OpenCloudOS-Kernel/kernel/livepatch/core.c

1277 lines
30 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
/*
* core.c - Kernel Live Patching Core
*
* Copyright (C) 2014 Seth Jennings <sjenning@redhat.com>
* Copyright (C) 2014 SUSE
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/mutex.h>
#include <linux/slab.h>
#include <linux/list.h>
#include <linux/kallsyms.h>
#include <linux/livepatch.h>
#include <linux/elf.h>
#include <linux/moduleloader.h>
livepatch: allow removal of a disabled patch Currently we do not allow patch module to unload since there is no method to determine if a task is still running in the patched code. The consistency model gives us the way because when the unpatching finishes we know that all tasks were marked as safe to call an original function. Thus every new call to the function calls the original code and at the same time no task can be somewhere in the patched code, because it had to leave that code to be marked as safe. We can safely let the patch module go after that. Completion is used for synchronization between module removal and sysfs infrastructure in a similar way to commit 942e443127e9 ("module: Fix mod->mkobj.kobj potentially freed too early"). Note that we still do not allow the removal for immediate model, that is no consistency model. The module refcount may increase in this case if somebody disables and enables the patch several times. This should not cause any harm. With this change a call to try_module_get() is moved to __klp_enable_patch from klp_register_patch to make module reference counting symmetric (module_put() is in a patch disable path) and to allow to take a new reference to a disabled module when being enabled. Finally, we need to be very careful about possible races between klp_unregister_patch(), kobject_put() functions and operations on the related sysfs files. kobject_put(&patch->kobj) must be called without klp_mutex. Otherwise, it might be blocked by enabled_store() that needs the mutex as well. In addition, enabled_store() must check if the patch was not unregisted in the meantime. There is no need to do the same for other kobject_put() callsites at the moment. Their sysfs operations neither take the lock nor they access any data that might be freed in the meantime. There was an attempt to use kobjects the right way and prevent these races by design. But it made the patch definition more complicated and opened another can of worms. See https://lkml.kernel.org/r/1464018848-4303-1-git-send-email-pmladek@suse.com [Thanks to Petr Mladek for improving the commit message.] Signed-off-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-03-07 01:20:29 +08:00
#include <linux/completion.h>
module: Fix livepatch/ftrace module text permissions race It's possible for livepatch and ftrace to be toggling a module's text permissions at the same time, resulting in the following panic: BUG: unable to handle page fault for address: ffffffffc005b1d9 #PF: supervisor write access in kernel mode #PF: error_code(0x0003) - permissions violation PGD 3ea0c067 P4D 3ea0c067 PUD 3ea0e067 PMD 3cc13067 PTE 3b8a1061 Oops: 0003 [#1] PREEMPT SMP PTI CPU: 1 PID: 453 Comm: insmod Tainted: G O K 5.2.0-rc1-a188339ca5 #1 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-20181126_142135-anatol 04/01/2014 RIP: 0010:apply_relocate_add+0xbe/0x14c Code: fa 0b 74 21 48 83 fa 18 74 38 48 83 fa 0a 75 40 eb 08 48 83 38 00 74 33 eb 53 83 38 00 75 4e 89 08 89 c8 eb 0a 83 38 00 75 43 <89> 08 48 63 c1 48 39 c8 74 2e eb 48 83 38 00 75 32 48 29 c1 89 08 RSP: 0018:ffffb223c00dbb10 EFLAGS: 00010246 RAX: ffffffffc005b1d9 RBX: 0000000000000000 RCX: ffffffff8b200060 RDX: 000000000000000b RSI: 0000004b0000000b RDI: ffff96bdfcd33000 RBP: ffffb223c00dbb38 R08: ffffffffc005d040 R09: ffffffffc005c1f0 R10: ffff96bdfcd33c40 R11: ffff96bdfcd33b80 R12: 0000000000000018 R13: ffffffffc005c1f0 R14: ffffffffc005e708 R15: ffffffff8b2fbc74 FS: 00007f5f447beba8(0000) GS:ffff96bdff900000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: ffffffffc005b1d9 CR3: 000000003cedc002 CR4: 0000000000360ea0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: klp_init_object_loaded+0x10f/0x219 ? preempt_latency_start+0x21/0x57 klp_enable_patch+0x662/0x809 ? virt_to_head_page+0x3a/0x3c ? kfree+0x8c/0x126 patch_init+0x2ed/0x1000 [livepatch_test02] ? 0xffffffffc0060000 do_one_initcall+0x9f/0x1c5 ? kmem_cache_alloc_trace+0xc4/0xd4 ? do_init_module+0x27/0x210 do_init_module+0x5f/0x210 load_module+0x1c41/0x2290 ? fsnotify_path+0x3b/0x42 ? strstarts+0x2b/0x2b ? kernel_read+0x58/0x65 __do_sys_finit_module+0x9f/0xc3 ? __do_sys_finit_module+0x9f/0xc3 __x64_sys_finit_module+0x1a/0x1c do_syscall_64+0x52/0x61 entry_SYSCALL_64_after_hwframe+0x44/0xa9 The above panic occurs when loading two modules at the same time with ftrace enabled, where at least one of the modules is a livepatch module: CPU0 CPU1 klp_enable_patch() klp_init_object_loaded() module_disable_ro() ftrace_module_enable() ftrace_arch_code_modify_post_process() set_all_modules_text_ro() klp_write_object_relocations() apply_relocate_add() *patches read-only code* - BOOM A similar race exists when toggling ftrace while loading a livepatch module. Fix it by ensuring that the livepatch and ftrace code patching operations -- and their respective permissions changes -- are protected by the text_mutex. Link: http://lkml.kernel.org/r/ab43d56ab909469ac5d2520c5d944ad6d4abd476.1560474114.git.jpoimboe@redhat.com Reported-by: Johannes Erdfelt <johannes@erdfelt.com> Fixes: 444d13ff10fb ("modules: add ro_after_init support") Acked-by: Jessica Yu <jeyu@kernel.org> Reviewed-by: Petr Mladek <pmladek@suse.com> Reviewed-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2019-06-14 09:07:22 +08:00
#include <linux/memory.h>
#include <linux/rcupdate.h>
#include <asm/cacheflush.h>
#include "core.h"
#include "patch.h"
livepatch: Allow to distinguish different version of system state changes The atomic replace runs pre/post (un)install callbacks only from the new livepatch. There are several reasons for this: + Simplicity: clear ordering of operations, no interactions between old and new callbacks. + Reliability: only new livepatch knows what changes can already be made by older livepatches and how to take over the state. + Testing: the atomic replace can be properly tested only when a newer livepatch is available. It might be too late to fix unwanted effect of callbacks from older livepatches. It might happen that an older change is not enough and the same system state has to be modified another way. Different changes need to get distinguished by a version number added to struct klp_state. The version can also be used to prevent loading incompatible livepatches. The check is done when the livepatch is enabled. The rules are: + Any completely new system state modification is allowed. + System state modifications with the same or higher version are allowed for already modified system states. + Cumulative livepatches must handle all system state modifications from already installed livepatches. + Non-cumulative livepatches are allowed to touch already modified system states. Link: http://lkml.kernel.org/r/20191030154313.13263-4-pmladek@suse.com To: Jiri Kosina <jikos@kernel.org> Cc: Kamalesh Babulal <kamalesh@linux.vnet.ibm.com> Cc: Nicolai Stange <nstange@suse.de> Cc: live-patching@vger.kernel.org Cc: linux-kernel@vger.kernel.org Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com>
2019-10-30 23:43:11 +08:00
#include "state.h"
livepatch: change to a per-task consistency model Change livepatch to use a basic per-task consistency model. This is the foundation which will eventually enable us to patch those ~10% of security patches which change function or data semantics. This is the biggest remaining piece needed to make livepatch more generally useful. This code stems from the design proposal made by Vojtech [1] in November 2014. It's a hybrid of kGraft and kpatch: it uses kGraft's per-task consistency and syscall barrier switching combined with kpatch's stack trace switching. There are also a number of fallback options which make it quite flexible. Patches are applied on a per-task basis, when the task is deemed safe to switch over. When a patch is enabled, livepatch enters into a transition state where tasks are converging to the patched state. Usually this transition state can complete in a few seconds. The same sequence occurs when a patch is disabled, except the tasks converge from the patched state to the unpatched state. An interrupt handler inherits the patched state of the task it interrupts. The same is true for forked tasks: the child inherits the patched state of the parent. Livepatch uses several complementary approaches to determine when it's safe to patch tasks: 1. The first and most effective approach is stack checking of sleeping tasks. If no affected functions are on the stack of a given task, the task is patched. In most cases this will patch most or all of the tasks on the first try. Otherwise it'll keep trying periodically. This option is only available if the architecture has reliable stacks (HAVE_RELIABLE_STACKTRACE). 2. The second approach, if needed, is kernel exit switching. A task is switched when it returns to user space from a system call, a user space IRQ, or a signal. It's useful in the following cases: a) Patching I/O-bound user tasks which are sleeping on an affected function. In this case you have to send SIGSTOP and SIGCONT to force it to exit the kernel and be patched. b) Patching CPU-bound user tasks. If the task is highly CPU-bound then it will get patched the next time it gets interrupted by an IRQ. c) In the future it could be useful for applying patches for architectures which don't yet have HAVE_RELIABLE_STACKTRACE. In this case you would have to signal most of the tasks on the system. However this isn't supported yet because there's currently no way to patch kthreads without HAVE_RELIABLE_STACKTRACE. 3. For idle "swapper" tasks, since they don't ever exit the kernel, they instead have a klp_update_patch_state() call in the idle loop which allows them to be patched before the CPU enters the idle state. (Note there's not yet such an approach for kthreads.) All the above approaches may be skipped by setting the 'immediate' flag in the 'klp_patch' struct, which will disable per-task consistency and patch all tasks immediately. This can be useful if the patch doesn't change any function or data semantics. Note that, even with this flag set, it's possible that some tasks may still be running with an old version of the function, until that function returns. There's also an 'immediate' flag in the 'klp_func' struct which allows you to specify that certain functions in the patch can be applied without per-task consistency. This might be useful if you want to patch a common function like schedule(), and the function change doesn't need consistency but the rest of the patch does. For architectures which don't have HAVE_RELIABLE_STACKTRACE, the user must set patch->immediate which causes all tasks to be patched immediately. This option should be used with care, only when the patch doesn't change any function or data semantics. In the future, architectures which don't have HAVE_RELIABLE_STACKTRACE may be allowed to use per-task consistency if we can come up with another way to patch kthreads. The /sys/kernel/livepatch/<patch>/transition file shows whether a patch is in transition. Only a single patch (the topmost patch on the stack) can be in transition at a given time. A patch can remain in transition indefinitely, if any of the tasks are stuck in the initial patch state. A transition can be reversed and effectively canceled by writing the opposite value to the /sys/kernel/livepatch/<patch>/enabled file while the transition is in progress. Then all the tasks will attempt to converge back to the original patch state. [1] https://lkml.kernel.org/r/20141107140458.GA21774@suse.cz Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Ingo Molnar <mingo@kernel.org> # for the scheduler changes Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-02-14 09:42:40 +08:00
#include "transition.h"
/*
livepatch: change to a per-task consistency model Change livepatch to use a basic per-task consistency model. This is the foundation which will eventually enable us to patch those ~10% of security patches which change function or data semantics. This is the biggest remaining piece needed to make livepatch more generally useful. This code stems from the design proposal made by Vojtech [1] in November 2014. It's a hybrid of kGraft and kpatch: it uses kGraft's per-task consistency and syscall barrier switching combined with kpatch's stack trace switching. There are also a number of fallback options which make it quite flexible. Patches are applied on a per-task basis, when the task is deemed safe to switch over. When a patch is enabled, livepatch enters into a transition state where tasks are converging to the patched state. Usually this transition state can complete in a few seconds. The same sequence occurs when a patch is disabled, except the tasks converge from the patched state to the unpatched state. An interrupt handler inherits the patched state of the task it interrupts. The same is true for forked tasks: the child inherits the patched state of the parent. Livepatch uses several complementary approaches to determine when it's safe to patch tasks: 1. The first and most effective approach is stack checking of sleeping tasks. If no affected functions are on the stack of a given task, the task is patched. In most cases this will patch most or all of the tasks on the first try. Otherwise it'll keep trying periodically. This option is only available if the architecture has reliable stacks (HAVE_RELIABLE_STACKTRACE). 2. The second approach, if needed, is kernel exit switching. A task is switched when it returns to user space from a system call, a user space IRQ, or a signal. It's useful in the following cases: a) Patching I/O-bound user tasks which are sleeping on an affected function. In this case you have to send SIGSTOP and SIGCONT to force it to exit the kernel and be patched. b) Patching CPU-bound user tasks. If the task is highly CPU-bound then it will get patched the next time it gets interrupted by an IRQ. c) In the future it could be useful for applying patches for architectures which don't yet have HAVE_RELIABLE_STACKTRACE. In this case you would have to signal most of the tasks on the system. However this isn't supported yet because there's currently no way to patch kthreads without HAVE_RELIABLE_STACKTRACE. 3. For idle "swapper" tasks, since they don't ever exit the kernel, they instead have a klp_update_patch_state() call in the idle loop which allows them to be patched before the CPU enters the idle state. (Note there's not yet such an approach for kthreads.) All the above approaches may be skipped by setting the 'immediate' flag in the 'klp_patch' struct, which will disable per-task consistency and patch all tasks immediately. This can be useful if the patch doesn't change any function or data semantics. Note that, even with this flag set, it's possible that some tasks may still be running with an old version of the function, until that function returns. There's also an 'immediate' flag in the 'klp_func' struct which allows you to specify that certain functions in the patch can be applied without per-task consistency. This might be useful if you want to patch a common function like schedule(), and the function change doesn't need consistency but the rest of the patch does. For architectures which don't have HAVE_RELIABLE_STACKTRACE, the user must set patch->immediate which causes all tasks to be patched immediately. This option should be used with care, only when the patch doesn't change any function or data semantics. In the future, architectures which don't have HAVE_RELIABLE_STACKTRACE may be allowed to use per-task consistency if we can come up with another way to patch kthreads. The /sys/kernel/livepatch/<patch>/transition file shows whether a patch is in transition. Only a single patch (the topmost patch on the stack) can be in transition at a given time. A patch can remain in transition indefinitely, if any of the tasks are stuck in the initial patch state. A transition can be reversed and effectively canceled by writing the opposite value to the /sys/kernel/livepatch/<patch>/enabled file while the transition is in progress. Then all the tasks will attempt to converge back to the original patch state. [1] https://lkml.kernel.org/r/20141107140458.GA21774@suse.cz Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Ingo Molnar <mingo@kernel.org> # for the scheduler changes Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-02-14 09:42:40 +08:00
* klp_mutex is a coarse lock which serializes access to klp data. All
* accesses to klp-related variables and structures must have mutex protection,
* except within the following functions which carefully avoid the need for it:
*
* - klp_ftrace_handler()
* - klp_update_patch_state()
*/
livepatch: change to a per-task consistency model Change livepatch to use a basic per-task consistency model. This is the foundation which will eventually enable us to patch those ~10% of security patches which change function or data semantics. This is the biggest remaining piece needed to make livepatch more generally useful. This code stems from the design proposal made by Vojtech [1] in November 2014. It's a hybrid of kGraft and kpatch: it uses kGraft's per-task consistency and syscall barrier switching combined with kpatch's stack trace switching. There are also a number of fallback options which make it quite flexible. Patches are applied on a per-task basis, when the task is deemed safe to switch over. When a patch is enabled, livepatch enters into a transition state where tasks are converging to the patched state. Usually this transition state can complete in a few seconds. The same sequence occurs when a patch is disabled, except the tasks converge from the patched state to the unpatched state. An interrupt handler inherits the patched state of the task it interrupts. The same is true for forked tasks: the child inherits the patched state of the parent. Livepatch uses several complementary approaches to determine when it's safe to patch tasks: 1. The first and most effective approach is stack checking of sleeping tasks. If no affected functions are on the stack of a given task, the task is patched. In most cases this will patch most or all of the tasks on the first try. Otherwise it'll keep trying periodically. This option is only available if the architecture has reliable stacks (HAVE_RELIABLE_STACKTRACE). 2. The second approach, if needed, is kernel exit switching. A task is switched when it returns to user space from a system call, a user space IRQ, or a signal. It's useful in the following cases: a) Patching I/O-bound user tasks which are sleeping on an affected function. In this case you have to send SIGSTOP and SIGCONT to force it to exit the kernel and be patched. b) Patching CPU-bound user tasks. If the task is highly CPU-bound then it will get patched the next time it gets interrupted by an IRQ. c) In the future it could be useful for applying patches for architectures which don't yet have HAVE_RELIABLE_STACKTRACE. In this case you would have to signal most of the tasks on the system. However this isn't supported yet because there's currently no way to patch kthreads without HAVE_RELIABLE_STACKTRACE. 3. For idle "swapper" tasks, since they don't ever exit the kernel, they instead have a klp_update_patch_state() call in the idle loop which allows them to be patched before the CPU enters the idle state. (Note there's not yet such an approach for kthreads.) All the above approaches may be skipped by setting the 'immediate' flag in the 'klp_patch' struct, which will disable per-task consistency and patch all tasks immediately. This can be useful if the patch doesn't change any function or data semantics. Note that, even with this flag set, it's possible that some tasks may still be running with an old version of the function, until that function returns. There's also an 'immediate' flag in the 'klp_func' struct which allows you to specify that certain functions in the patch can be applied without per-task consistency. This might be useful if you want to patch a common function like schedule(), and the function change doesn't need consistency but the rest of the patch does. For architectures which don't have HAVE_RELIABLE_STACKTRACE, the user must set patch->immediate which causes all tasks to be patched immediately. This option should be used with care, only when the patch doesn't change any function or data semantics. In the future, architectures which don't have HAVE_RELIABLE_STACKTRACE may be allowed to use per-task consistency if we can come up with another way to patch kthreads. The /sys/kernel/livepatch/<patch>/transition file shows whether a patch is in transition. Only a single patch (the topmost patch on the stack) can be in transition at a given time. A patch can remain in transition indefinitely, if any of the tasks are stuck in the initial patch state. A transition can be reversed and effectively canceled by writing the opposite value to the /sys/kernel/livepatch/<patch>/enabled file while the transition is in progress. Then all the tasks will attempt to converge back to the original patch state. [1] https://lkml.kernel.org/r/20141107140458.GA21774@suse.cz Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Ingo Molnar <mingo@kernel.org> # for the scheduler changes Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-02-14 09:42:40 +08:00
DEFINE_MUTEX(klp_mutex);
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
/*
* Actively used patches: enabled or in transition. Note that replaced
* or disabled patches are not listed even though the related kernel
* module still can be loaded.
