Re: [PATCH libata/for-3.16-fixes] libata: introduce ata_host->n_tags to avoid oops on SAS controllers

[Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>]

On Wednesday, July 23, 2014 12:36:01 PM Tejun Heo wrote:
> Hello,
> 
> On Wed, Jul 23, 2014 at 06:31:58PM +0200, Bartlomiej Zolnierkiewicz wrote:
> > Hmmm, wasn't ATA_MAX_QUEUE used before not ATA_MAX_QUEUE - 1?
> > 
> > It seems that after your patch the loop in the ata_qc_new() will use
> > only 30 tags and not 31 ones?
> 
> It was always 31 (ATA_MAX_QUEUE - 1) with the tag 31 reserved for EH
> commands.  The previous patch just used limit value which is one too
> high.  I'm planning to change that but this is the way it has always
> been.

I see, thanks for explaining this.

BTW:
		/* the last tag is reserved for internal command. */
		if (tag == ATA_TAG_INTERNAL)
			continue;

in ata_qc_new()'s loop now becomes a dead code (loop will be done
maximum ATA_MAX_QUEUE - 2 times and ATA_TAG_INTERNAL is defined
as ATA_MAX_QUEUE - 1).  Compiler can probably optimize it away so
as a bonus of your patch we may get one condition check less in
a hot-path. :)

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics