Skip to contentSkip to conversation

NFS problems, 2.5.5x

DISCUSSION David Ford 5 participants 8 messages Feb 1, 2003 – Feb 3, 2003

lore
Share & tools
lore

David Ford

author
#0 lore raw

Synopsis: nfsserver:/home/david mount, get dir. entries loops forever, 2.5.59 for client and server.

Example: ls -l /home/david

An strace will show the same directory entries flying by over and over until memory is exhausted or ^c comes along. It worked at first for about 30 minutes while I finished the new gentoo install on my desktop, but then things got weird. the nfs server spat out a big long callback trace (oops) and died hard. Had to reset the power. The looping started just minutes before that. I've rebooted, tried 2.5.53 on the client but no go.

Doing a stat on a single file works fine. Doing a glob, i.e. ls -l on the mount directory fails. Doing ls -l on any sub directory of the mount works fine.

ls -la /home/david/.xinitrc (file, works)
ls -la /home/david/.e (directory, works)
ls -la /home/david (loops forever on all directory entries until memory is exhausted and ls aborts)

Some things I've noted:

- not all directory entries are repeated, some only appear once and never again.
- ls only does this on the mount point of the nfs mounted directory, all other directories are fine

GLIBC 2.3.1, underlying filesystems are reiserfs.

Client:
nfsserver:/home/david on /home/david type nfs (rw,bg,hard,intr,timeo=7,rsize=16384,wsize=16384,addr=10.0.0.5)

Server (/var/lib/nfs/etab):
/raid/home/david hb.blue-labs.org(rw,async,wdelay,hide,secure,no_root_squash,no_all_squash,subtree_check,secure_locks,mapping=identity,anonuid=-2,anongid=-2)

David

#1 lore raw
Quoted 3 lines · David Ford

Synopsis: nfsserver:/home/david mount, get dir. entries loops forever, 2.5.59 for client and server.

If the server is ext3+htree then you've hit the htree dir cookie bug.

Use `dumpe2fs -h /dev/hda1 | grep index' to se if you're using htree.

Use `tune2fs -O ^dir_index /dev/hda1' to disable it.

author
#2 lore raw

Underlying filesystems on both client and server are reiserfs.

Got any more quickie suggestions? :)

Thanks,
David

Quoted 14 lines · Andrew Morton
Quoted 3 lines · David Ford

Synopsis: nfsserver:/home/david mount, get dir. entries loops forever, 2.5.59 for client and server.

If the server is ext3+htree then you've hit the htree dir cookie bug.

Use `dumpe2fs -h /dev/hda1 | grep index' to se if you're using htree.

Use `tune2fs -O ^dir_index /dev/hda1' to disable it.

#3 lore raw
In reply to David Ford
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

" " == David Ford <[email protected]> writes:

> Synopsis: nfsserver:/home/david mount, get dir. entries loops > forever,
> 2.5.59 for client and server.

> Example: ls -l /home/david

> An strace will show the same directory entries flying by over > and over until memory is exhausted or ^c comes along. It > worked at first for about 30 minutes while I finished the new > gentoo install on my desktop, but then things got weird. the > nfs server spat out a big long callback trace (oops) and died > hard. Had to reset the power. The looping started just > minutes before that. I've rebooted, tried 2.5.53 on the client > but no go.

AFAICR, there have been no changes to the NFS client readdir code since 2.5.30.

Cheers,
Trond

author
#4 lore raw

The last time NFS was working, I had 2.4.19 and 2.5.53 clients on a 2.5.59 server, that was yesterday. I had experienced a slight problem with it last week when my 2.5.53 client was booted for first time on 2.5.5x, it was previously a 2.4 kernel. The server OOPSed repeatedly shortly after bootup in NFS stuff then it never happened again and was rock solid until today.

David

Quoted 26 lines · Trond Myklebust
Quoted 10 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

" " == David Ford <[email protected]> writes:

Synopsis: nfsserver:/home/david mount, get dir. entries loops forever,
2.5.59 for client and server.

