Date: 17 Jul 1998 19:04:52 +0200 From: dag-erli@ifi.uio.no (Dag-Erling Coidan =?iso-8859-1?Q?Sm=F8rgrav?= ) To: HAMADA Naoki <hamada@astec.co.jp> Cc: smoergrd@oslo.geco-prakla.slb.com, mike@smith.net.au, sbabkin@dcn.att.com, hackers@FreeBSD.ORG Subject: Re: NIC drivers Message-ID: <xzppvf4v26j.fsf@hrotti.ifi.uio.no> In-Reply-To: HAMADA Naoki's message of "Fri, 17 Jul 1998 17:41:44 %2B0900 (JST)" References: <199807162021.NAA03930@antipodes.cdrom.com> <rx4lnps9a9m.fsf@oslo.geco-prakla.slb.com> <199807170841.RAA01128@stone.astec.co.jp>
next in thread | previous in thread | raw e-mail | index | archive | help
HAMADA Naoki <hamada@astec.co.jp> writes:
> Dag-Erling Sm,Ax(Brgrav writes:
> > Well, I don't have a -stable box right now, but from preliminary
> > testing on my -current boxes, I'd recommend against it. I haven't had
> > much time to track down the problem, but it would seem that a certain
> > percentage of incoming packets go down a black hole.
> Are the purged packets sane? If that is the case, the top packet
> should not be discarded when an ERR_RX_INCOMPLETE is reported. In my
> environment, an ERR_RX_INCOMPLETE is so rare that I cannot make
> sure...
Scratch that, I think the packet loss is caused by something else. I'm
only getting it on one of my computers, and it works no better without
the patch. My setup is as follows:
helen --- LPIP --- niobe --- Ether --- luna
|
PPP
|
Internet
Needless to say, Niobe is set up as a gateway. Both Niobe and Luna
have 3Com 3c509 NICs. Ssh from Helen or Niobe to Luna doesn't work
reliably; 'ssh -v luna' stops with "Waiting for server public key";
once in a while, it gets all the way to "requesting shell" and hangs
there. If I reboot Luna, it will work for some time after the reboot
(5-10 minutes) then gradually stop working. Ssh between Luna and Niobe
works just fine, 'ping -f luna' from Helen shows no packet loss. Luna
is headless, so I can't experiment much with ssh from Luna to Niobe or
Helen, but yesterday, when I had a monitor plugged in on Luna, I
noticed NFS and NIS were very flaky; not entirely gone, but flaky.
All three computers run -current, and all three had their world remade
yesterday. The problem manifested itself with the first kernel I
compiled with your patch, but since I cvsupped right before I applied
that patch, it may have been caused by something else. I have had
absolutely no network problems between Luna and Niobe (except for
forgetting to ifconfig lo0 on Luna) since I installed Luna on saturday
- in fact, I did a minimal installation from the February snapshot
CD-ROM, then cvsupped the entire tree from Niobe and made world.
Niobe and Luna have IP addresses 10.0.1.1 and 10.0.1.2 respectively;
on the LPIP link, Niobe and Helen have IP addresses 10.0.2.1 and
10.0.2.3 respectively. Niobe also has an (unused) Accton 2107 with IP
address 10.0.3.1. Netmask is 0xffffff00 on all interfaces. Niobe is
NFS, NIS, CVSup, FTP and HTTP server, and HTTP proxy. Luna is NFS and
NIS client. Helen, being a laptop, does not use NIS, but mounts the
repository over NFS from Niobe (not that I ever use it on Helen).
DES
--
One two, one two, one two.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?xzppvf4v26j.fsf>