Re: Promise SATA300 TX4 card issues

>I've got a ZFS pool setup, and when scrubbing the pool it comes up with
> erratic CKSUM errors (for example, scrub once and get 7 errors, scrub
> again immediately and get 53 more).

I did a buildworld and buildkernel just yesterday with the latest
CURRENT sources and I can confirm that I'm seeing the same problem.
What is interesting is that didn't notice any hard system lockups.

My system is amd64 with 2GB of RAM, four 500GB drives connected to my
Promise SATA300 TX4 card in a raidz1 pool. For testing purposes, I have
created a volume in my zpool, which houses a UFS file system created and
accessed via GELI. I then use samba to copy data to and from the volume
on another machine.

While writing data to my volume, I am getting erratic CKSUM errors being
"discovered" coming from my ZFS raidz1 (a look at zpool status after the
write confirms this), and scrubbing the zpool just seems to discover
more CKSUM errors. "zpool status" is telling me to replace some of my
disks. This never happened before.

I revert my kernel back to the old 200706 snapshot kernel and now the
CKSUM errors are all gone. A fsck on my GELI UFS volume now produces no
errors, and a scrub on my ZFS volume no longer produces CKSUM errors.
The data I wrote to the volume is intact - so no data corruption as far
as I can tell.

It would be great if these issues could be solved for this card, as it
is one of the cheapest 4-port SATA expansion cards out there for
creating a zpool with. *SIGH* In the meantime I am going to have to
stick with the 200706 snapshot kernel and userland, which is stable with
this card.