> umount -f should be just fine. I don't sit around waiting for things so
> you need to do it a few times sometimes but it works.
>
> The rpc cleanup is handled nicely. Olaf designed the layer right to support
> this. EIO is propogated back to all clients
Yup, I'm aware of that. I was thinking more about the NFS asynchronous
writebacks: In the write gathering / NFSv3 case, the requests spend a
certain amount of time in limbo for their sins waiting to be flushed
out. These requests will not be cleared by the rpc_killall_tasks,
since RPC doesn't know about them, and hence they may be a problem if
something decides to flush them, in particular since the RPC client is
likely to be dead...
Seems that stock linux-2.2.x should be more or less OK though, since
there each writeback request is immediately turned into an RPC
request.
Cheers,
Trond
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/