On Tue Jun 2, 2026 at 12:35 PM CEST, Gary Guo wrote: > I am not sure what the distinction even mean for a class device? > > There's 1 device per registration, so they're equal. Am I missing something?
The existing DRM device private data's lifetime is tied to the lifetime of the DRM device itself, which makes sense as long as IOCTLs are not guarded against driver unbind and can still reach into the driver after driver unbind. Once that is changed, this doesn't make sense anymore, the driver structures are then tied to the lifetime of the Registration. Any handles that userspace may keep open beyond this should be handled by the subsystem; at this point there is no more HW state the driver would need to take care of. Thus, the private data tied to the lifetime of the DRM device won't be needed anymore, see also [1]. [1] https://lore.kernel.org/dri-devel/[email protected]/
