drm/amd/mst: clean DP main link status only when unplug mst 1st link
Subsystems (3)
PATCH 1/1 drm/amd/mst: clean DP main link status only when unplug mst 1st link
[Why]
Under DP daisy chain scenario as below:
Src - Monitor_1 - Monitor_2
If unplug 2nd Monitor_2 and plug in again, observe that Monitor_1 doesn't light up.
When unplug 2nd monitor, we clear the dc_link->cur_link_settings.lane_count in dm_dp_destroy_mst_connector(). However this link status is a shared data structure by all connected mst monitors. Although the 2nd monitor is gone, this link status should still be retained for other connected mst monitors. Otherwise, when we plug the 2nd monitor in again, we find out that main link is not trained and do link training again. Payload ID Table for Monitor_1 is ruined and we don't reallocate it.
[How]
In dm_dp_destroy_mst_connector(), only clean the cur_link_settings when we no longer do mst mode.
--- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +4 -1
@@ -415,7 +415,10 @@ static void dm_dp_destroy_mst_connector(struct drm_dp_mst_topology_mgr *mgr, aconnector->dc_sink);dc_sink_release (aconnector->dc_sink); aconnector->dc_sink =NULL ; - aconnector->dc_link->cur_link_settings.lane_count =0 ; +mutex_lock (&mgr->lock); +if (!mgr->mst_state) + aconnector->dc_link->cur_link_settings.lane_count =0 ; +mutex_unlock (&mgr->lock); }drm_connector_unregister (connector);
Signature · 6 lines
2.17.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Quoted 31 lines · Wayne Lin
[Why]
Under DP daisy chain scenario as below:Src -
Monitor_1-Monitor_2If unplug 2nd
Monitor_2and plug in again, observe thatMonitor_1doesn't light up.When unplug 2nd monitor, we clear the
dc_link->cur_link_settings.lane_countindm_dp_destroy_mst_connector(). However this link status is a shared data structure by all connected mst monitors. Although the 2nd monitor is gone, this link status should still be retained for other connected mst monitors. Otherwise, when we plug the 2nd monitor in again, we find out that main link is not trained and do link training again. Payload ID Table forMonitor_1is ruined and we don't reallocate it.[How]
Indm_dp_destroy_mst_connector(), only clean thecur_link_settingswhen we no longer do mst mode.Signed-off-by Wayne Lin <[email protected]>--- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +0
-0
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c static void dm_dp_destroy_mst_connector(struct drm_dp_mst_topology_mgr *mgr, · line 421
Quoted 8 lines
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +4
-1@@ -415,7 +415,10 @@ static void dm_dp_destroy_mst_connector(struct drm_dp_mst_topology_mgr *mgr, aconnector->dc_sink);dc_sink_release (aconnector->dc_sink); aconnector->dc_sink =NULL ; - aconnector->dc_link->cur_link_settings.lane_count =0 ; +mutex_lock (&mgr->lock); +if (!mgr->mst_state) + aconnector->dc_link->cur_link_settings.lane_count =0 ; +mutex_unlock (&mgr->lock);
Hi Wayne,
The change looks good to me.
Just for curiosity, do you know why we use a mutex instead of spin_lock_irq for this case? FWIU, the spin_lock_irq behavior looks better for this sort of manipulation.
Thanks
Quoted 6 lines
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +0
-0}drm_connector_unregister (connector);Signature · 1 line
2.17.1
Signature · 2 lines
Rodrigo Siqueira https://siqueira.tech
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
[AMD Public Use]
Quoted 57 lines
-----Original Message-----
From: Siqueira, Rodrigo <[email protected]>
Sent: Wednesday, August 5, 2020 10:55 PM
To: Lin, Wayne <[email protected]>
Cc: [email protected]; Lakha, Bhawanpreet <[email protected]>; Wu, Hersen <[email protected]>; Kazlauskas, Nicholas <[email protected]>; Zuo, Jerry <[email protected]>
Subject: Re: [PATCH] drm/amd/mst: clean DP main link status only when unplug mst 1st linkQuoted 17 lines · Wayne Lin
[Why]
Under DP daisy chain scenario as below:Src -
Monitor_1-Monitor_2If unplug 2nd
Monitor_2and plug in again, observe thatMonitor_1doesn't light up.When unplug 2nd monitor, we clear the
dc_link->cur_link_settings.lane_countindm_dp_destroy_mst_connector(). However this link status is a shared data structure by all connected mst monitors. Although the 2nd monitor is gone, this link status should still be retained for other connected mst monitors. Otherwise, when we plug the 2nd monitor in again, we find out that main link is not trained and do link training again. Payload ID Table forMonitor_1is ruined and we don't reallocate it.Quoted 7 lines
[How]
Indm_dp_destroy_mst_connector(), only clean thecur_link_settingswhen we no longer do mst mode.Signed-off-by Wayne Lin <[email protected]>--- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c | 5++++-
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c static void dm_dp_destroy_mst_connector(struct
Quoted 9 lines
1 file changed, 4 insertions(+), 1 deletion(-)diff --git
a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.cb/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.cindex2c10352fa514..526f29598403100644drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +0
-0@@ -415,7 +415,10 @@ static void dm_dp_destroy_mst_connector(struct
drm_dp_mst_topology_mgr*mgr,drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c static void dm_dp_destroy_mst_connector(struct drm_dp_mst_topology_mgr *mgr, · line 421
Quoted 8 lines
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +4
-1aconnector->dc_sink);dc_sink_release (aconnector->dc_sink); aconnector->dc_sink =NULL ; - aconnector->dc_link->cur_link_settings.lane_count =0 ; +mutex_lock (&mgr->lock); +if (!mgr->mst_state) + aconnector->dc_link->cur_link_settings.lane_count =0 ; +mutex_unlock (&mgr->lock);Hi Wayne,
Quoted 7 lines
The change looks good to me.
Reviewed-by Rodrigo Siqueira <[email protected]>Just for curiosity, do you know why we use a mutex instead of
spin_lock_irqfor this case? FWIU, thespin_lock_irqbehavior looks better for this sort of manipulation.
Hi Siqueira,
Thanks for your time.
AFAIK, changing mst_state (e.g. enabling MST mode) involves some reading/writing steps on DPCD which might cost some time.
Quoted 5 lines
Thanks
Quoted 2 lines
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +0
-0}
Quoted 8 lines
Quoted 4 lines
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +0
-1drm_connector_unregister (connector); --2.17.1
--
Rodrigo Siqueira
https://siqueira.tech
--
Wayne Lin
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx