linux/Documentation/fault-injection/fault-injection.rst

552 lines
16 KiB
ReStructuredText
Raw Normal View History

===========================================
Fault injection capabilities infrastructure
===========================================
See also drivers/md/md-faulty.c and "every_nth" module option for scsi_debug.
Available fault injection capabilities
--------------------------------------
- failslab
injects slab allocation failures. (kmalloc(), kmem_cache_alloc(), ...)
- fail_page_alloc
injects page allocation failures. (alloc_pages(), get_free_pages(), ...)
lib, include/linux: add usercopy failure capability Patch series "add fault injection to user memory access", v3. The goal of this series is to improve testing of fault-tolerance in usages of user memory access functions, by adding support for fault injection. syzkaller/syzbot are using the existing fault injection modes and will use this particular feature also. The first patch adds failure injection capability for usercopy functions. The second changes usercopy functions to use this new failure capability (copy_from_user, ...). The third patch adds get/put/clear_user failures to x86. This patch (of 3): Add a failure injection capability to improve testing of fault-tolerance in usages of user memory access functions. Add CONFIG_FAULT_INJECTION_USERCOPY to enable faults in usercopy functions. The should_fail_usercopy function is to be called by these functions (copy_from_user, get_user, ...) in order to fail or not. Signed-off-by: Albert van der Linde <alinde@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Akinobu Mita <akinobu.mita@gmail.com> Reviewed-by: Alexander Potapenko <glider@google.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Andrey Konovalov <andreyknvl@google.com> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Marco Elver <elver@google.com> Cc: Christoph Hellwig <hch@lst.de> Link: http://lkml.kernel.org/r/20200831171733.955393-1-alinde@google.com Link: http://lkml.kernel.org/r/20200831171733.955393-2-alinde@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 11:13:46 +08:00
- fail_usercopy
injects failures in user memory access functions. (copy_from_user(), get_user(), ...)
- fail_futex
injects futex deadlock and uaddr fault errors.
- fail_sunrpc
injects kernel RPC client and server failures.
- fail_make_request
injects disk IO errors on devices permitted by setting
/sys/block/<device>/make-it-fail or
/sys/block/<device>/<partition>/make-it-fail. (submit_bio_noacct())
- fail_mmc_request
injects MMC data errors on devices permitted by setting
debugfs entries under /sys/kernel/debug/mmc0/fail_mmc_request
- fail_function
injects error return on specific functions, which are marked by
ALLOW_ERROR_INJECTION() macro, by setting debugfs entries
under /sys/kernel/debug/fail_function. No boot option supported.
- NVMe fault injection
inject NVMe status code and retry flag on devices permitted by setting
debugfs entries under /sys/kernel/debug/nvme*/fault_inject. The default
status code is NVME_SC_INVALID_OPCODE with no retry. The status code and
retry flag can be set via the debugfs.
- Null test block driver fault injection
inject IO timeouts by setting config items under
/sys/kernel/config/nullb/<disk>/timeout_inject,
inject requeue requests by setting config items under
/sys/kernel/config/nullb/<disk>/requeue_inject, and
inject init_hctx() errors by setting config items under
/sys/kernel/config/nullb/<disk>/init_hctx_fault_inject.
Configure fault-injection capabilities behavior
-----------------------------------------------
debugfs entries
^^^^^^^^^^^^^^^
fault-inject-debugfs kernel module provides some debugfs entries for runtime
configuration of fault-injection capabilities.
- /sys/kernel/debug/fail*/probability:
likelihood of failure injection, in percent.
Format: <percent>
Note that one-failure-per-hundred is a very high error rate
for some testcases. Consider setting probability=100 and configure
/sys/kernel/debug/fail*/interval for such testcases.
- /sys/kernel/debug/fail*/interval:
specifies the interval between failures, for calls to
should_fail() that pass all the other tests.
Note that if you enable this, by setting interval>1, you will
probably want to set probability=100.
- /sys/kernel/debug/fail*/times:
specifies how many times failures may happen at most. A value of -1
means "no limit".
- /sys/kernel/debug/fail*/space:
specifies an initial resource "budget", decremented by "size"
on each call to should_fail(,size). Failure injection is
suppressed until "space" reaches zero.
- /sys/kernel/debug/fail*/verbose
Format: { 0 | 1 | 2 }
specifies the verbosity of the messages when failure is
injected. '0' means no messages; '1' will print only a single
log line per failure; '2' will print a call trace too -- useful
to debug the problems revealed by fault injection.
- /sys/kernel/debug/fail*/task-filter:
Format: { 'Y' | 'N' }
A value of 'N' disables filtering by process (default).
