kern/64556: [sis] if_sis short cable fix problems with NetGear
volker at vwsoft.com
Mon Feb 18 20:40:03 UTC 2008
The following reply was made to PR kern/64556; it has been noted by GNATS.
From: Volker <volker at vwsoft.com>
To: Thomas Hurst <tom at hur.st>
Cc: bug-followup at FreeBSD.org
Subject: Re: kern/64556: [sis] if_sis short cable fix problems with NetGear
Date: Mon, 18 Feb 2008 21:35:24 +0100
On 02/18/08 15:07, Thomas Hurst wrote:
> Testing with Linux shows that receive seems to be throttled during
> bidirectional activity. Unidirectional sees 93Mbps in either direction,
> bidirectional sees 93Mbps sending and 35-60Mbps (50 average) receiving.
> There's no sign of connection dropping in an hour of testing.
> Linux is spewing APIC errors at me, not sure what, if anything, they're
> related to.
> A quick glance at the FreeBSD driver shows a DELAY(100000); which
> probably accounts for the performance drops; each burst of 'short cable
> fix' messages basically leaves the card idle for 300ms or so. The Linux
> driver (drivers/net/natsemi.c) has no sign of such a delay.
well, I'll leave it for the net-team to check if this DELAY can be
shortened (it's 100 msec delay).
> This all happens during card setup, sis_initl(), shouldn't this be only
> happening once when it aquires the link, not randomly during operation?
sis_initl is being called not just for initializing the card once, but
whenever RX errors or state changes are detected.
I'm wondering if you can give us the following information as I think
this will be needed for the net folks to further analyze your problem:
and the following when a bunch of network traffic has been generated:
Also just a wild guess as the problems might be related: PR kern/112179
http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/112179 contains a patch,
can you try if you see any difference with that patch applied?
Thanks a lot!
More information about the freebsd-bugs