2019-05-19 21:51:43 +08:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
2014-12-17 01:58:19 +08:00
|
|
|
/*
|
|
|
|
* 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>
|
2016-03-23 08:03:18 +08:00
|
|
|
#include <linux/elf.h>
|
|
|
|
#include <linux/moduleloader.h>
|
2017-03-07 01:20:29 +08:00
|
|
|
#include <linux/completion.h>
|
2019-06-14 09:07:22 +08:00
|
|
|
#include <linux/memory.h>
|
2015-12-04 06:33:26 +08:00
|
|
|
#include <asm/cacheflush.h>
|
2017-03-08 21:27:05 +08:00
|
|
|
#include "core.h"
|
2017-02-14 09:42:37 +08:00
|
|
|
#include "patch.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"
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2015-01-20 23:26:19 +08:00
|
|
|
/*
|
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()
|
2015-01-20 23:26:19 +08:00
|
|
|
*/
|
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);
|
2015-01-20 23:26:19 +08:00
|
|
|
|
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.
|
|
|
|
*/
|
2019-01-09 20:43:22 +08:00
|
|
|
LIST_HEAD(klp_patches);
|
2014-12-17 01:58:19 +08:00
|
|
|
|
|
|
|
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;
|
|
|
|
|
2014-12-17 01:58:19 +08:00
|
|
|
if (!klp_is_module(obj))
|
|
|
|
return;
|
|
|
|
|
|
|
|
mutex_lock(&module_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
|
|
|
* We do not want to block removal of patched modules and therefore
|
|
|
|
* we do not take a reference here. The patches are removed by
|
2016-03-17 08:55:39 +08:00
|
|
|
* 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);
|
|
|
|
/*
|
2016-03-17 08:55:39 +08:00
|
|
|
* 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.
|
2014-12-17 01:58:19 +08:00
|
|
|
*/
|
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;
|
|
|
|
|
2014-12-17 01:58:19 +08:00
|
|
|
mutex_unlock(&module_mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool klp_initialized(void)
|
|
|
|
{
|
2015-05-11 13:52:29 +08:00
|
|
|
return !!klp_root_kobj;
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2014-12-17 01:58:19 +08:00
|
|
|
struct klp_find_arg {
|
|
|
|
const char *objname;
|
|
|
|
const char *name;
|
|
|
|
unsigned long addr;
|
|
|
|
unsigned long count;
|
2015-12-02 10:40:54 +08:00
|
|
|
unsigned long pos;
|
2014-12-17 01:58:19 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
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++;
|
|
|
|
|
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;
|
|
|
|
|
2014-12-17 01:58:19 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int klp_find_object_symbol(const char *objname, const char *name,
|
2015-12-02 10:40:54 +08:00
|
|
|
unsigned long sympos, unsigned long *addr)
|
2014-12-17 01:58:19 +08:00
|
|
|
{
|
|
|
|
struct klp_find_arg args = {
|
|
|
|
.objname = objname,
|
|
|
|
.name = name,
|
|
|
|
.addr = 0,
|
2015-12-02 10:40:54 +08:00
|
|
|
.count = 0,
|
|
|
|
.pos = sympos,
|
2014-12-17 01:58:19 +08:00
|
|
|
};
|
|
|
|
|
2015-06-01 23:48:37 +08:00
|
|
|
mutex_lock(&module_mutex);
|
2017-03-28 21:10:35 +08:00
|
|
|
if (objname)
|
|
|
|
module_kallsyms_on_each_symbol(klp_find_callback, &args);
|
|
|
|
else
|
|
|
|
kallsyms_on_each_symbol(klp_find_callback, &args);
|
2015-06-01 23:48:37 +08:00
|
|
|
mutex_unlock(&module_mutex);
|
2014-12-17 01:58:19 +08:00
|
|
|
|
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)
|
2014-12-17 01:58:19 +08:00
|
|
|
pr_err("symbol '%s' not found in symbol table\n", name);
|
2015-12-02 10:40:54 +08:00
|
|
|
else if (args.count > 1 && sympos == 0) {
|
2016-03-09 22:20:59 +08:00
|
|
|
pr_err("unresolvable ambiguity for symbol '%s' in object '%s'\n",
|
|
|
|
name, objname);
|
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 {
|
2014-12-17 01:58:19 +08:00
|
|
|
*addr = args.addr;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
*addr = 0;
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2016-03-23 08:03:18 +08:00
|
|
|
static int klp_resolve_symbols(Elf_Shdr *relasec, struct module *pmod)
|
2014-12-17 01:58:19 +08:00
|
|
|
{
|
2016-03-23 08:03:18 +08:00
|
|
|
int i, cnt, vmlinux, ret;
|
|
|
|
char objname[MODULE_NAME_LEN];
|
|
|
|
char symname[KSYM_NAME_LEN];
|
|
|
|
char *strtab = pmod->core_kallsyms.strtab;
|
|
|
|
Elf_Rela *relas;
|
|
|
|
Elf_Sym *sym;
|
|
|
|
unsigned long sympos, addr;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2015-12-02 10:40:54 +08:00
|
|
|
/*
|
2016-03-23 08:03:18 +08:00
|
|
|
* Since the field widths for objname and symname 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.
