2002-11-27 01:01:00

by Christian Robottom Reis

[permalink] [raw]
Subject: Re: 2.4.19+trond and diskless locking problems

On Wed, Nov 20, 2002 at 06:02:34PM +0100, Trond Myklebust wrote:
> >>>>> " " == Christian Reis <[email protected]> writes:
>
> > I haven't forgotten this. It's just that I've been unable to
> > test: the problem just stopped showing up when I upgraded to
> > 2.4.20-pre11 with your NFS-ALL patches applied to it. Could
> > something have changed, or are we just lucky?
>
> The main changes have been the discovery of a couple of kmap()
> imbalances. Those are also fixed in 2.4.20-rc2.

Just following up. The boxes are all hanging at this very moment on
shutdown. Sample traffic between a client (nachocano) and the server
(anthem) follows. Interesting as every bit of data that goes through
stalls the boxes and makes them want to do ARP lookups, and they do a
lookup one at a time. Very very wierd. Anybody have clues?

22:53:52.565702 nachocano.async.com.br.732 > anthem.async.com.br.685: udp 56 (DF)
22:53:52.566343 anthem.async.com.br.685 > nachocano.async.com.br.732: udp 28 (DF)
22:53:57.564586 arp who-has anthem.async.com.br tell nachocano.async.com.br
22:53:57.564612 arp reply anthem.async.com.br is-at 0:50:4:c:68:f6
22:54:12.563654 nachocano.async.com.br.732 > anthem.async.com.br.685: udp 56 (DF)
22:54:12.563824 anthem.async.com.br.685 > nachocano.async.com.br.732: udp 28 (DF)
22:54:17.559228 arp who-has nachocano.async.com.br tell anthem.async.com.br
22:54:17.559322 arp reply nachocano.async.com.br is-at 0:2:2e:f4:73:31
22:54:32.561607 nachocano.async.com.br.732 > anthem.async.com.br.685: udp 56 (DF)
22:54:32.561788 anthem.async.com.br.685 > nachocano.async.com.br.732: udp 28 (DF)
22:54:52.559555 nachocano.async.com.br.732 > anthem.async.com.br.685: udp 56 (DF)
22:54:52.559721 anthem.async.com.br.685 > nachocano.async.com.br.732: udp 28 (DF)
22:54:57.549884 arp who-has nachocano.async.com.br tell anthem.async.com.br
22:54:57.549977 arp reply nachocano.async.com.br is-at 0:2:2e:f4:73:31
22:55:12.557503 nachocano.async.com.br.732 > anthem.async.com.br.685: udp 56 (DF)
22:55:12.557677 anthem.async.com.br.685 > nachocano.async.com.br.732: udp 28 (DF)
22:55:17.556386 arp who-has anthem.async.com.br tell nachocano.async.com.br
22:55:17.556402 arp reply anthem.async.com.br is-at 0:50:4:c:68:f6
22:55:32.555557 nachocano.async.com.br.733 > anthem.async.com.br.sunrpc: udp 56 (DF)
22:55:32.556037 anthem.async.com.br.sunrpc > nachocano.async.com.br.733: udp 28 (DF)
22:55:32.556350 nachocano.async.com.br.32904 > anthem.async.com.br.685: udp 88 (DF)
22:55:32.556483 anthem.async.com.br.685 > nachocano.async.com.br.32904:
udp 28 (DF)
22:55:37.550543 arp who-has nachocano.async.com.br tell anthem.async.com.br
22:55:37.550669 arp reply nachocano.async.com.br is-at 0:2:2e:f4:73:31
22:55:52.553435 nachocano.async.com.br.734 > anthem.async.com.br.685: udp 56 (DF)
22:55:52.553620 anthem.async.com.br.685 > nachocano.async.com.br.734: udp 28 (DF)
22:55:57.552283 arp who-has anthem.async.com.br tell nachocano.async.com.br
22:55:57.552301 arp reply anthem.async.com.br is-at 0:50:4:c:68:f6
22:56:12.551364 nachocano.async.com.br.734 > anthem.async.com.br.685: udp 56 (DF)
22:56:12.551535 anthem.async.com.br.685 > nachocano.async.com.br.734: udp 28 (DF)

I'm also attaching a full log of the previous tcpdump analysis I did for
the bootup hang.

Take care,
--
Christian Reis, Senior Engineer, Async Open Source, Brazil.
http://async.com.br/~kiko/ | [+55 16] 261 2331 | NMFL


Attachments:
(No filename) (3.34 kB)
violinux-hang (352.00 kB)
Download all attachments