After compiler update, infinite side effect free loops are replaced with trap
instructions. So use -fno-finite-loop to disable this behavior.
Bug: 196162833
Test: run debuggerd_test.
Change-Id: I057263360a5df64af18c17a025fab48887d0b470
When running debuggerd from the command line, it's possible that
the signal will happen on a side thread. The original intercept
in tombstoned is set to only handle crashes from the main thread
pid, so in this case, the intercept doesn't occur. To fix this,
modify the code so that running debuggerd always sends the signal
to the main pid. In addition, modify the signal handler is entered
due to the BIONIC_SIGNAL_DEBUGGER signal, then the crashing tid is
set to the main thread pid instead of the current thread.
Add unit test to cover this case.
Bug: 194346289
Test: All unit tests pass.
Test: Verify the new unit test is getting the signal on the non-main
Test: thread and still properly handling the intercept.
Test: Modify the debuggerd code to send the signal to the non main pid
Test: and verify the dump still occurs correctly.
Change-Id: I2dd1bd11fc8ef4a6fe87f05ecc67ae349a101c82
For the new kernel 5.13 heders, there is a new TRAP_PERF value that
needs to be handled.
Test: Builds.
Change-Id: I2c6658ca94423c210db9ad6692ec69f6be69b3f5
Dumping stack in ANR can fail, but error message is only printed
to logcat. To allow easier debugging of such cases we add the
error messages in the ANR file as well.
Also factor out some duplication, inline single-call functions to
their call sites, and make some of the lambdas clearer by only having
implicit state unrelated to their primary purpose in captures but
passing as arguments things they fundamentally need to do their job
(and actually adding some duplication for time_left() which was subtle
enough to fool me into thinking that we only needed one call of
set_timeout(), which I've renamed to update_timeout()).
Bug: 191172191
Test: Manual
Change-Id: I39a50ca5b72059bfeff48b010d3be44f19eb32fa
writepid command usage to join a cgroup has been deprecated in favor
of a more flexible approach using task_profiles. This way cgroup path
is not hardcoded and cgroup changes can be easily made. Replace
writepid with task_profiles command to migrate between cgroups.
Bug: 191283136
Test: build and boot
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Change-Id: I945c634dfa7621437d8ea3981bce370d680b7371
Using thread cache will cause SIGSEGV for 32bit+kernel4.9 device.
Bug: 190579082
Bug: 189803009
Test: run cts -m CtsSeccompHostTestCases
Change-Id: I47b13d02674aadbacd8dac36d8382eed0885413c
Signed-off-by: yidong zhang <yidong.zhang@amlogic.com>
I'm guessing that the original
F crash_dump64: crash_dump.cpp:460] failed to attach to thread 1671, already traced by 0 ()
was probably a race, where there _was_ a tracer but they disappeared?
Whatever, it doesn't seem helpful to show "already traced by nobody",
and we also don't want to clobber errno in the fallthrough case
(previously just where get_tracer() failed, but now also where
get_tracer() returns "nobody").
Bug: http://b/188668580
Test: treehugger
Change-Id: I3fa3b4f7e32531d48dfbb0ef946ff351ed5d9171
All but three files are Apache-2.0 already.
Bug: http://b/191499510
Test: /google/src/files/head/depot/google3/wireless/android/busytown/ayeaye/analyzers/copyright/tools/scan_android_project.sh ~/aosp/system/core/debuggerd/ | grep -v APACHE
Change-Id: I430c3382dd160e398f02470d7053ecea39c98f41
The code in the fallback path calls pthread_key_create when using the
normal thread cache. However, this code is executed out of the linker,
which means that the call doesn't see keys created by the libc version
of pthread_key_create. As of now, simply avoid using the thread cache
to avoid this problem.
Bug: 189803009
Test: debuggerd -b on a media process on a 32 bit Android Go device
Test: and observe no crash.
Test: debuggerd unit tests pass.
