Newer
Older
OUT_RING (chan, height);
OUT_RING (chan, 0x00000101);
OUT_RING (chan, 0x00000000);
BEGIN_NV04(chan, NvSubCopy, NV_MEMORY_TO_MEMORY_FORMAT_NOP, 1);
OUT_RING (chan, 0);
length -= amount;
src_offset += amount;
dst_offset += amount;
static int
nv04_bo_move_init(struct nouveau_channel *chan, u32 handle)
{
int ret = RING_SPACE(chan, 4);
if (ret == 0) {
BEGIN_NV04(chan, NvSubCopy, 0x0000, 1);
OUT_RING (chan, handle);
BEGIN_NV04(chan, NvSubCopy, 0x0180, 1);
OUT_RING (chan, chan->drm->ntfy.handle);
}
return ret;
}
static inline uint32_t
nouveau_bo_mem_ctxdma(struct ttm_buffer_object *bo,
struct nouveau_channel *chan, struct ttm_mem_reg *reg)
if (reg->mem_type == TTM_PL_TT)
return chan->vram.handle;
static int
nv04_bo_move_m2mf(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
struct ttm_mem_reg *old_reg, struct ttm_mem_reg *new_reg)
u32 src_offset = old_reg->start << PAGE_SHIFT;
u32 dst_offset = new_reg->start << PAGE_SHIFT;
u32 page_count = new_reg->num_pages;
int ret;
ret = RING_SPACE(chan, 3);
if (ret)
return ret;
BEGIN_NV04(chan, NvSubCopy, NV_MEMORY_TO_MEMORY_FORMAT_DMA_SOURCE, 2);
OUT_RING (chan, nouveau_bo_mem_ctxdma(bo, chan, old_reg));
OUT_RING (chan, nouveau_bo_mem_ctxdma(bo, chan, new_reg));
page_count = new_reg->num_pages;
while (page_count) {
int line_count = (page_count > 2047) ? 2047 : page_count;
ret = RING_SPACE(chan, 11);
if (ret)
return ret;
BEGIN_NV04(chan, NvSubCopy,
NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8);
OUT_RING (chan, src_offset);
OUT_RING (chan, dst_offset);
OUT_RING (chan, PAGE_SIZE); /* src_pitch */
OUT_RING (chan, PAGE_SIZE); /* dst_pitch */
OUT_RING (chan, PAGE_SIZE); /* line_length */
OUT_RING (chan, line_count);
OUT_RING (chan, 0x00000101);
OUT_RING (chan, 0x00000000);
BEGIN_NV04(chan, NvSubCopy, NV_MEMORY_TO_MEMORY_FORMAT_NOP, 1);
page_count -= line_count;
src_offset += (PAGE_SIZE * line_count);
dst_offset += (PAGE_SIZE * line_count);
}
nouveau_bo_move_prep(struct nouveau_drm *drm, struct ttm_buffer_object *bo,
struct ttm_mem_reg *reg)
struct nouveau_mem *old_mem = nouveau_mem(&bo->mem);
struct nouveau_mem *new_mem = nouveau_mem(reg);
struct nvif_vmm *vmm = &drm->client.vmm.vmm;
int ret;
ret = nvif_vmm_get(vmm, LAZY, false, old_mem->mem.page, 0,
old_mem->mem.size, &old_mem->vma[0]);
if (ret)
return ret;
ret = nvif_vmm_get(vmm, LAZY, false, new_mem->mem.page, 0,
new_mem->mem.size, &old_mem->vma[1]);
if (ret)
goto done;
ret = nouveau_mem_map(old_mem, vmm, &old_mem->vma[0]);
if (ret)
goto done;
ret = nouveau_mem_map(new_mem, vmm, &old_mem->vma[1]);
done:
if (ret) {
nvif_vmm_put(vmm, &old_mem->vma[1]);
nvif_vmm_put(vmm, &old_mem->vma[0]);
return 0;
}
