
|
|
|
* it.
|
|
|
|
*/
|
|
|
|
break;
|
|
|
|
|
powerpc/module: Handle R_PPC64_ENTRY relocations
GCC 6 will include changes to generated code with -mcmodel=large,
which is used to build kernel modules on powerpc64le. This was
necessary because the large model is supposed to allow arbitrary
sizes and locations of the code and data sections, but the ELFv2
global entry point prolog still made the unconditional assumption
that the TOC associated with any particular function can be found
within 2 GB of the function entry point:
func:
addis r2,r12,(.TOC.-func)@ha
addi r2,r2,(.TOC.-func)@l
.localentry func, .-func
To remove this assumption, GCC will now generate instead this global
entry point prolog sequence when using -mcmodel=large:
.quad .TOC.-func
func:
.reloc ., R_PPC64_ENTRY
ld r2, -8(r12)
add r2, r2, r12
.localentry func, .-func
The new .reloc triggers an optimization in the linker that will
replace this new prolog with the original code (see above) if the
linker determines that the distance between .TOC. and func is in
range after all.
Since this new relocation is now present in module object files,
the kernel module loader is required to handle them too. This
patch adds support for the new relocation and implements the
same optimization done by the GNU linker.
Cc: stable@vger.kernel.org
Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2016-01-12 20:14:23 +08:00
|
|
|
case R_PPC64_ENTRY:
|
|
|
|
/*
|
|
|
|
* Optimize ELFv2 large code model entry point if
|
|
|
|
* the TOC is within 2GB range of current location.
|
|
|
|
*/
|
|
|
|
value = my_r2(sechdrs, me) - (unsigned long)location;
|
|
|
|
if (value + 0x80008000 > 0xffffffff)
|
|
|
|
break;
|
|
|
|
/*
|
|
|
|
* Check for the large code model prolog sequence:
|
|
|
|
* ld r2, ...(r12)
|
|
|
|
* add r2, r2, r12
|
|
|
|
*/
|
|
|
|
if ((((uint32_t *)location)[0] & ~0xfffc)
|
|
|
|
!= 0xe84c0000)
|
|
|
|
break;
|
|
|
|
if (((uint32_t *)location)[1] != 0x7c426214)
|
|
|
|
break;
|
|
|
|
/*
|
|
|
|
* If found, replace it with:
|
|
|
|
* addis r2, r12, (.TOC.-func)@ha
|
|
|
|
* addi r2, r12, (.TOC.-func)@l
|
|
|
|
*/
|
|
|
|
((uint32_t *)location)[0] = 0x3c4c0000 + PPC_HA(value);
|
|
|
|
((uint32_t *)location)[1] = 0x38420000 + PPC_LO(value);
|
|
|
|
break;
|
|
|
|
|
2014-03-18 17:29:27 +08:00
|
|
|
case R_PPC64_REL16_HA:
|
|
|
|
/* Subtract location pointer */
|
|
|
|
value -= (unsigned long)location;
|
|
|
|
value = ((value + 0x8000) >> 16);
|
|
|
|
*((uint16_t *) location)
|
|
|
|
= (*((uint16_t *) location) & ~0xffff)
|
|
|
|
| (value & 0xffff);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case R_PPC64_REL16_LO:
|
|
|
|
/* Subtract location pointer */
|
|
|
|
value -= (unsigned long)location;
|
|
|
|
*((uint16_t *) location)
|
|
|
|
= (*((uint16_t *) location) & ~0xffff)
|
|
|
|
| (value & 0xffff);
|
|
|
|
break;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
default:
|
2014-09-17 12:39:35 +08:00
|
|
|
pr_err("%s: Unknown ADD relocation: %lu\n",
|
2005-04-17 06:20:36 +08:00
|
|
|
me->name,
|
|
|
|
(unsigned long)ELF64_R_TYPE(rela[i].r_info));
|
|
|
|
return -ENOEXEC;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-03-03 12:26:54 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-11-15 12:47:03 +08:00
|
|
|
#ifdef CONFIG_DYNAMIC_FTRACE
|
2016-03-03 12:26:56 +08:00
|
|
|
|
|
|
|
#ifdef CC_USING_MPROFILE_KERNEL
|
|
|
|
|
|
|
|
#define PACATOC offsetof(struct paca_struct, kernel_toc)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For mprofile-kernel we use a special stub for ftrace_caller() because we
|
|
|
|
* can't rely on r2 containing this module's TOC when we enter the stub.
