Re: [PATCH] mmc: host: sdhci: Fix the incorrect soft reset operation when runtime resuming

[Adrian Hunter]

On 15/07/19 2:37 PM, Baolin Wang wrote:
> Hi Adrian,
> 
> On Mon, 15 Jul 2019 at 19:20, Adrian Hunter <adrian.hunter@xxxxxxxxx> wrote:
>>
>> On 15/07/19 1:58 PM, Baolin Wang wrote:
>>> In sdhci_runtime_resume_host() function, we will always do software reset
>>> for all, but according to the specification, we should issue reset command
>>> and reinitialize the SD/eMMC card.
>>
>> Where does it say that?
> 
> I checked the SD host controller simplified specification Ver4.20, and
> in Page 75, Software Reset For All bit, it says "if this bit is set
> to1, the host driver should issue reset command and  reinitialize the
> SD card". (I did not check other versions).

That might simply be assuming that the bus power also controls the card power.

> 
>>
>>>                                    However, we only do reinitialize the
>>> SD/eMMC card when the SD/eMMC card are power down during runtime suspend.
>>>
>>> Thus for those platforms that do not power down the SD/eMMC card during
>>> runtime suspend, we should not do software reset for all.
>>>                                                           To fix this
>>> issue, we can add one condition to validate the MMC_CAP_AGGRESSIVE_PM
>>> to decide if we can do software reset for all or just reset command
>>> and data lines.
>>>
>>> Signed-off-by: Baolin Wang <baolin.wang@xxxxxxxxxx>
>>> ---
>>>  drivers/mmc/host/sdhci.c |    2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/mmc/host/sdhci.c b/drivers/mmc/host/sdhci.c
>>> index 9715834..470c5e0 100644
>>> --- a/drivers/mmc/host/sdhci.c
>>> +++ b/drivers/mmc/host/sdhci.c
>>> @@ -3333,7 +3333,7 @@ int sdhci_runtime_resume_host(struct sdhci_host *host)
>>>                       host->ops->enable_dma(host);
>>>       }
>>>
>>> -     sdhci_init(host, 0);
>>> +     sdhci_init(host, !(mmc->caps & MMC_CAP_AGGRESSIVE_PM));
>>
>> We have done a full reset for a long time, so it would be surprising to need
>> to change it.
>>
>> What problem is it causing?
> 
> If we did not power down the SD card during runtime suspend, and we
> reset for all when runtime resume, our SD host controller can not work
> well, will meet some strange behavior, like:
> 
> [    6.525397] mmc0: Got data interrupt 0x00000002 even though no data
> operation was in progress.
> [    6.534189] mmc0: sdhci: ============ SDHCI REGISTER DUMP ===========
> [    6.540797] mmc0: sdhci: Sys addr:  0x00000008 | Version:  0x00000004
> [    6.547413] mmc0: sdhci: Blk size:  0x00000200 | Blk cnt:  0x00000000
> [    6.554029] mmc0: sdhci: Argument:  0x03200101 | Trn mode: 0x00000033
> [    6.560645] mmc0: sdhci: Present:   0x01f000f0 | Host ctl: 0x00000030
> [    6.567262] mmc0: sdhci: Power:     0x00000000 | Blk gap:  0x00000000
> [    6.573877] mmc0: sdhci: Wake-up:   0x00000000 | Clock:    0x00000007
> [    6.580493] mmc0: sdhci: Timeout:   0x0000000e | Int stat: 0x00000000
> [    6.587109] mmc0: sdhci: Int enab:  0x037f000b | Sig enab: 0x037f000b
> [    6.593726] mmc0: sdhci: ACmd stat: 0x00000000 | Slot int: 0x00000000
> [    6.600342] mmc0: sdhci: Caps:      0x1c6d0080 | Caps_1:   0x08000007
> [    6.606959] mmc0: sdhci: Cmd:       0x0000061b | Max curr: 0x00ffffff
> [    6.613574] mmc0: sdhci: Resp[0]:   0x00001201 | Resp[1]:  0x00000000
> [    6.620190] mmc0: sdhci: Resp[2]:   0x00000000 | Resp[3]:  0x00000000
> [    6.626806] mmc0: sdhci: Host ctl2: 0x00003807
> [    6.631364] mmc0: sdhci: ADMA Err:  0x00000000 | ADMA Ptr: 0x00000000df062000
> [    6.638697] mmc0: sdhci: ============================================
> [    6.645379] mmc0: cache flush error -84
> 
> Got data interrupt but no data commands are processing now. With this
> patch, then our SD host controller can work well. Did I miss anything
> else? Thanks.

The response seems to show the card in state 9 bus-testing, which would
suggest the use of CMD19 for eMMC.  Perhaps the wrong command is used for
eMMC re-tuning?

The difficulty with changing long standing flow is that it might reveal
problems for other existing hardware.  Did you consider making a
driver-specific change?  The ->reset() callback could be used.