Skip to contentSkip to conversation

usb: typec: fusb302: Free log buffers on exit

PATCH In review Yuho Choi 3 participants 3 messages Aug 7 – Aug 26

Subsystems (2)
lore
Share & tools
lore

Yuho Choi

author
#0 lore raw

PATCH 1/1 usb: typec: fusb302: Free log buffers on exit

fusb302_log() lazily allocates entries in chip->logbuffer[], but fusb302_debugfs_exit() only removes the debugfs directory. The buffers are not part of the devm-managed chip allocation, so they leak when the driver is removed or probe fails after logging.

Free all log buffer entries during debugfs teardown.

Fixes c034a43e72dd ("staging: typec: Fairchild FUSB302 Type-c chip driver")
---
 drivers/usb/typec/tcpm/fusb302.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)
drivers/usb/typec/tcpm/fusb302.c +10 -1
index 3319f6a2b0c9..67ccbbd64caf 100644
@@ -223,7 +223,16 @@ static void fusb302_debugfs_init(struct fusb302_chip *chip)
 
 static void fusb302_debugfs_exit(struct fusb302_chip *chip)
 {
+	int i;
+
 	debugfs_remove(chip->dentry);
+
+	mutex_lock(&chip->logbuffer_lock);
+	for (i = 0; i < LOG_BUFFER_ENTRIES; i++) {
+		kfree(chip->logbuffer[i]);
+		chip->logbuffer[i] = NULL;
+	}
+	mutex_unlock(&chip->logbuffer_lock);
 }
 
 #else
@@ -1784,8 +1793,8 @@ static int fusb302_probe(struct i2c_client *client)
 fwnode_put:
 	fwnode_handle_put(chip->tcpc_dev.fwnode);
 destroy_workqueue:
-	fusb302_debugfs_exit(chip);
 	destroy_workqueue(chip->wq);
+	fusb302_debugfs_exit(chip);
 
 	return ret;
 }
Signature · 1 line
2.43.0
maintainer
#1 lore raw
Quoted 21 lines · Yuho Choi

fusb302_log() lazily allocates entries in chip->logbuffer[], but fusb302_debugfs_exit() only removes the debugfs directory. The buffers are not part of the devm-managed chip allocation, so they leak when the driver is removed or probe fails after logging.

Free all log buffer entries during debugfs teardown.

Fixes c034a43e72dd ("staging: typec: Fairchild FUSB302 Type-c chip driver")
---
 drivers/usb/typec/tcpm/fusb302.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)
drivers/usb/typec/tcpm/fusb302.c +0 -0
index 3319f6a2b0c9..67ccbbd64caf 100644
@@ -223,7 +223,16 @@ static void fusb302_debugfs_init(struct fusb302_chip *chip)
 
 static void fusb302_debugfs_exit(struct fusb302_chip *chip)
 {

drivers/usb/typec/tcpm/fusb302.c static void fusb302_debugfs_init(struct fusb302_chip *chip) · line 235

Quoted 10 lines
drivers/usb/typec/tcpm/fusb302.c +9 -0
+	int i;
+
 	debugfs_remove(chip->dentry);
+
+	mutex_lock(&chip->logbuffer_lock);
+	for (i = 0; i < LOG_BUFFER_ENTRIES; i++) {
+		kfree(chip->logbuffer[i]);
+		chip->logbuffer[i] = NULL;
+	}
+	mutex_unlock(&chip->logbuffer_lock);

As you are tearing things down here, and there is no actual user, why is the lock needed? And if so, can you just use a guard() instead?

thanks,

greg k-h

#2 lore raw
Quoted 4 lines · Greg KH
Quoted 4 lines
drivers/usb/typec/tcpm/fusb302.c +0 -0
@@ -223,7 +223,16 @@ static void fusb302_debugfs_init(struct fusb302_chip *chip)
 
Quoted 15 lines

drivers/usb/typec/tcpm/fusb302.c static void fusb302_debugfs_init(struct fusb302_chip *chip) · line 235

Quoted 12 lines
drivers/usb/typec/tcpm/fusb302.c +9 -0
 static void fusb302_debugfs_exit(struct fusb302_chip *chip)
 {
+	int i;
+
 	debugfs_remove(chip->dentry);
+
+	mutex_lock(&chip->logbuffer_lock);
+	for (i = 0; i < LOG_BUFFER_ENTRIES; i++) {
+		kfree(chip->logbuffer[i]);
+		chip->logbuffer[i] = NULL;
+	}
+	mutex_unlock(&chip->logbuffer_lock);

As you are tearing things down here, and there is no actual user, why is the lock needed? And if so, can you just use a guard() instead?

That is correct. The whole thing is about vanish so locking is not needed.
Looking at the actual user of that buffer, I'm curious if it wouldn't be better to use dev_err()/ dev_info() for some of the output and other which are just pure informative/ debug kind of information, hide behind a trace event which can be enabled if needed.

Quoted 3 lines

thanks,

greg k-h

Sebastian