About "panic: bufwrite: buffer is not busy???"

[Andrew Boyer]

On Feb 20, 2011, at 10:46 AM, Jeremy Chadwick wrote:

> On Sun, Feb 20, 2011 at 10:30:52AM -0500, Mike Tancsa wrote:
>> On 2/20/2011 9:33 AM, Andrey Smagin wrote:
>>> On week -current I have same problem, my box paniced every 2-15 min. I resolve problem by next steps - unplug network connectors from 2 intel em (82574L) cards. I think last time that mpd5 related panic, but mpd5 work with another re interface interated on MB. I think it may be em related panic, or em+mpd5.
>> 
>> The latest panic I saw didnt have anything to do with em.  Are you sure
>> your crashes are because of the nic drive ?
> 
> Not to mention, the error string the OP provided (see Subject) is only
> contained in one file: sys/ufs/ffs/ffs_vfsops.c, function
> ffs_bufwrite().  So, that would be some kind of weird filesystem-related
> issue, not NIC-specific.  I have no idea how to debug said problem.
> 

The issue is the file system activity occurring in parallel with the coredump, which is strange.  It seems like everything else should be halted before the dump begins but I couldn't find a place in the code that actually tries to stop the other CPUs.

My question isn't about the initial panic (I was using the sysctl to provoke one), but about the secondary panic.

This is on 8-core systems.

-Andrew

--------------------------------------------------
Andrew Boyer	aboyer at averesystems.com