Any positive value limits failures to only processes indicated by
/proc/<pid>/make-it-fail==1.
- /sys/kernel/debug/fail*/require-start,
/sys/kernel/debug/fail*/require-end,
/sys/kernel/debug/fail*/reject-start,
/sys/kernel/debug/fail*/reject-end:
specifies the range of virtual addresses tested during
stacktrace walking. Failure is injected only if some caller
in the walked stacktrace lies within the required range, and
none lies within the rejected range.
Default required range is [0,ULONG_MAX) (whole of virtual address space).
Default rejected range is [0,0).
- /sys/kernel/debug/fail*/stacktrace-depth:
specifies the maximum stacktrace depth walked during search
for a caller within [require-start,require-end) OR
[reject-start,reject-end).
- /sys/kernel/debug/fail_page_alloc/ignore-gfp-highmem:
Format: { 'Y' | 'N' }
default is 'Y', setting it to 'N' will also inject failures into
highmem/user allocations (__GFP_HIGHMEM allocations).
- /sys/kernel/debug/failslab/ignore-gfp-wait:
- /sys/kernel/debug/fail_page_alloc/ignore-gfp-wait:
Format: { 'Y' | 'N' }
default is 'Y', setting it to 'N' will also inject failures
into allocations that can sleep (__GFP_DIRECT_RECLAIM allocations).
- /sys/kernel/debug/fail_page_alloc/min-order:
specifies the minimum page allocation order to be injected
failures.
- /sys/kernel/debug/fail_futex/ignore-private:
Format: { 'Y' | 'N' }
default is 'N', setting it to 'Y' will disable failure injections
when dealing with private (address space) futexes.
- /sys/kernel/debug/fail_sunrpc/ignore-client-disconnect:
Format: { 'Y' | 'N' }
default is 'N', setting it to 'Y' will disable disconnect
injection on the RPC client.
- /sys/kernel/debug/fail_sunrpc/ignore-server-disconnect:
Format: { 'Y' | 'N' }
default is 'N', setting it to 'Y' will disable disconnect
injection on the RPC server.
- /sys/kernel/debug/fail_sunrpc/ignore-cache-wait:
Format: { 'Y' | 'N' }
default is 'N', setting it to 'Y' will disable cache wait
injection on the RPC server.
- /sys/kernel/debug/fail_function/inject:
Format: { 'function-name' | '!function-name' | '' }
specifies the target function of error injection by name.
If the function name leads '!' prefix, given function is
removed from injection list. If nothing specified ('')
injection list is cleared.
- /sys/kernel/debug/fail_function/injectable:
(read only) shows error injectable functions and what type of
error values can be specified. The error type will be one of
below;
- NULL: retval must be 0.
- ERRNO: retval must be -1 to -MAX_ERRNO (-4096).
- ERR_NULL: retval must be 0 or -1 to -MAX_ERRNO (-4096).
- /sys/kernel/debug/fail_function/<function-name>/retval:
specifies the "error" return value to inject to the given function.
This will be created when the user specifies a new injection entry.
Note that this file only accepts unsigned values. So, if you want to
use a negative errno, you better use 'printf' instead of 'echo', e.g.:
$ printf %#x -12 > retval
Boot option
^^^^^^^^^^^
In order to inject faults while debugfs is not available (early boot time),
use the boot option::
failslab=
fail_page_alloc=
lib, include/linux: add usercopy failure capability Patch series "add fault injection to user memory access", v3. The goal of this series is to improve testing of fault-tolerance in usages of user memory access functions, by adding support for fault injection. syzkaller/syzbot are using the existing fault injection modes and will use this particular feature also. The first patch adds failure injection capability for usercopy functions. The second changes usercopy functions to use this new failure capability (copy_from_user, ...). The third patch adds get/put/clear_user failures to x86. This patch (of 3): Add a failure injection capability to improve testing of fault-tolerance in usages of user memory access functions. Add CONFIG_FAULT_INJECTION_USERCOPY to enable faults in usercopy functions. The should_fail_usercopy function is to be called by these functions (copy_from_user, get_user, ...) in order to fail or not. Signed-off-by: Albert van der Linde <alinde@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Akinobu Mita <akinobu.mita@gmail.com> Reviewed-by: Alexander Potapenko <glider@google.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Andrey Konovalov <andreyknvl@google.com> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Marco Elver <elver@google.com> Cc: Christoph Hellwig <hch@lst.de> Link: http://lkml.kernel.org/r/20200831171733.955393-1-alinde@google.com Link: http://lkml.kernel.org/r/20200831171733.955393-2-alinde@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 11:13:46 +08:00
fail_usercopy=
fail_make_request=
fail_futex=
mmc_core.fail_request=<interval>,<probability>,<space>,<times>
proc entries
^^^^^^^^^^^^
