Open Bug 1424393 Opened 7 years ago Updated 5 months ago

Intermittent ASAN shutdown hang ==1709==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x7fd9769b9015 bp 0x7fd95d539df0 sp 0x7fd95d539de0 T76)

Categories

(Toolkit :: Async Tooling, defect, P5)

Product:
Toolkit
Component:
Async Tooling
Type:
defect

Tracking

()

People

(Reporter: intermittent-bug-filer, Unassigned)

References

(Blocks 2 open bugs)

Details

(Keywords: intermittent-failure) 

glandium: I'm lost triaging ASAN bugs and the component this belongs in is not obvious from the stack. Could you please triage it?
Flags: needinfo?(mh+mozilla)
Here's where the crash is happening: MOZ_CRASH("Shutdown too long, probably frozen, causing a crash.");

Sadly, ASAN doesn't say what other threads were doing, so there's not much actionable here.

Geoff, do you know what we do with shutdown hangs like this?
Flags: needinfo?(mh+mozilla) → needinfo?(gbrown)
(In reply to Mike Hommey [:glandium] from comment #2)
> Here's where the crash is happening: MOZ_CRASH("Shutdown too long, probably
> frozen, causing a crash.");
> 
> Sadly, ASAN doesn't say what other threads were doing, so there's not much
> actionable here.
> 
> Geoff, do you know what we do with shutdown hangs like this?

Without an actionable crash report, I would normally check to see if the shutdown hang is associated with a specific test, or test directory, and direct the investigation that way. In this case, there's lots of variation: of 4 known failures, 2 happen after browser/extensions/formautofill tests, 1 after browser/base/content/test/siteIdentity, and 1 after toolkit/content/tests/browser...I don't think that's enough to redirect. Given the low failure frequency (so far), I'd just wait and see what happens: Once we get more failures, maybe we'll see a pattern.
Flags: needinfo?(gbrown)
Product: Core → Firefox Build System
gbrown: we're still seeing this failure with some frequency. Would you mind revisiting your assessment from comment #5 to see if this is actionable now?
Flags: needinfo?(gbrown)
Thanks for the reminder.

I still do not see a significant pattern: This seems to happen after a variety of mochitests and reftests.

:ryanvm -- Any thoughts on who might be able to look into this intermittent linux-only, asan-only shutdown crash?
Flags: needinfo?(gbrown) → needinfo?(ryanvm)
This is just a dupe of bug 1358898, isn't it?
Flags: needinfo?(ryanvm)
Thanks Ryan, yes, that seems likely -- this is the asan expression of bug 1358898. We'd better keep this bug open since it is a distinct failure message, but let's align it more closely with 1358898 and hope that work on bug 1358898 eventually resolves this bug.
Blocks: RunWatchdogShutdownhang
Component: General → Async Tooling
Product: Firefox Build System → Toolkit
Bug 1358898 is more a meta bug about all those shutdown hang crashes. All the last months the most occurring hang for Marionette and wdspec tests at least is bug 1411908 for the QuotaManager. The bug will be fixed soon, so lets wait and see if that is related here.
Summary: Intermittent GECKO(1709) | ==1709==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x7fd9769b9015 bp 0x7fd95d539df0 sp 0x7fd95d539de0 T76) → Intermittent ==1709==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x7fd9769b9015 bp 0x7fd95d539df0 sp 0x7fd95d539de0 T76)
Severity: normal → S3

From this instance I understand that this is just a different (unexpected) way of categorizing a "normal" shutdown hang?

Flags: needinfo?(gsvelto)

Yes; IIRC ASAN disables the crash reporter so we get a different message in the log compared to builds w/ the crash reporter enabled.

Flags: needinfo?(gsvelto)
Summary: Intermittent ==1709==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x7fd9769b9015 bp 0x7fd95d539df0 sp 0x7fd95d539de0 T76) → Intermittent ASAN shutdown hang ==1709==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x7fd9769b9015 bp 0x7fd95d539df0 sp 0x7fd95d539de0 T76)
See Also: → 1805368
You need to log in before you can comment on or make changes to this bug.