On Wed, 5 Nov 2008, Oliver Neukum wrote:

this, I think, fixes everything.

Thanks a lot for working on this, Oliver. Could you please add a proper
changelog, describing all the changes that had been done and why they were
needed, so that this can be queued for 2.6.29?

 				schedule();
+				if (mutex_lock_interruptible(&list->thread_lock))
+					return -ERESTARTSYS;

Is ERESTARTSYS really appropriate here? Tail could have already moved from
within other thread, right?

 	case HIDIOCGSTRING:
-		return hiddev_ioctl_string(hiddev, cmd, user_arg);
+		lock_kernel();
+		if (!hiddev->exist)
+			r = hiddev_ioctl_string(hiddev, cmd, user_arg);
+		else
+			r = -ENODEV;
+		unlock_kernel();
+		return r;

Hmm, I know that this is a legacy interface, but I am really hesitant to
add a new synchronization using lock_kernel(), instead of removing it
gradually altogether.

Thanks,

-- 
Jiri Kosina
SUSE Labs