Microdumps are a very lightweight variant of minidumps. They are meant
to dump a minimal crash report on the system log (logcat on Android),
containing only the state of the crashing thread.
This is to deal with cases where the user has opted out from crash
uploading but we still want to generate meaningful information on the
device to pull a stacktrace for development purposes.
Conversely to conventional stack traces (e.g. the one generated by
Android's debuggerd or Chromium's base::stacktrace) microdumps do NOT
require unwind tables to be present in the target binary. This allows
to save precious binary size (~1.5 MB for Chrome on Arm, ~10 MB on
arm64).
More information and design doc on crbug.com/410294
BUG=chromium:410294
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1398 4c0a9323-5329-0410-9bdc-e9ce6186880e
This change introduces the necessary glue typedefs to deal with the
mismatch introduced by the latest Android NDK (w.r.t. desktop Linux):
- [x86_64] Rename fpregs mxcr_mask -> .mxcsr_mask
- [mips] uc_mcontext.fpregs.fp_r.fp_dregs -> uc_mcontext.fpreg
- [mips] restore the forked user.h
Submitting this on behalf of fdegans@chromium.orgTBR=mark@chromium.org
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1397 4c0a9323-5329-0410-9bdc-e9ce6186880e
This is a refactoring of the logic which determines the
module name and path for a given MappingInfo in minidump_writer.cc.
Such logic, which will be soon shared also with the upcoming
microdump_writer.cc, is simply being moved to linux_dumper.cc,
extracting a GetMappingEffectiveNameAndPath method.
No behavioral change is intended.
BUG=chromium:410294
R=thestig@chromium.org
Review URL: https://breakpad.appspot.com/7734002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1392 4c0a9323-5329-0410-9bdc-e9ce6186880e
This change is a pure refactoring of the common bits of minidump_writer.cc
that will be shared soon with the upcoming microdump_writer.cc.
In particular, this CL is extracting the following classes:
- ThreadInfo: handles the state of the threads in the crashing process.
- RawContextCPU: typedef for arch-specific CPU context structure.
- UContextReader: Fills out a dump RawContextCPU structure from the
ucontext struct provided by the kernel (arch-dependent).
- SeccompUnwinder: cleans out the stack frames of the Seccomp sandbox
on the supported architectures.
- MappingInfo: handles information about mappings
BUG=chromium:410294
R=mmandlis@chromium.org
Review URL: https://breakpad.appspot.com/4684002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1388 4c0a9323-5329-0410-9bdc-e9ce6186880e
When parsing /proc/pid/maps, current code merges adjacent entries that
refer to the same library and where the start of the second is equal to
the end of the first, for example:
40022000-40025000 r-xp 00000000 b3:11 827 /system/lib/liblog.so
40025000-40026000 r--p 00002000 b3:11 827 /system/lib/liblog.so
40026000-40027000 rw-p 00003000 b3:11 827 /system/lib/liblog.so
When the system linker loads a library it first reserves all the address
space required, from the smallest start to the largest end address, using
an anonymous mapping, and then maps loaded segments inside that reservation.
If the loaded segments do not fully occupy the reservation this leaves
gaps, and these gaps prevent merges that should occur from occurring:
40417000-4044a000 r-xp 00000000 b3:11 820 /system/lib/libjpeg.so
> 4044a000-4044b000 ---p 00000000 00:00 0
4044b000-4044c000 r--p 00033000 b3:11 820 /system/lib/libjpeg.so
4044c000-4044d000 rw-p 00034000 b3:11 820 /system/lib/libjpeg.so
Where the segments that follow this gap do not contain executable code
the failure to merge does not affect breakpad operation. However, where
they do then the merge needs to occur. Packing relocations in a large
library splits the executable segment into two, resulting in:
73b0c000-73b21000 r-xp 00000000 b3:19 786460 /data/.../libchrome.2160.0.so
> 73b21000-73d12000 ---p 00000000 00:00 0
73d12000-75a90000 r-xp 00014000 b3:19 786460 /data/.../libchrome.2160.0.so
75a90000-75c0d000 rw-p 01d91000 b3:19 786460 /data/.../libchrome.2160.0.so
Here the mapping at 73d12000-75a90000 must be merged into 73b0c000-73b21000
so that breakpad correctly calculates the base address for text.
This change enables the full merge by also merging anonymous maps which
result from unused reservation, identified as '---p' with offset 0, and
which follow on from an executable mapping, into that executable mapping.
BUG=chromium:394703
R=rmcilroy@chromium.org, thestig@chromium.org
Review URL: https://breakpad.appspot.com/7714003
Patch from Simon Baldwin <simonb@chromium.org>.
