> On Jun 3, 2022, at 6:28 AM, Gerd Hoffmann <kraxel@xxxxxxxxxx> wrote: > > ⚠ External Email > > Hi, > >> the legacy kms to atomic. This left virtualized drivers, all which >> are atomic, in a weird spot because all userspace compositors put >> those drivers on deny-lists for atomic kms due to the fact that mouse >> clicks were incorrectly routed, e.g: > >> - all userspace code needs to hardcore a list of drivers which require >> hotspots because there's no way to query from drm "does this driver >> require hotspot" > > So drivers move from one list to another. Not ideal, but also not worse > than it was before, so: Yea, that problem was always there because we never had a way of checking whether a particular driver requires hotspot info. Unfortunately I don’t think this can not be solved in the kernel anymore. Technically the hotspot properties solve it: does the driver expose hotspot properties on cursor planes? If yes, then it requires hotspot info. But we can’t differentiate between: - driver that doesn’t require hotspot info - driver that requires hotspot data but we’re running on a kernel before those properties were implemented We’d have to backport those properties to every kernel between now and when atomic kms was implemented. Same with any other check we could add to the kernel. Likely libdrm would be the place for this atm but I’m not sure in what form. z
