stable basic 4-port SATA card

stable basic 4-port SATA card

[Jeff Breidenbach]

I read with interest I. Straford's current trials and tribulations
with the Promise SATA300 TX4. Do people have a favorite
alternative to this card that plays well with Linux? I've read the
chipset compatibility list, but am not sure how to boil that
information down to an actual buyable SATA controller.

http://www.mail-archive.com/linux-ide@vger.kernel.org/msg12398.html
http://linux-ata.org/driver-status.html

Re: stable basic 4-port SATA card

[Tejun Heo]

Jeff Breidenbach wrote:
> I read with interest I. Straford's current trials and tribulations
> with the Promise SATA300 TX4. Do people have a favorite
> alternative to this card that plays well with Linux? I've read the
> chipset compatibility list, but am not sure how to boil that
> information down to an actual buyable SATA controller.
> 
> http://www.mail-archive.com/linux-ide@vger.kernel.org/msg12398.html
> http://linux-ata.org/driver-status.html

It's a sort of a difficult question to answer but I think it'll be nice
to develop a standard answer to this kind of question and put it on
linux-ata.org so that users can choose.  It can also hopefully work as
another pressure for vendors to enhance their linux support.

Anyways, here are what I know.

* ata_piix: Although it lacks shiny new features (interface limitation),
it should work pretty well.  No known serious bugs.

* ahci: Well supported.  Command switching PMP support is merged for
2.6.24 too.  Many (if not most) vendors now use ahci as programming
interface for storage controllers, so ahci driver supports lots of
controllers from many vendors.  Bugginess depends on which vendor or
chip you're actually using but in general all are well supported and if
you report a problem, it's very likely to get fixed soon.

* sata_sil: 3112 chips have problem with early seagate drives but both
3112 itself and those drives have been discontinued for a long time now.
 There are reports of data corruption when 3114 is connected to mobos
with NVidia chipsets.  This is still being investigated.  If you're on
intel mobo, 3112/3114/3512 should work good.

* sata_sil24: 3124/3132 chips don't have any outstanding serious
problems.  IRQ loss on PCI-X was the only recent serious known problem
but it's fixed now.  These chips are pretty advanced and all the
advanced features are supported by Linux including FIS switching PMP
support (will be included in 2.6.24 release); however, there is a known
hardware performance limitation so you can't use full SATA bandwidth
even if you use PMP but it's more than sufficient for most cases and
this chip is my personal favorite.

* sata_promise: Generally works okay; however there are still some
problems with recent 3Gbps chips. (Mikael, please pitch in)

* sata_nv: Generally works okay but ADMA support still seems to have
some problems (Robert?).

* sata_inic162x: Has half-working driver.  We need more info from the
vendor to proceed further.  Stay away from it for now.

* sata_via: Works okay.  No known serious problems but the chip is
quirky and not too dependable when errors occur.

* marvell ones: I don't have much idea.  Jeff?  Mark?

-- 
tejun

Re: stable basic 4-port SATA card

[Robert Hancock]

Tejun Heo wrote:
> Jeff Breidenbach wrote:
>> I read with interest I. Straford's current trials and tribulations
>> with the Promise SATA300 TX4. Do people have a favorite
>> alternative to this card that plays well with Linux? I've read the
>> chipset compatibility list, but am not sure how to boil that
>> information down to an actual buyable SATA controller.
>>
>> http://www.mail-archive.com/linux-ide@vger.kernel.org/msg12398.html
>> http://linux-ata.org/driver-status.html
> 
> It's a sort of a difficult question to answer but I think it'll be nice
> to develop a standard answer to this kind of question and put it on
> linux-ata.org so that users can choose.  It can also hopefully work as
> another pressure for vendors to enhance their linux support.
> 
> Anyways, here are what I know.
> 
> * ata_piix: Although it lacks shiny new features (interface limitation),
> it should work pretty well.  No known serious bugs.
> 
> * ahci: Well supported.  Command switching PMP support is merged for
> 2.6.24 too.  Many (if not most) vendors now use ahci as programming
> interface for storage controllers, so ahci driver supports lots of
> controllers from many vendors.  Bugginess depends on which vendor or
> chip you're actually using but in general all are well supported and if
> you report a problem, it's very likely to get fixed soon.
> 
> * sata_sil: 3112 chips have problem with early seagate drives but both
> 3112 itself and those drives have been discontinued for a long time now.
>  There are reports of data corruption when 3114 is connected to mobos
> with NVidia chipsets.  This is still being investigated.  If you're on
> intel mobo, 3112/3114/3512 should work good.
> 
> * sata_sil24: 3124/3132 chips don't have any outstanding serious
> problems.  IRQ loss on PCI-X was the only recent serious known problem
> but it's fixed now.  These chips are pretty advanced and all the
> advanced features are supported by Linux including FIS switching PMP
> support (will be included in 2.6.24 release); however, there is a known
> hardware performance limitation so you can't use full SATA bandwidth
> even if you use PMP but it's more than sufficient for most cases and
> this chip is my personal favorite.
> 
> * sata_promise: Generally works okay; however there are still some
> problems with recent 3Gbps chips. (Mikael, please pitch in)
> 
> * sata_nv: Generally works okay but ADMA support still seems to have
> some problems (Robert?).

