Skip to contentSkip to conversation

FAILED: patch "[PATCH] thunderbolt: Prevent XDomain delayed work use-after-free on" failed to apply to 6.18-stable tree

PATCH New [email protected] 2 participants 5 messages Jul 20 – Jul 26

Subsystems (2)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

The patch below does not apply to the 6.18-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.18.y
git checkout FETCH_HEAD
git cherry-pick -x 2c5d2d3c3f70cde2565d7b279b544893a2035842
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2026072000-karma-payable-51b3@gregkh' --subject-prefix 'PATCH 6.18.y' 'HEAD^..'

Possible dependencies:

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 2c5d2d3c3f70cde2565d7b279b544893a2035842 Mon Sep 17 00:00:00 2001
From: Michael Bommarito <[email protected]>
Date: Wed, 27 May 2026 07:46:04 -0400
Subject: [PATCH] thunderbolt: Prevent XDomain delayed work use-after-free on
disconnect

tb_xdp_handle_request() runs on system_wq and queues xd->state_work via queue_delayed_work() in three request handlers:
PROPERTIES_CHANGED_REQUEST, UUID_REQUEST (via start_handshake), and LINK_STATE_CHANGE_REQUEST. Similarly, update_xdomain() queues xd->properties_changed_work when local properties change.

Concurrently, tb_xdomain_remove() calls stop_handshake() which does cancel_delayed_work_sync() on both delayed works. Later, tb_xdomain_unregister() calls device_unregister() which eventually frees the xdomain. Since commit 559c1e1e0134 ("thunderbolt: Run tb_xdp_handle_request() in system workqueue") moved the request handler off tb->wq, the handler and the remove path are no longer serialized. If queue_delayed_work() executes after cancel_delayed_work_sync() but before the xdomain is freed, the delayed work fires on a freed object.

Add xd->removing that tb_xdomain_remove() sets under xd->lock before calling stop_handshake(). Each external queue site holds the same lock and checks removing before calling queue_delayed_work(). This provides the mutual exclusion needed:
either the queue site acquires the lock first and queues work that the subsequent cancel will see, or the remove path acquires the lock first and the queue site observes removing == true and skips the queue.

Fixes 559c1e1e0134 ("thunderbolt: Run tb_xdp_handle_request() in system workqueue")
Assisted-by Claude:claude-opus-4-7
drivers/thunderbolt/xdomain.c +31 -10
index 781d88d06b93..fe6c5ac703f4 100644
@@ -803,9 +803,13 @@ static void tb_xdp_handle_request(struct work_struct *work)
 		 * the xdomain related to this connection as well in
 		 * case there is a change in services it offers.
 		 */
-		if (xd && device_is_registered(&xd->dev))
-			queue_delayed_work(tb->wq, &xd->state_work,
-					   msecs_to_jiffies(XDOMAIN_SHORT_TIMEOUT));
+		if (xd) {
+			mutex_lock(&xd->lock);
+			if (!xd->removing && device_is_registered(&xd->dev))
+				queue_delayed_work(tb->wq, &xd->state_work,
+						   msecs_to_jiffies(XDOMAIN_SHORT_TIMEOUT));
+			mutex_unlock(&xd->lock);
+		}
 		break;
 
 	case UUID_REQUEST_OLD:
@@ -818,8 +822,12 @@ static void tb_xdp_handle_request(struct work_struct *work)
 		 * received UUID request from the remote host.
 		 */
 		if (!ret && xd && xd->state == XDOMAIN_STATE_ERROR) {
-			dev_dbg(&xd->dev, "restarting handshake\n");
-			start_handshake(xd);
+			mutex_lock(&xd->lock);
+			if (!xd->removing) {
+				dev_dbg(&xd->dev, "restarting handshake\n");
+				start_handshake(xd);
+			}
+			mutex_unlock(&xd->lock);
 		}
 		break;
 
