nfs: lock stuck after interrupt

From: Miklos Szeredi
Date: Thu Apr 17 2008 - 13:44:52 EST


1) on server lock file X
2) on client lock file X
blocks
3) on client send interrupt to process doing locking
locking syscall restarted, continues blocking
4) on server release lock on file X
on client lock is acquired
5) on client release lock on file X
6) on client lock file X
blocks

Everything up to the last step is fine, but something goes wrong
during the final unlock. Stopping the nfs-server removes the stray
lock.

Here's a trace on the server:

lockd: request from 192.168.1.1, port=862
lockd: LOCK called
lockd: nlm_lookup_host(192.168.1.2->192.168.1.1, p=6, v=4, my role=server, name=tucsk)
lockd: get host tucsk
lockd: nsm_monitor(tucsk)
lockd: nlm_lookup_file (02000001 00006200 00000002 0001783d 7f54ef79 00017801 d06c5915 00000000)
lockd: creating file for (02000001 00006200 00000002 0001783d 7f54ef79 00017801 d06c5915 00000000)
lockd: found file 0ad074c0 (count 0)
lockd: nlmsvc_lock(ubda/96317, ty=1, pi=1, 0-99, bl=1)
lockd: nlm_lookup_host(192.168.1.2->192.168.1.1, p=6, v=4, my role=server, name=tucsk)
lockd: get host tucsk
lockd: nlmsvc_lookup_block f=0ad074c0 pd=1 0-99 ty=1
lockd: get host tucsk
lockd: created block 0aeabf80...
lockd: vfs_lock_file returned 1
lockd: nlmsvc_insert_block(0aeabf80, -1)
lockd: release host tucsk
lockd: nlmsvc_lock returned 50331648
lockd: LOCK status 3
lockd: release host tucsk
lockd: nlm_release_file(0ad074c0, ct = 2)
lockd: request from 192.168.1.1, port=862
lockd: CANCEL called
lockd: nlm_lookup_host(192.168.1.2->192.168.1.1, p=6, v=4, my role=server, name=tucsk)
lockd: get host tucsk
lockd: nlm_lookup_file (02000001 00006200 00000002 0001783d 7f54ef79 00017801 d06c5915 00000000)
lockd: found file 0ad074c0 (count 1)
lockd: nlmsvc_cancel(ubda/96317, pi=1, 0-99)
lockd: nlmsvc_lookup_block f=0ad074c0 pd=1 0-99 ty=1
lockd: check f=0ad074c0 pd=1 0-99 ty=1 cookie=36120000
lockd: unlinking block 0aeabf80...
lockd: freeing block 0aeabf80...
lockd: release host tucsk
lockd: nlm_release_file(0ad074c0, ct = 2)
lockd: CANCEL status 0
lockd: release host tucsk
lockd: nlm_release_file(0ad074c0, ct = 1)
lockd: closing file ubda/96317
lockd: request from 192.168.1.1, port=862
lockd: LOCK called
lockd: nlm_lookup_host(192.168.1.2->192.168.1.1, p=6, v=4, my role=server, name=tucsk)
lockd: get host tucsk
lockd: nsm_monitor(tucsk)
lockd: nlm_lookup_file (02000001 00006200 00000002 0001783d 7f54ef79 00017801 d06c5915 00000000)
lockd: creating file for (02000001 00006200 00000002 0001783d 7f54ef79 00017801 d06c5915 00000000)
lockd: found file 0ad074c0 (count 0)
lockd: nlmsvc_lock(ubda/96317, ty=1, pi=2, 0-99, bl=1)
lockd: nlm_lookup_host(192.168.1.2->192.168.1.1, p=6, v=4, my role=server, name=tucsk)
lockd: get host tucsk
lockd: nlmsvc_lookup_block f=0ad074c0 pd=2 0-99 ty=1
lockd: get host tucsk
lockd: created block 0aeab240...
lockd: vfs_lock_file returned 0
lockd: freeing block 0aeab240...
lockd: release host tucsk
lockd: nlm_release_file(0ad074c0, ct = 2)
lockd: release host tucsk
lockd: nlmsvc_lock returned 0
lockd: LOCK status 0
lockd: release host tucsk
lockd: nlm_release_file(0ad074c0, ct = 1)
lockd: request from 192.168.1.1, port=862
lockd: UNLOCK called
lockd: nlm_lookup_host(192.168.1.2->192.168.1.1, p=6, v=4, my role=server, name=tucsk)
lockd: get host tucsk
lockd: nlm_lookup_file (02000001 00006200 00000002 0001783d 7f54ef79 00017801 d06c5915 00000000)
lockd: found file 0ad074c0 (count 0)
lockd: nlmsvc_unlock(ubda/96317, pi=3, 0-9223372036854775807)
lockd: nlmsvc_cancel(ubda/96317, pi=3, 0-9223372036854775807)
lockd: nlmsvc_lookup_block f=0ad074c0 pd=3 0-9223372036854775807 ty=2
lockd: UNLOCK status 0
lockd: release host tucsk
lockd: nlm_release_file(0ad074c0, ct = 1)


Everything looks normal, yet...

This is 100% reproducable for me (ext3 exported over nfs, server and
client: 2.6-git).

Miklos
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/