Chrome started hitting some crashes in v8 jitted code which happens to be
non ABI compliant and debuggers (including WinDBG) are unable to produce
meaningful stack traces.
The Breakpad stack walker has some builtin heuristics to deal with such cases.
More specifically, when unable to find a good parent frame, it scans the raw
stack to find a suitable parent frame. The max scan size was set at 30
pointers which was (apparently) not enough to recover in this case.
I'm increasing it to 40 pointers. I confirmed that at 34 pointers it was able
to recover however I'm setting it to 40 in order to it some slack.
I needed to update two unittests which were expecting the previous scan limit.
BUG=
R=mark@chromium.org
Review URL: https://codereview.chromium.org/1379433005 .
.raSearchStart in the cases where there are alignment operators in
the program string.
If alignment operators are found in the program string, the current
value of %ebp must be valid and it is the only reliable data point
that can be used for getting to the previous frame. Previously, the
.raSearchStart calculation was based on %esp and when %esp is aligned
in the current frame (which is a lossy operation) the resulting
.raSearchStart cannot was incorrect. There is code that is trying to
work around this problem (scanning of up to 3 words for a return
address) which is unreliable and it doesn't work in many cases (e.g.
when the alignment is on a 64-byte boundary).
This fix is already deployed in Google and it was measured to reduce
the number of wrong stack traces (for Windows crashes) by 45%. No
regressions have been found so far.
Here is an example of an issue that was fixed by this change (where
register %esp is aligned on the 64-byte boundary and the workarounds
that we already had didn't work):
https://code.google.com/p/chromium/issues/detail?id=311359
0:013> uf chrome_59630000!base::MessagePumpForIO::DoRunLoop
518 59685c39 55 push ebp
518 59685c3a 8bec mov ebp,esp
518 59685c3c 83e4c0 and esp,0FFFFFFC0h <== 64-byte boundary
518 59685c3f 83ec34 sub esp,34h
518 59685c42 53 push ebx
518 59685c43 56 push esi
Program string contains 64-byte alignment:
$T1 .raSearch = $T0 $T1 4 - 64 @ = $ebp $T1 4 - ^ = $eip $T1 ^ =
$esp $T1 4 + = $20 $T0 56 - ^ = $23 $T0 60 - ^ = $24 $T0 64 - ^ =
Review URL: https://breakpad.appspot.com/694002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1232 4c0a9323-5329-0410-9bdc-e9ce6186880e
More specifically:
- Detect corrupt symbols during minidump processing and provide the list of modules with corrupt symbols in the ProcessState. This will allow listing the corrupt symbol files in the final crash report.
- Skip and recover from symbol data parse errors - don't give up until 100 parse errors are seen.
- In order to recover from '\0' (null terminator) in the middle of a symbol file, a couple of methods have to be updated to require both buffer pointer and length. Previously they required only a buffer pointer (char *) and the size of the buffer was evaluated using strlen which is not reliable when the data is corrupt. Most of the changes are due to these signature updates.
- Added and updated unittests.
Also, updated minidump_stackwalk to show a WARNING for corrupt symbols. Output looks like this:
...
Loaded modules:
0x000da000 - 0x000dafff Google Chrome Canary ??? (main)
0x000e0000 - 0x0417dfff Google Chrome Framework 0.1500.0.3 (WARNING: Corrupt symbols, Google Chrome Framework, 4682A6B4136436C4BFECEB62D498020E0)
0x044a8000 - 0x04571fff IOBluetooth 0.1.0.0
...
Review URL: https://breakpad.appspot.com/613002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1200 4c0a9323-5329-0410-9bdc-e9ce6186880e
This is achieved by:
1. Extending the span of the scan for return address in the conext frame. Initially, I wanted to extend the span of the scan for all frames but then I noticed that there is code for ARM already that is extending the search only for the context frame. This kind of makes sense so I decided to reuse the same idea everywhere.
2. Attempting to restore the EBP chain after a successful scan for return address so that the stackwalker can switch back to FRAME_TRUST_CFI for the rest of the frames when possible.
I also fixed the lint errors in the files touched.
Review URL: https://breakpad.appspot.com/605002
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@1193 4c0a9323-5329-0410-9bdc-e9ce6186880e
do with FPO (frame-pointer-omission) optimized context
frames where the context frame represents a Windows
System call stub.
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@971 4c0a9323-5329-0410-9bdc-e9ce6186880e
The google_breakpad::TestAssembler classes are used in both the processor's
and the Linux dumper's test suites, and will soon be used in the Mac
dumper's tests as well. This patch moves their source files from
src/processor to src/common.
a=jimblandy, r=thestig
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@574 4c0a9323-5329-0410-9bdc-e9ce6186880e
Add a CFIFrameInfo class (named for symmetry with WindowsFrameInfo) to
represent the set of STACK CFI rules in effect at a given instruction,
and apply them to a set of register values. Provide a SimpleCFIWalker
class template, to allow the essential CFI code to be shared amongst
the different architectures.
Teach BasicSourceLineResolver to partially parse 'STACK CFI' records,
and produce the set of rules in effect at a given instruction on
demand, by combining the initial rule set and the appropriate rule
deltas in a CFIFrameInfo object.
Adapt StackwalkerX86 and StackFrameX86 to retrieve, store, and apply
CFI stack walking information.
Add validity flags for all the general-purpose registers to
StackFrameX86::ContextValidity.
a=jimblandy, r=mmentovai
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@549 4c0a9323-5329-0410-9bdc-e9ce6186880e
Issue 53001 (http://breakpad.appspot.com/53001) defines the
TestAssembler classes; those, along with a new set of mock classes
defined in stackwalker_unittest_utils.h, make it possible for us to
actually do proper unit testing of a stack walker. These tests get us
full code coverage for stackwalker_x86.cc.
a=jimblandy, r=mmentovai
git-svn-id: http://google-breakpad.googlecode.com/svn/trunk@548 4c0a9323-5329-0410-9bdc-e9ce6186880e