static int
nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, bool intr,
bool no_wait_gpu, struct ttm_mem_reg *new_reg)
struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
struct nouveau_channel *chan = drm->ttm.chan;
struct nouveau_cli *cli = (void *)chan->user.client;
struct nouveau_fence *fence;
/* create temporary vmas for the transfer and attach them to the
* old nvkm_mem node, these will get cleaned up after ttm has
* destroyed the ttm_mem_reg
if (drm->client.device.info.family >= NV_DEVICE_INFO_V0_TESLA) {
ret = nouveau_bo_move_prep(drm, bo, new_reg);
return ret;
mutex_lock_nested(&cli->mutex, SINGLE_DEPTH_NESTING);

Maarten Lankhorst
committed
ret = nouveau_fence_sync(nouveau_bo(bo), chan, true, intr);
if (ret == 0) {
ret = drm->ttm.move(chan, bo, &bo->mem, new_reg);
if (ret == 0) {
ret = nouveau_fence_new(chan, false, &fence);
if (ret == 0) {
ret = ttm_bo_move_accel_cleanup(bo,
&fence->base,
evict,
nouveau_fence_unref(&fence);
}
}
mutex_unlock(&cli->mutex);
return ret;
nouveau_bo_move_init(struct nouveau_drm *drm)
static const struct _method_table {
const char *name;
s32 oclass;
int (*exec)(struct nouveau_channel *,
struct ttm_buffer_object *,
struct ttm_mem_reg *, struct ttm_mem_reg *);
int (*init)(struct nouveau_channel *, u32 handle);
} _methods[] = {
{ "COPY", 4, 0xc5b5, nve0_bo_move_copy, nve0_bo_move_init },
{ "GRCE", 0, 0xc5b5, nve0_bo_move_copy, nvc0_bo_move_init },
{ "COPY", 4, 0xc3b5, nve0_bo_move_copy, nve0_bo_move_init },
{ "GRCE", 0, 0xc3b5, nve0_bo_move_copy, nvc0_bo_move_init },
{ "COPY", 4, 0xc1b5, nve0_bo_move_copy, nve0_bo_move_init },
{ "GRCE", 0, 0xc1b5, nve0_bo_move_copy, nvc0_bo_move_init },
{ "COPY", 4, 0xc0b5, nve0_bo_move_copy, nve0_bo_move_init },
{ "GRCE", 0, 0xc0b5, nve0_bo_move_copy, nvc0_bo_move_init },
{ "COPY", 4, 0xb0b5, nve0_bo_move_copy, nve0_bo_move_init },
{ "GRCE", 0, 0xb0b5, nve0_bo_move_copy, nvc0_bo_move_init },
{ "COPY", 4, 0xa0b5, nve0_bo_move_copy, nve0_bo_move_init },
{ "GRCE", 0, 0xa0b5, nve0_bo_move_copy, nvc0_bo_move_init },
{ "COPY1", 5, 0x90b8, nvc0_bo_move_copy, nvc0_bo_move_init },
{ "COPY0", 4, 0x90b5, nvc0_bo_move_copy, nvc0_bo_move_init },
{ "COPY", 0, 0x85b5, nva3_bo_move_copy, nv50_bo_move_init },
{ "CRYPT", 0, 0x74c1, nv84_bo_move_exec, nv50_bo_move_init },
{ "M2MF", 0, 0x9039, nvc0_bo_move_m2mf, nvc0_bo_move_init },
{ "M2MF", 0, 0x5039, nv50_bo_move_m2mf, nv50_bo_move_init },
{ "M2MF", 0, 0x0039, nv04_bo_move_m2mf, nv04_bo_move_init },
{ "CRYPT", 0, 0x88b4, nv98_bo_move_exec, nv50_bo_move_init },
};
const struct _method_table *mthd = _methods;
const char *name = "CPU";
int ret;
do {
struct nouveau_channel *chan;
if (mthd->engine)
