Skip to contentSkip to conversation

usb: typec: ucsi: possible deadlock in ucsi_pr_swap() and ucsi_handle_connector_change()

BUG Jia-Ju Bai 3 participants 6 messages Feb 9, 2022 – Mar 28, 2022

Subsystems (2)
lore
Share & tools
lore

Jia-Ju Bai

author
#0 lore raw

Hello,

My static analysis tool reports a possible deadlock in the ucsi driver in Linux 5.16:

ucsi_pr_swap()
  mutex_lock(&con->lock); --> Line 962 (Lock A) wait_for_completion_timeout(&con->complete, ...) --> Line 981 (Wait X)

ucsi_handle_connector_change()
  mutex_lock(&con->lock); --> Line 763 (Lock A)
  complete(&con->complete); --> Line 782 (Wake X)
  complete(&con->complete); --> Line 807 (Wake X)

When ucsi_pr_swap() is executed, "Wait X" is performed by holding "Lock A". If ucsi_handle_connector_change() is executed at this time, "Wake X" cannot be performed to wake up "Wait X" in ucsi_handle_connector_change(), because "Lock A" has been already held by ucsi_handle_connector_change(), causing a possible deadlock. I find that "Wait X" is performed with a timeout, to relieve the possible deadlock; but I think this timeout can cause inefficient execution.

I am not quite sure whether this possible problem is real. Any feedback would be appreciated, thanks :)

Best wishes,
Jia-Ju Bai

maintainer
#1 lore raw
Quoted 22 lines · Jia-Ju Bai

Hello,

My static analysis tool reports a possible deadlock in the ucsi driver in Linux 5.16:

ucsi_pr_swap()
  mutex_lock(&con->lock); --> Line 962 (Lock A) wait_for_completion_timeout(&con->complete, ...) --> Line 981 (Wait X)

ucsi_handle_connector_change()
  mutex_lock(&con->lock); --> Line 763 (Lock A)
  complete(&con->complete); --> Line 782 (Wake X)
  complete(&con->complete); --> Line 807 (Wake X)

When ucsi_pr_swap() is executed, "Wait X" is performed by holding "Lock A". If ucsi_handle_connector_change() is executed at this time, "Wake X" cannot be performed to wake up "Wait X" in ucsi_handle_connector_change(), because "Lock A" has been already held by ucsi_handle_connector_change(), causing a possible deadlock.
I find that "Wait X" is performed with a timeout, to relieve the possible deadlock; but I think this timeout can cause inefficient execution.

Quoted 2 lines

I am not quite sure whether this possible problem is real. Any feedback would be appreciated, thanks :)

This is probable a regression from commit ad74b8649bea ("usb: typec:
ucsi: Preliminary support for alternate modes"). Can you test does this patch fix the issue (attached)?

thanks,

Signature · 1 line
heikki

From 2ad06425a3df7be656f8a5b3c202aab45554fd17 Mon Sep 17 00:00:00 2001
From: Heikki Krogerus <[email protected]>
Date: Wed, 9 Feb 2022 17:27:19 +0300
Subject: [PATCH] usb: typec: ucsi: Test fix

Interim.

---
 drivers/usb/typec/ucsi/ucsi.c | 19 +++++++++++++------
 1 file changed, 13 insertions(+), 6 deletions(-)
drivers/usb/typec/ucsi/ucsi.c +13 -6
index f0c2fa19f3e0f..225104beda8be 100644
@@ -956,14 +956,18 @@ static int ucsi_dr_swap(struct typec_port *port, enum typec_data_role role)
 	if (ret < 0)
 		goto out_unlock;
 
+	mutex_unlock(&con->lock);
+
 	if (!wait_for_completion_timeout(&con->complete,
 					msecs_to_jiffies(UCSI_SWAP_TIMEOUT_MS)))
-		ret = -ETIMEDOUT;
+		return -ETIMEDOUT;
+
+	return 0;
 
 out_unlock:
 	mutex_unlock(&con->lock);
 