|
2015-12-02 10:40:54 +08:00
|
|
|
*/
|
2016-03-23 08:03:18 +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++) {
|
|
|
|
sym = pmod->core_kallsyms.symtab + ELF_R_SYM(relas[i].r_info);
|
|
|
|
if (sym->st_shndx != SHN_LIVEPATCH) {
|
2017-04-14 06:59:15 +08:00
|
|
|
pr_err("symbol %s is not marked as a livepatch symbol\n",
|
2016-03-23 08:03:18 +08:00
|
|
|
strtab + sym->st_name);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Format: .klp.sym.objname.symname,sympos */
|
|
|
|
cnt = sscanf(strtab + sym->st_name,
|
|
|
|
".klp.sym.%55[^.].%127[^,],%lu",
|
|
|
|
objname, symname, &sympos);
|
|
|
|
if (cnt != 3) {
|
2017-04-14 06:59:15 +08:00
|
|
|
pr_err("symbol %s has an incorrectly formatted name\n",
|
2016-03-23 08:03:18 +08:00
|
|
|
strtab + sym->st_name);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* klp_find_object_symbol() treats a NULL objname as vmlinux */
|
|
|
|
vmlinux = !strcmp(objname, "vmlinux");
|
|
|
|
ret = klp_find_object_symbol(vmlinux ? NULL : objname,
|
|
|
|
symname, sympos, &addr);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
sym->st_value = addr;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static int klp_write_object_relocations(struct module *pmod,
|
|
|
|
struct klp_object *obj)
|
|
|
|
{
|
2016-03-23 08:03:18 +08:00
|
|
|
int i, cnt, ret = 0;
|
|
|
|
const char *objname, *secname;
|
|
|
|
char sec_objname[MODULE_NAME_LEN];
|
|
|
|
Elf_Shdr *sec;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
|
|
|
if (WARN_ON(!klp_is_object_loaded(obj)))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2016-03-23 08:03:18 +08:00
|
|
|
objname = klp_is_module(obj) ? obj->name : "vmlinux";
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2016-03-23 08:03:18 +08:00
|
|
|
/* For each klp relocation section */
|
|
|
|
for (i = 1; i < pmod->klp_info->hdr.e_shnum; i++) {
|
|
|
|
sec = pmod->klp_info->sechdrs + i;
|
|
|
|
secname = pmod->klp_info->secstrings + sec->sh_name;
|
|
|
|
if (!(sec->sh_flags & SHF_RELA_LIVEPATCH))
|
|
|
|
continue;
|
2015-12-04 06:33:26 +08:00
|
|
|
|
2016-03-23 08:03:18 +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(secname, ".klp.rela.%55[^.]", sec_objname);
|
|
|
|
if (cnt != 1) {
|
2017-04-14 06:59:15 +08:00
|
|
|
pr_err("section %s has an incorrectly formatted name\n",
|
2016-03-23 08:03:18 +08:00
|
|
|
secname);
|
|
|
|
ret = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
2015-12-04 06:33:26 +08:00
|
|
|
|
2016-03-23 08:03:18 +08:00
|
|
|
if (strcmp(objname, sec_objname))
|
|
|
|
continue;
|
2015-12-04 06:33:26 +08:00
|
|
|
|
2016-03-23 08:03:18 +08:00
|
|
|
ret = klp_resolve_symbols(sec, pmod);
|
2015-12-02 10:40:55 +08:00
|
|
|
if (ret)
|
2016-03-23 08:03:18 +08:00
|
|
|
break;
|
2015-12-02 10:40:55 +08:00
|
|
|
|
2016-03-23 08:03:18 +08:00
|
|
|
ret = apply_relocate_add(pmod->klp_info->sechdrs,
|
|
|
|
pmod->core_kallsyms.strtab,
|
|
|
|
pmod->klp_info->symndx, i, pmod);
|
|
|
|
if (ret)
|
|
|
|
break;
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
2015-12-04 06:33:26 +08:00
|
|
|
return ret;
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* 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
|
2017-11-22 18:29:21 +08:00
|
|
|
* /sys/kernel/livepatch/<patch>/force
