On Fri, Feb 11, 2022 at 11:26:04AM +0200, Ville Syrjala wrote: > From: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx> > > Drop the locks around cursor plane register writes. The > lock isn't needed since each plane's register are neatly > contained on their own cachelines. > > The locking did have a secondary effect of disabling > interrupts around the cursor registers writes though. > If we drop that then we open outselves up for sceduling > delays and whatnot while on the middle of the register > writes. That increases the chance of not all the register > writes land during the same frame. For normal atomic > commits this is not a concern as the vblank evade mechanism > anyway disables interrupts around the update, but the legacy > cursor codepath does not. Technically we should do a vblank > evade there as well, but so far no one has bothered to hook > that up. So in the meantime let's put an explicit local irq > disable/enable around the legacy cursor update to keep the > race window minimal. > > v2: local_irq_{disable,enable}() for legacy cursor ioctl Guess, this will help our infamous atomic update evasion time exceeeded. I think I've even checked with similar patch. Good that its finally will make its way into kernel. Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@xxxxxxxxx> > > Signed-off-by: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx> > --- > drivers/gpu/drm/i915/display/intel_cursor.c | 20 ++++++++++---------- > 1 file changed, 10 insertions(+), 10 deletions(-) > > diff --git a/drivers/gpu/drm/i915/display/intel_cursor.c b/drivers/gpu/drm/i915/display/intel_cursor.c > index 2ade8fdd9bdd..b648be744cf2 100644 > --- a/drivers/gpu/drm/i915/display/intel_cursor.c > +++ b/drivers/gpu/drm/i915/display/intel_cursor.c > @@ -255,7 +255,6 @@ static void i845_cursor_update_arm(struct intel_plane *plane, > { > struct drm_i915_private *dev_priv = to_i915(plane->base.dev); > u32 cntl = 0, base = 0, pos = 0, size = 0; > - unsigned long irqflags; > > if (plane_state && plane_state->uapi.visible) { > unsigned int width = drm_rect_width(&plane_state->uapi.dst); > @@ -270,8 +269,6 @@ static void i845_cursor_update_arm(struct intel_plane *plane, > pos = intel_cursor_position(plane_state); > } > > - spin_lock_irqsave(&dev_priv->uncore.lock, irqflags); > - > /* On these chipsets we can only modify the base/size/stride > * whilst the cursor is disabled. > */ > @@ -290,8 +287,6 @@ static void i845_cursor_update_arm(struct intel_plane *plane, > } else { > intel_de_write_fw(dev_priv, CURPOS(PIPE_A), pos); > } > - > - spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); > } > > static void i845_cursor_disable_arm(struct intel_plane *plane, > @@ -492,7 +487,6 @@ static void i9xx_cursor_update_arm(struct intel_plane *plane, > struct drm_i915_private *dev_priv = to_i915(plane->base.dev); > enum pipe pipe = plane->pipe; > u32 cntl = 0, base = 0, pos = 0, fbc_ctl = 0; > - unsigned long irqflags; > > if (plane_state && plane_state->uapi.visible) { > int width = drm_rect_width(&plane_state->uapi.dst); > @@ -508,8 +502,6 @@ static void i9xx_cursor_update_arm(struct intel_plane *plane, > pos = intel_cursor_position(plane_state); > } > > - spin_lock_irqsave(&dev_priv->uncore.lock, irqflags); > - > /* > * On some platforms writing CURCNTR first will also > * cause CURPOS to be armed by the CURBASE write. > @@ -555,8 +547,6 @@ static void i9xx_cursor_update_arm(struct intel_plane *plane, > intel_de_write_fw(dev_priv, CURPOS(pipe), pos); > intel_de_write_fw(dev_priv, CURBASE(pipe), base); > } > - > - spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); > } > > static void i9xx_cursor_disable_arm(struct intel_plane *plane, > @@ -715,6 +705,14 @@ intel_legacy_cursor_update(struct drm_plane *_plane, > */ > crtc_state->active_planes = new_crtc_state->active_planes; > > + /* > + * Technically we should do a vblank evasion here to make > + * sure all the cursor registers update on the same frame. > + * For now just make sure the register writes happen as > + * quickly as possible to minimize the race window. > + */ > + local_irq_disable(); > + > if (new_plane_state->uapi.visible) { > intel_plane_update_noarm(plane, crtc_state, new_plane_state); > intel_plane_update_arm(plane, crtc_state, new_plane_state); > @@ -722,6 +720,8 @@ intel_legacy_cursor_update(struct drm_plane *_plane, > intel_plane_disable_arm(plane, crtc_state); > } > > + local_irq_enable(); > + > intel_plane_unpin_fb(old_plane_state); > > out_free: > -- > 2.34.1 >