Re: (Another?) Seagate / Sil3112a problem...

[Tejun Heo <htejun@gmail.com>]

philippe.grenard wrote:
[--snip--]
> Seagate 7200.8 drive. 
> Well, all went well for a week or two, but then the drive
> began to make strange noises, and i got some weird messages
> from dmesg output...
> I feared a drive failure, so I made a full Seagate diagnoses
> of the disk, but no errors...
> Well, maybe I got bad luck with that drive, so I decided to
> get another one. I took another Seagate, 250Go, 7200.10 this time.
> 
> I put this new Seagate (let's call it S_new, the other being
> S_old) to the first connector of the Sil 3112a chip, and put
> the "old" one on the second connector : thus I have sda for
> S_new, and sdb for S_old...
> 
> What is really surprising, is that i still got issues with
> sda, but none with sdb... so the believed faulty drive is not,
> as i got no dmesg errors from sdb...
> 
> thus i suspect either a faulty controller, or a problem with
> the driver (sata_sil) i use...(or even something with IRQ as I
> don't understand anything with IRQ...)
> I tried to put both disks numbers in the "blacklist" in
> sata_sil.c, but apart from a degraded speed, it didn't do
> anything...

Don't do that.  It has nothing to do with your problem.

> The other observation I made, was that these problems happens
> only when the computer is still "cold" : I mean, after an hour
> or two, no problem with this... and even if i reboot (I really
> mean reboot, not halt and restart : when the power still turns
> on), i got no problem...
> 
> Well since I use my computer for Desktop, it really is an
> issue for me at the moment, especially when the disk is making
> a noise...
> 
> I'm on Linux for about 2 or 3 years now, under Debian/SID,
> with kernel 2.6.17.13 from kernel.org (self-compiled)
> 
> Here is the output of dmesg if it helps...
> I can provide any information you would find useful, even make
> some tests, but if you could be not too technical, that would
> really be great, as I'm a real noob with Hardware problems...
> 
> Any help/links/infos/hints would really be appreciated!
> I've already googled a lot and found that Seagate/sil3112a is
> a problematic couple, but i didn't find any solution for that...
> I'll try this evening with an older kernel (if you have any
> suggestion for a kernel version...) to see if it's not related
> to a kernel upgrade...

All recent Seagate drives work fine w/ sil3112.  Only some old models of 
.7 drives are problematic.

> EXT3-fs: unable to read superblock
> ata1: command 0xec timeout, stat 0xd0 host_stat 0x0
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ
> 0xb/47/00
> ata1: status=0xd0 { Busy }
> ata1: command 0xec timeout, stat 0xd0 host_stat 0x0
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ
> 0xb/47/00
> ata1: status=0xd0 { Busy }
> ata1: command 0xb0 timeout, stat 0xd0 host_stat 0x0
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ
> 0xb/47/00
> ata1: status=0xd0 { Busy }
> ata1: command 0xb0 timeout, stat 0xd0 host_stat 0x0
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ
> 0xb/47/00
> ata1: status=0xd0 { Busy }
> ata1: command 0xec timeout, stat 0xd0 host_stat 0x0
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ
> 0xb/47/00
> ata1: status=0xd0 { Busy }
> ata1: command 0xec timeout, stat 0xd0 host_stat 0x0
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ
> 0xb/47/00
> ata1: status=0xd0 { Busy }
> ata1: command 0xb0 timeout, stat 0xd0 host_stat 0x0
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ
> 0xb/47/00
> ata1: status=0xd0 { Busy }
> ata1: command 0xb0 timeout, stat 0xd0 host_stat 0x0
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ
> 0xb/47/00
> ata1: status=0xd0 { Busy }

You'll probably get better result with 2.6.18-rc7 which includes new 
libata error handling and more detailed diagnostic messages.  However, 
your problem seems to be hardware issue.  Probably the controller or the 
cable.

I think your link is just flaky enough to cause problem from time to 
time and even when it does it oscillates between link established and 
broken.  Such condition can cause rapid continuous hardresets to the 
attached device, which, depending on drive, can result in weird noise.

So, several things to try...

* make sure the cables are well-plugged

* if you have a spare, swap cables.  if not, swap the first and second 
cable and see what happens.

* if the problem still persists && it doesn't follow the cable (ie. you 
swapped the first and second cable but the first port still fails), your 
controller or circuits on board is the problem.  Buy a add-on card, it's 
really cheap these days.

-- 
tejun