*/
LIST_HEAD(klp_patches);
static struct kobject *klp_root_kobj;
static bool klp_is_module(struct klp_object *obj)
{
return obj->name;
}
/* sets obj->mod if object is not vmlinux and module is found */
static void klp_find_object_module(struct klp_object *obj)
{
livepatch: Fix subtle race with coming and going modules There is a notifier that handles live patches for coming and going modules. It takes klp_mutex lock to avoid races with coming and going patches but it does not keep the lock all the time. Therefore the following races are possible: 1. The notifier is called sometime in STATE_MODULE_COMING. The module is visible by find_module() in this state all the time. It means that new patch can be registered and enabled even before the notifier is called. It might create wrong order of stacked patches, see below for an example. 2. New patch could still see the module in the GOING state even after the notifier has been called. It will try to initialize the related object structures but the module could disappear at any time. There will stay mess in the structures. It might even cause an invalid memory access. This patch solves the problem by adding a boolean variable into struct module. The value is true after the coming and before the going handler is called. New patches need to be applied when the value is true and they need to ignore the module when the value is false. Note that we need to know state of all modules on the system. The races are related to new patches. Therefore we do not know what modules will get patched. Also note that we could not simply ignore going modules. The code from the module could be called even in the GOING state until mod->exit() finishes. If we start supporting patches with semantic changes between function calls, we need to apply new patches to any still usable code. See below for an example. Finally note that the patch solves only the situation when a new patch is registered. There are no such problems when the patch is being removed. It does not matter who disable the patch first, whether the normal disable_patch() or the module notifier. There is nothing to do once the patch is disabled. Alternative solutions: ====================== + reject new patches when a patched module is coming or going; this is ugly + wait with adding new patch until the module leaves the COMING and GOING states; this might be dangerous and complicated; we would need to release kgr_lock in the middle of the patch registration to avoid a deadlock with the coming and going handlers; also we might need a waitqueue for each module which seems to be even bigger overhead than the boolean + stop modules from entering COMING and GOING states; wait until modules leave these states when they are already there; looks complicated; we would need to ignore the module that asked to stop the others to avoid a deadlock; also it is unclear what to do when two modules asked to stop others and both are in COMING state (situation when two new patches are applied) + always register/enable new patches and fix up the potential mess (registered patches order) in klp_module_init(); this is nasty and prone to regressions in the future development + add another MODULE_STATE where the kallsyms are visible but the module is not used yet; this looks too complex; the module states are checked on "many" locations Example of patch stacking breakage: =================================== The notifier could _not_ _simply_ ignore already initialized module objects. For example, let's have three patches (P1, P2, P3) for functions a() and b() where a() is from vmcore and b() is from a module M. Something like: a() b() P1 a1() b1() P2 a2() b2() P3 a3() b3(3) If you load the module M after all patches are registered and enabled. The ftrace ops for function a() and b() has listed the functions in this order: ops_a->func_stack -> list(a3,a2,a1) ops_b->func_stack -> list(b3,b2,b1) , so the pointer to b3() is the first and will be used. Then you might have the following scenario. Let's start with state when patches P1 and P2 are registered and enabled but the module M is not loaded. Then ftrace ops for b() does not exist. Then we get into the following race: CPU0 CPU1 load_module(M) complete_formation() mod->state = MODULE_STATE_COMING; mutex_unlock(&module_mutex); klp_register_patch(P3); klp_enable_patch(P3); # STATE 1 klp_module_notify(M) klp_module_notify_coming(P1); klp_module_notify_coming(P2); klp_module_notify_coming(P3); # STATE 2 The ftrace ops for a() and b() then looks: STATE1: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b3); STATE2: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b2,b1,b3); therefore, b2() is used for the module but a3() is used for vmcore because they were the last added. Example of the race with going modules: ======================================= CPU0 CPU1 delete_module() #SYSCALL try_stop_module() mod->state = MODULE_STATE_GOING; mutex_unlock(&module_mutex); klp_register_patch() klp_enable_patch() #save place to switch universe b() # from module that is going a() # from core (patched) mod->exit(); Note that the function b() can be called until we call mod->exit(). If we do not apply patch against b() because it is in MODULE_STATE_GOING, it will call patched a() with modified semantic and things might get wrong. [jpoimboe@redhat.com: use one boolean instead of two] Signed-off-by: Petr Mladek <pmladek@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-03-12 19:55:13 +08:00
struct module *mod;
if (!klp_is_module(obj))
return;
rcu_read_lock_sched();
/*
livepatch: Fix subtle race with coming and going modules There is a notifier that handles live patches for coming and going modules. It takes klp_mutex lock to avoid races with coming and going patches but it does not keep the lock all the time. Therefore the following races are possible: 1. The notifier is called sometime in STATE_MODULE_COMING. The module is visible by find_module() in this state all the time. It means that new patch can be registered and enabled even before the notifier is called. It might create wrong order of stacked patches, see below for an example. 2. New patch could still see the module in the GOING state even after the notifier has been called. It will try to initialize the related object structures but the module could disappear at any time. There will stay mess in the structures. It might even cause an invalid memory access. This patch solves the problem by adding a boolean variable into struct module. The value is true after the coming and before the going handler is called. New patches need to be applied when the value is true and they need to ignore the module when the value is false. Note that we need to know state of all modules on the system. The races are related to new patches. Therefore we do not know what modules will get patched. Also note that we could not simply ignore going modules. The code from the module could be called even in the GOING state until mod->exit() finishes. If we start supporting patches with semantic changes between function calls, we need to apply new patches to any still usable code. See below for an example. Finally note that the patch solves only the situation when a new patch is registered. There are no such problems when the patch is being removed. It does not matter who disable the patch first, whether the normal disable_patch() or the module notifier. There is nothing to do once the patch is disabled. Alternative solutions: ====================== + reject new patches when a patched module is coming or going; this is ugly + wait with adding new patch until the module leaves the COMING and GOING states; this might be dangerous and complicated; we would need to release kgr_lock in the middle of the patch registration to avoid a deadlock with the coming and going handlers; also we might need a waitqueue for each module which seems to be even bigger overhead than the boolean + stop modules from entering COMING and GOING states; wait until modules leave these states when they are already there; looks complicated; we would need to ignore the module that asked to stop the others to avoid a deadlock; also it is unclear what to do when two modules asked to stop others and both are in COMING state (situation when two new patches are applied) + always register/enable new patches and fix up the potential mess (registered patches order) in klp_module_init(); this is nasty and prone to regressions in the future development + add another MODULE_STATE where the kallsyms are visible but the module is not used yet; this looks too complex; the module states are checked on "many" locations Example of patch stacking breakage: =================================== The notifier could _not_ _simply_ ignore already initialized module objects. For example, let's have three patches (P1, P2, P3) for functions a() and b() where a() is from vmcore and b() is from a module M. Something like: a() b() P1 a1() b1() P2 a2() b2() P3 a3() b3(3) If you load the module M after all patches are registered and enabled. The ftrace ops for function a() and b() has listed the functions in this order: ops_a->func_stack -> list(a3,a2,a1) ops_b->func_stack -> list(b3,b2,b1) , so the pointer to b3() is the first and will be used. Then you might have the following scenario. Let's start with state when patches P1 and P2 are registered and enabled but the module M is not loaded. Then ftrace ops for b() does not exist. Then we get into the following race: CPU0 CPU1 load_module(M) complete_formation() mod->state = MODULE_STATE_COMING; mutex_unlock(&module_mutex); klp_register_patch(P3); klp_enable_patch(P3); # STATE 1 klp_module_notify(M) klp_module_notify_coming(P1); klp_module_notify_coming(P2); klp_module_notify_coming(P3); # STATE 2 The ftrace ops for a() and b() then looks: STATE1: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b3); STATE2: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b2,b1,b3); therefore, b2() is used for the module but a3() is used for vmcore because they were the last added. Example of the race with going modules: ======================================= CPU0 CPU1 delete_module() #SYSCALL try_stop_module() mod->state = MODULE_STATE_GOING; mutex_unlock(&module_mutex); klp_register_patch() klp_enable_patch() #save place to switch universe b() # from module that is going a() # from core (patched) mod->exit(); Note that the function b() can be called until we call mod->exit(). If we do not apply patch against b() because it is in MODULE_STATE_GOING, it will call patched a() with modified semantic and things might get wrong. [jpoimboe@redhat.com: use one boolean instead of two] Signed-off-by: Petr Mladek <pmladek@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-03-12 19:55:13 +08:00
* We do not want to block removal of patched modules and therefore
* we do not take a reference here. The patches are removed by
* klp_module_going() instead.
livepatch: Fix subtle race with coming and going modules There is a notifier that handles live patches for coming and going modules. It takes klp_mutex lock to avoid races with coming and going patches but it does not keep the lock all the time. Therefore the following races are possible: 1. The notifier is called sometime in STATE_MODULE_COMING. The module is visible by find_module() in this state all the time. It means that new patch can be registered and enabled even before the notifier is called. It might create wrong order of stacked patches, see below for an example. 2. New patch could still see the module in the GOING state even after the notifier has been called. It will try to initialize the related object structures but the module could disappear at any time. There will stay mess in the structures. It might even cause an invalid memory access. This patch solves the problem by adding a boolean variable into struct module. The value is true after the coming and before the going handler is called. New patches need to be applied when the value is true and they need to ignore the module when the value is false. Note that we need to know state of all modules on the system. The races are related to new patches. Therefore we do not know what modules will get patched. Also note that we could not simply ignore going modules. The code from the module could be called even in the GOING state until mod->exit() finishes. If we start supporting patches with semantic changes between function calls, we need to apply new patches to any still usable code. See below for an example. Finally note that the patch solves only the situation when a new patch is registered. There are no such problems when the patch is being removed. It does not matter who disable the patch first, whether the normal disable_patch() or the module notifier. There is nothing to do once the patch is disabled. Alternative solutions: ====================== + reject new patches when a patched module is coming or going; this is ugly + wait with adding new patch until the module leaves the COMING and GOING states; this might be dangerous and complicated; we would need to release kgr_lock in the middle of the patch registration to avoid a deadlock with the coming and going handlers; also we might need a waitqueue for each module which seems to be even bigger overhead than the boolean + stop modules from entering COMING and GOING states; wait until modules leave these states when they are already there; looks complicated; we would need to ignore the module that asked to stop the others to avoid a deadlock; also it is unclear what to do when two modules asked to stop others and both are in COMING state (situation when two new patches are applied) + always register/enable new patches and fix up the potential mess (registered patches order) in klp_module_init(); this is nasty and prone to regressions in the future development + add another MODULE_STATE where the kallsyms are visible but the module is not used yet; this looks too complex; the module states are checked on "many" locations Example of patch stacking breakage: =================================== The notifier could _not_ _simply_ ignore already initialized module objects. For example, let's have three patches (P1, P2, P3) for functions a() and b() where a() is from vmcore and b() is from a module M. Something like: a() b() P1 a1() b1() P2 a2() b2() P3 a3() b3(3) If you load the module M after all patches are registered and enabled. The ftrace ops for function a() and b() has listed the functions in this order: ops_a->func_stack -> list(a3,a2,a1) ops_b->func_stack -> list(b3,b2,b1) , so the pointer to b3() is the first and will be used. Then you might have the following scenario. Let's start with state when patches P1 and P2 are registered and enabled but the module M is not loaded. Then ftrace ops for b() does not exist. Then we get into the following race: CPU0 CPU1 load_module(M) complete_formation() mod->state = MODULE_STATE_COMING; mutex_unlock(&module_mutex); klp_register_patch(P3); klp_enable_patch(P3); # STATE 1 klp_module_notify(M) klp_module_notify_coming(P1); klp_module_notify_coming(P2); klp_module_notify_coming(P3); # STATE 2 The ftrace ops for a() and b() then looks: STATE1: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b3); STATE2: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b2,b1,b3); therefore, b2() is used for the module but a3() is used for vmcore because they were the last added. Example of the race with going modules: ======================================= CPU0 CPU1 delete_module() #SYSCALL try_stop_module() mod->state = MODULE_STATE_GOING; mutex_unlock(&module_mutex); klp_register_patch() klp_enable_patch() #save place to switch universe b() # from module that is going a() # from core (patched) mod->exit(); Note that the function b() can be called until we call mod->exit(). If we do not apply patch against b() because it is in MODULE_STATE_GOING, it will call patched a() with modified semantic and things might get wrong. [jpoimboe@redhat.com: use one boolean instead of two] Signed-off-by: Petr Mladek <pmladek@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-03-12 19:55:13 +08:00
*/
mod = find_module(obj->name);
/*
* Do not mess work of klp_module_coming() and klp_module_going().
* Note that the patch might still be needed before klp_module_going()
livepatch: Fix subtle race with coming and going modules There is a notifier that handles live patches for coming and going modules. It takes klp_mutex lock to avoid races with coming and going patches but it does not keep the lock all the time. Therefore the following races are possible: 1. The notifier is called sometime in STATE_MODULE_COMING. The module is visible by find_module() in this state all the time. It means that new patch can be registered and enabled even before the notifier is called. It might create wrong order of stacked patches, see below for an example. 2. New patch could still see the module in the GOING state even after the notifier has been called. It will try to initialize the related object structures but the module could disappear at any time. There will stay mess in the structures. It might even cause an invalid memory access. This patch solves the problem by adding a boolean variable into struct module. The value is true after the coming and before the going handler is called. New patches need to be applied when the value is true and they need to ignore the module when the value is false. Note that we need to know state of all modules on the system. The races are related to new patches. Therefore we do not know what modules will get patched. Also note that we could not simply ignore going modules. The code from the module could be called even in the GOING state until mod->exit() finishes. If we start supporting patches with semantic changes between function calls, we need to apply new patches to any still usable code. See below for an example. Finally note that the patch solves only the situation when a new patch is registered. There are no such problems when the patch is being removed. It does not matter who disable the patch first, whether the normal disable_patch() or the module notifier. There is nothing to do once the patch is disabled. Alternative solutions: ====================== + reject new patches when a patched module is coming or going; this is ugly + wait with adding new patch until the module leaves the COMING and GOING states; this might be dangerous and complicated; we would need to release kgr_lock in the middle of the patch registration to avoid a deadlock with the coming and going handlers; also we might need a waitqueue for each module which seems to be even bigger overhead than the boolean + stop modules from entering COMING and GOING states; wait until modules leave these states when they are already there; looks complicated; we would need to ignore the module that asked to stop the others to avoid a deadlock; also it is unclear what to do when two modules asked to stop others and both are in COMING state (situation when two new patches are applied) + always register/enable new patches and fix up the potential mess (registered patches order) in klp_module_init(); this is nasty and prone to regressions in the future development + add another MODULE_STATE where the kallsyms are visible but the module is not used yet; this looks too complex; the module states are checked on "many" locations Example of patch stacking breakage: =================================== The notifier could _not_ _simply_ ignore already initialized module objects. For example, let's have three patches (P1, P2, P3) for functions a() and b() where a() is from vmcore and b() is from a module M. Something like: a() b() P1 a1() b1() P2 a2() b2() P3 a3() b3(3) If you load the module M after all patches are registered and enabled. The ftrace ops for function a() and b() has listed the functions in this order: ops_a->func_stack -> list(a3,a2,a1) ops_b->func_stack -> list(b3,b2,b1) , so the pointer to b3() is the first and will be used. Then you might have the following scenario. Let's start with state when patches P1 and P2 are registered and enabled but the module M is not loaded. Then ftrace ops for b() does not exist. Then we get into the following race: CPU0 CPU1 load_module(M) complete_formation() mod->state = MODULE_STATE_COMING; mutex_unlock(&module_mutex); klp_register_patch(P3); klp_enable_patch(P3); # STATE 1 klp_module_notify(M) klp_module_notify_coming(P1); klp_module_notify_coming(P2); klp_module_notify_coming(P3); # STATE 2 The ftrace ops for a() and b() then looks: STATE1: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b3); STATE2: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b2,b1,b3); therefore, b2() is used for the module but a3() is used for vmcore because they were the last added. Example of the race with going modules: ======================================= CPU0 CPU1 delete_module() #SYSCALL try_stop_module() mod->state = MODULE_STATE_GOING; mutex_unlock(&module_mutex); klp_register_patch() klp_enable_patch() #save place to switch universe b() # from module that is going a() # from core (patched) mod->exit(); Note that the function b() can be called until we call mod->exit(). If we do not apply patch against b() because it is in MODULE_STATE_GOING, it will call patched a() with modified semantic and things might get wrong. [jpoimboe@redhat.com: use one boolean instead of two] Signed-off-by: Petr Mladek <pmladek@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-03-12 19:55:13 +08:00
* is called. Module functions can be called even in the GOING state
* until mod->exit() finishes. This is especially important for
* patches that modify semantic of the functions.
