Re: Bug in 2.6.29 ide-cd: Kernel freeze: bisected + unacceptable workaround

[Bartlomiej Zolnierkiewicz]

On Tuesday 31 March 2009, Borislav Petkov wrote:
> Hi,
> 
> > > @Bart: can you take at look at this. Somehow, if the device is behind a
> > > PCI IDE controller, the order of issuing the command and enabling DMA is
> > > important. Seen something like that before?
> > 
> > Spec is unclear on the ordering but empirically it seems that some hosts
> > may need packet command to start DMA.  I think that we should proceed with
> > your patch (please repost with patch description) and also apply the same
> > change to the rest of ATAPI devices in a subsequent patch.
> 
> Why two patches, do we really need to differentiate between ide-cd and
> other ATAPI devices? Besides the patch is so simple: 4 lines moved up.

Good engineering practice -- we've changed only ide-cd in 2.6.29 so we
should start with reverting the change and backporting it to -stable.

Fixing other ATAPI devices can be handled later and preferrably after
testing them first -- who knows if there are no devices which depend on
the reverse order of doing things or that it won't uncover/trigger some
other problem.  Feel free to call me (over-)paranoid... :)

However based on later comments from Alan and the current libata code
I feel that we should be pretty safe in this particular case so we may
as well fix everything in one go.

> Anyway, I lightly tested it with ide-cd and ide-floppy and both seem to
> take it ok. Also, if I remember correctly, the original ide-cd behavior
> was to issue the command and _then_ start DMA so we're back to that. I
> guess now we should be concerned whether the other ATAPI devices can
> handle the reverse situation where you first issue a command and _then_
> start DMA.. Hmm...
> 
> --
> From: Borislav Petkov <petkovbb@xxxxxxxxx>
> Date: Tue, 31 Mar 2009 09:36:44 +0200
> Subject: [PATCH] ide-atapi: start DMA after issuing a packet command
> MIME-Version: 1.0
> Content-Type: text/plain; charset=utf-8
> Content-Transfer-Encoding: 8bit
> 
> ApparentlyÂ, some ATAPI devices want to see the packet command first
> before enabling DMA otherwise they simply hang indefinitely. Reorder the
> two steps and start DMA only after having issued the command first.
> 
> [1] http://marc.info/?l=linux-kernel&m=123835520317235&w=2

I added "Reported-by:" here

> Signed-off-by: Borislav Petkov <petkovbb@xxxxxxxxx>

applied
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/