[U-Boot] [PATCH v2 1/3] mmc: sdhci: fix the wrong operation when response type is R1b

[Lei Wen]

Hi Jaehoon,

On Fri, Mar 30, 2012 at 12:36 PM, Jaehoon Chung <jh80.chung at samsung.com> wrote:
> Hi Lei.
>
> First, thanks for implemented the generic sdhci controller.

It is my pleasure to share this common code, and I'm glad that it is
used for other platforms now. :)

>
> On 03/30/2012 12:33 PM, Lei Wen wrote:
>
>> Hi Jaehoon,
>>
>> On Fri, Mar 30, 2012 at 10:39 AM, Jaehoon Chung <jh80.chung at samsung.com> wrote:
>>> When response type is R1b, mask value is added the SDHCI_INT_DAT_END.
>>> but in while(), didn't check that flag.
>>> So sdhci controller didn't work fine.
>>> CMD6 didn't always complete.
>>
>> Could you elaborate it more in details?
>>         do {
>>                 stat = sdhci_readl(host, SDHCI_INT_STATUS);
>>                 if (stat & SDHCI_INT_ERROR)
>>                         break;
>>         } while ((stat & mask) != mask);
>> Here in the while condition, if the status read out don't contain all mask,
>> then the looping would continue.
>> Do you mean you just need a retry max time set here?
>
> I found that didn't initialize the eMMC card.
> Because when send CMD6, running infinite loop in there.
> CMD6's mask is set to SDHCI_INT_RESPONSE and SDHCI_INT_DATA_END.
> (Because response type is R1B).
> Then mask value maybe is 0x3...
> stat = sdhci_readl(host, SDHCI_INT_STATUS);
> stat is 0x1.(cmd is done response).
> but in while(), stat & mask is 0x1, and mask is 0x3.
> ...doing while().
> If just add the timeout, then always produced the timeout error.

I see your problems. Your silicon only provide SDHCI_INT_RESPONSE for
r1b, while the standard
sdhci spec require both the flag is set when the r1b command is finished.
So my point is you could add a quirk condition check in the previously
endless loop checking.
If the quirk is true, and timeout count is met, then you could jump
out of original looping.
And for later checking, this timeout could be regarded as no error,
since the controller would never
return the SDHCI_INT_DATA_END in your case.

>
> How did you test? Is initialize the eMMC card?
> (I tested with eMMC4.41 (exynos4).)

I test on many Marvell platforms, with sd, mmc, emmc, they all works fine.

>
>>
>>
>>>
>>> Signed-off-by: Jaehoon Chung <jh80.chung at samsung.com>
>>> Signed-off-by: Kyungmin Park <kyungmin.park at samsung.com>
>>> ---
>>>  drivers/mmc/sdhci.c |   33 +++++++++++++++++++++++----------
>>>  1 files changed, 23 insertions(+), 10 deletions(-)
>>>
>>> diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c
>>> index fc904b5..0dd08b9 100644
>>> --- a/drivers/mmc/sdhci.c
>>> +++ b/drivers/mmc/sdhci.c
>>> @@ -124,10 +124,11 @@ int sdhci_send_command(struct mmc *mmc, struct mmc_cmd *cmd,
>>>  {
>>>        struct sdhci_host *host = (struct sdhci_host *)mmc->priv;
>>>        unsigned int stat = 0;
>>> -       int ret = 0;
>>> +       int i, ret = 0;
>>>        int trans_bytes = 0, is_aligned = 1;
>>>        u32 mask, flags, mode;
>>>        unsigned int timeout, start_addr = 0;
>>> +       unsigned int retry = 10000;
>>>
>>>        /* Wait max 10 ms */
>>>        timeout = 10;
>>> @@ -206,19 +207,31 @@ int sdhci_send_command(struct mmc *mmc, struct mmc_cmd *cmd,
>>>        flush_cache(start_addr, trans_bytes);
>>>  #endif
>>>        sdhci_writew(host, SDHCI_MAKE_CMD(cmd->cmdidx, flags), SDHCI_COMMAND);
>>> -       do {
>>> +
>>> +       for (i = 0; i < retry; i++) {
>>>                stat = sdhci_readl(host, SDHCI_INT_STATUS);
>>> -               if (stat & SDHCI_INT_ERROR)
>>> +               if (stat & (SDHCI_INT_RESPONSE | SDHCI_INT_DATA_END)) {
>>> +                       sdhci_cmd_done(host, cmd);
>>> +                       sdhci_writel(host, mask, SDHCI_INT_STATUS);
>>> +                       if (!data) {
>>> +                               sdhci_writel(host, stat,  SDHCI_INT_STATUS);
>>
>> Why do two write?
>
> We can remove the sdhci_writel(host, mask SDHCI_INT_STATUS).
> It's my mistake..i will fix that.
>
> Best Regards,
> Jaehoon Chung

Thanks,
Lei