Re: [Intel-gfx] [PATCH 1/6] drm/i915: Move psr unlock out from the pipe update critical section

2023-08-29 Thread Ville Syrjälä 
On Mon, Aug 28, 2023 at 11:16:13AM -0700, Manasi Navare wrote:
> By doing psr_unlock after the vblank evade, are we ensuring that even
> when VRR params change during the pipe update, psr unlock will
> happen after the actual vblank  based on newly programmed VRR params?

The unlock will happen as soon as the new register values have been
written. The vblank will happen when it happens, could be asap or could
be much later.

I don't actually even know what this PSR lock is protecting, I suppose
it's trying to prevent muckery with the PSR hw state while the update
is being programmed. Shrug.

> 
> Manasi
> 
> On Sun, Aug 27, 2023 at 10:41 PM Ville Syrjala
>  wrote:
> >
> > From: Ville Syrjälä 
> >
> > Do the PSR unlock after the vblank evade critcal section is
> > fully over, not before.
> >
> > Cc: Manasi Navare 
> > Signed-off-by: Ville Syrjälä 
> > ---
> >  drivers/gpu/drm/i915/display/intel_crtc.c | 9 +
> >  1 file changed, 5 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/display/intel_crtc.c 
> > b/drivers/gpu/drm/i915/display/intel_crtc.c
> > index 182c6dd64f47..5caa928e5ce9 100644
> > --- a/drivers/gpu/drm/i915/display/intel_crtc.c
> > +++ b/drivers/gpu/drm/i915/display/intel_crtc.c
> > @@ -646,10 +646,8 @@ void intel_pipe_update_end(struct intel_crtc_state 
> > *new_crtc_state)
> > ktime_t end_vbl_time = ktime_get();
> > struct drm_i915_private *dev_priv = to_i915(crtc->base.dev);
> >
> > -   intel_psr_unlock(new_crtc_state);
> > -
> > if (new_crtc_state->do_async_flip)
> > -   return;
> > +   goto out;
> >
> > trace_intel_pipe_update_end(crtc, end_vbl_count, scanline_end);
> >
> > @@ -709,7 +707,7 @@ void intel_pipe_update_end(struct intel_crtc_state 
> > *new_crtc_state)
> > local_irq_enable();
> >
> > if (intel_vgpu_active(dev_priv))
> > -   return;
> > +   goto out;
> >
> > if (crtc->debug.start_vbl_count &&
> > crtc->debug.start_vbl_count != end_vbl_count) {
> > @@ -724,4 +722,7 @@ void intel_pipe_update_end(struct intel_crtc_state 
> > *new_crtc_state)
> > }
> >
> > dbg_vblank_evade(crtc, end_vbl_time);
> > +
> > +out:
> > +   intel_psr_unlock(new_crtc_state);
> >  }
> > --
> > 2.41.0
> >

-- 
Ville Syrjälä
Intel

Re: [Intel-gfx] [PATCH 1/6] drm/i915: Move psr unlock out from the pipe update critical section

2023-08-28 Thread Manasi Navare 
By doing psr_unlock after the vblank evade, are we ensuring that even
when VRR params change during the pipe update, psr unlock will
happen after the actual vblank  based on newly programmed VRR params?

Manasi

On Sun, Aug 27, 2023 at 10:41 PM Ville Syrjala
 wrote:
>
> From: Ville Syrjälä 
>
> Do the PSR unlock after the vblank evade critcal section is
> fully over, not before.
>
> Cc: Manasi Navare 
> Signed-off-by: Ville Syrjälä 
> ---
>  drivers/gpu/drm/i915/display/intel_crtc.c | 9 +
>  1 file changed, 5 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/display/intel_crtc.c 
> b/drivers/gpu/drm/i915/display/intel_crtc.c
> index 182c6dd64f47..5caa928e5ce9 100644
> --- a/drivers/gpu/drm/i915/display/intel_crtc.c
> +++ b/drivers/gpu/drm/i915/display/intel_crtc.c
> @@ -646,10 +646,8 @@ void intel_pipe_update_end(struct intel_crtc_state 
> *new_crtc_state)
> ktime_t end_vbl_time = ktime_get();
> struct drm_i915_private *dev_priv = to_i915(crtc->base.dev);
>
> -   intel_psr_unlock(new_crtc_state);
> -
> if (new_crtc_state->do_async_flip)
> -   return;
> +   goto out;
>
> trace_intel_pipe_update_end(crtc, end_vbl_count, scanline_end);
>
> @@ -709,7 +707,7 @@ void intel_pipe_update_end(struct intel_crtc_state 
> *new_crtc_state)
> local_irq_enable();
>
> if (intel_vgpu_active(dev_priv))
> -   return;
> +   goto out;
>
> if (crtc->debug.start_vbl_count &&
> crtc->debug.start_vbl_count != end_vbl_count) {
> @@ -724,4 +722,7 @@ void intel_pipe_update_end(struct intel_crtc_state 
> *new_crtc_state)
> }
>
> dbg_vblank_evade(crtc, end_vbl_time);
> +
> +out:
> +   intel_psr_unlock(new_crtc_state);
>  }
> --
> 2.41.0
>

[Intel-gfx] [PATCH 1/6] drm/i915: Move psr unlock out from the pipe update critical section

2023-08-27 Thread Ville Syrjala 
From: Ville Syrjälä 

Do the PSR unlock after the vblank evade critcal section is
fully over, not before.

Cc: Manasi Navare 
Signed-off-by: Ville Syrjälä 
---
 drivers/gpu/drm/i915/display/intel_crtc.c | 9 +
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/i915/display/intel_crtc.c 
b/drivers/gpu/drm/i915/display/intel_crtc.c
index 182c6dd64f47..5caa928e5ce9 100644
--- a/drivers/gpu/drm/i915/display/intel_crtc.c
+++ b/drivers/gpu/drm/i915/display/intel_crtc.c
@@ -646,10 +646,8 @@ void intel_pipe_update_end(struct intel_crtc_state 
*new_crtc_state)
ktime_t end_vbl_time = ktime_get();
struct drm_i915_private *dev_priv = to_i915(crtc->base.dev);
 
-   intel_psr_unlock(new_crtc_state);
-
if (new_crtc_state->do_async_flip)
-   return;
+   goto out;
 
trace_intel_pipe_update_end(crtc, end_vbl_count, scanline_end);
 
@@ -709,7 +707,7 @@ void intel_pipe_update_end(struct intel_crtc_state 
*new_crtc_state)
local_irq_enable();
 
if (intel_vgpu_active(dev_priv))
-   return;
+   goto out;
 
if (crtc->debug.start_vbl_count &&
crtc->debug.start_vbl_count != end_vbl_count) {
@@ -724,4 +722,7 @@ void intel_pipe_update_end(struct intel_crtc_state 
*new_crtc_state)
}
 
dbg_vblank_evade(crtc, end_vbl_time);
+
+out:
+   intel_psr_unlock(new_crtc_state);
 }
-- 
2.41.0