The only way to get a bad architecture value in the protobuf is if
the data was corrupted or an unsupported architecture was added without
the register support.
If the protobuf is corrupted, this is strictly better since it
still produces a tombstone with the data present.
If there is an unsupported architecture, it will still result in a tombstone,
only the registers would not be present. It would also be very obviously
a problem that needs to be fixed. Again, this is strictly better since
the crash in generation is not necessarily visible unless you look at
the log. Here, the data is in the log and in the tombstone.
This also removes the only dependency in this file on the async_safe
library.
Test: Ran unit tests.
Test: Forced an invalid architecture and verified tombstone is present
Test: with error message, and error message printed in the log.
Change-Id: I8e4a2e3f778fafb5b7241c2f23d5f867f1341ed8
Timeouts in tombstoned.cpp and intercept_manager.cpp are scaled
by HwTimeoutMultiplier, but the timeouts in debuggerd_test.cpp
are not, which means the CrasherTest#intercept_timeout test will
fail for any platform that has a high enough HwTimeoutMultiplier.
Bug: 309532789
Test: debuggerd_test.CrasherTest#intercept_timeout
Change-Id: I83cd01e87644c011efa155a32fd5d92cc8a43a95
The new 6.6 kernel headers added a new segv type, SEGV_CPERR. Add this
to the switch statement.
Test: Unit tests pass.
Change-Id: I77eb4748e51c7e7d7291bfd2180b0ccb3b5a6ded
While doing this, refactor the intercept code to be easier to understand.
The primary use case for this is to perform a parallel stack dump (both Java and native) for specific ANRs.
Add tests for all of the different intercept conditions.
Modify the tests to display the error message from the intercept
response if there is an error.
Bug: 254634348
Test: All unit tests pass.
Test: Ran debuggerd on native and java processes.
Test: Created a bugreport without error.
Change-Id: Ic531ccee05b9a470748b815cf109e0076150a0b6
A clang update enabled -Wreorder-init-list by default. Since it doesn't
provide any benefit to the debuggerd code, disable the warning.
Test: Builds without warnings.
Change-Id: I75cfe064ba92c74312ba33f329b1364258eba06c
aosp/2734054 added socket timeouts for nonblocking liblog ops.
seccomp policy was not updated so tests failed when unallowed
socksetopt syscall was made.
Bug: 298420226
Test: atest debuggerd_test
Change-Id: Iace232ec8b94e5d316d344abc5d866fe314607e0
Signed-off-by: Andrei Diea <adiea@google.com>
Check for the log opening failing.
Add the ability to put error messages in the log and tombstone so
that it's clear if the log reading failed in some way.
Adjust test so that if there is a log or if no log exists, the test
will still pass.
Print an <unknown> if the command line is unreadable instead of nothing.
Test: Ran unit tests.
Test: Induced error and verified error message is save in tombstone.
Change-Id: I2fce8078573b40b9fed3cd453235f3824cadb5e3
Commit aosp/1259140 moved fdsan_table into debugger_process_info, which
is populated conditionally. This introduced a bug where the process that
receives BIONIC_SIGNAL_DEBUGGER (35) does not propagate the fdsan_table
pointer to crash_dump:
$ adb shell kill -SIG35 <pid>
$ adb logcat -s DEBUG
E DEBUG : failed to read fdsan table entry 0: I/O error
Fdsan in warn-only mode uses BIONIC_SIGNAL_DEBUGGER[1], so the generated
tombstones don't have any fd ownership info.
Fix it by calling get_process_info() irrespective of the signal being
handled, taking care to preserve the previous behavior of not showing
abort messages set by applications in non-fatal dumps.
Test: debuggerd_test
Test: send SIG35 to arbitrary process and inspect the log and tombstone
Test: crasher fdsan_file
[1] 20ad9129e7/libc/bionic/fdsan.cpp (166)
Change-Id: I76931ca4825e846fc99f26fa590c045130abb850
Also add the missing `.size` directives to all the assembler functions
for slightly improved backtraces.
Test: crasher64 pac; crasher64 bti
Change-Id: I8e0c127cbff56c33637e6ca8f1d927b971951807
When using the bootstrap linker, the get_gwp_asan_callbacks is
not set. Therefore, check it is not nullptr before calling it
during crash processing.
Bug: 284098779
Test: Ran crasher64 using /system/bin/bootstrap/linker64 and verify
Test: debuggerd code does not crash.
Test: All unit tests pass.
Change-Id: Ifc710fe4bef24661700444a1b69432bfc29d580f
The generate.sh script can generate the file, but current policy file does not match it.
And the rules are not appropriate, like missing "sysinfo", causing the
debuggerd_test to fail in system model. So we match the policy to
what it should be.
Test: make debuggerd_test
Change-Id: I57ebd7713f2ab939d01bfefcc7935e234fdd3e13
Signed-off-by: liwentao <liwentao@eswincomputing.com>
Some testing environments can have a test that is sending many
thousands of messages to the log. When this type of process crashes
all of these log messages are captured and can cause OOM errors
while creating the tombstone.
Added a test to verify the log messages are truncated. Leaving this
test disabled for now since it is inherently flaky due to having to
assume that 500 messages are in the log.
Added a test for a newline in a log message since it's somewhat
related to this change.
NOTE: The total number of messages is capped at 500, but if a message
contains multiple newlines, the total messages will exceed 500.
Counting messages this way seems to be in the spirit of the cap,
that a process logging a large message with multiple newlines does
not completely fill the tombstone log data.
Bug: 269182937
Bug: 282661754
Test: All unit tests pass.