|
|
|
|
*
|
|
|
|
* That can happen if the function calling us didn't need to use the toc. In
|
|
|
|
* that case it won't have setup r2, and the r2 value will be either the
|
|
|
|
* kernel's toc, or possibly another modules toc.
|
|
|
|
*
|
|
|
|
* To deal with that this stub uses the kernel toc, which is always accessible
|
|
|
|
* via the paca (in r13). The target (ftrace_caller()) is responsible for
|
|
|
|
* saving and restoring the toc before returning.
|
|
|
|
*/
|
|
|
|
static unsigned long create_ftrace_stub(const Elf64_Shdr *sechdrs, struct module *me)
|
|
|
|
{
|
|
|
|
struct ppc64_stub_entry *entry;
|
|
|
|
unsigned int i, num_stubs;
|
|
|
|
static u32 stub_insns[] = {
|
|
|
|
0xe98d0000 | PACATOC, /* ld r12,PACATOC(r13) */
|
|
|
|
0x3d8c0000, /* addis r12,r12,<high> */
|
|
|
|
0x398c0000, /* addi r12,r12,<low> */
|
|
|
|
0x7d8903a6, /* mtctr r12 */
|
|
|
|
0x4e800420, /* bctr */
|
|
|
|
};
|
|
|
|
long reladdr;
|
|
|
|
|
|
|
|
num_stubs = sechdrs[me->arch.stubs_section].sh_size / sizeof(*entry);
|
|
|
|
|
|
|
|
/* Find the next available stub entry */
|
|
|
|
entry = (void *)sechdrs[me->arch.stubs_section].sh_addr;
|
|
|
|
for (i = 0; i < num_stubs && stub_func_addr(entry->funcdata); i++, entry++);
|
|
|
|
|
|
|
|
if (i >= num_stubs) {
|
|
|
|
pr_err("%s: Unable to find a free slot for ftrace stub.\n", me->name);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
memcpy(entry->jump, stub_insns, sizeof(stub_insns));
|
|
|
|
|
|
|
|
/* Stub uses address relative to kernel toc (from the paca) */
|
|
|
|
reladdr = (unsigned long)ftrace_caller - kernel_toc_addr();
|
|
|
|
if (reladdr > 0x7FFFFFFF || reladdr < -(0x80000000L)) {
|
|
|
|
pr_err("%s: Address of ftrace_caller out of range of kernel_toc.\n", me->name);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
entry->jump[1] |= PPC_HA(reladdr);
|
|
|
|
entry->jump[2] |= PPC_LO(reladdr);
|
|
|
|
|
|
|
|
/* Eventhough we don't use funcdata in the stub, it's needed elsewhere. */
|
|
|
|
entry->funcdata = func_desc((unsigned long)ftrace_caller);
|
|
|
|
entry->magic = STUB_MAGIC;
|
|
|
|
|
|
|
|
return (unsigned long)entry;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static unsigned long create_ftrace_stub(const Elf64_Shdr *sechdrs, struct module *me)
|
|
|
|
{
|
|
|
|
return stub_for_addr(sechdrs, (unsigned long)ftrace_caller, me);
|
|
|
|
}
|
2008-11-15 12:47:03 +08:00
|
|
|
#endif
|
|
|
|
|
2016-03-03 12:26:54 +08:00
|
|
|
int module_finalize_ftrace(struct module *mod, const Elf_Shdr *sechdrs)
|
|
|
|
{
|
|
|
|
mod->arch.toc = my_r2(sechdrs, mod);
|
2016-03-03 12:26:56 +08:00
|
|
|
mod->arch.tramp = create_ftrace_stub(sechdrs, mod);
|
2016-03-03 12:26:54 +08:00
|
|
|
|
|
|
|
if (!mod->arch.tramp)
|
|
|
|
return -ENOENT;
|
2008-11-15 12:47:03 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
return 0;
|
|
|
|
}
|
2016-03-03 12:26:54 +08:00
|
|
|
#endif
|