On Sat, Nov 14, 2026 at 02:24:32AM +0000, [email protected] wrote: > Can allowing vrr.enable on joined pipes trigger a NULL pointer dereference > in intel_crtc_needs_joined_update() [...]
Yes. I can't reproduce it, no third pipe can be lit on the i915 side of my machine. v2 will have + /* A DDB change can pull in one joined pipe on its own */ + if (!new_pipe_crtc_state) + return false; > When a joined pipe's DDB allocation changes while VRR is enabled, does > falling back to per-pipe updates in skl_commit_modeset_enables() break VRR > push sequencing? It delays, it does not break: without a push the secondary latches at the vmax vblank, one frame of up to 1/min refresh later. The cover letter lists DDB changes among the commits that keep the per pipe sequence.
