On Wed, Apr 03, 2024 at 11:47:21AM +0200, Bartosz Golaszewski wrote: > On Wed, Apr 3, 2024 at 11:42 AM Kent Gibson <warthog618@xxxxxxxxx> wrote: > > > > > > It occurred to me that none of my tests cover this case, as they always > > request edges with the consumer set, so I added some and can confirm both > > the problem and the fix. > > > > In the process I found another bug - we overlooked setting up the irq > > label in debounce_setup() - the alternate path in edge_detector_setup() > > that performs sw debounce. That results in a double free of the > > req->label and memory corruption hilarity follows. > > > > I've got a patch for that - the unfortunate part being that > > debounce_setup() is earlier in the file than make_irq_label() and > > free_irq_label(). Those will need to be pushed earlier, so it is > > sure to conflict with this patch. > > How would you prefer to proceed? > > Can you take my patch and just make it part of your series? > Will do. Cheers, Kent.