chan = drm->cechan;
else
chan = drm->channel;
if (chan == NULL)
continue;
ret = nvif_object_init(&chan->user,
mthd->oclass | (mthd->engine << 16),
mthd->oclass, NULL, 0,
&drm->ttm.copy);
if (ret == 0) {
ret = mthd->init(chan, drm->ttm.copy.handle);
nvif_object_fini(&drm->ttm.copy);
drm->ttm.move = mthd->exec;
drm->ttm.chan = chan;
name = mthd->name;
break;
}
} while ((++mthd)->exec);
NV_INFO(drm, "MM: using %s for buffer copies\n", name);
static int
nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr,
bool no_wait_gpu, struct ttm_mem_reg *new_reg)
struct ttm_operation_ctx ctx = { intr, no_wait_gpu };
struct ttm_place placement_memtype = {
.fpfn = 0,
.lpfn = 0,
.flags = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING
};
struct ttm_placement placement;
struct ttm_mem_reg tmp_reg;
int ret;
placement.num_placement = placement.num_busy_placement = 1;

Francisco Jerez
committed
placement.placement = placement.busy_placement = &placement_memtype;
tmp_reg = *new_reg;
tmp_reg.mm_node = NULL;
ret = ttm_bo_mem_space(bo, &placement, &tmp_reg, &ctx);
if (ret)
return ret;
ret = ttm_tt_bind(bo->ttm, &tmp_reg, &ctx);
ret = nouveau_bo_move_m2mf(bo, true, intr, no_wait_gpu, &tmp_reg);
ret = ttm_bo_move_ttm(bo, &ctx, new_reg);
ttm_bo_mem_put(bo, &tmp_reg);
return ret;
}
static int
nouveau_bo_move_flips(struct ttm_buffer_object *bo, bool evict, bool intr,
bool no_wait_gpu, struct ttm_mem_reg *new_reg)
struct ttm_operation_ctx ctx = { intr, no_wait_gpu };
struct ttm_place placement_memtype = {
.fpfn = 0,
.lpfn = 0,
.flags = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING
};
struct ttm_placement placement;
struct ttm_mem_reg tmp_reg;
int ret;
placement.num_placement = placement.num_busy_placement = 1;

Francisco Jerez
committed
placement.placement = placement.busy_placement = &placement_memtype;
tmp_reg = *new_reg;
tmp_reg.mm_node = NULL;
ret = ttm_bo_mem_space(bo, &placement, &tmp_reg, &ctx);
if (ret)
return ret;
ret = ttm_bo_move_ttm(bo, &ctx, &tmp_reg);
ret = nouveau_bo_move_m2mf(bo, true, intr, no_wait_gpu, new_reg);
if (ret)
goto out;
out:
ttm_bo_mem_put(bo, &tmp_reg);
nouveau_bo_move_ntfy(struct ttm_buffer_object *bo, bool evict,
struct ttm_mem_reg *new_reg)
struct nouveau_mem *mem = new_reg ? nouveau_mem(new_reg) : NULL;
struct nouveau_bo *nvbo = nouveau_bo(bo);
struct nouveau_vma *vma;
/* ttm can now (stupidly) pass the driver bos it didn't create... */
if (bo->destroy != nouveau_bo_del_ttm)
return;
if (mem && new_reg->mem_type != TTM_PL_SYSTEM &&
mem->mem.page == nvbo->page) {
list_for_each_entry(vma, &nvbo->vma_list, head) {
nouveau_vma_map(vma, mem);
}
} else {
list_for_each_entry(vma, &nvbo->vma_list, head) {
