On Thu, 2008-09-11 at 12:26 +0300, Pasi Kärkkäinen wrote:
> Hello!
> 
> I have some IBM HS21-8853 blades, and they work all fine as long as I don't
> plug in IBM Qlogic QMC4052 iSCSI HBA..
> 
> After adding QMC4052 HBA, IRQ routing (?) gets messed up and the USB-based 
> console
> in IBM bladecenter/chassis (keyboard and mouse) don't work very well anymore..
> every keypress gets doubled, and mouse is really wild :) so can't really use
> the console for anything.. Other devices seem to work fine, ie. it _seems_ to 
> only 
> affect the USB console.
> 
> Screenshot of what kernel says during system startup:
> http://pasik.reaktio.net/ibm-hs21-blade-usb-problem-02.jpg
> 
> irq 74: nobody cared (try booting with the "irqpoll" option)
> ..
> Disabling IRQ #74
> 
> This problem gets fixed with the suggested "irqpoll" option to the kernel
> cmdline, ie. console works normally after giving "irqpoll" option to the
> kernel. Although the "nobody cared" error will still happen during bootup.. 
> 
> I'm running RHEL 5.2 (2.6.18-92) atm.
> 
> Any ideas how to fix this? I'm running latest IBM BIOS (v1.13) on the
> Blades.. Is there some option in BIOS that could affect this? 
> 
> Same QMC4052 HBAs work all fine in older IBM HS20-8843 blades..

We have exactly this same problem on the same hardware, IBM 8853 blades
with QMC4052 HBA's, and I've had a support case open with Redhat for
about a month.  So far the only feedback has been to try the latest
kernel version (we were running 2.6.18-92.1.6).  I haven't been able to
try the newest kernel because I've been waiting on a maintenance window
to reboot the system, which should happen this weekend.

I do have a workaround though, if you build the initrd using the
"--without-usb" option of the mkinitrd command, it seems to effectively
work around this bug.  The usb modules will still be loaded during
system initialization, just not until after the initrd is finished.  For
whatever reason, having the USB modules loaded after the qla4xxx module
seems to work around the issue.

You can actually also just add some commands in rc.local to unload and
reload the usb modules and they'll work after that.

Based on my experience the irqpoll option is not really usable in
production as it causes all interrupts to become polled and many drivers
and devices just don't perform very well, or behave very erratically,
with that option

We actually had nearly this identical problem with RHEL4 on some of our
very newest, model 7995 blades, but the most recent kernel did indeed
fix it.  We'll see if that hold for RHEL5 with the most recent kernel
this weekend.

Later,
Tom



_______________________________________________
rhelv5-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-list

Reply via email to