957540 - Hang on shutdown when restarting Firefox for an update

Status: RESOLVED WORKSFORME

(Toolkit :: Startup and Profile System, defect)

Description

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Attachment: sysprof.txt

I have already seen this a couple of times and it may be related to my strange memory conditions with unclassified heap memory as given on bug 751557.

I tried to apply the update to yesterdays Aurora build from my current build from 2014-01-02. The download and applying the update was successful, but when I triggered the restart Firefox never shutdown. I finally killed the process after 7 minutes. Firefox had a cpu load of about 400% all the time.

PID USER      PR  NI  VIRT  RES  SHR S  %CPU %MEM    TIME+ COMMAND                                                                                  
18440 henrik    20   0 3517m 2.1g  48m S 400.0 13.6 701:45.38 firefox 

Attached you can find the log file from sysprof.

Comment 1

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Interestingly the syslog log shows some inode mismatches, here for libnspr4.so and libxul.so. Not sure if those are related to this problem. But more interesting is the mutex lock. Looks like we entered an infinite loop here?

Comment 2

[Olli Pettay [:smaug][[e-mail]]]

I doubt this has anything to do with bug 751557, but the lock is suspicious.
Unfortunately that sysprof log doesn't tell much.

Comment 3

[Henrik Skupin [:whimboo][⌚️UTC+1]]

What else should I do next time I get this hang? I don't run debug builds, so would a gdb output be helpful?

Comment 4

[Olli Pettay [:smaug][[e-mail]]]

If the build has symbols, gdb would be still useful.

Comment 5

[Robert Strong (they/them - no direct email)]

Moving to the component that is responsible for restarting

Comment 6

[Lukas Blakk [:lsblakk] use ?needinfo]

How reproducible is this?  Not clear that this is something our users are hitting widely - can Sumo also weigh in on any input that might correlate to this?

Comment 7

[[:Cww]]

This sounds like Aurora only? We don't get much Aurora input... This is also going to be a hard thing for users to report because it's on shutdown when most people are already shutting down their computers. If this is also Linux only, then we have no data. The last time we had non-spam input from Linux on 28 or 29 was in November.

The only thing I've found that seems kinda close is:

https://input.mozilla.org/en-US/dashboard/response/4144473

There are also general complaints of slowness, but I don't have enough data to say if that's significantly more than usual (I don't think so):

https://input.mozilla.org/en-US/dashboard/response/4137206
https://input.mozilla.org/en-US/dashboard/response/4137152
https://input.mozilla.org/en-US/dashboard/response/4134275

Comment 8

[Ioana (away)]

I update Aurora every day on Ubuntu 13.04 32bit and didn't run into this issue once, so I guess it's not that reproducible. If anyone has more details about how to reproduce it than what's already in this report, let me know and I'll try that.

Comment 9

[Henrik Skupin [:whimboo][⌚️UTC+1]]

This bug might also be related to hibernation mode given that I'm updating Firefox not that regularly. I put my box into sleep mode a couple of times a day. So it might be that after a couple of days we end up with this behavior.

Comment 10

[Lukas Blakk [:lsblakk] use ?needinfo]

This sounds like an edge case and not something we need to track for release.

Comment 11

[Wayne Mery (:wsmwk)]

Henrik, is this now WFM for you?

Comment 12

[Henrik Skupin [:whimboo][⌚️UTC+1]]

I actually haven't seen this problem for a while. So I would close this bug as WFM, and reopen in case I see it again.