Re: Promise TX4 + SiI 3726 : only one disk visible

Re: Promise TX4 + SiI 3726 : only one disk visible

[Tejun Heo]

Jerome Vuarand wrote:
> Hi,
>  
> Abstract : SiI3726 based port multiplier on Promise SATA300 TX4 don't 
> work on Slackware 10.2 with 2.6.16.16 kernel and libata-tj-stable 
> combined patch : only one disk accessible behind the port multiplier. 
> Don't bother reading the long mail if that is a known issue.

 From README...

The following drivers support new features.

ata_piix:	new EH, irq-pio, warmplug (hardware restriction)
sata_sil:	new EH, irq-pio, hotplug
ahci:		new EH, irq-pio, NCQ, hotplug
sata_sil24:	new EH, irq-pio, NCQ, hotplug, Port Multiplier
^^^^^^^^^^^                                    ^^^^^^^^^^^^^^^

[--snip--]
> I don't know really what the problem is, and I don't know what else to 
> try. Your libata-tj README mention the SiI 4726 chipset, but there is no 
> such chipset on Silicon Image website, so I guessed you meant SiI 3726, 
> but I may be wrong on that point.

Yes, there is 4726.  I don't have 3726 yet but 4726 is probably 3726 + 
storage processor (so that it can do disk virtualization/RAID stuff). 
I'll add support for 3726 soon.

PMP support requires support from low-level drivers (sata_sil24, 
sata_promise,...) and sata_promise doesn't have it yet.  As noted above, 
the only supported card is SiI3124/32 and I'm working on AHCI.  As for 
sata_promise, I don't know.  I don't have the card nor the document, so 
it won't come out of me.  Jeff is planning to work on sata_promise and 
he might also do PM stuff, I guess.

-- 
tejun

Re: Promise TX4 + SiI 3726 : only one disk visible

[Jeff Garzik]

Tejun Heo wrote:
> PMP support requires support from low-level drivers (sata_sil24, 
> sata_promise,...) and sata_promise doesn't have it yet.  As noted above, 
> the only supported card is SiI3124/32 and I'm working on AHCI.  As for 
> sata_promise, I don't know.  I don't have the card nor the document, so 
> it won't come out of me.  Jeff is planning to work on sata_promise and 
> he might also do PM stuff, I guess.

FWIW, for sata_promise (PDC205xx):

I just looked in the docs, for Port Multiplier stuff.  There is a single 
vendor-proprietary 4-bit "PM port" MMIO register for each channel (aka 
each SATA port).

You may program this register only when the port is idle, no active 
commands nor DMA.

	Jeff

Re: Promise TX4 + SiI 3726 : only one disk visible

[Jeff Garzik]

Jeff Garzik wrote:
> I just looked in the docs, for Port Multiplier stuff.  There is a single 
> vendor-proprietary 4-bit "PM port" MMIO register for each channel (aka 
> each SATA port).

I take that back.  PDC205xx also has a 'PM Port' field in its 
per-command S/G DMA packets.

	Jeff

Re: Promise TX4 + SiI 3726 : only one disk visible

[Tejun Heo]

Jérôme VUARAND wrote:
> Thanks a lot for the quick answers. Sorry to have misread tejun's
> readme, I tought there was some abstraction layer between SATA
> protocol and SATAII extensions, so I didn't even suppose there could
> be a need for controller specific code to support port multiplier. I
> was clearly wrong.

Yeap, there is a layer, but it still needs low level support.

> So to be clear, there is not yet any support for my configuration
> (sata_promise+SiI3726) in any of your code branches ? If not which one
> should I start from to try to add the support myself (I never
> programmed anything in the kernel, so it would just be a newbie try) ?

No, not yet.  You can try but the biggest obstacle would be lack of 
document.  If you can access the doc, it shouldn't be too difficult, I 
guess.  You can analyze how sata_sil24 has been changed and update 
sata_promise similarly.

-- 
tejun

Re: Promise TX4 + SiI 3726 : only one disk visible

[Jeff Garzik]

Jérôme VUARAND wrote:
> Latest libata feature table mentions port multiplier support as "(patch 
> available)". Is that "patch" refering to tejun patch ? If not is that an 
> evolution or another branch ? Where is it available ? It don't seem to 
> be inside 2.6.17-rc4-git2-libata1.patch, but I had only a quick look at 
> the patch content.

That refers to Tejun's stuff, yes.


> Also Jeff mentionned information about port multiplier support in 
> PDC205xx. Is there any documentation freely available somewhere ? I 

No.


> looked on Promise site but found nothing. Should I contact Promise 
> myself as a customer to obtain documentation or maybe look at Promise 
> driver ?

I'll try to work up a patch with the necessary info.

	Jeff

Re: Promise TX4 + SiI 3726 : only one disk visible

[Tejun Heo]

Jérôme VUARAND wrote:
> Latest libata feature table mentions port multiplier support as "(patch 
> available)". Is that "patch" refering to tejun patch ? If not is that an 
> evolution or another branch ? Where is it available ? It don't seem to 
> be inside 2.6.17-rc4-git2-libata1.patch, but I had only a quick look at 
> the patch content.

Yeap, it probably is referring to my patch.  The PM part is not included 
in any other trees yet.  PM is gonna take some time to get merged. 
Also, 3726 needs a quirk patch to work with the current PMP support. 
I'm still trying to sort out hotplug spinup delaying, so it might take 
some time.

> If I should still use 2.6.16.16 <http://2.6.16.16> and tj patches, is 
> porting 0121-sata_sil24-implement-PM-support.patch to sata_promise 
> enough or are the other sata_sil24 patches also needed (0117 to 0122) ?

Actually, it's a bigger than the said patch.  To do PM, sata_promise 
must be converted to new EH and thus to use hotplug framework, then, the 
PM support.  Look at how many patches update sata_sil24 in my patch series.

> Also Jeff mentionned information about port multiplier support in 
> PDC205xx. Is there any documentation freely available somewhere ? I 
> looked on Promise site but found nothing. Should I contact Promise 
> myself as a customer to obtain documentation or maybe look at Promise 
> driver ?

The doc is available under NDA.  I don't know about Promise's NDA policy 
but in my experience getting docs is often difficult.

-- 
tejun