|
2014-12-17 01:58:19 +08:00
|
|
|
* /sys/kernel/livepatch/<patch>/<object>
|
2015-12-02 10:40:56 +08:00
|
|
|
* /sys/kernel/livepatch/<patch>/<object>/<function,sympos>
|
2014-12-17 01:58:19 +08:00
|
|
|
*/
|
2019-01-09 20:43:20 +08:00
|
|
|
static int __klp_disable_patch(struct klp_patch *patch);
|
2014-12-17 01:58:19 +08:00
|
|
|
|
|
|
|
static ssize_t enabled_store(struct kobject *kobj, struct kobj_attribute *attr,
|
|
|
|
const char *buf, size_t count)
|
|
|
|
{
|
|
|
|
struct klp_patch *patch;
|
|
|
|
int ret;
|
2017-02-14 09:42:38 +08:00
|
|
|
bool enabled;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2017-02-14 09:42:38 +08:00
|
|
|
ret = kstrtobool(buf, &enabled);
|
2014-12-17 01:58:19 +08:00
|
|
|
if (ret)
|
2017-02-14 09:42:38 +08:00
|
|
|
return ret;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
|
|
|
patch = container_of(kobj, struct klp_patch, kobj);
|
|
|
|
|
|
|
|
mutex_lock(&klp_mutex);
|
|
|
|
|
2017-02-14 09:42:38 +08:00
|
|
|
if (patch->enabled == enabled) {
|
2014-12-17 01:58:19 +08:00
|
|
|
/* already in requested state */
|
|
|
|
ret = -EINVAL;
|
2019-01-09 20:43:23 +08:00
|
|
|
goto out;
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
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();
|
2019-01-09 20:43:23 +08:00
|
|
|
else if (!enabled)
|
2014-12-17 01:58:19 +08:00
|
|
|
ret = __klp_disable_patch(patch);
|
2019-01-09 20:43:23 +08:00
|
|
|
else
|
|
|
|
ret = -EINVAL;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2019-01-09 20:43:23 +08:00
|
|
|
out:
|
2014-12-17 01:58:19 +08:00
|
|
|
mutex_unlock(&klp_mutex);
|
|
|
|
|
2019-01-09 20:43:23 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2014-12-17 01:58:19 +08:00
|
|
|
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);
|
2017-02-14 09:42:35 +08:00
|
|
|
return snprintf(buf, PAGE_SIZE-1, "%d\n", patch->enabled);
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
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);
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
2017-11-22 18:29:21 +08:00
|
|
|
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;
|
|
|
|
|
2017-12-21 21:40:43 +08:00
|
|
|
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);
|
2017-11-22 18:29:21 +08:00
|
|
|
|
|
|
|
return count;
|
|
|
|
}
|
|
|
|
|
2014-12-17 01:58:19 +08:00
|
|
|
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);
|
2017-11-22 18:29:21 +08:00
|
|
|
static struct kobj_attribute force_kobj_attr = __ATTR_WO(force);
|
2014-12-17 01:58:19 +08:00
|
|
|
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,
|
2017-11-22 18:29:21 +08:00
|
|
|
&force_kobj_attr.attr,
|
2014-12-17 01:58:19 +08:00
|
|
|
NULL
|
|
|
|
};
|
2019-04-02 10:51:58 +08:00
|
|
|
ATTRIBUTE_GROUPS(klp_patch);
|
2014-12-17 01:58:19 +08:00
|
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2019-05-03 21:26:25 +08:00
|
|
|
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);
|
2019-05-03 21:26:24 +08:00
|
|
|
|
2019-05-03 21:26:25 +08:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-05-03 21:26:25 +08:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-05-03 21:26:25 +08:00
|
|
|
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) {
|
