In case of a disconnect an ongoing flush() has to be made fail.
Nevertheless we cannot be sure that any pending URB has already
finished, so although they will never succeed, they still must
not be touched.
The clean solution for this is to check for WDM_IN_USE
and WDM_DISCONNECTED in flush(). There is no point in ever
clearing WDM_IN_USE, as no further writes make sense.
The issue is as old as the driver.
Fixes: afba937e540c9 ("USB: CDC WDM driver")
Reported-by: syzbot+d232cca6ec42c2edb3fc@xxxxxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: Oliver Neukum <oneukum@xxxxxxxx>
---
drivers/usb/class/cdc-wdm.c | 16 ++++++++++++----
1 file changed, 12 insertions(+), 4 deletions(-)
diff --git a/drivers/usb/class/cdc-wdm.c b/drivers/usb/class/cdc-wdm.c
index 1656f5155ab8..f9f7c8a5e091 100644
--- a/drivers/usb/class/cdc-wdm.c
+++ b/drivers/usb/class/cdc-wdm.c
@@ -588,10 +588,20 @@ static int wdm_flush(struct file *file, fl_owner_t id)
{
struct wdm_device *desc = file->private_data;
- wait_event(desc->wait, !test_bit(WDM_IN_USE, &desc->flags));
+ wait_event(desc->wait,
+ /*
+ * needs both flags. We cannot do with one
+ * because resetting it would cause a race
+ * with write() yet we need to signal
+ * a disconnect
+ */
+ !test_bit(WDM_IN_USE, &desc->flags) ||
+ test_bit(WDM_DISCONNECTING, &desc->flags));
/* cannot dereference desc->intf if WDM_DISCONNECTING */
- if (desc->werr < 0 && !test_bit(WDM_DISCONNECTING, &desc->flags))
+ if (test_bit(WDM_DISCONNECTING, &desc->flags))
+ return -ENODEV;
+ if (desc->werr < 0)
dev_err(&desc->intf->dev, "Error in flush path: %d\n",
desc->werr);
@@ -975,8 +985,6 @@ static void wdm_disconnect(struct usb_interface *intf)
spin_lock_irqsave(&desc->iuspin, flags);
set_bit(WDM_DISCONNECTING, &desc->flags);
set_bit(WDM_READ, &desc->flags);
- /* to terminate pending flushes */
- clear_bit(WDM_IN_USE, &desc->flags);
spin_unlock_irqrestore(&desc->iuspin, flags);
wake_up_all(&desc->wait);
mutex_lock(&desc->rlock);
--
2.16.4
