[mythtv-users] Seagate confirms problems with 1.5TB Barracuda drives

[Roger Heflin]

Martin Bene wrote:
>> I have one SD17 1.5TB, and I added it to a raid5 array, and
>> it built in in the array just fine, and then kicked it out
>> 3-4 hours later on a very odd looking error, and checking
>> the disk for errors showed no errors.
> 
>> Turning off write-cache, and again rebuilding it into the
>> array it has so far lasted 3-4 days without issues, so there
>> appears to be something to the write cache issue.
>> The error I got looked like the drive just did
>> not finish its write and the driver timed it out.
> 
> The typical Seagate 1.5TB error looks like this:
> 
> ata2.00: cmd ea/00:00:00:00:00/00:00:00:00:00/a0 tag 0 cdb 0x0 data 0
>          res 40/00:00:00:4f:c2/00:00:00:00:00/40 Emask 0x4 (timeout)
> 
> The failing command code is EA (Flush Cache Extended). So the drive 
> doesn't actually return an error to a write command, it fails to
> flush the write cache to disk. This also explains why
 > turning off the writecache avoids the issue.

That matches the error I got, after that error and the associated reset I also 
get this error about the sector that it had trouble with before it reset the
drive:

kernel: end_request: I/O error, dev sdd, sector 978534975
kernel: md: super_written gets error=-5, uptodate=0
kernel: raid5: Disk failure on sdd1, disabling device.
kernel: raid5: Operation continuing on 4 devices.

I wonder how many sectors need to be in the write cache for it to trigger?

Assuming that SD17 firmware on a 1TB drive is the same as on a 1.5TB drive,
then there would need parameters outside of the firmware that change the code 
paths of the firmware.   Either the size of the cache, or something else
that exposes the bug.