Skip to contentSkip to conversation

mmc: vub300: fix Use-After-Free of inactivity_timer on disconnect and probe failure

PATCH In review Muhammad Bilal 2 participants 4 messages Aug 18 – Aug 19

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mmc: vub300: fix Use-After-Free of inactivity_timer on disconnect and probe failure

The vub300 driver arms inactivity_timer during probe and acquires a kref reference for the running timer. However, vub300_disconnect() sets vub300->interface = NULL and drops the driver's kref without shutting down the timer via timer_shutdown_sync().

If the timer expires after the host is removed and vub300 is freed, or if concurrent work items re-arm the timer, vub300_inactivity_timer_expired() accesses the already freed vub300 structure, causing a Use-After-Free.

Fix this by explicitly shutting down inactivity_timer via timer_shutdown_sync() and releasing the timer's kref reference in vub300_disconnect() and probe error cleanup paths.

Fixes 88095e7b473a ("mmc: Add new VUB300 USB-to-SD/SDIO/MMC driver")
---
 drivers/mmc/host/vub300.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/mmc/host/vub300.c +4 -0
index 8e45dc1a4a12..9f84bd1531e2 100644
@@ -2350,6 +2350,8 @@ static int vub300_probe(struct usb_interface *interface,
 
 err_stop_io:
 	vub300->interface = NULL;
+	timer_shutdown_sync(&vub300->inactivity_timer);
+	kref_put(&vub300->kref, vub300_delete);
 	kref_put(&vub300->kref, vub300_delete);
 
 	return retval;
@@ -2385,6 +2387,8 @@ static void vub300_disconnect(struct usb_interface *interface)
 			usb_set_intfdata(interface, NULL);
 			/* prevent more I/O from starting */
 			vub300->interface = NULL;
+			timer_shutdown_sync(&vub300->inactivity_timer);
+			kref_put(&vub300->kref, vub300_delete);
 			mmc_remove_host(mmc);
 			kref_put(&vub300->kref, vub300_delete);
 			pr_info("USB vub300 remote SDIO host controller[%d]"
Signature · 1 line
2.43.0
#1 lore raw
Quoted 9 lines · Muhammad Bilal

The vub300 driver arms inactivity_timer during probe and acquires a kref reference for the running timer. However, vub300_disconnect() sets vub300->interface = NULL and drops the driver's kref without shutting down the timer via timer_shutdown_sync().

If the timer expires after the host is removed and vub300 is freed, or if concurrent work items re-arm the timer, vub300_inactivity_timer_expired() accesses the already freed vub300 structure, causing a Use-After-Free.

No, this cannot happen as the timer holds another reference.

Quoted 6 lines

Fix this by explicitly shutting down inactivity_timer via timer_shutdown_sync() and releasing the timer's kref reference in vub300_disconnect() and probe error cleanup paths.

Fixes 88095e7b473a ("mmc: Add new VUB300 USB-to-SD/SDIO/MMC driver")

How was this "issue" found? Are you using an LLM?

Quoted 9 lines
---
 drivers/mmc/host/vub300.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/mmc/host/vub300.c +0 -0
index 8e45dc1a4a12..9f84bd1531e2 100644

drivers/mmc/host/vub300.c static int vub300_probe(struct usb_interface *interface, · line 2355

Quoted 6 lines
drivers/mmc/host/vub300.c +2 -0
@@ -2350,6 +2350,8 @@ static int vub300_probe(struct usb_interface *interface,
 
 err_stop_io:
 	vub300->interface = NULL;
+	timer_shutdown_sync(&vub300->inactivity_timer);
+	kref_put(&vub300->kref, vub300_delete);
 	kref_put(&vub300->kref, vub300_delete);

And this results in a reference underflow.

Johan

#2 lore raw

Hi Johan,

You're right, thank you for catching this. I traced through vub300_probe() and vub300_inactivity_timer_expired() again: the kref_get() taken right before add_timer() is exactly the extra reference the timer holds, and the timer callback releases it itself and stops rearming once vub300->interface is NULL. So the UAF I described can't happen with the code as it stands.

I should also be upfront that this was flagged by Sashiko, and I did not verify the reference counting closely enough myself before sending it.
That's on me. Dropping this patch.

Sorry for the noise.
Muhammad Bilal

Quoted 42 lines · Johan Hovold
Quoted 9 lines · Muhammad Bilal

The vub300 driver arms inactivity_timer during probe and acquires a kref reference for the running timer. However, vub300_disconnect() sets vub300->interface = NULL and drops the driver's kref without shutting down the timer via timer_shutdown_sync().

If the timer expires after the host is removed and vub300 is freed, or if concurrent work items re-arm the timer, vub300_inactivity_timer_expired() accesses the already freed vub300 structure, causing a Use-After-Free.

No, this cannot happen as the timer holds another reference.

Quoted 6 lines

Fix this by explicitly shutting down inactivity_timer via timer_shutdown_sync() and releasing the timer's kref reference in vub300_disconnect() and probe error cleanup paths.

Fixes 88095e7b473a ("mmc: Add new VUB300 USB-to-SD/SDIO/MMC driver")

How was this "issue" found? Are you using an LLM?

Quoted 9 lines
---
 drivers/mmc/host/vub300.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/mmc/host/vub300.c +0 -0
index 8e45dc1a4a12..9f84bd1531e2 100644

drivers/mmc/host/vub300.c static int vub300_probe(struct usb_interface *interface, · line 2355

Quoted 6 lines
drivers/mmc/host/vub300.c +2 -0
@@ -2350,6 +2350,8 @@ static int vub300_probe(struct usb_interface *interface,
 
 err_stop_io:
      vub300->interface = NULL;
+     timer_shutdown_sync(&vub300->inactivity_timer);
+     kref_put(&vub300->kref, vub300_delete);
      kref_put(&vub300->kref, vub300_delete);

And this results in a reference underflow.

Johan

#3 lore raw

[ Please avoid top-posting when replying on the lists. ]

Quoted 10 lines · Muhammad Bilal

You're right, thank you for catching this. I traced through vub300_probe() and vub300_inactivity_timer_expired() again: the kref_get() taken right before add_timer() is exactly the extra reference the timer holds, and the timer callback releases it itself and stops rearming once vub300->interface is NULL. So the UAF I described can't happen with the code as it stands.

I should also be upfront that this was flagged by Sashiko, and I did not verify the reference counting closely enough myself before sending it.

Remember to mention when issues were found (or fixed) by tools like Sashiko or LLMs generally in the commit message of any future patches as its useful information for reviewers and maintainers (and as is also documented as being required).

Johan