Re: Marvell SATA Support

On 07/11/2005, at 10:20, Dima Dorfman wrote:

> Søren Schmidt <sos_at_DeepCore.dk> wrote:
>> Lawrence Farr wrote:
>>> Hello All,
>>>
>>> I have a Supermicro P4SCT+ with an onboard Marvell
>>> SATA controller, which also has the Adaptec Hostraid
>>> software raid functionality. Are there any patches to
>>> support the marvell controller as a plain controller
>>> anywhere?
>>>
>>> pciconf output if anyones interested:
>>>
>>> none4_at_pci2:4:0: class=0x010000 card=0x504111ab chip=0x504111ab  
>>> rev=0x03
>>> hdr=0x00
>>>    vendor   = 'Marvell Semiconductor (Was: Galileo Technology Ltd)'
>>>    device   = '88SX504 4-port SATA I PCI-X Controller'
>>>    class    = mass storage
>>>    subclass = SCSI
>>
>> Not yet, but I'm working on it, actually on the exact same board.
>> Maybe, just maybe, you could hack the Highpoint hptmv driver, I  
>> havn't
>> tried but unless they put in "tricks" to prohibit that it should  
>> work.
>
> Søren,
>
> Have you made any progress on this? I have one of these chips on a
> SuperMicro 5013MT-MT(B) board. It has a HostRAID BIOS, but I don't
> need the RAID functionality--just being able to see the drives would
> be enough.
>
> Here's my pciconf output (I forgot to write down the entire class
> value and PCI address, but everything else is accurate):
>
>  noneX_at_pciX:Y:Z: class=XXX card=0x518015d9 chip=0x504111ab rev=0x00  
> hdr=0x00
>      vendor   = 'Marvell Semiconductor (Was: Galileo Technology Ltd)'
>      device   = '88SX504 4-port SATA I PCI-X Controller'
>      class    = mass storage
>      subclass = RAID
>
> Since the subclass is RAID, the ata driver doesn't try to attach to it
> at all--not even using generic DMA. Would it make sense to force it to
> try generic DMA, or does the subclass indicate that the interface is
> completely different?
>
> In its current state, I can't see any disks attached to this
> controller. Is there an easy way to make it work without RAID? I'd
> hate to put another controller into this box.

The marvell chips doesn't look like a normal ATA controller *at all*,  
they are very different animals that actually could be dealt with in  
a seperate driver if need be.
Now I've gone to great lengths in ATA to be able to support chips  
like this (newer Promise chips and AHCI are also in this category),  
and support will come at some point, I just havn't had time to work  
on it due to other more urgent things (like getting ATA in 6.0  
working as well as possible). Since this is all spare time work I  
have to prioritize my time, and so far there's always been something  
else that popped up..
So patience is the way to go, or maybe hope for some vendor needing  
it badly enough to contract me working hours to do it...

Søren Schmidt
sos_at_FreeBSD.org