2019-05-03 21:26:25 +08:00
|
|
|
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;
|
|
|
|
|
2019-02-04 21:56:50 +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
|
|
|
klp_for_each_object(old_patch, old_obj) {
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = klp_add_object_nops(patch, old_obj);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-12-17 01:58:19 +08:00
|
|
|
static void klp_kobj_release_patch(struct kobject *kobj)
|
|
|
|
{
|
2017-03-07 01:20:29 +08:00
|
|
|
struct klp_patch *patch;
|
|
|
|
|
|
|
|
patch = container_of(kobj, struct klp_patch, kobj);
|
|
|
|
complete(&patch->finish);
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static struct kobj_type klp_ktype_patch = {
|
|
|
|
.release = klp_kobj_release_patch,
|
|
|
|
.sysfs_ops = &kobj_sysfs_ops,
|
2019-04-02 10:51:58 +08:00
|
|
|
.default_groups = klp_patch_groups,
|
2014-12-17 01:58:19 +08:00
|
|
|
};
|
|
|
|
|
2015-05-19 18:01:18 +08:00
|
|
|
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);
|
2015-05-19 18:01:18 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static struct kobj_type klp_ktype_object = {
|
|
|
|
.release = klp_kobj_release_object,
|
|
|
|
.sysfs_ops = &kobj_sysfs_ops,
|
|
|
|
};
|
|
|
|
|
2014-12-17 01:58:19 +08:00
|
|
|
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);
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static struct kobj_type klp_ktype_func = {
|
|
|
|
.release = klp_kobj_release_func,
|
|
|
|
.sysfs_ops = &kobj_sysfs_ops,
|
|
|
|
};
|
|
|
|
|
2019-01-09 20:43:26 +08:00
|
|
|
static void __klp_free_funcs(struct klp_object *obj, bool nops_only)
|
2014-12-17 01:58:19 +08:00
|
|
|
{
|
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;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
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) {
|
2019-01-09 20:43:26 +08:00
|
|
|
if (nops_only && !func->nop)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
list_del(&func->node);
|
2019-05-03 21:26:24 +08:00
|
|
|
kobject_put(&func->kobj);
|
2019-01-09 20:43:21 +08:00
|
|
|
}
|
2014-12-17 01:58:19 +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) {
|
2019-01-09 20:43:19 +08:00
|
|
|
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;
|
|
|
|
}
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
2019-01-09 20:43:26 +08:00
|
|
|
static void __klp_free_objects(struct klp_patch *patch, bool nops_only)
|
2014-12-17 01:58:19 +08:00
|
|
|
{
|
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;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
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) {
|
2019-01-09 20:43:26 +08:00
|
|
|
__klp_free_funcs(obj, nops_only);
|
|
|
|
|
|
|
|
if (nops_only && !obj->dynamic)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
list_del(&obj->node);
|
2019-05-03 21:26:24 +08:00
|
|
|
kobject_put(&obj->kobj);
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
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.
|
|
|
|
*/
|
2019-01-09 20:43:23 +08:00
|
|
|
void klp_free_patch_start(struct klp_patch *patch)
|
2014-12-17 01:58:19 +08:00
|
|
|
{
|
|
|
|
if (!list_empty(&patch->list))
|
|
|
|
list_del(&patch->list);
|
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.
