quota deadlock on 6.1-RC1

[Konstantin Belousov]

On Thu, May 04, 2006 at 04:59:33PM -0700, David Kirchner wrote:
> Here's how to reproduce the snapshot deadlock I'm seeing, with 6.1-RC2
> cvsup'd as of 5 or 6 hours ago:
> 
> 1) dd if=/dev/zero of=/usr/bigfile bs=1024 seek=209715200 count=0
> 2) mdconfig -a -t vnode -f /usr/bigfile
> 3) bsdlabel -w md0 auto
> 4) newfs -U md0a
> 5) fsck -v /dev/md0a # ^C this after a second or so, this makes the FS dirty
> 6) mount /dev/md0a /mnt
> 7) fsck -v -B /dev/md0a
> 
> in another window:
> 8) while true; do ls -al /mnt/.snap;sleep 1;done
> 
> It locks up every time for me, with no further disk activity.
> Unfortunately, for some reason, my server console became unaccessable,
> so I'm not able to get to the kdb prompt. If I can get to it later,
> what should I run other than "show lockedvnodes" and "show threads"?
> Also, can anyone else try these steps and verify if they cause the
> same problem for you?
I repeat you recipe on CURRENT.
What I got was the completely unresponsively system,
that was _not_ deadlocked. It has slowly made a progress. Slowness is
surely related to hole in the file backing fsck'ed (and snapshotted)
filesystem. Snapshotting slowly made a progress, with lot of disk
activity. After it had finished, system resumed normal operation.

Tor Egge committed several fixes into CURRENT, that certainly
help in this situation.
> 
> In my initial tests, filed in a PR, steps #1 and #2 were unnecessary
> as I was working with real disks. The result is the same here. Still,
> I am curious if anyone else can get the same result with a real disk
> >=200GB in size. I am unable to duplicate it with a 20GB partition,
> and I am not sure why.
> 
> --
> David 'dpk' Kirchner
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20060505/f3f2fe30/attachment.pgp