WARN_ON(ttm_bo_wait(bo, false, false));
nouveau_vma_unmap(vma);
nouveau_bo_vm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_reg,
struct nouveau_drm_tile **new_tile)
struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
struct drm_device *dev = drm->dev;
struct nouveau_bo *nvbo = nouveau_bo(bo);
u64 offset = new_reg->start << PAGE_SHIFT;
if (new_reg->mem_type != TTM_PL_VRAM)
if (drm->client.device.info.family >= NV_DEVICE_INFO_V0_CELSIUS) {
*new_tile = nv10_bo_set_tiling(dev, offset, new_reg->size,
nvbo->mode, nvbo->zeta);
return 0;
}
static void
nouveau_bo_vm_cleanup(struct ttm_buffer_object *bo,
struct nouveau_drm_tile *new_tile,
struct nouveau_drm_tile **old_tile)
struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
struct drm_device *dev = drm->dev;
struct dma_fence *fence = dma_resv_get_excl(bo->base.resv);
nv10_bo_put_tile_region(dev, *old_tile, fence);
*old_tile = new_tile;
}
static int
nouveau_bo_move(struct ttm_buffer_object *bo, bool evict,
struct ttm_operation_ctx *ctx,
struct ttm_mem_reg *new_reg)
struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
struct nouveau_bo *nvbo = nouveau_bo(bo);
struct ttm_mem_reg *old_reg = &bo->mem;
struct nouveau_drm_tile *new_tile = NULL;
ret = ttm_bo_wait(bo, ctx->interruptible, ctx->no_wait_gpu);
if (ret)
return ret;
if (nvbo->pin_refcnt)
NV_WARN(drm, "Moving pinned object %p!\n", nvbo);
if (drm->client.device.info.family < NV_DEVICE_INFO_V0_TESLA) {
ret = nouveau_bo_vm_bind(bo, new_reg, &new_tile);
if (ret)
return ret;
}
/* Fake bo copy. */
if (old_reg->mem_type == TTM_PL_SYSTEM && !bo->ttm) {
BUG_ON(bo->mem.mm_node != NULL);
bo->mem = *new_reg;
new_reg->mm_node = NULL;
/* Hardware assisted copy. */
if (drm->ttm.move) {
if (new_reg->mem_type == TTM_PL_SYSTEM)
ret = nouveau_bo_move_flipd(bo, evict,
ctx->interruptible,
ctx->no_wait_gpu, new_reg);
else if (old_reg->mem_type == TTM_PL_SYSTEM)
ret = nouveau_bo_move_flips(bo, evict,
ctx->interruptible,
ctx->no_wait_gpu, new_reg);
else
ret = nouveau_bo_move_m2mf(bo, evict,
ctx->interruptible,
ctx->no_wait_gpu, new_reg);
if (!ret)
goto out;
}
/* Fallback to software copy. */
ret = ttm_bo_wait(bo, ctx->interruptible, ctx->no_wait_gpu);
if (ret == 0)
ret = ttm_bo_move_memcpy(bo, ctx, new_reg);
if (drm->client.device.info.family < NV_DEVICE_INFO_V0_TESLA) {
if (ret)
nouveau_bo_vm_cleanup(bo, NULL, &new_tile);
else
nouveau_bo_vm_cleanup(bo, new_tile, &nvbo->tile);
}
}
static int
nouveau_bo_verify_access(struct ttm_buffer_object *bo, struct file *filp)
{
struct nouveau_bo *nvbo = nouveau_bo(bo);
return drm_vma_node_verify_access(&nvbo->bo.base.vma_node,
