Re: USB bugs in 2.6.11.8, 2.6.11.10 and 2.6.12.2

[mjt@nysv.org (Markus Törnqvist)]

(Gotta love replying to oneself ;)

On Thu, Jul 07, 2005 at 12:22:58PM +0300, Markus   Törnqvist wrote:
>Hi!
>
>I got some weird badness on two Dell PowerEdge 2850 (64-bit SMP Xeon)
>boxes. One of them is all identical with one that runs ok, so I
>attached that one's ksymoops information in this message.
>
>The one that works just fine was installed locally (iirc) so this
>may be an issue with Dell's Remote Administration Console's
>VGA console redirection and its emulation of USB for input, maybe?
>
>My first guess was just hardware failure, but it's a bit unlikely
>on two boxes, especially now that I installed the boxes with a 32-bit
>kernel and everything's fine.
>
>Needless to say both I and the customer would want the 64-bit
>one working :)
>
>The boxes are situated some 3000km from here so getting hard
>forensics is kind of hard, but the Dell system can do a bit,
>so please let me know ASAP if there's something more I can
>try out.
>
>Any ideas on why box number one works but these two don't?
>"Sunspots and lunar positions" is not a good theory ;)
>
>I put up everything I could easily dig out on
>http://mjt.nysv.org/kernelbugfest/
>
>It's just that the boxes don't boot far, as they get stuck in
>that suspend/wakeup loop, so there's only one really complete
>oops decode there from the identical working setup.

I reproduced this with vanilla 2.6.12.2 now, and I'm quoting
the entire message for the benefit of the guys over at
linux-poweredge@dell.com

To make it short, the suspend/wakeup loop didn't occur with 2.6.12.2
vanilla.

Still the bug happens, the box is apparently usable, but not
for the remote administration, so usability is quite shaky
a definition ;)

The relevant stuff is still at the url below, the two new interesting
files are:
http://mjt.nysv.org/kernelbugfest/usb_bug_2.6.12.2.ksymoops
http://mjt.nysv.org/kernelbugfest/netconsole_panic_2.6.12.2
(The netconsole issue is probably not related and I'll look into it
later, if ever, as we're not really using it...)

I also got a hint that disabling power management might help,
but that has to wait until tomorrow.. Anyone want to back
this theory? :)

Thanks!

-- 
mjt

Attachment: signature.asc
Description: Digital signature