On Mon, Jan 11, 2016 at 10:40:54PM +0100, Daniel Vetter wrote: > Hi all, > > Mostly just small changes from review feedback (plus a few misplaced hunks, > silly me). Plus an attempt at better kerneldoc to explain how this works. Since > that caused questions both from Thomas and Laurent let me explain things also > here: > > Currently anyone using drm_events (vblank code, atomic/legacy flips, drivers > like exynos) need to handle event setup, sending to the drm file and handling > when the drm file disappears themselves. This patch series provides functions to > encapsulate all these steps. Furthermore the file closing is handled > differently: Events are no longer completely destroyed in preclose hooks like > all current users do, but instead they're just unlinked from the file. The new > drm_send_event can still be called, but instead of transmitting the event to the > file it'll destroy it right away. > > This means that when the drm file disappears it's entirely transparent to > drivers, and we can remove all the special tracking and cleanup code around > preclose functions. > > The other consequence is that it's opt-in - drivers are still free to remove > events explicitly when the file disappears, it's just needless code. Which means > the patch series can be split up really nicely into core parts + driver patches. > The only patch I didn't split up in this fashion is the one that creates > drm_send_event - it's imo easier to review this way, and the driver changes are > really simple. -rc1 is out, so I pulled this all into drm-misc. Thanks a lot for all the reviews&comments. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
