if_igb crashes system

[matt]

bringing up igb0 is currently causing my 10-CURRENT box to become very
non-responsive (as though downclocked to some 100 KHZ...very, very, very
slow).

cmdwatch 'vmstat -i' shows the interrupts get assigned for igb0, and
about 1 second later the machine is essentially non-responsive.

anyone else seeing if_igb issues suddenly?
the machine is running off a cvsup from about 75 minutes ago.

I was only able to access the system after removing interface configs
from rc.conf, which made me suspect polling initially, however even
issuing "ifconfig igb0 up" with no media causes the issue to appear.

Let me know what additional diagnostic data may help?

Matt

_______________________________________________
freebsd...@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-curre...@freebsd.org"

[matt]

On 03/01/12 21:24, matt wrote:
> bringing up igb0 is currently causing my 10-CURRENT box to become very
> non-responsive (as though downclocked to some 100 KHZ...very, very, very
> slow).
>
> cmdwatch 'vmstat -i' shows the interrupts get assigned for igb0, and
> about 1 second later the machine is essentially non-responsive.
>
> anyone else seeing if_igb issues suddenly?
> the machine is running off a cvsup from about 75 minutes ago.
>
> I was only able to access the system after removing interface configs
> from rc.conf, which made me suspect polling initially, however even
> issuing "ifconfig igb0 up" with no media causes the issue to appear.
>
> Let me know what additional diagnostic data may help?
>
> Matt
>

Last time this machine was updated was last week.
I'm going to have to install something else to test the hardware, but it
had no problems until first reboot with new kernel.
I have blown away /usr/src and re fetched it...no change.
I've disabled pf and tried to bring it up...no change
I've tried to bring up the other interface in the pair...no change (igb1).
It doesn't look like the msix commit from jhb has anything to do with
it...didn't exactly revert but changed return(0); back to
return(msgs);...no change

I'm down to either the hardware conveniently failed during upgrade, or
some earlier commit.
Luckily I have an em-based port as well.
I'm testing hardware next and then going further back into the week on
the commit log?
Anyone else with igb out there and working with latest HEAD?