fault-inject: support systematic fault injection Add /proc/self/task/<current-tid>/fail-nth file that allows failing 0-th, 1-st, 2-nd and so on calls systematically. Excerpt from the added documentation: "Write to this file of integer N makes N-th call in the current task fail (N is 0-based). Read from this file returns a single char 'Y' or 'N' that says if the fault setup with a previous write to this file was injected or not, and disables the fault if it wasn't yet injected. Note that this file enables all types of faults (slab, futex, etc). This setting takes precedence over all other generic settings like probability, interval, times, etc. But per-capability settings (e.g. fail_futex/ignore-private) take precedence over it. This feature is intended for systematic testing of faults in a single system call. See an example below" Why add a new setting: 1. Existing settings are global rather than per-task. So parallel testing is not possible. 2. attr->interval is close but it depends on attr->count which is non reset to 0, so interval does not work as expected. 3. Trying to model this with existing settings requires manipulations of all of probability, interval, times, space, task-filter and unexposed count and per-task make-it-fail files. 4. Existing settings are per-failure-type, and the set of failure types is potentially expanding. 5. make-it-fail can't be changed by unprivileged user and aggressive stress testing better be done from an unprivileged user. Similarly, this would require opening the debugfs files to the unprivileged user, as he would need to reopen at least times file (not possible to pre-open before dropping privs). The proposed interface solves all of the above (see the example). We want to integrate this into syzkaller fuzzer. A prototype has found 10 bugs in kernel in first day of usage: https://groups.google.com/forum/#!searchin/syzkaller/%22FAULT_INJECTION%22%7Csort:relevance I've made the current interface work with all types of our sandboxes. For setuid the secret sauce was prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) to make /proc entries non-root owned. So I am fine with the current version of the code. [akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/20170328130128.101773-1-dvyukov@google.com Signed-off-by: Dmitry Vyukov <dvyukov@google.com> Cc: Akinobu Mita <akinobu.mita@gmail.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-13 05:34:35 +08:00
- /proc/<pid>/fail-nth,
/proc/self/task/<tid>/fail-nth:
fault-inject: support systematic fault injection Add /proc/self/task/<current-tid>/fail-nth file that allows failing 0-th, 1-st, 2-nd and so on calls systematically. Excerpt from the added documentation: "Write to this file of integer N makes N-th call in the current task fail (N is 0-based). Read from this file returns a single char 'Y' or 'N' that says if the fault setup with a previous write to this file was injected or not, and disables the fault if it wasn't yet injected. Note that this file enables all types of faults (slab, futex, etc). This setting takes precedence over all other generic settings like probability, interval, times, etc. But per-capability settings (e.g. fail_futex/ignore-private) take precedence over it. This feature is intended for systematic testing of faults in a single system call. See an example below" Why add a new setting: 1. Existing settings are global rather than per-task. So parallel testing is not possible. 2. attr->interval is close but it depends on attr->count which is non reset to 0, so interval does not work as expected. 3. Trying to model this with existing settings requires manipulations of all of probability, interval, times, space, task-filter and unexposed count and per-task make-it-fail files. 4. Existing settings are per-failure-type, and the set of failure types is potentially expanding. 5. make-it-fail can't be changed by unprivileged user and aggressive stress testing better be done from an unprivileged user. Similarly, this would require opening the debugfs files to the unprivileged user, as he would need to reopen at least times file (not possible to pre-open before dropping privs). The proposed interface solves all of the above (see the example). We want to integrate this into syzkaller fuzzer. A prototype has found 10 bugs in kernel in first day of usage: https://groups.google.com/forum/#!searchin/syzkaller/%22FAULT_INJECTION%22%7Csort:relevance I've made the current interface work with all types of our sandboxes. For setuid the secret sauce was prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) to make /proc entries non-root owned. So I am fine with the current version of the code. [akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/20170328130128.101773-1-dvyukov@google.com Signed-off-by: Dmitry Vyukov <dvyukov@google.com> Cc: Akinobu Mita <akinobu.mita@gmail.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-13 05:34:35 +08:00
Write to this file of integer N makes N-th call in the task fail.
Read from this file returns a integer value. A value of '0' indicates
that the fault setup with a previous write to this file was injected.
