aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c')
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c72
1 files changed, 33 insertions, 39 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index d897c4c61a01..b372d8d650a5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -161,7 +161,7 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
break;
case TTM_PL_TT:
man->func = &amdgpu_gtt_mgr_func;
- man->gpu_offset = adev->mc.gart_start;
+ man->gpu_offset = adev->gmc.gart_start;
man->available_caching = TTM_PL_MASK_CACHING;
man->default_caching = TTM_PL_FLAG_CACHED;
man->flags = TTM_MEMTYPE_FLAG_MAPPABLE | TTM_MEMTYPE_FLAG_CMA;
@@ -169,7 +169,7 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
case TTM_PL_VRAM:
/* "On-card" video ram */
man->func = &amdgpu_vram_mgr_func;
- man->gpu_offset = adev->mc.vram_start;
+ man->gpu_offset = adev->gmc.vram_start;
man->flags = TTM_MEMTYPE_FLAG_FIXED |
TTM_MEMTYPE_FLAG_MAPPABLE;
man->available_caching = TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_WC;
@@ -217,9 +217,9 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
adev->mman.buffer_funcs_ring &&
adev->mman.buffer_funcs_ring->ready == false) {
amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU);
- } else if (adev->mc.visible_vram_size < adev->mc.real_vram_size &&
+ } else if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size &&
!(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)) {
- unsigned fpfn = adev->mc.visible_vram_size >> PAGE_SHIFT;
+ unsigned fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
struct drm_mm_node *node = bo->mem.mm_node;
unsigned long pages_left;
@@ -638,9 +638,9 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_mem_
case TTM_PL_VRAM:
mem->bus.offset = mem->start << PAGE_SHIFT;
/* check if it's visible */
- if ((mem->bus.offset + mem->bus.size) > adev->mc.visible_vram_size)
+ if ((mem->bus.offset + mem->bus.size) > adev->gmc.visible_vram_size)
return -EINVAL;
- mem->bus.base = adev->mc.aper_base;
+ mem->bus.base = adev->gmc.aper_base;
mem->bus.is_iomem = true;
break;
default:
@@ -891,7 +891,7 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
placement.num_busy_placement = 1;
placement.busy_placement = &placements;
placements.fpfn = 0;
- placements.lpfn = adev->mc.gart_size >> PAGE_SHIFT;
+ placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
placements.flags = (bo->mem.placement & ~TTM_PL_MASK_MEM) |
TTM_PL_FLAG_TT;
@@ -997,9 +997,6 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm,
struct amdgpu_ttm_tt *gtt = (void *)ttm;
bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
- if (ttm->state != tt_unpopulated)
- return 0;
-
if (gtt && gtt->userptr) {
ttm->sg = kzalloc(sizeof(struct sg_table), GFP_KERNEL);
if (!ttm->sg)
@@ -1212,7 +1209,7 @@ static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo,
nodes = amdgpu_find_mm_node(&abo->tbo.mem, &offset);
pos = (nodes->start << PAGE_SHIFT) + offset;
- while (len && pos < adev->mc.mc_vram_size) {
+ while (len && pos < adev->gmc.mc_vram_size) {
uint64_t aligned_pos = pos & ~(uint64_t)3;
uint32_t bytes = 4 - (pos & 3);
uint32_t shift = (pos & 3) * 8;
@@ -1298,7 +1295,7 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev)
struct ttm_operation_ctx ctx = { false, false };
int r = 0;
int i;
- u64 vram_size = adev->mc.visible_vram_size;
+ u64 vram_size = adev->gmc.visible_vram_size;
u64 offset = adev->fw_vram_usage.start_offset;
u64 size = adev->fw_vram_usage.size;
struct amdgpu_bo *bo;
@@ -1312,7 +1309,7 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev)
r = amdgpu_bo_create(adev, adev->fw_vram_usage.size,
PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM,
AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
- AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, NULL, NULL, 0,
+ AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, NULL, NULL,
&adev->fw_vram_usage.reserved_bo);
if (r)
goto error_create;
@@ -1387,8 +1384,12 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
return r;
}
adev->mman.initialized = true;
+
+ /* We opt to avoid OOM on system pages allocations */
+ adev->mman.bdev.no_retry = true;
+
r = ttm_bo_init_mm(&adev->mman.bdev, TTM_PL_VRAM,
- adev->mc.real_vram_size >> PAGE_SHIFT);
+ adev->gmc.real_vram_size >> PAGE_SHIFT);
if (r) {
DRM_ERROR("Failed initializing VRAM heap.\n");
return r;
@@ -1397,11 +1398,11 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
