From 17c581d75514c32d1e8c1e416beb33b3ccce22a5 Mon Sep 17 00:00:00 2001 From: Sunny He Date: Tue, 15 Aug 2017 12:01:04 -0700 Subject: gpu: nvgpu: SGL passthrough implementation The basic nvgpu_mem_sgl implementation provides support for OS specific scatter-gather list implementations by simply copying them node by node. This is inefficient, taking extra time and memory. This patch implements an nvgpu_mem_sgt struct to act as a header which is inserted at the front of any scatter- gather list implementation. This labels every struct with a set of ops which can be used to interact with the attached scatter gather list. Since nvgpu common code only has to interact with these function pointers, any sgl implementation can be used. Initialization only requires the allocation of a single struct, removing the need to copy or iterate through the sgl being converted. Jira NVGPU-186 Change-Id: I2994f804a4a4cc141b702e987e9081d8560ba2e8 Signed-off-by: Sunny He Reviewed-on: https://git-master.nvidia.com/r/1541426 Reviewed-by: mobile promotions Tested-by: mobile promotions --- drivers/gpu/nvgpu/common/linux/vm.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'drivers/gpu/nvgpu/common/linux/vm.c') diff --git a/drivers/gpu/nvgpu/common/linux/vm.c b/drivers/gpu/nvgpu/common/linux/vm.c index 4a4429dc..2e29f0f7 100644 --- a/drivers/gpu/nvgpu/common/linux/vm.c +++ b/drivers/gpu/nvgpu/common/linux/vm.c @@ -69,19 +69,20 @@ static u64 nvgpu_get_buffer_alignment(struct gk20a *g, struct scatterlist *sgl, if (aperture == APERTURE_VIDMEM) { struct nvgpu_page_alloc *alloc = get_vidmem_page_alloc(sgl); - struct nvgpu_mem_sgl *sgl_vid = alloc->sgl; + struct nvgpu_sgt *sgt = &alloc->sgt; + void *sgl_vid = sgt->sgl; while (sgl_vid) { chunk_align = 1ULL << - __ffs(nvgpu_mem_sgl_phys(sgl_vid) | - nvgpu_mem_sgl_length(sgl_vid)); + __ffs(nvgpu_sgt_get_phys(sgt, sgl_vid)) | + nvgpu_sgt_get_length(sgt, sgl_vid); if (align) align = min(align, chunk_align); else align = chunk_align; - sgl_vid = nvgpu_mem_sgl_next(sgl_vid); + sgl_vid = nvgpu_sgt_get_next(sgt, sgl_vid); } return align; @@ -242,7 +243,7 @@ u64 nvgpu_vm_map(struct vm_gk20a *vm, struct nvgpu_vm_area *vm_area = NULL; u32 ctag_offset; enum nvgpu_aperture aperture; - struct nvgpu_mem_sgl *nvgpu_sgl; + struct nvgpu_sgt *nvgpu_sgt; /* * The kind used as part of the key for map caching. HW may @@ -399,12 +400,12 @@ u64 nvgpu_vm_map(struct vm_gk20a *vm, ctag_offset += buffer_offset >> ilog2(g->ops.fb.compression_page_size(g)); - nvgpu_sgl = nvgpu_mem_sgl_create(g, bfr.sgt); + nvgpu_sgt = nvgpu_linux_sgt_create(g, bfr.sgt); /* update gmmu ptes */ map_offset = g->ops.mm.gmmu_map(vm, map_offset, - nvgpu_sgl, + nvgpu_sgt, buffer_offset, /* sg offset */ mapping_size, bfr.pgsz_idx, @@ -419,7 +420,7 @@ u64 nvgpu_vm_map(struct vm_gk20a *vm, if (!map_offset) goto clean_up; - nvgpu_mem_sgl_free(g, nvgpu_sgl); + nvgpu_sgt_free(nvgpu_sgt, g); mapped_buffer = nvgpu_kzalloc(g, sizeof(*mapped_buffer)); if (!mapped_buffer) { -- cgit v1.2.2