[illumos-Developer] [REVIEW] 1299 Support SiI3531 chipset in si3124 driver

[Charles Stephens]

On Jul 29, 2011, at 2:53 PM, Garrett D'Amore wrote:

> On 07/29/2011 02:38 PM, Charles Stephens wrote:
>> On Jul 29, 2011, at 2:25 PM, Yuri Pankov wrote:
>> 
>>> On Fri, Jul 29, 2011 at 01:54:39PM -0700, Charles Stephens wrote:
>>>> On Jul 29, 2011, at 1:43 PM, Albert Lee wrote:
>>>> 
>>>>> Is there absolutely no way to tell how many ports present are on a device?
>>>> I have read the data sheets for both the 3124 and 3531 and it seems to be the
>>>> one thing you can't tell.  I used the Linux SiI driver as a guide to see how
>>>> they figure it out and it uses an array that lists port counts.
>>> Same in FreeBSD's siis(4) driver:
>>> 
>>> http://fxr.watson.org/fxr/source/dev/siis/siis.c#L96
>> So it's resolved, port counts are fixed per type of device and can not be determined at runtime.
>> 
>>> I wonder though if the quirks listed for 3531 and not listed for 3124
>>> are of any importance to us..
>> A cursory glance seems that these quirks don't really apply to Illumos's SATA framework.
> 
> I just reached the same conclusion:
> 
> a) We don't support MSI at all in the illumos driver (a potential RFE there, I think)



> b) We don't use the set device bits notification FIS in illumos -- we just log it.   I'm not entirely sure what this notification is intended for.   I don't have documentation for it.

Probably because port multiplying doesn't work with this driver, so it is probably a stern warning that you are going to one LUN's worth of disk seen on the system.

> Anyway, I think we can ship what we have now.

Yay, I haven't had anything put back into a gate since Solaris 8 and never outside of user-land.  Quite exciting.

cfs