There aren't many in the way of specific known problems that aren't 
worked around. There was a hotplug issue that was reported (kernel bug 
8421), but only seems to happen on a certain revision of MSI Neo4 
Platinum board so I suspect it's a board issue. Also there's the 
ATAPI/>4GB RAM issue that's being debugged. But I haven't seen many 
reports pointing to driver problems.

As far as the poster's question, these controllers are all built into 
the chipset, so it's not really relevant to their issue..

> 
> * sata_inic162x: Has half-working driver.  We need more info from the
> vendor to proceed further.  Stay away from it for now.
> 
> * sata_via: Works okay.  No known serious problems but the chip is
> quirky and not too dependable when errors occur.
> 
> * marvell ones: I don't have much idea.  Jeff?  Mark?
> 

Re: stable basic 4-port SATA card

[Mark Lord]

Tejun Heo wrote:
>
> * marvell ones: I don't have much idea.  Jeff?  Mark?
..

Right now, support is rather primitive and incomplete in sata_mv.

That may change over the next few months, as Marvell is beginning to
show some desire to have their chips fully supported, to/beyond the
current levels that sata_sil24 driver currently enjoys.

But right now it is rather spotty.

I don't know much about the other Marvell chips that run under AHCI,
but I hear that they're just fine.

Cheers

Re: stable basic 4-port SATA card

[Jeff Breidenbach]

Thanks for the excellent rundown.

>sata_sil24: 3124/3132 chips don't have any outstanding serious
>problems.  IRQ loss on PCI-X was the only recent serious known
>problem but it's fixed now.

I'm still a little confused how to translate this known-good chipset to
an actual buyable PCI card.  It isn't obvious from basic web searching.

Re: stable basic 4-port SATA card

[Tejun Heo]

Jeff Breidenbach wrote:
> Thanks for the excellent rundown.
> 
>> sata_sil24: 3124/3132 chips don't have any outstanding serious
>> problems.  IRQ loss on PCI-X was the only recent serious known
>> problem but it's fixed now.
> 
> I'm still a little confused how to translate this known-good chipset to
> an actual buyable PCI card.  It isn't obvious from basic web searching.

That I don't know either.  I usually magnify the product photograph and
try to read the chip number (I know shapes of several chips so it's
easier for me) or call the manufacturer.  Just ask for the PCI ID and
look it up in linux PCI IDs repository or modinfo result.

-- 
tejun

Re: stable basic 4-port SATA card

[Jim Paris]

Jeff Breidenbach wrote:
> Thanks for the excellent rundown.
> 
> >sata_sil24: 3124/3132 chips don't have any outstanding serious
> >problems.  IRQ loss on PCI-X was the only recent serious known
> >problem but it's fixed now.
> 
> I'm still a little confused how to translate this known-good chipset to
> an actual buyable PCI card.  It isn't obvious from basic web searching.

