Re: [PATCH] drm/i915/cmtg: Disable CMTG after PSR exit to avoid disable timeout
Jani Nikula <[email protected]>
| Newsgroups | org.freedesktop.lists.intel-xe,org.freedesktop.lists.intel-gfx |
|---|---|
| Organization | Intel Finland Oy - BIC 0357606-4 - c/o Alberga Business Park, 6 krs Bertel Jungin Aukio 5, 02600 Espoo, Finland |
| Message-ID | <[email protected]> |
On Fri, 07 Aug 2026, Animesh Manna <[email protected]> wrote: > CMTG is enabled in sync-to-port mode, so its running state (CMTG_STATE) > can only clear while the port is actively timing. In hsw_crtc_disable() > the CMTG was torn down at the very beginning of the sequence, before the > encoders were disabled. At that point PSR may still be active on the eDP, > leaving the port timing generator idle. As a result, after clearing > CMTG_ENABLE the wait for CMTG_STATE to clear never completes and triggers: > > xe 0000:00:02.0: [drm] CMTG: A disable timeout > WARNING: drivers/gpu/drm/i915/display/intel_cmtg.c:205 intel_cmtg_disable+0x331/0x3d0 [xe] > > This is intermittently seen with igt@kms_hdr@bpc-switch-suspend, which > repeatedly cycles the eDP in and out of PSR across suspend/resume. > > Move the CMTG disable to after intel_encoders_disable() (so PSR is > exited) but before intel_encoders_post_disable()/intel_dpll_disable(), > so the transcoder timing generator and port PLL are still active while > CMTG_STATE is polled. The ordering requirement of disabling CMTG before > the transcoder and port PLL are brought down is preserved. > > Cc: Dibin Moolakadan Subrahmanian <[email protected]> > Cc: Uma Shankar <[email protected]> > Signed-off-by: Animesh Manna <[email protected]> Fixes: ? Cc: stable ? > --- > drivers/gpu/drm/i915/display/intel_display.c | 18 ++++++++++++++---- > 1 file changed, 14 insertions(+), 4 deletions(-) > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c > index 829d7a411dcc..cc4d20fdda44 100644 > --- a/drivers/gpu/drm/i915/display/intel_display.c > +++ b/drivers/gpu/drm/i915/display/intel_display.c > @@ -1791,15 +1791,25 @@ static void hsw_crtc_disable(struct intel_atomic_state *state, > intel_atomic_get_old_crtc_state(state, crtc); > struct intel_crtc *pipe_crtc; > > - if (crtc->cmtg.enabled && intel_cmtg_is_allowed(old_crtc_state)) { > - intel_cmtg_set_clk_select(old_crtc_state); > - intel_cmtg_disable(old_crtc_state); > - } > /* > * FIXME collapse everything to one hook. > * Need care with mst->ddi interactions. > */ > intel_encoders_disable(state, crtc); > + > + /* > + * Disable CMTG after the encoders are disabled (so PSR is exited) but > + * before the transcoder timing generator and port PLL are torn down in > + * intel_encoders_post_disable()/intel_dpll_disable(). CMTG is synced to > + * the port, so its running state (CMTG_STATE) can only clear while the > + * port is actively timing. Doing this while PSR is still active leaves > + * the port idle and makes the CMTG_STATE clear wait time out. > + */ > + if (crtc->cmtg.enabled && intel_cmtg_is_allowed(old_crtc_state)) { > + intel_cmtg_set_clk_select(old_crtc_state); > + intel_cmtg_disable(old_crtc_state); > + } > + > intel_encoders_post_disable(state, crtc); > > intel_dpll_disable(old_crtc_state); -- Jani Nikula, Intel