minidump_stackwalk can't handle stack overflow crash (Breakpad generating bad minidump?)

NEW
Unassigned

Status

()

Toolkit
Breakpad Integration
8 years ago
a year ago

People

(Reporter: Jesse Ruderman, Unassigned)

Tracking

(Blocks: 1 bug)

Trunk
x86
Mac OS X
Points:
---
Dependency tree / graph

Firefox Tracking Flags

(Not tracked)

Details

(Whiteboard: [fuzzblocker])

(Reporter)

Description

8 years ago
For null derefs, I get stacks:
bp-3515922c-6c00-4fc3-8a02-069082090802 (bug 503709)

For stack exhaustion crashes, I don't get stacks:
bp-8746c593-d88f-478b-a5d4-30d5e2090802 (bug 459464)
bp-af65e09b-e0ea-4b44-8ee6-e76632090802 (bug 454465)

Example error message shown as "Processor Notes":

/home/processor/stackwalk/bin/stackwalk.sh returned no header lines for reportid: 52439766; No thread was identified as the cause of the crash; No signature could be created because we do not know which thread crashed; /home/processor/stackwalk/bin/stackwalk.sh returned no frame lines for reportid: 52439766; /home/processor/stackwalk/bin/stackwalk.sh failed with return code 1 when processing dump af65e09b-e0ea-4b44-8ee6-e76632090802

Btw, it's missing not only the stack but also the CPU+OS.
This is a breakpad problem, not a Socorro problem.  The error message reported by the processor are from the invocation of Breakpad's minidump_stackwalk program.  The crash dump submission received from the client was hosed before Socorro even received it.
Jesse: what platform are these crashes on? Is it the same on all platforms, or only some? I know I've seen stack overflow minidumps that we handled properly, since Lars added code to Socorro to elide the stack if it exceeded a certain depth at one point.
(Reporter)

Comment 3

8 years ago
These are on Mac.  I haven't tested other platforms.
((In reply to comment #2)
> I've seen stack overflow minidumps that we handled properly,
> since Lars added code to Socorro to elide the stack if it exceeded a certain
> depth at one point.

Jesse, given comment 2 and bug 459464 comment 1, is this WFM? Or is there room for improvement?
(Reporter)

Comment 5

7 years ago
Can't get that one to crash quickly any more, but bug 454465 gave me bp-3837780d-1233-407d-962d-bae4e2100515.  So still broken on Mac, it seems.
(Reporter)

Comment 6

7 years ago
Local breakpad says "** Unknown exception behavior" when I trip bug 550306, and then minidump_stackwalk doesn't print anything.
(Reporter)

Comment 7

7 years ago
When I trip bug 614172:

* Mac breakpad says "unknown exception behavior", as described above, and minidump_stackwalk says there's no thread info.

* Linux breakpad seems to not create a minidump at all.

* Windows breakpad works fine (EXCEPTION_STACK_OVERFLOW + top 1024 stack frames).
Presumably whatever we're doing on Linux/Mac isn't sufficient for handling stack overflow. We are using a separate handler thread with its own stack on Linux, and a separate exception handler thread on Mac, so I would expect things to work, but clearly they don't.

Of course, trying to search for anything related to the problem is nigh impossible, since stackoverflow.com has completely blown out Google results for "stack overflow".

Comment 9

6 years ago
I'm getting null signature crash reports pretty frequently on this one crash that happens in low memory situations:

bp-cacb678c-bdba-4b74-85f9-dff322110509
bp-2702e346-d22c-4893-8003-55e9e2110509
bp-586ff91d-6097-4e75-b173-d3fb62110509
bp-9f5fd6fa-7b9a-4be8-9ba8-a2c4a2110509
bp-eb2a921b-cc05-45e9-a0a1-d2f022110509
bp-cb33ad73-221f-4fca-a340-b9c5c2110509
bp-717ecbdf-5adc-4ab3-9548-e9d312110509

I'm on Windows so maybe this is a different problem?
Yeah, that's a different problem. We have known problems with writing dumps in OOM. It's probably bug 527095.
Unsurprisingly, this is still broken on Mac, as bug 808107 shows.
(Reporter)

Updated

5 years ago
Whiteboard: [fuzzblocker]
(Reporter)

Updated

4 years ago
Summary: Socorro can't handle stack overflow crash (Breakpad generating bad minidump?) → minidump_stackwalk can't handle stack overflow crash (Breakpad generating bad minidump?)
(Reporter)

Updated

4 years ago
Depends on: 527095
Blocks: 808107
You need to log in before you can comment on or make changes to this bug.