From 65270a6569710b42f5ab2073c1cc91fb90189eaa Mon Sep 17 00:00:00 2001 From: Ben Skeggs Date: Tue, 12 Aug 2014 14:30:52 +1000 Subject: drm/nouveau/core/mm: allow allocation to be confined to a specific slice of heap Signed-off-by: Ben Skeggs diff --git a/drivers/gpu/drm/nouveau/core/core/gpuobj.c b/drivers/gpu/drm/nouveau/core/core/gpuobj.c index 560b221..daee877 100644 --- a/drivers/gpu/drm/nouveau/core/core/gpuobj.c +++ b/drivers/gpu/drm/nouveau/core/core/gpuobj.c @@ -115,7 +115,7 @@ nouveau_gpuobj_create_(struct nouveau_object *parent, gpuobj->size = size; if (heap) { - ret = nouveau_mm_head(heap, 1, size, size, + ret = nouveau_mm_head(heap, 0, 1, size, size, max(align, (u32)1), &gpuobj->node); if (ret) return ret; diff --git a/drivers/gpu/drm/nouveau/core/core/mm.c b/drivers/gpu/drm/nouveau/core/core/mm.c index 02ce615..b4f5db6 100644 --- a/drivers/gpu/drm/nouveau/core/core/mm.c +++ b/drivers/gpu/drm/nouveau/core/core/mm.c @@ -98,6 +98,7 @@ region_head(struct nouveau_mm *mm, struct nouveau_mm_node *a, u32 size) b->offset = a->offset; b->length = size; + b->heap = a->heap; b->type = a->type; a->offset += size; a->length -= size; @@ -108,8 +109,8 @@ region_head(struct nouveau_mm *mm, struct nouveau_mm_node *a, u32 size) } int -nouveau_mm_head(struct nouveau_mm *mm, u8 type, u32 size_max, u32 size_min, - u32 align, struct nouveau_mm_node **pnode) +nouveau_mm_head(struct nouveau_mm *mm, u8 heap, u8 type, u32 size_max, + u32 size_min, u32 align, struct nouveau_mm_node **pnode) { struct nouveau_mm_node *prev, *this, *next; u32 mask = align - 1; @@ -119,6 +120,10 @@ nouveau_mm_head(struct nouveau_mm *mm, u8 type, u32 size_max, u32 size_min, BUG_ON(type == NVKM_MM_TYPE_NONE || type == NVKM_MM_TYPE_HOLE); list_for_each_entry(this, &mm->free, fl_entry) { + if (unlikely(heap != NVKM_MM_HEAP_ANY)) { + if (this->heap != heap) + continue; + } e = this->offset + this->length; s = this->offset; @@ -167,6 +172,7 @@ region_tail(struct nouveau_mm *mm, struct nouveau_mm_node *a, u32 size) a->length -= size; b->offset = a->offset + a->length; b->length = size; + b->heap = a->heap; b->type = a->type; list_add(&b->nl_entry, &a->nl_entry); @@ -176,8 +182,8 @@ region_tail(struct nouveau_mm *mm, struct nouveau_mm_node *a, u32 size) } int -nouveau_mm_tail(struct nouveau_mm *mm, u8 type, u32 size_max, u32 size_min, - u32 align, struct nouveau_mm_node **pnode) +nouveau_mm_tail(struct nouveau_mm *mm, u8 heap, u8 type, u32 size_max, + u32 size_min, u32 align, struct nouveau_mm_node **pnode) { struct nouveau_mm_node *prev, *this, *next; u32 mask = align - 1; @@ -188,6 +194,10 @@ nouveau_mm_tail(struct nouveau_mm *mm, u8 type, u32 size_max, u32 size_min, u32 e = this->offset + this->length; u32 s = this->offset; u32 c = 0, a; + if (unlikely(heap != NVKM_MM_HEAP_ANY)) { + if (this->heap != heap) + continue; + } prev = node(this, prev); if (prev && prev->type != type) @@ -262,7 +272,7 @@ nouveau_mm_init(struct nouveau_mm *mm, u32 offset, u32 length, u32 block) list_add_tail(&node->nl_entry, &mm->nodes); list_add_tail(&node->fl_entry, &mm->free); - mm->heap_nodes++; + node->heap = ++mm->heap_nodes; return 0; } diff --git a/drivers/gpu/drm/nouveau/core/include/core/mm.h b/drivers/gpu/drm/nouveau/core/include/core/mm.h index d4ef404..bfe6931 