Hi Owen,

Owen Martin wrote:
 > This looks like a timeout during a read command:
 >
 > ata3.00: cmd c8/00:08:90:3c:59
 >
 > Read dma of 8 blocks from 0x903c59
 >
 > Next time it happens, see if it is the same LBA. Since the drive came
 > back after the bus reset makes me think it was probably in error
 > recovery for an extended amount of time.

Sounds like a good idea. However, I had the drive swapped yesterday and 
have now reinstalled on a (seemingly) identical one which so far seems 
to be free from these messages. Hence, I keep my fingers crossed that 
this was indeed a hw error.

As it was on warranty I was not allowed to keep the bad drive for 
further experiments.

 > Sorry, but I am new to using smartmontools for decoding SMART
 > attributes. Your previous email showed:
 >
 > Device is:        Not in smartctl database [for details use: -P showall]
 >
 > Does that imply the tool will not know the exact meaning of all the
 > attributes? I am not familiar with Fujitsu's implementation.

I believe you are correct.

 >>From the data you sent about the attributes before, it looks like the
 > pending and reallocated sector counts are zero, so the block must have
 > not failed recovery. Can you try to dump the sector using hdparm-8.9 to
 > see if it reproduces?
 >
 > hdparm --read-sector 9452633 /dev/sda

Would if I could... The messages I sent were indeed only from cases 
where the driver succeeded write to the disk in the end (extracts from 
/var/log/messages). In the failure cases I did not make a hard copy.

 > What is the timeout set to?
 >
 > cat /sys/block/sda/device/timeout

30

 > Maybe try to increase that. You want to be sure that it is not a drive
 > issue by verifying the block is readable and the raw values from the
 > pending, uncorrectable or reallocated sector attributes don't change.

Will do if I ever see it again.

			Best / Jonas