The recent module patching code has exposed some problems with our data
structures. We currently patch the funcs and dynrelas individually,
which is kind of scary now that different objects can be patched at
different times. Instead it's cleaner and safer to group them by
patched object.
This patch implements per-object patching and relocations by refactoring
the interfaces:
- Completely separate the create-diff-object <-> patch module interface
from the patch module <-> core module interface. create-diff-object
will include "kpatch-patch.h" but not "kpatch.h". Thus,
create-diff-object has no knowledge about the core module's
interfaces, and the core module has no knowledge about the patch
module's special sections.
- Newly added kpatch-patch.h defines the format of the patch module
special sections. It's used by create-diff-object to create the
special sections and used by the patch module to read them.
- kpatch.h still defines the core module interfaces. Each kpatch_module
has a list of kpatch_objects for each module object to be patched.
Each kpatch_object has a list of kpatch_funcs and a list of
kpatch_dynrelas. The patch module creates these lists when populating
kpatch_module.
This way of structuring the data allows us to patch funcs and dynrelas
on a per patched object basis, which will allow us to catch more error
scenarios and make the code easier to manage going forward. It also
allows the use of much more common code between kpatch_register() and
kpatch_module_notify().
This allows a patch module to contain patched functions for modules
which haven't been loaded yet. If/when the module is loaded later, it
will be patched from the module notifier function.
In the replace case, stop calling module_put on a patch module before
we're potentially done with it.
This will also be needed for future module patching if we want to
properly replace a patch module which only patches a future loaded
module (that's a mouthful).
Fixes#165.
Create a list of registered patch modules, which will be used in the
module notifier to patch future loaded modules. It will also be used to
fix the kpatch replace race condition where it calls module_put too
early.
The kpatch_internal struct is a good idea, in that it documents which
parts of kpatch_module shouldn't be used by the patch module. But it
creates extra code and will require more extra code if we want to keep a
list of kpmods, which is needed to create a module notifier for module
patching of future loaded modules.
Embedding the private data directly in the public struct allows the code
to be simpler: no extra kmallocs/kfrees, no need to store pointers
between the public and private structs. I think the simpler code is
worth the tradeoff (exposing implementation detail). Kernel code
usually doesn't bother with hiding a internal struct data from other
kernel code anyway. For example, see ftrace_ops or struct kprobe.
The private fields are documented with a "private" comment.
Move all the ftrace filtering and registering logic into a couple of new
helper functions. Change kpatch_num_registered to kpatch_num_patched,
which now tracks the number of patched functions rather than the number
of patch modules.
This simplifies the code a bit and will also prevent a future loaded
module scenario where ftrace_ops can be registered with an empty filter,
resulting in _all_ kernel functions getting registered with ftrace.
Use single quotes when printing the name of a patch module, rather than
double quotes. This is more consistent with other printk messages, and
looks better too!
If kpatch_ftrace_add_func fails, num_funcs will be one less than what it
needs to be for kpatch_put_modules to work properly. Instead give it
the full array size, and it can figure out which modules to put based on
whether func->mod is nonzero.
This commit adds support for patching modules. If a patch or dynrela is
determined to be for a kernel module, the old_offset/src is not used and
the symbol location is looked up using kallsyms. The module being
patched is also references to keep if from disappearing from underneath
us.
This commit introduces early and limited support. The kernel module to
be patched must already be loaded or the patch module will not apply.
Signed-off-by: Seth Jennings <sjenning@redhat.com>
Make old addresses relative to the start address of the relocatable
kernel or module.
This commit has no functional effect; it just prepares the code for
future acceptance of the module patching support.
Signed-off-by: Seth Jennings <sjenning@redhat.com>
Ensure that dynrela destination addresses are within the patch module's
memory.
Also, use the module address ranges to check whether set_memory_rw() is
needed.
The -fdata-sections gcc flag doesn't work with objects in the
.data..percpu section. Any function which uses a percpu variable
references this section, causing the section to get incorrectly included
in the patch module.
Manually convert these section references to object symbol references so
that the needed symbol can be found in vmlinux.
Also, the core module symbol verification code will fail when looking up
a percpu variable, because sprint_symbol doesn't think a percpu address
is a valid kernel address. So rewrite the symbol verification code to
use kallsyms_on_each_symbol() instead. It's not ideal performance-wise:
it seems to cost about 1ms per symbol lookup. I think that's acceptable
for now. In the future we may want to try to get a better upstream
kallsyms interface.
This commit introduces functionality to verify the location of symbols
used in both the patch and dynrelas sections. It adds significant
protection from mismatches between the base and running kernels.
Signed-off-by: Seth Jennings <sjenning@redhat.com>
This adds dynamic linking support for the patch modules. It is the
first step toward supporting patching module code and relocatable
kernels.
Rela entries that reference non-included local and non-exported global
symbols are converted to "dynrelas". These dynrelas are relocations
that are done by the core module, not the kernel module linker. This
allows the core module to apply offsets to the base addresses found
in the base vmlinux or module.