A positive integer N indicates that the fault wasn't yet injected.
fault-inject: support systematic fault injection Add /proc/self/task/<current-tid>/fail-nth file that allows failing 0-th, 1-st, 2-nd and so on calls systematically. Excerpt from the added documentation: "Write to this file of integer N makes N-th call in the current task fail (N is 0-based). Read from this file returns a single char 'Y' or 'N' that says if the fault setup with a previous write to this file was injected or not, and disables the fault if it wasn't yet injected. Note that this file enables all types of faults (slab, futex, etc). This setting takes precedence over all other generic settings like probability, interval, times, etc. But per-capability settings (e.g. fail_futex/ignore-private) take precedence over it. This feature is intended for systematic testing of faults in a single system call. See an example below" Why add a new setting: 1. Existing settings are global rather than per-task. So parallel testing is not possible. 2. attr->interval is close but it depends on attr->count which is non reset to 0, so interval does not work as expected. 3. Trying to model this with existing settings requires manipulations of all of probability, interval, times, space, task-filter and unexposed count and per-task make-it-fail files. 4. Existing settings are per-failure-type, and the set of failure types is potentially expanding. 5. make-it-fail can't be changed by unprivileged user and aggressive stress testing better be done from an unprivileged user. Similarly, this would require opening the debugfs files to the unprivileged user, as he would need to reopen at least times file (not possible to pre-open before dropping privs). The proposed interface solves all of the above (see the example). We want to integrate this into syzkaller fuzzer. A prototype has found 10 bugs in kernel in first day of usage: https://groups.google.com/forum/#!searchin/syzkaller/%22FAULT_INJECTION%22%7Csort:relevance I've made the current interface work with all types of our sandboxes. For setuid the secret sauce was prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) to make /proc entries non-root owned. So I am fine with the current version of the code. [akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/20170328130128.101773-1-dvyukov@google.com Signed-off-by: Dmitry Vyukov <dvyukov@google.com> Cc: Akinobu Mita <akinobu.mita@gmail.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-13 05:34:35 +08:00
Note that this file enables all types of faults (slab, futex, etc).
This setting takes precedence over all other generic debugfs settings
like probability, interval, times, etc. But per-capability settings
(e.g. fail_futex/ignore-private) take precedence over it.
This feature is intended for systematic testing of faults in a single
system call. See an example below.
Error Injectable Functions
--------------------------
This part is for the kernel developers considering to add a function to
ALLOW_ERROR_INJECTION() macro.
Requirements for the Error Injectable Functions
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Since the function-level error injection forcibly changes the code path
and returns an error even if the input and conditions are proper, this can
cause unexpected kernel crash if you allow error injection on the function
which is NOT error injectable. Thus, you (and reviewers) must ensure;
- The function returns an error code if it fails, and the callers must check
it correctly (need to recover from it).
- The function does not execute any code which can change any state before
the first error return. The state includes global or local, or input
variable. For example, clear output address storage (e.g. `*ret = NULL`),
increments/decrements counter, set a flag, preempt/irq disable or get
a lock (if those are recovered before returning error, that will be OK.)
The first requirement is important, and it will result in that the release
(free objects) functions are usually harder to inject errors than allocate
functions. If errors of such release functions are not correctly handled
it will cause a memory leak easily (the caller will confuse that the object
has been released or corrupted.)
The second one is for the caller which expects the function should always
does something. Thus if the function error injection skips whole of the
function, the expectation is betrayed and causes an unexpected error.
Type of the Error Injectable Functions
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Each error injectable functions will have the error type specified by the
ALLOW_ERROR_INJECTION() macro. You have to choose it carefully if you add
a new error injectable function. If the wrong error type is chosen, the
kernel may crash because it may not be able to handle the error.
There are 4 types of errors defined in include/asm-generic/error-injection.h
EI_ETYPE_NULL
This function will return `NULL` if it fails. e.g. return an allocateed
object address.
EI_ETYPE_ERRNO
This function will return an `-errno` error code if it fails. e.g. return
-EINVAL if the input is wrong. This will include the functions which will
return an address which encodes `-errno` by ERR_PTR() macro.
EI_ETYPE_ERRNO_NULL
This function will return an `-errno` or `NULL` if it fails. If the caller
of this function checks the return value with IS_ERR_OR_NULL() macro, this
type will be appropriate.
EI_ETYPE_TRUE
This function will return `true` (non-zero positive value) if it fails.
If you specifies a wrong type, for example, EI_TYPE_ERRNO for the function
which returns an allocated object, it may cause a problem because the returned
value is not an object address and the caller can not access to the address.
How to add new fault injection capability
-----------------------------------------
- #include <linux/fault-inject.h>
- define the fault attributes
DECLARE_FAULT_ATTR(name);
Please see the definition of struct fault_attr in fault-inject.h
for details.
