Skip to contentSkip to conversation

USB: serial: keyspan: Fixing use-after-free in usa49_glocont_callback()

PATCH In review Rafael Alejandro Diaz Cruz 2 participants 2 messages Aug 20

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

PATCH 1/1 USB: serial: keyspan: Fixing use-after-free in usa49_glocont_callback()

usa49_glocont_callback() obtains the private data from all ports of a serial device using usb_get_serial_port_data(). This can happen during keyspan_port_remove() which frees the private data which can cause UAF.

Fix this by calling usb_set_serial_port_data(port, NULL) inside keyspan_port_remove() right before kfree() of private data. This in turn allows "if (!p_priv) continue;" condition within usa49_glocont_callback() to execute properly.

---
 drivers/usb/serial/keyspan.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/usb/serial/keyspan.c +2 -0
index 4d3746c7a94e..0e08acb99970 100644
@@ -2997,6 +2997,8 @@ static void keyspan_port_remove(struct usb_serial_port *port)
 	for (i = 0; i < ARRAY_SIZE(p_priv->in_buffer); ++i)
 		kfree(p_priv->in_buffer[i]);
 
+	usb_set_serial_port_data(port, NULL);
+
 	kfree(p_priv);
 }
 
Signature · 1 line
2.43.0
maintainer
#1 lore raw
Quoted 5 lines · Rafael Alejandro Diaz Cruz

usa49_glocont_callback() obtains the private data from all ports of a serial device using usb_get_serial_port_data(). This can happen during keyspan_port_remove() which frees the private data which can cause UAF.

Quoted 7 lines

Fix this by calling usb_set_serial_port_data(port, NULL) inside keyspan_port_remove() right before kfree() of private data. This in turn allows "if (!p_priv) continue;" condition within usa49_glocont_callback() to execute properly.

Reported-by [email protected]

Please include a Link to the report as well (and a Fixes tag).

Quoted 10 lines
---
 drivers/usb/serial/keyspan.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/usb/serial/keyspan.c +0 -0
index 4d3746c7a94e..0e08acb99970 100644

drivers/usb/serial/keyspan.c static void keyspan_port_remove(struct usb_serial_port *port) · line 3001

Quoted 5 lines
drivers/usb/serial/keyspan.c +2 -0
@@ -2997,6 +2997,8 @@ static void keyspan_port_remove(struct usb_serial_port *port)
 	for (i = 0; i < ARRAY_SIZE(p_priv->in_buffer); ++i)
 		kfree(p_priv->in_buffer[i]);
 
+	usb_set_serial_port_data(port, NULL);
+

This doesn't work as nothing prevents the completion handler from loading the pointer just before you set it to NULL here.

drivers/usb/serial/keyspan.c static void keyspan_port_remove(struct usb_serial_port *port) · line 3003

Quoted 2 lines
drivers/usb/serial/keyspan.c +0 -0
 	kfree(p_priv);
 }

The UAF syzbot reported is a symptom of a general problem with the disconnect handling. I've just sent a fix to address this here:

	https://lore.kernel.org/all/[email protected]/

Johan