Skip to contentSkip to conversation

usb: typec: ucsi: Fix command cancellation

PATCH In review Heikki Krogerus 3 participants 5 messages Jun 6, 2023 – Jun 13, 2023

Subsystems (4)
lore
Share & tools
lore

Heikki Krogerus

author maintainer
#0 lore raw

PATCH 1/1 usb: typec: ucsi: Fix command cancellation

The Cancel command was passed to the write callback as the offset instead of as the actual command which caused NULL pointer dereference.

Fixes 094902bc6a3c ("usb: typec: ucsi: Always cancel the command if PPM reports BUSY condition")
---
 drivers/usb/typec/ucsi/ucsi.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
drivers/usb/typec/ucsi/ucsi.c +7 -4
index 2b472ec01dc42..b664ecbb798be 100644
@@ -132,10 +132,8 @@ static int ucsi_exec_command(struct ucsi *ucsi, u64 cmd)
 	if (ret)
 		return ret;
 
-	if (cci & UCSI_CCI_BUSY) {
-		ucsi->ops->async_write(ucsi, UCSI_CANCEL, NULL, 0);
-		return -EBUSY;
-	}
+	if (cmd != UCSI_CANCEL && cci & UCSI_CCI_BUSY)
+		return ucsi_exec_command(ucsi, UCSI_CANCEL);
 
 	if (!(cci & UCSI_CCI_COMMAND_COMPLETE))
 		return -EIO;
@@ -149,6 +147,11 @@ static int ucsi_exec_command(struct ucsi *ucsi, u64 cmd)
 		return ucsi_read_error(ucsi);
 	}
 
+	if (cmd == UCSI_CANCEL && cci & UCSI_CCI_CANCEL_COMPLETE) {
+		ret = ucsi_acknowledge_command(ucsi);
+		return ret ? ret : -EBUSY;
+	}
+
 	return UCSI_CCI_LENGTH(cci);
 }
 
Signature · 1 line
2.39.2
#1 lore raw
Quoted 9 lines · Heikki Krogerus

The Cancel command was passed to the write callback as the offset instead of as the actual command which caused NULL pointer dereference.

Fixes 094902bc6a3c ("usb: typec: ucsi: Always cancel the command if PPM reports BUSY condition")

Gentle reminder that this made no progress for a week now. Or was there and I just missed it? Then apologies in advance.

I'm asking, as it afaics would be nice to have this (or some other fix for the regression linked above) mainlined before the next -rc. That would be ideal, as then it can get at least one week of testing before the final is released.

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) --
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.

Quoted 32 lines
 drivers/usb/typec/ucsi/ucsi.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
drivers/usb/typec/ucsi/ucsi.c +7 -4
index 2b472ec01dc42..b664ecbb798be 100644
@@ -132,10 +132,8 @@ static int ucsi_exec_command(struct ucsi *ucsi, u64 cmd)
 	if (ret)
 		return ret;
 
-	if (cci & UCSI_CCI_BUSY) {
-		ucsi->ops->async_write(ucsi, UCSI_CANCEL, NULL, 0);
-		return -EBUSY;
-	}
+	if (cmd != UCSI_CANCEL && cci & UCSI_CCI_BUSY)
+		return ucsi_exec_command(ucsi, UCSI_CANCEL);
 
 	if (!(cci & UCSI_CCI_COMMAND_COMPLETE))
 		return -EIO;
@@ -149,6 +147,11 @@ static int ucsi_exec_command(struct ucsi *ucsi, u64 cmd)
 		return ucsi_read_error(ucsi);
 	}
 
+	if (cmd == UCSI_CANCEL && cci & UCSI_CCI_CANCEL_COMPLETE) {
+		ret = ucsi_acknowledge_command(ucsi);
+		return ret ? ret : -EBUSY;
+	}
+
 	return UCSI_CCI_LENGTH(cci);
 }
 
#2 lore raw
Quoted 5 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 4 lines · Heikki Krogerus

The Cancel command was passed to the write callback as the offset instead of as the actual command which caused NULL pointer dereference.

Quoted 8 lines
Quoted 5 lines
Fixes 094902bc6a3c ("usb: typec: ucsi: Always cancel the command if PPM reports BUSY condition")

Gentle reminder that this made no progress for a week now. Or was there and I just missed it? Then apologies in advance.

This just landed in my usb-linus branch a few hours before you sent this, and will show up in linux-next tomorrow as:
c4a8bfabefed ("usb: typec: ucsi: Fix command cancellation")

Quoted 4 lines

I'm asking, as it afaics would be nice to have this (or some other fix for the regression linked above) mainlined before the next -rc. That would be ideal, as then it can get at least one week of testing before the final is released.

It will get there, sorry for the delay, now caught up on all pending USB and TTY/serial fixes.

greg k-h

#3 lore raw
Quoted 12 lines · Greg Kroah-Hartman
Quoted 11 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 9 lines · Heikki Krogerus

The Cancel command was passed to the write callback as the offset instead of as the actual command which caused NULL pointer dereference.

Fixes 094902bc6a3c ("usb: typec: ucsi: Always cancel the command if PPM reports BUSY condition")
Quoted 6 lines
Quoted 2 lines

Gentle reminder that this made no progress for a week now. Or was there and I just missed it? Then apologies in advance.

This just landed in my usb-linus branch a few hours before you sent this, and will show up in linux-next tomorrow as:
c4a8bfabefed ("usb: typec: ucsi: Fix command cancellation")

Ahh, great! Sorry, I check next in cases like this before sending mails, but not the subsystem trees directly. :-/

Quoted 7 lines
Quoted 4 lines

I'm asking, as it afaics would be nice to have this (or some other fix for the regression linked above) mainlined before the next -rc. That would be ideal, as then it can get at least one week of testing before the final is released.

It will get there, sorry for the delay, now caught up on all pending USB and TTY/serial fixes.

No worries and thx for the update. It just looked like something where a quick "what's up" seemed appropriate. Thx again.

Ciao, Thorsten

#4 lore raw
Quoted 16 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 15 lines · Greg Kroah-Hartman
Quoted 13 lines · Linux regression tracking (Thorsten Leemhuis)
Quoted 9 lines · Heikki Krogerus

The Cancel command was passed to the write callback as the offset instead of as the actual command which caused NULL pointer dereference.

Fixes 094902bc6a3c ("usb: typec: ucsi: Always cancel the command if PPM reports BUSY condition")

Gentle reminder that this made no progress for a week now. Or was there and I just missed it? Then apologies in advance.

Quoted 6 lines
Quoted 3 lines

This just landed in my usb-linus branch a few hours before you sent this, and will show up in linux-next tomorrow as:
c4a8bfabefed ("usb: typec: ucsi: Fix command cancellation")

Ahh, great! Sorry, I check next in cases like this before sending mails, but not the subsystem trees directly. :-/

I wouldn't expect you to look in subsystem trees, not a problem at all, thanks for the ping, you're doing great work here.

greg k-h