nouveau_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_mem_reg *reg)
struct ttm_mem_type_manager *man = &bdev->man[reg->mem_type];
struct nouveau_drm *drm = nouveau_bdev(bdev);
struct nvkm_device *device = nvxx_device(&drm->client.device);
struct nouveau_mem *mem = nouveau_mem(reg);
reg->bus.addr = NULL;
reg->bus.offset = 0;
reg->bus.size = reg->num_pages << PAGE_SHIFT;
reg->bus.base = 0;
reg->bus.is_iomem = false;
if (!(man->flags & TTM_MEMTYPE_FLAG_MAPPABLE))
return -EINVAL;
switch (reg->mem_type) {
case TTM_PL_SYSTEM:
/* System memory */
return 0;
case TTM_PL_TT:
if (drm->agp.bridge) {
reg->bus.offset = reg->start << PAGE_SHIFT;
reg->bus.base = drm->agp.base;
reg->bus.is_iomem = !drm->agp.cma;
}
#endif
if (drm->client.mem->oclass < NVIF_CLASS_MEM_NV50 || !mem->kind)
/* untiled */
break;
/* fall through - tiled memory */
case TTM_PL_VRAM:
reg->bus.offset = reg->start << PAGE_SHIFT;
reg->bus.base = device->func->resource_addr(device, 1);
reg->bus.is_iomem = true;
if (drm->client.mem->oclass >= NVIF_CLASS_MEM_NV50) {
union {
struct nv50_mem_map_v0 nv50;
struct gf100_mem_map_v0 gf100;
} args;
u64 handle, length;
u32 argc = 0;
int ret;
switch (mem->mem.object.oclass) {
case NVIF_CLASS_MEM_NV50:
args.nv50.version = 0;
args.nv50.ro = 0;
args.nv50.kind = mem->kind;
args.nv50.comp = mem->comp;

Thierry Reding
committed
argc = sizeof(args.nv50);
break;
case NVIF_CLASS_MEM_GF100:
args.gf100.version = 0;
args.gf100.ro = 0;
args.gf100.kind = mem->kind;

Thierry Reding
committed
argc = sizeof(args.gf100);
break;
default:
WARN_ON(1);
break;
}
ret = nvif_object_map_handle(&mem->mem.object,

Thierry Reding
committed
&args, argc,
&handle, &length);
if (ret != 1) {
if (WARN_ON(ret == 0))
return -EINVAL;
if (ret == -ENOSPC)
return -EAGAIN;
return ret;
}
reg->bus.base = 0;
reg->bus.offset = handle;
break;
default:
return -EINVAL;
}
return 0;
}
static void
nouveau_ttm_io_mem_free(struct ttm_bo_device *bdev, struct ttm_mem_reg *reg)
struct nouveau_drm *drm = nouveau_bdev(bdev);
struct nouveau_mem *mem = nouveau_mem(reg);
if (drm->client.mem->oclass >= NVIF_CLASS_MEM_NV50) {
switch (reg->mem_type) {
case TTM_PL_TT:
if (mem->kind)
nvif_object_unmap_handle(&mem->mem.object);
break;
case TTM_PL_VRAM:
nvif_object_unmap_handle(&mem->mem.object);
break;
default:
break;
}
}
}
static int
nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
{
struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
struct nouveau_bo *nvbo = nouveau_bo(bo);
struct nvkm_device *device = nvxx_device(&drm->client.device);
u32 mappable = device->func->resource_size(device, 1) >> PAGE_SHIFT;
/* as long as the bo isn't in vram, and isn't tiled, we've got
* nothing to do here.
