1663718 - Crash in [@ OOM | large | xul.dll | NS_internal_main]

Status: RESOLVED FIXED

(Core :: Networking: HTTP, defect, P2)

Description

[[:philipp]]

Crash report: https://crash-stats.mozilla.org/report/index/fdbeebaf-6ce6-4652-bc4e-638380200908

Top 10 frames of crashing thread:

0 xul.dll xul.dll@0x7e9e0c 
1 xul.dll xul.dll@0x3a9aa 
2 xul.dll xul.dll@0x1a3b12 
3 xul.dll xul.dll@0xd0378c 
4 xul.dll xul.dll@0xed22d1 
5 xul.dll xul.dll@0xd64d25 
6 xul.dll xul.dll@0xcde5ee 
7 xul.dll xul.dll@0xcdf51a 
8 xul.dll xul.dll@0x864d6d 
9 xul.dll xul.dll@0x869a66 

these browser crashes are regressing in volume in firefox 81, mostly on google web properties. in 81.0b7 related signatures are accounting for 3% of browser crashes.

this is one way i'm able to reproduce the crash:

• with a 32bit firefox on windows open https://www.google.com/search?hl=de&q=speedtest&gws_rd=ssl
• click on the button to perform a speedtest directly on google's widget
• while the upload speed is tested the browser crashes for me with the reported signature

bisecting with mozregression leads me to https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=0d22ad297b19d1653b679f731dd929e23d93d2a5&tochange=1cd6d2f9a603ad02a533b12c4602612e90f7da5a and bug 1497249 as having introduced the crash in this instance.

another url that's frequently showing up in crash reports is https://ogs.google.com/u/0/widget/app

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Kershaw, can you please take a look at this? Given where we are in the 81 Beta cycle, we may need a quick fix before next week's RC too (is this something we can pref off?).

Comment 2

[Kershaw Chang [:kershaw]]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #1)

Kershaw, can you please take a look at this? Given where we are in the 81 Beta cycle, we may need a quick fix before next week's RC too (is this something we can pref off?).

I don't know why, but the stacktrace of crash reports are meaningless. It's really difficult to debug.
I'd suggest to backout the patches in https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=0d22ad297b19d1653b679f731dd929e23d93d2a5&tochange=1cd6d2f9a603ad02a533b12c4602612e90f7da5a.

What do you think?

Comment 3

[Ryan VanderMeulen [:RyanVM]]

Backed out for 81.0b8. Fingers crossed!
https://hg.mozilla.org/releases/mozilla-beta/rev/0218c1751126fd28df953d37fb611a82362f9e56

Comment 4

[Kershaw Chang [:kershaw]]

Attachment: Bug 1663718 - Don't put too much data in buffer when the data can't be written to socket

Comment 5

[Ryan VanderMeulen [:RyanVM]]

The 81.0b8 crash rate is significantly lower than previous betas. Looks like the backout worked.

Comment 6

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/a2372b6285d7
Don't put too much data in buffer when the data can't be written to socket r=dragana

Comment 7

[Bogdan Tara[:bogdan_tara | bogdant]]

https://hg.mozilla.org/mozilla-central/rev/a2372b6285d7