*/
livepatch: Fix subtle race with coming and going modules There is a notifier that handles live patches for coming and going modules. It takes klp_mutex lock to avoid races with coming and going patches but it does not keep the lock all the time. Therefore the following races are possible: 1. The notifier is called sometime in STATE_MODULE_COMING. The module is visible by find_module() in this state all the time. It means that new patch can be registered and enabled even before the notifier is called. It might create wrong order of stacked patches, see below for an example. 2. New patch could still see the module in the GOING state even after the notifier has been called. It will try to initialize the related object structures but the module could disappear at any time. There will stay mess in the structures. It might even cause an invalid memory access. This patch solves the problem by adding a boolean variable into struct module. The value is true after the coming and before the going handler is called. New patches need to be applied when the value is true and they need to ignore the module when the value is false. Note that we need to know state of all modules on the system. The races are related to new patches. Therefore we do not know what modules will get patched. Also note that we could not simply ignore going modules. The code from the module could be called even in the GOING state until mod->exit() finishes. If we start supporting patches with semantic changes between function calls, we need to apply new patches to any still usable code. See below for an example. Finally note that the patch solves only the situation when a new patch is registered. There are no such problems when the patch is being removed. It does not matter who disable the patch first, whether the normal disable_patch() or the module notifier. There is nothing to do once the patch is disabled. Alternative solutions: ====================== + reject new patches when a patched module is coming or going; this is ugly + wait with adding new patch until the module leaves the COMING and GOING states; this might be dangerous and complicated; we would need to release kgr_lock in the middle of the patch registration to avoid a deadlock with the coming and going handlers; also we might need a waitqueue for each module which seems to be even bigger overhead than the boolean + stop modules from entering COMING and GOING states; wait until modules leave these states when they are already there; looks complicated; we would need to ignore the module that asked to stop the others to avoid a deadlock; also it is unclear what to do when two modules asked to stop others and both are in COMING state (situation when two new patches are applied) + always register/enable new patches and fix up the potential mess (registered patches order) in klp_module_init(); this is nasty and prone to regressions in the future development + add another MODULE_STATE where the kallsyms are visible but the module is not used yet; this looks too complex; the module states are checked on "many" locations Example of patch stacking breakage: =================================== The notifier could _not_ _simply_ ignore already initialized module objects. For example, let's have three patches (P1, P2, P3) for functions a() and b() where a() is from vmcore and b() is from a module M. Something like: a() b() P1 a1() b1() P2 a2() b2() P3 a3() b3(3) If you load the module M after all patches are registered and enabled. The ftrace ops for function a() and b() has listed the functions in this order: ops_a->func_stack -> list(a3,a2,a1) ops_b->func_stack -> list(b3,b2,b1) , so the pointer to b3() is the first and will be used. Then you might have the following scenario. Let's start with state when patches P1 and P2 are registered and enabled but the module M is not loaded. Then ftrace ops for b() does not exist. Then we get into the following race: CPU0 CPU1 load_module(M) complete_formation() mod->state = MODULE_STATE_COMING; mutex_unlock(&module_mutex); klp_register_patch(P3); klp_enable_patch(P3); # STATE 1 klp_module_notify(M) klp_module_notify_coming(P1); klp_module_notify_coming(P2); klp_module_notify_coming(P3); # STATE 2 The ftrace ops for a() and b() then looks: STATE1: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b3); STATE2: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b2,b1,b3); therefore, b2() is used for the module but a3() is used for vmcore because they were the last added. Example of the race with going modules: ======================================= CPU0 CPU1 delete_module() #SYSCALL try_stop_module() mod->state = MODULE_STATE_GOING; mutex_unlock(&module_mutex); klp_register_patch() klp_enable_patch() #save place to switch universe b() # from module that is going a() # from core (patched) mod->exit(); Note that the function b() can be called until we call mod->exit(). If we do not apply patch against b() because it is in MODULE_STATE_GOING, it will call patched a() with modified semantic and things might get wrong. [jpoimboe@redhat.com: use one boolean instead of two] Signed-off-by: Petr Mladek <pmladek@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-03-12 19:55:13 +08:00
if (mod && mod->klp_alive)
obj->mod = mod;
rcu_read_unlock_sched();
}
static bool klp_initialized(void)
{
return !!klp_root_kobj;
}
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
static struct klp_func *klp_find_func(struct klp_object *obj,
struct klp_func *old_func)
{
struct klp_func *func;
klp_for_each_func(obj, func) {
if ((strcmp(old_func->old_name, func->old_name) == 0) &&
(old_func->old_sympos == func->old_sympos)) {
return func;
}
}
return NULL;
}
static struct klp_object *klp_find_object(struct klp_patch *patch,
struct klp_object *old_obj)
{
struct klp_object *obj;
klp_for_each_object(patch, obj) {
if (klp_is_module(old_obj)) {
if (klp_is_module(obj) &&
strcmp(old_obj->name, obj->name) == 0) {
return obj;
}
} else if (!klp_is_module(obj)) {
return obj;
}
}
return NULL;
}
struct klp_find_arg {
const char *objname;
const char *name;
unsigned long addr;
unsigned long count;
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
unsigned long pos;
};
static int klp_find_callback(void *data, const char *name,
struct module *mod, unsigned long addr)
{
struct klp_find_arg *args = data;
if ((mod && !args->objname) || (!mod && args->objname))
return 0;
if (strcmp(args->name, name))
return 0;
if (args->objname && strcmp(args->objname, mod->name))
return 0;
args->addr = addr;
args->count++;
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
/*
* Finish the search when the symbol is found for the desired position
* or the position is not defined for a non-unique symbol.
*/
if ((args->pos && (args->count == args->pos)) ||
(!args->pos && (args->count > 1)))
return 1;
return 0;
}
static int klp_find_object_symbol(const char *objname, const char *name,
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
unsigned long sympos, unsigned long *addr)
{
struct klp_find_arg args = {
.objname = objname,
.name = name,
.addr = 0,
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
.count = 0,
.pos = sympos,
};
if (objname)
module_kallsyms_on_each_symbol(klp_find_callback, &args);
else
kallsyms_on_each_symbol(klp_find_callback, &args);
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
/*
* Ensure an address was found. If sympos is 0, ensure symbol is unique;
* otherwise ensure the symbol position count matches sympos.
*/
if (args.addr == 0)
pr_err("symbol '%s' not found in symbol table\n", name);
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
else if (args.count > 1 && sympos == 0) {
pr_err("unresolvable ambiguity for symbol '%s' in object '%s'\n",
name, objname);
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
} else if (sympos != args.count && sympos > 0) {
pr_err("symbol position %lu for symbol '%s' in object '%s' not found\n",
sympos, name, objname ? objname : "vmlinux");
} else {
*addr = args.addr;
return 0;
}
*addr = 0;
return -EINVAL;
}
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
static int klp_resolve_symbols(Elf64_Shdr *sechdrs, const char *strtab,
unsigned int symndx, Elf_Shdr *relasec,
const char *sec_objname)
{
int i, cnt, ret;
char sym_objname[MODULE_NAME_LEN];
char sym_name[KSYM_NAME_LEN];
Elf_Rela *relas;
Elf_Sym *sym;
unsigned long sympos, addr;
bool sym_vmlinux;
bool sec_vmlinux = !strcmp(sec_objname, "vmlinux");
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
/*
* Since the field widths for sym_objname and sym_name in the sscanf()
* call are hard-coded and correspond to MODULE_NAME_LEN and
* KSYM_NAME_LEN respectively, we must make sure that MODULE_NAME_LEN
* and KSYM_NAME_LEN have the values we expect them to have.
*
* Because the value of MODULE_NAME_LEN can differ among architectures,
* we use the smallest/strictest upper bound possible (56, based on
* the current definition of MODULE_NAME_LEN) to prevent overflows.
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
*/
BUILD_BUG_ON(MODULE_NAME_LEN < 56 || KSYM_NAME_LEN != 128);
relas = (Elf_Rela *) relasec->sh_addr;
/* For each rela in this klp relocation section */
for (i = 0; i < relasec->sh_size / sizeof(Elf_Rela); i++) {
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
sym = (Elf64_Sym *)sechdrs[symndx].sh_addr + ELF_R_SYM(relas[i].r_info);
if (sym->st_shndx != SHN_LIVEPATCH) {
pr_err("symbol %s is not marked as a livepatch symbol\n",
strtab + sym->st_name);
return -EINVAL;
}
/* Format: .klp.sym.sym_objname.sym_name,sympos */
cnt = sscanf(strtab + sym->st_name,
".klp.sym.%55[^.].%127[^,],%lu",
sym_objname, sym_name, &sympos);
if (cnt != 3) {
pr_err("symbol %s has an incorrectly formatted name\n",
strtab + sym->st_name);
return -EINVAL;
}
sym_vmlinux = !strcmp(sym_objname, "vmlinux");
/*
* Prevent module-specific KLP rela sections from referencing
* vmlinux symbols. This helps prevent ordering issues with
* module special section initializations. Presumably such
* symbols are exported and normal relas can be used instead.
*/
if (!sec_vmlinux && sym_vmlinux) {
pr_err("invalid access to vmlinux symbol '%s' from module-specific livepatch relocation section",
sym_name);
return -EINVAL;
}
/* klp_find_object_symbol() treats a NULL objname as vmlinux */
ret = klp_find_object_symbol(sym_vmlinux ? NULL : sym_objname,
sym_name, sympos, &addr);
if (ret)
return ret;
sym->st_value = addr;
}
return 0;
}
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
/*
* At a high-level, there are two types of klp relocation sections: those which
* reference symbols which live in vmlinux; and those which reference symbols
* which live in other modules. This function is called for both types:
*
* 1) When a klp module itself loads, the module code calls this function to
* write vmlinux-specific klp relocations (.klp.rela.vmlinux.* sections).
* These relocations are written to the klp module text to allow the patched
* code/data to reference unexported vmlinux symbols. They're written as
* early as possible to ensure that other module init code (.e.g.,
* jump_label_apply_nops) can access any unexported vmlinux symbols which
* might be referenced by the klp module's special sections.
*
* 2) When a to-be-patched module loads -- or is already loaded when a
* corresponding klp module loads -- klp code calls this function to write
* module-specific klp relocations (.klp.rela.{module}.* sections). These
* are written to the klp module text to allow the patched code/data to
* reference symbols which live in the to-be-patched module or one of its
* module dependencies. Exported symbols are supported, in addition to
* unexported symbols, in order to enable late module patching, which allows
* the to-be-patched module to be loaded and patched sometime *after* the
* klp module is loaded.
*/
int klp_apply_section_relocs(struct module *pmod, Elf_Shdr *sechdrs,
const char *shstrtab, const char *strtab,
unsigned int symndx, unsigned int secndx,
const char *objname)
{
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
int cnt, ret;
char sec_objname[MODULE_NAME_LEN];
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
Elf_Shdr *sec = sechdrs + secndx;
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
/*
* Format: .klp.rela.sec_objname.section_name
* See comment in klp_resolve_symbols() for an explanation
* of the selected field width value.
*/
cnt = sscanf(shstrtab + sec->sh_name, ".klp.rela.%55[^.]",
sec_objname);
if (cnt != 1) {
pr_err("section %s has an incorrectly formatted name\n",
shstrtab + sec->sh_name);
return -EINVAL;
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
}
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
if (strcmp(objname ? objname : "vmlinux", sec_objname))
return 0;
ret = klp_resolve_symbols(sechdrs, strtab, symndx, sec, sec_objname);
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
if (ret)
return ret;
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
return apply_relocate_add(sechdrs, strtab, symndx, secndx, pmod);
}
/*
* Sysfs Interface
*
* /sys/kernel/livepatch
* /sys/kernel/livepatch/<patch>
* /sys/kernel/livepatch/<patch>/enabled
livepatch: change to a per-task consistency model Change livepatch to use a basic per-task consistency model. This is the foundation which will eventually enable us to patch those ~10% of security patches which change function or data semantics. This is the biggest remaining piece needed to make livepatch more generally useful. This code stems from the design proposal made by Vojtech [1] in November 2014. It's a hybrid of kGraft and kpatch: it uses kGraft's per-task consistency and syscall barrier switching combined with kpatch's stack trace switching. There are also a number of fallback options which make it quite flexible. Patches are applied on a per-task basis, when the task is deemed safe to switch over. When a patch is enabled, livepatch enters into a transition state where tasks are converging to the patched state. Usually this transition state can complete in a few seconds. The same sequence occurs when a patch is disabled, except the tasks converge from the patched state to the unpatched state. An interrupt handler inherits the patched state of the task it interrupts. The same is true for forked tasks: the child inherits the patched state of the parent. Livepatch uses several complementary approaches to determine when it's safe to patch tasks: 1. The first and most effective approach is stack checking of sleeping tasks. If no affected functions are on the stack of a given task, the task is patched. In most cases this will patch most or all of the tasks on the first try. Otherwise it'll keep trying periodically. This option is only available if the architecture has reliable stacks (HAVE_RELIABLE_STACKTRACE). 2. The second approach, if needed, is kernel exit switching. A task is switched when it returns to user space from a system call, a user space IRQ, or a signal. It's useful in the following cases: a) Patching I/O-bound user tasks which are sleeping on an affected function. In this case you have to send SIGSTOP and SIGCONT to force it to exit the kernel and be patched. b) Patching CPU-bound user tasks. If the task is highly CPU-bound then it will get patched the next time it gets interrupted by an IRQ. c) In the future it could be useful for applying patches for architectures which don't yet have HAVE_RELIABLE_STACKTRACE. In this case you would have to signal most of the tasks on the system. However this isn't supported yet because there's currently no way to patch kthreads without HAVE_RELIABLE_STACKTRACE. 3. For idle "swapper" tasks, since they don't ever exit the kernel, they instead have a klp_update_patch_state() call in the idle loop which allows them to be patched before the CPU enters the idle state. (Note there's not yet such an approach for kthreads.) All the above approaches may be skipped by setting the 'immediate' flag in the 'klp_patch' struct, which will disable per-task consistency and patch all tasks immediately. This can be useful if the patch doesn't change any function or data semantics. Note that, even with this flag set, it's possible that some tasks may still be running with an old version of the function, until that function returns. There's also an 'immediate' flag in the 'klp_func' struct which allows you to specify that certain functions in the patch can be applied without per-task consistency. This might be useful if you want to patch a common function like schedule(), and the function change doesn't need consistency but the rest of the patch does. For architectures which don't have HAVE_RELIABLE_STACKTRACE, the user must set patch->immediate which causes all tasks to be patched immediately. This option should be used with care, only when the patch doesn't change any function or data semantics. In the future, architectures which don't have HAVE_RELIABLE_STACKTRACE may be allowed to use per-task consistency if we can come up with another way to patch kthreads. The /sys/kernel/livepatch/<patch>/transition file shows whether a patch is in transition. Only a single patch (the topmost patch on the stack) can be in transition at a given time. A patch can remain in transition indefinitely, if any of the tasks are stuck in the initial patch state. A transition can be reversed and effectively canceled by writing the opposite value to the /sys/kernel/livepatch/<patch>/enabled file while the transition is in progress. Then all the tasks will attempt to converge back to the original patch state. [1] https://lkml.kernel.org/r/20141107140458.GA21774@suse.cz Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Ingo Molnar <mingo@kernel.org> # for the scheduler changes Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-02-14 09:42:40 +08:00
* /sys/kernel/livepatch/<patch>/transition
* /sys/kernel/livepatch/<patch>/force
* /sys/kernel/livepatch/<patch>/<object>
* /sys/kernel/livepatch/<patch>/<object>/<function,sympos>
*/
static int __klp_disable_patch(struct klp_patch *patch);
static ssize_t enabled_store(struct kobject *kobj, struct kobj_attribute *attr,
const char *buf, size_t count)
{
struct klp_patch *patch;
int ret;
bool enabled;
ret = kstrtobool(buf, &enabled);
if (ret)
return ret;
patch = container_of(kobj, struct klp_patch, kobj);
mutex_lock(&klp_mutex);
if (patch->enabled == enabled) {
/* already in requested state */
ret = -EINVAL;
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
goto out;
}
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
/*
* Allow to reverse a pending transition in both ways. It might be
* necessary to complete the transition without forcing and breaking
* the system integrity.
*
* Do not allow to re-enable a disabled patch.
