Subject: Re: pcmcia ep0 performance awful in 1.6/-current
To: Nathan J. Williams <nathanw@wasabisystems.com>
From: Greywolf <greywolf@starwolf.com>
List: current-users
Date: 07/09/2002 13:26:29
This might be related somehow, I don't know.
OS: NetBSD/sparc 1.5ZC
Hardware: SPARCstation 5/170MHz.
Networking: le0 [10BT]
be0 [100TXd]
hme0 [100TX]
Comment: Gods, I love my hme card.
Other: Hawking 8-port N-way switch [very small switch].
Netopia 8-port 10BT R3100-I
Results as soon as I switched to the hawking from using the Netopia's
built-in 10baseT hub:
NFS default size NFS [rw]size=8192
le0 Stopped working Worked fine
be0 Does not work at all Does not work at all - no networking
hme0 Works fine Works fine
I'm wondering what was wrong with le0 and the 32k I/O size. Watching
the traffic showed a VERY SLOW snd/ack sequence. It was on the order
of SND/ACK...[wait about 30 seconds]ACK...[wait ...]ACK [received! yay!]
SND/ACK...[wait...]ACK...
and like that. Needless to say, I was less than thrilled at having to
reboot my box because at 1.5k/minute, I wasn't going to be able to save
that 1.2MB file any time soon...
I have NO clue as to what was up.
On 9 Jul 2002, Nathan J. Williams wrote:
# Date: 09 Jul 2002 12:02:25 -0400
# From: Nathan J. Williams <nathanw@wasabisystems.com>
# To: Ben Harris <bjh21@netbsd.org>
# Cc: current-users@netbsd.org
# Subject: Re: pcmcia ep0 performance awful in 1.6/-current
#
# Ben Harris <bjh21@netbsd.org> writes:
#
# > Hmm. What does "netstat -p ip" have to say about incoming fragments
# > (especially dropped after timeout vs received)? It sounds as if the network
# > card is losing fragments, which usually means its receive buffer is too
# > small.
#
# 8747 fragments received, 0 fragments dropped (with -r and -w set to
# 4096). I don't want to wedge the machine remotely to test the other
# configuration, so that'll have to wait.
#
# >
# > >Server is a 1.5ZC Alpha pc164 with a tlp0 set to 100baseTX-FDX, and
# > >there's a little 5-port 10/100 switch between them.
# >
# > What speed is ep0 running at? If it's at 10 Mbit/s, I wonder how much
# > buffering the switch has.
#
# Yes, it's running at 10 Mbit/s, half-duplex. The switch is a GigaFast
# EZ500-S, if that tells you anything.
#
# - Nathan
#
--*greywolf;
--
NetBSD: Power Your Net.