100644 --- a/drivers/gpu/drm/nouveau/core/include/core/mm.h +++ b/drivers/gpu/drm/nouveau/core/include/core/mm.h @@ -6,6 +6,8 @@ struct nouveau_mm_node { struct list_head fl_entry; struct list_head rl_entry; +#define NVKM_MM_HEAP_ANY 0x00 + u8 heap; #define NVKM_MM_TYPE_NONE 0x00 #define NVKM_MM_TYPE_HOLE 0xff u8 type; @@ -29,10 +31,10 @@ nouveau_mm_initialised(struct nouveau_mm *mm) int nouveau_mm_init(struct nouveau_mm *, u32 offset, u32 length, u32 block); int nouveau_mm_fini(struct nouveau_mm *); -int nouveau_mm_head(struct nouveau_mm *, u8 type, u32 size_max, u32 size_min, - u32 align, struct nouveau_mm_node **); -int nouveau_mm_tail(struct nouveau_mm *, u8 type, u32 size_max, u32 size_min, - u32 align, struct nouveau_mm_node **); +int nouveau_mm_head(struct nouveau_mm *, u8 heap, u8 type, u32 size_max, + u32 size_min, u32 align, struct nouveau_mm_node **); +int nouveau_mm_tail(struct nouveau_mm *, u8 heap, u8 type, u32 size_max, + u32 size_min, u32 align, struct nouveau_mm_node **); void nouveau_mm_free(struct nouveau_mm *, struct nouveau_mm_node **); #endif diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/nv20.c b/drivers/gpu/drm/nouveau/core/subdev/fb/nv20.c index f003c1b..2209ade 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/fb/nv20.c +++ b/drivers/gpu/drm/nouveau/core/subdev/fb/nv20.c @@ -45,7 +45,7 @@ nv20_fb_tile_comp(struct nouveau_fb *pfb, int i, u32 size, u32 flags, { u32 tiles = DIV_ROUND_UP(size, 0x40); u32 tags = round_up(tiles / pfb->ram->parts, 0x40); - if (!nouveau_mm_head(&pfb->tags, 1, tags, tags, 1, &tile->tag)) { + if (!nouveau_mm_head(&pfb->tags, 0, 1, tags, tags, 1, &tile->tag)) { if (!(flags & 2)) tile->zcomp = 0x00000000; /* Z16 */ else tile->zcomp = 0x04000000; /* Z24S8 */ tile->zcomp |= tile->tag->offset; diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/nv25.c b/drivers/gpu/drm/nouveau/core/subdev/fb/nv25.c index f34f422..e2a66c3 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/fb/nv25.c +++ b/drivers/gpu/drm/nouveau/core/subdev/fb/nv25.c @@ -32,7 +32,7 @@ nv25_fb_tile_comp(struct nouveau_fb *pfb, int i, u32 size, u32 flags, { u32 tiles = DIV_ROUND_UP(size, 0x40); u32 tags = round_up(tiles / pfb->ram->parts, 0x40); - if (!nouveau_mm_head(&pfb->tags, 1, tags, tags, 1, &tile->tag)) { + if (!nouveau_mm_head(&pfb->tags, 0, 1, tags, tags, 1, &tile->tag)) { if (!(flags & 2)) tile->zcomp = 0x00100000; /* Z16 */ else tile->zcomp = 0x00200000; /* Z24S8 */ tile->zcomp |= tile->tag->offset; diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/nv30.c b/drivers/gpu/drm/nouveau/core/subdev/fb/nv30.c index 69093f7..cbec402 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/fb/nv30.c +++ b/drivers/gpu/drm/nouveau/core/subdev/fb/nv30.c @@ -51,7 +51,7 @@ nv30_fb_tile_comp(struct nouveau_fb *pfb, int i, u32 size, u32 flags, { u32 tiles = DIV_ROUND_UP(size, 0x40); u32 tags = round_up(tiles / pfb->ram->parts, 0x40); - if (!nouveau_mm_head(&pfb->tags, 1, tags, tags, 1, &tile->tag)) { + if (!nouveau_mm_head(&pfb->tags, 0, 1, tags, tags, 1, &tile->tag)) { if (flags & 2) tile->zcomp |= 0x01000000; /* Z16 */ else tile->zcomp |= 0x02000000; /* Z24S8 */ tile->zcomp |= ((tile->tag->offset ) >> 6); diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/nv35.c