Re: [PATCH V2 3/5] mmc: mmci: fix clear of busy detect status

[Ludovic BARRE <ludovic.barre@st.com>]

hi Ulf

On 5/27/19 8:17 PM, Ulf Hansson wrote:
> On Fri, 26 Apr 2019 at 09:46, Ludovic Barre <ludovic.Barre@st.com> wrote:
>>
>> From: Ludovic Barre <ludovic.barre@st.com>
>>
>> The "busy_detect_flag" is used to read/clear busy value of
>> mmci status. The "busy_detect_mask" is used to manage busy irq of
>> mmci mask.
>> For sdmmc variant, the 2 properties have not the same offset.
>> To clear the busyd0 status bit, we must add busy detect flag,
>> the mmci mask is not enough.
>>
>> Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
> 
> Ludovic, again, apologies for the delay.
> 
>> ---
>>   drivers/mmc/host/mmci.c | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/mmc/host/mmci.c b/drivers/mmc/host/mmci.c
>> index a040f54..3cd52e8 100644
>> --- a/drivers/mmc/host/mmci.c
>> +++ b/drivers/mmc/host/mmci.c
>> @@ -1517,7 +1517,8 @@ static irqreturn_t mmci_irq(int irq, void *dev_id)
>>                   * to make sure that both start and end interrupts are always
>>                   * cleared one after the other.
>>                   */
>> -               status &= readl(host->base + MMCIMASK0);
>> +               status &= readl(host->base + MMCIMASK0) |
>> +                       host->variant->busy_detect_flag;
> 
> I think this is not entirely correct, because it would mean we check
> for busy even if we haven't unmasked the busy IRQ via the
> variant->busy_detect_mask.

if the variant is busy_detect false:
  => no problem because the busy_detect_flag or busy_detect_mask is not
     defined.

if variant is busy_detect true:
the busy handle is split in 3 steps (see mmci_cmd_irq):
step 1: detection of busy line => unmasked the busy irq end
step 2: in busy wait => ignore cmd irq while current busy flag is
enabled.
step 3: end of busy => clear and mask busy irq

To detect the first step (see mmci_cmd_irq: which unmasks the busy irq)
we need to know the current busy state. Actually, the status register is
re-read in mmci_cmd_irq, why not used the status read in mmci_irq and in
parameter ?

Regards,
Ludo

> 
> I suggest to store a new bool in the host (call it
> "busy_detect_unmasked" or whatever makes sense to you), to track
> whether we have unmasked the busy IRQ or not. Then take this flag into
> account, before ORing the value of host->variant->busy_detect_flag,
> according to above.
> 
>>                  if (host->variant->busy_detect)
>>                          writel(status & ~host->variant->busy_detect_mask,
>>                                 host->base + MMCICLEAR);
>> --
>> 2.7.4
>>
> 
> Kind regards
> Uffe
>