"TIMEOUT - READ_DMA48 retrying" has suddenly started appearing (RELENG_6)

[Pertti Kosunen]

Daniel Eriksson wrote:
> I'm running RELENG_6 on two machines, one SMP and one UP (both using SCHED_4BSD). I cvsup/update the OS a few times a week, and after an update on Saturday (24/9) I've seen multiple "TIMEOUT - READ_DMA48 retrying" errors on both machines, causing all sorts of trouble (broken arrays, panics, ...). After a reboot the disks have come back to life and seem to work just fine.
> 
> It could just be a coincidence and multiple disks on two separate machines have started to go bad, but since the disks are different models on different ATA/SATA controllers in different machines it could also be an indication that something has happened with FBSD. If it's a FBSD problem then the "offending" code was added some time between Monday 19/9 and Saturday 24/9.
> 
> Is anyone else experiencing similar problems?

I have had same problems with >UDMA100 since 5.3-RELEASE, UDMA66 works 
perfectly. Machine boots ok with higher modes, but some disk load 
triggers the timeouts. Problems just revisited after upgrade to 
6.0-BETA5 because atacontrol mode syntax changed.

cat /etc/rc.local
#!/bin/sh
atacontrol mode ad0 UDMA66
atacontrol mode ad4 UDMA66
atacontrol mode ad5 UDMA66
atacontrol mode ad6 UDMA66