Re: Playing with SATA NCQ

[Jens Axboe <axboe@suse.de>]

On Thu, May 26 2005, Jeff Garzik wrote:
> Jens Axboe wrote:
> >Re-reading AHCI spec, it does indicate that you want to set SActive
> >after building the command. I'll move it back, but keep the conditional
> >of setting SActive on queued commands.
> 
> SActive is intentionally used for non-NCQ devices.  The SATA registers 
> are -host- registers not -device- registers, remember.

But the host sets SActive only, the device clears it. And at least the
maxtor drives don't clear SActive on non-NCQ commands, which makes
things really confusing once you have completed a non-NCQ command and
start doing some NCQ ones. Page 59 of the AHCI 1.1 spec reads:

4. If it is a queued command, software shall first set
PxSACT.DS(pFreeSlot). [...]

So I really do think this is an error in ahci.c since the beginning.

> At the very least, I would like to see a lot of testing before you make 
> the current unconditional code conditional.

Perhaps you could include such a patch in libata-dev for a while, if you
wish? I really cannot remove it from the NCQ patch.

-- 
Jens Axboe