Re: LibPATA code issues / 2.6.15.4

From: "Eric D. Mudama" <edmudama@gmail.com>
To: "Nicolas Mailhot" <nicolas.mailhot@gmail.com>
Cc: "Mark Lord" <liml@rtr.ca>,
	linux-ide@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: LibPATA code issues / 2.6.15.4
Date: Wed, 1 Mar 2006 18:19:30 -0700	[thread overview]
Message-ID: <311601c90603011719k43af0fbbg889f47d798e22839@mail.gmail.com> (raw)
In-Reply-To: <1141254762.11543.10.camel@rousalka.dyndns.org>

On 3/1/06, Nicolas Mailhot <nicolas.mailhot@gmail.com> wrote:
> Le mercredi 01 mars 2006 à 14:22 -0500, Mark Lord a écrit :
> > Nicolas Mailhot wrote:
> > >>
> > > How about the drives that got blacklisted following :
> > > http://bugzilla.kernel.org/show_bug.cgi?id=5914 ?
> > > and
> > > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=177951 ?
> > >
> > > Device Model:     Maxtor 6L300S0
> > > Firmware Version: BANC1G10
> > >
> > > on Silicon Image, Inc. SiI 3114 [SATALink/SATARaid] Serial ATA Controller (rev 02)
> >
> > Mmm.. somebody with one of those controllers should check
> > to see if *any* drives work with FUA, and blacklist the controller
> > instead of the drives if everything is failing.
>
> I'm a someone with such a controller (that's my boog here)
> But I only have these drives.
> So I can only confirm the combo it deadly.
> (I could possibly try to plug one on the nforce4 controller, not sure if
> extracting the box from the tangle of cables and hardware he's part of
> is worth it. sata_nv is rev-eng, while the siI docs are public, right?)
>
> I do suspect Eric D. Mudama knows if the problem is on the hard-drive
> side though
>
> Regards,
>
> --
> Nicolas Mailhot
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.2.1 (GNU/Linux)
>
> iEYEABECAAYFAkQGKmoACgkQI2bVKDsp8g0veQCggJkweq1nQn7YNSEIobOHitk0
> QXsAn0TnHI/6LBG9nezBnS0MTskLml0W
> =s1TM
> -----END PGP SIGNATURE-----
>

I didn't know offhand so we plugged in a bus analzyer and took a look
here in the lab... We didn't have a 3114 lying around, but issuing the
Write DMA FUA (0x3D) opcode on a 3112 resulted in a D0h soft hang.  I
think they're related (4-port vs 2-port).

Looking at the bus trace, the command is issued on the SATA bus, the
drive generates a DMA Activate FIS which is accepted by the 3112, and
then the 3112 generates a Data Payload FIS (46h) with no contents.

The first DWORD of the payload is a HOLD primitive, to which the
device promptly responds with HOLDA, and the two are in a soft bus
lock and will sit forever.  No data is ever generated by the host
(stopped capture after 4 seconds).

I believe this core should not be part of the FUA whitelist.  If I
remember correctly, there are other implementations out there with
similar limitations to opcodes this "new" to ATA.

--eric