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1380 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch makes two changes to breakpad to enable crash reporting to work correctly when the library is inside another file (an archive):
- Do not filter mappings which map an executable at a non-zero offset.
- If such an executable is mapped look in the ELF information for the
shared object name and use that name in the minidump.
Note this change doesn't care about the archive format and isn't Android
specific (though loading the shared library this way is currently only done on Android).
BUG=390618
R=thestig@chromium.org
Review URL: https://breakpad.appspot.com/7684002
Patch from Anton Carver <anton@chromium.org>.
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1355 4c0a9323-5329-0410-9bdc-e9ce6186880e
__system_property_get has been removed from the Android NDK for 'L' so Breakpad
no longer links. This CL just removes the call in WriteOSInformation because
Chrome already passes the build fingerprint as a crash key called "android_build_fp" in the crash report HTTP POST message.
BUG=394841,393903
R=mark@chromium.org
Review URL: https://breakpad.appspot.com/1694002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1351 4c0a9323-5329-0410-9bdc-e9ce6186880e
This bug manifests itself on Android x64 (in ExceptionHandlerTest::ExternalDumper), but is possible on any platform.
The compiler is unaware of the code which runs in the signal handler which reads this variable. Without volatile the compiler is free to optimise away writes to this variable which it believes are never read.
BUG=381142, 346626
R=thestig@chromium.org
Review URL: https://breakpad.appspot.com/8664002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1343 4c0a9323-5329-0410-9bdc-e9ce6186880e
This is an initial attempt to add Arm64 (aarch64) support to Breakpad for
Linux / Android platforms. This CL adds the Arm64 data structures, but does
not yet implement the Android getcontext support or CPUFillFromThreadInfo /
CPUFillFromUContext.
BUG=354405,335641
R=mark@chromium.org
Review URL: https://breakpad.appspot.com/1354002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1301 4c0a9323-5329-0410-9bdc-e9ce6186880e
The current CreateChildCrash logic is racy when it comes to creating a
crash dump for two reasons:
The main thread that calls kill() on a different thread is guaranteed
the signal will be *queued* when it returns, but not *delivered*. If
the kernel doesn't automatically schedule the receiving thread, but
instead lets the main thread run to the exit() call, then the signal
never triggers a coredump and the whole process simply exits.
The main thread is using kill() to try to deliver a signal to a
specific thread, but that function is for sending signals to a
process. That means the kernel is free to deliver the signal to
any thread in the process and not just the one requested. This
manifests itself as the pr_pid in the coredump not being the one
expected. Instead, we must use tkill() with the tid (which we
already took care of gathering) to deliver to a specific thread.
These are a lot easier to see on a UMP system as contention is heavier.
BUG=chromium:207918
TEST=`dumper_unittest` still passes, and doesn't flake out in a UMP system
TEST=`linux_client_unittest` still passes
R=benchan@chromium.org
Review URL: https://breakpad.appspot.com/1304005
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1299 4c0a9323-5329-0410-9bdc-e9ce6186880e
There are a bunch of tests that use invalid memory acesses (on purpose)
to trigger a crash so that we can detect things are dumped correctly.
When we run under ASAN, it catches those accesses and the breaks the
testing flow.
For now, use the existing ADDRESS_SANITIZER symbol to disable more tests.
Ideally we'd use a compile-time attribute to disable ASAN on a few funcs,
but that seems to be broken atm.
BUG=chromium:293519
BUG=chromium:304575
TEST=ran unittests under ASAN and they now pass
TEST=ran unittests w/out asan/clang and they still pass
R=benchan@chromium.org
Review URL: https://breakpad.appspot.com/884002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1255 4c0a9323-5329-0410-9bdc-e9ce6186880e
For CPUs that don't support the MMX instruction set, such pre-Pentium III or industrial x86 embedded PCs, the minidump fails when it tries to retrieve MMX specific registers.
This patch adds MMX detection for that call.
Tested on Ubuntu 12.04 with i686, and on a custom Linux distro on a Vortex86DX microcontroller.
Original review: https://breakpad.appspot.com/455002/
A=aras.vaichas
BUG=495
Review URL: https://breakpad.appspot.com/864002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1248 4c0a9323-5329-0410-9bdc-e9ce6186880e
It is incorrect to wrap close in HANDLE_EINTR on Linux.