*/
if (patch == klp_transition_patch)
livepatch: change to a per-task consistency model Change livepatch to use a basic per-task consistency model. This is the foundation which will eventually enable us to patch those ~10% of security patches which change function or data semantics. This is the biggest remaining piece needed to make livepatch more generally useful. This code stems from the design proposal made by Vojtech [1] in November 2014. It's a hybrid of kGraft and kpatch: it uses kGraft's per-task consistency and syscall barrier switching combined with kpatch's stack trace switching. There are also a number of fallback options which make it quite flexible. Patches are applied on a per-task basis, when the task is deemed safe to switch over. When a patch is enabled, livepatch enters into a transition state where tasks are converging to the patched state. Usually this transition state can complete in a few seconds. The same sequence occurs when a patch is disabled, except the tasks converge from the patched state to the unpatched state. An interrupt handler inherits the patched state of the task it interrupts. The same is true for forked tasks: the child inherits the patched state of the parent. Livepatch uses several complementary approaches to determine when it's safe to patch tasks: 1. The first and most effective approach is stack checking of sleeping tasks. If no affected functions are on the stack of a given task, the task is patched. In most cases this will patch most or all of the tasks on the first try. Otherwise it'll keep trying periodically. This option is only available if the architecture has reliable stacks (HAVE_RELIABLE_STACKTRACE). 2. The second approach, if needed, is kernel exit switching. A task is switched when it returns to user space from a system call, a user space IRQ, or a signal. It's useful in the following cases: a) Patching I/O-bound user tasks which are sleeping on an affected function. In this case you have to send SIGSTOP and SIGCONT to force it to exit the kernel and be patched. b) Patching CPU-bound user tasks. If the task is highly CPU-bound then it will get patched the next time it gets interrupted by an IRQ. c) In the future it could be useful for applying patches for architectures which don't yet have HAVE_RELIABLE_STACKTRACE. In this case you would have to signal most of the tasks on the system. However this isn't supported yet because there's currently no way to patch kthreads without HAVE_RELIABLE_STACKTRACE. 3. For idle "swapper" tasks, since they don't ever exit the kernel, they instead have a klp_update_patch_state() call in the idle loop which allows them to be patched before the CPU enters the idle state. (Note there's not yet such an approach for kthreads.) All the above approaches may be skipped by setting the 'immediate' flag in the 'klp_patch' struct, which will disable per-task consistency and patch all tasks immediately. This can be useful if the patch doesn't change any function or data semantics. Note that, even with this flag set, it's possible that some tasks may still be running with an old version of the function, until that function returns. There's also an 'immediate' flag in the 'klp_func' struct which allows you to specify that certain functions in the patch can be applied without per-task consistency. This might be useful if you want to patch a common function like schedule(), and the function change doesn't need consistency but the rest of the patch does. For architectures which don't have HAVE_RELIABLE_STACKTRACE, the user must set patch->immediate which causes all tasks to be patched immediately. This option should be used with care, only when the patch doesn't change any function or data semantics. In the future, architectures which don't have HAVE_RELIABLE_STACKTRACE may be allowed to use per-task consistency if we can come up with another way to patch kthreads. The /sys/kernel/livepatch/<patch>/transition file shows whether a patch is in transition. Only a single patch (the topmost patch on the stack) can be in transition at a given time. A patch can remain in transition indefinitely, if any of the tasks are stuck in the initial patch state. A transition can be reversed and effectively canceled by writing the opposite value to the /sys/kernel/livepatch/<patch>/enabled file while the transition is in progress. Then all the tasks will attempt to converge back to the original patch state. [1] https://lkml.kernel.org/r/20141107140458.GA21774@suse.cz Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Ingo Molnar <mingo@kernel.org> # for the scheduler changes Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-02-14 09:42:40 +08:00
klp_reverse_transition();
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
else if (!enabled)
ret = __klp_disable_patch(patch);
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
else
ret = -EINVAL;
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
out:
mutex_unlock(&klp_mutex);
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
if (ret)
return ret;
return count;
}
static ssize_t enabled_show(struct kobject *kobj,
struct kobj_attribute *attr, char *buf)
{
struct klp_patch *patch;
patch = container_of(kobj, struct klp_patch, kobj);
return snprintf(buf, PAGE_SIZE-1, "%d\n", patch->enabled);
}
livepatch: change to a per-task consistency model Change livepatch to use a basic per-task consistency model. This is the foundation which will eventually enable us to patch those ~10% of security patches which change function or data semantics. This is the biggest remaining piece needed to make livepatch more generally useful. This code stems from the design proposal made by Vojtech [1] in November 2014. It's a hybrid of kGraft and kpatch: it uses kGraft's per-task consistency and syscall barrier switching combined with kpatch's stack trace switching. There are also a number of fallback options which make it quite flexible. Patches are applied on a per-task basis, when the task is deemed safe to switch over. When a patch is enabled, livepatch enters into a transition state where tasks are converging to the patched state. Usually this transition state can complete in a few seconds. The same sequence occurs when a patch is disabled, except the tasks converge from the patched state to the unpatched state. An interrupt handler inherits the patched state of the task it interrupts. The same is true for forked tasks: the child inherits the patched state of the parent. Livepatch uses several complementary approaches to determine when it's safe to patch tasks: 1. The first and most effective approach is stack checking of sleeping tasks. If no affected functions are on the stack of a given task, the task is patched. In most cases this will patch most or all of the tasks on the first try. Otherwise it'll keep trying periodically. This option is only available if the architecture has reliable stacks (HAVE_RELIABLE_STACKTRACE). 2. The second approach, if needed, is kernel exit switching. A task is switched when it returns to user space from a system call, a user space IRQ, or a signal. It's useful in the following cases: a) Patching I/O-bound user tasks which are sleeping on an affected function. In this case you have to send SIGSTOP and SIGCONT to force it to exit the kernel and be patched. b) Patching CPU-bound user tasks. If the task is highly CPU-bound then it will get patched the next time it gets interrupted by an IRQ. c) In the future it could be useful for applying patches for architectures which don't yet have HAVE_RELIABLE_STACKTRACE. In this case you would have to signal most of the tasks on the system. However this isn't supported yet because there's currently no way to patch kthreads without HAVE_RELIABLE_STACKTRACE. 3. For idle "swapper" tasks, since they don't ever exit the kernel, they instead have a klp_update_patch_state() call in the idle loop which allows them to be patched before the CPU enters the idle state. (Note there's not yet such an approach for kthreads.) All the above approaches may be skipped by setting the 'immediate' flag in the 'klp_patch' struct, which will disable per-task consistency and patch all tasks immediately. This can be useful if the patch doesn't change any function or data semantics. Note that, even with this flag set, it's possible that some tasks may still be running with an old version of the function, until that function returns. There's also an 'immediate' flag in the 'klp_func' struct which allows you to specify that certain functions in the patch can be applied without per-task consistency. This might be useful if you want to patch a common function like schedule(), and the function change doesn't need consistency but the rest of the patch does. For architectures which don't have HAVE_RELIABLE_STACKTRACE, the user must set patch->immediate which causes all tasks to be patched immediately. This option should be used with care, only when the patch doesn't change any function or data semantics. In the future, architectures which don't have HAVE_RELIABLE_STACKTRACE may be allowed to use per-task consistency if we can come up with another way to patch kthreads. The /sys/kernel/livepatch/<patch>/transition file shows whether a patch is in transition. Only a single patch (the topmost patch on the stack) can be in transition at a given time. A patch can remain in transition indefinitely, if any of the tasks are stuck in the initial patch state. A transition can be reversed and effectively canceled by writing the opposite value to the /sys/kernel/livepatch/<patch>/enabled file while the transition is in progress. Then all the tasks will attempt to converge back to the original patch state. [1] https://lkml.kernel.org/r/20141107140458.GA21774@suse.cz Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Ingo Molnar <mingo@kernel.org> # for the scheduler changes Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-02-14 09:42:40 +08:00
static ssize_t transition_show(struct kobject *kobj,
struct kobj_attribute *attr, char *buf)
{
struct klp_patch *patch;
patch = container_of(kobj, struct klp_patch, kobj);
return snprintf(buf, PAGE_SIZE-1, "%d\n",
patch == klp_transition_patch);
}
static ssize_t force_store(struct kobject *kobj, struct kobj_attribute *attr,
const char *buf, size_t count)
{
struct klp_patch *patch;
int ret;
bool val;
ret = kstrtobool(buf, &val);
if (ret)
return ret;
if (!val)
return count;
mutex_lock(&klp_mutex);
patch = container_of(kobj, struct klp_patch, kobj);
if (patch != klp_transition_patch) {
mutex_unlock(&klp_mutex);
return -EINVAL;
}
klp_force_transition();
mutex_unlock(&klp_mutex);
return count;
}
static struct kobj_attribute enabled_kobj_attr = __ATTR_RW(enabled);
livepatch: change to a per-task consistency model Change livepatch to use a basic per-task consistency model. This is the foundation which will eventually enable us to patch those ~10% of security patches which change function or data semantics. This is the biggest remaining piece needed to make livepatch more generally useful. This code stems from the design proposal made by Vojtech [1] in November 2014. It's a hybrid of kGraft and kpatch: it uses kGraft's per-task consistency and syscall barrier switching combined with kpatch's stack trace switching. There are also a number of fallback options which make it quite flexible. Patches are applied on a per-task basis, when the task is deemed safe to switch over. When a patch is enabled, livepatch enters into a transition state where tasks are converging to the patched state. Usually this transition state can complete in a few seconds. The same sequence occurs when a patch is disabled, except the tasks converge from the patched state to the unpatched state. An interrupt handler inherits the patched state of the task it interrupts. The same is true for forked tasks: the child inherits the patched state of the parent. Livepatch uses several complementary approaches to determine when it's safe to patch tasks: 1. The first and most effective approach is stack checking of sleeping tasks. If no affected functions are on the stack of a given task, the task is patched. In most cases this will patch most or all of the tasks on the first try. Otherwise it'll keep trying periodically. This option is only available if the architecture has reliable stacks (HAVE_RELIABLE_STACKTRACE). 2. The second approach, if needed, is kernel exit switching. A task is switched when it returns to user space from a system call, a user space IRQ, or a signal. It's useful in the following cases: a) Patching I/O-bound user tasks which are sleeping on an affected function. In this case you have to send SIGSTOP and SIGCONT to force it to exit the kernel and be patched. b) Patching CPU-bound user tasks. If the task is highly CPU-bound then it will get patched the next time it gets interrupted by an IRQ. c) In the future it could be useful for applying patches for architectures which don't yet have HAVE_RELIABLE_STACKTRACE. In this case you would have to signal most of the tasks on the system. However this isn't supported yet because there's currently no way to patch kthreads without HAVE_RELIABLE_STACKTRACE. 3. For idle "swapper" tasks, since they don't ever exit the kernel, they instead have a klp_update_patch_state() call in the idle loop which allows them to be patched before the CPU enters the idle state. (Note there's not yet such an approach for kthreads.) All the above approaches may be skipped by setting the 'immediate' flag in the 'klp_patch' struct, which will disable per-task consistency and patch all tasks immediately. This can be useful if the patch doesn't change any function or data semantics. Note that, even with this flag set, it's possible that some tasks may still be running with an old version of the function, until that function returns. There's also an 'immediate' flag in the 'klp_func' struct which allows you to specify that certain functions in the patch can be applied without per-task consistency. This might be useful if you want to patch a common function like schedule(), and the function change doesn't need consistency but the rest of the patch does. For architectures which don't have HAVE_RELIABLE_STACKTRACE, the user must set patch->immediate which causes all tasks to be patched immediately. This option should be used with care, only when the patch doesn't change any function or data semantics. In the future, architectures which don't have HAVE_RELIABLE_STACKTRACE may be allowed to use per-task consistency if we can come up with another way to patch kthreads. The /sys/kernel/livepatch/<patch>/transition file shows whether a patch is in transition. Only a single patch (the topmost patch on the stack) can be in transition at a given time. A patch can remain in transition indefinitely, if any of the tasks are stuck in the initial patch state. A transition can be reversed and effectively canceled by writing the opposite value to the /sys/kernel/livepatch/<patch>/enabled file while the transition is in progress. Then all the tasks will attempt to converge back to the original patch state. [1] https://lkml.kernel.org/r/20141107140458.GA21774@suse.cz Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Ingo Molnar <mingo@kernel.org> # for the scheduler changes Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-02-14 09:42:40 +08:00
static struct kobj_attribute transition_kobj_attr = __ATTR_RO(transition);
static struct kobj_attribute force_kobj_attr = __ATTR_WO(force);
static struct attribute *klp_patch_attrs[] = {
&enabled_kobj_attr.attr,
livepatch: change to a per-task consistency model Change livepatch to use a basic per-task consistency model. This is the foundation which will eventually enable us to patch those ~10% of security patches which change function or data semantics. This is the biggest remaining piece needed to make livepatch more generally useful. This code stems from the design proposal made by Vojtech [1] in November 2014. It's a hybrid of kGraft and kpatch: it uses kGraft's per-task consistency and syscall barrier switching combined with kpatch's stack trace switching. There are also a number of fallback options which make it quite flexible. Patches are applied on a per-task basis, when the task is deemed safe to switch over. When a patch is enabled, livepatch enters into a transition state where tasks are converging to the patched state. Usually this transition state can complete in a few seconds. The same sequence occurs when a patch is disabled, except the tasks converge from the patched state to the unpatched state. An interrupt handler inherits the patched state of the task it interrupts. The same is true for forked tasks: the child inherits the patched state of the parent. Livepatch uses several complementary approaches to determine when it's safe to patch tasks: 1. The first and most effective approach is stack checking of sleeping tasks. If no affected functions are on the stack of a given task, the task is patched. In most cases this will patch most or all of the tasks on the first try. Otherwise it'll keep trying periodically. This option is only available if the architecture has reliable stacks (HAVE_RELIABLE_STACKTRACE). 2. The second approach, if needed, is kernel exit switching. A task is switched when it returns to user space from a system call, a user space IRQ, or a signal. It's useful in the following cases: a) Patching I/O-bound user tasks which are sleeping on an affected function. In this case you have to send SIGSTOP and SIGCONT to force it to exit the kernel and be patched. b) Patching CPU-bound user tasks. If the task is highly CPU-bound then it will get patched the next time it gets interrupted by an IRQ. c) In the future it could be useful for applying patches for architectures which don't yet have HAVE_RELIABLE_STACKTRACE. In this case you would have to signal most of the tasks on the system. However this isn't supported yet because there's currently no way to patch kthreads without HAVE_RELIABLE_STACKTRACE. 3. For idle "swapper" tasks, since they don't ever exit the kernel, they instead have a klp_update_patch_state() call in the idle loop which allows them to be patched before the CPU enters the idle state. (Note there's not yet such an approach for kthreads.) All the above approaches may be skipped by setting the 'immediate' flag in the 'klp_patch' struct, which will disable per-task consistency and patch all tasks immediately. This can be useful if the patch doesn't change any function or data semantics. Note that, even with this flag set, it's possible that some tasks may still be running with an old version of the function, until that function returns. There's also an 'immediate' flag in the 'klp_func' struct which allows you to specify that certain functions in the patch can be applied without per-task consistency. This might be useful if you want to patch a common function like schedule(), and the function change doesn't need consistency but the rest of the patch does. For architectures which don't have HAVE_RELIABLE_STACKTRACE, the user must set patch->immediate which causes all tasks to be patched immediately. This option should be used with care, only when the patch doesn't change any function or data semantics. In the future, architectures which don't have HAVE_RELIABLE_STACKTRACE may be allowed to use per-task consistency if we can come up with another way to patch kthreads. The /sys/kernel/livepatch/<patch>/transition file shows whether a patch is in transition. Only a single patch (the topmost patch on the stack) can be in transition at a given time. A patch can remain in transition indefinitely, if any of the tasks are stuck in the initial patch state. A transition can be reversed and effectively canceled by writing the opposite value to the /sys/kernel/livepatch/<patch>/enabled file while the transition is in progress. Then all the tasks will attempt to converge back to the original patch state. [1] https://lkml.kernel.org/r/20141107140458.GA21774@suse.cz Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Ingo Molnar <mingo@kernel.org> # for the scheduler changes Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-02-14 09:42:40 +08:00
&transition_kobj_attr.attr,
&force_kobj_attr.attr,
NULL
};
ATTRIBUTE_GROUPS(klp_patch);
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
static void klp_free_object_dynamic(struct klp_object *obj)
{
kfree(obj->name);
kfree(obj);
}
static void klp_init_func_early(struct klp_object *obj,
struct klp_func *func);
static void klp_init_object_early(struct klp_patch *patch,
struct klp_object *obj);
static struct klp_object *klp_alloc_object_dynamic(const char *name,
struct klp_patch *patch)
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
{
struct klp_object *obj;
obj = kzalloc(sizeof(*obj), GFP_KERNEL);
if (!obj)
return NULL;
if (name) {
obj->name = kstrdup(name, GFP_KERNEL);
if (!obj->name) {
kfree(obj);
return NULL;
}
}
klp_init_object_early(patch, obj);
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
obj->dynamic = true;
return obj;
}
static void klp_free_func_nop(struct klp_func *func)
{
kfree(func->old_name);
kfree(func);
}
static struct klp_func *klp_alloc_func_nop(struct klp_func *old_func,
struct klp_object *obj)
{
struct klp_func *func;
func = kzalloc(sizeof(*func), GFP_KERNEL);
if (!func)
return NULL;
if (old_func->old_name) {
func->old_name = kstrdup(old_func->old_name, GFP_KERNEL);
if (!func->old_name) {
kfree(func);
return NULL;
}
}
klp_init_func_early(obj, func);
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
/*
* func->new_func is same as func->old_func. These addresses are
* set when the object is loaded, see klp_init_object_loaded().
*/
func->old_sympos = old_func->old_sympos;
func->nop = true;
return func;
}
static int klp_add_object_nops(struct klp_patch *patch,
struct klp_object *old_obj)
{
struct klp_object *obj;
struct klp_func *func, *old_func;
obj = klp_find_object(patch, old_obj);
if (!obj) {
obj = klp_alloc_object_dynamic(old_obj->name, patch);
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
if (!obj)
return -ENOMEM;
}
klp_for_each_func(old_obj, old_func) {
func = klp_find_func(obj, old_func);
if (func)
continue;
func = klp_alloc_func_nop(old_func, obj);
if (!func)
return -ENOMEM;
}
return 0;
}
/*
* Add 'nop' functions which simply return to the caller to run
* the original function. The 'nop' functions are added to a
* patch to facilitate a 'replace' mode.
