[PATCH 0/8] drm/mxsfb/lcdif: use DRM_BRIDGE_ATTACH_NO_CONNECTOR and the bridge-connector

[PATCH 0/8] drm/mxsfb/lcdif: use DRM_BRIDGE_ATTACH_NO_CONNECTOR and the bridge-connector

[Luca Ceresoli]

This series modernizes the i.mx8mp LCDIF driver to use the
bridge-connector, which is the current best practice in DRM.

== Call for testing on i.MX8MP boards (especially those using HDMI)!

This series applies changes to how video output devices are probed on
i.MX8MP, especially those using HDMI. Even though I have put care in not
breaking anything, there could potentially be pitfalls I haven't realized,
causing regressions on existing boards.

I have thus added in Cc all developers which appeared active on dts files
for imx8mp boards involving video. I would appreciate testing on as many
boards as possible, along with a Tested-by tag, or a report about any
issues encountered.

Thanks in advance to all testers!

== Review recommendation

I recommend reviewing patches in this order to be understood more
effectively:

 * Cover letter
 * Patches 1-5 are small preliminary cleanups/improvements
 * Patch 8 is the goal of this series, but would not work alone
 * Patch 7 this lets patch 8 work; but in turn it can't work alone
 * Patch 6 lets patch 7 work

== Series description

This series is not strictly related to DRM bridge hotplug, it is rather a
preparation step. Introducing hotplug would need two different approaches:
one for the new way, for drivers using bridge-connector and
DRM_BRIDGE_ATTACH_NO_CONNECTOR, another for drivers using the "old, legacy
way" where the last bridge is supposed to instantiate the
drm_connector. Hotplug is complicated enough in one case, so it makes sense
to only support the new way.

The hardware I'm working on is an i.MX8MP, whose LCDIF driver is still
using the old way. So this series converts to the new way as a preparation
step.

Patch 8 does the conversion, which is simple. However this would introduce
a regression on some boards. Here's why:

There are 3 instances of the LCDIF in i.MX8MP:

 * LCDIF1, driving the DSI output
 * LCDIF2, driving the LVDS output
 * LCDIF3, driving the HDMI output

The device drivers of peripherals connected to LCDIF1 and LCDIF2 already
support the DRM_BRIDGE_ATTACH_NO_CONNECTOR flag. So far so good.

LCDIF3 is more tricky. The HDMI pipeline is:

  LCDIF3 -> fsl,imx8mp-hdmi-pvi -> fsl,imx8mp-hdmi-tx -> HDMI connector

The fsl,imx8mp-hdmi-tx (hdmi-tx) component supports both cases (with or
without DRM_BRIDGE_ATTACH_NO_CONNECTOR), but in the
DRM_BRIDGE_ATTACH_NO_CONNECTOR case it does not create a drm_connector,
thus preventing the creation of the pipeline. To make it work a connector
must be created, and the way to do so is describing the connector in the
device tree (compatible = "hdmi-connector"), so the display-driver will add
the connector along with a wrapping bridge.

Unfortunately not all device trees in mainline have an hdmi-connector
node. Adding one is easy, but would break existing hardware upgrading to a
newer kernel without upgrading the device tree blob. This is addressed by
patch 7 reusing an existing approach to add such a node to the live device
tree at init time using a device tree overlay for boards which don't have
one.

Finally, patch 7 cannot work alone because of a bad interaction between
devlink and device tree overlays. Patch 6 solves that.

== Grand plan

This is part of the work to support hotplug of DRM bridges. The grand plan
was discussed in [0].

Here's the work breakdown (➜ marks the current series):

 1. … add refcounting to DRM bridges struct drm_bridge,
      based on devm_drm_bridge_alloc()
    A. ✔ add new alloc API and refcounting (v6.16)
    B. ✔ convert all bridge drivers to new API (v6.17)
    C. ✔ kunit tests (v6.17)
    D. ✔ add get/put to drm_bridge_add/remove() + attach/detach()
         and warn on old allocation pattern (v6.17)
    E. … add get/put on drm_bridge accessors
       1. ✔ drm_bridge_chain_get_first_bridge(), add cleanup action (v6.18)
       2. ✔ drm_bridge_get_prev_bridge() (v6.18)
       3. ✔ drm_bridge_get_next_bridge() (v6.19)
       4. ✔ drm_for_each_bridge_in_chain() (v6.19)
       5. ✔ drm_bridge_connector_init (v6.19)
       6. … protect encoder bridge chain with a mutex
       7. … of_drm_find_bridge
          a. ✔ add of_drm_get_bridge() (v7.0),
	       convert basic direct users (v7.0-v7.1)
	  b. ✔ convert direct of_drm_get_bridge() users, part 2 (v7.0)
	  c. ✔ convert direct of_drm_get_bridge() users, part 3 (v7.0)
	  d. ✔… convert direct of_drm_get_bridge() users, part 4
	        (some v7.1, some pending)
	  e.   convert bridge-only drm_of_find_panel_or_bridge() users
       8. drm_of_find_panel_or_bridge, *_of_get_bridge
       9. ✔ enforce drm_bridge_add before drm_bridge_attach (v6.19)
    F. ✔ debugfs improvements
       1. ✔ add top-level 'bridges' file (v6.16)
       2. ✔ show refcount and list lingering bridges (v6.19)
 2. … handle gracefully atomic updates during bridge removal
    A. ✔ Add drm_bridge_enter/exit() to protect device resources (v7.0)
    B. … protect private_obj removal from list
    C. ✔ Add drm_bridge_clear_and_put() (v7.1)
 3. … DSI host-device driver interaction
 4. ✔ removing the need for the "always-disconnected" connector
 5. ➜ Migrate i.MX LCDIF driver to bridge-connector
 6.   DRM bridge hotplug
    A.   Bridge hotplug management in the DRM core
    B.   Device tree description

[0] https://lore.kernel.org/lkml/20250206-hotplug-drm-bridge-v6-0-9d6f2c9c3058@bootlin.com/#t

Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
---
Luca Ceresoli (8):
      drm/mxsfb/lcdif: simplify remote pointer management using __free
      drm/mxsfb/lcdif: don't unnecessarily loop over ports
      drm/mxsfb/lcdif: use dev_err_probe() consistently in lcdif_attach_bridge
      drm/bridge: dw-hdmi: document the output_port field
      drm/bridge: dw-hdmi: warn on unsupported attach combination
      drm/bridge: dw-hdmi: move next_bridge lookup to attach time
      drm/bridge: imx8mp-hdmi-tx: add an hdmi-connector when missing using a DT overlay at boot time
      drm/mxsfb/lcdif: use DRM_BRIDGE_ATTACH_NO_CONNECTOR and the bridge-connector

 drivers/gpu/drm/bridge/imx/Kconfig                 | 17 +++++
 drivers/gpu/drm/bridge/imx/Makefile                |  2 +
 .../bridge/imx/imx8mp-hdmi-tx-connector-fixup.c    | 60 +++++++++++++++
 .../bridge/imx/imx8mp-hdmi-tx-connector-fixup.dtso | 38 ++++++++++
 drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx.c        |  1 +
 drivers/gpu/drm/bridge/synopsys/dw-hdmi.c          | 45 ++++-------
 drivers/gpu/drm/mxsfb/Kconfig                      |  1 +
 drivers/gpu/drm/mxsfb/lcdif_drv.c                  | 88 +++++++++-------------
 include/drm/bridge/dw_hdmi.h                       |  5 ++
 9 files changed, 174 insertions(+), 83 deletions(-)
---
base-commit: 8402cf4fc8f8d5756dc81cf9fda1dccdb3622634
change-id: 20260306-drm-lcdif-dbanc-83dd948327de

Best regards,
-- 
Luca Ceresoli <luca.ceresoli@bootlin.com>

[PATCH 1/8] drm/mxsfb/lcdif: simplify remote pointer management using __free

[Luca Ceresoli]

Putting the remote device_node reference requires a of_node_put(ep) in both
error return points. Use a cleanup action to simplify the code.

Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
---
 drivers/gpu/drm/mxsfb/lcdif_drv.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/mxsfb/lcdif_drv.c b/drivers/gpu/drm/mxsfb/lcdif_drv.c
index 47da1d9336b9..756ca96373c8 100644
--- a/drivers/gpu/drm/mxsfb/lcdif_drv.c
+++ b/drivers/gpu/drm/mxsfb/lcdif_drv.c
@@ -53,16 +53,13 @@ static int lcdif_attach_bridge(struct lcdif_drm_private *lcdif)
 	int ret;
 
 	for_each_endpoint_of_node(dev->of_node, ep) {
-		struct device_node *remote;
+		struct device_node *remote __free(drm_bridge_put) =
+			of_graph_get_remote_port_parent(ep);
 		struct of_endpoint of_ep;
 		struct drm_encoder *encoder;
 
-		remote = of_graph_get_remote_port_parent(ep);
-		if (!of_device_is_available(remote)) {
-			of_node_put(remote);
+		if (!of_device_is_available(remote))
 			continue;
-		}
-		of_node_put(remote);
 
 		ret = of_graph_parse_endpoint(ep, &of_ep);
 		if (ret < 0) {

-- 
2.53.0

[PATCH 2/8] drm/mxsfb/lcdif: don't unnecessarily loop over ports

[Luca Ceresoli]

According to the bindings [0] there can be only one port. The in-tree board
device trees also don't contain multiple ports (searched thos matching
'fsl,imx(23|28|6sx|8mp|93)-lcdif').

Avoid an unnecessary loop around multipltle ports. This allows to greatly
simplify the code.

[0] Documentation/devicetree/bindings/display/fsl,lcdif.yaml

Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>

---

Viewing this patch with '--ignore-all-space' is recommended
---
 drivers/gpu/drm/mxsfb/lcdif_drv.c | 77 ++++++++++++++-------------------------
 1 file changed, 27 insertions(+), 50 deletions(-)

diff --git a/drivers/gpu/drm/mxsfb/lcdif_drv.c b/drivers/gpu/drm/mxsfb/lcdif_drv.c
index 756ca96373c8..83e134c04882 100644
--- a/drivers/gpu/drm/mxsfb/lcdif_drv.c
+++ b/drivers/gpu/drm/mxsfb/lcdif_drv.c
@@ -48,61 +48,38 @@ static const struct drm_encoder_funcs lcdif_encoder_funcs = {
 static int lcdif_attach_bridge(struct lcdif_drm_private *lcdif)
 {
 	struct device *dev = lcdif->drm->dev;
-	struct device_node *ep;
+	struct drm_encoder *encoder;
 	struct drm_bridge *bridge;
 	int ret;
 
