Re: MMC error on Exynos4210 board

[Jaehoon Chung <jh80.chung@samsung.com>]

On 06/19/2014 07:40 PM, Sachin Kamat wrote:
> On Thu, Jun 19, 2014 at 2:40 PM, Tim Kryger <tim.kryger@gmail.com> wrote:
>> On Thu, Jun 19, 2014 at 1:49 AM, Sachin Kamat <spk.linux@gmail.com> wrote:
>>> +cc Some relevant guys from Samsung
>>>
>>> On Thu, Jun 19, 2014 at 2:12 PM, Tim Kryger <tim.kryger@gmail.com> wrote:
>>>> On Wed, Jun 18, 2014 at 8:54 PM, Sachin Kamat <spk.linux@gmail.com> wrote:
>>>>
>>>>>> On Wed, Jun 18, 2014 at 4:33 AM, Sachin Kamat <spk.linux@gmail.com> wrote:
>>>>
>>>>>>> I see the below error on Exynos4210 based Origen board with linux-next
>>>>>>> (20140618).
>>>>>>> Reverting the below commit works fine.
>>>>>>>
>>>>>>> Commit: 8d02e775a6 "mmc: sdhci: Use mmc core regulator infrastucture"
>>>>
>>>>>>>
>>>>>>> -- [    2.068992] sdhci: Secure Digital Host Controller Interface driver
>>>>>>> [    2.075059] sdhci: Copyright(c) Pierre Ossman
>>>>>>> [    2.079762] of_get_named_gpiod_flags: can't parse gpios property of
>>>>>>> node '/sdhci@12510000[0]'
>>>>>>> [    2.088021] s3c-sdhci 12510000.sdhci: clock source 2: mmc_busclk.2
>>>>>>> (50000000 Hz)
>>>>>>> [    2.095322] of_get_named_gpiod_flags: can't parse gpios property of
>>>>>>> node '/sdhci@12510000[0]'
>>>>>>> [    2.103794] of_get_named_gpiod_flags: can't parse gpios property of
>>>>>>> node '/sdhci@12510000[0]'
>>>>>>> [    2.112478] s3c-sdhci 12510000.sdhci: No vqmmc regulator found
>>>>>>> [    2.118117] mmc0: Hardware doesn't report any support voltages.
>>>>>>> [    2.124004] s3c-sdhci 12510000.sdhci: sdhci_add_host() failed
>>>>>>> [    2.130080] of_get_named_gpiod_flags: can't parse gpios property of
>>>>>>> node '/sdhci@12530000[0]'
>>>>>>> [    2.138352] s3c-sdhci 12530000.sdhci: clock source 2: mmc_busclk.2
>>>>>>> (16666667 Hz)
>>>>>>> [    2.145661] of_get_named_gpiod_flags: can't parse gpios property of
>>>>>>> node '/sdhci@12530000[0]'
>>>>>>> [    2.154139] of_get_named_gpiod_flags: can't parse gpios property of
>>>>>>> node '/sdhci@12530000[0]'
>>>>>>> [    2.162834] s3c-sdhci 12530000.sdhci: No vqmmc regulator found
>>>>>>> [    2.168464] mmc0: Hardware doesn't report any support voltages.
>>>>>>> [    2.174349] s3c-sdhci 12530000.sdhci: sdhci_add_host() failed
>>>>
>>>>>>> [    2.336148] Waiting for root device /dev/mmcblk0p1...
>>>>
>>>>> FYI, the board has a 2.8V fixed regulator supply connected to the MMC.
>>>>> You may refer to arch/arm/boot/dts/exynos4210-origen.dts for more details.
>>>>
>>>> A 2.8v regulator results in mmc->ocr_avail being set to MMC_VDD_27_28
>>>> | MMC_VDD_28_29.
>>>>
>>>> The SDHCI capabilities register only indicates support of three voltage levels
>>>>   - 1.8v: SDHCI_CAN_VDD_180 => MMC_VDD_165_195
>>>>   - 3.0v: SDHCI_CAN_VDD_300 => MMC_VDD_29_30 | MMC_VDD_30_31
>>>>   - 3.3v: SDHCI_CAN_VDD_330 => MMC_VDD_32_33 | MMC_VDD_33_34

Right. sdhci capabilities only indicated them.
But I think SoC can be support the specific VDD range.

>>>>
>>>> Even if all capability bits of the host controller were set, there
>>>> still wouldn't be any overlap.  Thus you see a "Hardware doesn't
>>>> report any support voltages" message.
>>>>
>>>> Previously, this issue was being swept under the rug by cec2e21 mmc:
>>>> sdhci: Use regulator min/max voltage range according to spec.  That
>>>> change hacked up the voltage range checks such that with your 2.8v
>>>> fixed regulator, the driver would believe the host could support
>>>> MMC_VDD_29_30 | MMC_VDD_30_31 | MMC_VDD_32_33 | MMC_VDD_33_34.  The
>>>> driver would start down the path of commanding 3.3v-3.4v (the highest
>>>> voltage range believed to be supported).  At the last second, the
>>>> driver would see the regulator was fixed and blindly skip over the set
>>>> voltage operation, saving it from failure.
>>>>
>>>> Since my patch eliminates the bogus voltage range checks, your board
>>>> is now getting caught playing too loose with the SDHCI regulator
>>>> voltages.
>>>>
>>>> Furthermore, the fixed regulator special-case logic that helped hide
>>>> your issue should also be considered for removal given that fixed
>>>> regulators now behave properly thanks to c00dc35 regulator: core:
>>>> Allow regulator_set_voltage for fixed regulators.
>>>
>>> Thanks for the detailed explanation. What do you propose to get this fixed?
>>
>> I'm not really sure of the best path forward.  I suppose you could
>> modify your device tree to lie about the voltage of the fixed
>> regulator.  Changing it to 3.0v should allow it to boot up but that is
>> definitely a hack.
I don't want to change the 3.0V at dt file...is it hack? i don't think so.
Almost exynos4 Series is used the fixed regulator(2.8V).
I didn't know exactly why we used the 2.8V. But i guess there is some reason.(H/W design).

If we need to change ocr value, 
then i will add the quirk or controlling function for exynos into sdhci-s3c.c
How about?

> 
> Or I could simply remove the vmmc-supply property altogether (as it is optional)
> and get the board to work.
Then is it supported the power "always-on"?
> 
>> It would be nice if the driver could be extended
>> to handle the peculiarities of your board in a deliberate manner but
>> limiting the common sdhci driver to supporting only the three voltages
>> from the spec also seems sensible.
> 
> Until such time that the driver gets fixed to handle 2.8V fixed supply your
> current patch leaves several of Exynos boards broken for now.

the all of exynos used the fixed-regulator(2.8v) should be broken.
(Maybe exynos4 series??)

>