Unnecessary #includes of eintr_wrapper.h are also removed. The variable naming
within the macro is also updated per Chromium r178174.
einter_wrapper.h contains a non-mechanical change. Mechanical changes were
generated by running:
sed -E -i '' \
-e 's/((=|if|return|CHECK|EXPECT|ASSERT).*)HANDLE(_EINTR\(.*close)/\1IGNORE\3/' \
-e 's/(ignore_result|void ?)\(HANDLE_EINTR\((.*close\(.*)\)\)/\2/' \
-e 's/(\(void\) ?)?HANDLE_EINTR\((.*close\(.*)\)/\2/' \
$(grep -rl HANDLE_EINTR.*close . --exclude-dir=.svn)
sed -E -i '' -e '/#include.*eintr_wrapper\.h"/d' \
$(grep -EL '(HANDLE|IGNORE)_EINTR' \
$(grep -Elr '#include.*eintr_wrapper\.h"' . --exclude-dir=.svn))
BUG=chromium:269623
R=ted.mielczarek@gmail.com
Review URL: https://breakpad.appspot.com/784002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1239 4c0a9323-5329-0410-9bdc-e9ce6186880e
SIGABRT can be generated internally, usually by calling abort(),
or externally by another process. When the signal is generated
by the kernel, info->si_pid is 0 and the signal is treated in the
same way as an exception (SIGSEGV, etc.), but the assumption
that the exception happens again upon return from the handler
is wrong, so we must have a special case for this.
Original CL: https://breakpad.appspot.com/734002/
BUG=chromium:303075
TEST=tested with Alt-VolumeUp-X on Chrome OS
A=semenzato@chromium.orgR=semenzato@google.com
Review URL: https://breakpad.appspot.com/754002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1233 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch improves several things for Linux/ARM:
- Better detection of the number of CPUs on the target
device. The content of /proc/cpuinfo only matches the
number of "online" CPUs, which varies over time with
recent Android devices.
- Reconstruct the CPUID and ELF hwcaps values from
/proc/cpuinfo, this is useful to better identify
target devices in minidumps.
- Make minidump_dump display the new information
in useful ways.
- Write a small helper class to parse /proc/cpuinfo
and also use it for x86/64.
- Write a small helper class to parse sysfds cpu lists.
- Add a my_memchr() implementation.
- Add unit tests.
Tested on a Nexus S (1 CPU), Galaxy Nexus (2 CPUs)
and a Nexus 4 (4 CPUs).
Review URL: https://breakpad.appspot.com/540003
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1160 4c0a9323-5329-0410-9bdc-e9ce6186880e
Three unit tests were failing on recent ARM devices (e.g. Galaxy Nexus
or Nexus 4), while ran properly on older ones (e.g. Nexus S).
The main issue is that the instruction cache needs to be explicitely
cleared on ARM after writing machine code bytes to a malloc()-ed
page with PROT_EXEC.
Review URL: https://breakpad.appspot.com/540002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1132 4c0a9323-5329-0410-9bdc-e9ce6186880e
If the stack sizes for threads in the MinidumpSizeLimit test are too big,
then subtracting 64KB from the normal minidump file size is not enough to
trigger the size-limiting logic. Instead of basing the arbitrary limit off
of the normal file size, make it relative to the 8KB stack size the logic
assumes.
BUG=google-breakpad:510
TEST=Ran unittests
Review URL: https://breakpad.appspot.com/504002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1090 4c0a9323-5329-0410-9bdc-e9ce6186880e
When there are upwards of 200 threads in a crashing process, each having an
8KB stack, this can result in a huge, 1.8MB minidump file. So I added a
parameter that, if set, can compel the minidump writer to dump less stack.
More specifically, if the writer expects to go over the limit (due to the
number of threads), then it will dump less of a thread's stack after the
first 20 threads.
There are two ways to specify the limit, depending on how you write minidumps:
1) If you call WriteMinidump() directly, there's now a version of the
function that takes the minidump size limit as an argument.
2) If you use the ExceptionHandler class, the MinidumpDescriptor object you
pass to it now has a set_size_limit() method you would call before
passing it to the constructor.
BUG=chromium-os:31447, chromium:154546
TEST=Wrote a size-limit unittest; Ran unittests
Review URL: https://breakpad.appspot.com/487002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1082 4c0a9323-5329-0410-9bdc-e9ce6186880e
Breakpad can be used on processes where a mistaken
library saves then restores one of our signal handlers
with 'signal' instead of 'sigaction'.
This loses the SA_SIGINFO flag associated with the
Breakpad handler, and in some cases (e.g. Android/ARM
kernels), the values of the 'info' and 'uc' parameters
that ExceptionHandler::SignalHandler() receives will
be completely bogus, leading to a crash when the function
is executed (and of course, no minidump generation).
To work-around this, have SignalHandler() check the state
of the flag. If it is incorrectly unset, re-register with
'sigaction' and the correct flag, then return. The signal
will be re-thrown, and this time the function will be
called with the correct values.
Review URL: https://breakpad.appspot.com/481002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1067 4c0a9323-5329-0410-9bdc-e9ce6186880e
- One of the unit test binaries refused to link due to
missing linker flags.
