Skip to contentSkip to conversation

Potential order bug in 'drivers/usb/misc/ldusb.c', mainly in 'ld_usb_disconnect()'

bug report Ginger 2 participants 2 messages Jun 1 USB

lore
Share & tools
lore

Ginger

author
#0 lore raw

Dear Linux kernel maintainers,

My research-based static analyzer found a potential order bug within the 'drivers/usb/misc' subsystem, more specifically, in 'drivers/usb/misc/ldusb.c'.

This potential issue is similar to a previously reported one
(https://lore.kernel.org/linux-usb/2026042444-freeing-consumer-468b@gregkh/).
It is present as of git commit eb3f4b7426cfd2b79d65b7d37155480b32259a11.

Potential concurrent triggering executions:
T0:
ld_usb_disconnect

    --> usb_set_intfdata(interface, NULL); [t0]
    --> usb_deregister_dev(interface, &ld_usb_class);
         --> usb_minors[intf->minor] = NULL; [t2]
T1:
usb_open
   --> new_fops = fops_get(usb_minors[iminor(inode)]); [t1]
   --> err = file->f_op->open(inode, file);
       --> ...
       --> ld_usb_open
             --> dev = usb_get_intfdata(interface); [t3]

In T0, the interface is nullified before its get deregistered. Thus, it is possible for T1 to still get the usb dev and access it via the interface, which, however, has already been nullified. The concurrent buggy order is t0 -> t1 -> t2 -> t3.

Thank you for your time and consideration.

Sincerely,
Ginger

maintainer
#1 lore raw
Quoted 24 lines · Ginger

Dear Linux kernel maintainers,

My research-based static analyzer found a potential order bug within the 'drivers/usb/misc' subsystem, more specifically, in 'drivers/usb/misc/ldusb.c'.

This potential issue is similar to a previously reported one
(https://lore.kernel.org/linux-usb/2026042444-freeing-consumer-468b@gregkh/).
It is present as of git commit eb3f4b7426cfd2b79d65b7d37155480b32259a11.

Potential concurrent triggering executions:
T0:
ld_usb_disconnect

    --> usb_set_intfdata(interface, NULL); [t0]
    --> usb_deregister_dev(interface, &ld_usb_class);
         --> usb_minors[intf->minor] = NULL; [t2]
T1:
usb_open
   --> new_fops = fops_get(usb_minors[iminor(inode)]); [t1]
   --> err = file->f_op->open(inode, file);
       --> ...
       --> ld_usb_open
             --> dev = usb_get_intfdata(interface); [t3]
Quoted 6 lines

In T0, the interface is nullified before its get deregistered. Thus, it is possible for T1 to still get the usb dev and access it via the interface, which, however, has already been nullified. The concurrent buggy order is t0 -> t1 -> t2 -> t3.

Thank you for your time and consideration.

If you think this needs to be fixed, please just send a patch for it and we can evaluate it that way.

thanks,

greg k-h