09:28karolherbst: uhm...
09:28karolherbst: imirkin: I think we waste a lot of memory in our mm_slab code ...
09:29karolherbst: nouveau_mm_allocate called with size == 0x200000, but we allocate a 0x400000 big slab
09:34karolherbst: ehh wait, no that's fine
09:34karolherbst: the code is just confusing
10:00karolherbst: *sigh*, two threads mapping the same bo
10:00karolherbst: lol
10:01HdkR: Seems fine to me :)
10:01karolherbst: not if they race on the VM address
10:01karolherbst: race around drm_mmap :/
10:02HdkR: Time for a mutex?
10:02karolherbst: not at this level
10:02karolherbst: I am actually wondering why we even need nouveau_bo_map...
10:02karolherbst: because.. I think all bos are already mapped anyway
10:05karolherbst: huh actually.. I can lock a different lock earlier
16:19fling: No crases yet!
16:19fling: lua hack in awesome helped
16:19fling: llvmpipe rocks!