Skip to contentSkip to conversation

drm/i915/dp: add support for load detect Apple DP->VGA dongles

RFC PATCH In review Dave Airlie 3 participants 3 messages Jun 6, 2014 – Jun 19, 2014

Subsystems (3)
lore
Share & tools
lore

Dave Airlie

author maintainer
#0 lore raw

RFC PATCH 1/1 drm/i915/dp: add support for load detect Apple DP->VGA dongles

From: Dave Airlie <[email protected]>

The DP1.2 spec has some bits for forced load sensing on VGA dongles, however the apple dongle I have doesn't do DP 1.2 yet, so I dug into its vendor specific area and found a register that seems to correspond to load detected on the outputs.

The main reason I need this was at LCA this year the slide capture system failed to provide EDID, but I realised OSX worked. Really need to add support to nouveau, but hey i915 is a start.

The dongle also appears to use short IRQs to denote a plug/unplug event, though something seems to be failing in reading back the dpcd values from it. (also this is based on my MST work just because.)

---
 drivers/gpu/drm/i915/intel_dp.c  | 29 +++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 include/drm/drm_dp_helper.h      | 19 +++++++++++++++++++
 3 files changed, 49 insertions(+)
drivers/gpu/drm/i915/intel_dp.c +29 -0
index c500f63..49afd7d 100644
@@ -3173,6 +3173,16 @@ intel_dp_probe_oui(struct intel_dp *intel_dp)
 		DRM_DEBUG_KMS("Branch OUI: %02hx%02hx%02hx\n",
 			      buf[0], buf[1], buf[2]);
 
+	intel_dp->is_apple_vga = false;
+	if (drm_dp_branch_is_apple(buf)) {
+		u8 mybuf[8];
+		if (intel_dp_dpcd_read_wake(&intel_dp->aux, DP_BRANCH_OUI + 3, mybuf, 8)) {
+			if (drm_dp_apple_has_load_detect(mybuf)) {
+				DRM_DEBUG_KMS("detected Apple DP VGA dongle\n");
+				intel_dp->is_apple_vga = true;
+			}
+		}
+	}
 	edp_panel_vdd_off(intel_dp, false);
 }
 
@@ -3404,6 +3414,21 @@ intel_dp_detect_dpcd(struct intel_dp *intel_dp)
 	if (drm_probe_ddc(&intel_dp->aux.ddc))
 		return connector_status_connected;
 
+	if (intel_dp->is_apple_vga) {
+		u8 detect_buf;
+		int ret;
+		ret = drm_dp_dpcd_readb(&intel_dp->aux, DP_APPLE_LOAD_DETECT,
+					&detect_buf);
+
+		if (ret == 1) {
+			DRM_DEBUG_KMS("Apple VGA detect is 0x%x\n", detect_buf);
+			/* I suspect 4 == load, 8 == edid */
+			if (detect_buf)
+				return connector_status_connected;
+			else
+				return connector_status_disconnected;
+		}
+	}
 	/* Well we tried, say unknown for unreliable port types */
 	if (intel_dp->dpcd[DP_DPCD_REV] >= 0x11) {
 		type = intel_dp->downstream_ports[0] & DP_DS_PORT_TYPE_MASK;
@@ -3854,6 +3879,10 @@ intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd)
 			 * but for short hpds we should check it now
 			 */
 			intel_dp_check_link_status(intel_dp);
+
+			/* if we get a short hpd on apple vga - its a hotplug */
+			if (intel_dp->is_apple_vga)
+				return true;
 		}
 	}
 	return false;
drivers/gpu/drm/i915/intel_drv.h +1 -0
index 1edb38a..23f767c 100644
@@ -549,6 +549,7 @@ struct intel_dp {
 	bool use_tps3;
 	bool can_mst; /* this port supports mst */
 	bool is_mst;
+	bool is_apple_vga;
 	int active_mst_links;
 	/* connector directly attached - won't be use for modeset in mst world */
 	struct intel_connector *attached_connector;
include/drm/drm_dp_helper.h +19 -0
index 517d6c1..c8ebd27 100644
@@ -612,4 +612,23 @@ int drm_dp_link_configure(struct drm_dp_aux *aux, struct drm_dp_link *link);
 int drm_dp_aux_register(struct drm_dp_aux *aux);
 void drm_dp_aux_unregister(struct drm_dp_aux *aux);
 
