Message-Id: <1180230362.23126.18.camel@obelisk.thedillows.org>
Date:	Sat, 26 May 2007 21:46:02 -0400
From:	Dave Dillow <dave@...dillows.org>
To:	Jeff Garzik <jeff@...zik.org>
Cc:	dean gaudet <dean@...tic.org>, linux-ide@...r.kernel.org,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH, RFT, v4] sata_mv: convert to new EH

On Sat, 2007-05-26 at 17:38 -0400, Jeff Garzik wrote:
> Dave Dillow wrote:
> > It is likely I can come up with at least one, and maybe two if I
> > cannibalize my main machine... do you want results under all three
> > kernels or just the v4 patch?
> 
> Mainly the v4 patch, but an it-works test on 2.6.22-rc3 just for 
> sanity's sake would be useful as well.
> 
> If you have time, check both new and old drives with SMART
> 	smartctl -d ata -t long /dev/blahblah

Whew, this has been fun. Nothing like having your 4yr old complain about
not being able to get to his email... "Daddy's working on it, why don't
you play outside?" :)

So, I added WDC WD1600JS-00N to the mix. This worked much better.

2.6.22-rc3-new.log is the baseline boot -- walking the tree with find
worked well, and I got 61MB/s reading the first 10/11GB or so.

2.6.22-rc3-mv4-new.log is a boot using the new driver, and only has the
WD attached. Again, everything is happy -- no warnings.

2.6.22-rc3-mv4-new2.log -- I put the Maxtor disks back online, and now I
get tons of WARNING: at drivers/ata/sata_mv.c:1287 mv_qc_issue() and
WARNING: at drivers/ata/sata_mv.c:1333 mv_get_crpb_status()

It eventually booted, but I had stopped the console log at that point.
So, enter 2.6.22-rc3-mv4-new3.log wherein I try to reboot and let it
come all the way up to try testing in the presence of the warnings.
However, this time, md0 mounts, but I immediately get I/O errors. I
tried dmesg to see if something got lost, but the log is faithful. This
could be ext3 corruption from previous testing, but there is no record
of problems anywhere. And I've forced fsck since the last known
corruption.

Still, I've not been able to recreate the issue I had with the
corruption before, but I've been turning off the machine each time, so I
try a warm-boot. 2.6.22-rc3-mv4-new4-warm.log gives the now familiar
warnings, and wedges at the end of the log.

I run smart long tests on my drives every two weeks, and short tests
every day using smartd's scheduling options. Just to be sure, I ran the
long tests again, and had no errors. Of course, shortly after that, I
now have had to correct an uncorrectable error on one of the drives, and
am currently resyncing the RAID.

At least Tommy could get his email, if he were still awake... :)

I've seen some chatter on forums about certain versions of the 6L200S0's
firmware having an issue with NCQ and nForce chipsets, but I don't think
it applies to the firmware I have, nor can I seem to find an update on
Maxtor/Seagate's site. Is there a command line option I can give to
disable NCQ, or will I need to play with the blacklist to turn it off?
Do you think that is an avenue worth pursuing?

Also, my main machine has a Maxtor 6Y120M0. It is more of a pain to test
it on the Marvell chipset, but I'd rather not unless you think it will
be worth it.

Dave



View attachment "2.6.22-rc3-mv4-new.log" of type "text/x-log" (17715 bytes)

View attachment "2.6.22-rc3-mv4-new2.log" of type "text/x-log" (56954 bytes)

View attachment "2.6.22-rc3-mv4-new3.log" of type "text/x-log" (20542 bytes)

View attachment "2.6.22-rc3-mv4-new4-warm.log" of type "text/x-log" (23668 bytes)

View attachment "2.6.22-rc3-new.log" of type "text/x-log" (20815 bytes)

Powered by blists - more mailing lists