Hi Ulf
On 01/17/2018 10:34 AM, Ulf Hansson wrote:
> [...]
>
>> /*
>> @@ -1616,6 +1625,32 @@ static int mmci_probe(struct amba_device *dev,
>> host = mmc_priv(mmc);
>> host->mmc = mmc;
>>
>> + /*
>> + * Some variant (STM32) doesn't have opendrain bit, nevertheless
>> + * pins can be set accordingly using pinctrl
>> + */
>> + if (!variant->opendrain) {
>> + host->pinctrl = devm_pinctrl_get(&dev->dev);
>> + if (IS_ERR(host->pinctrl)) {
>> + dev_err(&dev->dev, "failed to get pinctrl");
>> + goto host_free;
>> + }
>> +
>> + host->pins_default = pinctrl_lookup_state(host->pinctrl,
>> + PINCTRL_STATE_DEFAULT);
>> + if (IS_ERR(host->pins_default)) {
>> + dev_warn(mmc_dev(mmc), "Can't select default pins\n");
>> + host->pins_default = NULL;
>
> This is wrong, I think you should bail out and return the error code instead.
Ok
>
> Moreover, calling pinctrl_select_state() from ->set_ios by using a
> NULL state, will likely trigger a NULL pointer deference bug in the
> pinctrl layer.
Regarding pinctrl_select_state() call with a NULL state, this case is
managed inside pinctrl_state(), but ok, it will be more elegant to exit
directly in case of no DT pins definition found.
>
>> + }
>> +
>> + host->pins_opendrain = pinctrl_lookup_state(host->pinctrl,
>> + MMCI_PINCTRL_STATE_OPENDRAIN);
>> + if (IS_ERR(host->pins_opendrain)) {
>> + dev_warn(mmc_dev(mmc), "Can't select opendrain pins\n");
>> + host->pins_opendrain = NULL;
>
> Ditto.
ok
Thanks
Patrice
>
>> + }
>> + }
>> +
>
> [...]
>
> Kind regards
> Uffe
> ��.n��������+%������w��{.n�����{��i��)�� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
