Skip to contentSkip to conversation

usb: typec: class: Don't use port parent for getting mux handles

PATCH In review Heikki Krogerus 2 participants 2 messages Mar 28, 2019 – Apr 1, 2019

Subsystems (3)
lore
Share & tools
lore

Heikki Krogerus

author maintainer
#0 lore raw

PATCH 1/1 usb: typec: class: Don't use port parent for getting mux handles

commit 23481121c81d984193edf1532f5e123637e50903 upstream.

It is not possible to use the parent of the port device when requesting mux handles as the parent may be a multiport USB Type-C or PD controller. The muxes must be assigned to the ports, not the controllers.

This will also move the requesting of the muxes after the port device is initialized.

[ heikki: This patch fixes an issue where port->id was used uninitialized. ]

Cc <[email protected]> # 4.19.x: cd7753d: drivers: base: Helpers for adding device connection descriptions
Cc <[email protected]> # 4.19.x: 140a4ec: platform: x86: intel_cht_int33fe: Register all connections at once
Cc <[email protected]> # 4.19.x: 78d2b54: platform: x86: intel_cht_int33fe: Add connection for the DP alt mode
Cc <[email protected]> # 4.19.x: 495965a: platform: x86: intel_cht_int33fe: Add connections for the USB Type-C port
Cc <[email protected]> # 4.19.x
Cc <[email protected]> # 4.19.x: 148b0aa: platform: x86: intel_cht_int33fe: Remove the old connections for the muxes
Fixes bdecb33af34f ("usb: typec: API for controlling USB Type-C Multiplexers")

---
Hi,

This is a request to backport commit 23481121c81d984193edf1532f5e123637e50903

It was not marked as a fix, but it does fix an issue where a variable "port->id" was used in function typec_register_port() without initializing it.

The patch has additional prerequisites which are listed in the sign-off area. Please note that there is also one patch listed there after this one.

thanks,

heikki

---
 drivers/usb/typec/class.c | 38 +++++++++++++++-----------------------
 1 file changed, 15 insertions(+), 23 deletions(-)
drivers/usb/typec/class.c +15 -23
index e61dffb27a0c..00141e05bc72 100644
@@ -1500,7 +1500,7 @@ typec_port_register_altmode(struct typec_port *port,
 
 	sprintf(id, "id%04xm%02x", desc->svid, desc->mode);
 
-	mux = typec_mux_get(port->dev.parent, id);
+	mux = typec_mux_get(&port->dev, id);
 	if (IS_ERR(mux))
 		return ERR_CAST(mux);
 
@@ -1540,18 +1540,6 @@ struct typec_port *typec_register_port(struct device *parent,
 		return ERR_PTR(id);
 	}
 
-	port->sw = typec_switch_get(cap->fwnode ? &port->dev : parent);
-	if (IS_ERR(port->sw)) {
-		ret = PTR_ERR(port->sw);
-		goto err_switch;
-	}
-
-	port->mux = typec_mux_get(parent, "typec-mux");
-	if (IS_ERR(port->mux)) {
-		ret = PTR_ERR(port->mux);
-		goto err_mux;
-	}
-
 	switch (cap->type) {
 	case TYPEC_PORT_SRC:
 		port->pwr_role = TYPEC_SOURCE;
@@ -1592,13 +1580,26 @@ struct typec_port *typec_register_port(struct device *parent,
 	port->port_type = cap->type;
 	port->prefer_role = cap->prefer_role;
 
+	device_initialize(&port->dev);
 	port->dev.class = typec_class;
 	port->dev.parent = parent;
 	port->dev.fwnode = cap->fwnode;
 	port->dev.type = &typec_port_dev_type;
 	dev_set_name(&port->dev, "port%d", id);
 
-	ret = device_register(&port->dev);
+	port->sw = typec_switch_get(&port->dev);
+	if (IS_ERR(port->sw)) {
+		put_device(&port->dev);
+		return ERR_CAST(port->sw);
+	}
+
+	port->mux = typec_mux_get(&port->dev, "typec-mux");
+	if (IS_ERR(port->mux)) {
+		put_device(&port->dev);
+		return ERR_CAST(port->mux);
+	}
+
+	ret = device_add(&port->dev);
 	if (ret) {
 		dev_err(parent, "failed to register port (%d)\n", ret);
 		put_device(&port->dev);
@@ -1606,15 +1607,6 @@ struct typec_port *typec_register_port(struct device *parent,
 	}
 
 	return port;
-
-err_mux:
-	typec_switch_put(port->sw);
-
-err_switch:
-	ida_simple_remove(&typec_index_ida, port->id);
-	kfree(port);
-
-	return ERR_PTR(ret);
 }
 EXPORT_SYMBOL_GPL(typec_register_port);
 
Signature · 1 line
2.20.1
maintainer
#1 lore raw
Quoted 27 lines · Heikki Krogerus

commit 23481121c81d984193edf1532f5e123637e50903 upstream.

It is not possible to use the parent of the port device when requesting mux handles as the parent may be a multiport USB Type-C or PD controller. The muxes must be assigned to the ports, not the controllers.

This will also move the requesting of the muxes after the port device is initialized.

[ heikki: This patch fixes an issue where port->id was used uninitialized. ]

Cc <[email protected]> # 4.19.x: cd7753d: drivers: base: Helpers for adding device connection descriptions
Cc <[email protected]> # 4.19.x: 140a4ec: platform: x86: intel_cht_int33fe: Register all connections at once
Cc <[email protected]> # 4.19.x: 78d2b54: platform: x86: intel_cht_int33fe: Add connection for the DP alt mode
Cc <[email protected]> # 4.19.x: 495965a: platform: x86: intel_cht_int33fe: Add connections for the USB Type-C port
Cc <[email protected]> # 4.19.x
Cc <[email protected]> # 4.19.x: 148b0aa: platform: x86: intel_cht_int33fe: Remove the old connections for the muxes
Fixes bdecb33af34f ("usb: typec: API for controlling USB Type-C Multiplexers")

---
Hi,

This is a request to backport commit 23481121c81d984193edf1532f5e123637e50903

Quoted 7 lines

It was not marked as a fix, but it does fix an issue where a variable "port->id" was used in function typec_register_port() without initializing it.

The patch has additional prerequisites which are listed in the sign-off area. Please note that there is also one patch listed there after this one.

Ok, all are now queued up. Please let me know if I missed anything.

thanks,

greg k-h