@@ -885,9 +893,13 @@ static void tb_xdp_handle_request(struct work_struct *work)
 
 			ret = tb_xdp_link_state_change_response(ctl, route,
 								sequence, 0);
-			xd->target_link_width = lsc->tlw;
-			queue_delayed_work(tb->wq, &xd->state_work,
-					   msecs_to_jiffies(XDOMAIN_SHORT_TIMEOUT));
+			mutex_lock(&xd->lock);
+			if (!xd->removing) {
+				xd->target_link_width = lsc->tlw;
+				queue_delayed_work(tb->wq, &xd->state_work,
+						   msecs_to_jiffies(XDOMAIN_SHORT_TIMEOUT));
+			}
+			mutex_unlock(&xd->lock);
 		} else {
 			tb_xdp_error_response(ctl, route, sequence,
 					      ERROR_NOT_READY);
@@ -971,8 +983,12 @@ static int update_xdomain(struct device *dev, void *data)
 
 	xd = tb_to_xdomain(dev);
 	if (xd) {
-		queue_delayed_work(xd->tb->wq, &xd->properties_changed_work,
-				   msecs_to_jiffies(50));
+		mutex_lock(&xd->lock);
+		if (!xd->removing)
+			queue_delayed_work(xd->tb->wq,
+					   &xd->properties_changed_work,
+					   msecs_to_jiffies(50));
+		mutex_unlock(&xd->lock);
 	}
 
 	return 0;
@@ -2200,6 +2216,11 @@ static int unregister_service(struct device *dev, void *data)
 void tb_xdomain_remove(struct tb_xdomain *xd)
 {
 	tb_xdomain_debugfs_remove(xd);
+
+	mutex_lock(&xd->lock);
+	xd->removing = true;
+	mutex_unlock(&xd->lock);
+
 	stop_handshake(xd);
 	tb_xdomain_link_exit(xd);
 
include/linux/thunderbolt.h +3 -0
index b5659f883517..feb1af175cfd 100644
@@ -213,6 +213,8 @@ enum tb_link_width {
  * @link_width: Width of the downstream facing link
  * @link_usb4: Downstream link is USB4
  * @is_unplugged: The XDomain is unplugged
+ * @removing: Set by tb_xdomain_remove() under @lock to prevent
+ *	      concurrent delayed work queueing
  * @needs_uuid: If the XDomain does not have @remote_uuid it will be
  *		queried first
  * @service_ids: Used to generate IDs for the services
@@ -262,6 +264,7 @@ struct tb_xdomain {
 	enum tb_link_width link_width;
 	bool link_usb4;
 	bool is_unplugged;
+	bool removing;
 	bool needs_uuid;
 	struct ida service_ids;
 	struct ida in_hopids;
maintainer
#1 lore raw

PATCH 1/4 thunderbolt: Keep XDomain reference during the lifetime of a service

From: Mika Westerberg <[email protected]>

[ Upstream commit 8b4060998637f06975fceee9b73845d8672d411e ]

This is needed because we release the service ID in tb_service_release() and the ID array is owned by the parent XDomain.

Stable-dep-of: 2c5d2d3c3f70 ("thunderbolt: Prevent XDomain delayed work use-after-free on disconnect")

---
 drivers/thunderbolt/xdomain.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/thunderbolt/xdomain.c +2 -1
index 458476907eab05..fd2356db934fe3 100644
@@ -1012,6 +1012,7 @@ static void tb_service_release(struct device *dev)
 	ida_free(&xd->service_ids, svc->id);
 	kfree(svc->key);
 	kfree(svc);
+	tb_xdomain_put(xd);
 }
 
 const struct device_type tb_service_type = {
@@ -1120,7 +1121,7 @@ static void enumerate_services(struct tb_xdomain *xd)
 		svc->id = id;
 		svc->dev.bus = &tb_bus_type;
 		svc->dev.type = &tb_service_type;
-		svc->dev.parent = &xd->dev;
+		svc->dev.parent = get_device(&xd->dev);
 		dev_set_name(&svc->dev, "%s.%d", dev_name(&xd->dev), svc->id);
 
 		tb_service_debugfs_init(svc);
Signature · 1 line
2.53.0
maintainer
#2 lore raw

PATCH 2/4 thunderbolt: Remove service debugfs entries during unregister

From: Mika Westerberg <[email protected]>

[ Upstream commit 4d5fc3f4068568dfcb8cbe2852b4adc56394aa26 ]

We add them as part of the register path so to keep it symmetric remove them as part of the unregister path. This also removes them even if the service itself is not yet released (but is unregistered), thus allowing new register with the same service name to happen.

