Re: stuck in nfsfsync

[Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>]

:Matthew Dillon wrote:
:
:> 
:> :Same result with OpenBSD on the server. tar quickly hangs in nfsfsync
:> :when I untar the heimdal distribution in nfs on the DragonFly client.
:> :
:> :        -Richard
:> 
:>     Hmm.  This might indicate that vp->v_numoutput tracking is getting
:>     hosed. If you can get a kernel crash dump while the system is stuck in
:>     this state (e.g. ctl-alt-esc + "panic" + return + return) I think I
:>     should be able to track down the problem quickly.  Upload it to your
:>     leaf  account.
:
:Any progress with that? It still doesn't work for me with a HEAD kernel built 
:on 20/1. I've got a tcpdump during a stuck cp -R and a kernel dump if that 
:would help. Let me know if there's anything else I should provide.
:
:Aggelos

    I haven't been able to find any problem in the codebase and I can't
    reproduce the problem here.

    What I need from you is to use a UDP mounted NFS filesystem and do a
    tcpdump on both the client and server machines showing the point 
    where it gets stuck (plus some time before that and packet traffic for
    at least 60 seconds after it gets stuck too).

    Right now my best guess is that it is an MTU issue or a
    maximum-packet-size issue on one of the machines or that your network
    interface simply can't handle a whole lot of IP fragments coming in at
    once.  The tcpdump should make things clearer.

					-Matt
					Matthew Dillon 
					<dillon@xxxxxxxxxxxxx>