On Mon 03 Apr 2017, Jason Ekstrand wrote: > On Mon, Apr 3, 2017 at 5:19 PM, Chad Versace <chadvers...@chromium.org> > wrote: > > > On Wed 15 Mar 2017, Jason Ekstrand wrote: > > > This cache allows us to easily ensure that we have a unique anv_bo for > > > each gem handle. We'll need this in order to support multiple-import of > > > memory objects and semaphores. > > > > > > v2 (Jason Ekstrand): > > > - Reject BO imports if the size doesn't match the prime fd size as > > > reported by lseek(). > > > > > > v3 (Jason Ekstrand): > > > - Fix reference counting around cache_release (Chris Willson) > > > - Move the mutex_unlock() later in cache_release > > > --- > > > src/intel/vulkan/anv_allocator.c | 261 ++++++++++++++++++++++++++++++ > > +++++++++ > > > src/intel/vulkan/anv_private.h | 26 ++++ > > > 2 files changed, 287 insertions(+) > > > > > > > > > +VkResult anv_bo_cache_alloc(struct anv_device *device, > > > + struct anv_bo_cache *cache, > > > + uint64_t size, struct anv_bo **bo, > > > + VkAllocationCallbacks *alloc); > > > > > +VkResult anv_bo_cache_import(struct anv_device *device, > > > + struct anv_bo_cache *cache, > > > + int fd, uint64_t size, struct anv_bo **bo, > > > + VkAllocationCallbacks *alloc); > > > > > +void anv_bo_cache_release(struct anv_device *device, > > > + struct anv_bo_cache *cache, > > > + struct anv_bo *bo, > > > + VkAllocationCallbacks *alloc); > > > + > > > > The app may do this: > > > > // fd1 != fd2 > > // fd1 and fd2 wrap the same gem handle > > > > mem1 = vkAllocateMemory(import fd1, pAllocator=alloc1); // anv_cached_bo > > is allocated here > > mem2 = vkAllocateMemory(import fd2, pAllocator=alloc2); > > > > ... > > > > vkFreeMemory(mem1, alloc1); > > vkFreeMemory(mem2, alloc2); // anv_cached_bo is freed here > > > > So we can't use either pAllocator to allocate/free the anv_cached_bo, > > unless anv_cached_bo also cached its original allocator. > > If we don't cache the allocator (and please, let's not), then the > > 'alloc' params to anv_bo_cache_alloc/import/release are dead params. > > We must use an allocator that is not bound > > to the anv_cached_bo, such as device->alloc. > > > > > > Patch 8/18 correctly passes device->alloc into these functions. But, > > since the 'alloc' params aren't really used, they should be dropped to > > prevent accidental misuse. > > > > I'm not quite sure what you're asking for here. Do you want me to just use > device->alloc and drop the parameter? I'm happy to do that.
Yeah, I meant that. _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