*/
static int klp_add_nops(struct klp_patch *patch)
{
struct klp_patch *old_patch;
struct klp_object *old_obj;
klp_for_each_patch(old_patch) {
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
klp_for_each_object(old_patch, old_obj) {
int err;
err = klp_add_object_nops(patch, old_obj);
if (err)
return err;
}
}
return 0;
}
static void klp_kobj_release_patch(struct kobject *kobj)
{
livepatch: allow removal of a disabled patch Currently we do not allow patch module to unload since there is no method to determine if a task is still running in the patched code. The consistency model gives us the way because when the unpatching finishes we know that all tasks were marked as safe to call an original function. Thus every new call to the function calls the original code and at the same time no task can be somewhere in the patched code, because it had to leave that code to be marked as safe. We can safely let the patch module go after that. Completion is used for synchronization between module removal and sysfs infrastructure in a similar way to commit 942e443127e9 ("module: Fix mod->mkobj.kobj potentially freed too early"). Note that we still do not allow the removal for immediate model, that is no consistency model. The module refcount may increase in this case if somebody disables and enables the patch several times. This should not cause any harm. With this change a call to try_module_get() is moved to __klp_enable_patch from klp_register_patch to make module reference counting symmetric (module_put() is in a patch disable path) and to allow to take a new reference to a disabled module when being enabled. Finally, we need to be very careful about possible races between klp_unregister_patch(), kobject_put() functions and operations on the related sysfs files. kobject_put(&patch->kobj) must be called without klp_mutex. Otherwise, it might be blocked by enabled_store() that needs the mutex as well. In addition, enabled_store() must check if the patch was not unregisted in the meantime. There is no need to do the same for other kobject_put() callsites at the moment. Their sysfs operations neither take the lock nor they access any data that might be freed in the meantime. There was an attempt to use kobjects the right way and prevent these races by design. But it made the patch definition more complicated and opened another can of worms. See https://lkml.kernel.org/r/1464018848-4303-1-git-send-email-pmladek@suse.com [Thanks to Petr Mladek for improving the commit message.] Signed-off-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-03-07 01:20:29 +08:00
struct klp_patch *patch;
patch = container_of(kobj, struct klp_patch, kobj);
complete(&patch->finish);
}
static struct kobj_type klp_ktype_patch = {
.release = klp_kobj_release_patch,
.sysfs_ops = &kobj_sysfs_ops,
.default_groups = klp_patch_groups,
};
static void klp_kobj_release_object(struct kobject *kobj)
{
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
struct klp_object *obj;
obj = container_of(kobj, struct klp_object, kobj);
if (obj->dynamic)
klp_free_object_dynamic(obj);
}
static struct kobj_type klp_ktype_object = {
.release = klp_kobj_release_object,
.sysfs_ops = &kobj_sysfs_ops,
};
static void klp_kobj_release_func(struct kobject *kobj)
{
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
struct klp_func *func;
func = container_of(kobj, struct klp_func, kobj);
if (func->nop)
klp_free_func_nop(func);
}
static struct kobj_type klp_ktype_func = {
.release = klp_kobj_release_func,
.sysfs_ops = &kobj_sysfs_ops,
};
livepatch: Remove Nop structures when unused Replaced patches are removed from the stack when the transition is finished. It means that Nop structures will never be needed again and can be removed. Why should we care? + Nop structures give the impression that the function is patched even though the ftrace handler has no effect. + Ftrace handlers do not come for free. They cause slowdown that might be visible in some workloads. The ftrace-related slowdown might actually be the reason why the function is no longer patched in the new cumulative patch. One would expect that cumulative patch would help solve these problems as well. + Cumulative patches are supposed to replace any earlier version of the patch. The amount of NOPs depends on which version was replaced. This multiplies the amount of scenarios that might happen. One might say that NOPs are innocent. But there are even optimized NOP instructions for different processors, for example, see arch/x86/kernel/alternative.c. And klp_ftrace_handler() is much more complicated. + It sounds natural to clean up a mess that is no longer needed. It could only be worse if we do not do it. This patch allows to unpatch and free the dynamic structures independently when the transition finishes. The free part is a bit tricky because kobject free callbacks are called asynchronously. We could not wait for them easily. Fortunately, we do not have to. Any further access can be avoided by removing them from the dynamic lists. Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:26 +08:00
static void __klp_free_funcs(struct klp_object *obj, bool nops_only)
{
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
struct klp_func *func, *tmp_func;
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
klp_for_each_func_safe(obj, func, tmp_func) {
livepatch: Remove Nop structures when unused Replaced patches are removed from the stack when the transition is finished. It means that Nop structures will never be needed again and can be removed. Why should we care? + Nop structures give the impression that the function is patched even though the ftrace handler has no effect. + Ftrace handlers do not come for free. They cause slowdown that might be visible in some workloads. The ftrace-related slowdown might actually be the reason why the function is no longer patched in the new cumulative patch. One would expect that cumulative patch would help solve these problems as well. + Cumulative patches are supposed to replace any earlier version of the patch. The amount of NOPs depends on which version was replaced. This multiplies the amount of scenarios that might happen. One might say that NOPs are innocent. But there are even optimized NOP instructions for different processors, for example, see arch/x86/kernel/alternative.c. And klp_ftrace_handler() is much more complicated. + It sounds natural to clean up a mess that is no longer needed. It could only be worse if we do not do it. This patch allows to unpatch and free the dynamic structures independently when the transition finishes. The free part is a bit tricky because kobject free callbacks are called asynchronously. We could not wait for them easily. Fortunately, we do not have to. Any further access can be avoided by removing them from the dynamic lists. Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:26 +08:00
if (nops_only && !func->nop)
continue;
list_del(&func->node);
kobject_put(&func->kobj);
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
}
}
/* Clean up when a patched object is unloaded */
static void klp_free_object_loaded(struct klp_object *obj)
{
struct klp_func *func;
obj->mod = NULL;
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
klp_for_each_func(obj, func) {
func->old_func = NULL;
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
if (func->nop)
func->new_func = NULL;
}
}
livepatch: Remove Nop structures when unused Replaced patches are removed from the stack when the transition is finished. It means that Nop structures will never be needed again and can be removed. Why should we care? + Nop structures give the impression that the function is patched even though the ftrace handler has no effect. + Ftrace handlers do not come for free. They cause slowdown that might be visible in some workloads. The ftrace-related slowdown might actually be the reason why the function is no longer patched in the new cumulative patch. One would expect that cumulative patch would help solve these problems as well. + Cumulative patches are supposed to replace any earlier version of the patch. The amount of NOPs depends on which version was replaced. This multiplies the amount of scenarios that might happen. One might say that NOPs are innocent. But there are even optimized NOP instructions for different processors, for example, see arch/x86/kernel/alternative.c. And klp_ftrace_handler() is much more complicated. + It sounds natural to clean up a mess that is no longer needed. It could only be worse if we do not do it. This patch allows to unpatch and free the dynamic structures independently when the transition finishes. The free part is a bit tricky because kobject free callbacks are called asynchronously. We could not wait for them easily. Fortunately, we do not have to. Any further access can be avoided by removing them from the dynamic lists. Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:26 +08:00
static void __klp_free_objects(struct klp_patch *patch, bool nops_only)
{
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
struct klp_object *obj, *tmp_obj;
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
klp_for_each_object_safe(patch, obj, tmp_obj) {
livepatch: Remove Nop structures when unused Replaced patches are removed from the stack when the transition is finished. It means that Nop structures will never be needed again and can be removed. Why should we care? + Nop structures give the impression that the function is patched even though the ftrace handler has no effect. + Ftrace handlers do not come for free. They cause slowdown that might be visible in some workloads. The ftrace-related slowdown might actually be the reason why the function is no longer patched in the new cumulative patch. One would expect that cumulative patch would help solve these problems as well. + Cumulative patches are supposed to replace any earlier version of the patch. The amount of NOPs depends on which version was replaced. This multiplies the amount of scenarios that might happen. One might say that NOPs are innocent. But there are even optimized NOP instructions for different processors, for example, see arch/x86/kernel/alternative.c. And klp_ftrace_handler() is much more complicated. + It sounds natural to clean up a mess that is no longer needed. It could only be worse if we do not do it. This patch allows to unpatch and free the dynamic structures independently when the transition finishes. The free part is a bit tricky because kobject free callbacks are called asynchronously. We could not wait for them easily. Fortunately, we do not have to. Any further access can be avoided by removing them from the dynamic lists. Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:26 +08:00
__klp_free_funcs(obj, nops_only);
if (nops_only && !obj->dynamic)
continue;
list_del(&obj->node);
kobject_put(&obj->kobj);
}
}
livepatch: Remove Nop structures when unused Replaced patches are removed from the stack when the transition is finished. It means that Nop structures will never be needed again and can be removed. Why should we care? + Nop structures give the impression that the function is patched even though the ftrace handler has no effect. + Ftrace handlers do not come for free. They cause slowdown that might be visible in some workloads. The ftrace-related slowdown might actually be the reason why the function is no longer patched in the new cumulative patch. One would expect that cumulative patch would help solve these problems as well. + Cumulative patches are supposed to replace any earlier version of the patch. The amount of NOPs depends on which version was replaced. This multiplies the amount of scenarios that might happen. One might say that NOPs are innocent. But there are even optimized NOP instructions for different processors, for example, see arch/x86/kernel/alternative.c. And klp_ftrace_handler() is much more complicated. + It sounds natural to clean up a mess that is no longer needed. It could only be worse if we do not do it. This patch allows to unpatch and free the dynamic structures independently when the transition finishes. The free part is a bit tricky because kobject free callbacks are called asynchronously. We could not wait for them easily. Fortunately, we do not have to. Any further access can be avoided by removing them from the dynamic lists. Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:26 +08:00
static void klp_free_objects(struct klp_patch *patch)
{
__klp_free_objects(patch, false);
}
static void klp_free_objects_dynamic(struct klp_patch *patch)
{
__klp_free_objects(patch, true);
}
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
/*
* This function implements the free operations that can be called safely
* under klp_mutex.
*
* The operation must be completed by calling klp_free_patch_finish()
* outside klp_mutex.
*/
static void klp_free_patch_start(struct klp_patch *patch)
{
if (!list_empty(&patch->list))
list_del(&patch->list);
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
klp_free_objects(patch);
}
/*
* This function implements the free part that must be called outside
* klp_mutex.
*
* It must be called after klp_free_patch_start(). And it has to be
* the last function accessing the livepatch structures when the patch
* gets disabled.
*/
static void klp_free_patch_finish(struct klp_patch *patch)
{
/*
* Avoid deadlock with enabled_store() sysfs callback by
* calling this outside klp_mutex. It is safe because
* this is called when the patch gets disabled and it
* cannot get enabled again.
*/
kobject_put(&patch->kobj);
wait_for_completion(&patch->finish);
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
/* Put the module after the last access to struct klp_patch. */
if (!patch->forced)
module_put(patch->mod);
}
/*
* The livepatch might be freed from sysfs interface created by the patch.
* This work allows to wait until the interface is destroyed in a separate
* context.
*/
static void klp_free_patch_work_fn(struct work_struct *work)
{
struct klp_patch *patch =
container_of(work, struct klp_patch, free_work);
klp_free_patch_finish(patch);
}
void klp_free_patch_async(struct klp_patch *patch)
{
klp_free_patch_start(patch);
schedule_work(&patch->free_work);
}
void klp_free_replaced_patches_async(struct klp_patch *new_patch)
{
struct klp_patch *old_patch, *tmp_patch;
klp_for_each_patch_safe(old_patch, tmp_patch) {
if (old_patch == new_patch)
return;
klp_free_patch_async(old_patch);
}
}
static int klp_init_func(struct klp_object *obj, struct klp_func *func)
{
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
if (!func->old_name)
return -EINVAL;
/*
* NOPs get the address later. The patched module must be loaded,
* see klp_init_object_loaded().
*/
if (!func->new_func && !func->nop)
return -EINVAL;
if (strlen(func->old_name) >= KSYM_NAME_LEN)
return -EINVAL;
INIT_LIST_HEAD(&func->stack_node);
func->patched = false;
livepatch: change to a per-task consistency model Change livepatch to use a basic per-task consistency model. This is the foundation which will eventually enable us to patch those ~10% of security patches which change function or data semantics. This is the biggest remaining piece needed to make livepatch more generally useful. This code stems from the design proposal made by Vojtech [1] in November 2014. It's a hybrid of kGraft and kpatch: it uses kGraft's per-task consistency and syscall barrier switching combined with kpatch's stack trace switching. There are also a number of fallback options which make it quite flexible. Patches are applied on a per-task basis, when the task is deemed safe to switch over. When a patch is enabled, livepatch enters into a transition state where tasks are converging to the patched state. Usually this transition state can complete in a few seconds. The same sequence occurs when a patch is disabled, except the tasks converge from the patched state to the unpatched state. An interrupt handler inherits the patched state of the task it interrupts. The same is true for forked tasks: the child inherits the patched state of the parent. Livepatch uses several complementary approaches to determine when it's safe to patch tasks: 1. The first and most effective approach is stack checking of sleeping tasks. If no affected functions are on the stack of a given task, the task is patched. In most cases this will patch most or all of the tasks on the first try. Otherwise it'll keep trying periodically. This option is only available if the architecture has reliable stacks (HAVE_RELIABLE_STACKTRACE). 2. The second approach, if needed, is kernel exit switching. A task is switched when it returns to user space from a system call, a user space IRQ, or a signal. It's useful in the following cases: a) Patching I/O-bound user tasks which are sleeping on an affected function. In this case you have to send SIGSTOP and SIGCONT to force it to exit the kernel and be patched. b) Patching CPU-bound user tasks. If the task is highly CPU-bound then it will get patched the next time it gets interrupted by an IRQ. c) In the future it could be useful for applying patches for architectures which don't yet have HAVE_RELIABLE_STACKTRACE. In this case you would have to signal most of the tasks on the system. However this isn't supported yet because there's currently no way to patch kthreads without HAVE_RELIABLE_STACKTRACE. 3. For idle "swapper" tasks, since they don't ever exit the kernel, they instead have a klp_update_patch_state() call in the idle loop which allows them to be patched before the CPU enters the idle state. (Note there's not yet such an approach for kthreads.) All the above approaches may be skipped by setting the 'immediate' flag in the 'klp_patch' struct, which will disable per-task consistency and patch all tasks immediately. This can be useful if the patch doesn't change any function or data semantics. Note that, even with this flag set, it's possible that some tasks may still be running with an old version of the function, until that function returns. There's also an 'immediate' flag in the 'klp_func' struct which allows you to specify that certain functions in the patch can be applied without per-task consistency. This might be useful if you want to patch a common function like schedule(), and the function change doesn't need consistency but the rest of the patch does. For architectures which don't have HAVE_RELIABLE_STACKTRACE, the user must set patch->immediate which causes all tasks to be patched immediately. This option should be used with care, only when the patch doesn't change any function or data semantics. In the future, architectures which don't have HAVE_RELIABLE_STACKTRACE may be allowed to use per-task consistency if we can come up with another way to patch kthreads. The /sys/kernel/livepatch/<patch>/transition file shows whether a patch is in transition. Only a single patch (the topmost patch on the stack) can be in transition at a given time. A patch can remain in transition indefinitely, if any of the tasks are stuck in the initial patch state. A transition can be reversed and effectively canceled by writing the opposite value to the /sys/kernel/livepatch/<patch>/enabled file while the transition is in progress. Then all the tasks will attempt to converge back to the original patch state. [1] https://lkml.kernel.org/r/20141107140458.GA21774@suse.cz Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Ingo Molnar <mingo@kernel.org> # for the scheduler changes Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-02-14 09:42:40 +08:00
func->transition = false;
/* The format for the sysfs directory is <function,sympos> where sympos
* is the nth occurrence of this symbol in kallsyms for the patched
* object. If the user selects 0 for old_sympos, then 1 will be used
* since a unique symbol will be the first occurrence.