Stable-dep-of: 2c5d2d3c3f70 ("thunderbolt: Prevent XDomain delayed work use-after-free on disconnect")

---
 drivers/thunderbolt/xdomain.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)
drivers/thunderbolt/xdomain.c +11 -3
index fd2356db934fe3..56344260b4f2c6 100644
@@ -1008,7 +1008,6 @@ static void tb_service_release(struct device *dev)
 	struct tb_service *svc = container_of(dev, struct tb_service, dev);
 	struct tb_xdomain *xd = tb_service_parent(svc);
 
-	tb_service_debugfs_remove(svc);
 	ida_free(&xd->service_ids, svc->id);
 	kfree(svc->key);
 	kfree(svc);
@@ -1023,6 +1022,14 @@ const struct device_type tb_service_type = {
 };
 EXPORT_SYMBOL_GPL(tb_service_type);
 
+static void __unregister_service(struct device *dev)
+{
+	struct tb_service *svc = tb_to_service(dev);
+
+	tb_service_debugfs_remove(svc);
+	device_unregister(&svc->dev);
+}
+
 static int remove_missing_service(struct device *dev, void *data)
 {
 	struct tb_xdomain *xd = data;
@@ -1034,7 +1041,7 @@ static int remove_missing_service(struct device *dev, void *data)
 
 	if (!tb_property_find(xd->remote_properties, svc->key,
 			      TB_PROPERTY_TYPE_DIRECTORY))
-		device_unregister(dev);
+		__unregister_service(dev);
 
 	return 0;
 }
@@ -1127,6 +1134,7 @@ static void enumerate_services(struct tb_xdomain *xd)
 		tb_service_debugfs_init(svc);
 
 		if (device_register(&svc->dev)) {
+			tb_service_debugfs_remove(svc);
 			put_device(&svc->dev);
 			break;
 		}
@@ -2059,7 +2067,7 @@ void tb_xdomain_add(struct tb_xdomain *xd)
 
 static int unregister_service(struct device *dev, void *data)
 {
-	device_unregister(dev);
+	__unregister_service(dev);
 	return 0;
 }
 
Signature · 1 line
2.53.0
maintainer
#3 lore raw

PATCH 3/4 thunderbolt: Remove XDomain from the bus without holding tb->lock

In reply to Sasha Levin

From: Mika Westerberg <[email protected]>

[ Upstream commit a8937f35cf39c39c64325aa84d0463d866850857 ]

Currently we call device_unregister() for services and the XDomain itself with tb->lock held. This prevents the service drivers from calling any functions that may take it. For this reason separate removing the XDomain from the topology data structures (where we need the lock) from unregistering the device from the bus (where remove callbacks of the drivers are being called).

Stable-dep-of: 2c5d2d3c3f70 ("thunderbolt: Prevent XDomain delayed work use-after-free on disconnect")

---
 drivers/thunderbolt/debugfs.c |  2 ++
 drivers/thunderbolt/domain.c  | 30 ++++++++++++++++++
 drivers/thunderbolt/icm.c     |  5 +++
 drivers/thunderbolt/switch.c  | 14 +++++++++
 drivers/thunderbolt/tb.c      | 59 +++++++++++++++++------------------
 drivers/thunderbolt/tb.h      |  2 ++
 drivers/thunderbolt/xdomain.c | 53 +++++++++++++++++++------------
 7 files changed, 115 insertions(+), 50 deletions(-)
drivers/thunderbolt/debugfs.c +2 -0
index 71c9833563c709..2968965c76d3f6 100644
@@ -1786,6 +1786,8 @@ static void margining_port_remove(struct tb_port *port)
 
 	if (!port->usb4)
 		return;
+	if (!port->usb4->margining)
+		return;
 
 	snprintf(dir_name, sizeof(dir_name), "port%d", port->port);
 	parent = debugfs_lookup(dir_name, port->sw->debugfs_dir);
drivers/thunderbolt/domain.c +30 -0
index 83defc915d3381..7210415cd269df 100644
@@ -850,6 +850,36 @@ int tb_domain_disconnect_all_paths(struct tb *tb)
 	return bus_for_each_dev(&tb_bus_type, NULL, tb, disconnect_xdomain);
 }
 
