|
Home > Archive > Linux Debian support > September 2007 > Random read errors on Samsung SATA disks via SIL3114 (sata_sil)
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Random read errors on Samsung SATA disks via SIL3114 (sata_sil)
|
|
| Daniel Buus 2007-09-06, 1:14 pm |
| Hey there 
I'm trying to set up this (mainly) file serving server with Kubuntu
7.04. It's pretty old hardware, an Abit KT7-RAID mainboard with just
384 MB VMC memory, a Matrox G200 AGP card, a 100Mb ether, a 1Gb ether,
an old sound card, and three new PCI SATA controllers equipped with
the SIL3114 chip.
There are 12 500GB Samsung T166 Spinpoint drives attached to the
cards, and if I didn't know better, I'd think everything was working
just fine. I can format, write, read, no errors reported, only thing
is, the data being read back is not the same as was written. Actually,
I don't even know how the data that was written is stored, as the data
being read back is apparently different on each read. I say this,
because I tried creating an ISO image of a DVD on a different
computer, then created an sfv checksum file for it, then wrote both to
the each SATA drive on the server, and started checking the read
results. Here're results of two sequential runs on the same file, as
reported by cfv:
D0001.iso : crc does not match (08c5fb69!=ea27237c)
D0001.iso : crc does not match (08c5fb69!=d9f280e5)
I also have two 120 GB Maxtor ATA drives, set up in mirror MDs, and if
I copy the image to here, no corruption occurs. Only on the SATA
drives.
The Samsungs tell you to use a utility to choose a slower transfer
mode if your controller does not support SATA2, which mine don't, so I
ran the utility and set the transfer rate to UDMA6, or "something
133" (forget what it said exactly). This made no difference.
I've also tried rmmod'ing sata_sil, then insmod'ing it with the
slow_down option set to on, makes no difference.
Just now I tried disabling PCI burst in the BIOS, no difference
either.
I'm willing to try a lot of things here, I'd just like some good
ideas, or a "hey stupid, you forgot to..." if one such exists That
would be nice
So have any of you experienced something like this?
I've tried different formattings of the drives, the first go I created
reisers with 1k blocks, which has proved buggy for me in the past, so
my first fix attempt was the reformat with 4k blocks, which didn't
help. Then I tried ext3 with 4k blocks, but that didn't help either.
Hope someone has some good tips
Thanks,
Daniel
| |
| Daniel Buus 2007-09-11, 7:11 am |
| Okay, it seems that this issue is very hardware related indeed. I
threw in another ATA disk and put XP on it, tried there, and the same
pattern of corruption appeared. I then tried different setups with the
BIOS, relocation of PCI cards, removing any card that I could possibly
start the system without, even tried out different versions of the
BIOS, I've been able to make the problem worse (fewer errors), but
I've been unable to fix the issue entirely.
The Abit KT7-RAID has never been the most stable board IME, guess it
was a bad idea to hope for it to support this setup  Throwing in the
towel.
Daniel
|
|
|
|
|