*/
return kobject_add(&func->kobj, &obj->kobj, "%s,%lu",
func->old_name,
func->old_sympos ? func->old_sympos : 1);
}
static int klp_apply_object_relocs(struct klp_patch *patch,
struct klp_object *obj)
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
{
int i, ret;
struct klp_modinfo *info = patch->mod->klp_info;
for (i = 1; i < info->hdr.e_shnum; i++) {
Elf_Shdr *sec = info->sechdrs + i;
if (!(sec->sh_flags & SHF_RELA_LIVEPATCH))
continue;
ret = klp_apply_section_relocs(patch->mod, info->sechdrs,
info->secstrings,
patch->mod->core_kallsyms.strtab,
info->symndx, i, obj->name);
if (ret)
return ret;
}
return 0;
}
/* parts of the initialization that is done only when the object is loaded */
static int klp_init_object_loaded(struct klp_patch *patch,
struct klp_object *obj)
{
struct klp_func *func;
int ret;
livepatch: Apply vmlinux-specific KLP relocations early KLP relocations are livepatch-specific relocations which are applied to a KLP module's text or data. They exist for two reasons: 1) Unexported symbols: replacement functions often need to access unexported symbols (e.g. static functions), which "normal" relocations don't allow. 2) Late module patching: this is the ability for a KLP module to bypass normal module dependencies, such that the KLP module can be loaded *before* a to-be-patched module. This means that relocations which need to access symbols in the to-be-patched module might need to be applied to the KLP module well after it has been loaded. Non-late-patched KLP relocations are applied from the KLP module's init function. That usually works fine, unless the patched code wants to use alternatives, paravirt patching, jump tables, or some other special section which needs relocations. Then we run into ordering issues and crashes. In order for those special sections to work properly, the KLP relocations should be applied *before* the special section init code runs, such as apply_paravirt(), apply_alternatives(), or jump_label_apply_nops(). You might think the obvious solution would be to move the KLP relocation initialization earlier, but it's not necessarily that simple. The problem is the above-mentioned late module patching, for which KLP relocations can get applied well after the KLP module is loaded. To "fix" this issue in the past, we created .klp.arch sections: .klp.arch.{module}..altinstructions .klp.arch.{module}..parainstructions Those sections allow KLP late module patching code to call apply_paravirt() and apply_alternatives() after the module-specific KLP relocations (.klp.rela.{module}.{section}) have been applied. But that has a lot of drawbacks, including code complexity, the need for arch-specific code, and the (per-arch) danger that we missed some special section -- for example the __jump_table section which is used for jump labels. It turns out there's a simpler and more functional approach. There are two kinds of KLP relocation sections: 1) vmlinux-specific KLP relocation sections .klp.rela.vmlinux.{sec} These are relocations (applied to the KLP module) which reference unexported vmlinux symbols. 2) module-specific KLP relocation sections .klp.rela.{module}.{sec}: These are relocations (applied to the KLP module) which reference unexported or exported module symbols. Up until now, these have been treated the same. However, they're inherently different. Because of late module patching, module-specific KLP relocations can be applied very late, thus they can create the ordering headaches described above. But vmlinux-specific KLP relocations don't have that problem. There's nothing to prevent them from being applied earlier. So apply them at the same time as normal relocations, when the KLP module is being loaded. This means that for vmlinux-specific KLP relocations, we no longer have any ordering issues. vmlinux-referencing jump labels, alternatives, and paravirt patching will work automatically, without the need for the .klp.arch hacks. All that said, for module-specific KLP relocations, the ordering problems still exist and we *do* still need .klp.arch. Or do we? Stay tuned. Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:44 +08:00
if (klp_is_module(obj)) {
/*
* Only write module-specific relocations here
* (.klp.rela.{module}.*). vmlinux-specific relocations were
* written earlier during the initialization of the klp module
* itself.
*/
ret = klp_apply_object_relocs(patch, obj);
livepatch: Remove .klp.arch After the previous patch, vmlinux-specific KLP relocations are now applied early during KLP module load. This means that .klp.arch sections are no longer needed for *vmlinux-specific* KLP relocations. One might think they're still needed for *module-specific* KLP relocations. If a to-be-patched module is loaded *after* its corresponding KLP module is loaded, any corresponding KLP relocations will be delayed until the to-be-patched module is loaded. If any special sections (.parainstructions, for example) rely on those relocations, their initializations (apply_paravirt) need to be done afterwards. Thus the apparent need for arch_klp_init_object_loaded() and its corresponding .klp.arch sections -- it allows some of the special section initializations to be done at a later time. But... if you look closer, that dependency between the special sections and the module-specific KLP relocations doesn't actually exist in reality. Looking at the contents of the .altinstructions and .parainstructions sections, there's not a realistic scenario in which a KLP module's .altinstructions or .parainstructions section needs to access a symbol in a to-be-patched module. It might need to access a local symbol or even a vmlinux symbol; but not another module's symbol. When a special section needs to reference a local or vmlinux symbol, a normal rela can be used instead of a KLP rela. Since the special section initializations don't actually have any real dependency on module-specific KLP relocations, .klp.arch and arch_klp_init_object_loaded() no longer have a reason to exist. So remove them. As Peter said much more succinctly: So the reason for .klp.arch was that .klp.rela.* stuff would overwrite paravirt instructions. If that happens you're doing it wrong. Those RELAs are core kernel, not module, and thus should've happened in .rela.* sections at patch-module loading time. Reverting this removes the two apply_{paravirt,alternatives}() calls from the late patching path, and means we don't have to worry about them when removing module_disable_ro(). [ jpoimboe: Rewrote patch description. Tweaked klp_init_object_loaded() error path. ] Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2020-04-29 23:24:45 +08:00
if (ret)
return ret;
}
module: Fix livepatch/ftrace module text permissions race It's possible for livepatch and ftrace to be toggling a module's text permissions at the same time, resulting in the following panic: BUG: unable to handle page fault for address: ffffffffc005b1d9 #PF: supervisor write access in kernel mode #PF: error_code(0x0003) - permissions violation PGD 3ea0c067 P4D 3ea0c067 PUD 3ea0e067 PMD 3cc13067 PTE 3b8a1061 Oops: 0003 [#1] PREEMPT SMP PTI CPU: 1 PID: 453 Comm: insmod Tainted: G O K 5.2.0-rc1-a188339ca5 #1 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-20181126_142135-anatol 04/01/2014 RIP: 0010:apply_relocate_add+0xbe/0x14c Code: fa 0b 74 21 48 83 fa 18 74 38 48 83 fa 0a 75 40 eb 08 48 83 38 00 74 33 eb 53 83 38 00 75 4e 89 08 89 c8 eb 0a 83 38 00 75 43 <89> 08 48 63 c1 48 39 c8 74 2e eb 48 83 38 00 75 32 48 29 c1 89 08 RSP: 0018:ffffb223c00dbb10 EFLAGS: 00010246 RAX: ffffffffc005b1d9 RBX: 0000000000000000 RCX: ffffffff8b200060 RDX: 000000000000000b RSI: 0000004b0000000b RDI: ffff96bdfcd33000 RBP: ffffb223c00dbb38 R08: ffffffffc005d040 R09: ffffffffc005c1f0 R10: ffff96bdfcd33c40 R11: ffff96bdfcd33b80 R12: 0000000000000018 R13: ffffffffc005c1f0 R14: ffffffffc005e708 R15: ffffffff8b2fbc74 FS: 00007f5f447beba8(0000) GS:ffff96bdff900000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: ffffffffc005b1d9 CR3: 000000003cedc002 CR4: 0000000000360ea0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: klp_init_object_loaded+0x10f/0x219 ? preempt_latency_start+0x21/0x57 klp_enable_patch+0x662/0x809 ? virt_to_head_page+0x3a/0x3c ? kfree+0x8c/0x126 patch_init+0x2ed/0x1000 [livepatch_test02] ? 0xffffffffc0060000 do_one_initcall+0x9f/0x1c5 ? kmem_cache_alloc_trace+0xc4/0xd4 ? do_init_module+0x27/0x210 do_init_module+0x5f/0x210 load_module+0x1c41/0x2290 ? fsnotify_path+0x3b/0x42 ? strstarts+0x2b/0x2b ? kernel_read+0x58/0x65 __do_sys_finit_module+0x9f/0xc3 ? __do_sys_finit_module+0x9f/0xc3 __x64_sys_finit_module+0x1a/0x1c do_syscall_64+0x52/0x61 entry_SYSCALL_64_after_hwframe+0x44/0xa9 The above panic occurs when loading two modules at the same time with ftrace enabled, where at least one of the modules is a livepatch module: CPU0 CPU1 klp_enable_patch() klp_init_object_loaded() module_disable_ro() ftrace_module_enable() ftrace_arch_code_modify_post_process() set_all_modules_text_ro() klp_write_object_relocations() apply_relocate_add() *patches read-only code* - BOOM A similar race exists when toggling ftrace while loading a livepatch module. Fix it by ensuring that the livepatch and ftrace code patching operations -- and their respective permissions changes -- are protected by the text_mutex. Link: http://lkml.kernel.org/r/ab43d56ab909469ac5d2520c5d944ad6d4abd476.1560474114.git.jpoimboe@redhat.com Reported-by: Johannes Erdfelt <johannes@erdfelt.com> Fixes: 444d13ff10fb ("modules: add ro_after_init support") Acked-by: Jessica Yu <jeyu@kernel.org> Reviewed-by: Petr Mladek <pmladek@suse.com> Reviewed-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
2019-06-14 09:07:22 +08:00
klp_for_each_func(obj, func) {
livepatch: add old_sympos as disambiguator field to klp_func Currently, patching objects with duplicate symbol names fail because the creation of the sysfs function directory collides with the previous attempt. Appending old_addr to the function name is problematic as it reveals the address of the function being patch to a normal user. Using the symbol's occurrence in kallsyms to postfix the function name in the sysfs directory solves the issue of having consistent unique names and ensuring that the address is not exposed to a normal user. In addition, using the symbol position as the user's method to disambiguate symbols instead of addr allows for disambiguating symbols in modules as well for both function addresses and for relocs. This also simplifies much of the code. Special handling for kASLR is no longer needed and can be removed. The klp_find_verify_func_addr function can be replaced by klp_find_object_symbol, and klp_verify_vmlinux_symbol and its callback can be removed completely. In cases of duplicate symbols, old_sympos will be used to disambiguate instead of old_addr. By default old_sympos will be 0, and patching will only succeed if the symbol is unique. Specifying a positive value will ensure that occurrence of the symbol in kallsyms for the patched object will be used for patching if it is valid. In addition, make old_addr an internal structure field not to be specified by the user. Finally, remove klp_find_verify_func_addr as it can be replaced by klp_find_object_symbol directly. Support for symbol position disambiguation for relocations is added in the next patch in this series. Signed-off-by: Chris J Arges <chris.j.arges@canonical.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-12-02 10:40:54 +08:00
ret = klp_find_object_symbol(obj->name, func->old_name,
func->old_sympos,
(unsigned long *)&func->old_func);
if (ret)
return ret;
ret = kallsyms_lookup_size_offset((unsigned long)func->old_func,
&func->old_size, NULL);
if (!ret) {
pr_err("kallsyms size lookup failed for '%s'\n",
func->old_name);
return -ENOENT;
}
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
if (func->nop)
func->new_func = func->old_func;
ret = kallsyms_lookup_size_offset((unsigned long)func->new_func,
&func->new_size, NULL);
if (!ret) {
pr_err("kallsyms size lookup failed for '%s' replacement\n",
func->old_name);
return -ENOENT;
}
}
return 0;
}
static int klp_init_object(struct klp_patch *patch, struct klp_object *obj)
{
struct klp_func *func;
int ret;
const char *name;
if (klp_is_module(obj) && strlen(obj->name) >= MODULE_NAME_LEN)
return -EINVAL;
obj->patched = false;
livepatch: Fix subtle race with coming and going modules There is a notifier that handles live patches for coming and going modules. It takes klp_mutex lock to avoid races with coming and going patches but it does not keep the lock all the time. Therefore the following races are possible: 1. The notifier is called sometime in STATE_MODULE_COMING. The module is visible by find_module() in this state all the time. It means that new patch can be registered and enabled even before the notifier is called. It might create wrong order of stacked patches, see below for an example. 2. New patch could still see the module in the GOING state even after the notifier has been called. It will try to initialize the related object structures but the module could disappear at any time. There will stay mess in the structures. It might even cause an invalid memory access. This patch solves the problem by adding a boolean variable into struct module. The value is true after the coming and before the going handler is called. New patches need to be applied when the value is true and they need to ignore the module when the value is false. Note that we need to know state of all modules on the system. The races are related to new patches. Therefore we do not know what modules will get patched. Also note that we could not simply ignore going modules. The code from the module could be called even in the GOING state until mod->exit() finishes. If we start supporting patches with semantic changes between function calls, we need to apply new patches to any still usable code. See below for an example. Finally note that the patch solves only the situation when a new patch is registered. There are no such problems when the patch is being removed. It does not matter who disable the patch first, whether the normal disable_patch() or the module notifier. There is nothing to do once the patch is disabled. Alternative solutions: ====================== + reject new patches when a patched module is coming or going; this is ugly + wait with adding new patch until the module leaves the COMING and GOING states; this might be dangerous and complicated; we would need to release kgr_lock in the middle of the patch registration to avoid a deadlock with the coming and going handlers; also we might need a waitqueue for each module which seems to be even bigger overhead than the boolean + stop modules from entering COMING and GOING states; wait until modules leave these states when they are already there; looks complicated; we would need to ignore the module that asked to stop the others to avoid a deadlock; also it is unclear what to do when two modules asked to stop others and both are in COMING state (situation when two new patches are applied) + always register/enable new patches and fix up the potential mess (registered patches order) in klp_module_init(); this is nasty and prone to regressions in the future development + add another MODULE_STATE where the kallsyms are visible but the module is not used yet; this looks too complex; the module states are checked on "many" locations Example of patch stacking breakage: =================================== The notifier could _not_ _simply_ ignore already initialized module objects. For example, let's have three patches (P1, P2, P3) for functions a() and b() where a() is from vmcore and b() is from a module M. Something like: a() b() P1 a1() b1() P2 a2() b2() P3 a3() b3(3) If you load the module M after all patches are registered and enabled. The ftrace ops for function a() and b() has listed the functions in this order: ops_a->func_stack -> list(a3,a2,a1) ops_b->func_stack -> list(b3,b2,b1) , so the pointer to b3() is the first and will be used. Then you might have the following scenario. Let's start with state when patches P1 and P2 are registered and enabled but the module M is not loaded. Then ftrace ops for b() does not exist. Then we get into the following race: CPU0 CPU1 load_module(M) complete_formation() mod->state = MODULE_STATE_COMING; mutex_unlock(&module_mutex); klp_register_patch(P3); klp_enable_patch(P3); # STATE 1 klp_module_notify(M) klp_module_notify_coming(P1); klp_module_notify_coming(P2); klp_module_notify_coming(P3); # STATE 2 The ftrace ops for a() and b() then looks: STATE1: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b3); STATE2: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b2,b1,b3); therefore, b2() is used for the module but a3() is used for vmcore because they were the last added. Example of the race with going modules: ======================================= CPU0 CPU1 delete_module() #SYSCALL try_stop_module() mod->state = MODULE_STATE_GOING; mutex_unlock(&module_mutex); klp_register_patch() klp_enable_patch() #save place to switch universe b() # from module that is going a() # from core (patched) mod->exit(); Note that the function b() can be called until we call mod->exit(). If we do not apply patch against b() because it is in MODULE_STATE_GOING, it will call patched a() with modified semantic and things might get wrong. [jpoimboe@redhat.com: use one boolean instead of two] Signed-off-by: Petr Mladek <pmladek@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-03-12 19:55:13 +08:00
obj->mod = NULL;
klp_find_object_module(obj);
name = klp_is_module(obj) ? obj->name : "vmlinux";
ret = kobject_add(&obj->kobj, &patch->kobj, "%s", name);
if (ret)
return ret;
klp_for_each_func(obj, func) {
ret = klp_init_func(obj, func);
if (ret)
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
return ret;
}
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
if (klp_is_object_loaded(obj))
ret = klp_init_object_loaded(patch, obj);
return ret;
}
static void klp_init_func_early(struct klp_object *obj,
struct klp_func *func)
{
kobject_init(&func->kobj, &klp_ktype_func);
list_add_tail(&func->node, &obj->func_list);
}
static void klp_init_object_early(struct klp_patch *patch,
struct klp_object *obj)
{
INIT_LIST_HEAD(&obj->func_list);
kobject_init(&obj->kobj, &klp_ktype_object);
list_add_tail(&obj->node, &patch->obj_list);
}
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
static int klp_init_patch_early(struct klp_patch *patch)
{
struct klp_object *obj;
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
struct klp_func *func;
if (!patch->objs)
return -EINVAL;
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
INIT_LIST_HEAD(&patch->list);
INIT_LIST_HEAD(&patch->obj_list);
kobject_init(&patch->kobj, &klp_ktype_patch);
patch->enabled = false;
patch->forced = false;
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
INIT_WORK(&patch->free_work, klp_free_patch_work_fn);
livepatch: allow removal of a disabled patch Currently we do not allow patch module to unload since there is no method to determine if a task is still running in the patched code. The consistency model gives us the way because when the unpatching finishes we know that all tasks were marked as safe to call an original function. Thus every new call to the function calls the original code and at the same time no task can be somewhere in the patched code, because it had to leave that code to be marked as safe. We can safely let the patch module go after that. Completion is used for synchronization between module removal and sysfs infrastructure in a similar way to commit 942e443127e9 ("module: Fix mod->mkobj.kobj potentially freed too early"). Note that we still do not allow the removal for immediate model, that is no consistency model. The module refcount may increase in this case if somebody disables and enables the patch several times. This should not cause any harm. With this change a call to try_module_get() is moved to __klp_enable_patch from klp_register_patch to make module reference counting symmetric (module_put() is in a patch disable path) and to allow to take a new reference to a disabled module when being enabled. Finally, we need to be very careful about possible races between klp_unregister_patch(), kobject_put() functions and operations on the related sysfs files. kobject_put(&patch->kobj) must be called without klp_mutex. Otherwise, it might be blocked by enabled_store() that needs the mutex as well. In addition, enabled_store() must check if the patch was not unregisted in the meantime. There is no need to do the same for other kobject_put() callsites at the moment. Their sysfs operations neither take the lock nor they access any data that might be freed in the meantime. There was an attempt to use kobjects the right way and prevent these races by design. But it made the patch definition more complicated and opened another can of worms. See https://lkml.kernel.org/r/1464018848-4303-1-git-send-email-pmladek@suse.com [Thanks to Petr Mladek for improving the commit message.] Signed-off-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-03-07 01:20:29 +08:00
init_completion(&patch->finish);
klp_for_each_object_static(patch, obj) {
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
if (!obj->funcs)
return -EINVAL;
klp_init_object_early(patch, obj);
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
klp_for_each_func_static(obj, func) {
klp_init_func_early(obj, func);
}
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
}
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
if (!try_module_get(patch->mod))
return -ENODEV;
