Marvell 88SX[56]0[48]1 libata progress?

Marvell 88SX[56]0[48]1 libata progress?

[linux]

I was just wondering if there's been any progress.
I'm about to invest in an Abit SU-2S with an 88sx6081 on board, and was
wondering how things were going.

I can use the Marvell binary driver as a stopgap, but I'm hoping that
an open-source driver (and eventually NCQ support) will appear before
too long.

I'm afraid I can't afford to sponsor development personally, so I don't
have the right to complain too loudly, but could I politely inquire?

Thanks!

Re: Marvell 88SX[56]0[48]1 libata progress?

[Jeff Garzik]

linux@horizon.com wrote:
> I was just wondering if there's been any progress.
> I'm about to invest in an Abit SU-2S with an 88sx6081 on board, and was
> wondering how things were going.

No progress to report, I'm afraid.  I've been fighting bugs and working 
towards hotplug/queueing/error handling improvements.

After that, I'll do a round of driver writing for Marvell and several 
other boards.

If anybody has the hardware docs and time, they are welcome to contribute.

	Jeff

Re: Marvell 88SX[56]0[48]1 libata progress?

[Andy Warner]

Jeff Garzik wrote:
> [...]
> If anybody has the hardware docs and time, they are welcome to contribute.

A few months ago, it seems like things had already been started,
got any code to share, or are we at ground zero for the Marvell
chipset ?
-- 
andyw@pobox.com

Andy Warner		Voice: (612) 801-8549	Fax: (208) 575-5634

Re: Marvell 88SX[56]0[48]1 libata progress?

[Jeff Garzik]

Andy Warner wrote:
> Jeff Garzik wrote:
> 
>>[...]
>>If anybody has the hardware docs and time, they are welcome to contribute.
> 
> 
> A few months ago, it seems like things had already been started,
> got any code to share, or are we at ground zero for the Marvell
> chipset ?

I've got a skeleton of a driver that I'm embarrassed to show in public :)

	Jeff

Re: Marvell 88SX[56]0[48]1 libata progress?

[Bogdan Costescu]

On Thu, 27 Jan 2005, Jeff Garzik wrote:

> After that, I'll do a round of driver writing for Marvell and several other 
> boards.

Any tentative time-frame ? I'm also interested due to the presence of 
this controller disguised as Adaptec AIC-8110 on my Asus mainboard.

Unfortunately Asus only provides binary drivers for RH 8 & 9 and 
equally ancient SuSE kernels on the CD that accompanied the board, so 
no hope form there.

I have however found a driver with complete source (no .o files or I
have missed them) at:

http://file.abit.com.tw/pub/download/drivers/linux/marvell/mvsata340.zip

It compiled without any problem for the 2.4 kernel of TaoLinux, but
the performance is very low: hdparm shows only about 9Mb/s for a WD
Raptor disk. I refuse to believe that this is a normal level of
performance from this controller, the driver must be at fault...

-- 
Bogdan Costescu

IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu@IWR.Uni-Heidelberg.De

Re: Marvell 88SX[56]0[48]1 libata progress?

[Jeff Garzik]

Bogdan Costescu wrote:
> On Thu, 27 Jan 2005, Jeff Garzik wrote:
> 
> 
>>After that, I'll do a round of driver writing for Marvell and several other 
>>boards.
> 
> 
> Any tentative time-frame ? I'm also interested due to the presence of 
> this controller disguised as Adaptec AIC-8110 on my Asus mainboard.
> 
> Unfortunately Asus only provides binary drivers for RH 8 & 9 and 
> equally ancient SuSE kernels on the CD that accompanied the board, so 
> no hope form there.
> 
> I have however found a driver with complete source (no .o files or I
> have missed them) at:
> 
> http://file.abit.com.tw/pub/download/drivers/linux/marvell/mvsata340.zip
> 
> It compiled without any problem for the 2.4 kernel of TaoLinux, but
> the performance is very low: hdparm shows only about 9Mb/s for a WD
> Raptor disk. I refuse to believe that this is a normal level of
> performance from this controller, the driver must be at fault...

That driver has also been rejected...  it's vastly overengineered for 
what it needs to do, and should not be directly using the SCSI 
infrastructure (which makes it divergent from other SATA drives, which 
only indirectly use it) -- which causes incompatibilities down the line.

	Jeff