what on earth is firefox up to?

[Scott Karlin]

On 06/30/12 22:17, Tom Horsley wrote:
> Every time I start firefox after recent updates, top
> shows it periodically taking up to 50% of the CPU
> even if I'm just looking at a simple page of plain
> html (no scripts, not even any images) on my local
> web server.
>
> This is ff 13.0.1 on x86_64 fedora 17.
>
> Any clues? Anyone see anything similar?
>

I'm experiencing similar behavior with Firefox and Thunderbird.
While I don't have it nailed down, I've got some hunches that
may (or may not) apply to your situation.

I'm running F17 on x86_64 with
firefox-13.0.1-1.fc17.x86_64
thunderbird-13.0.1-1.fc17.x86_64

The issues I have are high CPU utilization and frequent
freezes of these applications.  The freeze lasts for a bit
and then the applications become responsive again.

I thought I had it licked when I switched from nouveau to
the nVidia proprietary drivers.  This did result in a significant
drop in CPU utilization but did not address the frequent
pauses in TB and FF.

Using strace, I found that a delay happened (to TB) when trying
to get a lock:

open("/u/scott/.pulse-cookie", O_RDWR|O_CREAT|O_NOCTTY|O_CLOEXEC, 0600) = 65
fcntl(65, F_GETFD)                      = 0x1 (flags FD_CLOEXEC)
fcntl(65, F_SETLKW, {type=F_WRLCK, whence=SEEK_SET, start=0, len=0}^C

Process 4385 detached
  <detached ...>

As you can see, I killed the strace.  Of note, my home directory is
NFS mounted.  It seems that I may be hitting this issue in pulse-audio:
http://lists.freedesktop.org/archives/pulseaudio-discuss/2011-August/011036.html
https://bugs.launchpad.net/ubuntu/+source/pulseaudio/+bug/817269
I followed the suggestion on comment 9:
https://bugs.launchpad.net/ubuntu/+source/pulseaudio/+bug/817269/comments/9
to move the pulse-cookie to /tmp and this improved things.  However,
I still get some freezes and more strace-ing led to the NFS mounted
~/.cache/event-sound-cache.tbd.{long_hex_string}.x86_64-redhat-linux-gnu
as another possible culprit.  I created a symlink from ~/.cache to
local disk and now things seem OK (with about 30 minutes of testing).

Scott