-	for_each_endpoint_of_node(dev->of_node, ep) {
-		struct device_node *remote __free(drm_bridge_put) =
-			of_graph_get_remote_port_parent(ep);
-		struct of_endpoint of_ep;
-		struct drm_encoder *encoder;
-
-		if (!of_device_is_available(remote))
-			continue;
-
-		ret = of_graph_parse_endpoint(ep, &of_ep);
-		if (ret < 0) {
-			dev_err(dev, "Failed to parse endpoint %pOF\n", ep);
-			of_node_put(ep);
-			return ret;
-		}
-
-		bridge = devm_drm_of_get_bridge(dev, dev->of_node, 0, of_ep.id);
-		if (IS_ERR(bridge)) {
-			of_node_put(ep);
-			return dev_err_probe(dev, PTR_ERR(bridge),
-					     "Failed to get bridge for endpoint%u\n",
-					     of_ep.id);
-		}
-
-		encoder = devm_kzalloc(dev, sizeof(*encoder), GFP_KERNEL);
-		if (!encoder) {
-			dev_err(dev, "Failed to allocate encoder for endpoint%u\n",
-				of_ep.id);
-			of_node_put(ep);
-			return -ENOMEM;
-		}
-
-		encoder->possible_crtcs = drm_crtc_mask(&lcdif->crtc);
-		ret = drm_encoder_init(lcdif->drm, encoder, &lcdif_encoder_funcs,
-				       DRM_MODE_ENCODER_NONE, NULL);
-		if (ret) {
-			dev_err(dev, "Failed to initialize encoder for endpoint%u: %d\n",
-				of_ep.id, ret);
-			of_node_put(ep);
-			return ret;
-		}
-
-		ret = drm_bridge_attach(encoder, bridge, NULL, 0);
-		if (ret) {
-			of_node_put(ep);
-			return dev_err_probe(dev, ret,
-					     "Failed to attach bridge for endpoint%u\n",
-					     of_ep.id);
-		}
+	struct device_node *remote __free(device_node) =
+		of_graph_get_remote_node(dev->of_node, 0, -1);
+
+	if (!of_device_is_available(remote))
+		return 0;
+
+	bridge = devm_drm_of_get_bridge(dev, dev->of_node, 0, -1);
+	if (IS_ERR(bridge))
+		return dev_err_probe(dev, PTR_ERR(bridge), "Failed to get bridge\n");
+
+	encoder = devm_kzalloc(dev, sizeof(*encoder), GFP_KERNEL);
+	if (!encoder) {
+		dev_err(dev, "Failed to allocate encoder\n");
+		return -ENOMEM;
 	}
 
+	encoder->possible_crtcs = drm_crtc_mask(&lcdif->crtc);
+	ret = drm_encoder_init(lcdif->drm, encoder, &lcdif_encoder_funcs,
+			       DRM_MODE_ENCODER_NONE, NULL);
+	if (ret) {
+		dev_err(dev, "Failed to initialize encoder: %d\n", ret);
+		return ret;
+	}
+
+	ret = drm_bridge_attach(encoder, bridge, NULL, 0);
+	if (ret)
+		return dev_err_probe(dev, ret, "Failed to attach bridge\n");
+
 	return 0;
 }
 

-- 
2.53.0

[PATCH 3/8] drm/mxsfb/lcdif: use dev_err_probe() consistently in lcdif_attach_bridge

[Luca Ceresoli]

lcdif_attach_bridge() uses dev_err_probe() in some error paths, dev_err() +
return in others. Use dev_err_probe() for all of them to make code
consistent and simpler.

Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
---
 drivers/gpu/drm/mxsfb/lcdif_drv.c | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/mxsfb/lcdif_drv.c b/drivers/gpu/drm/mxsfb/lcdif_drv.c
index 83e134c04882..e40253ad607b 100644
--- a/drivers/gpu/drm/mxsfb/lcdif_drv.c
+++ b/drivers/gpu/drm/mxsfb/lcdif_drv.c
@@ -63,18 +63,14 @@ static int lcdif_attach_bridge(struct lcdif_drm_private *lcdif)
 		return dev_err_probe(dev, PTR_ERR(bridge), "Failed to get bridge\n");
 
 	encoder = devm_kzalloc(dev, sizeof(*encoder), GFP_KERNEL);
-	if (!encoder) {
-		dev_err(dev, "Failed to allocate encoder\n");
-		return -ENOMEM;
-	}
+	if (!encoder)
+		return dev_err_probe(dev, -ENOMEM, "Failed to allocate encoder\n");
 
 	encoder->possible_crtcs = drm_crtc_mask(&lcdif->crtc);
 	ret = drm_encoder_init(lcdif->drm, encoder, &lcdif_encoder_funcs,
 			       DRM_MODE_ENCODER_NONE, NULL);
-	if (ret) {
-		dev_err(dev, "Failed to initialize encoder: %d\n", ret);
-		return ret;
-	}
+	if (ret)
+		return dev_err_probe(dev, ret, "Failed to initialize encoder\n");
 
 	ret = drm_bridge_attach(encoder, bridge, NULL, 0);
 	if (ret)

-- 
2.53.0

[PATCH 4/8] drm/bridge: dw-hdmi: document the output_port field

[Luca Ceresoli]

The meaning of this flag may not be obvious at first sight.

Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
---
 include/drm/bridge/dw_hdmi.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/include/drm/bridge/dw_hdmi.h b/include/drm/bridge/dw_hdmi.h
index 336f062e1f9d..45f6ba1a8ee1 100644
--- a/include/drm/bridge/dw_hdmi.h
+++ b/include/drm/bridge/dw_hdmi.h
@@ -126,6 +126,11 @@ struct dw_hdmi_phy_ops {
 struct dw_hdmi_plat_data {
 	struct regmap *regm;
 
+	/*
+	 * The HDMI output port number (which must be 1) if it is described
+	 * in the device tree. 0 if the device tree does not describe the
+	 * next component (legacy mode).
+	 */
 	unsigned int output_port;
 
 	unsigned long input_bus_encoding;

-- 
2.53.0

[PATCH 5/8] drm/bridge: dw-hdmi: warn on unsupported attach combination

[Luca Ceresoli]

dw-hdmi can operate in two different modes, depending on the platform data
as set by the driver:

 A. hdmi->plat_data->output_port = 0:
    the HDMI output (port@1) in device tree is not used

 B. hdmi->plat_data->output_port = 1:
    the HDMI output (port@1) is parsed to find the next bridge

Only case B is supported when the DRM_BRIDGE_ATTACH_NO_CONNECTOR flag is
passed to the attach callback. Emit a warning when this is violated. Also
return -EINVAL which would be returned by drm_bridge_attach() right after
anyway.

Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>

---

Note: Returning when teh warning triggers does not change the functional
behaviour of this function. It is not strictly necessary in this patch but
it will have to be done anyway in the following patch.
---
 drivers/gpu/drm/bridge/synopsys/dw-hdmi.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c
index ee88c0e793b0..a668d66aeece 100644
--- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c
+++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c
@@ -2910,6 +2910,10 @@ static int dw_hdmi_bridge_attach(struct drm_bridge *bridge,
 {
 	struct dw_hdmi *hdmi = bridge->driver_private;
 
+	/* DRM_BRIDGE_ATTACH_NO_CONNECTOR requires a remote-endpoint to the next bridge */
+	if (WARN_ON((flags & DRM_BRIDGE_ATTACH_NO_CONNECTOR) && !hdmi->plat_data->output_port))
+		return -EINVAL;
+
 	if (flags & DRM_BRIDGE_ATTACH_NO_CONNECTOR)
 		return drm_bridge_attach(encoder, hdmi->bridge.next_bridge,
 					 bridge, flags);

-- 
2.53.0

[PATCH 6/8] drm/bridge: dw-hdmi: move next_bridge lookup to attach time

[Luca Ceresoli]

This driver looks up the next_bridge at probe time and stores it in
hdmi->bridge.next_bridge, but only uses the stored value when attaching,
and only in the DRM_BRIDGE_ATTACH_NO_CONNECTOR case.

This will be problematic with an upcoming change, adding an hdmi-connector
using a device tree overlay when not present. That change is in turn
necessary to migrate the i.MX LCDIF driver to the bridge-connector.

The problem is that, adding the hdmi-connector via an overlay, devlink
considers hdmi-connector a consumer of the dw-hdmi device, generating a
chicken-egg problem:

 * hdmi-connector probe won't be tried until dw-hdmi is probed (devlink)
 * dw-hdmi probe will defer until it finds the next_bridge (the
   hdmi-connector wrapper bridge)

In preparation for those changes, move the next_bridge lookup from probe to
attach, when it is actually used. This allows dw-hdmi to probe, so that the
hdmi-connector can probe as well.

Also avoid storing the pointer in hdmi->bridge.next_bridge: the value is
computed when needed, thus a local variable is enough.

Finally, this also allows to slightly improve the code by not doing any DT
lookup in the !DRM_BRIDGE_ATTACH_NO_CONNECTOR case.

Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
---
 drivers/gpu/drm/bridge/synopsys/dw-hdmi.c | 41 +++++++++----------------------
 1 file changed, 12 insertions(+), 29 deletions(-)

diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c
index a668d66aeece..4ee865a1a6c8 100644
--- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c
+++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c
@@ -2914,9 +2914,18 @@ static int dw_hdmi_bridge_attach(struct drm_bridge *bridge,
 	if (WARN_ON((flags & DRM_BRIDGE_ATTACH_NO_CONNECTOR) && !hdmi->plat_data->output_port))
 		return -EINVAL;
 
-	if (flags & DRM_BRIDGE_ATTACH_NO_CONNECTOR)
-		return drm_bridge_attach(encoder, hdmi->bridge.next_bridge,
-					 bridge, flags);
+	if (flags & DRM_BRIDGE_ATTACH_NO_CONNECTOR) {
+		struct device_node *remote __free(device_node) =
+			of_graph_get_remote_node(hdmi->dev->of_node,
+						 hdmi->plat_data->output_port, -1);
+		if (!remote)
+			return -EPROBE_DEFER;
+
+		struct drm_bridge *next_bridge __free(drm_bridge_put) =
+			of_drm_find_and_get_bridge(remote);
+
+		return drm_bridge_attach(encoder, next_bridge, bridge, flags);
+	}
 
 	return dw_hdmi_connector_create(hdmi);
 }
@@ -3307,28 +3316,6 @@ static void dw_hdmi_init_hw(struct dw_hdmi *hdmi)
  * Probe/remove API, used from platforms based on the DRM bridge API.
  */
 
-static int dw_hdmi_parse_dt(struct dw_hdmi *hdmi)
-{
-	struct device_node *remote;
-
-	if (!hdmi->plat_data->output_port)
-		return 0;
-
-
-	remote = of_graph_get_remote_node(hdmi->dev->of_node,
-					  hdmi->plat_data->output_port,
-					  -1);
-	if (!remote)
-		return -ENODEV;
-
-	hdmi->bridge.next_bridge = of_drm_find_and_get_bridge(remote);
-	of_node_put(remote);
-	if (!hdmi->bridge.next_bridge)
-		return -EPROBE_DEFER;
-
-	return 0;
-}
-
 bool dw_hdmi_bus_fmt_is_420(struct dw_hdmi *hdmi)
 {
 	return hdmi_bus_fmt_is_yuv420(hdmi->hdmi_data.enc_out_bus_format);
@@ -3373,10 +3360,6 @@ struct dw_hdmi *dw_hdmi_probe(struct platform_device *pdev,
 	mutex_init(&hdmi->cec_notifier_mutex);
 	spin_lock_init(&hdmi->audio_lock);
 
-	ret = dw_hdmi_parse_dt(hdmi);
-	if (ret < 0)
-		return ERR_PTR(ret);
-
 	ddc_node = of_parse_phandle(np, "ddc-i2c-bus", 0);
 	if (ddc_node) {
 		hdmi->ddc = of_get_i2c_adapter_by_node(ddc_node);

-- 
2.53.0

[PATCH 7/8] drm/bridge: imx8mp-hdmi-tx: add an hdmi-connector when missing using a DT overlay at boot time

[Luca Ceresoli]

The imx8mp-hdmi-tx one of many drivers based on dw-hdmi. dw-hdmi in turn
can operate in two different modes, depending on the platform data as set
by the driver:

 A. hdmi->plat_data->output_port = 0:
    the HDMI output (port@1) in device tree is not used [0]

 B. hdmi->plat_data->output_port = 1:
    the HDMI output (port@1) is parsed to find the next bridge

The imx8mp-hdmi-tx driver falls in case A. This implies next_bridge will
always be NULL, and so dw_hdmi_bridge_attach() [1] will always fail if
called with the DRM_BRIDGE_ATTACH_NO_CONNECTOR flag.

In fact case A assumes that DRM_BRIDGE_ATTACH_NO_CONNECTOR is not set and
in that case it adds the connector programmatically at bridge attach time.

