Am 24.09.20 um 07:18 schrieb Dave Airlie:
From: Dave Airlie airlied@redhat.com
Signed-off-by: Dave Airlie airlied@redhat.com
NAK, completely unnecessary.
We should rather do the remaining accounting in the already existing release_notify() callback.
That makes much more sense and if I'm not completely mistaken could actually fix a bug in amdgpu.
Christian.
drivers/gpu/drm/ttm/ttm_bo.c | 4 +++- include/drm/ttm/ttm_bo_driver.h | 7 +++++++ 2 files changed, 10 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index a2a61a8d1394..ba69c682e53b 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -282,7 +282,9 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
static void ttm_bo_cleanup_memtype_use(struct ttm_buffer_object *bo) {
- if (bo->bdev->driver->move_notify)
if (bo->bdev->driver->invalidate_notify)
bo->bdev->driver->invalidate_notify(bo);else if (bo->bdev->driver->move_notify) bo->bdev->driver->move_notify(bo, false, NULL);
ttm_bo_tt_destroy(bo);
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h index cfb151dbb2d0..da4afe669664 100644 --- a/include/drm/ttm/ttm_bo_driver.h +++ b/include/drm/ttm/ttm_bo_driver.h @@ -165,6 +165,13 @@ struct ttm_bo_driver { void (*move_notify)(struct ttm_buffer_object *bo, bool evict, struct ttm_resource *new_mem);
- /**
* Hook to notify driver about a bo being torn down.* can be used for invalidation instead of move_notify.*/- void (*invalidate_notify)(struct ttm_buffer_object *bo);
- /* notify the driver we are taking a fault on this BO
int (*fault_reserve_notify)(struct ttm_buffer_object *bo);
- and have reserved it */