|
|
|
|
*/
|
2019-05-03 21:26:24 +08:00
|
|
|
kobject_put(&patch->kobj);
|
|
|
|
wait_for_completion(&patch->finish);
|
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);
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
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)
|
2016-04-28 22:34:08 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
2018-07-20 17:46:42 +08:00
|
|
|
if (strlen(func->old_name) >= KSYM_NAME_LEN)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2015-01-20 23:26:19 +08:00
|
|
|
INIT_LIST_HEAD(&func->stack_node);
|
2017-02-14 09:42:35 +08:00
|
|
|
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;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2015-12-02 10:40:56 +08:00
|
|
|
/* 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.
|
|
|
|
*/
|
2019-05-03 21:26:24 +08:00
|
|
|
return kobject_add(&func->kobj, &obj->kobj, "%s,%lu",
|
|
|
|
func->old_name,
|
|
|
|
func->old_sympos ? func->old_sympos : 1);
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
2016-08-18 08:58:28 +08:00
|
|
|
/* Arches may override this to finish any remaining arch-specific tasks */
|
|
|
|
void __weak arch_klp_init_object_loaded(struct klp_patch *patch,
|
|
|
|
struct klp_object *obj)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
2014-12-17 01:58:19 +08:00
|
|
|
/* 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;
|
|
|
|
|
2019-06-14 09:07:22 +08:00
|
|
|
mutex_lock(&text_mutex);
|
|
|
|
|
2016-08-18 08:58:28 +08:00
|
|
|
module_disable_ro(patch->mod);
|
2016-03-23 08:03:18 +08:00
|
|
|
ret = klp_write_object_relocations(patch->mod, obj);
|
2016-08-18 08:58:28 +08:00
|
|
|
if (ret) {
|
|
|
|
module_enable_ro(patch->mod, true);
|
2019-06-14 09:07:22 +08:00
|
|
|
mutex_unlock(&text_mutex);
|
2016-03-23 08:03:18 +08:00
|
|
|
return ret;
|
2016-08-18 08:58:28 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
arch_klp_init_object_loaded(patch, obj);
|
|
|
|
module_enable_ro(patch->mod, true);
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2019-06-14 09:07:22 +08:00
|
|
|
mutex_unlock(&text_mutex);
|
|
|
|
|
2015-05-19 18:01:19 +08:00
|
|
|
klp_for_each_func(obj, func) {
|
2015-12-02 10:40:54 +08:00
|
|
|
ret = klp_find_object_symbol(obj->name, func->old_name,
|
|
|
|
func->old_sympos,
|
2019-01-09 20:43:19 +08:00
|
|
|
(unsigned long *)&func->old_func);
|
2014-12-17 01:58:19 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2017-02-14 09:42:39 +08:00
|
|
|
|
2019-01-09 20:43:19 +08:00
|
|
|
ret = kallsyms_lookup_size_offset((unsigned long)func->old_func,
|
2017-02-14 09:42:39 +08:00
|
|
|
&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;
|
|
|
|
|
2017-02-14 09:42:39 +08:00
|
|
|
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;
|
|
|
|
}
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int klp_init_object(struct klp_patch *patch, struct klp_object *obj)
|
|
|
|
{
|
|
|
|
struct klp_func *func;
|
|
|
|
int ret;
|
|
|
|
const char *name;
|
|
|
|
|
2018-07-20 17:46:42 +08:00
|
|
|
if (klp_is_module(obj) && strlen(obj->name) >= MODULE_NAME_LEN)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2017-02-14 09:42:35 +08:00
|
|
|
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;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
|
|
|
klp_find_object_module(obj);
|
|
|
|
|
|
|
|
name = klp_is_module(obj) ? obj->name : "vmlinux";
|
2019-05-03 21:26:24 +08:00
|
|
|
ret = kobject_add(&obj->kobj, &patch->kobj, "%s", name);
|
2015-05-19 18:01:18 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2015-05-19 18:01:19 +08:00