+struct unregister_context {
+	const struct tb *tb;
+	int n;
+};
+
+static int unregister_unplugged_xdomain(struct device *dev, void *data)
+{
+	struct unregister_context *ctx = data;
+	struct tb_xdomain *xd;
+
+	xd = tb_to_xdomain(dev);
+	if (xd && xd->tb == ctx->tb && xd->is_unplugged) {
+		tb_xdomain_unregister(xd);
+		ctx->n++;
+	}
+	return 0;
+}
+
+int tb_domain_unregister_unplugged_xdomains(struct tb *tb)
+{
+	struct unregister_context ctx;
+
+	ctx.tb = tb_domain_get(tb);
+	ctx.n = 0;
+	bus_for_each_dev(&tb_bus_type, NULL, &ctx, unregister_unplugged_xdomain);
+	tb_domain_put(tb);
+
+	return ctx.n;
+}
+
 int tb_domain_init(void)
 {
 	int ret;
drivers/thunderbolt/icm.c +5 -0
index f213d9174dc574..0a17b2d0b6e8d3 100644
@@ -738,6 +738,7 @@ static void remove_xdomain(struct tb_xdomain *xd)
 
 	sw = tb_to_switch(xd->dev.parent);
 	tb_port_at(xd->route, sw)->xdomain = NULL;
+	xd->is_unplugged = true;
 	tb_xdomain_remove(xd);
 }
 
@@ -1762,6 +1763,8 @@ static void icm_handle_notification(struct work_struct *work)
 
 	kfree(n->pkg);
 	kfree(n);
+
+	tb_domain_unregister_unplugged_xdomains(tb);
 }
 
 static void icm_handle_event(struct tb *tb, enum tb_cfg_pkg_type type,
@@ -2112,6 +2115,8 @@ static void icm_rescan_work(struct work_struct *work)
 	if (tb->root_switch)
 		icm_free_unplugged_children(tb->root_switch);
 	mutex_unlock(&tb->lock);
+
+	tb_domain_unregister_unplugged_xdomains(tb);
 }
 
 static void icm_complete(struct tb *tb)
drivers/thunderbolt/switch.c +14 -0
index 0e07904aa73b45..b628ca702717e3 100644
@@ -3603,6 +3603,20 @@ int tb_switch_resume(struct tb_switch *sw, bool runtime)
 				tb_port_warn(port,
 					     "lost during suspend, disconnecting\n");
 				tb_sw_set_unplugged(port->remote->sw);
+			} else if (port->xdomain) {
+				/*
+				 * If the user replaced the XDomain with
+				 * another router, this will succeed in
+				 * which case we must remove the XDomain
+				 * before adding the new router.
+				 */
+				err = tb_cfg_get_upstream_port(sw->tb->ctl,
+							       port->xdomain->route);
+				if (err > 0) {
+					tb_port_warn(port,
+						     "XDomain was disconnected\n");
+					port->xdomain->is_unplugged = true;
+				}
 			}
 		}
 	}
drivers/thunderbolt/tb.c +29 -30

