Re: [PATCH 3/3] mmc: core: Default to generic_cmd6_time as timeout in __mmc_switch()

[Adrian Hunter <adrian.hunter@intel.com>]

On 23/02/21 1:54 pm, Ulf Hansson wrote:
> On Tue, 23 Feb 2021 at 12:19, Paul Fertser <fercerpav@gmail.com> wrote:
>>
>> Hello Adrian,
>>
>> On Tue, Feb 23, 2021 at 01:01:09PM +0200, Adrian Hunter wrote:
>>> On 23/02/21 11:32 am, Paul Fertser wrote:
>>>> Hello Ulf,
>>>>
>>>> On Tue, Feb 23, 2021 at 10:23:28AM +0100, Ulf Hansson wrote:
>>>>>> diff --git a/drivers/mmc/core/mmc.c b/drivers/mmc/core/mmc.c
>>>>>> index f5dedb7f9b27..9adf735391fa 100644
>>>>>> --- a/drivers/mmc/core/mmc.c
>>>>>> +++ b/drivers/mmc/core/mmc.c
>>>>>> @@ -426,8 +426,7 @@ static int mmc_decode_ext_csd(struct mmc_card *card, u8 *ext_csd)
>>>>>>                 /* EXT_CSD value is in units of 10ms, but we store in ms */
>>>>>>                 card->ext_csd.part_time = 10 * ext_csd[EXT_CSD_PART_SWITCH_TIME];
>>>>>>                 /* Some eMMC set the value too low so set a minimum */
>>>>>> -               if (card->ext_csd.part_time &&
>>>>>> -                   card->ext_csd.part_time < MMC_MIN_PART_SWITCH_TIME)
>>>>>> +               if (card->ext_csd.part_time < MMC_MIN_PART_SWITCH_TIME)
>>>>>>                         card->ext_csd.part_time = MMC_MIN_PART_SWITCH_TIME;
>>>>>>
>>>>>>                 /* Sleep / awake timeout in 100ns units */
>>>>>>
>>>>>> I do not see any more warnings on my system.
>>>>>
>>>>> That looks like the correct fix to the problem. Do you want to send a
>>>>> proper patch that I can pick up or do you prefer if help to do it?
>>>>
>>>> I've sent this as a diff precisely because 1c447116d017 was so
>>>> explicit about special-casing zero ext_csd timeout value, so I thought
>>>> probably Adrian can provide the rationale for that. I'd prefer to wait
>>>> for his feedback before sending a formal patch. Does this make sense?
>>>
>>> Zero means indefinite.  Might be safer to use a higher value than
>>> MMC_MIN_PART_SWITCH_TIME for that case.  The maximum GENERIC_CMD6_TIME is
>>> 2550 ms.
>>
>> Thanks for the clarification! I would guess that most likely than not
>> when whoever defines that value to be zero it means "I do not
>> care/know" rather than "the timeout must be set to more than 2550 ms,
>> too bad 8 bits are not enough to represent that". I'd say setting it
>> to DEFAULT_CMD6_TIMEOUT_MS should be safe enough since it worked
>> before.
> 
> Hmm.
> 
> The DEFAULT_CMD6_TIMEOUT_MS is intended to override the
> ext_csd->generic_cmd6_time, in case it's not defined in the register.
> 
> Perhaps it's reasonable to think that eMMC vendors specify the
> GENERIC_CMD6_TIME, but may skip to specify other timeouts, like the
> PARTITION_SWITCH_TIME. In that case, should we use the specified
> GENERIC_CMD6_TIME, rather than always default to
> DEFAULT_CMD6_TIMEOUT_MS?

Sounds reasonable, but perhaps still enforce a minimum, for some of the same
reasons as commit 1c447116d017 ?
e.g.

	if (!card->ext_csd.part_time)
		card->ext_csd.part_time = card->ext_csd.generic_cmd6_time;
	if (card->ext_csd.part_time < MMC_MIN_PART_SWITCH_TIME)
		card->ext_csd.part_time = MMC_MIN_PART_SWITCH_TIME;