|
|
|
klp_for_each_func(obj, func) {
|
2014-12-17 01:58:19 +08:00
|
|
|
ret = klp_init_func(obj, func);
|
|
|
|
if (ret)
|
2019-01-09 20:43:21 +08:00
|
|
|
return ret;
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
2019-01-09 20:43:21 +08:00
|
|
|
if (klp_is_object_loaded(obj))
|
2014-12-17 01:58:19 +08:00
|
|
|
ret = klp_init_object_loaded(patch, obj);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2019-05-03 21:26:25 +08:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2019-01-09 20:43:21 +08:00
|
|
|
static int klp_init_patch_early(struct klp_patch *patch)
|
2014-12-17 01:58:19 +08:00
|
|
|
{
|
|
|
|
struct klp_object *obj;
|
2019-01-09 20:43:21 +08:00
|
|
|
struct klp_func *func;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
|
|
|
if (!patch->objs)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2019-01-09 20:43:21 +08:00
|
|
|
INIT_LIST_HEAD(&patch->list);
|
2019-01-09 20:43:24 +08:00
|
|
|
INIT_LIST_HEAD(&patch->obj_list);
|
2019-05-03 21:26:24 +08:00
|
|
|
kobject_init(&patch->kobj, &klp_ktype_patch);
|
2017-02-14 09:42:35 +08:00
|
|
|
patch->enabled = false;
|
2019-01-09 20:43:22 +08:00
|
|
|
patch->forced = false;
|
2019-01-09 20:43:23 +08:00
|
|
|
INIT_WORK(&patch->free_work, klp_free_patch_work_fn);
|
2017-03-07 01:20:29 +08:00
|
|
|
init_completion(&patch->finish);
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2019-01-09 20:43:24 +08:00
|
|
|
klp_for_each_object_static(patch, obj) {
|
2019-01-09 20:43:21 +08:00
|
|
|
if (!obj->funcs)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2019-05-03 21:26:25 +08:00
|
|
|
klp_init_object_early(patch, obj);
|
2019-01-09 20:43:21 +08:00
|
|
|
|
2019-01-09 20:43:24 +08:00
|
|
|
klp_for_each_func_static(obj, func) {
|
2019-05-03 21:26:25 +08:00
|
|
|
klp_init_func_early(obj, func);
|
2019-01-09 20:43:24 +08:00
|
|
|
}
|
2019-01-09 20:43:21 +08:00
|
|
|
}
|
|
|
|
|
2019-01-09 20:43:23 +08:00
|
|
|
if (!try_module_get(patch->mod))
|
|
|
|
return -ENODEV;
|
|
|
|
|
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;
|
|
|
|
|
2019-05-03 21:26:24 +08:00
|
|
|
ret = kobject_add(&patch->kobj, klp_root_kobj, "%s", patch->mod->name);
|
2019-01-09 20:43:23 +08:00
|
|
|
if (ret)
|
2017-03-07 01:20:29 +08:00
|
|
|
return ret;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2015-05-19 18:01:19 +08:00
|
|
|
klp_for_each_object(patch, obj) {
|
2014-12-17 01:58:19 +08:00
|
|
|
ret = klp_init_object(patch, obj);
|
|
|
|
if (ret)
|
2019-01-09 20:43:23 +08:00
|
|
|
return ret;
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
2015-01-10 04:03:04 +08:00
|
|
|
list_add_tail(&patch->list, &klp_patches);
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2017-03-07 01:20:29 +08:00
|
|
|
return 0;
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
2019-01-09 20:43:20 +08:00
|
|
|
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;
|
2019-01-09 20:43:23 +08:00
|
|
|
klp_try_complete_transition();
|
2019-01-09 20:43:20 +08:00
|
|
|
|
|
|
|
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;
|
2019-01-09 20:43:23 +08:00
|
|
|
klp_try_complete_transition();
|
2019-01-09 20:43:20 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
err:
|
|
|
|
pr_warn("failed to enable patch '%s'\n", patch->mod->name);
|
|
|
|
|
|
|
|
klp_cancel_transition();
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2019-01-09 20:43:23 +08:00
|
|
|
* klp_enable_patch() - enable the livepatch
|
|
|
|
* @patch: patch to be enabled
|
2019-01-09 20:43:20 +08:00
|
|
|
*
|
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.
