cvs commit: src/sys/dev/em if_em.c
Eric Anderson
anderson at centtech.com
Tue Aug 22 15:35:21 UTC 2006
On 08/22/06 10:32, Gleb Smirnoff wrote:
> On Tue, Aug 22, 2006 at 10:26:02AM -0500, Eric Anderson wrote:
> E> On 08/22/06 10:23, Gleb Smirnoff wrote:
> E> >On Tue, Aug 22, 2006 at 02:32:48AM +0000, Pyun YongHyeon wrote:
> E> >P> yongari 2006-08-22 02:32:48 UTC
> E> >P>
> E> >P> FreeBSD src repository
> E> >P>
> E> >P> Modified files:
> E> >P> sys/dev/em if_em.c
> E> >P> Log:
> E> >P> It seems that em(4) misses Tx completion interrupts under certain
> E> >P> conditions. The cause of missing Tx completion interrupts comes from
> E> >P> Tx interrupt moderation mechanism(delayed interrupts) or chipset bug.
> E> >P> If Tx interrupt moderation mechanism is the cause of false watchdog
> E> >P> timeout error we should have to fix all device drivers that have Tx
> E> >P> interrupt moderation capability. We may need more investigation
> E> >P> for this issue. Anyway, the fix is the same for both cases.
> E> >P>
> E> >P> This should fix occasional watchdog timeout errors seen on a few
> E> >P> systems.
> E> >P>
> E> >P> Reported by: -net, Patrick M. Hausen < hausen AT punkt DOT de >
> E> >P> Tested by: Patrick M. Hausen < hausen AT punkt DOT de >
> E> >
> E> >This look like a workaround, not a fix the root of the problem. Several
> E> >people on net said that this problem disappears if debug.mpsafenet=0.
> E> >So I think there is a problem in FreeBSD or driver, not in chip.
> E>
> E> And it also worked perfectly for a very very long time until 6.x tree.
> E> I went from 5-STABLE to 6-STABLE and started seeing it a lot ( a few
> E> times per day) on a couple servers.
>
> In 5-STABLE the watchdog events were not logged, but the watchdog was
> processed.
>
But my interface never seemed to disappear (I page if the interface goes
down), and I'd still notice a link up/down, correct?
Eric
--
------------------------------------------------------------------------
Eric Anderson Sr. Systems Administrator Centaur Technology
Anything that works is better than anything that doesn't.
------------------------------------------------------------------------
More information about the cvs-src
mailing list