Hi Heikki,
On 10/23/2025 8:03 PM, Heikki Krogerus wrote:
diff --git a/drivers/gpu/drm/bridge/Makefile b/drivers/gpu/drm/bridge/Makefile
index 245e8a27e3fc..e91736829167 100644
--- a/drivers/gpu/drm/bridge/Makefile
+++ b/drivers/gpu/drm/bridge/Makefile
@@ -1,6 +1,7 @@
# SPDX-License-Identifier: GPL-2.0
obj-$(CONFIG_DRM_AUX_BRIDGE) += aux-bridge.o
obj-$(CONFIG_DRM_AUX_HPD_BRIDGE) += aux-hpd-bridge.o
+obj-$(CONFIG_DRM_AUX_TYPEC_DP_HPD_BRIDGE) += aux-hpd-typec-dp-bridge.o
Instead, why not just make that a part of aux-hpd-bridge
conditionally:
ifneq ($(CONFIG_TYPEC),)
aux-hpd-bridge-y += aux-hpd-typec-dp-bridge.o
endif
Oh, I did consider that! But I noticed that aux-hpd-bridge.c contains the
following statement module_auxiliary_driver(drm_aux_hpd_bridge_drv), which
already includes a module_init. In the newly added file, in order to call the
register function, another module_init was also added. If the two files are
each made into a module separately, would there be a problem?
You would not call module_init() from the new file. Instead you would
call drm_aux_hpd_typec_dp_bridge_init() and what ever directly from
aux-hpd-bridge.c:
diff --git a/drivers/gpu/drm/bridge/aux-bridge.h
b/drivers/gpu/drm/bridge/aux-bridge.h
new file mode 100644
index 000000000000..ae689a7778fa
--- /dev/null
+++ b/drivers/gpu/drm/bridge/aux-hpd-bridge.h
@@ -0,0 +1,13 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef AUX_HPD_BRIDGE_H
+#define AUX_HPD_BRIDGE_H
+
+#if IS_ENABLED(CONFIG_TYPEC)
+int drm_aux_hpd_typec_dp_bridge_init(void);
+void drm_aux_hpd_typec_dp_bridge_exit(void);
+#else
+static inline int drm_aux_hpd_typec_dp_bridge_init(void) { return 0; }
+static inline void drm_aux_hpd_typec_dp_bridge_exit(void) { }
+#endif /* IS_ENABLED(CONFIG_TYPEC) */
+
+#endif /* AUX_HPD_BRIDGE_H */
diff --git a/drivers/gpu/drm/bridge/aux-hpd-bridge.c
b/drivers/gpu/drm/bridge/aux-hpd-bridge.c
index 2e9c702c7087..3578df1df78a 100644
--- a/drivers/gpu/drm/bridge/aux-hpd-bridge.c
+++ b/drivers/gpu/drm/bridge/aux-hpd-bridge.c
@@ -12,6 +12,8 @@
#include <drm/drm_bridge.h>
#include <drm/bridge/aux-bridge.h>
+#include "aux-hpd-bridge.h"
+
static DEFINE_IDA(drm_aux_hpd_bridge_ida);
struct drm_aux_hpd_bridge_data {
@@ -190,9 +192,16 @@ static int drm_aux_hpd_bridge_probe(struct
auxiliary_device *auxdev,
auxiliary_set_drvdata(auxdev, data);
+ drm_aux_hpd_typec_dp_bridge_init();
+
return devm_drm_bridge_add(data->dev, &data->bridge);
}
+static void drm_aux_hpd_bridge_remove(struct auxiliary_device *auxdev)
+{
+ drm_aux_hpd_typec_dp_bridge_exit();
+}
+
static const struct auxiliary_device_id drm_aux_hpd_bridge_table[] = {
{ .name = KBUILD_MODNAME ".dp_hpd_bridge", .driver_data =
DRM_MODE_CONNECTOR_DisplayPort, },
{},
@@ -203,6 +212,7 @@ static struct auxiliary_driver drm_aux_hpd_bridge_drv = {
.name = "aux_hpd_bridge",
.id_table = drm_aux_hpd_bridge_table,
.probe = drm_aux_hpd_bridge_probe,
+ .remove = drm_aux_hpd_bridge_remove,
};
module_auxiliary_driver(drm_aux_hpd_bridge_drv);
Yes, if we don't distinguish them through Kconfig, we need to use the
IS_ENABLED macro in the code. Thanks again for you code.
Another thing is that CONFIG_DRM_AUX_HPD_BRIDGE originally needed to be
selected by other modules. With this change, we also need to expose it in
Kconfig.
--
Best,
Chaoyi
