Recurring panic

[Gleb Smirnoff]

On Wed, Jun 05, 2013 at 10:18:21AM +0200, Ian FREISLICH wrote:
I> I have the following recurring panic on all my heavily network
I> loaded -CURRENT routers.  The current process is always different.
I> 
I> Gleb, can you please chime in with what you've managed to uncover.

The panics appear on selfd mutex. The mtx_lock value is a free mutex, but
it has 1 extra bit set:

(kgdb) p/x sfp->sf_mtx->mtx_lock
$3 = 0x1000004

Rarely (only one panic observed) more than one bit is set:

$3 = 0x21000004

It is important that selfd mutexes are taken from mtxpool(9), which
is allocated at a early boot stage. Thus, across reboots all possible
sfp->sf_mtx mutexes usually fall into the same virtual memory region.
I'm not sure, but I suppose, they fall into same physical region.

This can lead one to idea that RAM in the box has problems. But it
is running ECC memory, and it doesn't experience other random panics.

The only special about the box is that it is running pf(4) with huge
ruleset and a lot of traffic. So the pf(4) is the number one suspected,
albeit it isn't closely related to selfds.

-- 
Totus tuus, Glebius.