On Sat, Feb 20, 2010 at 7:26 PM, Benjamin S. <sbenni@xxxxxx> wrote:
>
>> > Does somebody have an idea what I can test before I have to bisect
>> > it?
>>
>> Could be some kind of block layer or IO scheduler behavior change.. you
>> could try testing after doing:
>>
>> echo noop > /sys/block/sda/queue/scheduler
>>
>> or
>>
>> echo deadline > /sys/block/sda/queue/scheduler
>
> Thanks for your reply. I have already tested both schedulers.
> Unfortunately it does not make any difference which of them is
> used.
It could be a problem with the recently added FPDMA auto-activate
support on AHCI and this drive. It basically allows two messages from
the drive during DMA transfer setup to be combined into one in some
cases. Your drive indicates support for it, and it seems to be
working, but it could be it's tickling some firmware bug that kills
the performance.
>From looking at the Super Talent web site it looks like 1819 is
already the latest firmware available. My next suggestion would be to
try changing this line in drivers/ata/ahci.c:
pi.flags |= ATA_FLAG_NCQ | ATA_FLAG_FPDMA_AA;
to
pi.flags |= ATA_FLAG_NCQ;
and rebuild, and see if you see the same performance as before. If it
helps,we can create a proper blacklist entry for FPDMA AA on that
drive model/firmware version. In this case, reporting the issue to
Super Talent might be a good idea, since this is likely a firmware bug
of some type..
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
