Merged requested to merge zzyiwei/virglrenderer:gbm-alloc into master
We scope down the prior
FORCE_ENABLE_DMABUF given the workaround is only required for host visible mem type, and also sanity check the externam memory export feature before force the export path.
Only non-import path is affected by this fallback:
- drop the export info since the fd type can only be dma_buf here
- allocate gbm bo and export a fd for import as external memory
- track the gbm bo in vkr_device_memory
- vkr_context_get_blob exports fd from the bo and we destroy bo here
- vkr_device_memory_release ensures the gbm bo gets destroyed
- 1st commit is tiny fix on the sign of error return codes.
- 2-4 commits refactor so that the special object releases are synced. The one for device memory will be enriched later thus not made inline.
- 5-7 commits refactor so that fd export can be relocated inside vkr_device_memory with a clean interface.
- 8th commit scopes down the prior hack to just host visible memory types and physical device supporting memory export.
- 9th commit let vkr_physical_device track the gbm device for later gbm bo allocation and fd export.
- 10th commit add the gbm bo import fallback path with necessary docs.
- dEQP-VK.memory.* (pass with mesa/mesa!13874 (merged))
No perf regressions on anv, radv since they will stay on the force dma_buf path. As tested before, there's also no perf regression even if forcing them to use the import path.