- provide a way to configure fault attributes
- boot option
If you need to enable the fault injection capability from boot time, you can
provide boot option to configure it. There is a helper function for it:
setup_fault_attr(attr, str);
- debugfs entries
lib, include/linux: add usercopy failure capability Patch series "add fault injection to user memory access", v3. The goal of this series is to improve testing of fault-tolerance in usages of user memory access functions, by adding support for fault injection. syzkaller/syzbot are using the existing fault injection modes and will use this particular feature also. The first patch adds failure injection capability for usercopy functions. The second changes usercopy functions to use this new failure capability (copy_from_user, ...). The third patch adds get/put/clear_user failures to x86. This patch (of 3): Add a failure injection capability to improve testing of fault-tolerance in usages of user memory access functions. Add CONFIG_FAULT_INJECTION_USERCOPY to enable faults in usercopy functions. The should_fail_usercopy function is to be called by these functions (copy_from_user, get_user, ...) in order to fail or not. Signed-off-by: Albert van der Linde <alinde@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Akinobu Mita <akinobu.mita@gmail.com> Reviewed-by: Alexander Potapenko <glider@google.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Andrey Konovalov <andreyknvl@google.com> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Marco Elver <elver@google.com> Cc: Christoph Hellwig <hch@lst.de> Link: http://lkml.kernel.org/r/20200831171733.955393-1-alinde@google.com Link: http://lkml.kernel.org/r/20200831171733.955393-2-alinde@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 11:13:46 +08:00
failslab, fail_page_alloc, fail_usercopy, and fail_make_request use this way.
Helper functions:
fault_create_debugfs_attr(name, parent, attr);
- module parameters
If the scope of the fault injection capability is limited to a
single kernel module, it is better to provide module parameters to
configure the fault attributes.
- add a hook to insert failures
Upon should_fail() returning true, client code should inject a failure:
should_fail(attr, size);
Application Examples
--------------------
- Inject slab allocation failures into module init/exit code::
#!/bin/bash
FAILTYPE=failslab
echo Y > /sys/kernel/debug/$FAILTYPE/task-filter
echo 10 > /sys/kernel/debug/$FAILTYPE/probability
echo 100 > /sys/kernel/debug/$FAILTYPE/interval
echo -1 > /sys/kernel/debug/$FAILTYPE/times
echo 0 > /sys/kernel/debug/$FAILTYPE/space
echo 2 > /sys/kernel/debug/$FAILTYPE/verbose
echo Y > /sys/kernel/debug/$FAILTYPE/ignore-gfp-wait
faulty_system()
{
bash -c "echo 1 > /proc/self/make-it-fail && exec $*"
}
if [ $# -eq 0 ]
then
echo "Usage: $0 modulename [ modulename ... ]"
exit 1
fi
for m in $*
do
echo inserting $m...
faulty_system modprobe $m
echo removing $m...
faulty_system modprobe -r $m
done
------------------------------------------------------------------------------
- Inject page allocation failures only for a specific module::
#!/bin/bash
FAILTYPE=fail_page_alloc
module=$1
if [ -z $module ]
then
echo "Usage: $0 <modulename>"
exit 1
fi
modprobe $module
if [ ! -d /sys/module/$module/sections ]
then
echo Module $module is not loaded
exit 1
fi
cat /sys/module/$module/sections/.text > /sys/kernel/debug/$FAILTYPE/require-start
cat /sys/module/$module/sections/.data > /sys/kernel/debug/$FAILTYPE/require-end
echo N > /sys/kernel/debug/$FAILTYPE/task-filter
echo 10 > /sys/kernel/debug/$FAILTYPE/probability
echo 100 > /sys/kernel/debug/$FAILTYPE/interval
echo -1 > /sys/kernel/debug/$FAILTYPE/times
echo 0 > /sys/kernel/debug/$FAILTYPE/space
echo 2 > /sys/kernel/debug/$FAILTYPE/verbose
echo Y > /sys/kernel/debug/$FAILTYPE/ignore-gfp-wait
echo Y > /sys/kernel/debug/$FAILTYPE/ignore-gfp-highmem
echo 10 > /sys/kernel/debug/$FAILTYPE/stacktrace-depth
trap "echo 0 > /sys/kernel/debug/$FAILTYPE/probability" SIGINT SIGTERM EXIT
echo "Injecting errors into the module $module... (interrupt to stop)"
sleep 1000000
------------------------------------------------------------------------------
- Inject open_ctree error while btrfs mount::
#!/bin/bash
rm -f testfile.img
dd if=/dev/zero of=testfile.img bs=1M seek=1000 count=1
DEVICE=$(losetup --show -f testfile.img)
mkfs.btrfs -f $DEVICE
mkdir -p tmpmnt
FAILTYPE=fail_function
FAILFUNC=open_ctree
echo $FAILFUNC > /sys/kernel/debug/$FAILTYPE/inject
printf %#x -12 > /sys/kernel/debug/$FAILTYPE/$FAILFUNC/retval
echo N > /sys/kernel/debug/$FAILTYPE/task-filter
echo 100 > /sys/kernel/debug/$FAILTYPE/probability
echo 0 > /sys/kernel/debug/$FAILTYPE/interval
echo -1 > /sys/kernel/debug/$FAILTYPE/times
echo 0 > /sys/kernel/debug/$FAILTYPE/space
echo 1 > /sys/kernel/debug/$FAILTYPE/verbose
mount -t btrfs $DEVICE tmpmnt
if [ $? -ne 0 ]
then
echo "SUCCESS!"
