Hi
See patch 4/4 for a more detailed explanation of this series. I basically kill off the whole drm_mm pre-alloc code as it really doesn't make any sense with todays infrastructure. No drm_mm user runs in atomic context. We use pre-alloc only to allow allocation while holding a spin-lock. But we can easily kzalloc() the node before taking the spinlock and use the drm_mm_insert_*() helpers directly.
This series converts the last pre-alloc users (ttm and i915-gem-stolen) to use the already established kzalloc()+drm_mm_insert_*() helpers.
The last patch removes a bunch of old drm_mm code, so Daniel can tackle his "drm_mm documentation" TODO list.
Cheers David
David Herrmann (4): drm/mm: add "best_match" to drm_mm_insert_node() drm/ttm: replace drm_mm_pre_get() by direct alloc drm/i915: pre-alloc instead of drm_mm search/get_block drm/mm: remove unused API
drivers/gpu/drm/drm_mm.c | 183 +++++---------------------------- drivers/gpu/drm/drm_vma_manager.c | 4 +- drivers/gpu/drm/i915/i915_gem.c | 3 +- drivers/gpu/drm/i915/i915_gem_stolen.c | 72 ++++++++----- drivers/gpu/drm/sis/sis_mm.c | 4 +- drivers/gpu/drm/ttm/ttm_bo_manager.c | 40 ++++--- drivers/gpu/drm/via/via_mm.c | 4 +- include/drm/drm_mm.h | 131 +++++------------------ 8 files changed, 123 insertions(+), 318 deletions(-)