Re: PATCH: rtsx_pci driver - don't disable the rts5229 card reader on Intel NUC boxes

[Chris Clayton]

Hi Ricky

On 05/08/2020 06:51, Chris Clayton wrote:
> Thanks, Ricky.
> 
> On 05/08/2020 03:35, 吳昊澄 Ricky wrote:
>>
>>
>>> -----Original Message-----
>>> From: Chris Clayton [mailto:chris2553@xxxxxxxxxxxxxx]
>>> Sent: Tuesday, August 04, 2020 7:52 PM
>>> To: 吳昊澄 Ricky; gregkh@xxxxxxxxxxxxxxxxxxx
>>> Cc: LKML; rdunlap@xxxxxxxxxxxxx; philquadra@xxxxxxxxx; Arnd Bergmann
>>> Subject: Re: PATCH: rtsx_pci driver - don't disable the rts5229 card reader on
>>> Intel NUC boxes
>>>
>>>
>>>
>>> On 04/08/2020 11:46, 吳昊澄 Ricky wrote:
>>>>> -----Original Message-----
>>>>> From: Chris Clayton [mailto:chris2553@xxxxxxxxxxxxxx]
>>>>> Sent: Tuesday, August 04, 2020 4:51 PM
>>>>> To: 吳昊澄 Ricky; gregkh@xxxxxxxxxxxxxxxxxxx
>>>>> Cc: LKML; rdunlap@xxxxxxxxxxxxx; philquadra@xxxxxxxxx; Arnd Bergmann
>>>>> Subject: Re: PATCH: rtsx_pci driver - don't disable the rts5229 card reader on
>>>>> Intel NUC boxes
>>>>>
>>>>>
>>>>>
>>>>> On 04/08/2020 09:08, 吳昊澄 Ricky wrote:
>>>>>>> -----Original Message-----
>>>>>>> From: gregkh@xxxxxxxxxxxxxxxxxxx [mailto:gregkh@xxxxxxxxxxxxxxxxxxx]
>>>>>>> Sent: Tuesday, August 04, 2020 3:49 PM
>>>>>>> To: 吳昊澄 Ricky
>>>>>>> Cc: Chris Clayton; LKML; rdunlap@xxxxxxxxxxxxx; philquadra@xxxxxxxxx;
>>>>> Arnd
>>>>>>> Bergmann
>>>>>>> Subject: Re: PATCH: rtsx_pci driver - don't disable the rts5229 card reader
>>> on
>>>>>>> Intel NUC boxes
>>>>>>>
>>>>>>> On Tue, Aug 04, 2020 at 02:44:41AM +0000, 吳昊澄 Ricky wrote:
>>>>>>>> Hi Chris,
>>>>>>>>
>>>>>>>> rtsx_pci_write_register(pcr, FPDTL, OC_POWER_DOWN,
>>>>> OC_POWER_DOWN);
>>>>>>>> This register operation saved power under 1mA, so if do not care the 1mA
>>>>>>> power we can have a patch to remove it, make compatible with NUC6
>>>>>>>> We tested others our card reader that remove it, we did not see any side
>>>>> effect
>>>>>>>>
>>>>>>>> Hi Greg k-h,
>>>>>>>>
>>>>>>>> Do you have any comments?
>>>>>>>
>>>>>>> comments on what?  I don't know what you are responding to here, sorry.
>>>>>>>
>>>>>> Can we have a patch to kernel for NUC6? It may cause more power(1mA) but
>>> it
>>>>> will have more compatibility
>>>>>>
>>>>>
>>>>> Ricky,
>>>>>
>>>>> I don't understand why you want to completely remove the code that sets up
>>> the
>>>>> 1mA power saving. That code was there
>>>>> even before your patch (bede03a579b3b4a036003c4862cc1baa4ddc351f), so I
>>>>> assume it benefits some of the Realtek card
>>>>> readers. Before your patch however, rtsx_pci_init_ocp() was not called for the
>>>>> rts5229 reader, but the patch introduced
>>>>> an unconditional call to that function into rtsx_pci_init_hw(), which is run for
>>> the
>>>>> rts5229. That is what now disables
>>>>> the card reader.
>>>>>
>>>>> Now, I don't know whether other cards are affected, although I don't recall
>>>>> seeing any reported as I searched the kernel
>>>>> and ubuntu bugzillas for any analysis of the problem. I know this is not what
>>> the
>>>>> patch I sent does, but having thought
>>>>> about it more, seems to me that the simplest fix is to skip the new call to
>>>>> rtsx_pci_init_ocp() if the reader is an rts5229.
>>>>>
>>>>
>>>> Because we are thinking about if others our card reader that not belong A
>>> series(my ocp patch coverage) also on NUC6 platform maybe have the same
>>> problem...
>>>>
>>>
>>> OK. What if we do make the new call but only for the card readers that are in the
>>> A series? Are they the ones that have
>>> PID_5nnn defines in include/linux/rtsx_pci.h? Or is there another simple way of
>>> identifying that a reader is a member of
>>> the A series?
>>>
>>> I'm thinking of something like:
>>> static bool rtsx_pci_is_series_A(pcr)
>>> {
>>> 	unsigned short device = pcr->pci->device;
>>>
>>> 	return device == PID524A || device == PID_5249 || device == PID_5250 ||
>>> device == PID_525A
>>> 			|| device == PID_525A || device == PID_5260 || device ==
>>> PID_5261;
>>> }
>>>
>>> then in rtsx_pci_init_hw() change the unconditional call to:
>>>
>>> 	if rtsx_pci_is_series_A(pcr)
>>> 		rtsx_pci_init_ocp();
>>>
>>> Does that seem OK?
>>>
>> Previously, I want to remove
>> else {
>> 		/* OC power down */
>> 		rtsx_pci_write_register(pcr, FPDCTL, OC_POWER_DOWN,
>> 			OC_POWER_DOWN);
>> }
>> Because in our A-series card Reader we already assigned option->ocp_en to 1 in self init_params() , this is an easy way to fix this problem
>>
> 
> Ah, OK. I'll prepare the patch and send it to you once I've tested it.
> 

Please see the patch included below. As you suggested, it removes the code that does the OC power down on card readers
that are not members of your A series. The patch is against a fresh pull of Linus's tree this morning
(v5.8-2768-g4da9f3302615).

I've tested the resultant kernel on my Intel NUC6CAYH box (which contains an NUC66AYB board) and the rts5229 works fine.
I've also tested it on my laptop which also has a card reader supported by the rtsx_pci driver (an RTL8411B) and that
also works fine.

As I mentioned yesterday, I think it's a candidate for the 5.4 ans 5.7 stable series.

Thanks for your patience!

Chris

Signed-off-by: Chris Clayton <chris2553@xxxxxxxxxxxxxx>

--- a/drivers/misc/cardreader/rtsx_pcr.c        2020-08-05 07:10:21.752072515 +0100
+++ b/drivers/misc/cardreader/rtsx_pcr.c        2020-08-05 07:11:05.568074278 +0100
@@ -1172,10 +1172,6 @@ void rtsx_pci_init_ocp(struct rtsx_pcr *
                        rtsx_pci_write_register(pcr, REG_OCPGLITCH,
                                SD_OCP_GLITCH_MASK, pcr->hw_param.ocp_glitch);
                        rtsx_pci_enable_ocp(pcr);
-               } else {
-                       /* OC power down */
-                       rtsx_pci_write_register(pcr, FPDCTL, OC_POWER_DOWN,
-                               OC_POWER_DOWN);
                }
        }
 }