kernel panics with htb on centos5

kernel panics with htb on centos5

[Mike Schmidt]

HI,

We have a number of production systems running partly as a large
router, using a few variants of centos5[32bit] with IMQ (using the
clearos kernel). These systems are generally loaded with a large
number of clients 500-2000+) and provide bandwidth management in both
directions via IMQ (or in one case ifb) devices. Most clients have
individual htb leaf classes, so there are a lot of htb classes. The
problem is that we have a few of these systems that kernel panic in
exactly the same way:

I only have photographs of the console, but all the kernel panics
(with different stack traces) end up at:

__qdisc_run+0x1f8/0x1d0 SS:ESP 0068:c0754fb8
<0Kernel panic - not syncing: Fatal exception in interrupt

This has happenned on the original Clearos kernel (2.6.18-6), then
2.6.18-308 on both the standard Centos 5.8 kernel with ifb, or the
centosplus kernel.  I can provide if anyone is interested the
photographs of the console showing the details of the kernel panic.

We have been unable to duplicate the problem in house, even with an
emulator bomabrding the test system with tc rule changes and traffic.

Searches have provided some hints, especially that this seems related
to htb, but they also seem to indicate that a very similar problem
others have encountered is resolved in 2.6.18-164-11, which does not
seem to be our case, since we crashed with a standard 2.6.18-308 from
the centos repo.

Does anyone have any idea as to what this may be, or how we can go
about determining more precisely what's going oon?


Thanks a lot,

Mike


--
Mike SCHMIDT
CTO
Intello Technologies Inc.
mike.schmidt@intello.com
Canada: 1-888-404-6261 x320
USA: 1-888-404-6268 x320
Mobile: 514-409-6898
www.intello.com