Change-Id: I9ca1a55e44d3bb69d49450826d7d64d7a64145c3
(cherry picked from commit 49e5a76544)
This information clearly meets the bar for being dumped to logcat. If we
omit the info, we may confuse the user into thinking that it's not
available at all, especially if it's their first time seeing an MTE
report.
This also adds some functionality to the integration testing library to
pull logcat messages and scan them to make sure the contents are in both
places.
Bug: 187881237
Test: atest debuggerd_test # on QEMU w/ MTE.
Change-Id: Icc17ea45bda7628331cc4812eaad3bc5c949b7a7
This type of error is unlikely and attempting to detect it with MTE
is likely to produce false positive reports. Make sure that this type
of error is not detected by the allocator.
Change-Id: I90676d1a031411d6b725890311317802bc24b459
This does not currently cause any problems but it does block progress
on the referenced bugs.
Bug: 187910671
Bug: 187914588
Test: m droid
Change-Id: I977cd842101187441ddbc873eac25598295aab06
When moving to the proto-ized tombstones, the note about unreadable
elf files in a backtrace got lost. This re-adds it and adds a test
to verify that the note properly shows up.
Bug: 185428454
Test: Ran unit tests.
Change-Id: I1150cc737772e1b79fd73ec5c782caadc4629421
A change was made so that pthread_create is calling
prctl(PR_PAC_RESET_KEYS, ...) on aarch64. It's possible that other
seccomp policies might need to change to allow this.
Test: CrasherTest.seccomp_backtrace passes on aarch64.
Change-Id: I9c4d1b3dca5f19a6285bf904bb942f1f52e42bd0
Talk of "gdb" when we currently mean "gdb or lldb" and will soon mean
"lldb" is starting to confuse people. Let's use the more neutral
"debugger" in places where it really doesn't matter.
The switch from gdbclient.py to lldbclient.py is a change for another
day...
Test: treehugger
Change-Id: If39ca7e1cdf4c8bb9475f1791cdaf201fbea50e0
Bug: http://b/181927912
Clang already has -Wfree-nonheap-object but it became a default warning
with clang-r416183
Test: compile crasher.cpp
Change-Id: Ice532e9f373a628e07acd08a4fc7bfa7cf5d4e08
Proto tombstones were missing tagged fault addresses, tagged_addr_ctrl,
tags in memory dumps and Scudo and GWP-ASan error reports. Since text
tombstones now go via protos, all of these features broke when we
switched to text tombstones generated from protos by default. Fix
the features by adding support for them to the proto format,
tombstone_proto and tombstone_proto_to_text.
Bug: 135772972
Bug: 182489365
Change-Id: I3ca854546c38755b1f6410a1f6198a44d25ed1c5
Looks like we unintentionally had a breakage after aosp/1595302, where
both GWP-ASan and MTE tests started failing because the extra
information wasn't plumbed through the tombstones. MTE has end-to-end
tests but aren't run continuously, and GWP-ASan was missing the e2e
tests.
Also remove some unique wording for GWP-ASan, a UaF on the free'd
pointer is now "0 bytes into a 16-byte allocation" instead of "on a
16-byte allocation". The former is more descriptive and is more
ubiquitously used in our tooling.
This patch adds the E2E tests, but the underlying problem needs to be
fixed as well, before this patch can land.
Bug: 182489365
Test: atest debuggerd_test
Change-Id: I0fe8aba7ea443b3071724987f46b19a6525cda3c
In order to test the platform in emulators that are orders of magnitude
slower than real hardware we need to be able to avoid hitting timeouts
that prevent it from coming up properly. For this purpose introduce
a system property, ro.hw_timeout_multiplier, which may be set to
an integer value that acts as a multiplier for various timeouts on
the system.
Bug: 178231152
Change-Id: I6d7710beed0c4c5b1720e74e7abe3a586778c678
Merged-In: I6d7710beed0c4c5b1720e74e7abe3a586778c678
Application developers would like to know how long their process has
been alive for to distinguish between crashes that happen immediately
upon startup and crashes in regular operation.
Test: manual
Change-Id: Ia31eeadfcced358b478c7a7c7bb2e8a0252e30f4