*/
if (bo->mem.mem_type != TTM_PL_VRAM) {
if (drm->client.device.info.family < NV_DEVICE_INFO_V0_TESLA ||
!nvbo->kind)
if (bo->mem.mem_type == TTM_PL_SYSTEM) {
nouveau_bo_placement_set(nvbo, TTM_PL_TT, 0);
ret = nouveau_bo_validate(nvbo, false, false);
if (ret)
return ret;
}
return 0;
}
/* make sure bo is in mappable vram */
if (drm->client.device.info.family >= NV_DEVICE_INFO_V0_TESLA ||
bo->mem.start + bo->mem.num_pages < mappable)
for (i = 0; i < nvbo->placement.num_placement; ++i) {
nvbo->placements[i].fpfn = 0;
nvbo->placements[i].lpfn = mappable;
}
for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
nvbo->busy_placements[i].fpfn = 0;
nvbo->busy_placements[i].lpfn = mappable;
}
nouveau_bo_placement_set(nvbo, TTM_PL_FLAG_VRAM, 0);
return nouveau_bo_validate(nvbo, false, false);
static int
nouveau_ttm_tt_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
{
struct ttm_dma_tt *ttm_dma = (void *)ttm;
struct nouveau_drm *drm;
struct device *dev;
unsigned i;
int r;
bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
if (ttm->state != tt_unpopulated)
return 0;
if (slave && ttm->sg) {
/* make userspace faulting work */
drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages,
ttm_dma->dma_address, ttm->num_pages);
ttm->state = tt_unbound;
return 0;
}
drm = nouveau_bdev(ttm->bdev);
dev = drm->dev->dev;
if (drm->agp.bridge) {
return ttm_agp_tt_populate(ttm, ctx);
#if IS_ENABLED(CONFIG_SWIOTLB) && IS_ENABLED(CONFIG_X86)
if (swiotlb_nr_tbl()) {
return ttm_dma_populate((void *)ttm, dev, ctx);
}
#endif
r = ttm_pool_populate(ttm, ctx);
if (r) {
return r;
}
for (i = 0; i < ttm->num_pages; i++) {
addr = dma_map_page(dev, ttm->pages[i], 0, PAGE_SIZE,
if (dma_mapping_error(dev, addr)) {
dma_unmap_page(dev, ttm_dma->dma_address[i],
PAGE_SIZE, DMA_BIDIRECTIONAL);
}
ttm_pool_unpopulate(ttm);
return -EFAULT;
}
ttm_dma->dma_address[i] = addr;
}
return 0;
}
static void
nouveau_ttm_tt_unpopulate(struct ttm_tt *ttm)
{
struct ttm_dma_tt *ttm_dma = (void *)ttm;
struct nouveau_drm *drm;
struct device *dev;
unsigned i;
bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
if (slave)
return;
drm = nouveau_bdev(ttm->bdev);
dev = drm->dev->dev;
if (drm->agp.bridge) {
ttm_agp_tt_unpopulate(ttm);
return;
}
#endif
#if IS_ENABLED(CONFIG_SWIOTLB) && IS_ENABLED(CONFIG_X86)
if (swiotlb_nr_tbl()) {
ttm_dma_unpopulate((void *)ttm, dev);
return;
}
#endif
for (i = 0; i < ttm->num_pages; i++) {
if (ttm_dma->dma_address[i]) {
dma_unmap_page(dev, ttm_dma->dma_address[i], PAGE_SIZE,
}
}
ttm_pool_unpopulate(ttm);
}
void
nouveau_bo_fence(struct nouveau_bo *nvbo, struct nouveau_fence *fence, bool exclusive)
{
struct dma_resv *resv = nvbo->bo.base.resv;

Maarten Lankhorst
committed
dma_resv_add_excl_fence(resv, &fence->base);
else if (fence)
dma_resv_add_shared_fence(resv, &fence->base);
}
struct ttm_bo_driver nouveau_bo_driver = {
.ttm_tt_create = &nouveau_ttm_tt_create,
.ttm_tt_populate = &nouveau_ttm_tt_populate,
.ttm_tt_unpopulate = &nouveau_ttm_tt_unpopulate,
.init_mem_type = nouveau_bo_init_mem_type,
.eviction_valuable = ttm_bo_eviction_valuable,
.evict_flags = nouveau_bo_evict_flags,
.move_notify = nouveau_bo_move_ntfy,
.move = nouveau_bo_move,
.verify_access = nouveau_bo_verify_access,
.fault_reserve_notify = &nouveau_ttm_fault_reserve_notify,
.io_mem_reserve = &nouveau_ttm_io_mem_reserve,
.io_mem_free = &nouveau_ttm_io_mem_free,