-	return ret < 0 ? ret : 0;
+	return ret;
 }
 
 static int ucsi_pr_swap(struct typec_port *port, enum typec_role role)
@@ -992,11 +996,13 @@ static int ucsi_pr_swap(struct typec_port *port, enum typec_role role)
 	if (ret < 0)
 		goto out_unlock;
 
+	mutex_unlock(&con->lock);
+
 	if (!wait_for_completion_timeout(&con->complete,
-				msecs_to_jiffies(UCSI_SWAP_TIMEOUT_MS))) {
-		ret = -ETIMEDOUT;
-		goto out_unlock;
-	}
+				msecs_to_jiffies(UCSI_SWAP_TIMEOUT_MS)))
+		return -ETIMEDOUT;
+
+	mutex_lock(&con->lock);
 
 	/* Something has gone wrong while swapping the role */
 	if (UCSI_CONSTAT_PWR_OPMODE(con->status.flags) !=
@@ -1372,6 +1378,7 @@ void ucsi_unregister(struct ucsi *ucsi)
 	ucsi->ops->async_write(ucsi, UCSI_CONTROL, &cmd, sizeof(cmd));
 
 	for (i = 0; i < ucsi->cap.num_connectors; i++) {
+		complete(&ucsi->connector[i].complete);
 		cancel_work_sync(&ucsi->connector[i].work);
 		ucsi_unregister_partner(&ucsi->connector[i]);
 		ucsi_unregister_altmodes(&ucsi->connector[i],
Signature · 1 line
2.34.1
author
#2 lore raw
Quoted 23 lines · Heikki Krogerus
Quoted 22 lines · Jia-Ju Bai

Hello,

My static analysis tool reports a possible deadlock in the ucsi driver in Linux 5.16:

ucsi_pr_swap()
  mutex_lock(&con->lock); --> Line 962 (Lock A) wait_for_completion_timeout(&con->complete, ...) --> Line 981 (Wait X)

ucsi_handle_connector_change()
  mutex_lock(&con->lock); --> Line 763 (Lock A) complete(&con->complete); --> Line 782 (Wake X) complete(&con->complete); --> Line 807 (Wake X)

When ucsi_pr_swap() is executed, "Wait X" is performed by holding "Lock A". If ucsi_handle_connector_change() is executed at this time, "Wake X" cannot be performed to wake up "Wait X" in ucsi_handle_connector_change(), because "Lock A" has been already held by ucsi_handle_connector_change(), causing a possible deadlock.
I find that "Wait X" is performed with a timeout, to relieve the possible deadlock; but I think this timeout can cause inefficient execution.

Quoted 5 lines
Quoted 2 lines

I am not quite sure whether this possible problem is real. Any feedback would be appreciated, thanks :)

This is probable a regression from commit ad74b8649bea ("usb: typec:
ucsi: Preliminary support for alternate modes"). Can you test does this patch fix the issue (attached)?

Hi Heikki,

Thanks for the reply and patch.
After the patch is used, my tool does not report this deadlock in the ucsi driver. Thus, I think this patch should be okay to fix the deadlock :)

Best wishes,
Jia-Ju Bai

maintainer
#3 lore raw
Quoted 32 lines · Jia-Ju Bai
Quoted 23 lines · Heikki Krogerus
Quoted 22 lines · Jia-Ju Bai

Hello,

My static analysis tool reports a possible deadlock in the ucsi driver in Linux 5.16:

ucsi_pr_swap()
  mutex_lock(&con->lock); --> Line 962 (Lock A) wait_for_completion_timeout(&con->complete, ...) --> Line 981 (Wait X)

ucsi_handle_connector_change()
  mutex_lock(&con->lock); --> Line 763 (Lock A) complete(&con->complete); --> Line 782 (Wake X) complete(&con->complete); --> Line 807 (Wake X)

When ucsi_pr_swap() is executed, "Wait X" is performed by holding "Lock A". If ucsi_handle_connector_change() is executed at this time, "Wake X" cannot be performed to wake up "Wait X" in ucsi_handle_connector_change(), because "Lock A" has been already held by ucsi_handle_connector_change(), causing a possible deadlock.
I find that "Wait X" is performed with a timeout, to relieve the possible deadlock; but I think this timeout can cause inefficient execution.

