Subject: Re: Should Alpha PCI code manage latency timers?
To: None <port-alpha@NetBSD.ORG, tech-kern@NetBSD.ORG>
From: List Mail User <track@Plectere.com>
Date: 01/26/2005 00:38:12
>From firstname.lastname@example.org Wed Jan 26 00:17:07 2005
>Date: Wed, 26 Jan 2005 03:16:13 -0500 (EST)
>Content-Type: text/plain; charset=us-ascii
>[ On Tuesday, January 25, 2005 at 21:26:52 (-0800), List Mail User wrote: ]
>> Subject: Re: Should Alpha PCI code manage latency timers?
>> Here, unlike the Sun Ultra 10 reported before, both of the values for
>> your "Q Logic" cards look incorrect.
>Hmmm.... they're the few things on this machine that actually seem to
>work exactly as expected. (unlike any GigE card I've tried!)
>Though I haven't yet tried to pull data off the FC arrays while pushing
>it out the network at the same time.....
> Greg A. Woods
>H:+1 416 218-0098 W:+1 416 489-5852 x122 VE3TCP RoboHack <email@example.com>
>Planix, Inc. <firstname.lastname@example.org> Secrets of the Weird <email@example.com>
I would expect the Q logic controllers to work since everybody else
gives them "enough" time. I'd simply expect them to interfere with other
devices. Unfortunately most common the expected failure mode for a bad
latency setting is that the "bad" device works, just others may not.
The test would be to see if you can generate a high and continuous load
using the Q Logic FC cards, and see problems with other devices. Likely the GigE
cards would show problems if they were also running full out - even though their
setting look correct, they would miscalculate the required buffer size, and you
would expect underruns to occur - does this explanation match the problems which
you have seen? Your comment about all GigE cards misbehaving would seem to back
this up - Do you know what kind of errors you get on the network cards? I'd bet
on short packets and/or underruns (i.e. the grant is removed before they "expect"
to be complete with the transaction).