livepatch: Consolidate klp_free functions The code for freeing livepatch structures is a bit scattered and tricky: + direct calls to klp_free_*_limited() and kobject_put() are used to release partially initialized objects + klp_free_patch() removes the patch from the public list and releases all objects except for patch->kobj + object_put(&patch->kobj) and the related wait_for_completion() are called directly outside klp_mutex; this code is duplicated; Now, we are going to remove the registration stage to simplify the API and the code. This would require handling more situations in klp_enable_patch() error paths. More importantly, we are going to add a feature called atomic replace. It will need to dynamically create func and object structures. We will want to reuse the existing init() and free() functions. This would create even more error path scenarios. This patch implements more straightforward free functions: + checks kobj_added flag instead of @limit[*] + initializes patch->list early so that the check for empty list always works + The action(s) that has to be done outside klp_mutex are done in separate klp_free_patch_finish() function. It waits only when patch->kobj was really released via the _start() part. The patch does not change the existing behavior. [*] We need our own flag to track that the kobject was successfully added to the hierarchy. Note that kobj.state_initialized only indicates that kobject has been initialized, not whether is has been added (and needs to be removed on cleanup). Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Miroslav Benes <mbenes@suse.cz> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Jason Baron <jbaron@akamai.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:21 +08:00
return 0;
}
static int klp_init_patch(struct klp_patch *patch)
{
struct klp_object *obj;
int ret;
ret = kobject_add(&patch->kobj, klp_root_kobj, "%s", patch->mod->name);
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
if (ret)
livepatch: allow removal of a disabled patch Currently we do not allow patch module to unload since there is no method to determine if a task is still running in the patched code. The consistency model gives us the way because when the unpatching finishes we know that all tasks were marked as safe to call an original function. Thus every new call to the function calls the original code and at the same time no task can be somewhere in the patched code, because it had to leave that code to be marked as safe. We can safely let the patch module go after that. Completion is used for synchronization between module removal and sysfs infrastructure in a similar way to commit 942e443127e9 ("module: Fix mod->mkobj.kobj potentially freed too early"). Note that we still do not allow the removal for immediate model, that is no consistency model. The module refcount may increase in this case if somebody disables and enables the patch several times. This should not cause any harm. With this change a call to try_module_get() is moved to __klp_enable_patch from klp_register_patch to make module reference counting symmetric (module_put() is in a patch disable path) and to allow to take a new reference to a disabled module when being enabled. Finally, we need to be very careful about possible races between klp_unregister_patch(), kobject_put() functions and operations on the related sysfs files. kobject_put(&patch->kobj) must be called without klp_mutex. Otherwise, it might be blocked by enabled_store() that needs the mutex as well. In addition, enabled_store() must check if the patch was not unregisted in the meantime. There is no need to do the same for other kobject_put() callsites at the moment. Their sysfs operations neither take the lock nor they access any data that might be freed in the meantime. There was an attempt to use kobjects the right way and prevent these races by design. But it made the patch definition more complicated and opened another can of worms. See https://lkml.kernel.org/r/1464018848-4303-1-git-send-email-pmladek@suse.com [Thanks to Petr Mladek for improving the commit message.] Signed-off-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-03-07 01:20:29 +08:00
return ret;
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
if (patch->replace) {
ret = klp_add_nops(patch);
if (ret)
return ret;
}
klp_for_each_object(patch, obj) {
ret = klp_init_object(patch, obj);
if (ret)
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
return ret;
}
list_add_tail(&patch->list, &klp_patches);
livepatch: allow removal of a disabled patch Currently we do not allow patch module to unload since there is no method to determine if a task is still running in the patched code. The consistency model gives us the way because when the unpatching finishes we know that all tasks were marked as safe to call an original function. Thus every new call to the function calls the original code and at the same time no task can be somewhere in the patched code, because it had to leave that code to be marked as safe. We can safely let the patch module go after that. Completion is used for synchronization between module removal and sysfs infrastructure in a similar way to commit 942e443127e9 ("module: Fix mod->mkobj.kobj potentially freed too early"). Note that we still do not allow the removal for immediate model, that is no consistency model. The module refcount may increase in this case if somebody disables and enables the patch several times. This should not cause any harm. With this change a call to try_module_get() is moved to __klp_enable_patch from klp_register_patch to make module reference counting symmetric (module_put() is in a patch disable path) and to allow to take a new reference to a disabled module when being enabled. Finally, we need to be very careful about possible races between klp_unregister_patch(), kobject_put() functions and operations on the related sysfs files. kobject_put(&patch->kobj) must be called without klp_mutex. Otherwise, it might be blocked by enabled_store() that needs the mutex as well. In addition, enabled_store() must check if the patch was not unregisted in the meantime. There is no need to do the same for other kobject_put() callsites at the moment. Their sysfs operations neither take the lock nor they access any data that might be freed in the meantime. There was an attempt to use kobjects the right way and prevent these races by design. But it made the patch definition more complicated and opened another can of worms. See https://lkml.kernel.org/r/1464018848-4303-1-git-send-email-pmladek@suse.com [Thanks to Petr Mladek for improving the commit message.] Signed-off-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Reviewed-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-03-07 01:20:29 +08:00
return 0;
}
static int __klp_disable_patch(struct klp_patch *patch)
{
struct klp_object *obj;
if (WARN_ON(!patch->enabled))
return -EINVAL;
if (klp_transition_patch)
return -EBUSY;
klp_init_transition(patch, KLP_UNPATCHED);
klp_for_each_object(patch, obj)
if (obj->patched)
klp_pre_unpatch_callback(obj);
/*
* Enforce the order of the func->transition writes in
* klp_init_transition() and the TIF_PATCH_PENDING writes in
* klp_start_transition(). In the rare case where klp_ftrace_handler()
* is called shortly after klp_update_patch_state() switches the task,
* this ensures the handler sees that func->transition is set.
*/
smp_wmb();
klp_start_transition();
patch->enabled = false;
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
klp_try_complete_transition();
return 0;
}
static int __klp_enable_patch(struct klp_patch *patch)
{
struct klp_object *obj;
int ret;
if (klp_transition_patch)
return -EBUSY;
if (WARN_ON(patch->enabled))
return -EINVAL;
pr_notice("enabling patch '%s'\n", patch->mod->name);
klp_init_transition(patch, KLP_PATCHED);
/*
* Enforce the order of the func->transition writes in
* klp_init_transition() and the ops->func_stack writes in
* klp_patch_object(), so that klp_ftrace_handler() will see the
* func->transition updates before the handler is registered and the
* new funcs become visible to the handler.
*/
smp_wmb();
klp_for_each_object(patch, obj) {
if (!klp_is_object_loaded(obj))
continue;
ret = klp_pre_patch_callback(obj);
if (ret) {
pr_warn("pre-patch callback failed for object '%s'\n",
klp_is_module(obj) ? obj->name : "vmlinux");
goto err;
}
ret = klp_patch_object(obj);
if (ret) {
pr_warn("failed to patch object '%s'\n",
klp_is_module(obj) ? obj->name : "vmlinux");
goto err;
}
}
klp_start_transition();
patch->enabled = true;
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
klp_try_complete_transition();
return 0;
err:
pr_warn("failed to enable patch '%s'\n", patch->mod->name);
klp_cancel_transition();
return ret;
}
/**
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
* klp_enable_patch() - enable the livepatch
* @patch: patch to be enabled
*
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
* Initializes the data structure associated with the patch, creates the sysfs
* interface, performs the needed symbol lookups and code relocations,
* registers the patched functions with ftrace.
*
* This function is supposed to be called from the livepatch module_init()
* callback.
*
* Return: 0 on success, otherwise error
*/
int klp_enable_patch(struct klp_patch *patch)
{
int ret;
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
if (!patch || !patch->mod)
return -EINVAL;
if (!is_livepatch_module(patch->mod)) {
pr_err("module %s is not marked as a livepatch module\n",
patch->mod->name);
return -EINVAL;
}
if (!klp_initialized())
return -ENODEV;
if (!klp_have_reliable_stack()) {
pr_warn("This architecture doesn't have support for the livepatch consistency model.\n");
pr_warn("The livepatch transition may never complete.\n");
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
}
mutex_lock(&klp_mutex);
livepatch: Allow to distinguish different version of system state changes The atomic replace runs pre/post (un)install callbacks only from the new livepatch. There are several reasons for this: + Simplicity: clear ordering of operations, no interactions between old and new callbacks. + Reliability: only new livepatch knows what changes can already be made by older livepatches and how to take over the state. + Testing: the atomic replace can be properly tested only when a newer livepatch is available. It might be too late to fix unwanted effect of callbacks from older livepatches. It might happen that an older change is not enough and the same system state has to be modified another way. Different changes need to get distinguished by a version number added to struct klp_state. The version can also be used to prevent loading incompatible livepatches. The check is done when the livepatch is enabled. The rules are: + Any completely new system state modification is allowed. + System state modifications with the same or higher version are allowed for already modified system states. + Cumulative livepatches must handle all system state modifications from already installed livepatches. + Non-cumulative livepatches are allowed to touch already modified system states. Link: http://lkml.kernel.org/r/20191030154313.13263-4-pmladek@suse.com To: Jiri Kosina <jikos@kernel.org> Cc: Kamalesh Babulal <kamalesh@linux.vnet.ibm.com> Cc: Nicolai Stange <nstange@suse.de> Cc: live-patching@vger.kernel.org Cc: linux-kernel@vger.kernel.org Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com>
2019-10-30 23:43:11 +08:00
if (!klp_is_patch_compatible(patch)) {
pr_err("Livepatch patch (%s) is not compatible with the already installed livepatches.\n",
patch->mod->name);
mutex_unlock(&klp_mutex);
return -EINVAL;
}
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
ret = klp_init_patch_early(patch);
if (ret) {
mutex_unlock(&klp_mutex);
return ret;
}
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
ret = klp_init_patch(patch);
if (ret)
goto err;
ret = __klp_enable_patch(patch);
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
if (ret)
goto err;
mutex_unlock(&klp_mutex);
return 0;
err:
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
klp_free_patch_start(patch);
mutex_unlock(&klp_mutex);
livepatch: Simplify API by removing registration step The possibility to re-enable a registered patch was useful for immediate patches where the livepatch module had to stay until the system reboot. The improved consistency model allows to achieve the same result by unloading and loading the livepatch module again. Also we are going to add a feature called atomic replace. It will allow to create a patch that would replace all already registered patches. The aim is to handle dependent patches more securely. It will obsolete the stack of patches that helped to handle the dependencies so far. Then it might be unclear when a cumulative patch re-enabling is safe. It would be complicated to support the many modes. Instead we could actually make the API and code easier to understand. Therefore, remove the two step public API. All the checks and init calls are moved from klp_register_patch() to klp_enabled_patch(). Also the patch is automatically freed, including the sysfs interface when the transition to the disabled state is completed. As a result, there is never a disabled patch on the top of the stack. Therefore we do not need to check the stack in __klp_enable_patch(). And we could simplify the check in __klp_disable_patch(). Also the API and logic is much easier. It is enough to call klp_enable_patch() in module_init() call. The patch can be disabled by writing '0' into /sys/kernel/livepatch/<patch>/enabled. Then the module can be removed once the transition finishes and sysfs interface is freed. The only problem is how to free the structures and kobjects safely. The operation is triggered from the sysfs interface. We could not put the related kobject from there because it would cause lock inversion between klp_mutex and kernfs locks, see kn->count lockdep map. Therefore, offload the free task to a workqueue. It is perfectly fine: + The patch can no longer be used in the livepatch operations. + The module could not be removed until the free operation finishes and module_put() is called. + The operation is asynchronous already when the first klp_try_complete_transition() fails and another call is queued with a delay. Suggested-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:23 +08:00
klp_free_patch_finish(patch);
return ret;
}
EXPORT_SYMBOL_GPL(klp_enable_patch);
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
/*
* This function unpatches objects from the replaced livepatches.
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
*
* We could be pretty aggressive here. It is called in the situation where
* these structures are no longer accessed from the ftrace handler.
* All functions are redirected by the klp_transition_patch. They
* use either a new code or they are in the original code because
* of the special nop function patches.
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
*
* The only exception is when the transition was forced. In this case,
* klp_ftrace_handler() might still see the replaced patch on the stack.
* Fortunately, it is carefully designed to work with removed functions
* thanks to RCU. We only have to keep the patches on the system. Also
* this is handled transparently by patch->module_put.
*/
void klp_unpatch_replaced_patches(struct klp_patch *new_patch)
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
{
struct klp_patch *old_patch;
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
klp_for_each_patch(old_patch) {
livepatch: Add atomic replace Sometimes we would like to revert a particular fix. Currently, this is not easy because we want to keep all other fixes active and we could revert only the last applied patch. One solution would be to apply new patch that implemented all the reverted functions like in the original code. It would work as expected but there will be unnecessary redirections. In addition, it would also require knowing which functions need to be reverted at build time. Another problem is when there are many patches that touch the same functions. There might be dependencies between patches that are not enforced on the kernel side. Also it might be pretty hard to actually prepare the patch and ensure compatibility with the other patches. Atomic replace && cumulative patches: A better solution would be to create cumulative patch and say that it replaces all older ones. This patch adds a new "replace" flag to struct klp_patch. When it is enabled, a set of 'nop' klp_func will be dynamically created for all functions that are already being patched but that will no longer be modified by the new patch. They are used as a new target during the patch transition. The idea is to handle Nops' structures like the static ones. When the dynamic structures are allocated, we initialize all values that are normally statically defined. The only exception is "new_func" in struct klp_func. It has to point to the original function and the address is known only when the object (module) is loaded. Note that we really need to set it. The address is used, for example, in klp_check_stack_func(). Nevertheless we still need to distinguish the dynamically allocated structures in some operations. For this, we add "nop" flag into struct klp_func and "dynamic" flag into struct klp_object. They need special handling in the following situations: + The structures are added into the lists of objects and functions immediately. In fact, the lists were created for this purpose. + The address of the original function is known only when the patched object (module) is loaded. Therefore it is copied later in klp_init_object_loaded(). + The ftrace handler must not set PC to func->new_func. It would cause infinite loop because the address points back to the beginning of the original function. + The various free() functions must free the structure itself. Note that other ways to detect the dynamic structures are not considered safe. For example, even the statically defined struct klp_object might include empty funcs array. It might be there just to run some callbacks. Also note that the safe iterator must be used in the free() functions. Otherwise already freed structures might get accessed. Special callbacks handling: The callbacks from the replaced patches are _not_ called by intention. It would be pretty hard to define a reasonable semantic and implement it. It might even be counter-productive. The new patch is cumulative. It is supposed to include most of the changes from older patches. In most cases, it will not want to call pre_unpatch() post_unpatch() callbacks from the replaced patches. It would disable/break things for no good reasons. Also it should be easier to handle various scenarios in a single script in the new patch than think about interactions caused by running many scripts from older patches. Not to say that the old scripts even would not expect to be called in this situation. Removing replaced patches: One nice effect of the cumulative patches is that the code from the older patches is no longer used. Therefore the replaced patches can be removed. It has several advantages: + Nops' structs will no longer be necessary and might be removed. This would save memory, restore performance (no ftrace handler), allow clear view on what is really patched. + Disabling the patch will cause using the original code everywhere. Therefore the livepatch callbacks could handle only one scenario. Note that the complication is already complex enough when the patch gets enabled. It is currently solved by calling callbacks only from the new cumulative patch. + The state is clean in both the sysfs interface and lsmod. The modules with the replaced livepatches might even get removed from the system. Some people actually expected this behavior from the beginning. After all a cumulative patch is supposed to "completely" replace an existing one. It is like when a new version of an application replaces an older one. This patch does the first step. It removes the replaced patches from the list of patches. It is safe. The consistency model ensures that they are no longer used. By other words, each process works only with the structures from klp_transition_patch. The removal is done by a special function. It combines actions done by __disable_patch() and klp_complete_transition(). But it is a fast track without all the transaction-related stuff. Signed-off-by: Jason Baron <jbaron@akamai.com> [pmladek@suse.com: Split, reuse existing code, simplified] Signed-off-by: Petr Mladek <pmladek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Jessica Yu <jeyu@kernel.org> Cc: Jiri Kosina <jikos@kernel.org> Cc: Miroslav Benes <mbenes@suse.cz> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:25 +08:00
if (old_patch == new_patch)
return;
old_patch->enabled = false;
klp_unpatch_objects(old_patch);
}
}
livepatch: Remove Nop structures when unused Replaced patches are removed from the stack when the transition is finished. It means that Nop structures will never be needed again and can be removed. Why should we care? + Nop structures give the impression that the function is patched even though the ftrace handler has no effect. + Ftrace handlers do not come for free. They cause slowdown that might be visible in some workloads. The ftrace-related slowdown might actually be the reason why the function is no longer patched in the new cumulative patch. One would expect that cumulative patch would help solve these problems as well. + Cumulative patches are supposed to replace any earlier version of the patch. The amount of NOPs depends on which version was replaced. This multiplies the amount of scenarios that might happen. One might say that NOPs are innocent. But there are even optimized NOP instructions for different processors, for example, see arch/x86/kernel/alternative.c. And klp_ftrace_handler() is much more complicated. + It sounds natural to clean up a mess that is no longer needed. It could only be worse if we do not do it. This patch allows to unpatch and free the dynamic structures independently when the transition finishes. The free part is a bit tricky because kobject free callbacks are called asynchronously. We could not wait for them easily. Fortunately, we do not have to. Any further access can be avoided by removing them from the dynamic lists. Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Miroslav Benes <mbenes@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2019-01-09 20:43:26 +08:00
/*
* This function removes the dynamically allocated 'nop' functions.
*
* We could be pretty aggressive. NOPs do not change the existing
* behavior except for adding unnecessary delay by the ftrace handler.