- The WriteDSODebug() function now works on Android, so
do not special-case it anymore.
- Ensure android/run-checks.sh will complain properly if
the client unit test suite fails on Android. It used to
consider that such failures were acceptable. Note that
it still considers failures when running the tools and
processor test suite on the device normal (fixing this
is a lot harder, and these parts of Breakpad typically
never run on a device, but on the host).
Review URL: https://breakpad.appspot.com/482002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1066 4c0a9323-5329-0410-9bdc-e9ce6186880e
Currently, if a thread's stack pointer is not within a valid memory page,
the minidump writing will fail with an error. This change allows an invalid
stack pointer by simply setting the memory size to zero in the minidump.
The processing code already checks for the size being zero, although it
currently just gives an error (see https://breakpad.appspot.com/413002/).
BUG=google-breakpad:499, chromium-os:34880
TEST=make check, manually ran minidump-2-core and core2md
Review URL: https://breakpad.appspot.com/478002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1065 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch updates README.ANDROID, since the client library
now supports directly generating a minidump without raising
a signal, and that it builds and runs fine on Android/x86.
+ Formatting fixes in minidump_writer_unittest.cc
Review URL: https://breakpad.appspot.com/446002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1026 4c0a9323-5329-0410-9bdc-e9ce6186880e
If none of the installed ExceptionHandlers handle a signal (their
FilterCallbacks or HandlerCallbacks all return false), then the signal
should be delivered to the signal handlers that were previously
installed.
This requires that old_handlers_ become a static vector so that we can
restore the handlers in the static HandleSignal.
Currently it is also restoring signals in ~ExceptionHandler (if there
are no others). This should not be required since our documentation
states that a process can only have one ExceptionHandler for which
install_handlers is true (and so we get the correct behavior if we
simply leave our handlers installed forever), but even the tests
themselves violate that.
Patch by Chris Hopman <cjhopman@chromium.org>
Review URL: https://breakpad.appspot.com/440002/
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1025 4c0a9323-5329-0410-9bdc-e9ce6186880e
This adds a minimalistic implementation of getcontext()
for Android/ARM and Android/x86. The provided code is
in assembly and only implements the bare minimum required
by Breakpad to get the current processor state.
Note that:
- The FPU state is not saved to the ucontext_t on ARM.
(that's actually the main difference with a normal
getcontext() implementation).
This is normal. On Linux/ARM, such state must be
obtained with PTRACE_GETVFPREGS instead. This will
be implemented in a future patch.
- On x86, only the 'regular' FPU state is saved, to
mimic the GLibc/i386 implementation. The state of
SSE/SSE2/etc registers is not part of the upstream
getcontext() implementation.
Review URL: https://breakpad.appspot.com/444002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1024 4c0a9323-5329-0410-9bdc-e9ce6186880e
This small patch allows the build of the tools and processor
when targetting Android with the Automake/Autconf build.
Not that these necessarily work correctly at the moment,
but there is no need for --disable-tools --disable-processor
now when using --host=arm-linux-androideabi or
--host=i686-linux-android.
+ Modify android/run-checks.sh to build all binaries with
the Automake build.
+ Tiny fix for --abi=x86 in android/run-checks.sh
Review URL: https://breakpad.appspot.com/438002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1018 4c0a9323-5329-0410-9bdc-e9ce6186880e
This patch remove many Android-specific #ifdefs from the Breakpad
source code. This is achieved by providing "fixed-up" platform
headers (e.g. <signal.h> or <sys/user,h>), in the new directory
src/common/android/include/, which masks differences between
the NDK and GLibc headers.
The old "android_link.h" and "android_ucontext.h" are moved
and renamed.
This also requires putting this directory as the first
include path during Android-hosted builds, hence the
modification of Makefile.am and configure.ac
Review URL: https://breakpad.appspot.com/434002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1017 4c0a9323-5329-0410-9bdc-e9ce6186880e
where the minidump should be created, without the need of opening any other
file.
BUG=None
TEST=Run unit-tests.
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1007 4c0a9323-5329-0410-9bdc-e9ce6186880e
Ted Mielczarek:
> You could try backing out r989, although Mozilla has been running with that
> patch for months without issue.
Me:
> src/client/windows/handler/exception_handler.cc in r989 appears to have
> formatting problems, an unwanted property change, and no real Breakpad review
> history, so maybe we should back it out anyway until the proper process is
> followed.
NACL Tests nacl_integration failures:
http://build.chromium.org/p/chromium/builders/NACL%20Tests/builds/30138
chrome src/native_client/tests/inbrowser_crash_test/crash_dump_tester.py says
that the observed failures are a symptom of crash_service.exe itself crashing.
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@998 4c0a9323-5329-0410-9bdc-e9ce6186880e