else
echo "FAILED!"
umount tmpmnt
fi
echo > /sys/kernel/debug/$FAILTYPE/inject
rmdir tmpmnt
losetup -d $DEVICE
rm testfile.img
Tool to run command with failslab or fail_page_alloc
----------------------------------------------------
In order to make it easier to accomplish the tasks mentioned above, we can use
tools/testing/fault-injection/failcmd.sh. Please run a command
"./tools/testing/fault-injection/failcmd.sh --help" for more information and
see the following examples.
Examples:
Run a command "make -C tools/testing/selftests/ run_tests" with injecting slab
allocation failure::
# ./tools/testing/fault-injection/failcmd.sh \
-- make -C tools/testing/selftests/ run_tests
Same as above except to specify 100 times failures at most instead of one time
at most by default::
# ./tools/testing/fault-injection/failcmd.sh --times=100 \
-- make -C tools/testing/selftests/ run_tests
Same as above except to inject page allocation failure instead of slab
allocation failure::
# env FAILCMD_TYPE=fail_page_alloc \
./tools/testing/fault-injection/failcmd.sh --times=100 \
-- make -C tools/testing/selftests/ run_tests
fault-inject: support systematic fault injection Add /proc/self/task/<current-tid>/fail-nth file that allows failing 0-th, 1-st, 2-nd and so on calls systematically. Excerpt from the added documentation: "Write to this file of integer N makes N-th call in the current task fail (N is 0-based). Read from this file returns a single char 'Y' or 'N' that says if the fault setup with a previous write to this file was injected or not, and disables the fault if it wasn't yet injected. Note that this file enables all types of faults (slab, futex, etc). This setting takes precedence over all other generic settings like probability, interval, times, etc. But per-capability settings (e.g. fail_futex/ignore-private) take precedence over it. This feature is intended for systematic testing of faults in a single system call. See an example below" Why add a new setting: 1. Existing settings are global rather than per-task. So parallel testing is not possible. 2. attr->interval is close but it depends on attr->count which is non reset to 0, so interval does not work as expected. 3. Trying to model this with existing settings requires manipulations of all of probability, interval, times, space, task-filter and unexposed count and per-task make-it-fail files. 4. Existing settings are per-failure-type, and the set of failure types is potentially expanding. 5. make-it-fail can't be changed by unprivileged user and aggressive stress testing better be done from an unprivileged user. Similarly, this would require opening the debugfs files to the unprivileged user, as he would need to reopen at least times file (not possible to pre-open before dropping privs). The proposed interface solves all of the above (see the example). We want to integrate this into syzkaller fuzzer. A prototype has found 10 bugs in kernel in first day of usage: https://groups.google.com/forum/#!searchin/syzkaller/%22FAULT_INJECTION%22%7Csort:relevance I've made the current interface work with all types of our sandboxes. For setuid the secret sauce was prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) to make /proc entries non-root owned. So I am fine with the current version of the code. [akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/20170328130128.101773-1-dvyukov@google.com Signed-off-by: Dmitry Vyukov <dvyukov@google.com> Cc: Akinobu Mita <akinobu.mita@gmail.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-13 05:34:35 +08:00
Systematic faults using fail-nth
---------------------------------
The following code systematically faults 0-th, 1-st, 2-nd and so on
capabilities in the socketpair() system call::
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <sys/syscall.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>
#include <errno.h>
int main()
{
fault-inject: support systematic fault injection Add /proc/self/task/<current-tid>/fail-nth file that allows failing 0-th, 1-st, 2-nd and so on calls systematically. Excerpt from the added documentation: "Write to this file of integer N makes N-th call in the current task fail (N is 0-based). Read from this file returns a single char 'Y' or 'N' that says if the fault setup with a previous write to this file was injected or not, and disables the fault if it wasn't yet injected. Note that this file enables all types of faults (slab, futex, etc). This setting takes precedence over all other generic settings like probability, interval, times, etc. But per-capability settings (e.g. fail_futex/ignore-private) take precedence over it. This feature is intended for systematic testing of faults in a single system call. See an example below" Why add a new setting: 1. Existing settings are global rather than per-task. So parallel testing is not possible. 