Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 16 Nov 1997 13:14:43 -0800
From:      David Greenman <dg@root.com>
To:        Mark Mayo <mark@vmunix.com>
Cc:        hackers@FreeBSD.ORG
Subject:   Re: de underflow errors. huh? 
Message-ID:  <199711162114.NAA21266@implode.root.com>
In-Reply-To: Your message of "Sun, 16 Nov 1997 14:59:10 EST." <19971116145910.31953@vmunix.com> 

next in thread | previous in thread | raw e-mail | index | archive | help
>On Sun, Nov 16, 1997 at 12:45:44AM -0800, David Greenman wrote:
> >
>> >de0: abnormal interrupt: transmit underflow
>> >
>> >I have no clue what this means. Packets seem to be flowing through
>> >the interface nicely, and there is no noticeable packet loss.
>> >If anyone has any ideas what could be causing this, or if I should
>> >give a hoot, please let me know.
>> 
>>    It's indicating that the PCI bus was sufficiently busy enough to cause
>> the transmit DMA to be stalled too long. The proper response to this condition
>> in the device driver it to dynamically increase the transmit threshold (the
>> number of bytes that are DMA'ed onto the card before the transmission is
>> actually started on the wire). The fxp driver does this, the de driver
>> doesn't. I think Matt Thomas might have fixed this in a later rev., but I'm
>> not sure about that.
>>    Anyway, other than the annoying console messages and perhaps a packet
>> drop when it happens, the problem can be ignored.
>
>Do you think increasing the bus speed will make a difference? This machine
>is a P54C 150 - I originally wanted to run it at 75MHz * 2, but the
>dealer gave me one of the aftermarket Intel CPUs with the integrated
>heat sink, and the machine just plain won't make it past the memory
>check when I run the bus at 75... I've done this (75*2) setup on several
>other machines with no problems - but they had the OEM'ed version of the
>CPU with no heatsink, and a grey bottom. I'm guessing that this black
>bottomed jobby is preventing itself from being "overclocked" somehow.
>I may try and get a replacement CPU..  But before I do that do you think
>the jump from 60MHz to 75MHz will help the de card?

   I don't know.

> Or is the 100Mb fxp
>simply delivering data too quickly for it to handle. Perhaps replacing
>the de with a fxp...  

   That would get rid of the problem. You didn't mention which chip was on the
de card, but if it is the original 21140, then that might also be a problem -
I seem to recall that it didn't do the memory-read-multiple PCI operation,
resulting in very poor PCI performance.

-DG

David Greenman
Core-team/Principal Architect, The FreeBSD Project



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199711162114.NAA21266>