On Tue, Aug 5, 2014 at 7:32 PM, Christian König deathsimple@vodafone.de wrote:
Christian wrote some patches to validate the interfaces, but I'm not sure he ever sent them out. We haven't yet done a full implementation in the usermode drivers to take advantage of this yet.
Well right now I've consistently rejected all patches that don't yet come with the full thing (libdrm, usermode drivers and tests for it all as not ready). And I do that at least once per week since we have blob userspace separate from mesa, too. So if we toss that rule overboard (and my understanding is that Dave's been fairly strict here) I'll look rather bad. As in really, really bad.
I strongly prefer that userptr gets postponed until it's ready. Dave?
That's just my fault. Wanted to wait with the mesa patches till we have the kernel interface accepted.
I've just send them out,
Ah, seen them now, thanks. Personally I prefer if kernel and usermode parts get reviewed in parallel. But that's mostly since we have 2 separate teams. So no objection to userptr from my side any more ;-)
Cheers, Daniel