Re: [PATCH #upstream] ahci: add HFLAG_YES_FBS and apply it to 88SE9128

[Anssi Hannula <anssi.hannula@iki.fi>]

On 30.11.2010 22:10, Anssi Hannula wrote:
> On 30.11.2010 15:31, Tejun Heo wrote:
>> Hello,

Hi!

>> On 11/30/2010 03:04 AM, Anssi Hannula wrote:
>>> On 24.07.2010 17:53, Tejun Heo wrote:
>>>> 88SE9128 can do FBS and sets it in HOST_CAP but forgets to set FBSCP
>>>> in PORT_CMD.  Implement AHCI_HFLAG_YES_FBS and apply it to 88SE9128.
>>>>
>>>> Signed-off-by: Tejun Heo <tj@kernel.org>
>>> [...]
>>>> diff --git a/drivers/ata/ahci.c b/drivers/ata/ahci.c
>>>> index f252253..41fa0a3 100644
>>> [...]
>>>> @@ -362,6 +371,8 @@ static const struct pci_device_id ahci_pci_tbl[] = {
>>>>  	/* Marvell */
>>>>  	{ PCI_VDEVICE(MARVELL, 0x6145), board_ahci_mv },	/* 6145 */
>>>>  	{ PCI_VDEVICE(MARVELL, 0x6121), board_ahci_mv },	/* 6121 */
>>>> +	{ PCI_DEVICE(0x1b4b, 0x9123),
>>>> +	  .driver_data = board_ahci_yes_fbs },			/* 88se9128 */
>>>
>>> The device contains an IDE interface as well, and the above
>>> pci_device_id matches them both:
>>> 08:00.0 SATA controller [0106]: Device [1b4b:9123] (rev 10)
>>> 08:00.1 IDE interface [0101]: Device [1b4b:9123] (rev 10)
>>
>> *GASP* Come on Marvell.... :-(
>>
>>> This results in some (apparently harmless) mess [1].
>>>
>>> Attached is a patch to make the id entry class-specific. The patch is
>>> untested (the issue happens in a production machine).
>>
>> Thanks a lot for the patch.  It looks correct but it would be great if
>> it can be veified.
> 
> I think I can do that later, but probably not this week or so.

I've now verified that the patch works on a system with 1b4b:9123
controller.

>> Also, there's a pending patch to add another PCI
>> ID for similar marvell controller.  I wonder whether similar
>> workaround should be applied.  Hmmm... ISTR the other one reporting
>> IDE class even though it works in IDE mode.
> 
> IDE class in IDE mode? Isn't that kind of expected? :)
> 
> Also, just to avoid any misunderstanding: I see both the SATA and IDE
> interfaces at the same time.
> 
>> Can someone with marvell
>> documentation access check what's going on?
>>
>> Thanks.
>>
> 


-- 
Anssi Hannula