*
* It is safe even when the transition was forced. The ftrace handler
* will see a valid ops->func_stack entry thanks to RCU.
*
* We could even free the NOPs structures. They must be the last entry
* in ops->func_stack. Therefore unregister_ftrace_function() is called.
* It does the same as klp_synchronize_transition() to make sure that
* nobody is inside the ftrace handler once the operation finishes.
*
* IMPORTANT: It must be called right after removing the replaced patches!
*/
void klp_discard_nops(struct klp_patch *new_patch)
{
klp_unpatch_objects_dynamic(klp_transition_patch);
klp_free_objects_dynamic(klp_transition_patch);
}
/*
* Remove parts of patches that touch a given kernel module. The list of
* patches processed might be limited. When limit is NULL, all patches
* will be handled.
*/
static void klp_cleanup_module_patches_limited(struct module *mod,
struct klp_patch *limit)
{
struct klp_patch *patch;
struct klp_object *obj;
klp_for_each_patch(patch) {
if (patch == limit)
break;
klp_for_each_object(patch, obj) {
if (!klp_is_module(obj) || strcmp(obj->name, mod->name))
continue;
if (patch != klp_transition_patch)
klp_pre_unpatch_callback(obj);
pr_notice("reverting patch '%s' on unloading module '%s'\n",
patch->mod->name, obj->mod->name);
klp_unpatch_object(obj);
klp_post_unpatch_callback(obj);
klp_free_object_loaded(obj);
break;
}
}
}
int klp_module_coming(struct module *mod)
{
int ret;
struct klp_patch *patch;
struct klp_object *obj;
if (WARN_ON(mod->state != MODULE_STATE_COMING))
return -EINVAL;
if (!strcmp(mod->name, "vmlinux")) {
pr_err("vmlinux.ko: invalid module name");
return -EINVAL;
}
mutex_lock(&klp_mutex);
/*
* Each module has to know that klp_module_coming()
* has been called. We never know what module will
* get patched by a new patch.
*/
mod->klp_alive = true;
klp_for_each_patch(patch) {
klp_for_each_object(patch, obj) {
if (!klp_is_module(obj) || strcmp(obj->name, mod->name))
continue;
obj->mod = mod;
ret = klp_init_object_loaded(patch, obj);
if (ret) {
pr_warn("failed to initialize patch '%s' for module '%s' (%d)\n",
patch->mod->name, obj->mod->name, ret);
goto err;
}
pr_notice("applying patch '%s' to loading module '%s'\n",
patch->mod->name, obj->mod->name);
ret = klp_pre_patch_callback(obj);
if (ret) {
pr_warn("pre-patch callback failed for object '%s'\n",
obj->name);
goto err;
}
ret = klp_patch_object(obj);
if (ret) {
pr_warn("failed to apply patch '%s' to module '%s' (%d)\n",
patch->mod->name, obj->mod->name, ret);
livepatch: Correctly call klp_post_unpatch_callback() in error paths The post_unpatch_enabled flag in struct klp_callbacks is set when a pre-patch callback successfully executes, indicating that we need to call a corresponding post-unpatch callback when the patch is reverted. This is true for ordinary patch disable as well as the error paths of klp_patch_object() callers. As currently coded, we inadvertently execute the post-patch callback twice in klp_module_coming() when klp_patch_object() fails: - We explicitly call klp_post_unpatch_callback() for the failed object - We call it again for the same object (and all the others) via klp_cleanup_module_patches_limited() We should clear the flag in klp_post_unpatch_callback() to make sure that the callback is not called twice. It makes the API more safe. (We could have removed the callback from the former error path as it would be covered by the latter call, but I think that is is cleaner to clear the post_unpatch_enabled after its invoked. For example, someone might later decide to call the callback only when obj->patched flag is set.) There is another mistake in the error path of klp_coming_module() in which it skips the post-unpatch callback for the klp_transition_patch. However, the pre-patch callback was called even for this patch, so be sure to make the corresponding callbacks for all patches. Finally, I used this opportunity to make klp_pre_patch_callback() more readable. [jkosina@suse.cz: incorporate changelog wording changes proposed by Joe Lawrence] Signed-off-by: Petr Mladek <pmladek@suse.com> Acked-by: Joe Lawrence <joe.lawrence@redhat.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2017-10-20 22:56:50 +08:00
klp_post_unpatch_callback(obj);
goto err;
}
if (patch != klp_transition_patch)
klp_post_patch_callback(obj);
break;
}
}
mutex_unlock(&klp_mutex);
return 0;
err:
/*
* If a patch is unsuccessfully applied, return
* error to the module loader.
*/
pr_warn("patch '%s' failed for module '%s', refusing to load module '%s'\n",
patch->mod->name, obj->mod->name, obj->mod->name);
mod->klp_alive = false;
obj->mod = NULL;
klp_cleanup_module_patches_limited(mod, patch);
mutex_unlock(&klp_mutex);
return ret;
}
void klp_module_going(struct module *mod)
{
if (WARN_ON(mod->state != MODULE_STATE_GOING &&
mod->state != MODULE_STATE_COMING))
return;
mutex_lock(&klp_mutex);
livepatch: Fix subtle race with coming and going modules There is a notifier that handles live patches for coming and going modules. It takes klp_mutex lock to avoid races with coming and going patches but it does not keep the lock all the time. Therefore the following races are possible: 1. The notifier is called sometime in STATE_MODULE_COMING. The module is visible by find_module() in this state all the time. It means that new patch can be registered and enabled even before the notifier is called. It might create wrong order of stacked patches, see below for an example. 2. New patch could still see the module in the GOING state even after the notifier has been called. It will try to initialize the related object structures but the module could disappear at any time. There will stay mess in the structures. It might even cause an invalid memory access. This patch solves the problem by adding a boolean variable into struct module. The value is true after the coming and before the going handler is called. New patches need to be applied when the value is true and they need to ignore the module when the value is false. Note that we need to know state of all modules on the system. The races are related to new patches. Therefore we do not know what modules will get patched. Also note that we could not simply ignore going modules. The code from the module could be called even in the GOING state until mod->exit() finishes. If we start supporting patches with semantic changes between function calls, we need to apply new patches to any still usable code. See below for an example. Finally note that the patch solves only the situation when a new patch is registered. There are no such problems when the patch is being removed. It does not matter who disable the patch first, whether the normal disable_patch() or the module notifier. There is nothing to do once the patch is disabled. Alternative solutions: ====================== + reject new patches when a patched module is coming or going; this is ugly + wait with adding new patch until the module leaves the COMING and GOING states; this might be dangerous and complicated; we would need to release kgr_lock in the middle of the patch registration to avoid a deadlock with the coming and going handlers; also we might need a waitqueue for each module which seems to be even bigger overhead than the boolean + stop modules from entering COMING and GOING states; wait until modules leave these states when they are already there; looks complicated; we would need to ignore the module that asked to stop the others to avoid a deadlock; also it is unclear what to do when two modules asked to stop others and both are in COMING state (situation when two new patches are applied) + always register/enable new patches and fix up the potential mess (registered patches order) in klp_module_init(); this is nasty and prone to regressions in the future development + add another MODULE_STATE where the kallsyms are visible but the module is not used yet; this looks too complex; the module states are checked on "many" locations Example of patch stacking breakage: =================================== The notifier could _not_ _simply_ ignore already initialized module objects. For example, let's have three patches (P1, P2, P3) for functions a() and b() where a() is from vmcore and b() is from a module M. Something like: a() b() P1 a1() b1() P2 a2() b2() P3 a3() b3(3) If you load the module M after all patches are registered and enabled. The ftrace ops for function a() and b() has listed the functions in this order: ops_a->func_stack -> list(a3,a2,a1) ops_b->func_stack -> list(b3,b2,b1) , so the pointer to b3() is the first and will be used. Then you might have the following scenario. Let's start with state when patches P1 and P2 are registered and enabled but the module M is not loaded. Then ftrace ops for b() does not exist. Then we get into the following race: CPU0 CPU1 load_module(M) complete_formation() mod->state = MODULE_STATE_COMING; mutex_unlock(&module_mutex); klp_register_patch(P3); klp_enable_patch(P3); # STATE 1 klp_module_notify(M) klp_module_notify_coming(P1); klp_module_notify_coming(P2); klp_module_notify_coming(P3); # STATE 2 The ftrace ops for a() and b() then looks: STATE1: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b3); STATE2: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b2,b1,b3); therefore, b2() is used for the module but a3() is used for vmcore because they were the last added. Example of the race with going modules: ======================================= CPU0 CPU1 delete_module() #SYSCALL try_stop_module() mod->state = MODULE_STATE_GOING; mutex_unlock(&module_mutex); klp_register_patch() klp_enable_patch() #save place to switch universe b() # from module that is going a() # from core (patched) mod->exit(); Note that the function b() can be called until we call mod->exit(). If we do not apply patch against b() because it is in MODULE_STATE_GOING, it will call patched a() with modified semantic and things might get wrong. [jpoimboe@redhat.com: use one boolean instead of two] Signed-off-by: Petr Mladek <pmladek@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-03-12 19:55:13 +08:00
/*
* Each module has to know that klp_module_going()
* has been called. We never know what module will
* get patched by a new patch.
livepatch: Fix subtle race with coming and going modules There is a notifier that handles live patches for coming and going modules. It takes klp_mutex lock to avoid races with coming and going patches but it does not keep the lock all the time. Therefore the following races are possible: 1. The notifier is called sometime in STATE_MODULE_COMING. The module is visible by find_module() in this state all the time. It means that new patch can be registered and enabled even before the notifier is called. It might create wrong order of stacked patches, see below for an example. 2. New patch could still see the module in the GOING state even after the notifier has been called. It will try to initialize the related object structures but the module could disappear at any time. There will stay mess in the structures. It might even cause an invalid memory access. This patch solves the problem by adding a boolean variable into struct module. The value is true after the coming and before the going handler is called. New patches need to be applied when the value is true and they need to ignore the module when the value is false. Note that we need to know state of all modules on the system. The races are related to new patches. Therefore we do not know what modules will get patched. Also note that we could not simply ignore going modules. The code from the module could be called even in the GOING state until mod->exit() finishes. If we start supporting patches with semantic changes between function calls, we need to apply new patches to any still usable code. See below for an example. Finally note that the patch solves only the situation when a new patch is registered. There are no such problems when the patch is being removed. It does not matter who disable the patch first, whether the normal disable_patch() or the module notifier. There is nothing to do once the patch is disabled. Alternative solutions: ====================== + reject new patches when a patched module is coming or going; this is ugly + wait with adding new patch until the module leaves the COMING and GOING states; this might be dangerous and complicated; we would need to release kgr_lock in the middle of the patch registration to avoid a deadlock with the coming and going handlers; also we might need a waitqueue for each module which seems to be even bigger overhead than the boolean + stop modules from entering COMING and GOING states; wait until modules leave these states when they are already there; looks complicated; we would need to ignore the module that asked to stop the others to avoid a deadlock; also it is unclear what to do when two modules asked to stop others and both are in COMING state (situation when two new patches are applied) + always register/enable new patches and fix up the potential mess (registered patches order) in klp_module_init(); this is nasty and prone to regressions in the future development + add another MODULE_STATE where the kallsyms are visible but the module is not used yet; this looks too complex; the module states are checked on "many" locations Example of patch stacking breakage: =================================== The notifier could _not_ _simply_ ignore already initialized module objects. For example, let's have three patches (P1, P2, P3) for functions a() and b() where a() is from vmcore and b() is from a module M. Something like: a() b() P1 a1() b1() P2 a2() b2() P3 a3() b3(3) If you load the module M after all patches are registered and enabled. The ftrace ops for function a() and b() has listed the functions in this order: ops_a->func_stack -> list(a3,a2,a1) ops_b->func_stack -> list(b3,b2,b1) , so the pointer to b3() is the first and will be used. Then you might have the following scenario. Let's start with state when patches P1 and P2 are registered and enabled but the module M is not loaded. Then ftrace ops for b() does not exist. Then we get into the following race: CPU0 CPU1 load_module(M) complete_formation() mod->state = MODULE_STATE_COMING; mutex_unlock(&module_mutex); klp_register_patch(P3); klp_enable_patch(P3); # STATE 1 klp_module_notify(M) klp_module_notify_coming(P1); klp_module_notify_coming(P2); klp_module_notify_coming(P3); # STATE 2 The ftrace ops for a() and b() then looks: STATE1: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b3); STATE2: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b2,b1,b3); therefore, b2() is used for the module but a3() is used for vmcore because they were the last added. Example of the race with going modules: ======================================= CPU0 CPU1 delete_module() #SYSCALL try_stop_module() mod->state = MODULE_STATE_GOING; mutex_unlock(&module_mutex); klp_register_patch() klp_enable_patch() #save place to switch universe b() # from module that is going a() # from core (patched) mod->exit(); Note that the function b() can be called until we call mod->exit(). If we do not apply patch against b() because it is in MODULE_STATE_GOING, it will call patched a() with modified semantic and things might get wrong. [jpoimboe@redhat.com: use one boolean instead of two] Signed-off-by: Petr Mladek <pmladek@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-03-12 19:55:13 +08:00
*/
mod->klp_alive = false;
livepatch: Fix subtle race with coming and going modules There is a notifier that handles live patches for coming and going modules. It takes klp_mutex lock to avoid races with coming and going patches but it does not keep the lock all the time. Therefore the following races are possible: 1. The notifier is called sometime in STATE_MODULE_COMING. The module is visible by find_module() in this state all the time. It means that new patch can be registered and enabled even before the notifier is called. It might create wrong order of stacked patches, see below for an example. 2. New patch could still see the module in the GOING state even after the notifier has been called. It will try to initialize the related object structures but the module could disappear at any time. There will stay mess in the structures. It might even cause an invalid memory access. This patch solves the problem by adding a boolean variable into struct module. The value is true after the coming and before the going handler is called. New patches need to be applied when the value is true and they need to ignore the module when the value is false. Note that we need to know state of all modules on the system. The races are related to new patches. Therefore we do not know what modules will get patched. Also note that we could not simply ignore going modules. The code from the module could be called even in the GOING state until mod->exit() finishes. If we start supporting patches with semantic changes between function calls, we need to apply new patches to any still usable code. See below for an example. Finally note that the patch solves only the situation when a new patch is registered. There are no such problems when the patch is being removed. It does not matter who disable the patch first, whether the normal disable_patch() or the module notifier. There is nothing to do once the patch is disabled. Alternative solutions: ====================== + reject new patches when a patched module is coming or going; this is ugly + wait with adding new patch until the module leaves the COMING and GOING states; this might be dangerous and complicated; we would need to release kgr_lock in the middle of the patch registration to avoid a deadlock with the coming and going handlers; also we might need a waitqueue for each module which seems to be even bigger overhead than the boolean + stop modules from entering COMING and GOING states; wait until modules leave these states when they are already there; looks complicated; we would need to ignore the module that asked to stop the others to avoid a deadlock; also it is unclear what to do when two modules asked to stop others and both are in COMING state (situation when two new patches are applied) + always register/enable new patches and fix up the potential mess (registered patches order) in klp_module_init(); this is nasty and prone to regressions in the future development + add another MODULE_STATE where the kallsyms are visible but the module is not used yet; this looks too complex; the module states are checked on "many" locations Example of patch stacking breakage: =================================== The notifier could _not_ _simply_ ignore already initialized module objects. For example, let's have three patches (P1, P2, P3) for functions a() and b() where a() is from vmcore and b() is from a module M. Something like: a() b() P1 a1() b1() P2 a2() b2() P3 a3() b3(3) If you load the module M after all patches are registered and enabled. The ftrace ops for function a() and b() has listed the functions in this order: ops_a->func_stack -> list(a3,a2,a1) ops_b->func_stack -> list(b3,b2,b1) , so the pointer to b3() is the first and will be used. Then you might have the following scenario. Let's start with state when patches P1 and P2 are registered and enabled but the module M is not loaded. Then ftrace ops for b() does not exist. Then we get into the following race: CPU0 CPU1 load_module(M) complete_formation() mod->state = MODULE_STATE_COMING; mutex_unlock(&module_mutex); klp_register_patch(P3); klp_enable_patch(P3); # STATE 1 klp_module_notify(M) klp_module_notify_coming(P1); klp_module_notify_coming(P2); klp_module_notify_coming(P3); # STATE 2 The ftrace ops for a() and b() then looks: STATE1: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b3); STATE2: ops_a->func_stack -> list(a3,a2,a1); ops_b->func_stack -> list(b2,b1,b3); therefore, b2() is used for the module but a3() is used for vmcore because they were the last added. Example of the race with going modules: ======================================= CPU0 CPU1 delete_module() #SYSCALL try_stop_module() mod->state = MODULE_STATE_GOING; mutex_unlock(&module_mutex); klp_register_patch() klp_enable_patch() #save place to switch universe b() # from module that is going a() # from core (patched) mod->exit(); Note that the function b() can be called until we call mod->exit(). If we do not apply patch against b() because it is in MODULE_STATE_GOING, it will call patched a() with modified semantic and things might get wrong. [jpoimboe@redhat.com: use one boolean instead of two] Signed-off-by: Petr Mladek <pmladek@suse.cz> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2015-03-12 19:55:13 +08:00
klp_cleanup_module_patches_limited(mod, NULL);
mutex_unlock(&klp_mutex);
}
static int __init klp_init(void)
{
klp_root_kobj = kobject_create_and_add("livepatch", kernel_kobj);
if (!klp_root_kobj)
return -ENOMEM;
return 0;
}
module_init(klp_init);