On Thu, Oct 10, 2024 at 08:33:11AM +0300, Jouni Högander wrote: > Currently vrr code is overwriting possibly set PSR PR Frame Change Enable > bit in TRANS_PUSH register. Avoid this by using rmw instead of write. RMWs are not good if we ever want to do this from DSB/etc. We should know what the state should be and just program that in. Does the PSR bit here have to match the PSR_CTL enable bit or could we just always set the bit here based on has_psr whether PSR is currently active or not? > > Signed-off-by: Jouni Högander <jouni.hogander@xxxxxxxxx> > --- > drivers/gpu/drm/i915/display/intel_vrr.c | 8 +++++--- > 1 file changed, 5 insertions(+), 3 deletions(-) > > diff --git a/drivers/gpu/drm/i915/display/intel_vrr.c b/drivers/gpu/drm/i915/display/intel_vrr.c > index 9a51f5bac3071..8b4e3f938efea 100644 > --- a/drivers/gpu/drm/i915/display/intel_vrr.c > +++ b/drivers/gpu/drm/i915/display/intel_vrr.c > @@ -339,8 +339,8 @@ void intel_vrr_enable(const struct intel_crtc_state *crtc_state) > if (!crtc_state->vrr.enable) > return; > > - intel_de_write(display, TRANS_PUSH(display, cpu_transcoder), > - TRANS_PUSH_EN); > + intel_de_rmw(display, TRANS_PUSH(display, cpu_transcoder), 0, > + TRANS_PUSH_EN); > > if (HAS_AS_SDP(display)) > intel_de_write(display, > @@ -371,7 +371,9 @@ void intel_vrr_disable(const struct intel_crtc_state *old_crtc_state) > intel_de_wait_for_clear(display, > TRANS_VRR_STATUS(display, cpu_transcoder), > VRR_STATUS_VRR_EN_LIVE, 1000); > - intel_de_write(display, TRANS_PUSH(display, cpu_transcoder), 0); > + > + intel_de_rmw(display, TRANS_PUSH(display, cpu_transcoder), TRANS_PUSH_EN, > + 0); > > if (HAS_AS_SDP(display)) > intel_de_write(display, > -- > 2.34.1 -- Ville Syrjälä Intel
