On Tue, Jul 29, 2025 at 12:44:47PM +0300, Jani Nikula wrote: > On Thu, 24 Jul 2025, Maarten Lankhorst <[email protected]> > wrote: > > Hey, > > > > Den 2025-07-23 kl. 15:19, skrev Rodrigo Vivi: > >> On Wed, Jul 23, 2025 at 12:38:08PM +0300, Imre Deak wrote: > >>> On Wed, Jul 23, 2025 at 02:59:46PM +0530, Dibin Moolakadan Subrahmanian > >>> wrote: > >>>> It has been observed that during `xe_display_pm_suspend()` execution, > >>>> an HPD interrupt can still be triggered, resulting in `dig_port_work` > >>>> being scheduled. The issue arises when this work executes after > >>>> `xe_display_pm_suspend_late()`, by which time the display is fully > >>>> suspended. > >>>> > >>>> This can lead to errors such as "DC state mismatch", as the dig_port > >>>> work accesses display resources that are no longer available or > >>>> powered. > >>>> > >>>> To address this, introduce 'intel_encoder_block_all_hpds' and > >>>> 'intel_encoder_unblock_all_hpds' functions, which iterate over all > >>>> encoders and block/unblock HPD respectively. > >>>> > >>>> These are used to: > >>>> - Block HPD IRQs before calling 'intel_hpd_cancel_work' in suspend > >>>> and shutdown > >>>> - Unblock HPD IRQs after 'intel_hpd_init' in resume > >>>> > >>>> This will prevent 'dig_port_work' being scheduled during display > >>>> suspend. > >>>> > >>>> Continuation of previous patch discussion: > >>>> https://patchwork.freedesktop.org/patch/663964/ > >>>> > >>>> Changes in v2: > >>>> - Add 'intel_encoder_block_all_hpds' to 'xe_display_pm_shutdown' > >>>> - Add 'intel_hpd_cancel_work' to 'xe_display_fini_early' to cancel > >>>> any HPD pending work at late driver removal > >>>> > >>>> Signed-off-by: Dibin Moolakadan Subrahmanian > >>>> <[email protected]> > >>>> --- > >>>> drivers/gpu/drm/i915/display/intel_encoder.c | 23 ++++++++++++++++++++ > >>>> drivers/gpu/drm/i915/display/intel_encoder.h | 3 +++ > >>>> drivers/gpu/drm/i915/display/intel_hotplug.c | 2 -- > >>>> drivers/gpu/drm/xe/display/xe_display.c | 6 +++++ > >>>> 4 files changed, 32 insertions(+), 2 deletions(-) > >>>> > >>>> diff --git a/drivers/gpu/drm/i915/display/intel_encoder.c > >>>> b/drivers/gpu/drm/i915/display/intel_encoder.c > >>>> index 0b7bd26f4339..4e2b77b87678 100644 > >>>> --- a/drivers/gpu/drm/i915/display/intel_encoder.c > >>>> +++ b/drivers/gpu/drm/i915/display/intel_encoder.c > >>>> @@ -8,6 +8,7 @@ > >>>> #include "intel_display_core.h" > >>>> #include "intel_display_types.h" > >>>> #include "intel_encoder.h" > >>>> +#include "intel_hotplug.h" > >>>> > >>>> static void intel_encoder_link_check_work_fn(struct work_struct *work) > >>>> { > >>>> @@ -37,6 +38,28 @@ void intel_encoder_link_check_queue_work(struct > >>>> intel_encoder *encoder, int dela > >>>> &encoder->link_check_work, > >>>> msecs_to_jiffies(delay_ms)); > >>>> } > >>>> > >>>> +void intel_encoder_unblock_all_hpds(struct intel_display *display) > >>>> +{ > >>>> + struct intel_encoder *encoder; > >>>> + > >>>> + if (!HAS_DISPLAY(display)) > >>>> + return; > >>>> + > >>>> + for_each_intel_encoder(display->drm, encoder) > >>>> + intel_hpd_unblock(encoder); > >>>> +} > >>>> + > >>>> +void intel_encoder_block_all_hpds(struct intel_display *display) > >>>> +{ > >>>> + struct intel_encoder *encoder; > >>>> + > >>>> + if (!HAS_DISPLAY(display)) > >>>> + return; > >>>> + > >>>> + for_each_intel_encoder(display->drm, encoder) > >>>> + intel_hpd_block(encoder); > >>>> +} > >>>> + > >>>> void intel_encoder_suspend_all(struct intel_display *display) > >>>> { > >>>> struct intel_encoder *encoder; > >>>> diff --git a/drivers/gpu/drm/i915/display/intel_encoder.h > >>>> b/drivers/gpu/drm/i915/display/intel_encoder.h > >>>> index 3fa5589f0b1c..e1d3aeab7c00 100644 > >>>> --- a/drivers/gpu/drm/i915/display/intel_encoder.h > >>>> +++ b/drivers/gpu/drm/i915/display/intel_encoder.h > >>>> @@ -17,4 +17,7 @@ void intel_encoder_link_check_flush_work(struct > >>>> intel_encoder *encoder); > >>>> void intel_encoder_suspend_all(struct intel_display *display); > >>>> void intel_encoder_shutdown_all(struct intel_display *display); > >>>> > >>>> +void intel_encoder_block_all_hpds(struct intel_display *display); > >>>> +void intel_encoder_unblock_all_hpds(struct intel_display *display); > >>>> + > >>>> #endif /* __INTEL_ENCODER_H__ */ > >>>> diff --git a/drivers/gpu/drm/i915/display/intel_hotplug.c > >>>> b/drivers/gpu/drm/i915/display/intel_hotplug.c > >>>> index 265aa97fcc75..c69b535497bf 100644 > >>>> --- a/drivers/gpu/drm/i915/display/intel_hotplug.c > >>>> +++ b/drivers/gpu/drm/i915/display/intel_hotplug.c > >>>> @@ -971,8 +971,6 @@ void intel_hpd_cancel_work(struct intel_display > >>>> *display) > >>>> > >>>> spin_lock_irq(&display->irq.lock); > >>>> > >>>> - drm_WARN_ON(display->drm, get_blocked_hpd_pin_mask(display)); > >>>> - > >>>> display->hotplug.long_hpd_pin_mask = 0; > >>>> display->hotplug.short_hpd_pin_mask = 0; > >>>> display->hotplug.event_bits = 0; > >>>> diff --git a/drivers/gpu/drm/xe/display/xe_display.c > >>>> b/drivers/gpu/drm/xe/display/xe_display.c > >>>> index e2e0771cf274..9e984a045059 100644 > >>>> --- a/drivers/gpu/drm/xe/display/xe_display.c > >>>> +++ b/drivers/gpu/drm/xe/display/xe_display.c > >>>> @@ -96,6 +96,7 @@ static void xe_display_fini_early(void *arg) > >>>> if (!xe->info.probe_display) > >>>> return; > >>>> > >>>> + intel_hpd_cancel_work(display); > >>>> intel_display_driver_remove_nogem(display); > >>>> intel_display_driver_remove_noirq(display); > >>>> intel_opregion_cleanup(display); > >>>> @@ -340,6 +341,8 @@ void xe_display_pm_suspend(struct xe_device *xe) > >>>> > >>>> xe_display_flush_cleanup_work(xe); > >>>> > >>>> + intel_encoder_block_all_hpds(display); > >>>> + > >>>> intel_hpd_cancel_work(display); > >>>> > >>>> if (has_display(xe)) { > >>>> @@ -369,6 +372,7 @@ void xe_display_pm_shutdown(struct xe_device *xe) > >>>> } > >>>> > >>>> xe_display_flush_cleanup_work(xe); > >>>> + intel_encoder_block_all_hpds(display); > >>> > >>> MST still needs HPD IRQs for side-band messaging, so the HPD IRQs must > >>> be blocked only after intel_dp_mst_suspend(). > >>> > >>> Otherwise the patch looks ok to me, so with the above fixed and provided > >>> that Maarten is ok to disable all display IRQs only later: > >> > >> Also probably good to identify the patch as both xe and i915 in the subject > >> drm/{i915,xe}/display: > >> > >> and Maarten or Imre, any preference on which branch to go? any chance of > >> conflicting with any of work you might be doing in any side? > >> > >> From my side I believe that any conflict might be easy to handle, so > >> > >> Acked-by: Rodrigo Vivi <[email protected]> > >> > >> from either side... > >> > >>> > >>> Reviewed-by: Imre Deak <[email protected]> > > We had a discussion on this approach, and it seems that completely > > disabling interrupts here like in i915 would fail too. > > > > Reviewed-by: Maarten Lankhorst <[email protected]> > > > > I don't mind either branch. As long as it applies. :-) > > Please do not merge through *any* tree. > > How come you all think it's okay to add this xe specific thing, and make > unification harder?
I lost any moral or rights to complain here since I couldn't move with my tasks of unification of the pm flow :( double sorry! > > intel_encoder_block_all_hpds() is *way* too specific for a high level > function. Neither xe nor i915 should never call something like that > directly. that's a valid point indeed. But I cannot see another way to fix the current issue right now without trying to move with the full unification faster. Do you? > > > BR, > Jani. > > > -- > Jani Nikula, Intel