b/drivers/gpu/drm/nouveau/core/subdev/fb/nv35.c index 161b06e..b2cf8c6 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/fb/nv35.c +++ b/drivers/gpu/drm/nouveau/core/subdev/fb/nv35.c @@ -32,7 +32,7 @@ nv35_fb_tile_comp(struct nouveau_fb *pfb, int i, u32 size, u32 flags, { u32 tiles = DIV_ROUND_UP(size, 0x40); u32 tags = round_up(tiles / pfb->ram->parts, 0x40); - if (!nouveau_mm_head(&pfb->tags, 1, tags, tags, 1, &tile->tag)) { + if (!nouveau_mm_head(&pfb->tags, 0, 1, tags, tags, 1, &tile->tag)) { if (flags & 2) tile->zcomp |= 0x04000000; /* Z16 */ else tile->zcomp |= 0x08000000; /* Z24S8 */ tile->zcomp |= ((tile->tag->offset ) >> 6); diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/nv36.c b/drivers/gpu/drm/nouveau/core/subdev/fb/nv36.c index 2dd3d0a..b4cdae2a 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/fb/nv36.c +++ b/drivers/gpu/drm/nouveau/core/subdev/fb/nv36.c @@ -32,7 +32,7 @@ nv36_fb_tile_comp(struct nouveau_fb *pfb, int i, u32 size, u32 flags, { u32 tiles = DIV_ROUND_UP(size, 0x40); u32 tags = round_up(tiles / pfb->ram->parts, 0x40); - if (!nouveau_mm_head(&pfb->tags, 1, tags, tags, 1, &tile->tag)) { + if (!nouveau_mm_head(&pfb->tags, 0, 1, tags, tags, 1, &tile->tag)) { if (flags & 2) tile->zcomp |= 0x10000000; /* Z16 */ else tile->zcomp |= 0x20000000; /* Z24S8 */ tile->zcomp |= ((tile->tag->offset ) >> 6); diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/nv40.c b/drivers/gpu/drm/nouveau/core/subdev/fb/nv40.c index 95a115a..5281425 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/fb/nv40.c +++ b/drivers/gpu/drm/nouveau/core/subdev/fb/nv40.c @@ -33,7 +33,7 @@ nv40_fb_tile_comp(struct nouveau_fb *pfb, int i, u32 size, u32 flags, u32 tiles = DIV_ROUND_UP(size, 0x80); u32 tags = round_up(tiles / pfb->ram->parts, 0x100); if ( (flags & 2) && - !nouveau_mm_head(&pfb->tags, 1, tags, tags, 1, &tile->tag)) { + !nouveau_mm_head(&pfb->tags, 0, 1, tags, tags, 1, &tile->tag)) { tile->zcomp = 0x28000000; /* Z24S8_SPLIT_GRAD */ tile->zcomp |= ((tile->tag->offset ) >> 8); tile->zcomp |= ((tile->tag->offset + tags - 1) >> 8) << 13; diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/ramnv50.c b/drivers/gpu/drm/nouveau/core/subdev/fb/ramnv50.c index e5d12c2..514b078 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/fb/ramnv50.c +++ b/drivers/gpu/drm/nouveau/core/subdev/fb/ramnv50.c @@ -280,7 +280,7 @@ nv50_ram_get(struct nouveau_fb *pfb, u64 size, u32 align, u32 ncmin, if (align == 16) { int n = (max >> 4) * comp; - ret = nouveau_mm_head(tags, 1, n, n, 1, &mem->tag); + ret = nouveau_mm_head(tags, 0, 1, n, n, 1, &mem->tag); if (ret) mem->tag = NULL; } @@ -296,9 +296,9 @@ nv50_ram_get(struct nouveau_fb *pfb, u64 size, u32 align, u32 ncmin, type = nv50_fb_memtype[type]; do { if (back) - ret = nouveau_mm_tail(heap, type, max, min, align, &r); + ret = nouveau_mm_tail(heap, 0, type, max, min, align, &r); else - ret = nouveau_mm_head(heap, type, max, min, align, &r); + ret = nouveau_mm_head(heap, 0, type, max, min, align, &r); if (ret) { mutex_unlock(&pfb->base.mutex); pfb->ram->put(pfb, &mem); diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/ramnvc0.c b/drivers/gpu/drm/nouveau/core/subdev/fb/ramnvc0.c index 8007f61..fc9de88 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/fb/ramnvc0.c +++ b/drivers/gpu/drm/nouveau/core/subdev/fb/ramnvc0.c @@ -483,9 +483,9 @@ nvc0_ram_get(struct nouveau_fb *pfb, u64 size, u32 align, u32 ncmin, do { if (back) - ret = nouveau_mm_tail(mm, 1, size, ncmin, align, &r); + ret = nouveau_mm_tail(mm, 0, 1, size, ncmin, align, &r); else - ret = nouveau_mm_head(mm, 1, size, ncmin, align, &r); + ret = nouveau_mm_head(mm, 0, 1, size, ncmin, align, &r); if (ret) { mutex_unlock(&pfb->base.mutex); pfb->ram->put(pfb, &mem); diff --git a/drivers/gpu/drm/nouveau/core/subdev/instmem/nv04.c b/drivers/gpu/drm/nouveau/core/subdev/instmem/nv04.c index 7b64bef..e8b1401 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/instmem/nv04.c +++ b/drivers/gpu/drm/nouveau/core/subdev/instmem/nv04.c @@ -69,7 +69,7 @@ nv04_instobj_ctor(struct nouveau_object *parent, struct nouveau_object *engine, if (ret) return ret; - ret = nouveau_mm_head(&priv->heap, 1, args->size, args->size, + ret = nouveau_mm_head(&priv->heap, 0, 1, args->size, args->size, args->align, &node->mem); if (ret) return ret; diff --git a/drivers/gpu/drm/nouveau/core/subdev/ltc/base.c b/drivers/gpu/drm/nouveau/core/subdev/ltc/base.c index 32ed442..7fa3315 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/ltc/base.c +++ b/drivers/gpu/drm/nouveau/core/subdev/ltc/base.c @@ -31,7 +31,7 @@ nvkm_ltc_tags_alloc(struct nouveau_ltc *ltc, u32 n, struct nvkm_ltc_priv *priv = (void *)ltc; int ret; - ret = nouveau_mm_head(&priv->tags, 1, n, n, 1, pnode); + ret = nouveau_mm_head(&priv->tags, 0, 1, n, n, 1, pnode); if (ret) *pnode = NULL; diff --git a/drivers/gpu/drm/nouveau/core/subdev/ltc/gf100.c b/drivers/gpu/drm/nouveau/core/subdev/ltc/gf100.c index 7f9dd2b..8857d41 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/ltc/gf100.c +++ b/drivers/gpu/drm/nouveau/core/subdev/ltc/gf100.c @@ -168,7 +168,7 @@ gf100_ltc_init_tag_ram(struct nouveau_fb *pfb, struct nvkm_ltc_priv *priv) tag_size += tag_align; tag_size = (tag_size + 0xfff) >> 12; /* round up */ - ret = nouveau_mm_tail(&pfb->vram, 1, tag_size, tag_size, 1, + ret = nouveau_mm_tail(&pfb->vram, 0, 1, tag_size, tag_size, 1, &priv->tag_ram); if (ret) { priv->num_tags = 0; diff --git a/drivers/gpu/drm/nouveau/core/subdev/vm/base.c b/drivers/gpu/drm/nouveau/core/subdev/vm/base.c index 7dd680f..f75a683 100644 --- a/drivers/gpu/drm/nouveau/core/subdev/vm/base.c +++ b/drivers/gpu/drm/nouveau/core/subdev/vm/base.c @@ -296,7 +296,7 @@ nouveau_vm_get(struct nouveau_vm *vm, u64 size, u32 page_shift, int ret; mutex_lock(&nv_subdev(vmm)->mutex); - ret = nouveau_mm_head(&vm->mm, page_shift, msize, msize, align, + ret = nouveau_mm_head(&vm->mm, 0, page_shift, msize, msize, align, &vma->node); if (unlikely(ret != 0)) { mutex_unlock(&nv_subdev(vmm)->mutex); diff --git a/drivers/gpu/drm/nouveau/nouveau_abi16.c b/drivers/gpu/drm/nouveau/nouveau_abi16.c index 615714c..a24faa5 100644 --- a/drivers/gpu/drm/nouveau/nouveau_abi16.c +++ b/drivers/gpu/drm/nouveau/nouveau_abi16.c @@ -448,7 +448,7 @@ nouveau_abi16_ioctl_notifierobj_alloc(ABI16_IOCTL_ARGS) list_add(&ntfy->head, &chan->notifiers); ntfy->handle = info->handle; - ret = nouveau_mm_head(&chan->heap, 1, info->size, info->size, 1, + ret = nouveau_mm_head(&chan->heap, 0, 1, info->size, info->size, 1, &ntfy->node); if (ret) goto done; -- cgit v0.10.2