Re: Processes spinning forever, apparently in lock_timer_base()?

[Bruno Wolff III <bruno@wolff.to>]

On Fri, Sep 21, 2007 at 16:58:15 +0100,
  Hugh Dickins <hugh@veritas.com> wrote:
> 
> But once I look harder at it, I wonder what would have kept
> 2.6.18 to 2.6.23 safe from the same issue: per-cpu deltas from
> the global vm stats too low to get synched back to global, yet
> adding up to something which misleads balance_dirty_pages into
> an indefinite loop e.g. total nr_writeback actually 0, but
> appearing more than dirty_thresh in the global approximation.

I have been seeing this problem since Fedora Core 5's 2.6.20-1.2300.fc5
kernel. The last Fedora kernel that didn't have this problem was
2.6.19-1.2288.2.4.fc5, which I am still running on one machine.

I have put up vmstat and proc/meminfo data at: http://wolff.to/bruno/vmlog
There are a number of iterations of the output and they are mixed together,
though the vmstat info was written when a buffer was full, so there are
chunks of lines together. This was from a machine running Fedora 7's
2.6.23.1-4.fc7 kernel.

The machine is a single processor P3 with 512 MB of memory. There are
three hard drives, two of which are used in software raid 1 arrays.
I have ext3 file systems mounted noatime. I can provide more details if
you need anything else.

I can get lockups to happen fairly often. Often lots of firefox activity
and/or some other disk heavy task such as yum update will get it to happen.

I am not subscribed to LMKL, so please keep me copied on replies.