There is no underrun expected (that I know of) when coming out of training on recent product generations. You should undo this masking and debug.
-----Original Message----- From: Vivi, Rodrigo <[email protected]> Sent: Wednesday, June 26, 2024 8:13 AM To: Nautiyal, Ankit K <[email protected]>; Runyan, Arthur J <[email protected]> Cc: [email protected]; [email protected] Subject: Re: [PATCH] drm/i915/dp: Wait more before rearming FIFO underrun during retrain On Wed, Jun 26, 2024 at 02:06:24PM +0530, Ankit Nautiyal wrote: > During Link re-training reporting underrun is disabled and then > renabled after re-training is completed. For BMG its seen that we get > FIFO underrun just after the retraining is completed and the underrun > reporting is re-enabled. > Add one more intel_crtc_wait_for_next_vblank before re-arming the > underruns. Cc: Arthur Runyan <[email protected]> Art, any new workaround in BMG for this issue? > > Signed-off-by: Ankit Nautiyal <[email protected]> > --- > drivers/gpu/drm/i915/display/intel_dp.c | 7 ++++++- > 1 file changed, 6 insertions(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/i915/display/intel_dp.c > b/drivers/gpu/drm/i915/display/intel_dp.c > index 3903f6ead6e6..25af51499383 100644 > --- a/drivers/gpu/drm/i915/display/intel_dp.c > +++ b/drivers/gpu/drm/i915/display/intel_dp.c > @@ -5325,7 +5325,12 @@ static int intel_dp_retrain_link(struct intel_encoder > *encoder, > const struct intel_crtc_state *crtc_state = > to_intel_crtc_state(crtc->base.state); > > - /* Keep underrun reporting disabled until things are stable */ > + /* > + * Keep underrun reporting disabled until things are stable. > + * Wait for some more time, as we see (at least on BMG) that > + * underrun gets reported just after the reporting is enabled. > + */ > + intel_crtc_wait_for_next_vblank(crtc); > intel_crtc_wait_for_next_vblank(crtc); > > intel_set_cpu_fifo_underrun_reporting(dev_priv, crtc->pipe, > true); > -- > 2.40.1 >