I also look for 3124/3132 and these are some that I've found:
(determined by specs, product reviews, looking closely at photos, etc,
it's worth a double-check before deciding)

3132 PCIe:
http://www.newegg.com/Product/Product.aspx?Item=N82E16816132010
http://www.newegg.com/Product/Product.aspx?Item=N82E16816855002
http://www.newegg.com/Product/Product.aspx?Item=N82E16816132011
http://www.newegg.com/Product/Product.aspx?Item=N82E16816132008

3132 ExpressCard:
http://www.newegg.com/Product/Product.aspx?Item=N82E16839150001

3124 PCI:
http://www.newegg.com/Product/Product.aspx?Item=N82E16816124003

3124 PCI-X:
http://www.newegg.com/Product/Product.aspx?Item=N82E16816318004

Hope this helps

-jim

Re: stable basic 4-port SATA card

[Mikael Pettersson]

On Thu, 15 Nov 2007 12:32:14 +0900, Tejun Heo wrote:
> * sata_promise: Generally works okay; however there are still some
> problems with recent 3Gbps chips. (Mikael, please pitch in)

Right, 2nd-gen 3Gbps chips have had "intermittent" issues,
which we hope are cured by the recent ASIC bug workaround,
but it will take a while for that fix to propagate out.
To speed up that process I'm considering backporting the fix
to 2.6.23 and 2.6.22.

NCQ and PMP are supported in the hardware and in the vendor's
driver, but not yet in sata_promise. My intention is to add
NCQ soon, but there's no time-plan yet for this.

/Mikael

Re: stable basic 4-port SATA card

[john]

On Thu, 15 Nov 2007, Jim Paris wrote:

> Jeff Breidenbach wrote:
>> Thanks for the excellent rundown.
>>
>>> sata_sil24: 3124/3132 chips don't have any outstanding serious
>>> problems.  IRQ loss on PCI-X was the only recent serious known
>>> problem but it's fixed now.
>>
>> I'm still a little confused how to translate this known-good chipset to
>> an actual buyable PCI card.  It isn't obvious from basic web searching.
>
> I also look for 3124/3132 and these are some that I've found:
> (determined by specs, product reviews, looking closely at photos, etc,
> it's worth a double-check before deciding)
>
> 3132 PCIe:
> http://www.newegg.com/Product/Product.aspx?Item=N82E16816132010
> http://www.newegg.com/Product/Product.aspx?Item=N82E16816855002
> http://www.newegg.com/Product/Product.aspx?Item=N82E16816132011
> http://www.newegg.com/Product/Product.aspx?Item=N82E16816132008

I have used the Syba without any problems:

http://www.newegg.com/Product/Product.aspx?Item=N82E16815124027


John

Re: stable basic 4-port SATA card

[Tejun Heo]

Mikael Pettersson wrote:
> On Thu, 15 Nov 2007 12:32:14 +0900, Tejun Heo wrote:
>> * sata_promise: Generally works okay; however there are still some
>> problems with recent 3Gbps chips. (Mikael, please pitch in)
> 
> Right, 2nd-gen 3Gbps chips have had "intermittent" issues,
> which we hope are cured by the recent ASIC bug workaround,
> but it will take a while for that fix to propagate out.
> To speed up that process I'm considering backporting the fix
> to 2.6.23 and 2.6.22.
> 
> NCQ and PMP are supported in the hardware and in the vendor's
> driver, but not yet in sata_promise. My intention is to add
> NCQ soon, but there's no time-plan yet for this.

Thanks all.  I'll incorporate the info here into linux-ata.org when it
gets wikified.

-- 
tejun

Re: stable basic 4-port SATA card

[Patric Karlsson]

Mikael Pettersson wrote:
> On Thu, 15 Nov 2007 12:32:14 +0900, Tejun Heo wrote:
>> * sata_promise: Generally works okay; however there are still some
>> problems with recent 3Gbps chips. (Mikael, please pitch in)
> 
> Right, 2nd-gen 3Gbps chips have had "intermittent" issues,
> which we hope are cured by the recent ASIC bug workaround,
> but it will take a while for that fix to propagate out.
> To speed up that process I'm considering backporting the fix
> to 2.6.23 and 2.6.22.
> 
> NCQ and PMP are supported in the hardware and in the vendor's
> driver, but not yet in sata_promise. My intention is to add
> NCQ soon, but there's no time-plan yet for this.
> 
> /Mikael
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ide" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Backporting would be nice, since we're still some ways from a stable 
2.6.24, and I'm not comfortable running -rc kernels in production 
enviroments.

