From mboxrd@z Thu Jan  1 00:00:00 1970
From: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Subject: Re: libata vs ATAPI
Date: Wed, 13 Oct 2004 01:03:58 +0200
Sender: linux-ide-owner@vger.kernel.org
Message-ID: <58cb370e04101216036013d3fe@mail.gmail.com>
References: <58cb370e04101213215ce9a23c@mail.gmail.com>
	 <416C438C.4010902@pobox.com>
	 <58cb370e0410121417591ecff0@mail.gmail.com>
	 <20041012212225.GA2625@havoc.gtf.org>
	 <58cb370e0410121436749e2bfb@mail.gmail.com>
	 <416C5093.7050304@pobox.com>
Reply-To: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Return-path: <linux-ide-owner@vger.kernel.org>
Received: from rproxy.gmail.com ([64.233.170.207]:31885 "EHLO mproxy.gmail.com")
	by vger.kernel.org with ESMTP id S268045AbUJLXD6 (ORCPT
	<rfc822;linux-ide@vger.kernel.org>); Tue, 12 Oct 2004 19:03:58 -0400
Received: by mproxy.gmail.com with SMTP id 77so394780rnk
        for <linux-ide@vger.kernel.org>; Tue, 12 Oct 2004 16:03:58 -0700 (PDT)
In-Reply-To: <416C5093.7050304@pobox.com>
List-Id: linux-ide@vger.kernel.org
To: Jeff Garzik <jgarzik@pobox.com>
Cc: Linux IDE <linux-ide@vger.kernel.org>, Jens Axboe <axboe@suse.de>

On Tue, 12 Oct 2004 17:45:55 -0400, Jeff Garzik <jgarzik@pobox.com> wrote:
> Bartlomiej Zolnierkiewicz wrote:
> > INQUIRY is done in PIO in ide-scsi and it works for this device
> 
> I wouldn't mind doing non-read/write commands with PIO.  As long as the
> "fast path" commands are done via DMA...

Exactly my POV. 

> >>>>>Unfortunately it doesn't help because atapi_pio_sector() assumes
> >>>>>transfer length to be % SECTOR_SIZE and transfer length for INQUIRY
> >>>>>is mere 36 bytes so this fails.
> >>>>
> >>>>Yes, you are welcome to fix this assumption.  :)
> >>>
> >>>Another bug in this area: bcount is set to 64 * 1024 (0xFFFF) but
> >>>ATA/ATAPI spec defines max bcount as 0xFFFE and there must
> >>>be some reason why ide-scsi limits it to 63 * 1024 and ide-cd to
> >>>32 * 1024.
> >>
> >>bcount should be set to 8K (one SATA FIS) for PIO,
> >>and 0xFFFF for DMA.  It's not used for DMA.
> >
> >
> > According to ATA/ATAPI-5 spec bcount is ignored for
> > non-PIO transfers.
> 
> Right.  In practice for some host controllers it should be set to
> 0xffff.  It looks like I should change atapi_xlat() to do that.
> 
> 
> > It sets bcount also for no-data which seems wrong.
> 
> well some "non-data" commands could still return sense, so I preferred
> to err on the side of caution.
> 
> 
> >>Where is this code in libata?
> >
> > atapi_xlat() in libata-scsi.c
> 
> I do not see code in atapi_xlat() that sets bcount to 0xffff?

My mistake, it is setting bcount to 8K (SATA FIS).  Sorry for noise.

>         Jeff
> 
>