| Message ID | 20250529-hpd_display_off-v1-0-ce33bac2987c@oss.qualcomm.com |
|---|---|
| Headers | show |
| Series | drm/msm/dp: ST_DISPLAY_OFF hpd cleanup | expand |
On Fri, 30 May 2025 at 02:15, Jessica Zhang <jessica.zhang@oss.qualcomm.com> wrote: > > HPD state machine in msm dp display driver manages the state transitions > between various HPD events and the expected state of driver to make sure > both match up. > > Although originally done with the intent of managing userspace interactions > and interactions with compliance equipment, over period of time, > changes to this piece of code has become quite difficult to manage. > > Although, unwinding this logic will take some time and will be spread over > various changes, to start things, this series tries to get rid of the > ST_DISPLAY_OFF state as firstly, its really not an hpd state but a state > of the display overall. Coupled with this, there are quite a few checks > in the current code, the origins of which need to be re-visited OR are unclear > which seem unlikely or redundant. With DP controller on newer chipsets supporting > multiple streams, this has become increasingly difficult to work with. > > This series removes the redundant state checks and simplifies the logic as an > attempt to get rid of this ST_DISPLAY_OFF state. > > Note: This series has been tested with sa8775p and sc7180 devices with multiple > monitors and also multiple dongles with no noticeable regressions. > Both of these devices use native DP PHY though. Hence, if this series can > be verified on some devices with USBC-DP combo PHY with the help of the other > developers, that will be great. > > --- > Changes in v2: The series is not marked as v2 though. > - Rebased on top of next-20250523 > - Change atomic_enable() to return early if ST_DISCONENCT_PENDING > instead of completely dropping the > if (state != ST_DISPLAY_OFF && state != ST_MAINLINK_READY) check (Dmitry) > > --- > Abhinav Kumar (4): > drm/msm/dp: remove redundant checks related to ST_DISPLAY_OFF in plug/irq_ipd handlers > drm/msm/dp: Return early from atomic_enable() if ST_DISCONNECT_PENDING > drm/msm/dp: replace ST_DISPLAY_OFF with power_on in msm_dp_hpd_unplug_handle() > drm/msm/dp: remove ST_DISPLAY_OFF as a hpd_state > > drivers/gpu/drm/msm/dp/dp_display.c | 19 +++---------------- > 1 file changed, 3 insertions(+), 16 deletions(-) > --- > base-commit: daf70030586cf0279a57b58a94c32cfe901df23d > change-id: 20241202-hpd_display_off-6051aa510f23 > > Best regards, > -- > Jessica Zhang <jessica.zhang@oss.qualcomm.com> > -- With best wishes Dmitry
On 5/30/2025 9:05 AM, Dmitry Baryshkov wrote: > On Fri, 30 May 2025 at 02:15, Jessica Zhang > <jessica.zhang@oss.qualcomm.com> wrote: >> >> HPD state machine in msm dp display driver manages the state transitions >> between various HPD events and the expected state of driver to make sure >> both match up. >> >> Although originally done with the intent of managing userspace interactions >> and interactions with compliance equipment, over period of time, >> changes to this piece of code has become quite difficult to manage. >> >> Although, unwinding this logic will take some time and will be spread over >> various changes, to start things, this series tries to get rid of the >> ST_DISPLAY_OFF state as firstly, its really not an hpd state but a state >> of the display overall. Coupled with this, there are quite a few checks >> in the current code, the origins of which need to be re-visited OR are unclear >> which seem unlikely or redundant. With DP controller on newer chipsets supporting >> multiple streams, this has become increasingly difficult to work with. >> >> This series removes the redundant state checks and simplifies the logic as an >> attempt to get rid of this ST_DISPLAY_OFF state. >> >> Note: This series has been tested with sa8775p and sc7180 devices with multiple >> monitors and also multiple dongles with no noticeable regressions. >> Both of these devices use native DP PHY though. Hence, if this series can >> be verified on some devices with USBC-DP combo PHY with the help of the other >> developers, that will be great. >> >> --- >> Changes in v2: > > The series is not marked as v2 though. Hi Dmitry, Sorry for the confusion -- had pulled the v1 [1] using `b4 prep -F` but forgot to force the revision number to v2. [1] https://patchwork.freedesktop.org/series/142010/#rev1 Thanks, Jessica Zhang > >> - Rebased on top of next-20250523 >> - Change atomic_enable() to return early if ST_DISCONENCT_PENDING >> instead of completely dropping the >> if (state != ST_DISPLAY_OFF && state != ST_MAINLINK_READY) check (Dmitry) >> >> --- >> Abhinav Kumar (4): >> drm/msm/dp: remove redundant checks related to ST_DISPLAY_OFF in plug/irq_ipd handlers >> drm/msm/dp: Return early from atomic_enable() if ST_DISCONNECT_PENDING >> drm/msm/dp: replace ST_DISPLAY_OFF with power_on in msm_dp_hpd_unplug_handle() >> drm/msm/dp: remove ST_DISPLAY_OFF as a hpd_state >> >> drivers/gpu/drm/msm/dp/dp_display.c | 19 +++---------------- >> 1 file changed, 3 insertions(+), 16 deletions(-) >> --- >> base-commit: daf70030586cf0279a57b58a94c32cfe901df23d >> change-id: 20241202-hpd_display_off-6051aa510f23 >> >> Best regards, >> -- >> Jessica Zhang <jessica.zhang@oss.qualcomm.com> >> > > > -- > With best wishes > > Dmitry
HPD state machine in msm dp display driver manages the state transitions between various HPD events and the expected state of driver to make sure both match up. Although originally done with the intent of managing userspace interactions and interactions with compliance equipment, over period of time, changes to this piece of code has become quite difficult to manage. Although, unwinding this logic will take some time and will be spread over various changes, to start things, this series tries to get rid of the ST_DISPLAY_OFF state as firstly, its really not an hpd state but a state of the display overall. Coupled with this, there are quite a few checks in the current code, the origins of which need to be re-visited OR are unclear which seem unlikely or redundant. With DP controller on newer chipsets supporting multiple streams, this has become increasingly difficult to work with. This series removes the redundant state checks and simplifies the logic as an attempt to get rid of this ST_DISPLAY_OFF state. Note: This series has been tested with sa8775p and sc7180 devices with multiple monitors and also multiple dongles with no noticeable regressions. Both of these devices use native DP PHY though. Hence, if this series can be verified on some devices with USBC-DP combo PHY with the help of the other developers, that will be great. --- Changes in v2: - Rebased on top of next-20250523 - Change atomic_enable() to return early if ST_DISCONENCT_PENDING instead of completely dropping the if (state != ST_DISPLAY_OFF && state != ST_MAINLINK_READY) check (Dmitry) --- Abhinav Kumar (4): drm/msm/dp: remove redundant checks related to ST_DISPLAY_OFF in plug/irq_ipd handlers drm/msm/dp: Return early from atomic_enable() if ST_DISCONNECT_PENDING drm/msm/dp: replace ST_DISPLAY_OFF with power_on in msm_dp_hpd_unplug_handle() drm/msm/dp: remove ST_DISPLAY_OFF as a hpd_state drivers/gpu/drm/msm/dp/dp_display.c | 19 +++---------------- 1 file changed, 3 insertions(+), 16 deletions(-) --- base-commit: daf70030586cf0279a57b58a94c32cfe901df23d change-id: 20241202-hpd_display_off-6051aa510f23 Best regards,