Skip to contentSkip to conversation

NFS problems with 2.6.4-rc1-mm2

DISCUSSION Peter Zijlstra 2 participants 4 messages Mar 4, 2004 – Mar 6, 2004 1 message not in the archive

lore
Share & tools
lore
#0 lore raw

Hi,

I've just build and booted 2.6.4-rc1-mm2, and mounting my NFS exports works. however when I try to unzip a file over those mount the process freezes over and any other IO to that mount results in more stuck processes. SIGKILL will not remove the processes, only reboot will manage.

dmesg reports like:

nfs: server 192.168.0.1 not responding, timed out

which is total nonsense, because all other hosts on the network can access the exports just fine.

I'm about to back out all nfs patches from the broken-out patch set to see what that does for me.

Peter Zijlstra

#1 lore raw

I'm amazed at the total lack of information I managed to put into it ;-)

Its a dual athlon machine, nfs v3 over tcp. nfsd is run on vanilla 2.4.22 and reports no problems. (yes I know I should upgrade that machine but its not exposed to the outside world).

anyway,. I played about a bit and it's in one of these patches:

#nfs-remove-XID-spinlock.patch
#nfs-misc-rpc-fixes.patch
#nfs-improved-writeback-strategy.patch
#nfs-simplify-config-options.patch
#nfs-fix-msync.patch
#nfs-mount-return-useful-errors.patch
#nfs-misc-minor-fixes.patch
#nfs-lockd-sync-01.patch
#nfs-lockd-sync-02.patch
#nfs-lockd-sync-03.patch
#nfs-lockd-sync-04.patch
#nfs-rpc-remove-redundant-memset.patch
#nfs-tunable-rpc-slot-table.patch
#nfs-short-read-fix.patch

I'll do a binary search through these patches and report those that seem to cause the problem.

Peter Zijlstra

Quoted 19 lines · Peter Zijlstra

Hi,

I've just build and booted 2.6.4-rc1-mm2, and mounting my NFS exports works. however when I try to unzip a file over those mount the process freezes over and any other IO to that mount results in more stuck processes. SIGKILL will not remove the processes, only reboot will manage.

dmesg reports like:

nfs: server 192.168.0.1 not responding, timed out

which is total nonsense, because all other hosts on the network can access the exports just fine.

I'm about to back out all nfs patches from the broken-out patch set to see what that does for me.

Peter Zijlstra

A message that is not in the archive (sent off-list or not yet received): [email protected]

#3 lore raw

The bad guy is:
nfs-tunable-rpc-slot-table.patch
without that patch nfs works fine for me.

As said I use NFSv3 over TCP, and its a SMP machine that has the problem, I haven't tried any UP box.

Peter

Quoted 23 lines · Andrew Morton
Quoted 18 lines · Peter Zijlstra

I played about a bit and it's in one of these patches:

#nfs-remove-XID-spinlock.patch
#nfs-misc-rpc-fixes.patch
#nfs-improved-writeback-strategy.patch
#nfs-simplify-config-options.patch
#nfs-fix-msync.patch
#nfs-mount-return-useful-errors.patch
#nfs-misc-minor-fixes.patch
#nfs-lockd-sync-01.patch
#nfs-lockd-sync-02.patch
#nfs-lockd-sync-03.patch
#nfs-lockd-sync-04.patch
#nfs-rpc-remove-redundant-memset.patch
#nfs-tunable-rpc-slot-table.patch
#nfs-short-read-fix.patch

Quoted 2 lines

I'll do a binary search through these patches and report those that seem to cause the problem.

Thanks, that would be muchly appreciated.

#4 lore raw

På lau , 06/03/2004 klokka 06:10, skreiv Peter Zijlstra:

Quoted 6 lines

The bad guy is:
nfs-tunable-rpc-slot-table.patch
without that patch nfs works fine for me.

As said I use NFSv3 over TCP, and its a SMP machine that has the problem, I haven't tried any UP box.

I have no such problems on my SMP boxes. Could you try just resetting the default TCP slot size to 16?

sysctl -w sunrpc.tcp_slot_table_entries=16

that should in theory give you the same settings as without the above patch.

Cheers,
Trond