Example: ls -l /home/david

Quoted 8 lines

An strace will show the same directory entries flying by over and over until memory is exhausted or ^c comes along. It worked at first for about 30 minutes while I finished the new gentoo install on my desktop, but then things got weird. the nfs server spat out a big long callback trace (oops) and died hard. Had to reset the power. The looping started just minutes before that. I've rebooted, tried 2.5.53 on the client but no go.

AFAICR, there have been no changes to the NFS client readdir code since 2.5.30.

Cheers,
Trond

#5 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

" " == David Ford <[email protected]> writes:

> The last time NFS was working, I had 2.4.19 and 2.5.53 clients > on a
> 2.5.59 server, that was yesterday. I had experienced a slight
> problem
> with it last week when my 2.5.53 client was booted for first > time on 2.5.5x, it was previously a 2.4 kernel. The server > OOPSed repeatedly shortly after bootup in NFS stuff then it > never happened again and was rock solid until today.

So have you tried out the 2.5.53 client since you noticed this problem?

Cheers,
Trond

#6 lore raw

Yes. Today I haven't experienced the loop problem. On the other hand, when I reboot back and forth between 2.5.53 and 2.5.59, I have to restart the server nfs programs or I get permission denied on the client and "rpc.mountd: getfh failed: Operation not permitted" on the server.

I have also had to restart 2.4 clients because NFS silently hangs. I believe there's a few patches on the list that I need to apply regarding this.

David

Quoted 20 lines · Trond Myklebust
Quoted 4 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

" " == David Ford <[email protected]> writes:

Quoted 8 lines

The last time NFS was working, I had 2.4.19 and 2.5.53 clients on a
2.5.59 server, that was yesterday. I had experienced a slight
problem
with it last week when my 2.5.53 client was booted for first time on 2.5.5x, it was previously a 2.4 kernel. The server OOPSed repeatedly shortly after bootup in NFS stuff then it never happened again and was rock solid until today.

So have you tried out the 2.5.53 client since you noticed this problem?

Cheers,
Trond

Signature · 4 lines
I may have the information you need and I may choose only HTML.  It's up to you. Disclaimer: I am not responsible for any email that you send me nor am I bound to any obligation to deal with any received email in any given fashion.  If you send me spam or a virus, I may in whole or part send you 50,000 return copies of it. I may also publically announce any and all emails and post them to message boards, news sites, and even parody sites.  I may also mark them up, cut and paste, print, and staple them to telephone poles for the enjoyment of people without internet access.  This is not a confidential medium and your assumption that your email can or will be handled confidentially is akin to baring your backside, burying your head in the ground, and thinking nobody can see you butt nekkid and in plain view for miles away.  Don't be a cluebert, buy one from K-mart today.

When it absolutely, positively, has to be destroyed overnight.
                          AIR FORCE
#7 lore raw
In reply to Trond Myklebust

Hello!

Quoted 10 lines · Trond Myklebust

> The last time NFS was working, I had 2.4.19 and 2.5.53 clients > on a
> 2.5.59 server, that was yesterday. I had experienced a slight
> problem
> with it last week when my 2.5.53 client was booted for first > time on 2.5.5x, it was previously a 2.4 kernel. The server > OOPSed repeatedly shortly after bootup in NFS stuff then it > never happened again and was rock solid until today.
So have you tried out the 2.5.53 client since you noticed this problem?

While trying to reproduce mounting of reiserfs FS over NFS from 2.5.59 server, trying to see if there is something to do with reiserfs itself (and it worked perfectly with 2.4.19 client), I decided to try to mount it from this same workstation as I did not had 2.5.59 client at hand.
This way I learned that mount localhost:/exportedfs /mnt -t nfs (localhost can be replaced by any local IP) results in mount hanging in D state:
100 0 775 770 15 0 1532 620 rpc_ex D pts/1 0:00 mount 212.16.7.78:/home /mnt -t nfs

At the same time I still can mount it from external hosts (but cannot kill this hung mount, obviously).

This is on today's 2.5.529 bk snapshot.

Bye,
Oleg