Bug#317798: kernel-image-2.6.11-1-686-smp: nfs locking can cause a process to hang forever

[Marc Horowitz <marc@mit.edu>]

maximilian attems <debian@sternwelten.at> writes:

>> On Mon, 11 Jul 2005, Marc Horowitz wrote:
>
>> can you try out if newer 2.6.12 fixed that nfs issue?
>> http://charm.itp.tuwien.ac.at/~mattems/
>
>> from the client side, but i guess you better update your server side.

I can't update the server side trivially, since it's a production
fileserver.

I see the same problem with 2.6.12:

evince        D C0391F20     0  3814   3715          3815       (NOTLB)
f750db38 00000086 df9bda20 c0391f20 df87c060 00000000 f7641a00 00000286 
       00000000 d34336c0 000f420f df9bda20 df9bdb44 c194f680 f750db74 c194f70c 
       f750db54 f8c56c7a c194f680 f713de00 00000236 f750c000 00000000 00000000 
Call Trace:
 [<f8c56c7a>] __rpc_execute+0x12a/0x340 [sunrpc]
 [<c012c520>] autoremove_wake_function+0x0/0x60
 [<c012c520>] autoremove_wake_function+0x0/0x60
 [<f8c52730>] rpc_call_sync+0x70/0xb0 [sunrpc]
 [<f8c40d6e>] nlmclnt_call+0x8e/0x1e0 [lockd]
 [<f8c41265>] nlmclnt_lock+0x55/0xf0 [lockd]
 [<c016811a>] locks_copy_lock+0x8a/0x90
 [<f8c40a80>] nlmclnt_proc+0x240/0x330 [lockd]
 [<c0115b60>] default_wake_function+0x0/0x20
 [<c0115b60>] default_wake_function+0x0/0x20
 [<c0115b60>] default_wake_function+0x0/0x20
 [<c0115b60>] default_wake_function+0x0/0x20
 [<f8cb6d63>] do_setlk+0x83/0x1a0 [nfs]
 [<c016a727>] fcntl_setlk+0x2a7/0x300
 [<c0146e68>] vma_link+0x48/0xc0
 [<c0147af6>] do_mmap_pgoff+0x476/0x7b0
 [<c0165f67>] do_fcntl+0xd7/0x190
 [<c0166158>] sys_fcntl64+0xa8/0xc0
 [<c0102f89>] syscall_call+0x7/0xb

                Marc