Re: [PATCH V2 2/4] mmc: sdhci: Avoid unnecessary re-configuration

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

On 30/11/22 21:56, Robert Marko wrote:
> On Wed, 30 Nov 2022 at 19:39, Florian Fainelli <f.fainelli@gmail.com> wrote:
>>
>> On 11/30/22 09:24, Robert Marko wrote:
>>> On Wed, 30 Nov 2022 at 15:16, Adrian Hunter <adrian.hunter@intel.com> wrote:
>>>>
>>>> On 30/11/22 15:00, Robert Marko wrote:
>>>>> On Wed, 30 Nov 2022 at 13:46, Adrian Hunter <adrian.hunter@intel.com> wrote:
>>>>>>
>>>>>> On 30/11/22 13:54, Robert Marko wrote:
>>>>>>>
>>>>>>> On 28. 11. 2022. 14:32, Adrian Hunter wrote:
>>>>>>>> Avoid re-configuring UHS and preset settings when the settings have not
>>>>>>>> changed, irrespective of whether the clock is turning on.
>>>>>>>>
>>>>>>>> Tested-by: Haibo Chen <haibo.chen@nxp.com>
>>>>>>>> Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
>>>>>>>
>>>>>>> Hi, this is breaking sdhci-msm on IPQ8074 in next-20221130 for me
>>>>>>> and reverting it makes the eMMC work again.
>>>>>>>
>>>>>>> I get a lot of:
>>>>>>>
>>>>>>> [    2.727287] mmc0: tuning execution failed: -5
>>>>>>> [    2.727323] mmc0: error -5 whilst initialising MMC card
>>>>>>> [    3.846540] mmc0: tuning execution failed: -5
>>>>>>> [    3.846564] mmc0: error -5 whilst initialising MMC card
>>>>>>> [    4.966517] mmc0: tuning execution failed: -5
>>>>>>> [    4.966539] mmc0: error -5 whilst initialising MMC card
>>>>>>> [    6.096486] mmc0: tuning execution failed: -5
>>>>>>> [    6.096508] mmc0: error -5 whilst initialising MMC card
>>>>>>> [    7.206431] mmc0: tuning execution failed: -5
>>>>>>> [    7.206454] mmc0: error -5 whilst initialising MMC card
>>>>>>
>>>>>> Thanks for the report!  Are you able to debug this any more?
>>>>>> What transfer mode is it? e.g. HS400?  Can you enable debug
>>>>>> messages and get more information?
>>>>>
>>>>> With some guidance yes, it's in HS200 as there is an issue with HS400
>>>>> to HS200 switch on this SoC so I have HS400 disabled.
>>>>>
>>>>> With CONFIG_MMC_DEBUG and loglevel=8 I dont have any new
>>>>> messages.
>>>>
>>>> You should get a lot more with:
>>>>
>>>> CONFIG_DYNAMIC_DEBUG=y
>>>>
>>>> and kernel commandline option:
>>>>
>>>> dyndbg="file drivers/mmc/core/* +p;file drivers/mmc/host/* +p"
>>>
>>> Unfortunatelly not:
>>
>> Are you sure you have debug messages enabled with your current console
>> loglevel? Might want to add "debug" at the end of your kernel command
>> line and try again.
> 
> Ok, so indeed debug was required, here is the huge bootlog now:
> https://gist.github.com/robimarko/e370bce66d0d2e7e54a2f5daf9784ee4

Thanks for the log!  It shows everything is OK up until the first
(HS200) tuning.

sdhci-msm takes the clock frequency into account when setting UHS
signaling, refer sdhci_msm_set_uhs_signaling(), so that is
presumably why the UHS signalling needs to be re-done even if
only the clock frequency changes.

It would be possible to change sdhci-msm to hook ->set_ios() and
set host->reinit_uhs before calling sdhci_set_ios(), which should
put back the original behaviour for sdhci-msm.

However "mmc: sdhci: Avoid unnecessary re-configuration" is really
a "nice-to-have" and it is not impossible other drivers are affected
by something similar, but just haven't noticed.

Consequently, I tend to think we should just drop the patch
"mmc: sdhci: Avoid unnecessary re-configuration" ?