2. attr->interval is close but it depends on attr->count which is non reset to 0, so interval does not work as expected. 3. Trying to model this with existing settings requires manipulations of all of probability, interval, times, space, task-filter and unexposed count and per-task make-it-fail files. 4. Existing settings are per-failure-type, and the set of failure types is potentially expanding. 5. make-it-fail can't be changed by unprivileged user and aggressive stress testing better be done from an unprivileged user. Similarly, this would require opening the debugfs files to the unprivileged user, as he would need to reopen at least times file (not possible to pre-open before dropping privs). The proposed interface solves all of the above (see the example). We want to integrate this into syzkaller fuzzer. A prototype has found 10 bugs in kernel in first day of usage: https://groups.google.com/forum/#!searchin/syzkaller/%22FAULT_INJECTION%22%7Csort:relevance I've made the current interface work with all types of our sandboxes. For setuid the secret sauce was prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) to make /proc entries non-root owned. So I am fine with the current version of the code. [akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/20170328130128.101773-1-dvyukov@google.com Signed-off-by: Dmitry Vyukov <dvyukov@google.com> Cc: Akinobu Mita <akinobu.mita@gmail.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-13 05:34:35 +08:00
int i, err, res, fail_nth, fds[2];
char buf[128];
system("echo N > /sys/kernel/debug/failslab/ignore-gfp-wait");
sprintf(buf, "/proc/self/task/%ld/fail-nth", syscall(SYS_gettid));
fail_nth = open(buf, O_RDWR);
for (i = 1;; i++) {
fault-inject: support systematic fault injection Add /proc/self/task/<current-tid>/fail-nth file that allows failing 0-th, 1-st, 2-nd and so on calls systematically. Excerpt from the added documentation: "Write to this file of integer N makes N-th call in the current task fail (N is 0-based). Read from this file returns a single char 'Y' or 'N' that says if the fault setup with a previous write to this file was injected or not, and disables the fault if it wasn't yet injected. Note that this file enables all types of faults (slab, futex, etc). This setting takes precedence over all other generic settings like probability, interval, times, etc. But per-capability settings (e.g. fail_futex/ignore-private) take precedence over it. This feature is intended for systematic testing of faults in a single system call. See an example below" Why add a new setting: 1. Existing settings are global rather than per-task. So parallel testing is not possible. 2. attr->interval is close but it depends on attr->count which is non reset to 0, so interval does not work as expected. 3. Trying to model this with existing settings requires manipulations of all of probability, interval, times, space, task-filter and unexposed count and per-task make-it-fail files. 4. Existing settings are per-failure-type, and the set of failure types is potentially expanding. 5. make-it-fail can't be changed by unprivileged user and aggressive stress testing better be done from an unprivileged user. Similarly, this would require opening the debugfs files to the unprivileged user, as he would need to reopen at least times file (not possible to pre-open before dropping privs). The proposed interface solves all of the above (see the example). We want to integrate this into syzkaller fuzzer. A prototype has found 10 bugs in kernel in first day of usage: https://groups.google.com/forum/#!searchin/syzkaller/%22FAULT_INJECTION%22%7Csort:relevance I've made the current interface work with all types of our sandboxes. For setuid the secret sauce was prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) to make /proc entries non-root owned. So I am fine with the current version of the code. [akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/20170328130128.101773-1-dvyukov@google.com Signed-off-by: Dmitry Vyukov <dvyukov@google.com> Cc: Akinobu Mita <akinobu.mita@gmail.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-13 05:34:35 +08:00
sprintf(buf, "%d", i);
write(fail_nth, buf, strlen(buf));
res = socketpair(AF_LOCAL, SOCK_STREAM, 0, fds);
err = errno;
pread(fail_nth, buf, sizeof(buf), 0);