Support for DRM_BRIDGE_ATTACH_NO_CONNECTOR is implemented by dw-hdmi.c in
case B. So, in preparation to support DRM_BRIDGE_ATTACH_NO_CONNECTOR in
imx8mp-hdmi-tx, move to case B by setting hdmi->plat_data->output_port = 1.

However this change requires that port@1 is connected to a "next
bridge" DT node, typically the HDMI connector, because dw-hdmi won't add
the connector when using DRM_BRIDGE_ATTACH_NO_CONNECTOR.

Many dts files for imx8mp-based boards in the kernel have such a connector
described and linked to port@1, so a connector is added by the
display-connector driver along with a bridge wrapping it. Sadly some of
those dts files don't have the connector described. Adding it would solve
the problem easily, but this would break existing devices which do not
update the dtb when upgrading to a newer kernel.

To preserve backward compatibility for such devices, introduce a module
adding the hdmi-connector node to the live device tree at init time. This
allows the dw-hdmi code to find the next bridge (the one wrapping the
hdmi-connector) and let the pipeline work as before.

[0] https://elixir.bootlin.com/linux/v7.0-rc1/source/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c#L3310
[1] https://elixir.bootlin.com/linux/v7.0-rc1/source/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c#L2907

Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>

---

This patch is inspired by commit 0ff223d99147 ("drm/tilcdc: Convert legacy
panel binding via DT overlay at boot time")
---
 drivers/gpu/drm/bridge/imx/Kconfig                 | 17 ++++++
 drivers/gpu/drm/bridge/imx/Makefile                |  2 +
 .../bridge/imx/imx8mp-hdmi-tx-connector-fixup.c    | 60 ++++++++++++++++++++++
 .../bridge/imx/imx8mp-hdmi-tx-connector-fixup.dtso | 38 ++++++++++++++
 drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx.c        |  1 +
 5 files changed, 118 insertions(+)

diff --git a/drivers/gpu/drm/bridge/imx/Kconfig b/drivers/gpu/drm/bridge/imx/Kconfig
index b9028a5e5a06..b9ce140a93dc 100644
--- a/drivers/gpu/drm/bridge/imx/Kconfig
+++ b/drivers/gpu/drm/bridge/imx/Kconfig
@@ -25,6 +25,23 @@ config DRM_IMX8MP_DW_HDMI_BRIDGE
 	  Choose this to enable support for the internal HDMI encoder found
 	  on the i.MX8MP SoC.
 
+config DRM_IMX8MP_DW_HDMI_BRIDGE_CONNECTOR_FIXUP
+	bool "Support device tree blobs without an hdmi-connector node"
+	default y
+	depends on DRM_IMX8MP_DW_HDMI_BRIDGE
+	depends on OF
+	select OF_OVERLAY
+	select DRM_DISPLAY_CONNECTOR
+	help
+	  Modifies at early boot the live device tree of boards using the
+	  i.MX8MP fsl,imx8mp-hdmi-tx adding a hdmi-connector node linked to
+	  the htmi-tx. This is needed to support bridge-connector usage in
+	  the i.MX8MP LCDIF driver.
+
+	  You need this if you use the i.MX8MP HDMI output and your board
+	  device tree file does not have an hdmi-connector node connected
+	  to it.
+
 config DRM_IMX8MP_HDMI_PAI
 	tristate "Freescale i.MX8MP HDMI PAI bridge support"
 	depends on OF
diff --git a/drivers/gpu/drm/bridge/imx/Makefile b/drivers/gpu/drm/bridge/imx/Makefile
index 8d01fda25451..84499fe2e444 100644
--- a/drivers/gpu/drm/bridge/imx/Makefile
+++ b/drivers/gpu/drm/bridge/imx/Makefile
@@ -1,6 +1,8 @@
 obj-$(CONFIG_DRM_IMX_LDB_HELPER) += imx-ldb-helper.o
 obj-$(CONFIG_DRM_IMX_LEGACY_BRIDGE) += imx-legacy-bridge.o
 obj-$(CONFIG_DRM_IMX8MP_DW_HDMI_BRIDGE) += imx8mp-hdmi-tx.o
+obj-$(CONFIG_DRM_IMX8MP_DW_HDMI_BRIDGE_CONNECTOR_FIXUP) += imx8mp-hdmi-tx-connector-fixup.o \
+							   imx8mp-hdmi-tx-connector-fixup.dtbo.o
 obj-$(CONFIG_DRM_IMX8MP_HDMI_PAI) += imx8mp-hdmi-pai.o
 obj-$(CONFIG_DRM_IMX8MP_HDMI_PVI) += imx8mp-hdmi-pvi.o
 obj-$(CONFIG_DRM_IMX8QM_LDB) += imx8qm-ldb.o
diff --git a/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx-connector-fixup.c b/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx-connector-fixup.c
new file mode 100644
index 000000000000..8c423b9bfa50
--- /dev/null
+++ b/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx-connector-fixup.c
@@ -0,0 +1,60 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Add an hdmi-connector node to boards using the imx8mp hdmi_tx which
+ * don't have one. This is needed for the i.MX LCDIF to work with
+ * DRM_BRIDGE_ATTACH_NO_CONNECTOR.
+ *
+ * Copyright (C) 2026 GE HealthCare
+ * Author: Luca Ceresoli <luca.ceresoli@bootlin.com>
+ */
+
+#include <linux/kernel.h>
+#include <linux/of.h>
+#include <linux/of_fdt.h>
+#include <linux/slab.h>
+
+/* Embedded dtbo symbols created by cmd_wrap_S_dtb in scripts/Makefile.dtbs */
+extern char __dtbo_imx8mp_hdmi_tx_connector_fixup_begin[];
+extern char __dtbo_imx8mp_hdmi_tx_connector_fixup_end[];
+
+static int __init imx8mp_hdmi_tx_connector_fixup_init(void)
+{
+	struct device_node *hdmi_tx __free(device_node) = NULL;
+	struct device_node *endpoint __free(device_node) = NULL;
+	struct device_node *hdmi_conn __free(device_node) = NULL;
+	void *dtbo_start;
+	u32 dtbo_size;
+	int ovcs_id;
+	int err;
+
+	hdmi_tx = of_find_node_by_path("/soc@0/bus@32c00000/hdmi@32fd8000");
+	if (!of_device_is_available(hdmi_tx))
+		return 0;
+
+	/* If endpoint exists, assume an hdmi-connector exists already */
+	endpoint = of_find_node_by_path("/soc@0/bus@32c00000/hdmi@32fd8000/ports/port@1/endpoint");
+	if (endpoint)
+		return 0;
+
+	dtbo_start = __dtbo_imx8mp_hdmi_tx_connector_fixup_begin;
+	dtbo_size = __dtbo_imx8mp_hdmi_tx_connector_fixup_end -
+		    __dtbo_imx8mp_hdmi_tx_connector_fixup_begin;
+
+	err = of_overlay_fdt_apply(dtbo_start, dtbo_size, &ovcs_id, NULL);
+	if (err)
+		return err;
+
+	hdmi_conn = of_find_node_by_name(NULL, "fixup-hdmi-connector");
+	if (!hdmi_conn) {
+		err = -ENODEV;
+		goto overlay_remove;
+	}
+
+	return 0;
+
+overlay_remove:
+	of_overlay_remove(&ovcs_id);
+	return err;
+}
+
+subsys_initcall(imx8mp_hdmi_tx_connector_fixup_init);
diff --git a/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx-connector-fixup.dtso b/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx-connector-fixup.dtso
new file mode 100644
index 000000000000..ee718ca1b11b
--- /dev/null
+++ b/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx-connector-fixup.dtso
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * DTS overlay adding an hdmi-connector node to boards using the imx8mp hdmi_tx
+ *
+ * Copyright (C) 2026 GE HealthCare
+ * Author: Luca Ceresoli <luca.ceresoli@bootlin.com>
+ */
+
+/dts-v1/;
+/plugin/;
+
+&{/} {
+	fixup-hdmi-connector {
+		compatible = "hdmi-connector";
+		label = "HDMI";
+		type = "a";
+
+		port {
+			fixup_hdmi_connector_in: endpoint {
+				remote-endpoint = <&hdmi_tx_out>;
+			};
+		};
+	};
+
+	soc@0 {
+		bus@32c00000 {
+			hdmi@32fd8000 {
+				ports {
+					port@1 {
+						hdmi_tx_out: endpoint {
+							remote-endpoint = <&fixup_hdmi_connector_in>;
+						};
+					};
+				};
+			};
+		};
+	};
+};
diff --git a/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx.c b/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx.c
index 32fd3554e267..8e8cfd66f23b 100644
--- a/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx.c
+++ b/drivers/gpu/drm/bridge/imx/imx8mp-hdmi-tx.c
@@ -138,6 +138,7 @@ static int imx8mp_dw_hdmi_probe(struct platform_device *pdev)
 	plat_data->phy_name = "SAMSUNG HDMI TX PHY";
 	plat_data->priv_data = hdmi;
 	plat_data->phy_force_vendor = true;
+	plat_data->output_port = 1;
 