Quoted 5 lines
Quoted 2 lines

I am not quite sure whether this possible problem is real. Any feedback would be appreciated, thanks :)

This is probable a regression from commit ad74b8649bea ("usb: typec:
ucsi: Preliminary support for alternate modes"). Can you test does this patch fix the issue (attached)?

Quoted 6 lines

Hi Heikki,

Thanks for the reply and patch.
After the patch is used, my tool does not report this deadlock in the ucsi driver.
Thus, I think this patch should be okay to fix the deadlock :)

Thanks for testing it. I'll tag you as the reporter and tester in the patch.

Br,

Signature · 1 line
heikki
#4 lore raw
In reply to Heikki Krogerus

Hi Heikki,

Quoted 23 lines · Heikki Krogerus
Quoted 22 lines · Jia-Ju Bai

Hello,

My static analysis tool reports a possible deadlock in the ucsi driver in Linux 5.16:

ucsi_pr_swap()
  mutex_lock(&con->lock); --> Line 962 (Lock A) wait_for_completion_timeout(&con->complete, ...) --> Line 981 (Wait X)

ucsi_handle_connector_change()
  mutex_lock(&con->lock); --> Line 763 (Lock A)
  complete(&con->complete); --> Line 782 (Wake X)
  complete(&con->complete); --> Line 807 (Wake X)

When ucsi_pr_swap() is executed, "Wait X" is performed by holding "Lock A". If ucsi_handle_connector_change() is executed at this time, "Wake X" cannot be performed to wake up "Wait X" in ucsi_handle_connector_change(), because "Lock A" has been already held by ucsi_handle_connector_change(), causing a possible deadlock.
I find that "Wait X" is performed with a timeout, to relieve the possible deadlock; but I think this timeout can cause inefficient execution.

Quoted 6 lines
Quoted 2 lines

I am not quite sure whether this possible problem is real. Any feedback would be appreciated, thanks :)

This is probable a regression from commit ad74b8649bea ("usb: typec:
ucsi: Preliminary support for alternate modes"). Can you test does this patch fix the issue (attached)?

We encountered a slightly different twist to this bug. Instead of deadlocking, we see that the dr_swap() / pr_swap() operations actually jump out of the wait_for_completion_timeout() immediately, even before any partner change occurs. This is because the con->complete may already have its done flag set to true from the first time ucsi_handle_connector_change() runs, and is never reset after that.

In addition to the unlocking below, I think we need to also add reinit_completion() calls at the start of ucsi_{pr,dr}_swap().

Thanks,
Jack

Quoted 66 lines

From 2ad06425a3df7be656f8a5b3c202aab45554fd17 Mon Sep 17 00:00:00 2001
From: Heikki Krogerus <[email protected]>
Date: Wed, 9 Feb 2022 17:27:19 +0300
Subject: [PATCH] usb: typec: ucsi: Test fix

Interim.

