On 24-10-17 08:00:44, Johan Hovold wrote: > On Tue, Oct 15, 2024 at 04:03:53PM +0300, Abel Vesa wrote: > > On 24-10-15 14:41:25, Johan Hovold wrote: > > > On Fri, Oct 04, 2024 at 04:57:36PM +0300, Abel Vesa wrote: > > > > The Parade PS8830 is a Type-C multi-protocol retimer that is controlled > > > > via I2C. It provides altmode and orientation handling and usually sits > > > > between the Type-C port and the PHY. > > > > > This retimer is a LTTPR (Link-Training Tunable PHY Repeater) which means > > > > it can support link training from source to itself. This means that the > > > > DP driver needs to be aware of the repeater presence and to handle > > > > the link training accordingly. This is currently missing from msm dp > > > > driver, but there is already effort going on to add it. Once done, > > > > full external DP will be working on all X1E laptops that make use of > > > > this retimer. > > > > > > I was gonna ask you to include the devicetree changes that enables the > > > retimers as part of this series (to facilitate review and testing), but > > > perhaps you should indeed not post them again until LTTPR support is in > > > place. > > > > I was thinking maybe we should not wait for LTTPR support as this series > > brings orientation support as is. I still need to figure out how to > > strip out the DP parts of it in such a way that orientation should still > > be working but DP should not (until LTTPR is in). > > Yeah, possible, or you can at least include the DT patches here but mark > them as do-not-merge-yet or similar. Sure, will do that. Will have to split the DP part of it into separate patches and mark only those as do-not-merge-yet. > > > > > Signed-off-by: Abel Vesa <abel.vesa@xxxxxxxxxx> > > > > --- > > > > Changes in v2: > > > > > - Fixed coldplug (on boot) orientation detection. > > > > > > Coldplug orientation detection still does not work here with this series > > > applied. > > > > > > I'm not entirely sure this whether worked better with v1, but with v2 > > > my SuperSpeed ethernet device shows up as a HighSpeed device in one > > > orientation. It is also not disconnected an re-enumerated as SS as is > > > the case on the X13s (and possibly with v1): > > > > > > usb 1-1: new high-speed USB device number 2 using xhci-hcd > > > > For coldplug, this series does the right thing as it leaves the retimer > > initialized if it was left enabled at boot. There is a second part > > needed for the coldplug to work. That is the regulator-boot-on property > > in retimer's vregs nodes. That will ensure that the regulator is not > > disabled until retimer driver probes and will keep the retimer initialized > > until USB device is enumerated. > > I can confirm that marking the regulators as having been left on by the > bootloader so that they are not disabled temporarily during boot indeed > fixes the coldplug issue here. > > That however makes me wonder whether something is missing in the driver > so that it still relies on setup having been done by the boot firmware. > > Have you tried actually asserting reset during probe to verify that > driver can configure the retimers itself without relying on the boot > firmware? We do not want to reset the retimers on probe because we won't be able to figure out the orientation config until next pmic glink notify comes. The pmic glink notify only triggers on USB event, which never comes until you replug the device. So in order to have coldplug orientation configured correctly in the retimer, we need to make sure the retimer holds state until unplug. > > Johan