 	platform_set_drvdata(pdev, hdmi);
 

-- 
2.53.0

[PATCH 8/8] drm/mxsfb/lcdif: use DRM_BRIDGE_ATTACH_NO_CONNECTOR and the bridge-connector

[Luca Ceresoli]

Convert this driver to DRM_BRIDGE_ATTACH_NO_CONNECTOR and to the
drm_bridge_connector framework which is the current DRM bridge best
practice.

Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
---
 drivers/gpu/drm/mxsfb/Kconfig     |  1 +
 drivers/gpu/drm/mxsfb/lcdif_drv.c | 12 +++++++++++-
 2 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/mxsfb/Kconfig b/drivers/gpu/drm/mxsfb/Kconfig
index 264e74f45554..1a8a5d1ba8c0 100644
--- a/drivers/gpu/drm/mxsfb/Kconfig
+++ b/drivers/gpu/drm/mxsfb/Kconfig
@@ -33,6 +33,7 @@ config DRM_IMX_LCDIF
 	select DRM_GEM_DMA_HELPER
 	select DRM_PANEL
 	select DRM_PANEL_BRIDGE
+	select DRM_BRIDGE_CONNECTOR
 	help
 	  Choose this option if you have an LCDIFv3 LCD controller.
 	  Those devices are found in various i.MX SoC (i.MX8MP,
diff --git a/drivers/gpu/drm/mxsfb/lcdif_drv.c b/drivers/gpu/drm/mxsfb/lcdif_drv.c
index e40253ad607b..3388d959c93d 100644
--- a/drivers/gpu/drm/mxsfb/lcdif_drv.c
+++ b/drivers/gpu/drm/mxsfb/lcdif_drv.c
@@ -17,6 +17,7 @@
 #include <drm/clients/drm_client_setup.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_bridge.h>
+#include <drm/drm_bridge_connector.h>
 #include <drm/drm_drv.h>
 #include <drm/drm_encoder.h>
 #include <drm/drm_fbdev_dma.h>
@@ -50,6 +51,7 @@ static int lcdif_attach_bridge(struct lcdif_drm_private *lcdif)
 	struct device *dev = lcdif->drm->dev;
 	struct drm_encoder *encoder;
 	struct drm_bridge *bridge;
+	struct drm_connector *connector;
 	int ret;
 
 	struct device_node *remote __free(device_node) =
@@ -72,10 +74,18 @@ static int lcdif_attach_bridge(struct lcdif_drm_private *lcdif)
 	if (ret)
 		return dev_err_probe(dev, ret, "Failed to initialize encoder\n");
 
