NFSv4 client lock-ups

[Fredrik Tolf]

I've been noticing some strange lock-ups in the NFSv4 client. I'm
running kernel 2.6.18-gentoo-r3, and every once in a while (say around 5
times a day that I notice), a process (most often firefox or evolution,
but not at all limited to these) will lock up sleeping for around 60-90
seconds or so. The wait time is indicated in I/O wait, but I don't
remember whether the process sleeps uninterruptibly.

"ps -o pid,wchan=WIDE-WCHAN-COLUMN,comm" on the process says that WCHAN
is nfs_wait_on_write, which I'm guessing should be
nfs_wait_on_write_congestion, but shortened to the column width.

During that time, it seems completely random whether other processes
that also use NFS will lock up on the same thing. Sometimes they do,
other times they don't. I've never seen *all* NFS-using processes lock
up at the same time. When several processes lock up, it seems they all
continue running at the same time.

Is there anything I can do to pinpoint the problem?

Fredrik Tolf