SATA diagnoses (+ 2nd drive on Shuttle SD37P2)

[iaw]

dear linux wizards:

I am running a shuttle sd37p2. with one samsung SATA 400gb drive,
the system is rock-steady.

I have two problems.

[1] I want to learn what hard drive speed I am getting, and why my
second identical drive is giving me trouble. on /dev/hda devices,
hdparm -i gave some neat info. sdparm output seems rather spare
in comparison. I also would like to learn more about DMA vs. PIO
modes, and I would like to learn the SMART status of my device. How?

[2] A second hard drive in the system fails. that is, after heavy disk use,
there is suddenly some IO error, then there are suddenly a lot of kernel
messages about errors, and the drive disconnects. the drive itself is the
same as the #1 drive and works perfectly in an external USB enclosure.

so, the most likely problems are now that either the linux SATA driver is
unstable (which would surprise me), or that the sd37p2 has a problem
on the second SATA channel--but this is weird, too. A standard problem
rather than the "after much use" problem is more often software than
hardware. are there any known kernel issues under 2.6.18 x86_64 (gcc-4.1.1) ?

help appreciated.

sincerely,

/ivo welch
_________________
/iaw

[radagast]

in answer to (2),
i have a 120G seagate drive which i was using in a raid array with another identical drive. the array was crashing about once a month, IO error and kernel messages (i could dig them up if you want to compare), and eventually i retired the dodgy drive to a backup, where it spins along quite happily.

if both your drives are SATA, it's very unlikely to be the driver. switching the drive channels is pretty easy too. and to test the kernel you could install another kernel - maybe even a generic x86 one - and run something to see if it crashes overnight.

in my experience though, it was the disk.

[iaw]

thanks. is there any way to learn the smart status or pio/dma usage on the sata channel?
_________________
/iaw

[radagast]

the only success i've ever had with SMART on windows or linux is through the bios. the drive i described in my last post had no warning, even though i can reproduce the failure.
trouble with SMART is, you never know it's working until it tells you it is.

hdparm -tT works fine for sd devices, and that's all you need to compare speeds.

[wynn]

If you run smartd (from smartmontools) then it will post to the kernel log when it starts up and when any critical attribute changes, this includes temperature.

The entries in /etc/smartd.conf here are

[iaw]

great! mille grazie.

regards,

/iaw
_________________
/iaw