On Wed, 2006-02-15 at 21:44 -0800, Nicholas Miell wrote:

quoted

From time to time, I've been experiencing timeouts with my SATA disk (a

Seagate ST3300831AS) attached to a VIA VT6420 controller on a VIA
K8T800-based MSI motherboard. This has happened somewhat rarely on a
variety of Fedora kernel versions.

Basically, sometimes IO will get really slow and I'll start getting
"ata1: command 0x35 timeout, stat 0x50 host_stat 0x4" in my logs, with
the occasional "ata1: command 0x25 timeout, stat 0x50 host_stat 0x4" and
"ata1: command 0xb0 timeout, stat 0x50 host_stat 0x0" thrown in for good
measure.

The disk (and/or controller) isn't unresponsive -- I can do a smartctl
-a -d ata /dev/sda and it'll complete (eventually), and IO to the disk
continues (very slowly), but it's generally unusable until the next
reboot.

I actually managed to successfully unmount the filesystem and deactivate
the LVM volume group for once, which allowed me to remove and reload
sata_via, libata, sd_mod and scsi_mod.

After the drivers were reloaded, IO to the disk worked fine, which leads
me to think that sata_via and/or libata needs the ability to reset the
hardware when errors occur.

-- 
Nicholas Miell [off-list ref]