Re: wapbl lockups

[Chuck Silvers <chuq%chuq.com@localhost>]

On Sat, Jul 21, 2012 at 10:59:17AM -0400, Greg Troxel wrote:
> 
> Greg Troxel <gdt%ir.bbn.com@localhost> writes:
> 
> > Date: Wed, 27 Jun 2012 08:18:08 -0400 (3 weeks, 3 days, 2 hours ago)
> > I have seen two machines lock up, and some experimentation implicates
> > WAPBL and USB-connected big disks.  I am curious if anyone else has seen
> > this.
> 
> Perhaps http://gnats.netbsd.org/46325 is related to this.  Opinions?

the changes I made there only change the I/O priority of some I/O requests,
which would at best mask a deadlock problem rather than fix one,
and probably not even that.

in your previous mail, you mentioned that these lockups were of the "tstile"
variety, which I'll guess is the vnode-locking-related kind that were
quite common some time back.  have you analyzed a crash dump from one of these
to see what everything is ultimately waiting for?  ... I guess you won't have
from the "system 2" that you described since that's powerpc and crash dumps
still aren't implemented there.  can you reproduce this with 6.0_BETA2
on a platform where crash dumps work (eg. x86) and?

-Chuck