[altq 1233] Starfire driver problems with Adaptec Quad Cards under FreeBSD (fwd)

[Nicolas Christin <nicolas@cs.virginia.edu>]

	Hi all,

	I'm having some problems with the sf driver (Starfire).  I
contacted the author of the driver (see below) but did not get any
response. Anybody interested in helping? I thought I'd just post this here
since I guess most people able to fix this are browsing this list anyways.
Sorry if it's not directly ALTQ-related (even though it could be, if
someone was to propose a patch I guess it could be included...).

	--Nick

P.S.: Kenjiro, is there any chance you could set up a filter on your
mailserver that prevents that annoying spammer (the guy who spammed
3-4 times in the past couple of days) from staying in business?

---------- Forwarded message ----------
Date: Thu, 01 Nov 2001 15:03:55 -0500
From: Nicolas Christin <nicolas@cs.virginia.edu>
To: wpaul@ctr.columbia.edu
Subject: Starfire driver problems with Adaptec Quad Cards under FreeBSD

	Dear Bill,

	I am contacting you because you appear to be the author of the Starfire
(sf) driver for FreeBSD. (I found your email address in
/usr/src/sys/pci/if_sf.c) I am experiencing some problems with this
driver, which I use on the following quad card:

Adaptec ANA-62044 SGLQuartet 64  10/100Base TX

	The driver works just fine most of the time, I can transmit/receive data
on all four ports. However, when the link governed by the sf driver gets
overloaded, I sometimes get the following error:

nc2y@tin </home/nc2y>$ ping 10.0.4.2
PING 10.0.4.2 (10.0.4.2): 56 data bytes
ping: sendto: No buffer space available

where sf0 is configured as follows:

nc2y@tin </home/nc2y>$ ifconfig sf0
sf0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
         inet 10.0.4.1 netmask 0xffffff00 broadcast 10.0.4.255
         inet6 fe80::200:d1ff:feee:3ff1%sf0 prefixlen 64 scopeid 0x3
         ether 00:00:d1:ee:3f:f1
         media: 100baseTX <full-duplex> status: active
         supported media: autoselect 100baseTX <full-duplex> 100baseTX
10baseT/UTP <full-duplex> 10baseT/UTP none

Interestingly enough, when pinging 10.0.4.2 from 10.0.4.1, I do not get
any error, and ICMP replies actually manage to make it.

This experiment was run on an Intel Pentium III running FreeBSD 4.3. (I
don't think there are any changes to the sf driver in FreeBSD 4.4.)

	I would be extremely interested in getting this problem fixed,
since I use these Adaptec quad-cards quite frequently. I have some
experience doing kernel-level development (I recently implemented a new
queuing discipline for ALTQ as part of my PhD research), but none
regarding device drivers development, which is why I haven't yet tried to
dig in your source code. However, if you are willing to try to address
this driver problem, I can run some tests here (I have a testbed of 6 such
PC's available) and help you as much as I can.

	Best regards,
	--Nick