Test: The disabled max_log_messages test passes.
Change-Id: If18e62b29f899c2c4670101b402e37762bffbec6
Just noticed some opportunities while skimming.
Test: adb shell debuggerd $(adb shell pidof com.android.systemui)
Test: All unit tests pass (both 32 bit and 64 bit).
Test: Ran unit tests in a loop hundreds of times.
Change-Id: I428d0cf599ed603a21944b084b95594db893cbd5
Make sure that all the threads have started up, otherwise the main part
of the test might not be testing as stressful a situation as expected.
Note that the "race" moniker is still valid because of the debuggerd
timeout.
The test is now faster (405ms) when run under good conditions.
Test: atest 'debuggerd_test:debuggerd_client#race'
Test: Ran debuggerd_client.race 1000 times on its own.
Test: Ran the whole suite of debuggerd unit tests 1000 times.
Change-Id: I487e7654a71df9f1799f09c6f385c929ddf2f234
Also add new unit tests to verify this behavior.
Bug: 276934420
Test: New unit tests pass.
Test: Ran new unit tests without pthread_setname_np call and verified
Test: the tests fail.
Test: Force crash logd and verify log messages are not gathered.
Test: Force crash a logd thread and verify log messages are not gathered.
Change-Id: If8effef68f629432923cdc89e57d28ef5b8b4ce2
liblog can drop data when debuggerd is overloaded, which leads to
truncated tombstones. by adding the count separately, automation can
easily see whether it is dealing with a truncated tombstone or not.
Bug: 269537146
Change-Id: Ia991537efc0d6b57cbff23ee45af6521467aa20d
This change adds a missing pointer dereference to the InterceptResponse
when checking for a size mismatch.
Test: build
Bug: N/A
Change-Id: I88afed6f1c0f33fe237d337b0fb8fc0a0c0e3bac
This adds the missing assembler for riscv64, even though I don't have a
working tombstoned yet to test it with. There's a distinct possibility
we'll be back to fix the test (because although "register 1" is harmless
for the other architectures, it's the ra register on riscv64; the default
link register), but at least this lets us build the test.
I've also simplified all the assembly to be the simplest sequence I
know that writes 0 to address 0 (because if there was a reason to use
so many instructions before, I want to know what it is so I can write
the missing comment!).
Test: treehugger
Change-Id: I10d117eaedf361d9759a450e0973d07c4f97090e
r29 is the stack pointer on mips, but it's x2 on riscv64 (and the git
history shows that this was indeed copy & pasted from the mips code)
and since bionic always sets up a signal stack with sigaltstack() I
doubt the comment was relevant even on mips (but no-one ever used it,
so who'd know?).
While I'm here, stop using decimal arithmetic --- the whole point was to
have each register contain the value that was obviously appropriate for
that register. (riscv64's mips-like mess of registers all over the place
means that's not going to be super readable, but there's no reason to
make it worse.)
Also, even though I personally prefer the 0xdead from the old mips code,
everyone else is using 0xa5a5, so let's make riscv64 match the others.
Test: treehugger
Change-Id: Ibbae821bc0a02e07164147d621e342224528c2c9
This changes the crash export mechanism in Microdroid. For this, we
create module tombstone_handler which exports methods very similar to
tombstoned.h
For Microdroid (detected using prop: ro.hardware): It calls newly
introduces microdroid specific methods to connect/notify completion of
crash.
Individual methods:
connect -> For Android, it would connect to
tombstoned which would send it the fd corresponding to newly created
file on /data/tombstone_ . For Microdroid, we connect to tombstone
server on host via vsock & populate these sockets as the output fd.
crash_dump, in the later case, would directly write on the socket(s).
notify_completion: For Microdroid, it would simply shutdown the vsock
connections.
Note when OS is not Microdroid: It calls corresponding methods of
tombstoned_client, essentially serving as a proxy.
Detailed design: go/vm-tombstone
Test: atest MicrodroidHostTests#testTombstonesAreGeneratedUponUserspaceCrash
Bug: 243494912
Change-Id: I68537b967f2ee48c1647f0f923aa79e8bcc66942
When this was translated to riscv64, someone "fixed" the crashing bugs
that were the whole point of these two functions. Fix them back so they
actually crash, and add the CFI directives.
Test: treehugger
Change-Id: I312c51fa4c893d27b0f4e39383521657a5870a0d
See aosp/2485756 for more details, but this patch introduces a new
protocol between debuggerd and ActivityManager. This new protocol allows
ActivityManager to correctly handle recoverable crashes.
Bug: 270720163
Test: atest CtsGwpAsanTestCases
Change-Id: Icac6262d608dd57a5daf51699064ab28b0c4703f
They're the same script right now, but gdbclient.py is a bit misleading,
even if we're not likely to ever actually remove it.
Test: treehugger
Change-Id: Ic514f98bf13b3e699be4dbad2bafef22d41d9ffd
1. Fixes this test under clang coverage, which is run under presubmit
for TEST_MAPPING files. When we spawn under a minijail, and the
process exited normally (which is the case for recoverable), clang
coverage would use atexit handlers to dump some stuff using banned
prctl's and other syscalls. Instead of allow-listing them all which
sounds like a huge pain, call _exit() which skips those handlers.
2. Extends the invariant testing to make sure that recoverable GWP-ASan
recovers both the first time, and a second time in a different slot.
Bug: N/A
Test: CLANG_COVERAGE=true NATIVE_COVERAGE_PATHS="*" atest debuggerd_test
Change-Id: I6059e21db4c2898b1c9777a00d2a54497d80ef79
