Question: Does drm-backend requires vblank interrupts to be enabled in drm driver all the time?

Dear weston experts,

Recently I observed an issue when I tried to run weston (with desktop shell, FADE-IN animation on) on drm-backend, I found weston didn't get the right vblank timestamp from kernel, which makes the weston_spring_done never achieved, and the animation layer can't be destroyed.

But based on my investigation, the most possibly touched path of updating the timestamp is for user to call drmWaitVBlank, which will call drm_vblank_get and enable vblank. However, in drm-backend, I found drmWaitVBlank would be skipped on first frame which triggered by drm_output_start_repaint_loop.

So I would like to know if weston expects kernel to support vblank interrupts all the time (which could be fulfilled by setting drm_vblank_offdelay to 0, refers to https://elixir.bootlin.com/linux/v6.4-rc6/source/drivers/gpu/drm/drm_vblank.c#L1508), or there supposed to be some other ways to trigger the first vblank enablement?

Thanks in advance!