Re: [PATCH V2 1/2] mmc: sdhci-msm: Add support to store supported vdd-io voltages

[Doug Anderson]

Hi,

On Tue, Mar 6, 2018 at 11:13 PM, Vijay Viswanath
<vviswana@xxxxxxxxxxxxxx> wrote:
> Hi Dough, Jeremy,
>
>
> On 3/3/2018 4:38 AM, Jeremy McNicoll wrote:
>>
>> On 2018-03-02 10:23 AM, Doug Anderson wrote:
>>>
>>> Hi,
>>>
>>> On Sun, Feb 11, 2018 at 10:01 PM, Vijay Viswanath
>>> <vviswana@xxxxxxxxxxxxxx> wrote:
>>>>
>>>> During probe check whether the vdd-io regulator of sdhc platform device
>>>> can support 1.8V and 3V and store this information as a capability of
>>>> platform device.
>>>>
>>>> Signed-off-by: Vijay Viswanath <vviswana@xxxxxxxxxxxxxx>
>>>> ---
>>>>   drivers/mmc/host/sdhci-msm.c | 38
>>>> ++++++++++++++++++++++++++++++++++++++
>>>>   1 file changed, 38 insertions(+)
>>>>
>>>> diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c
>>>> index c283291..5c23e92 100644
>>>> --- a/drivers/mmc/host/sdhci-msm.c
>>>> +++ b/drivers/mmc/host/sdhci-msm.c
>>>> @@ -23,6 +23,7 @@
>>>>   #include <linux/iopoll.h>
>>>>
>>>>   #include "sdhci-pltfm.h"
>>>> +#include <linux/regulator/consumer.h>
>>>
>>>
>>> This is a strange sort order for this include file.  Why is it after
>>> the local include?
>>>
>>>
>>>>   #define CORE_MCI_VERSION               0x50
>>>>   #define CORE_VERSION_MAJOR_SHIFT       28
>>>> @@ -81,6 +82,9 @@
>>>>   #define CORE_HC_SELECT_IN_HS400        (6 << 19)
>>>>   #define CORE_HC_SELECT_IN_MASK (7 << 19)
>>>>
>>>> +#define CORE_3_0V_SUPPORT      (1 << 25)
>>>> +#define CORE_1_8V_SUPPORT      (1 << 26)
>>>> +
>>>
>>>
>>> Is there something magical about 25 and 26?  This is a new caps field,
>>> so I'd have expected 0 and 1.
>>>
>>>
>
> Yes, these bits are the same corresponding to the capabilities in the
> Capabilities Register (offset 0x40). The bit positions become important when
> capabilities register doesn't show support to some voltages, but we can
> support those voltages. At that time, we will have to fake capabilities. The
> changes for those are currently not yet pushed up.
>
>
>>>>   #define CORE_CSR_CDC_CTLR_CFG0         0x130
>>>>   #define CORE_SW_TRIG_FULL_CALIB                BIT(16)
>>>>   #define CORE_HW_AUTOCAL_ENA            BIT(17)
>>>> @@ -148,6 +152,7 @@ struct sdhci_msm_host {
>>>>          u32 curr_io_level;
>>>>          wait_queue_head_t pwr_irq_wait;
>>>>          bool pwr_irq_flag;
>>>> +       u32 caps_0;
>>>>   };
>>>>
>>>>   static unsigned int msm_get_clock_rate_for_bus_mode(struct sdhci_host
>>>> *host,
>>>> @@ -1313,6 +1318,35 @@ static void sdhci_msm_writeb(struct sdhci_host
>>>> *host, u8 val, int reg)
>>>>                  sdhci_msm_check_power_status(host, req_type);
>>>>   }
>>>>
>>>> +static int sdhci_msm_set_regulator_caps(struct sdhci_msm_host
>>>> *msm_host)
>>>> +{
>>>> +       struct mmc_host *mmc = msm_host->mmc;
>>>> +       struct regulator *supply = mmc->supply.vqmmc;
>>>> +       int i, count;
>>>> +       u32 caps = 0, vdd_uV;
>>>> +
>>>> +       if (!IS_ERR(mmc->supply.vqmmc)) {
>>>> +               count = regulator_count_voltages(supply);
>>>> +               if (count < 0)
>>>> +                       return count;
>>>> +               for (i = 0; i < count; i++) {
>>>> +                       vdd_uV = regulator_list_voltage(supply, i);
>>>> +                       if (vdd_uV <= 0)
>>>> +                               continue;
>>>> +                       if (vdd_uV > 2700000)
>>>> +                               caps |= CORE_3_0V_SUPPORT;
>>>> +                       if (vdd_uV < 1950000)
>>>> +                               caps |= CORE_1_8V_SUPPORT;
>>>> +               }
>>>
>>>
>>> Shouldn't you be using regulator_is_supported_voltage() rather than
>>> open coding?  Also: I've never personally worked on a device where it
>>> was used, but there is definitely a concept floating about of a
>>> voltage level of 1.2V.  Maybe should copy the ranges from
>>> mmc_regulator_set_vqmmc()?
>>>
>>>
>
> regulator_is_supported_voltage() checks for a range and it also uses
> regulator_list_voltage() internally. regulator_list_voltage() is also an
> exported API for use by drivers AFAIK. Please correct if it is not.