/* Reduce size of CPU-visible VRAM if requested */
vis_vram_limit = (u64)amdgpu_vis_vram_limit * 1024 * 1024;
if (amdgpu_vis_vram_limit > 0 &&
- vis_vram_limit <= adev->mc.visible_vram_size)
- adev->mc.visible_vram_size = vis_vram_limit;
+ vis_vram_limit <= adev->gmc.visible_vram_size)
+ adev->gmc.visible_vram_size = vis_vram_limit;
/* Change the size here instead of the init above so only lpfn is affected */
- amdgpu_ttm_set_active_vram_size(adev, adev->mc.visible_vram_size);
+ amdgpu_ttm_set_active_vram_size(adev, adev->gmc.visible_vram_size);
/*
*The reserved vram for firmware must be pinned to the specified
@@ -1412,21 +1413,21 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
return r;
}
- r = amdgpu_bo_create_kernel(adev, adev->mc.stolen_size, PAGE_SIZE,
+ r = amdgpu_bo_create_kernel(adev, adev->gmc.stolen_size, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->stolen_vga_memory,
NULL, NULL);
if (r)
return r;
DRM_INFO("amdgpu: %uM of VRAM memory ready\n",
- (unsigned) (adev->mc.real_vram_size / (1024 * 1024)));
+ (unsigned) (adev->gmc.real_vram_size / (1024 * 1024)));
if (amdgpu_gtt_size == -1) {
struct sysinfo si;
si_meminfo(&si);
gtt_size = min(max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20),
- adev->mc.mc_vram_size),
+ adev->gmc.mc_vram_size),
((uint64_t)si.totalram * si.mem_unit * 3/4));
}
else
@@ -1559,7 +1560,7 @@ static int amdgpu_map_buffer(struct ttm_buffer_object *bo,
BUG_ON(adev->mman.buffer_funcs->copy_max_bytes <
AMDGPU_GTT_MAX_TRANSFER_SIZE * 8);
- *addr = adev->mc.gart_start;
+ *addr = adev->gmc.gart_start;
*addr += (u64)window * AMDGPU_GTT_MAX_TRANSFER_SIZE *
AMDGPU_GPU_PAGE_SIZE;
@@ -1677,13 +1678,12 @@ error_free:
}
int amdgpu_fill_buffer(struct amdgpu_bo *bo,
- uint64_t src_data,
+ uint32_t src_data,
struct reservation_object *resv,
struct dma_fence **fence)
{
struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
- uint32_t max_bytes = 8 *
- adev->vm_manager.vm_pte_funcs->set_max_nums_pte_pde;
+ uint32_t max_bytes = adev->mman.buffer_funcs->fill_max_bytes;
struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
struct drm_mm_node *mm_node;
@@ -1714,9 +1714,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
num_pages -= mm_node->size;
++mm_node;
}
-
- /* num of dwords for each SDMA_OP_PTEPDE cmd */
- num_dw = num_loops * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw;
+ num_dw = num_loops * adev->mman.buffer_funcs->fill_num_dw;
/* for IB padding */
num_dw += 64;
@@ -1741,16 +1739,12 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
uint32_t byte_count = mm_node->size << PAGE_SHIFT;
uint64_t dst_addr;
- WARN_ONCE(byte_count & 0x7, "size should be a multiple of 8");
-
dst_addr = amdgpu_mm_node_addr(&bo->tbo, mm_node, &bo->tbo.mem);
while (byte_count) {
uint32_t cur_size_in_bytes = min(byte_count, max_bytes);
- amdgpu_vm_set_pte_pde(adev, &job->ibs[0],
- dst_addr, 0,
- cur_size_in_bytes >> 3, 0,
- src_data);
+ amdgpu_emit_fill_buffer(adev, &job->ibs[0], src_data,
+ dst_addr, cur_size_in_bytes);
dst_addr += cur_size_in_bytes;
byte_count -= cur_size_in_bytes;
@@ -1811,14 +1805,14 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf,
if (size & 0x3 || *pos & 0x3)
return -EINVAL;
- if (*pos >= adev->mc.mc_vram_size)
+ if (*pos >= adev->gmc.mc_vram_size)
return -ENXIO;
while (size) {
unsigned long flags;
uint32_t value;
- if (*pos >= adev->mc.mc_vram_size)
+ if (*pos >= adev->gmc.mc_vram_size)
return result;
spin_lock_irqsave(&adev->mmio_idx_lock, flags);
@@ -1850,14 +1844,14 @@ static ssize_t amdgpu_ttm_vram_write(struct file *f, const char __user *buf,
if (size & 0x3 || *pos & 0x3)
return -EINVAL;
- if (*pos >= adev->mc.mc_vram_size)
+ if (*pos >= adev->gmc.mc_vram_size)
return -ENXIO;
while (size) {
unsigned long flags;
uint32_t value;
- if (*pos >= adev->mc.mc_vram_size)
+ if (*pos >= adev->gmc.mc_vram_size)
return result;
r = get_user(value, (uint32_t *)buf);
@@ -2001,9 +1995,9 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
if (IS_ERR(ent))
return PTR_ERR(ent);
if (ttm_debugfs_entries[count].domain == TTM_PL_VRAM)
- i_size_write(ent->d_inode, adev->mc.mc_vram_size);
+ i_size_write(ent->d_inode, adev->gmc.mc_vram_size);
else if (ttm_debugfs_entries[count].domain == TTM_PL_TT)
- i_size_write(ent->d_inode, adev->mc.gart_size);
+ i_size_write(ent->d_inode, adev->gmc.gart_size);
adev->mman.debugfs_entries[count] = ent;
}
generated by cgit v1.2.3 (git 2.39.1) at 2026-01-06 05:33:54 +0000