6.2 SHOWSTOPPER - em completely unusable on 6.2

Oliver Brandmueller ob at e-Gitt.NET
Wed Sep 27 08:29:41 PDT 2006


Hi Scott,

On Wed, Sep 27, 2006 at 03:16:57AM -0600, Scott Long wrote:
> Well, the best I can say at the moment is, "Wow."  =-(  I guess the 
> thing to do here is to figure out if the problem lies with the em 
> interrupt handler not getting run, or the taskqueue not getting run.
> Since you've stated that it seems to be related to shared interrupts,
> the first possibility is more likely.  However, I'm not sure why the
> symptom would only be showing up now.  The Intel docs say that the
> 82547EI are a bit interesting, and I wonder if assumptions that we
> make about PCI ordering aren't true (or if there are bugs that make
> our assumptions invalid).
> 
> Does this happen after there has been a lot of disk activity, like a
> large tar extraction?  Are you using the SMBus interface at all, or is
> it sitting completely idle?

Disk activity does not trigger the problem, I hammered the disk with 
around 85 MB/s (dd) for about half an hour without seeing any effect. A 
CPU bound thing like a buildworld triggered the problem.

The SMBus Interface is not used at all (it's not even really usable). 
Anyway, as soon as I unload the ichsmb module I cannot triger the 
problem anymore. If I load it again, the problem cann again be triggered 
by a buildworld. Statistical relevance: I did 4 buildworlds, alternating 
the load/unload of ichsmb - both times with ichsmb loaded I saw 3 
watchdog timeouts during the buildworld was running, while ichsmb was 
not loaded I did not see a single watchdog timeout. The use of the 
interface was around the same during all the time (constant NFS traffic 
of around 1-2 MBit/s).

Since we all seem to see this on only the interfaces sharing interrupts 
(as I read the other poster's mails) and the problem can be worked 
around by using polling, it seems to become pretty clear, that it has to 
to with interrupt handling.

The UP/SMP idea seems to be only of interest, because on an UP machine 
it's more likely to share interrupts than on SMP machines, it has 
nothing to do with the fact of UP or SMP itself.

- Oliver


-- 
| Oliver Brandmueller | Offenbacher Str. 1  | Germany       D-14197 Berlin |
| Fon +49-172-3130856 | Fax +49-172-3145027 | WWW:   http://the.addict.de/ |
|               Ich bin das Internet. Sowahr ich Gott helfe.               |
| Eine gewerbliche Nutzung aller enthaltenen Adressen ist nicht gestattet! |
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20060927/2c2f4c2c/attachment.pgp


More information about the freebsd-stable mailing list