[PATCH v2] phy: phy-imx8mq-usb: add vbus regulator support

[Tim Harvey]

On Mon, Jul 10, 2023 at 9:33 AM Marek Vasut <marex at denx.de> wrote:
>
> On 7/10/23 17:58, Tim Harvey wrote:
> > On Mon, Jul 10, 2023 at 8:44 AM Marek Vasut <marex at denx.de> wrote:
> >>
> >> On 7/10/23 17:26, Tim Harvey wrote:
> >>> On Sat, Jul 8, 2023 at 1:55 PM Marek Vasut <marex at denx.de> wrote:
> >>>>
> >>>> On 6/9/23 19:28, Tim Harvey wrote:
> >>>>> Add support for enabling and disabling vbus-supply regulator found
> >>>>> on several imx8mp boards in the usb3_phy0 and usb3_phy1 nodes.
> >>>>>
> >>>>> Signed-off-by: Tim Harvey <tharvey at gateworks.com>
> >>>>> Reviewed-by: Adam Ford <aford173 at gmail.com>
> >>>>> ---
> >>>>> v2:
> >>>>>     - protect ret with __maybe_unused in case CONFIG_CLK and
> >>>>>       CONFIG_DM_REGULATOR not defined
> >>>>>     - add error prints on failures
> >>>>>     - add Adam's rb tag
> >>>>> ---
> >>>>>     drivers/phy/phy-imx8mq-usb.c | 32 ++++++++++++++++++++++++++++++--
> >>>>>     1 file changed, 30 insertions(+), 2 deletions(-)
> >>>>>
> >>>>> diff --git a/drivers/phy/phy-imx8mq-usb.c b/drivers/phy/phy-imx8mq-usb.c
> >>>>> index 69f01de55538..53099436b04b 100644
> >>>>> --- a/drivers/phy/phy-imx8mq-usb.c
> >>>>> +++ b/drivers/phy/phy-imx8mq-usb.c
> >>>>> @@ -14,6 +14,7 @@
> >>>>>     #include <linux/delay.h>
> >>>>>     #include <linux/err.h>
> >>>>>     #include <clk.h>
> >>>>> +#include <power/regulator.h>
> >>>>>
> >>>>>     #define PHY_CTRL0                   0x0
> >>>>>     #define PHY_CTRL0_REF_SSP_EN                BIT(2)
> >>>>> @@ -81,6 +82,7 @@ struct imx8mq_usb_phy {
> >>>>>     #endif
> >>>>>         void __iomem *base;
> >>>>>         enum imx8mpq_phy_type type;
> >>>>
> >>>> Shouldn't this be in #if CONFIG_IS_ENABLED(DM_REGULATOR) ?
> >>>>
> >>>>> +     struct udevice *vbus_supply;
> >>>
> >>> Hi Marek,
> >>>
> >>> No, the usage of it is within an 'if (CONFIG_IS_ENABLED(DM_REGULATOR)
> >>> && imx_phy->vbus_supply)' statement
> >>
> >> Except if this is not ifdef'd out, the structure is larger for no good
> >> reason if DM_REGULATOR is DISABLED.
> >
> > ok, I see your point.
> >
> >>
> >>>>>     };
> >>>>>
> >>>>>     static const struct udevice_id imx8mq_usb_phy_of_match[] = {
> >>>>> @@ -172,10 +174,10 @@ static int imx8mq_usb_phy_power_on(struct phy *usb_phy)
> >>>>>     {
> >>>>>         struct udevice *dev = usb_phy->dev;
> >>>>>         struct imx8mq_usb_phy *imx_phy = dev_get_priv(dev);
> >>>>> +     __maybe_unused int ret;
> >>>>>         u32 value;
> >>>>>
> >>>>>     #if CONFIG_IS_ENABLED(CLK)
> >>>>> -     int ret;
> >>>>>         ret = clk_enable(&imx_phy->phy_clk);
> >>>>>         if (ret) {
> >>>>>                 printf("Failed to enable usb phy clock\n");
> >>>>> @@ -183,6 +185,14 @@ static int imx8mq_usb_phy_power_on(struct phy *usb_phy)
> >>>>>         }
> >>>>>     #endif
> >>>>>
> >>>>> +     if (CONFIG_IS_ENABLED(DM_REGULATOR) && imx_phy->vbus_supply) {
> >>>>> +             ret = regulator_set_enable(imx_phy->vbus_supply, true);
> >>>>> +             if (ret) {
> >>>>> +                     pr_err("Failed to enable VBUS regulator: %d\n", ret);
> >>>>
> >>>> You do need to disable clock enabled above.
> >>>
> >>> it does get disabled in imx8mq_usb_phy_power_off()
> >>
> >> If that called if power_on returns -ERRNO ?
> >
> > sorry, I don't quite understand.
> >
> > Are you asking me to add a regulator disable in a remove function to
> > make sure it's disabled if removed or are you saying that I shouldn't
> > return the error from regulator_set_enable from
> > imx8mq_usb_phy_power_on?
>
> Actually, neither. There is clk_enable() a bit higher up in this
> power_on function. You need to stop the clock if this
> regulator_set_enable() return -ERRNO.
>
> >>>> This likely needs a fail path with "err: ... return ret;"
> >>>
> >>> there are no further conditional return paths in this function.
> >>
> >> I suspect there should be.
> >
> > again, I don't understand. The only thing remaining in this function
> > is disable rx term override below so I don't see the point in adding a
> > goto around that if that is what you are suggesting.
> >
> > I don't really know what 'rx term override' is either... perhaps the
> > regulator_set_enable should go after that?
>
> Its only a matter of disabling the clock if the regulator enable fails
> and we bail out, that's all. Does it make sense now ?

Yes - that makes sense. Thanks for explaining... not sure how I
misunderstood that. You said 'clock' and I kept reading 'regulator' :)

Thanks,

Tim