MFC of "Large set of CAM improvements" breaks I/O to Adaptec 29160 SCSI controller

[Scott Long]

On Apr 27, 2010, at 2:08 PM, Andy Farkas wrote:
> On Wed, Apr 28, 2010 at 12:19 AM, Matthew Jacob <mj at feral.com> wrote:
> 
>> What do you mean by 'deadlock'?
> 
> All disk I/O stops.
> 
> eg. bonnie++ will get stuck in 'getblk' state, or rm will get stuck in
> 'biord' state.
> bufdaemon gets stuck in 'wdrain' state.
> 
> The system will respond to pings but any process requiring disk access hangs.
> 
> Any help is appreciated, thanks.
> 

ACK.  It'll take a little bit to unwind the problem out of that big commit, and I'm consumed with finals and work at the moment.  If someone else doesn't fix it in the next few days, I'm going to propose that the entire commit be reverted.

Scott