Signed-off-by: Seth Jennings <sjenning@redhat.com>
Conflicts:
kpatch-build/kpatch-build
We merged PR #186 a little too hastily. It seg faults with the new
parainstructions-section.patch in the integration test suite. Reverting
it for now until we get it figured out.
This reverts commit e1177e3a03.
This reverts commit 880e271841.
This reverts commit 2de5f6cbfb.
This reverts commit 38b7ac74ad.
This reverts commit 108cd9f95e.
This adds dynamic linking support for the patch modules. It is the
first step toward supporting patching module code and relocatable
kernels.
Rela entries that reference non-included local and non-exported global
symbols are converted to "dynrelas". These dynrelas are relocations
that are done by the core module, not the kernel module linker. This
allows the core module to apply offsets to the base addresses found
in the base vmlinux or module.
Signed-off-by: Seth Jennings <sjenning@redhat.com>
Make kpatch_funs truly internal by:
Defining it in core.c
Adding a struct kpatch_internal, declared in kpatch.h and defined in
core.c, that contains per patch module internal data.
Adding an "internal" field to struct kpatch_modules.
Allocating internal and funcs data in core.c, not in the patch module,
since the patch module has no knowledge of kpatch_func anymore.
Adding a "patch" field to kpatch_func that points directly to the
kpatch_patch provided by the module (rather than a field-by-field copy)
Signed-off-by: Seth Jennings <sjenning@redhat.com>
The error likes:
make -C /lib/modules/3.11.0-12-generic/build M=/home/ryan/kpatch/kmod/core kpatch.ko
make[3]: Entering directory `/usr/src/linux-headers-3.11.0-12-generic'
CC [M] /home/ryan/kpatch/kmod/core/core.o
/home/ryan/kpatch/kmod/core/core.c:42:32: fatal error: linux/preempt_mask.h: No such file or directory
#include <linux/preempt_mask.h>
I feel sorry to introduce this problem from my laster commit 6c2d6444.
Some old kernel doesn't have header file preempt_mask.h, and the safe
way is using hardirq.h to find in_nmi().
Signed-off-by: Jincheng Miao <jincheng.miao@gmail.com>
It's probably cleaner to just return -EINVAL if the kpmod isn't enabled,
instead of warning and continuing, which would be dangerous.
I think the reason I had it WARN before is because this condition
shouldn't be possible given the source for the patch module. But the
core module can't necessarily assume that it's our trustworthy patch
module code on the other side.
Allow the user to atomically replace all existing modules with a new
"kpatch replace" command. This provides a safe way to do atomic
upgrades for cumulative patch module updates.
- checkpatch doesn't like the FSF address since it's subject to change
- checkpatch doesn't like strings split by line
- whitespace fix
- sparse suggested to change some variables and functions to static
This is an attempt to both simplify and improve the correctness of the
NMI synchronization code.
There's a race in kpatch_ftrace_handler() between the kpatch_get_func()
and kpatch_finish_status() calls which could result in func being NULL.
The retry was supposed to fix this. However, this race would still be a
problem in the repatching case (if the function had already been
previously patched), in which case func would not be NULL, but could
instead point to the previously patched version of the function. In
this case it wouldn't retry and it would be possible for the previous
version of the function to run.
The fix is to use a memory barrier between accesses of the func hash and
the status variable, and then just call kpatch_get_func() *after*
accessing the status variable. For OP_PATCH, if status is SUCCESS, then
func is guaranteed to point to the new function. If status is FAILURE,
func might point to the new function, in which case we can use
get_prev_func to get the previous version of the function.
I also made some pretty big changes to try to simplify the design so
that there are less moving parts and so that it's hopefully easier to
understand. I moved the OP field into the kpatch_func struct. This
allows us to merge the two global state variables (status + op) into a
single global state variable (state), which helps make the code quite a
bit simpler. I turned it into a proper state machine and documented the
meaning of each state in the comments.
Moving the OP field to the kpatch_func struct also paves the way for an
upcoming pull request which will allow patch modules to be atomically
replaced ("kpatch load --replace <module>").
In kpatch_unregister(), if kpatch_remove_patch succeeds but one of the
subsequent ftrace unregistering calls fails, it returns an error and
fails to module_put() the patch module, even though the patch has been
removed. This causes the patch module to get stuck in a weird place
where its patch has been unregistered but the patch module can't ever be
removed.
These errors aren't serious and wouldn't cause any real problems if they
did somehow fail, so instead just WARN if they fail.
Currently the patch module calls kpatch_unregister in the patch module
exit path. If the activeness safety check fails in kpatch_unregister,
it's too late for the patch module to stop exiting, so all it can do is
panic.
Prevent this scenario by requiring the user to disable the patch module
via sysfs before allowing the module to be unloaded. The sysfs write
will fail if the activeness safety check fails. An rmmod will fail if
the patch is still enabled.
Also add support for this new unloading model in "kpatch unload".
When compiling core.c, it may report error like:
"error: implicit declaration of function ‘in_nmi’"
Adding header file in_nmi defined could avoid this.
Signed-off-by: Jincheng Miao <jincheng.miao@gmail.com>
