185287 - 100% cpu usage but then sleep wake cycle resets to normal

Status: RESOLVED WORKSFORME

(Camino Graveyard :: General, defect)

Description

[S Woodside]

I just noticed that Chimera 2002121304 was chewing up 100% cpu time in top. I
put the system to sleep and then woke it up and chimera immediately returned to
normal (minimal) cpu usage.

Comment 1

[S Woodside]

whoops should have left this UNCO

Comment 2

[saari (gone)]

I could swear this is a dupe

Comment 3

[Simon Fraser [no longer active]]

Hrm, I fixed a bug with big cpu usage when waking from sleep. Simon: if you see
this again, please run sampler.

Comment 4

[S Woodside]

before/after/both?

Comment 5

[Simon Fraser [no longer active]]

During high cpu usage. Also, note what windows chimera is showing. Is the caret
blinking in a text field in the page content? Are there animated GIFs?

Comment 6

[S Woodside]

Attachment: a sample

Frontmost window is http://www.w3.org/TR/xpath I don't think there's any
textboxes or animated gifs in there. I have five tabs open, also:
http://www.w3.org/TR/xslt#named-templates
http://localhost/form.html?=&= -- has a form but no cursor, or blinking
http://www.w3.org/TR/html401/interact/forms.html#h-17.5
http://www.oasis-open.org/committees/relax-ng/tutorial-20011203.html

Comment 7

[S Woodside]

Attachment: another sample

Taken just after the first one. Note that as I write in in another chimera
window, chimera is still sucking the CPU. There's also another process running
high. it's odd, it's PID 0 : kernel_tas (that's all that top shows, I guess
it's kernel_task). This does not show up in ps -aux, and I cannot run sample on
it from the command line. top -u says e.g.:

  452 Navigator   68.8% 30:18.86   8   152   565  25.9M  30.1M+ 32.9M	130M 
    0 kernel_tas  26.6%  8:18.09  28	 0     -      -      -	26.0M	325M

Comment 8

[S Woodside]

See ifyou can reconstruct what I was doing based on the pages I had open in
comment #6 ;-)

Comment 9

[Simon Fraser [no longer active]]

That is this thread?

    150 Thread_5303
      150 _pthread_body
        150 CAPThread::Entry(CAPThread*)
          150 HALRunLoop::OwnThread(void*)
            150 CFRunLoopRunSpecific
              150 __CFRunLoopRun
                150 mach_msg
                  150 mach_msg_trap
                    150 mach_msg_trap [STACK TOP]

I don't see that on my machine.

Comment 10

[S Woodside]

Agreed. I don't have that on my navigator right now either (behaving normally).
also - based on a mach trap corrolates with the high kernel CPU usage - HAL is
for hardware abstraction layer ... something to do with audio?

Comment 11

[S Woodside]

The plot thickens. I just woke up my PB and saw the kernel process meter on
"Spy" (in the toolbar) was at 100% but the user process meter was normal. I went
into top and saw kernel_process (PID 0) was sucking all the CPU up, but the
computer was still usable. Navigator was /not/ sucking the CPU. On a hunch I
quit Navigator and the kernel CPU usage dropped back to normal. Sorry, I didn't
get a sample.

Comment 12

[S Woodside]

Attachment: sample3a - after wake from sleep

Comment 13

[S Woodside]

Attachment: sample3b - after 2nd sleep cycle

Comment 14

[S Woodside]

Attachment: sample3c - after 3rd sleep cycle - back to normal CPU usage

Comment 15

[S Woodside]

Attachment: sample3d - Mail.app misbehaving

Comment 16

[S Woodside]

Attachment: 'top' output for sample3a and sample3b

Comment 17

[S Woodside]

I just had this again. I woke the system from sleep and found chimera was
sucking CPU cycles. I sampled (3a) and then cycled sleep. To my surprise it was
again spinning (3b) and so was Mail.app (3d). I killed off Mail which was not
responding, although chimera was responding OK. Then I sleep cycled again and
navigator was back to normal (3c). I forgot to sample chimera after killing mail
and before sleeping.

I also grabbed some screen scapes of top with expanded -uw output.

Comment 18

[S Woodside]

Note that my machine is a normal production machine, but it has some slight
flakiness in that the fan never comes on when it overheats and sometimes that
causes a kernel panic if I'm running DVDs, iTunes and charging the battery.
However since this occurred after sleeping, I doubt the computer was hot. The
system is a TiBook/400 with 256 ram, 10.2.3 currently, and 2003010804 currently.

Comment 19

[saari (gone)]

nominating topembed as it may also affect embedding customers

Comment 20

[Michael Buckland]

Discussed in edt.  Plussing.

Comment 21

[Simon Fraser [no longer active]]

I think bug 197863 is the one you really want to topembed+

Comment 22

[Simon Fraser [no longer active]]

We need more info on this bug, in particular, how it relates to bug 197863, and
bug 197889.

Comment 23

[Mike Pinkerton (not reading bugmail)]

if other apps have some issues (mail.app), and it's at the kernel level, and
your fan doesn't always know to come on, i think you've got bigger problems :(

Comment 24

[S Woodside]

Could be. It's definitely flaky in the overheating thing.

Comment 25

[Simon Fraser [no longer active]]

Removing topembed+, since this is not obviously a problem for most people, and
bug 197863 is the one we should care about.

Comment 26

[Michael Buckland]

Discussed in topembed bug triage.  Minusing.

Comment 27

[Ludovic Hirlimann [:Usul]]

Simon could you update this issue with a newer build ?

Comment 28

[S Woodside]

I haven't seen this since I got a new computer. I'll close it, and reopen if I
ever see it again.