On Mon, Sep 23, 2024 at 11:09:19AM -0700, Matt Roper wrote: > On Mon, Sep 23, 2024 at 08:38:35PM +0300, Ville Syrjälä wrote: > > On Mon, Sep 23, 2024 at 08:40:07AM +0530, Suraj Kandpal wrote: > > > Reduce SHPD_CNT to 250us for display version 12 as it lines up > > > with DP1.4a(Table3-4) spec. > > > > > > --v2 > > > -Update commit message and comment [Matt] > > > > > > Signed-off-by: Suraj Kandpal <[email protected]> > > > --- > > > drivers/gpu/drm/i915/display/intel_hotplug_irq.c | 6 +++++- > > > 1 file changed, 5 insertions(+), 1 deletion(-) > > > > > > diff --git a/drivers/gpu/drm/i915/display/intel_hotplug_irq.c > > > b/drivers/gpu/drm/i915/display/intel_hotplug_irq.c > > > index 2c4e946d5575..8427386132e6 100644 > > > --- a/drivers/gpu/drm/i915/display/intel_hotplug_irq.c > > > +++ b/drivers/gpu/drm/i915/display/intel_hotplug_irq.c > > > @@ -849,7 +849,11 @@ static void icp_hpd_irq_setup(struct > > > drm_i915_private *dev_priv) > > > enabled_irqs = intel_hpd_enabled_irqs(dev_priv, > > > dev_priv->display.hotplug.pch_hpd); > > > hotplug_irqs = intel_hpd_hotplug_irqs(dev_priv, > > > dev_priv->display.hotplug.pch_hpd); > > > > > > - if (INTEL_PCH_TYPE(dev_priv) <= PCH_TGP) > > > + /* > > > + * We reduce the value to 250us to be able to detect SHPD when an > > > external display > > > + * is connected. This is also expected of us as stated in DP1.4a Table > > > 3-4. > > > + */ > > > + if (INTEL_PCH_TYPE(dev_priv) < PCH_TGP) > > > intel_uncore_write(&dev_priv->uncore, SHPD_FILTER_CNT, > > > SHPD_FILTER_CNT_500_ADJ); > > > else > > > intel_uncore_write(&dev_priv->uncore, SHPD_FILTER_CNT, > > > SHPD_FILTER_CNT_250); > > > > IMO if we start reducing this for older platforms then we > > should just do it for all of them, instead of based on some > > random cutoff. > > There is a note in the bspec (page 4342) that says > > "Program SHPD_FILTER_CNT with the "500 microseconds adjusted" > value before enabling hotplug detection." > > for CNP and ICP which is where the cutoff originally came from. I'm not > sure about CML (which uses CNP), but ICL at least supports DP1.4a, so I > think the note in the bspec is probably just outdated and we'd still > want 250 to ensure we align with the DP spec.
The 250usec has been in the DP spec forever. DP 1.4a vs. older is a moot point. Also, this is about filtering out glitches generated by the sink, and anyone can plug any DP sink to any DP source. So if the 500us is a problem for TGP+ then it'll be a problem for LPT+ as well when using the same sink. For IBX/CPT it seems we can't adjust this to a more sensible value (the only choice seems to be 500us vs. 50us), so LPT+ is the hard cutoff unfortunately. -- Ville Syrjälä Intel
