You might be running into the cfs_wait_on_write_congestion() bug.  It happens in 1.9.1 too, but you just happen to run into it now.  I've checked in a temporary workaround to keep you from hanging forever.  Check your kernel debug log for messages looking like cfs_writepages(): deadlock...  Let me know.

I might have an idea how to fix this if there really is a race between test_and_set_bit() in cfs_writepages() and some other function(s) - and cfs_wait_on_write_congestion() happens to be in the middle of it.

Roger


On 10/26/05, J.H. <warthog19@eaglescrag.net> wrote:
Upgraded to CVS (OPENSSI-FC) yesterday and am trying to tweak the kernel
a little (mainly just recompiling it), and noticed that untaring the
linux-2.6.10.tar.bz2 file was taking an inordinate amount of time.
Doing some checking I noticed that it's (seemingly) getting stuck in the
kernel inside of cfs_wait in a D+ state (from ps -alx).  This didn't
seem to be the case in 1.9.1.  It eventually continues without issue
(maybe a minute or so later) but I thought I'd see if anyone else is
seeing this.

- John


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)

iD8DBQBDX/m9/E3kyWU9dicRAhgrAJ9xiewC1HyJ1qIj7jDJ2CUgOsjb5wCgghK3
pkVaKHerj/bzc24yStaY8gw=
=7v2Z
-----END PGP SIGNATURE-----