100 lines not on this page · raw message

drivers/thunderbolt/tb.h +2 -0
index 8e2762ff8d5178..76958a801be3d2 100644
@@ -792,6 +792,7 @@ int tb_domain_disconnect_xdomain_paths(struct tb *tb, struct tb_xdomain *xd,
 				       int transmit_path, int transmit_ring,
 				       int receive_path, int receive_ring);
 int tb_domain_disconnect_all_paths(struct tb *tb);
+int tb_domain_unregister_unplugged_xdomains(struct tb *tb);
 
 static inline struct tb *tb_domain_get(struct tb *tb)
 {
@@ -1262,6 +1263,7 @@ struct tb_xdomain *tb_xdomain_alloc(struct tb *tb, struct device *parent,
 				    const uuid_t *remote_uuid);
 void tb_xdomain_add(struct tb_xdomain *xd);
 void tb_xdomain_remove(struct tb_xdomain *xd);
+void tb_xdomain_unregister(struct tb_xdomain *xd);
 struct tb_xdomain *tb_xdomain_find_by_link_depth(struct tb *tb, u8 link,
 						 u8 depth);
 
drivers/thunderbolt/xdomain.c +33 -20

73 lines not on this page · raw message

Signature · 1 line
2.53.0
maintainer
#4 lore raw

PATCH 4/4 thunderbolt: Prevent XDomain delayed work use-after-free on disconnect

In reply to Sasha Levin

From: Michael Bommarito <[email protected]>

[ Upstream commit 2c5d2d3c3f70cde2565d7b279b544893a2035842 ]

tb_xdp_handle_request() runs on system_wq and queues xd->state_work via queue_delayed_work() in three request handlers:
PROPERTIES_CHANGED_REQUEST, UUID_REQUEST (via start_handshake), and LINK_STATE_CHANGE_REQUEST. Similarly, update_xdomain() queues xd->properties_changed_work when local properties change.

Concurrently, tb_xdomain_remove() calls stop_handshake() which does cancel_delayed_work_sync() on both delayed works. Later, tb_xdomain_unregister() calls device_unregister() which eventually frees the xdomain. Since commit 559c1e1e0134 ("thunderbolt: Run tb_xdp_handle_request() in system workqueue") moved the request handler off tb->wq, the handler and the remove path are no longer serialized. If queue_delayed_work() executes after cancel_delayed_work_sync() but before the xdomain is freed, the delayed work fires on a freed object.

Add xd->removing that tb_xdomain_remove() sets under xd->lock before calling stop_handshake(). Each external queue site holds the same lock and checks removing before calling queue_delayed_work(). This provides the mutual exclusion needed:
either the queue site acquires the lock first and queues work that the subsequent cancel will see, or the remove path acquires the lock first and the queue site observes removing == true and skips the queue.

Fixes 559c1e1e0134 ("thunderbolt: Run tb_xdp_handle_request() in system workqueue")
Assisted-by Claude:claude-opus-4-7
---
 drivers/thunderbolt/xdomain.c | 26 +++++++++++++-------------
 1 file changed, 13 insertions(+), 13 deletions(-)
drivers/thunderbolt/xdomain.c +13 -13
index 8035dd51d62f10..5efb0540c1d15c 100644
@@ -909,6 +909,19 @@ void tb_unregister_service_driver(struct tb_service_driver *drv)
 }
 EXPORT_SYMBOL_GPL(tb_unregister_service_driver);
 
+static int update_xdomain(struct device *dev, void *data)
+{
+	struct tb_xdomain *xd;
+
+	xd = tb_to_xdomain(dev);
+	if (xd) {
+		queue_delayed_work(xd->tb->wq, &xd->properties_changed_work,
+				   msecs_to_jiffies(50));
+	}
+
+	return 0;
+}
+
 static ssize_t key_show(struct device *dev, struct device_attribute *attr,
 			char *buf)
 {
@@ -2500,19 +2513,6 @@ bool tb_xdomain_handle_request(struct tb *tb, enum tb_cfg_pkg_type type,
 	return ret > 0;
 }
 
-static int update_xdomain(struct device *dev, void *data)
-{
-	struct tb_xdomain *xd;
-
-	xd = tb_to_xdomain(dev);
-	if (xd) {
-		queue_delayed_work(xd->tb->wq, &xd->properties_changed_work,
-				   msecs_to_jiffies(50));
-	}
-
-	return 0;
-}
-
 static void update_all_xdomains(void)
 {
 	bus_for_each_dev(&tb_bus_type, NULL, NULL, update_xdomain);
Signature · 1 line
2.53.0