fault-inject: support systematic fault injection Add /proc/self/task/<current-tid>/fail-nth file that allows failing 0-th, 1-st, 2-nd and so on calls systematically. Excerpt from the added documentation: "Write to this file of integer N makes N-th call in the current task fail (N is 0-based). Read from this file returns a single char 'Y' or 'N' that says if the fault setup with a previous write to this file was injected or not, and disables the fault if it wasn't yet injected. Note that this file enables all types of faults (slab, futex, etc). This setting takes precedence over all other generic settings like probability, interval, times, etc. But per-capability settings (e.g. fail_futex/ignore-private) take precedence over it. This feature is intended for systematic testing of faults in a single system call. See an example below" Why add a new setting: 1. Existing settings are global rather than per-task. So parallel testing is not possible. 2. attr->interval is close but it depends on attr->count which is non reset to 0, so interval does not work as expected. 3. Trying to model this with existing settings requires manipulations of all of probability, interval, times, space, task-filter and unexposed count and per-task make-it-fail files. 4. Existing settings are per-failure-type, and the set of failure types is potentially expanding. 5. make-it-fail can't be changed by unprivileged user and aggressive stress testing better be done from an unprivileged user. Similarly, this would require opening the debugfs files to the unprivileged user, as he would need to reopen at least times file (not possible to pre-open before dropping privs). The proposed interface solves all of the above (see the example). We want to integrate this into syzkaller fuzzer. A prototype has found 10 bugs in kernel in first day of usage: https://groups.google.com/forum/#!searchin/syzkaller/%22FAULT_INJECTION%22%7Csort:relevance I've made the current interface work with all types of our sandboxes. For setuid the secret sauce was prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) to make /proc entries non-root owned. So I am fine with the current version of the code. [akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/20170328130128.101773-1-dvyukov@google.com Signed-off-by: Dmitry Vyukov <dvyukov@google.com> Cc: Akinobu Mita <akinobu.mita@gmail.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-13 05:34:35 +08:00
if (res == 0) {
close(fds[0]);
close(fds[1]);
}
printf("%d-th fault %c: res=%d/%d\n", i, atoi(buf) ? 'N' : 'Y',
res, err);
if (atoi(buf))
fault-inject: support systematic fault injection Add /proc/self/task/<current-tid>/fail-nth file that allows failing 0-th, 1-st, 2-nd and so on calls systematically. Excerpt from the added documentation: "Write to this file of integer N makes N-th call in the current task fail (N is 0-based). Read from this file returns a single char 'Y' or 'N' that says if the fault setup with a previous write to this file was injected or not, and disables the fault if it wasn't yet injected. Note that this file enables all types of faults (slab, futex, etc). This setting takes precedence over all other generic settings like probability, interval, times, etc. But per-capability settings (e.g. fail_futex/ignore-private) take precedence over it. This feature is intended for systematic testing of faults in a single system call. See an example below" Why add a new setting: 1. Existing settings are global rather than per-task. So parallel testing is not possible. 2. attr->interval is close but it depends on attr->count which is non reset to 0, so interval does not work as expected. 3. Trying to model this with existing settings requires manipulations of all of probability, interval, times, space, task-filter and unexposed count and per-task make-it-fail files. 4. Existing settings are per-failure-type, and the set of failure types is potentially expanding. 5. make-it-fail can't be changed by unprivileged user and aggressive stress testing better be done from an unprivileged user. Similarly, this would require opening the debugfs files to the unprivileged user, as he would need to reopen at least times file (not possible to pre-open before dropping privs). The proposed interface solves all of the above (see the example). We want to integrate this into syzkaller fuzzer. A prototype has found 10 bugs in kernel in first day of usage: https://groups.google.com/forum/#!searchin/syzkaller/%22FAULT_INJECTION%22%7Csort:relevance I've made the current interface work with all types of our sandboxes. For setuid the secret sauce was prctl(PR_SET_DUMPABLE, 1, 0, 0, 0) to make /proc entries non-root owned. So I am fine with the current version of the code. [akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/20170328130128.101773-1-dvyukov@google.com Signed-off-by: Dmitry Vyukov <dvyukov@google.com> Cc: Akinobu Mita <akinobu.mita@gmail.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-13 05:34:35 +08:00
break;
}
return 0;
}
An example output::
1-th fault Y: res=-1/23
2-th fault Y: res=-1/23
3-th fault Y: res=-1/12
4-th fault Y: res=-1/12
5-th fault Y: res=-1/23
6-th fault Y: res=-1/23
7-th fault Y: res=-1/23
8-th fault Y: res=-1/12
9-th fault Y: res=-1/12
10-th fault Y: res=-1/12
11-th fault Y: res=-1/12
12-th fault Y: res=-1/12
13-th fault Y: res=-1/12
14-th fault Y: res=-1/12
15-th fault Y: res=-1/12
16-th fault N: res=0/12