Sure, regulator_list_voltage() is valid to call.  I'm not saying that
your code is wrong or violates abstractions, just that it's
essentially re-implementing regulator_is_supported_voltage() for very
little gain.  Calling regulator_is_supported_voltage() is better
because:

1. In theory, it should generate less code.  Sure, it might loop twice
with the current implementation of regulator_is_supported_voltage(),
but for a non-time-critical section like this smaller code is likely
better than faster code (decreases kernel size / uses up less cache
space, etc).

2. If regulator_is_supported_voltage() is ever improved to be more
efficient you'll get that improvement automatically.  If someone
happened to source vqmmc from a PWM regulator, for instance, trying to
enumerate all voltages like this would be a disaster.

3. Code will be simpler to understand.

You can replace your whole loop with:

if (regulator_is_supported_voltage(mmc->supply.vqmmc, 1700000, 1950000))
  caps |= CORE_1_8V_SUPPORT
if (regulator_is_supported_voltage(mmc->supply.vqmmc, 2700000, 3600000))
  caps |= CORE_3_0V_SUPPORT


>>> Also: seems like you should have some way to deal with "caps" ending
>>> up w/ no bits set.  IIRC you can have a regulator that can be enabled
>>> / disabled but doesn't list a voltage, so if someone messed up their
>>> device tree you could end up in this case.  Should you print a
>>> warning?  ...or treat it as if we support "3.0V"?  ...or ?  I guess it
>>> depends on how do you want patch #2 to behave in that case.
>>
>>
>> Both, initialize it to sane value and print something.  This way at
>> least you have a good chance of booting and not hard hanging and you
>> are given a reasonable message indicating what needs to be fixed.
>>
>> -jeremy
>>
>>>
>>>
>>>> +       }
>>>
>>>
>>> How should things behave if vqmmc is an error?  In that case is it
>>> important to not set "CORE_IO_PAD_PWR_SWITCH_EN" in patch set #2?
>>> ...or should you set "CORE_IO_PAD_PWR_SWITCH_EN" but then make sure
>>> you don't set "CORE_IO_PAD_PWR_SWITCH"?
>>>
>>>
>
> Thanks for the suggestion. If the regulators exit and doesn't list the
> voltages, then I believe initialization itself will not happen. We will not
> have any available ocr and in sdhci_setup_host it should fail.
> But these enhancements can be incorporated. Since this patch is already
> acknowledged, I will incorporate these changes in a subsequent patch.

It's already acknowledged?  I saw that your RFC was acknowledged by
Adrian Hunter but then you didn't include that tag in the posting of
v2, so I assumed for some reason it no longer applied.  If you're
thinking that Ulf would be the one to apply this patch, he probably
doesn't know that it's Acked either.

Perhaps Adrian or Ulf can give direction for how they see this patch proceeding.


>>>> +       msm_host->caps_0 |= caps;
>>>> +       pr_debug("%s: %s: supported caps: 0x%08x\n", mmc_hostname(mmc),
>>>> +                       __func__, caps);
>>>> +
>>>> +       return 0;
>>>> +}
>>>> +
>>>> +
>>>>   static const struct of_device_id sdhci_msm_dt_match[] = {
>>>>          { .compatible = "qcom,sdhci-msm-v4" },
>>>>          {},
>>>> @@ -1530,6 +1564,10 @@ static int sdhci_msm_probe(struct platform_device
>>>> *pdev)
>>>>          ret = sdhci_add_host(host);
>>>>          if (ret)
>>>>                  goto pm_runtime_disable;
>>>> +       ret = sdhci_msm_set_regulator_caps(msm_host);
>>>> +       if (ret)
>>>> +               dev_err(&pdev->dev, "%s: Failed to set regulator caps:
>>>> %d\n",
>>>> +                               __func__, ret);
>>>
>>>
>>> Why do you need __func__ here?  You're already using dev_err(), that
>>> gives an idea of where we are.
>>>
>
> dev_err() doesn't give information of where it is getting called.

It gives you the driver and the error message should be unique to the
driver and easy to find.  Including "__func__ in messages like this is
discouraged unless you are in a context where you somehow can't get
access to the device pointer.  I suppose ultimately it's up the the
maintainer for individual cases but overall I've seen this to be a
consistently applied rule in the kernel.

In any case, why would this particular print be special that it should
include __func__ but all others (in this file, or in dev_err in
general) shouldn't?


>>>>          pm_runtime_mark_last_busy(&pdev->dev);
>>>>          pm_runtime_put_autosuspend(&pdev->dev);
>>>> --
>>>>   Qualcomm India Private Limited, on behalf of Qualcomm Innovation
>>>> Center, Inc.
>>>> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a
>>>> Linux Foundation Collaborative Project.
>>>>
>>>> --
>>>> To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
>>>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>
> Thanks,
> Vijay