---
 drivers/usb/typec/ucsi/ucsi.c | 19 +++++++++++++------
 1 file changed, 13 insertions(+), 6 deletions(-)
drivers/usb/typec/ucsi/ucsi.c +13 -6
index f0c2fa19f3e0f..225104beda8be 100644
@@ -956,14 +956,18 @@ static int ucsi_dr_swap(struct typec_port *port, enum typec_data_role role)
 	if (ret < 0)
 		goto out_unlock;
 
+	mutex_unlock(&con->lock);
+
 	if (!wait_for_completion_timeout(&con->complete,
 					msecs_to_jiffies(UCSI_SWAP_TIMEOUT_MS)))
-		ret = -ETIMEDOUT;
+		return -ETIMEDOUT;
+
+	return 0;
 
 out_unlock:
 	mutex_unlock(&con->lock);
 
-	return ret < 0 ? ret : 0;
+	return ret;
 }
 
 static int ucsi_pr_swap(struct typec_port *port, enum typec_role role)
@@ -992,11 +996,13 @@ static int ucsi_pr_swap(struct typec_port *port, enum typec_role role)
 	if (ret < 0)
 		goto out_unlock;
 
+	mutex_unlock(&con->lock);
+
 	if (!wait_for_completion_timeout(&con->complete,
-				msecs_to_jiffies(UCSI_SWAP_TIMEOUT_MS))) {
-		ret = -ETIMEDOUT;
-		goto out_unlock;
-	}
+				msecs_to_jiffies(UCSI_SWAP_TIMEOUT_MS)))
+		return -ETIMEDOUT;
+
+	mutex_lock(&con->lock);
 
 	/* Something has gone wrong while swapping the role */
 	if (UCSI_CONSTAT_PWR_OPMODE(con->status.flags) !=
@@ -1372,6 +1378,7 @@ void ucsi_unregister(struct ucsi *ucsi)
 	ucsi->ops->async_write(ucsi, UCSI_CONTROL, &cmd, sizeof(cmd));
 
 	for (i = 0; i < ucsi->cap.num_connectors; i++) {
+		complete(&ucsi->connector[i].complete);
 		cancel_work_sync(&ucsi->connector[i].work);
 		ucsi_unregister_partner(&ucsi->connector[i]);
 		ucsi_unregister_altmodes(&ucsi->connector[i],
Signature · 1 line
2.34.1
maintainer
#5 lore raw
Quoted 40 lines · Jack Pham

Hi Heikki,

Quoted 23 lines · Heikki Krogerus
Quoted 22 lines · Jia-Ju Bai

Hello,

My static analysis tool reports a possible deadlock in the ucsi driver in Linux 5.16:

ucsi_pr_swap()
  mutex_lock(&con->lock); --> Line 962 (Lock A) wait_for_completion_timeout(&con->complete, ...) --> Line 981 (Wait X)

ucsi_handle_connector_change()
  mutex_lock(&con->lock); --> Line 763 (Lock A)
  complete(&con->complete); --> Line 782 (Wake X)
  complete(&con->complete); --> Line 807 (Wake X)

When ucsi_pr_swap() is executed, "Wait X" is performed by holding "Lock A". If ucsi_handle_connector_change() is executed at this time, "Wake X" cannot be performed to wake up "Wait X" in ucsi_handle_connector_change(), because "Lock A" has been already held by ucsi_handle_connector_change(), causing a possible deadlock.
I find that "Wait X" is performed with a timeout, to relieve the possible deadlock; but I think this timeout can cause inefficient execution.

Quoted 6 lines
Quoted 2 lines

I am not quite sure whether this possible problem is real. Any feedback would be appreciated, thanks :)

This is probable a regression from commit ad74b8649bea ("usb: typec:
ucsi: Preliminary support for alternate modes"). Can you test does this patch fix the issue (attached)?

We encountered a slightly different twist to this bug. Instead of deadlocking, we see that the dr_swap() / pr_swap() operations actually jump out of the wait_for_completion_timeout() immediately, even before any partner change occurs. This is because the con->complete may already have its done flag set to true from the first time ucsi_handle_connector_change() runs, and is never reset after that.

Quoted 2 lines

In addition to the unlocking below, I think we need to also add reinit_completion() calls at the start of ucsi_{pr,dr}_swap().

OK. I'll add that to the patch.

thanks,

Signature · 1 line
heikki