425929 - Rare hangs on shutdown. SIGILL: [@0xffffe410], [pt_PostNotifies], [libpthread-2.7.so@0x8df1] with high CPU

Status: RESOLVED WORKSFORME

(Core :: XPCOM, defect)

Description

[[:Aleksej]]

Mozilla/5.0 (X11; U; Linux i686; eo; rv:1.9pre) Gecko/2008032804 Minefield/3.0pre and builds before it occasionally hang on shutdown, using 100% of CPU resources.

Didn’t reproduce with a clean profile.

Reports after kill -ILL pid:

http://crash-stats.mozilla.com/report/index/9fca3b5c-f278-11dc-b93a-001a4bd43ef6
http://crash-stats.mozilla.com/report/index/7796adc7-f4fd-11dc-9cfa-001a4bd43ef6
http://crash-stats.mozilla.com/report/index/240500bd-fd91-11dc-9574-001a4bd43ef6

Comment 1

[[:Aleksej]]

Attachment: The top repeating part of report 7796adc7-f4fd-11dc-9cfa-001a4bd43ef6.

Top part of report
http://crash-stats.mozilla.com/report/index/7796adc7-f4fd-11dc-9cfa-001a4bd43ef6
that seems to repeat (except for a few final lines at the top of it).

Comment 2

[[:Aleksej]]

The summary of http://crash-stats.mozilla.com/report/index/7796adc7-f4fd-11dc-9cfa-001a4bd43ef6 :

Signature	pt_PostNotifies
UUID	7796adc7-f4fd-11dc-9cfa-001a4bd43ef6
Time	2008-03-18 08:10:08-07:00
Uptime	0
Product	Firefox
Version	3.0b5pre
Build ID	2008031704
OS	Linux
OS Version	0.0.0 Linux 2.6.22-3-k7 #1 SMP Sun Feb 10 21:04:14 UTC 2008 i686 GNU/Linux
CPU	x86
CPU Info	AuthenticAMD family 1 model 44 stepping 2
Crash Reason	SIGILL
Crash Address	0xb72272d0

Of http://crash-stats.mozilla.com/report/index/9fca3b5c-f278-11dc-b93a-001a4bd43ef6 :

Signature	libpthread-2.7.so@0x8df1
UUID	9fca3b5c-f278-11dc-b93a-001a4bd43ef6
Time	2008-03-15 03:13:56-07:00
Uptime	0
Product	Firefox
Version	3.0b5pre
Build ID	2008031404
OS	Linux
OS Version	0.0.0 Linux 2.6.22-3-k7 #1 SMP Sun Feb 10 21:04:14 UTC 2008 i686 GNU/Linux
CPU	x86
CPU Info	AuthenticAMD family 1 model 44 stepping 2
Crash Reason	SIGILL
Crash Address	0xb7f59df1

Of http://crash-stats.mozilla.com/report/index/240500bd-fd91-11dc-9574-001a4bd43ef6 :

Signature	@0xffffe410
UUID	240500bd-fd91-11dc-9574-001a4bd43ef6
Time	2008-03-29 06:07:21-07:00
Uptime	77752
Product	Firefox
Version	3.0pre
Build ID	2008032804
OS	Linux
OS Version	0.0.0 Linux 2.6.22-3-k7 #1 SMP Sun Feb 10 21:04:14 UTC 2008 i686 GNU/Linux
CPU	x86
CPU Info	AuthenticAMD family 1 model 44 stepping 2
Crash Reason	SIGILL
Crash Address	0xffffe410

Comment 3

[Andrew Schultz]

> Didn’t reproduce with a clean profile.

Can you determine which part of the profile is needed to hit the bug?  If nothing else, you can back the whole thing up and just delete files until the bug goes away.

Comment 4

[[:Aleksej]]

(In reply to comment #3)
> If nothing else, you can back the whole thing up and just delete files until
> the bug goes away.

No, it happens too rarely for that.

Comment 5

[Jonathan ROTH]

Hi,

I got the same symptoms, but reproductible by closing tab(s).
CPU usage goes about 100%, and memory usage somewhere between 350 and 700MB.

I'm investigating.

Comment 6

[[:Aleksej]]

This happened, and still happens as I've seen yesterday, when I run the Firefox from xfce4-terminal.

Now I run Firefox using gksu, and this never happens, but also crashreporter never appears.

Comment 7

[[:Aleksej]]

(In reply to comment #6)
It happened twice that day, when running using gksu. crashreporter appeared once, but soon disappeared.

Comment 8

[Wayne Mery (:wsmwk)]

Jonathan in comment #5)
> I got the same symptoms, but reproductible by closing tab(s).
> CPU usage goes about 100%, and memory usage somewhere between 350 and 700MB.
> 
> I'm investigating.

Same symptoms = same stacks?
What steps to reproduce?

Comment 9

[[:Aleksej]]

During the winter I’ve switched to GNU IceCat, and I think I’ve never seen this hang happen with it.

Comment 10

[[:Aleksej]]

(In reply to comment #9)
On the other hand, I am now using an older computer (256 MiB of RAM), and so restart the browser much more often.

Comment 11

[Wayne Mery (:wsmwk)]

but you no longer see the problem, correct?

Comment 12

[[:Aleksej]]

I am using Debian Iceweasel 3.0.x and no longer see the problem.
Although it appears to crash on shutdown sometimes (the "Clear private data" dialog appears when starting it again), no hangs on shutdown.

Comment 13

[Wayne Mery (:wsmwk)]

=> WFM per reporter