303972 - High CPU usage/churn

Status: RESOLVED DUPLICATE of bug 265172

(Firefox :: General, defect)

Description

[emerson.clarke]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-GB; rv:1.7.10) Gecko/20050717 Firefox/1.0.6
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-GB; rv:1.7.10) Gecko/20050717 Firefox/1.0.6

After suspending and resuming my laptop over several days i almost always
encounter a situation where firefox begins to flatline the cpu for no apparent
reason.

Closing the browser and re-opening all of the same pages seems to fix the
problem, indicating that its not page specific.

Reproducible: Always

Steps to Reproduce:
1. Open lots of pages 5-10, and suspend.
2. Resume. Then continue browsing.
3. Wash rinse and repeat for several days.

Actual Results:  
Firefox churned my cpu into the dirt.

Expected Results:  
Firefox should have been well behaved and used only idle cpu.

I have a stack trace from process explorer which shows 85.71% of cpu in the
jpeg_free_large call from firefox.exe.  That same thread also seems to be doing
a lot of context switching, so id guess its locking/unlocking a lot, probably
memory  allocation stuff.

Possibly related to this bug, which has another screenshot showing a different
thread chewing cpu, but jpeg_free_small is in the picture also.

https://bugzilla.mozilla.org/show_bug.cgi?id=271636

If i was to hazard a guess id say there might be a deadlock issue, and its
almost definately related to that jpeg_free* code.

Comment 1

[emerson.clarke]

To provide some aditional detail, from the stack trace.  Process explorer shows
the following items in the detailed stack dump for the churning thread.

0 NPSWF32.dll+0x24ed7
1 NPSWF32.dll+0x25061

I guess that implicates flash somewhat also...

Oh and if its useful the context switch delta was at 335, atleast three times
that of the other threads. 

Comment 2

[Jo Hermans]

probably bug 76831, although jpeg_free_large wasn't mentioned there

Comment 3

[emerson.clarke]

I don't think its related to 76831, that bug appears to be all about stuff being
swapped out and back in and would only do momentary damage.  This one is much
more of a permanent hosing...

And i was running on a 2GHz Centrino with 1GB of ram, so theres no reson for
delay.  The CPU usuage is not recoverable, it just stays like that till the
applicatoin is closed.

Comment 4

[Tim A. Miller]

Pretty sure this is a duplicate of bug 265172.  

Comment 5

[Masatoshi Kimura [:emk]]

*** This bug has been marked as a duplicate of 265172 ***