+#define DP_APPLE_OUI 0x10fa
+
+#define DP_APPLE_LOAD_DETECT (DP_BRANCH_OUI + 12)
+
+static inline bool drm_dp_branch_is_apple(const u8 buf[3])
+{
+	if (buf[0] == ((DP_APPLE_OUI >> 16) & 0xff) &&
+	    buf[1] == ((DP_APPLE_OUI >> 8) & 0xff) &&
+	    buf[2] == ((DP_APPLE_OUI & 0xff)))
+		return true;
+	return false;
+}
+
+static inline bool drm_dp_apple_has_load_detect(const u8 buf[8])
+{
+	if (!memcmp((const char *)buf, "mVGAa", 5))
+		return true;
+	return false;
+}
 #endif /* _DRM_DP_HELPER_H_ */
Signature · 1 line
1.9.3
maintainer
#1 lore raw
Quoted 7 lines · Dave Airlie

From: Dave Airlie <[email protected]>

The DP1.2 spec has some bits for forced load sensing on VGA dongles, however the apple dongle I have doesn't do DP 1.2 yet, so I dug into its vendor specific area and found a register that seems to correspond to load detected on the outputs.

Quoted 7 lines

The main reason I need this was at LCA this year the slide capture system failed to provide EDID, but I realised OSX worked. Really need to add support to nouveau, but hey i915 is a start.

The dongle also appears to use short IRQs to denote a plug/unplug event, though something seems to be failing in reading back the dpcd values from it. (also this is based on my MST work just because.)

Good timing for making use of the OUI! See

http://mid.gmane.org/[email protected]

Quoted 12 lines
---
 drivers/gpu/drm/i915/intel_dp.c  | 29 +++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 include/drm/drm_dp_helper.h      | 19 +++++++++++++++++++
 3 files changed, 49 insertions(+)
drivers/gpu/drm/i915/intel_dp.c +0 -0
index c500f63..49afd7d 100644

drivers/gpu/drm/i915/intel_dp.c intel_dp_probe_oui(struct intel_dp *intel_dp) · line 3179

Quoted 7 lines
drivers/gpu/drm/i915/intel_dp.c +4 -0
@@ -3173,6 +3173,16 @@ intel_dp_probe_oui(struct intel_dp *intel_dp)
 		DRM_DEBUG_KMS("Branch OUI: %02hx%02hx%02hx\n",
 			      buf[0], buf[1], buf[2]);
 
+	intel_dp->is_apple_vga = false;
+	if (drm_dp_branch_is_apple(buf)) {
+		u8 mybuf[8];
+		if (intel_dp_dpcd_read_wake(&intel_dp->aux, DP_BRANCH_OUI + 3, mybuf, 8)) {

That may return negative values for errors.

FWIW I've thought about reading the sink/branch device identification strings unconditionally along with the OUI just for debug purposes. I wouldn't be opposed to such a change here.

drivers/gpu/drm/i915/intel_dp.c intel_dp_probe_oui(struct intel_dp *intel_dp) · line 3185

Quoted 6 lines
drivers/gpu/drm/i915/intel_dp.c +6 -0
+			if (drm_dp_apple_has_load_detect(mybuf)) {
+				DRM_DEBUG_KMS("detected Apple DP VGA dongle\n");
+				intel_dp->is_apple_vga = true;
+			}
+		}
+	}

I think I'd like this abstracted to a dedicated function and doing:

	intel_dp->is_apple_vga = is_apple_vga(intel_dp, buf);

or something.

Quoted 53 lines
drivers/gpu/drm/i915/intel_dp.c +19 -0
 	edp_panel_vdd_off(intel_dp, false);
 }
 
@@ -3404,6 +3414,21 @@ intel_dp_detect_dpcd(struct intel_dp *intel_dp)
 	if (drm_probe_ddc(&intel_dp->aux.ddc))
 		return connector_status_connected;
 
