> Not fully having understood the issue, I was careless and tried disabling the
> write cache explicitly with "hdparm -W 0 /dev/hda", and ran the bonnie test
> again.
>
> At the first investigation it looks good, at least all the messages go away. But
> there are still phases during write in which the system becomes totally locked
> (not even the mouse cursor can be moved) for a few seconds. This time NOT
> followed by the stated messages.
>
> Re-Enabling write caching (with hdparm -W 1) caused all the messages to appear
> again.
How old is the drive and have you ever gotten "Uncorrectable Error"?
I rethought my earlier statements and they were wrong because I was
thinking of another problem when answering yours.
The problem in reality is that the drive is taking to long to perform the
DMAing with write-cache on/off. I am wondering if your drive is out of
slip sectors and is wildly doing head seeks to grab data from relocated
sectors.
The only way I can think to test this without OEM vender specific calls,
is to run "bonnie" in such a way that it would occur only if we hit a
certain region of the media.
If you have partitioned up the drive as to give some zone tests, it my
allow me to reduce the variables of the problem. It may be that the drive
is going to fail because of running out of relocation-slip sectors.
This is only a WAG!
If it is only that you are having to hit the backup-sector blocks on the
hidden part of the drive, the only solution is to allow for expanded
timeouts. I have to build a counter field that discrimates against the
error type and then adjusts timers accordingly.
What partition are you running bonnie, and how many partitions are there?
Does the drive support SMART?
Andre Hedrick
The Linux ATA/IDE guy
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/