Hello,
With the latest RHEL 5.2 installed on one of the machines, I noticed a
NFS/RPC hang. Looking at
the stack of all the process, I noticed the following wait by rpciod thread.
PID: 7672 TASK: ed3bd550 CPU: 0 COMMAND: "rpciod/0"
#0 [cc9c8e1c] schedule at c05facb9
#1 [cc9c8e84] rpc_wait_bit_interruptible at f8cff9ed
#2 [cc9c8e88] __wait_on_bit at c05fb52d
#3 [cc9c8ea0] out_of_line_wait_on_bit at c05fb5b2
#4 [cc9c8ed8] __rpc_wait_for_completion_task at f8cffa16
#5 [cc9c8ee4] nfs_complete_unlink at f90a872b
#6 [cc9c8efc] nfs_dentry_iput at f90a0e0c
#7 [cc9c8f08] dentry_iput at c047de77
#8 [cc9c8f14] dput at c047e996
#9 [cc9c8f20] put_nfs_open_context at f90a2bad
#10 [cc9c8f38] nfs_writeback_done_full at f90aa450
#11 [cc9c8f60] __rpc_execute at f8cffe87
#12 [cc9c8f8c] worker_thread at c04329bb
#13 [cc9c8fcc] kthread at c0434d97
#14 [cc9c8fe4] kernel_thread_helper at c0404c39
This was never an issue earlier and looking at 5.1 nfs_complete_unlink() source:
void
nfs_complete_unlink(struct dentry *dentry)
{
. . . .
dentry->d_flags &= ~DCACHE_NFSFS_RENAMED;
spin_unlock(&dentry->d_lock);
rpc_wake_up_task(&data->task);
nfs_put_unlinkdata(data);
}
So, it seems like __rpc_wait_for_completion_task() called is added to
nfs_comlete_unlink() in 5.2
Also looking at the RHEL 5.2 announcement, the following bug seems to
have pulled this into 5.2
https://bugzilla.redhat.com/show_bug.cgi?id=253663
The only way to resolve this is by rebooting the node, I don't have a
reproduction scenario, but it
hanged couple of time in last 2 days.
Any comments on how we can fix this now ?
Thanks,
- vnr