On Sun, Mar 17, 2013 at 08:50:03PM +0100, Daniel Vetter wrote: > Doesn't that mean that we need these checks everywhere? Or at least a > fixup in drm core proper? > > And I think we need to add trinity to our test setup eventually ;-) Note that trinity's ioctl fuzzing is still very new (added in just the last few weeks), and for drm isn't very advanced at all yet. I was pretty surprised when Tommi's changes started turning up bugs so quickly, but I guess a lot of the ioctl paths have just never been audited for these kinds of bugs. As you can see at https://github.com/kernelslacker/trinity/blob/master/ioctls/drm.c It's literally just enumerating the known ioctl's, and using the generic fuzzing routines (so it just guesses what the argument is, and hence passes crap like NULL, or a page of garbage). Eventually I'd like to have routines for each of the individual ioctl cases to pass something that looks slightly more realistic to what it's expecting to see. (Compare to say, the SCSI SG_IO routines here: https://github.com/kernelslacker/trinity/blob/master/ioctls/scsi.c [still kinda dumb, but gives an idea of the direction]) Lots of work ahead. Dave
