On Wed, 12 Nov 2025 15:12:53 +0100 Boris Brezillon <[email protected]> wrote:
> > OK. timeout_wq sharing is intended and allowed, so if that's what > > you're doing, good. But I agree that accessing the wq through the > > driver's struct is then cleaner and more obviously correct. > > Will do. Actually, after looking into that, it seems I'd have to go back to dev_get_drvdata(drm_gpu_scheduler::dev)::reset_wq to get the timeout workqueue, so I think I'd prefer to keep using drm_gpu_scheduler::timeout_wq unless that's really seen as a layering violation (but I'd argue that accessing drm_gpu_scheduler::dev looks like the same kind of layering violation to me).