|
2019-01-09 20:43:20 +08:00
|
|
|
*
|
|
|
|
* Return: 0 on success, otherwise error
|
|
|
|
*/
|
|
|
|
int klp_enable_patch(struct klp_patch *patch)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
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()) {
|
2019-04-24 16:55:48 +08:00
|
|
|
pr_warn("This architecture doesn't have support for the livepatch consistency model.\n");
|
|
|
|
pr_warn("The livepatch transition may never complete.\n");
|
2019-01-09 20:43:23 +08:00
|
|
|
}
|
|
|
|
|
2019-01-09 20:43:20 +08:00
|
|
|
mutex_lock(&klp_mutex);
|
|
|
|
|
2019-01-09 20:43:23 +08:00
|
|
|
ret = klp_init_patch_early(patch);
|
|
|
|
if (ret) {
|
|
|
|
mutex_unlock(&klp_mutex);
|
|
|
|
return ret;
|
2019-01-09 20:43:20 +08:00
|
|
|
}
|
|
|
|
|
2019-01-09 20:43:23 +08:00
|
|
|
ret = klp_init_patch(patch);
|
|
|
|
if (ret)
|
|
|
|
goto err;
|
|
|
|
|
2019-01-09 20:43:20 +08:00
|
|
|
ret = __klp_enable_patch(patch);
|
2019-01-09 20:43:23 +08:00
|
|
|
if (ret)
|
|
|
|
goto err;
|
|
|
|
|
|
|
|
mutex_unlock(&klp_mutex);
|
|
|
|
|
|
|
|
return 0;
|
2019-01-09 20:43:20 +08:00
|
|
|
|
|
|
|
err:
|
2019-01-09 20:43:23 +08:00
|
|
|
klp_free_patch_start(patch);
|
|
|
|
|
2019-01-09 20:43:20 +08:00
|
|
|
mutex_unlock(&klp_mutex);
|
2019-01-09 20:43:23 +08:00
|
|
|
|
|
|
|
klp_free_patch_finish(patch);
|
|
|
|
|
2019-01-09 20:43:20 +08:00
|
|
|
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 removes replaced patches.
|
|
|
|
*
|
|
|
|
* We could be pretty aggressive here. It is called in the situation where
|
|
|
|
* these structures are no longer accessible. 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.
|
|
|
|
*
|
|
|
|
* 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_discard_replaced_patches(struct klp_patch *new_patch)
|
|
|
|
{
|
|
|
|
struct klp_patch *old_patch, *tmp_patch;
|
|
|
|
|
2019-02-04 21:56:50 +08:00
|
|
|
klp_for_each_patch_safe(old_patch, tmp_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);
|
|
|
|
klp_free_patch_start(old_patch);
|
|
|
|
schedule_work(&old_patch->free_work);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2017-10-02 23:56:48 +08:00
|
|
|
/*
|
|
|
|
* 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;
|
|
|
|
|
2019-02-04 21:56:50 +08:00
|
|
|
klp_for_each_patch(patch) {
|
2017-10-02 23:56:48 +08:00
|
|
|
if (patch == limit)
|
|
|
|
break;
|
|
|
|
|
|
|
|
klp_for_each_object(patch, obj) {
|
|
|
|
if (!klp_is_module(obj) || strcmp(obj->name, mod->name))
|
|
|
|
continue;
|
|
|
|
|
2019-02-04 21:56:53 +08:00
|
|
|
if (patch != klp_transition_patch)
|
|
|
|
klp_pre_unpatch_callback(obj);
|
2017-11-15 17:53:24 +08:00
|
|
|
|
2019-02-04 21:56:53 +08:00
|
|
|
pr_notice("reverting patch '%s' on unloading module '%s'\n",
|
|
|
|
patch->mod->name, obj->mod->name);
|
|
|
|
klp_unpatch_object(obj);
|
2017-11-15 17:53:24 +08:00
|
|
|
|
2019-02-04 21:56:53 +08:00
|
|
|
klp_post_unpatch_callback(obj);
|
2017-10-02 23:56:48 +08:00
|
|
|
|
|
|
|
klp_free_object_loaded(obj);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
int klp_module_coming(struct module *mod)
|
2014-12-17 01:58:19 +08:00
|
|
|
{
|
|
|
|
int ret;
|
2016-03-17 08:55:39 +08:00
|
|
|
struct klp_patch *patch;
|
|
|
|