-	ret = drm_bridge_attach(encoder, bridge, NULL, 0);
+	ret = drm_bridge_attach(encoder, bridge, NULL, DRM_BRIDGE_ATTACH_NO_CONNECTOR);
 	if (ret)
 		return dev_err_probe(dev, ret, "Failed to attach bridge\n");
 
+	connector = drm_bridge_connector_init(lcdif->drm, encoder);
+	if (IS_ERR(connector))
+		return dev_err_probe(dev, PTR_ERR(connector), "Failed to init bridge_connector\n");
+
+	ret = drm_connector_attach_encoder(connector, encoder);
+	if (ret)
+		return dev_err_probe(dev, ret, "Failed to attach connector\n");
+
 	return 0;
 }
 

-- 
2.53.0

Re: [PATCH 1/8] drm/mxsfb/lcdif: simplify remote pointer management using __free

[Luca Ceresoli]

Hello,

On Fri Mar 20, 2026 at 11:46 AM CET, Luca Ceresoli wrote:
> Putting the remote device_node reference requires a of_node_put(ep) in both
> error return points. Use a cleanup action to simplify the code.
>
> Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
> ---
>  drivers/gpu/drm/mxsfb/lcdif_drv.c | 9 +++------
>  1 file changed, 3 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/mxsfb/lcdif_drv.c b/drivers/gpu/drm/mxsfb/lcdif_drv.c
> index 47da1d9336b9..756ca96373c8 100644
> --- a/drivers/gpu/drm/mxsfb/lcdif_drv.c
> +++ b/drivers/gpu/drm/mxsfb/lcdif_drv.c
> @@ -53,16 +53,13 @@ static int lcdif_attach_bridge(struct lcdif_drm_private *lcdif)
>  	int ret;
>
>  	for_each_endpoint_of_node(dev->of_node, ep) {
> -		struct device_node *remote;
> +		struct device_node *remote __free(drm_bridge_put) =
                                                  ^^^^^^^^^^^^^^

I just realized there's a mistake here, this should be
__free(device_node). However this does not prevent testing the series in
its entirety because patch 2 fixes this mistake.

Will be fixed in v2.

Sorry about the noise, did too many rebases today!

Luca

--
Luca Ceresoli, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Claude review: drm/mxsfb/lcdif: use DRM_BRIDGE_ATTACH_NO_CONNECTOR and the bridge-connector

[Claude Code Review Bot]

Overall Series Review

Subject: drm/mxsfb/lcdif: use DRM_BRIDGE_ATTACH_NO_CONNECTOR and the bridge-connector
Author: Luca Ceresoli <luca.ceresoli@bootlin.com>
Patches: 10
Reviewed: 2026-03-22T03:48:36.879364

---

This is a well-structured and well-motivated series that modernizes the i.MX8MP LCDIF driver to use `DRM_BRIDGE_ATTACH_NO_CONNECTOR` with `drm_bridge_connector`, which is the current DRM best practice. The cover letter is excellent and clearly explains the dependency chain between patches. The backward compatibility approach using a DT overlay (patch 7) is a reasonable reuse of the pattern from `drm/tilcdc`.

There is one clear bug in patch 1 (wrong cleanup type), and a few minor issues in other patches. The core changes in patches 6 and 8 look correct and well-reasoned.

---
Generated by Claude Code Patch Reviewer

Claude review: drm/mxsfb/lcdif: simplify remote pointer management using __free

[Claude Code Review Bot]

Patch Review

**Bug: Wrong cleanup specifier.** The patch uses `__free(drm_bridge_put)` on a `struct device_node *`:

```c
struct device_node *remote __free(drm_bridge_put) =
    of_graph_get_remote_port_parent(ep);
```

`drm_bridge_put` is defined for `struct drm_bridge *` (see `include/drm/drm_bridge.h:1295`), not `struct device_node *`. This should be `__free(device_node)` which calls `of_node_put()` — matching what the original code did manually. As-is, this would either fail to compile (type mismatch) or invoke undefined behavior.

The simplification logic itself is correct — the two manual `of_node_put(remote)` calls are properly replaced by the auto-cleanup.

---
Generated by Claude Code Patch Reviewer

Claude review: drm/mxsfb/lcdif: don't unnecessarily loop over ports

[Claude Code Review Bot]

Patch Review

This is a good simplification. Using `of_graph_get_remote_node(dev->of_node, 0, -1)` to directly get port 0's remote is cleaner than iterating endpoints.

**Minor nit:** The commit message has typos: "thos" should be "those", and "multipltle" should be "multiple".

One observation: the original code used `devm_drm_of_get_bridge(dev, dev->of_node, 0, of_ep.id)` where `of_ep.id` was the endpoint id. The new code uses `devm_drm_of_get_bridge(dev, dev->of_node, 0, -1)`. Since there's only one port with one endpoint, this should be equivalent, but worth confirming the `-1` semantics for `devm_drm_of_get_bridge` are "any endpoint" (which they are — it passes through to `of_graph_get_remote_node`).

The `__free(device_node)` usage here is correct — good, this was fixed from patch 1's bug.

---
Generated by Claude Code Patch Reviewer

Claude review: drm/mxsfb/lcdif: use dev_err_probe() consistently in lcdif_attach_bridge

[Claude Code Review Bot]

Patch Review

Straightforward cleanup. Using `dev_err_probe()` for all error paths is the right approach. One minor observation:

```c
if (!encoder)
    return dev_err_probe(dev, -ENOMEM, "Failed to allocate encoder\n");
```

Using `dev_err_probe()` for `-ENOMEM` is slightly unusual since `-ENOMEM` is never `-EPROBE_DEFER`, but it's acceptable for code consistency and still prints the error message correctly.

