Re: [PATCH 00/16] omap_hsmmc: Add ADMA support and UHS/HS200/DDR support

[Ulf Hansson]

+Adrian

On 20 July 2017 at 15:36, Kishon Vijay Abraham I <kishon@xxxxxx> wrote:
> Hi Ulf,
>
> On Tuesday 11 July 2017 07:27 PM, Kishon Vijay Abraham I wrote:
>> Hi Ulf,
>>
>> On Tuesday 11 July 2017 06:40 PM, Ulf Hansson wrote:
>>> On 16 June 2017 at 14:45, Kishon Vijay Abraham I <kishon@xxxxxx> wrote:
>>>> This is the final part of the series originally sent as
>>>> part of [2].
>>>>
>>>> This series adds UHS, HS200, DDR mode and ADMA support to
>>>> omap_hsmmc driver used to improve the throughput of MMC/SD in dra7
>>>> SoCs.
>>>>
>>>> Changes from [2]:
>>>> *) No more updating omap2plus_defconfig or multi_v7_defconfig is
>>>>    required, so those patches are removed.
>>>> *) Addressed Rob Herring's comment on implementing a function
>>>>    instead of having a macro while getting pinctrl state.
>>>>
>>>> This series is created on top of [3], [4], [5] AND [6]
>>>> (i.e after
>>>>   ARM: dts: omap3-overo: Remove "vqmmc-supply" property from MMC dt node
>>>>   omap_hsmmc: use mmc_regulator_get_supply() to get regulators
>>>>   omap*: Fixes/Cleanups for MMC devicetree node
>>>>   ARM: dts: Add iodelay data for MMC)
>>>>
>>>> The functionality implemented in this series was sent before ([1]) but
>>>> was never followed up since supporting high speed modes in dra7 required
>>>> IODelay values to be configured. With IODelay driver being merged into
>>>> kernel, sending it as a fresh series with support for configuring IODelay
>>>> values.
>>>
>>> Is it safe to queue this via mmc tree for 4.14 or is there a
>>> dependency I must consider? The above didn't quite tell me that, could
>>> you please elaborate.
>>
>> There is a dependency with https://www.spinics.net/lists/arm-kernel/msg586215.html.
>>
>> I'll resend the series after all the platform data specific stuff is merged.
>>>
>>>>
>>>> Suggestions of migrating to sdhci driver (from omap_hsmmc driver) is not
>>>> addressed since
>>>> *) tuning ratio of MMC in dra7 is different from sdhci
>>>> *) IOdelay is required for dra7
>>>> *) GPIO based card detect is not supported in sdhci
>>>
>>> Lots of sdhci drivers calls mmc_of_parse(), and uses the mmc slot gpio
>>> APIs, so I don't this this is correct statement.
>>>
>>>> *) Some of the registers don't always have correct values as in sdhci
>>>>    (like PSTATE).
>>>> Supporting all of these in sdhci will result in adding lot of quirks in
>>>> sdhci driver.
>>>
>>> Is it really that much different? It would be nice if some really took
>>> on the challenge of converting omap_hsmmc into an sdhci variant.
>>
>> Okay. I'll give that a shot and see how far I can get.
>
> I created a sdhci variant and was able to get it working with high speed mode
> in dra7.

That's very nice!

>
> I have a query on sdhci_set_power_noreg() in sdhci.c. sdhci_set_power_noreg
> sets SDHCI_POWER_CONTROL based on *vdd* value. However SDHCI_POWER_CONTROL is
> for setting bus voltage and need not be based on *vdd*. IMHO
> SDHCI_POWER_CONTROL should be set based on "ios->signal_voltage".

According to the implementation in sdhci_set_power_noreg(), I
understand it as the SDHCI_POWER_CONTROL register actually controls
the power to the card (VDD) and the not the I/O voltage level
(ios->signal_voltage).

However, if I read the SDHCI spec, it seems to be mentioning the "SD
bus voltage", which to me is about the I/O voltage level. So either
the spec is unclear or the code is wrong. :-)

Are you saying that the omap_hsmmc variant is using the
SDHCI_POWER_CONTROL to control the I/O voltage level?

I am wondering how other variants use it. Perhaps most variants
actually have an external regulator and thus calls
sdhci_set_power_reg() instead.

Looping in Adrian, to see if he has some thoughts on this.

Kind regards
Uffe