On Tue, 24 Oct 2023 at 01:47, Bjorn Andersson <quic_bjorande@xxxxxxxxxxx> wrote:
>
> On Tue, Oct 24, 2023 at 12:47:26AM +0300, Dmitry Baryshkov wrote:
> > On sevral Qualcomm platforms (SC8180X, SM8350, SC8280XP) a call to
> > UCSI_GET_PDOS for non-PD partners will cause a firmware crash with no
> > easy way to recover from it. Since we have no easy way to determine
> > whether the partner really has PD support, shortcut UCSI_GET_PDOS on
> > such platforms. This allows us to enable UCSI support on such devices.
> >
>
> Really nice to see this. Thanks.
>
> > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxx>
> > ---
> > drivers/usb/typec/ucsi/ucsi.c | 3 +++
> > drivers/usb/typec/ucsi/ucsi.h | 3 +++
> > drivers/usb/typec/ucsi/ucsi_glink.c | 3 +++
> > 3 files changed, 9 insertions(+)
> >
> > diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c
> > index 61b64558f96c..5392ec698959 100644
> > --- a/drivers/usb/typec/ucsi/ucsi.c
> > +++ b/drivers/usb/typec/ucsi/ucsi.c
> > @@ -578,6 +578,9 @@ static int ucsi_read_pdos(struct ucsi_connector *con,
> > u64 command;
> > int ret;
> >
> > + if (ucsi->quirks & UCSI_NO_PARTNER_PDOS)
> > + return 0;
> > +
> > command = UCSI_COMMAND(UCSI_GET_PDOS) | UCSI_CONNECTOR_NUMBER(con->num);
> > command |= UCSI_GET_PDOS_PARTNER_PDO(is_partner);
> > command |= UCSI_GET_PDOS_PDO_OFFSET(offset);
> > diff --git a/drivers/usb/typec/ucsi/ucsi.h b/drivers/usb/typec/ucsi/ucsi.h
> > index 474315a72c77..6478016d5cb8 100644
> > --- a/drivers/usb/typec/ucsi/ucsi.h
> > +++ b/drivers/usb/typec/ucsi/ucsi.h
> > @@ -317,6 +317,9 @@ struct ucsi {
> > #define EVENT_PENDING 0
> > #define COMMAND_PENDING 1
> > #define ACK_PENDING 2
> > +
> > + unsigned long quirks;
> > +#define UCSI_NO_PARTNER_PDOS BIT(0) /* Don't read partner's PDOs */
> > };
> >
> > #define UCSI_MAX_SVID 5
> > diff --git a/drivers/usb/typec/ucsi/ucsi_glink.c b/drivers/usb/typec/ucsi/ucsi_glink.c
> > index db6e248f8208..5c159e7b2b65 100644
> > --- a/drivers/usb/typec/ucsi/ucsi_glink.c
> > +++ b/drivers/usb/typec/ucsi/ucsi_glink.c
> > @@ -327,6 +327,8 @@ static int pmic_glink_ucsi_probe(struct auxiliary_device *adev,
> > if (ret)
> > return ret;
> >
> > + ucsi->ucsi->quirks = id->driver_data;
> > +
> > ucsi_set_drvdata(ucsi->ucsi, ucsi);
> >
> > device_for_each_child_node(dev, fwnode) {
> > @@ -379,6 +381,7 @@ static void pmic_glink_ucsi_remove(struct auxiliary_device *adev)
> >
> > static const struct auxiliary_device_id pmic_glink_ucsi_id_table[] = {
> > { .name = "pmic_glink.ucsi", },
> > + { .name = "pmic_glink.ucsi-no-pdos", .driver_data = UCSI_NO_PARTNER_PDOS, },
>
> In altmode and battmgr drivers we apply quirks based on the compatible
> of the pmic_glink of_node.
... and I can't say that I like that. In typical drivers we perform
driver tuning by looking at the device's data (e.g. by using
of_device_is_compatible or by of_device_get_match_data. Checking the
parent device seems like breaking the layering.
But if you insist, I can follow that approach.
>
> Could we do the same here, instead of mixing the two schemes?
>
> Regards,
> Bjorn
>
> > {},
> > };
> > MODULE_DEVICE_TABLE(auxiliary, pmic_glink_ucsi_id_table);
> > --
> > 2.42.0
> >
--
With best wishes
Dmitry
