NFS hang on 11.31

[Tanuj Vohra]

Some of my processes regularly hang accessing NFS directories. The NFS client is 11.31 rx2620.

Here is the kernel stack trace for one such process:

#0 0xe000000000890880:0 in swtch_to_thread+0x2e0 ()
at /ux/core/kern/common/pm/sched/pm_swtch.c:466
#1 0xe0000000007f3a40:0 in inline slpq_swtch_core+0x34e ()
at /ux/core/kern/common/pm/sched/sched_sleep.c:3268
#2 0xe0000000007f36f0:2 in slpq_swtch+0x52 ()
at /ux/core/kern/common/pm/sched/sched_sleep.c:3039
#3 0xe0000000006d1fc0:0 in inline cv_sleep+0x90 ()
at /ux/core/kern/common/svc/sync_sema/condvar.c:786
#4 0xe0000000006d1f30:0 in cv_wait+0x50 ()
at /ux/core/kern/common/svc/sync_sema/condvar.c:947
#5 0xe000000001358800:0 in $cold_vas_fcache_unmap+0x170 ()
#6 0xe00000000073b860:0 in inline fcache_as_ctl_remove+0x3c0 ()
at /ux/core/kern/common/vm/vmfs/fcache_as.c:1572
#7 0xe00000000073b4a0:0 in fcache_as_remove+0x160 ()
at /ux/core/kern/common/vm/vmfs/fcache_as.c:1565
#8 0xe0000000007a9f30:0 in fcache_vn_destroy+0x2d0 ()
at /ux/core/kern/common/vm/vmfs/fcache_vn.c:877
#9 0xe000000001c82270:0 in hpnfs_vnode_cleanup+0x1c0 ()
at /ux/core/isu/onc/kern/src/common/fs/onc/nfs/hpnfs_vm_wrappers.c:480
#10 0xe00000000243aa10:0 in $cold_make_rnode+0x2a0 ()
#11 0xe00000000061fde0:0 in makenfs3node+0x80 ()
at /ux/core/isu/onc/kern/src/common/fs/onc/nfs/nfs_subr.c:2195
#12 0xe000000001c96060:0 in nfs3create+0x3f0 ()
at /ux/core/isu/onc/kern/src/common/fs/onc/nfs/nfs3_vnops.c:2658
#13 0xe000000001c96d40:0 in nfs3_create+0x390 ()
at /ux/core/isu/onc/kern/src/common/fs/onc/nfs/nfs3_vnops.c:2499
#14 0xe000000001c654e0:0 in hpnfs3_create+0x180 ()
at /ux/core/isu/onc/kern/src/common/fs/onc/nfs/hpnfs3_vnops.c:1901
#15 0xe000000000926270:0 in inline vns_create+0xf0 ()
at /ux/core/kern/common/fs/vfs/vfs_vnode.c:953
#16 0xe000000000926180:0 in inline vn_create+0x14f ()
at /ux/core/kern/common/fs/vfs/vfs_vnode.c:882
#17 0xe000000000926030:1 in vn_open+0x5b1 ()
at /ux/core/kern/common/fs/vfs/vfs_vnode.c:465
#18 0xe000000000951c40:0 in inline copen+0xfe ()
at /ux/core/kern/common/fs/vfs/vfs_scalls.c:693
#19 0xe000000000951b40:2 in open+0x82 ()
at /ux/core/kern/common/fs/vfs/vfs_scalls.c:606
#20 0xe00000000088be50:0 in syscall+0x570 ()
at /ux/core/kern/common/svc/scall/syscall.c:1731
#21 0xe00000010868c480 in ?? ()

Is this a known problem for which patches are available ?

I never see any hangs with 11.23 clients mounting off the same NFS server.

[Dave Olker]

Hi,

The problem here is an interaction between NFS and the Unified File Cache, which is new for 11.31. This explains why the problem doesn't exist for 11.23 clients, as UFC didn't exist on 11.23.

This appears to be similar to a known problem on 11.31. I'll check with the lab to see what the current status is of the problem and post an update when I hear back from them.

Regards,

Dave

[Tanuj Vohra]

I also took a TOC dump. In case you guys need
it, just let me know the place where I can upload to.

I am also OK with scrounging around with kwdb in case you need more info.

[Dave Olker · Accepted solution]

Hi Tanuj,

I researched this further and identified a defect in our database that is very similar to the one you're reporting. The known problem I've found is fixed in ONCPlus 11.31.02.

My suggestion to you would be to install the 11.31.02 version of ONCPlus on your 11.31 system and see if the problem is resolved.

You can obtain this ISU update at the following website:

http://h20392.www2.hp.com/portal/swdepot/displayProductInfo.do?productNumber=ONCplus

Please try this and let me know if it resolves your problem.

Regards,

Dave

[Tanuj Vohra]

Bingo. It fixed the problem.

[Tanuj Vohra]

new ONC depot fixes it.