---
Generated by Claude Code Patch Reviewer

Claude review: drm/bridge: dw-hdmi: document the output_port field

[Claude Code Review Bot]

Patch Review

Fine. The comment is accurate and helpful:

```c
/*
 * The HDMI output port number (which must be 1) if it is described
 * in the device tree. 0 if the device tree does not describe the
 * next component (legacy mode).
 */
```

---
Generated by Claude Code Patch Reviewer

Claude review: drm/bridge: dw-hdmi: warn on unsupported attach combination

[Claude Code Review Bot]

Patch Review

The WARN_ON + early return is appropriate:

```c
if (WARN_ON((flags & DRM_BRIDGE_ATTACH_NO_CONNECTOR) && !hdmi->plat_data->output_port))
    return -EINVAL;
```

The commit message has a typo: "teh" should be "the".

The note about functional behavior is correct — in the current code, if `DRM_BRIDGE_ATTACH_NO_CONNECTOR` is set and `output_port` is 0, `next_bridge` would be NULL, so `drm_bridge_attach()` would fail with `-EINVAL` anyway.

---
Generated by Claude Code Patch Reviewer

Claude review: drm/bridge: dw-hdmi: move next_bridge lookup to attach time

[Claude Code Review Bot]

Patch Review

This is the key enabling patch. Moving the lookup from probe to attach breaks the devlink chicken-and-egg problem described in the commit message.

The implementation looks correct. The new code in `dw_hdmi_bridge_attach()`:

```c
struct device_node *remote __free(device_node) =
    of_graph_get_remote_node(hdmi->dev->of_node,
                 hdmi->plat_data->output_port, -1);
if (!remote)
    return -EPROBE_DEFER;

struct drm_bridge *next_bridge __free(drm_bridge_put) =
    of_drm_find_and_get_bridge(remote);

return drm_bridge_attach(encoder, next_bridge, bridge, flags);
```

**Potential issue:** If `of_drm_find_and_get_bridge()` returns NULL (bridge not yet registered), `next_bridge` will be NULL and `drm_bridge_attach()` will be called with a NULL bridge, returning `-EINVAL`. The old `dw_hdmi_parse_dt()` returned `-EPROBE_DEFER` in this case. Consider adding a NULL check on `next_bridge` and returning `-EPROBE_DEFER` to maintain the probe deferral behavior. Without this, the attach would fail hard instead of deferring.

**Semantic change:** The old code returned `-ENODEV` when the remote node was missing in DT; the new code returns `-EPROBE_DEFER`. Since the WARN_ON in patch 5 guards the `!output_port` case, and if `output_port` is set but there's no remote in DT, `-EPROBE_DEFER` may not be the right error — the node might simply not exist. However, with the overlay fixup in patch 7, this becomes a timing issue, so `-EPROBE_DEFER` is arguably more appropriate.

---
Generated by Claude Code Patch Reviewer

Claude review: drm/bridge: imx8mp-hdmi-tx: add an hdmi-connector when missing using a DT overlay at boot time

[Claude Code Review Bot]

Patch Review

This is the most complex and controversial patch. The approach is reasonable and has precedent (`drm/tilcdc`).

**Observations on the C code (`imx8mp-hdmi-tx-connector-fixup.c`):**

1. The hardcoded path `/soc@0/bus@32c00000/hdmi@32fd8000` is fragile — it only works for i.MX8MP and assumes a specific DT structure. However, since the Kconfig already ties this to `DRM_IMX8MP_DW_HDMI_BRIDGE`, this is acceptable.

2. The `hdmi_conn` lookup after overlay application seems unnecessary:
   ```c
   hdmi_conn = of_find_node_by_name(NULL, "fixup-hdmi-connector");
   if (!hdmi_conn) {
       err = -ENODEV;
       goto overlay_remove;
   }
   ```
   If the overlay applied successfully, the node should always exist. This looks like a sanity check, but the `hdmi_conn` variable is declared with `__free(device_node)` and then never used beyond the NULL check. This is fine for validation but could use a comment explaining it's a sanity check.

3. The overlay is never removed on module exit / cleanup (no `module_exit`). Since this is `bool` (not tristate) in Kconfig, there's no module unload path, which is fine. But this means the overlay ID `ovcs_id` is a dead store if the function succeeds — not a real issue but worth noting.

4. Using `subsys_initcall` is the right level to ensure this runs before device probing.

**On the DT overlay (`imx8mp-hdmi-tx-connector-fixup.dtso`):**
The overlay looks correct. It adds the `fixup-hdmi-connector` node at the root and links it bidirectionally to `port@1` of the HDMI TX.

**One-line change in `imx8mp-hdmi-tx.c`:**
```c
plat_data->output_port = 1;
```
This enables the "case B" mode in dw-hdmi. This is correct and is the key change that makes the HDMI pipeline work with `DRM_BRIDGE_ATTACH_NO_CONNECTOR`. However, this change also affects the non-`DRM_BRIDGE_ATTACH_NO_CONNECTOR` path — in the current tree, dw-hdmi will now call `dw_hdmi_parse_dt()` at probe time, which will look for a remote on port@1. For boards that already have an hdmi-connector in DT this is fine, but for boards that don't (and don't have the fixup), `dw_hdmi_parse_dt()` would return `-ENODEV` and **probe would fail**. This is only safe because patch 6 removes `dw_hdmi_parse_dt()` entirely — so patches 6 and 7 must be applied together (bisectability concern if the `output_port = 1` line were in a separate patch from patch 6, but since patch 6 comes first, bisecting should be fine).

---
Generated by Claude Code Patch Reviewer

Claude review: drm/mxsfb/lcdif: use DRM_BRIDGE_ATTACH_NO_CONNECTOR and the bridge-connector

[Claude Code Review Bot]

Patch Review

This is the goal patch. The change is clean and follows the standard pattern:

```c
ret = drm_bridge_attach(encoder, bridge, NULL, DRM_BRIDGE_ATTACH_NO_CONNECTOR);
...
connector = drm_bridge_connector_init(lcdif->drm, encoder);
...
ret = drm_connector_attach_encoder(connector, encoder);
```

The Kconfig `select DRM_BRIDGE_CONNECTOR` is correct.

This looks good. All the preparatory work in patches 1-7 makes this final patch straightforward.

---
Generated by Claude Code Patch Reviewer