On Wed, May 15, 2013 at 4:06 PM, Rob Clark <robdclark@xxxxxxxxx> wrote: > So while it seems nice and orthogonal/clean to couple cache and > synchronization and handle dma->cpu and cpu->cpu and cpu->dma in the > same generic way, but I think in practice we have to make things more > complex than they otherwise need to be to do this. Otherwise I think > we'll be having problems with badly behaved or crashing userspace. I haven't read through the entire thread careful but imo this is very important. If we add a fence interface which allows userspace to block dma this is a no-go. The only thing we need is to sync up with all outstanding dma operations and flush caches for cpu access. If broken userspace starts to issue new dma (or multiple thread stomp onto each another) that's not a problem dma fences/syncpoints should try to solve. This way we can concentrate on solving the (already challenging) device-to-device sync issues without additional complexities which cpu->cpu sync would impose. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/dri-devel
