| Message ID | 20200124144216.4045-1-stanislav.lisovskiy@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v2] drm/i915: Fix inconsistance between pfit.enable and scaler freeing | expand |
On Fri, Jan 24, 2020 at 04:42:16PM +0200, Stanislav Lisovskiy wrote: > Despite that during hw readout we seem to have scalers assigned > to pipes, then call atomic_setup_scalers, at the commit stage in > skl_update_scaler there is a check, that if we have fb src and > dest of same size, we stage freeing of that scaler. > > However we don't update pfit.enabled flag then, which makes > the state inconsistent, which in turn triggers a WARN_ON > in skl_pfit_enable, because we have pfit enabled, > but no assigned scaler. > > To me this looks weird that we kind of do the decision > to use or not use the scaler at skl_update_scaler stage > but not in intel_atomic_setup_scalers, moreover > not updating the whole state consistently. > > This fix is to not free the scaler if we have pfit.enabled > flag set, so that the state is now consistent > and the warnings are gone. > > Bugzilla: https://gitlab.freedesktop.org/drm/intel/issues/577 > Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com> > --- > drivers/gpu/drm/i915/display/intel_display.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c > index 5768cfcf71c4..23221b8d244d 100644 > --- a/drivers/gpu/drm/i915/display/intel_display.c > +++ b/drivers/gpu/drm/i915/display/intel_display.c > @@ -5953,7 +5953,7 @@ skl_update_scaler(struct intel_crtc_state *crtc_state, bool force_detach, > * the 90/270 degree plane rotation cases (to match the > * GTT mapping), hence no need to account for rotation here. > */ > - if (src_w != dst_w || src_h != dst_h) > + if (src_w != dst_w || src_h != dst_h || crtc_state->pch_pfit.enabled) This is called for both pfit and planes. So you need to put that pfit check into the pfit specific function. > need_scaler = true; > > /* > -- > 2.24.1.485.gad05a3d8e5
diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c index 5768cfcf71c4..23221b8d244d 100644 --- a/drivers/gpu/drm/i915/display/intel_display.c +++ b/drivers/gpu/drm/i915/display/intel_display.c @@ -5953,7 +5953,7 @@ skl_update_scaler(struct intel_crtc_state *crtc_state, bool force_detach, * the 90/270 degree plane rotation cases (to match the * GTT mapping), hence no need to account for rotation here. */ - if (src_w != dst_w || src_h != dst_h) + if (src_w != dst_w || src_h != dst_h || crtc_state->pch_pfit.enabled) need_scaler = true; /*
Despite that during hw readout we seem to have scalers assigned to pipes, then call atomic_setup_scalers, at the commit stage in skl_update_scaler there is a check, that if we have fb src and dest of same size, we stage freeing of that scaler. However we don't update pfit.enabled flag then, which makes the state inconsistent, which in turn triggers a WARN_ON in skl_pfit_enable, because we have pfit enabled, but no assigned scaler. To me this looks weird that we kind of do the decision to use or not use the scaler at skl_update_scaler stage but not in intel_atomic_setup_scalers, moreover not updating the whole state consistently. This fix is to not free the scaler if we have pfit.enabled flag set, so that the state is now consistent and the warnings are gone. Bugzilla: https://gitlab.freedesktop.org/drm/intel/issues/577 Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com> --- drivers/gpu/drm/i915/display/intel_display.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)