[PATCH 02/13] drm/i915/cdclk: Fix voltage_level programming edge case

[Ville Syrjälä]

On Fri, Mar 29, 2024 at 02:04:55PM -0300, Gustavo Sousa wrote:
> Quoting Ville Syrjala (2024-03-27 14:45:33-03:00)
> >From: Ville Syrjälä <ville.syrjala at linux.intel.com>
> >
> >Currently we only consider the relationship of the
> >old and new CDCLK frequencies when determining whether
> >to do the repgramming from intel_set_cdclk_pre_plane_update()
> >or intel_set_cdclk_post_plane_update().
> >
> >It is technically possible to have a situation where the
> >CDCLK frequency is decreasing, but the voltage_level is
> >increasing due a DDI port. In this case we should bump
> >the voltage level already in intel_set_cdclk_pre_plane_update()
> >(so that the voltage_level will have been increased by the
> >time the port gets enabled), while leaving the CDCLK frequency
> >unchanged (as active planes/etc. may still depend on it).
> >We can then reduce the CDCLK frequency to its final value
> >from intel_set_cdclk_post_plane_update().
> >
> >In order to handle that correctly we shall construct a
> >suitable amalgam of the old and new cdclk states in
> >intel_set_cdclk_pre_plane_update().
> >
> >And we can simply call intel_set_cdclk() unconditionally
> >in both places as it will not do anything if nothing actually
> >changes vs. the current hw state.
> >
> >v2: Handle cdclk_state->disable_pipes
> >
> >Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> >---
> > drivers/gpu/drm/i915/display/intel_cdclk.c | 27 +++++++++++++---------
> > 1 file changed, 16 insertions(+), 11 deletions(-)
> >
> >diff --git a/drivers/gpu/drm/i915/display/intel_cdclk.c b/drivers/gpu/drm/i915/display/intel_cdclk.c
> >index 619529dba095..504c5cbbcfff 100644
> >--- a/drivers/gpu/drm/i915/display/intel_cdclk.c
> >+++ b/drivers/gpu/drm/i915/display/intel_cdclk.c
> >@@ -2600,6 +2600,7 @@ intel_set_cdclk_pre_plane_update(struct intel_atomic_state *state)
> >                 intel_atomic_get_old_cdclk_state(state);
> >         const struct intel_cdclk_state *new_cdclk_state =
> >                 intel_atomic_get_new_cdclk_state(state);
> >+        struct intel_cdclk_config cdclk_config;
> > 
> >         if (!intel_cdclk_changed(&old_cdclk_state->actual,
> >                                  &new_cdclk_state->actual))
> >@@ -2608,13 +2609,21 @@ intel_set_cdclk_pre_plane_update(struct intel_atomic_state *state)
> >         if (IS_DG2(i915))
> >                 intel_cdclk_pcode_pre_notify(state);
> > 
> >-        if (new_cdclk_state->disable_pipes ||
> >-            old_cdclk_state->actual.cdclk <= new_cdclk_state->actual.cdclk) {
> >-                drm_WARN_ON(&i915->drm, !new_cdclk_state->base.changed);
> >+        if (new_cdclk_state->disable_pipes) {
> >+                cdclk_config = new_cdclk_state->actual;
> >+        } else {
> >+                if (new_cdclk_state->actual.cdclk >= old_cdclk_state->actual.cdclk)
> >+                        cdclk_config = new_cdclk_state->actual;
> >+                else
> >+                        cdclk_config = old_cdclk_state->actual;
> > 
> >-                intel_set_cdclk(i915, &new_cdclk_state->actual,
> >-                                new_cdclk_state->pipe);
> >+                cdclk_config.voltage_level = max(new_cdclk_state->actual.voltage_level,
> >+                                                 old_cdclk_state->actual.voltage_level);
> >         }
> >+
> >+        drm_WARN_ON(&i915->drm, !new_cdclk_state->base.changed);
> >+
> >+        intel_set_cdclk(i915, &cdclk_config, new_cdclk_state->pipe);
> 
> Not sure if there could be unwanted side effects with passing
> new_cdclk_state->pipe when using old_cdclk_state->actual. Because
> voltage_level might have changed, parts of the cdclk change sequence end
> up being exercised even when cdclk_config == old_cdclk_state->actual.
> 
> Well, even if those side effects might be harmless, I wonder if it would
> be better if we used INVALID_PIPE when using old_cdclk_state->actual.

Yeah. I doubt there should be any really bad side effects, but
probably a good idea to sidestep the whole question by passing
in INVALID_PIPE.

-- 
Ville Syrjälä
Intel