Hello ， Heiko Stuebner <[email protected]> 于2026年7月16日周四 03:30写道： > > Am Mittwoch, 15. Juli 2026, 21:00:12 Mitteleuropäische Sommerzeit schrieb > Sebastian Reichel: > > Hello Andy, > > > > On Tue, Jul 14, 2026 at 06:08:06PM +0800, Andy Yan wrote: > > > > @@ -58,6 +59,8 @@ static void dw_dp_rockchip_hpd_sw_sel(void *data, bool > > > > force_hpd_from_sw) > > > > > > > > dp->hpd_sel = force_hpd_from_sw; > > > > > > > > + ACQUIRE(pm_runtime_active_auto, pm)(dp->dev); > > > > > > This can triger HPD irq for USB-C DP AltMode, but for a standard > > > DP port, this is no way to acquire runtime, so there will no irq > > > when plug in. > > > > Right. I don't have any RK3588 or RK3576 system exposing a DP > > connector directly. This this is non-trivial to test for me :/ > > > > I see the following options: > > > > 1. Only enable runtime PM when running in USB-C mode > > 2. Mux the HPD pin as GPIO to support waking up with the controller > > being suspended; either remux when the controller is awake or use > > the same mechanism as USB-C to do software based HPD on top of the > > GPIO (which would allow using any GPIO as HPD). > > 3. Have a poll function, which wakes up the controller regularly to > > check the status > > > > I consider going for option 1 for now, which allows option 2 to be > > implemented as a follow up. Mostly to avoid that the series becomes > > as big and messy as the USBDP PHY series became. > > I'd agree with that solution. > I think we can all agree that option 3 is baaaad :-) .
I'm also in favor of moving ahead with 1. About Option 2: we've got a GPIO-based HPD workaround in our BSP tree, but it's a bit flaky in some use-cases： According to the DP spec, there are two types of interrupts: 1. HOTPLUG_IRQ: A level-transition pulse that must be sustained for at least 2 ms. 2. HPD_IRQ: A low-level pulse with a duration between 0.25 ms ~ 2 ms. With the GPIO-based solution, it is difficult to reliably detect such short pulses. > > So with 1, functionality keeps working, and then someone can put the > special HPD handling on top later - the Rock-5-ITX does expose the DP1 > (on a hdmi port) > > > Heiko > >