Re: stable basic 4-port SATA card

[Mikael Pettersson]

Patric Karlsson writes:
 > Mikael Pettersson wrote:
 > > On Thu, 15 Nov 2007 12:32:14 +0900, Tejun Heo wrote:
 > >> * sata_promise: Generally works okay; however there are still some
 > >> problems with recent 3Gbps chips. (Mikael, please pitch in)
 > > 
 > > Right, 2nd-gen 3Gbps chips have had "intermittent" issues,
 > > which we hope are cured by the recent ASIC bug workaround,
 > > but it will take a while for that fix to propagate out.
 > > To speed up that process I'm considering backporting the fix
 > > to 2.6.23 and 2.6.22.
 > > 
 > > NCQ and PMP are supported in the hardware and in the vendor's
 > > driver, but not yet in sata_promise. My intention is to add
 > > NCQ soon, but there's no time-plan yet for this.
 > > 
 > > /Mikael
 > > -
 > > To unsubscribe from this list: send the line "unsubscribe linux-ide" in
 > > the body of a message to majordomo@vger.kernel.org
 > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
 > > 
 > 
 > Backporting would be nice, since we're still some ways from a stable 
 > 2.6.24, and I'm not comfortable running -rc kernels in production 
 > enviroments.

Ok.

For 2.6.23 there is one patch:

<http://user.it.uu.se/~mikpe/linux/patches/sata_promise/patch-sata_promise-1-asic-sg-bug-fix-v3-2.6.23>
This is the workaround for the ASIC PRD/SG bug.

For 2.6.22 there are two patches:

<http://user.it.uu.se/~mikpe/linux/patches/sata_promise/patch-sata_promise-1-ft_tx4200-is-gen2-2.6.22>
This corrects the classification of FastTrack TX4200 cards,
which in turn fixes nasty errors in how they are accessed.

<http://user.it.uu.se/~mikpe/linux/patches/sata_promise/patch-sata_promise-2-asic-sg-bug-fix-v3-2.6.22>
This is the workaround for the ASIC PRD/SG bug.

/Mikael

Re: stable basic 4-port SATA card

[Krzysztof Oledzki]

On 2007-12-13 15:09, Mikael Pettersson wrote:
> Patric Karlsson writes:
>  > Mikael Pettersson wrote:
>  > > On Thu, 15 Nov 2007 12:32:14 +0900, Tejun Heo wrote:
>  > >> * sata_promise: Generally works okay; however there are still some
>  > >> problems with recent 3Gbps chips. (Mikael, please pitch in)
>  > > 
>  > > Right, 2nd-gen 3Gbps chips have had "intermittent" issues,
>  > > which we hope are cured by the recent ASIC bug workaround,
>  > > but it will take a while for that fix to propagate out.
>  > > To speed up that process I'm considering backporting the fix
>  > > to 2.6.23 and 2.6.22.
>  > > 
>  > > NCQ and PMP are supported in the hardware and in the vendor's
>  > > driver, but not yet in sata_promise. My intention is to add
>  > > NCQ soon, but there's no time-plan yet for this.
>  > > 
>  > > /Mikael
>  > > -
>  > > To unsubscribe from this list: send the line "unsubscribe linux-ide" in
>  > > the body of a message to majordomo@vger.kernel.org
>  > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
>  > > 
>  > 
>  > Backporting would be nice, since we're still some ways from a stable 
>  > 2.6.24, and I'm not comfortable running -rc kernels in production 
>  > enviroments.
> 
> Ok.
> 
> For 2.6.23 there is one patch:
> 
> <http://user.it.uu.se/~mikpe/linux/patches/sata_promise/patch-sata_promise-1-asic-sg-bug-fix-v3-2.6.23>
> This is the workaround for the ASIC PRD/SG bug.
> 
> For 2.6.22 there are two patches:
> 
> <http://user.it.uu.se/~mikpe/linux/patches/sata_promise/patch-sata_promise-1-ft_tx4200-is-gen2-2.6.22>
> This corrects the classification of FastTrack TX4200 cards,
> which in turn fixes nasty errors in how they are accessed.
> 
> <http://user.it.uu.se/~mikpe/linux/patches/sata_promise/patch-sata_promise-2-asic-sg-bug-fix-v3-2.6.22>
> This is the workaround for the ASIC PRD/SG bug.

Maybe -stable should get this?

Best regards,

				Krzysztof Olędzki