[PATCH] Re: [openib-general] Re: IPoIB Failure CQ overrun

Tziporet Koren tziporet at mellanox.co.il
Thu Dec 23 01:07:14 PST 2004


Hi,
We found an issue with the FW that causing this overrun.
The bug happened when incrementing CQ consumer index in more then 1 and a
CQE is written at the same time on this CQ.
This bug is the same in 3.3.1 and 4.6.1 versions.

A FW with a fix will be provided next week.

Tziporet

-----Original Message-----
From: Michael S. Tsirkin [mailto:mst at mellanox.co.il]
Sent: Wednesday, December 22, 2004 11:21 PM
To: Roland Dreier
Cc: openib-general at openib.org
Subject: Re: [PATCH] Re: [openib-general] Re: IPoIB Failure CQ overrun


Hello!
Quoting r. Roland Dreier (roland at topspin.com) "Re: [PATCH] Re:
[openib-general] Re: IPoIB Failure CQ overrun":
>     Roland> Testing on PCI Express/Arbel systems (with dual 3.2 GHz
>     Roland> Xeons) with FW 4.5.3, I've seen somewhat different
>     Roland> behavior.  Even after setting IPOIB_NUM_WC to 1 so that
>     Roland> IPoIB never polls more than 1 CQE at a time (and so we
>     Roland> always inc the CI by exactly 1), I still get the CQ
>     Roland> overrun.
> 
> FW 4.6.1 on these systems behaves the way Tavor FW 3.3.1 did on my
> systems: if I modify IPoIB to always poll 1 CQE at a time, it works
> fine.  If IPoIB polls more than 1 CQE at a time then we see an
> overrun.
> 
>  - Roland

Thanks, Roland!
I'll be working on it.
mst
_______________________________________________
openib-general mailing list
openib-general at openib.org
http://openib.org/mailman/listinfo/openib-general

To unsubscribe, please visit
http://openib.org/mailman/listinfo/openib-general
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20041223/11d77388/attachment.html>


More information about the general mailing list