+	if (intel_dp->is_apple_vga) {
+		u8 detect_buf;
+		int ret;
+		ret = drm_dp_dpcd_readb(&intel_dp->aux, DP_APPLE_LOAD_DETECT,
+					&detect_buf);
+
+		if (ret == 1) {
+			DRM_DEBUG_KMS("Apple VGA detect is 0x%x\n", detect_buf);
+			/* I suspect 4 == load, 8 == edid */
+			if (detect_buf)
+				return connector_status_connected;
+			else
+				return connector_status_disconnected;
+		}
+	}
 	/* Well we tried, say unknown for unreliable port types */
 	if (intel_dp->dpcd[DP_DPCD_REV] >= 0x11) {
 		type = intel_dp->downstream_ports[0] & DP_DS_PORT_TYPE_MASK;
@@ -3854,6 +3879,10 @@ intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd)
 			 * but for short hpds we should check it now
 			 */
 			intel_dp_check_link_status(intel_dp);
+
+			/* if we get a short hpd on apple vga - its a hotplug */
+			if (intel_dp->is_apple_vga)
+				return true;
 		}
 	}
 	return false;
drivers/gpu/drm/i915/intel_drv.h +1 -0
index 1edb38a..23f767c 100644
@@ -549,6 +549,7 @@ struct intel_dp {
 	bool use_tps3;
 	bool can_mst; /* this port supports mst */
 	bool is_mst;
+	bool is_apple_vga;
 	int active_mst_links;
 	/* connector directly attached - won't be use for modeset in mst world */
 	struct intel_connector *attached_connector;
include/drm/drm_dp_helper.h +0 -0
index 517d6c1..c8ebd27 100644

include/drm/drm_dp_helper.h int drm_dp_link_configure(struct drm_dp_aux *aux, struct drm_dp_link *link); · line 619

Quoted 8 lines
include/drm/drm_dp_helper.h +5 -0
@@ -612,4 +612,23 @@ int drm_dp_link_configure(struct drm_dp_aux *aux, struct drm_dp_link *link);
 int drm_dp_aux_register(struct drm_dp_aux *aux);
 void drm_dp_aux_unregister(struct drm_dp_aux *aux);
 
+#define DP_APPLE_OUI 0x10fa
+
+#define DP_APPLE_LOAD_DETECT (DP_BRANCH_OUI + 12)
+
+static inline bool drm_dp_branch_is_apple(const u8 buf[3])

Naming buf and its size similar to other drm_dp_* helpers use would be nice.

include/drm/drm_dp_helper.h int drm_dp_link_configure(struct drm_dp_aux *aux, struct drm_dp_link *link); · line 628

Quoted 9 lines
include/drm/drm_dp_helper.h +9 -0
+{
+	if (buf[0] == ((DP_APPLE_OUI >> 16) & 0xff) &&
+	    buf[1] == ((DP_APPLE_OUI >> 8) & 0xff) &&
+	    buf[2] == ((DP_APPLE_OUI & 0xff)))
+		return true;
+	return false;
+}
+
+static inline bool drm_dp_apple_has_load_detect(const u8 buf[8])

Ditto.

include/drm/drm_dp_helper.h int drm_dp_link_configure(struct drm_dp_aux *aux, struct drm_dp_link *link); · line 630

Quoted 2 lines
include/drm/drm_dp_helper.h +2 -0
+{
+	if (!memcmp((const char *)buf, "mVGAa", 5))

Why the cast?

BR,
Jani.

Quoted 11 lines
include/drm/drm_dp_helper.h +3 -0
+		return true;
+	return false;
+}
 #endif /* _DRM_DP_HELPER_H_ */
Signature · 6 lines
1.9.3

_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 1 line
Jani Nikula, Intel Open Source Technology Center
#2 lore raw
Quoted 12 lines · Jani Nikula
Quoted 11 lines · Dave Airlie

From: Dave Airlie <[email protected]>

The DP1.2 spec has some bits for forced load sensing on VGA dongles, however the apple dongle I have doesn't do DP 1.2 yet, so I dug into its vendor specific area and found a register that seems to correspond to load detected on the outputs.

The main reason I need this was at LCA this year the slide capture system failed to provide EDID, but I realised OSX worked. Really need to add support to nouveau, but hey i915 is a start.

Quoted 9 lines
Quoted 3 lines

The dongle also appears to use short IRQs to denote a plug/unplug event, though something seems to be failing in reading back the dpcd values from it. (also this is based on my MST work just because.)

Good timing for making use of the OUI! See

http://mid.gmane.org/[email protected]

Quoted 1 line

Just a high-level bikeshed: Now that we have this nice dp helper library shouldn't we add a dp_detect function there which fills out a bunch of sink state in the core dp struct and also implements tricks like this? Maybe as the plain dp version of the relevant dp mst handler?

Definitely not much point in copy&pasting this for radeon, noveau, tegra and all the others.
-Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch