Am Mittwoch, 15. August 2018, 12:34:42 CEST schrieb Enric Balletbo i Serra:

Hi Heiko,

On 15/08/18 12:18, Heiko Stuebner wrote:

quoted

Hi Enric,

Am Mittwoch, 15. August 2018, 11:59:33 CEST schrieb Enric Balletbo i Serra:

quoted

The OTG disconnection event is generated after the presence/abscense of
an ID connection, but some platforms doesn't have the ID pin connected, so
the event is not generated. In such case, for detecting the disconnection
event, we can get the cable state from an extcon driver. We need, though,
to force to set the B-Device Session Valid bit on the PHY to have the
device respond to setup address. Otherwise, the following error is
shown:

    usb 2-2: Device not responding to setup address.
    usb 2-2: device not accepting address 14, error -71
    usb usb2-port2: unable to enumerate USB device

The patch allows to tell the PHY to force the B-Device Session Valid bit
when the OTG role is device and clear that bit if the OTG role is host.

Signed-off-by: Enric Balletbo i Serra <redacted>

quoted

@@ -1024,6 +1051,12 @@ static int rockchip_usb2phy_otg_port_init(struct rockchip_usb2phy *rphy,
 	INIT_DELAYED_WORK(&rport->chg_work, rockchip_chg_detect_work);
 	INIT_DELAYED_WORK(&rport->otg_sm_work, rockchip_usb2phy_otg_sm_work);
 
+	rport->force_bvalid = false;
+	if (of_device_is_compatible(rphy->dev->of_node,
+				    "rockchip,rk3399-usb2phy"))
+		rport->force_bvalid = of_property_read_bool(child_np,
+						"rockchip,force-bvalid");

That feels a bit clumsy, especially as the rk3399 seems to have the id

Agree, It is, let me explain :)

Ideally we shouldn't have this check, to get rid of this check I only need the
offsets for bvalid_session register for all the compatibles and fill in phy
configuration data (rk3228_phy_cfgs, &rk3328_phy_cfgs, rk3366_phy_cfgs,
rv1108_phy_cfgs)

To be honest I didn't look if all the datasheets are public available, let me do
some research. Or, if anyone has the datasheet and can tell where the
bvalid_session bit is I can fill all the data.

Or just always check for the presence of the property and just make the
driver warn if that bvalid_session setting is not available for that soc yet.

From the TRMs I have your rk3399 bvalid_session settings seems to live

in the depths of the undocumented inno-phy regs, so that will probably
be the same for other socs using that phy.


Heiko