git.ipfire.org Git - thirdparty/kernel/stable-queue.git/blob

   1 From 8bd29dbe03fc5b0f039ab2395ff37b64236d2f0c Mon Sep 17 00:00:00 2001
   2 From: Jeongjun Park <aha310510@gmail.com>
   3 Date: Sun, 11 Jan 2026 00:17:53 +0900
   4 Subject: media: as102: fix to not free memory after the device is registered in as102_usb_probe()
   5
   6 From: Jeongjun Park <aha310510@gmail.com>
   7
   8 commit 8bd29dbe03fc5b0f039ab2395ff37b64236d2f0c upstream.
   9
  10 In as102_usb driver, the following race condition occurs:
  11 ```
  12                 CPU0                                            CPU1
  13 as102_usb_probe()
  14   kzalloc(); // alloc as102_dev_t
  15   ....
  16   usb_register_dev();
  17                                                 fd = sys_open("/path/to/dev"); // open as102 fd
  18                                                 ....
  19   usb_deregister_dev();
  20   ....
  21   kfree(); // free as102_dev_t
  22   ....
  23                                                 sys_close(fd);
  24                                                   as102_release() // UAF!!
  25                                                     as102_usb_release()
  26                                                       kfree(); // DFB!!
  27 ```
  28
  29 When a USB character device registered with usb_register_dev() is later
  30 unregistered (via usb_deregister_dev() or disconnect), the device node is
  31 removed so new open() calls fail. However, file descriptors that are
  32 already open do not go away immediately: they remain valid until the last
  33 reference is dropped and the driver's .release() is invoked.
  34
  35 In as102, as102_usb_probe() calls usb_register_dev() and then, on an
  36 error path, does usb_deregister_dev() and frees as102_dev_t right away.
  37 If userspace raced a successful open() before the deregistration, that
  38 open FD will later hit as102_release() --> as102_usb_release() and access
  39 or free as102_dev_t again, occur a race to use-after-free and
  40 double-free vuln.
  41
  42 The fix is to never kfree(as102_dev_t) directly once usb_register_dev()
  43 has succeeded. After deregistration, defer freeing memory to .release().
  44
  45 In other words, let release() perform the last kfree when the final open
  46 FD is closed.
  47
  48 Cc: <stable@vger.kernel.org>
  49 Reported-by: syzbot+47321e8fd5a4c84088db@syzkaller.appspotmail.com
  50 Closes: https://syzkaller.appspot.com/bug?extid=47321e8fd5a4c84088db
  51 Fixes: cd19f7d3e39b ("[media] as102: fix leaks at failure paths in as102_usb_probe()")
  52 Signed-off-by: Jeongjun Park <aha310510@gmail.com>
  53 Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
  54 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  55 ---
  56  drivers/media/usb/as102/as102_usb_drv.c |    2 ++
  57  1 file changed, 2 insertions(+)
  58
  59 --- a/drivers/media/usb/as102/as102_usb_drv.c
  60 +++ b/drivers/media/usb/as102/as102_usb_drv.c
  61 @@ -403,7 +403,9 @@ static int as102_usb_probe(struct usb_in
  62  failed_dvb:
  63         as102_free_usb_stream_buffer(as102_dev);
  64  failed_stream:
  65 +       usb_set_intfdata(intf, NULL);
  66         usb_deregister_dev(intf, &as102_usb_class_driver);
  67 +       return ret;
  68  failed:
  69         usb_put_dev(as102_dev->bus_adap.usb_dev);
  70         usb_set_intfdata(intf, NULL);