Dave Airlie skrev:
> From: Dave Airlie <[email protected]>
>
> This adds color tiling support for buffers in VRAM, it enables
> a color tiled fbcon and a color tiled X frontbuffer.
>
> It changes the API:
> adds two new parameters to the object creation API (is this better than
> a set/get tiling?) we probably still need a get but not sure for what yet.
> relocs are required for 2D DST_PITCH_OFFSET and SRC_PITCH_OFFSET type-0,
> and 3D COLORPITCH registers.
>
> TTM:
> adds a new check_tiling call to TTM, gets called at fault and around
> bo moves.
>
>
Dave,
Since I'm working on a driver with driver-private GPU maps, I wonder
whether we could use some more generic callbacks here. Also we should
protect them with a NULL check.
> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
> index c1c407f..e440b8b 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
> @@ -304,6 +304,9 @@ static int ttm_bo_handle_move_mem(struct
> ttm_buffer_object *bo,
>
> }
>
> + /* drop tiling around bo move */
> + bdev->driver->check_tiling(bo, 0, 1);
> +
> if (!(old_man->flags & TTM_MEMTYPE_FLAG_FIXED) &&
> !(new_man->flags & TTM_MEMTYPE_FLAG_FIXED))
> ret = ttm_bo_move_ttm(bo, evict, no_wait, mem);
>
What about naming the callback
bdev::driver::move_notify(bo, mem);
The driver can then compare bo::mem and mem to see any action is needed.
In my case it would take down the driver-private GPU maps. In your case
it would take down tiling.
> diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c b/drivers/gpu/drm/ttm/ttm_bo_vm.c
> index 27b146c..c10d96c 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo_vm.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c
> @@ -154,6 +154,7 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma,
> struct vm_fault *vmf)
> */
>
> if (is_iomem) {
> + bdev->driver->check_tiling(bo, 0, 0);
> vma->vm_page_prot = ttm_io_prot(bo->mem.placement,
> vma->vm_page_prot);
> } else {
> diff --git a/include/drm/radeon_drm.h b/include/drm/radeon_drm.h
> index 41862e9..1ed43c7 100644
>
>
A driver may want to move a buffer in the fault handler or set up, for
example, a VRAM aperture to simulate that the buffer is always mappable.
That would happen just after bo::reserve. Could we move this callback to
just after bo::reserve and call it bo_driver::fault_notify_reserved or
something like that?
One important to remember with callbacks from the fault handler is that
they will be subject to the infamous mmap_sem locking inversion, So if
there are any driver-private locks within check_tiling their locking
order will be _after_ mmap sem, and _after_ bo::reserve.
> --- a/include/drm/ttm/ttm_bo_driver.h
> +++ b/include/drm/ttm/ttm_bo_driver.h
> @@ -353,6 +353,14 @@ struct ttm_bo_driver {
> int (*sync_obj_flush) (void *sync_obj, void *sync_arg);
> void (*sync_obj_unref) (void **sync_obj);
> void *(*sync_obj_ref) (void *sync_obj);
> +
> + /* hook to check tiling registers are correct
> + * @bo : Pointer to a buffer object
> + * @has_moved : buffer has just been moved
> + * @force_drop: drop the surface
> + */
> + int (*check_tiling)(struct ttm_buffer_object *bo,
> + bool has_moved, bool force_drop);
> };
>
> #define TTM_NUM_MEM_TYPES 8
>
Thanks,
Thomas
------------------------------------------------------------------------------
--
_______________________________________________
Dri-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dri-devel
