[Intel-gfx] [PATCH RESEND] drm: i915: Don't try detecting sinks on ports already in use

[Gabriel Krisman Bertazi]

Chris Wilson <chris at chris-wilson.co.uk> writes:

> The key problem here is say a race between DP unplug and HDMI plug, and
> users are evil enough (or common enough) for it to happen.
>
> I thought the idea was reasonable though, and perhaps we could make
> more use of the knowlege of the shared ports to improve detection of
> common DP/HDMI DDIs (i.e. run detection once for a ddi and have it
> decide whether it is hdmi or dp).

Hi Chris and Manasi, thanks for the feedback and sorry for the late
response.

I see your point about being racy.  I think this is not an issue in my
system since it appears that the disconnect of the first connector
triggers another scheduling of the hotplug_work, that further detects
the other connector.

I'm still working on this, but can you provide more details on your
suggestion about using more knowledge of the shared ports to improve the
detection of DDIs?  I still don't see how we could prevent this race
once and for all.

In the mean time, I wrote another patch with a different way to reduce
the overhead of attempting to detect connectors of shared ports already
in use.  My understanding is that this version is no longer racy but it
doesn't catch every condition where we could avoid detection.  Though,
it feels like something we want to get upstream.  Can you please take a
look and let me know?

I think this also addresses the cloned encoders issue raised by Manasi,
if I understood it correctly.

>8
Subject: [PATCH] drm: i915: Stop further detection on shared pins if a sink is
 connected

HPD pins may be shared among connectors but, according to documentation,
only one connector can be enebled at any given time.  This avoids
waisting time with attempting to detect further connectors on the same
pin if the first one was already detected.

Since this reduces the overhead of the i915_hotplug_work_func, it may
address the following vblank misses detected by the CI:

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100215
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100558

Signed-off-by: Gabriel Krisman Bertazi <krisman at collabora.co.uk>
---
 drivers/gpu/drm/i915/intel_hotplug.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_hotplug.c b/drivers/gpu/drm/i915/intel_hotplug.c
index f1200272a699..913e8523b89d 100644
--- a/drivers/gpu/drm/i915/intel_hotplug.c
+++ b/drivers/gpu/drm/i915/intel_hotplug.c
@@ -319,6 +319,7 @@ static void i915_hotplug_work_func(struct work_struct *work)
 	struct drm_connector_list_iter conn_iter;
 	bool changed = false;
 	u32 hpd_event_bits;
+	u32 enc_hpd_pin_mask;
 
 	mutex_lock(&dev->mode_config.mutex);
 	DRM_DEBUG_KMS("running encoder hotplug functions\n");
@@ -339,13 +340,18 @@ static void i915_hotplug_work_func(struct work_struct *work)
 		if (!intel_connector->encoder)
 			continue;
 		intel_encoder = intel_connector->encoder;
-		if (hpd_event_bits & (1 << intel_encoder->hpd_pin)) {
+		enc_hpd_pin_mask = (1 << intel_encoder->hpd_pin);
+		if (hpd_event_bits & enc_hpd_pin_mask) {
 			DRM_DEBUG_KMS("Connector %s (pin %i) received hotplug event.\n",
 				      connector->name, intel_encoder->hpd_pin);
 			if (intel_encoder->hot_plug)
 				intel_encoder->hot_plug(intel_encoder);
-			if (intel_hpd_irq_event(dev, connector))
+			if (intel_hpd_irq_event(dev, connector)) {
 				changed = true;
+				if (connector->status ==
+				    connector_status_connected)
+					hpd_event_bits &= ~(enc_hpd_pin_mask);
+			}
 		}
 	}
 	drm_connector_list_iter_end(&conn_iter);
-- 
2.11.0