struct klp_object *obj;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
if (WARN_ON(mod->state != MODULE_STATE_COMING))
|
|
|
|
return -EINVAL;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
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;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2019-02-04 21:56:50 +08:00
|
|
|
klp_for_each_patch(patch) {
|
2016-03-17 08:55:39 +08:00
|
|
|
klp_for_each_object(patch, obj) {
|
|
|
|
if (!klp_is_module(obj) || strcmp(obj->name, mod->name))
|
|
|
|
continue;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
obj->mod = mod;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
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;
|
|
|
|
}
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
pr_notice("applying patch '%s' to loading module '%s'\n",
|
|
|
|
patch->mod->name, obj->mod->name);
|
|
|
|
|
2017-10-14 03:08:41 +08:00
|
|
|
ret = klp_pre_patch_callback(obj);
|
|
|
|
if (ret) {
|
|
|
|
pr_warn("pre-patch callback failed for object '%s'\n",
|
|
|
|
obj->name);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
2017-02-14 09:42:35 +08:00
|
|
|
ret = klp_patch_object(obj);
|
2016-03-17 08:55:39 +08:00
|
|
|
if (ret) {
|
|
|
|
pr_warn("failed to apply patch '%s' to module '%s' (%d)\n",
|
|
|
|
patch->mod->name, obj->mod->name, ret);
|
2017-10-14 03:08:41 +08:00
|
|
|
|
2017-10-20 22:56:50 +08:00
|
|
|
klp_post_unpatch_callback(obj);
|
2016-03-17 08:55:39 +08:00
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
2017-10-14 03:08:41 +08:00
|
|
|
if (patch != klp_transition_patch)
|
|
|
|
klp_post_patch_callback(obj);
|
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
mutex_unlock(&klp_mutex);
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
return 0;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
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;
|
2019-07-19 20:28:39 +08:00
|
|
|
obj->mod = NULL;
|
2017-10-02 23:56:48 +08:00
|
|
|
klp_cleanup_module_patches_limited(mod, patch);
|
2016-03-17 08:55:39 +08:00
|
|
|
mutex_unlock(&klp_mutex);
|
|
|
|
|
|
|
|
return ret;
|
2014-12-17 01:58:19 +08:00
|
|
|
}
|
|
|
|
|
2016-03-17 08:55:39 +08:00
|
|
|
void klp_module_going(struct module *mod)
|
2014-12-17 01:58:19 +08:00
|
|
|
{
|
2016-03-17 08:55:39 +08:00
|
|
|
if (WARN_ON(mod->state != MODULE_STATE_GOING &&
|
|
|
|
mod->state != MODULE_STATE_COMING))
|
|
|
|
return;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
|
|
|
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
|
|
|
/*
|
2016-03-17 08:55:39 +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
|
|
|
*/
|
2016-03-17 08:55:39 +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
|
|
|
|
2017-10-02 23:56:48 +08:00
|
|
|
klp_cleanup_module_patches_limited(mod, NULL);
|
2014-12-17 01:58:19 +08:00
|
|
|
|
|
|
|
mutex_unlock(&klp_mutex);
|
|
|
|
}
|
|
|
|
|
2015-05-22 22:26:29 +08:00
|
|
|
static int __init klp_init(void)
|
2014-12-17 01:58:19 +08:00
|
|
|
{
|
|
|
|
klp_root_kobj = kobject_create_and_add("livepatch", kernel_kobj);
|
2016-03-17 08:55:39 +08:00
|
|
|
if (!klp_root_kobj)
|
|
|
|
return -ENOMEM;
|
2014-12-17 01:58:19 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
module_init(klp_init);
|