From: Christian K?nig <christian.koe...@amd.com>
Signed-off-by: Christian K?nig <christian.koenig at amd.com>
---
drivers/gpu/drm/radeon/radeon.h | 12 +-
drivers/gpu/drm/radeon/radeon_cs.c | 5 -
drivers/gpu/drm/radeon/radeon_device.c | 1 -
drivers/gpu/drm/radeon/radeon_vm.c | 389 ++++++++++++---------------------
4 files changed, 141 insertions(+), 266 deletions(-)
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index f4fb984..375dbf5 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -854,8 +854,12 @@ struct radeon_mec {
#define R600_PTE_READABLE (1 << 5)
#define R600_PTE_WRITEABLE (1 << 6)
+struct radeon_vm_pt {
+ struct radeon_bo *bo;
+ uint64_t addr;
+};
+
struct radeon_vm {
- struct list_head list;
struct list_head va;
unsigned id;
@@ -864,7 +868,7 @@ struct radeon_vm {
uint64_t pd_gpu_addr;
/* array of page tables, one for each page directory entry */
- struct radeon_sa_bo **page_tables;
+ struct radeon_vm_pt *page_tables;
struct mutex mutex;
/* last fence for cs using this vm */
@@ -877,9 +881,7 @@ struct radeon_vm {
struct radeon_vm_manager {
struct mutex lock;
- struct list_head lru_vm;
struct radeon_fence *active[RADEON_NUM_VM];
- struct radeon_sa_manager sa_manager;
uint32_t max_pfn;
/* number of VMIDs */
unsigned nvm;
@@ -2793,11 +2795,9 @@ int radeon_vm_manager_init(struct radeon_device *rdev);
void radeon_vm_manager_fini(struct radeon_device *rdev);
int radeon_vm_init(struct radeon_device *rdev, struct radeon_vm *vm);
void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm);
-int radeon_vm_alloc_pt(struct radeon_device *rdev, struct radeon_vm *vm);
struct radeon_bo_list *radeon_vm_add_bos(struct radeon_device *rdev,
struct radeon_vm *vm,
struct list_head *head);
-void radeon_vm_add_to_lru(struct radeon_device *rdev, struct radeon_vm *vm);
struct radeon_fence *radeon_vm_grab_id(struct radeon_device *rdev,
struct radeon_vm *vm, int ring);
void radeon_vm_flush(struct radeon_device *rdev,
diff --git a/drivers/gpu/drm/radeon/radeon_cs.c
b/drivers/gpu/drm/radeon/radeon_cs.c
index 335cc84..849f997 100644
--- a/drivers/gpu/drm/radeon/radeon_cs.c
+++ b/drivers/gpu/drm/radeon/radeon_cs.c
@@ -420,10 +420,6 @@ static int radeon_cs_ib_vm_chunk(struct radeon_device
*rdev,
mutex_lock(&rdev->vm_manager.lock);
mutex_lock(&vm->mutex);
- r = radeon_vm_alloc_pt(rdev, vm);
- if (r) {
- goto out;
- }
r = radeon_bo_vm_update_pte(parser, vm);
if (r) {
goto out;
@@ -441,7 +437,6 @@ static int radeon_cs_ib_vm_chunk(struct radeon_device *rdev,
}
out:
- radeon_vm_add_to_lru(rdev, vm);
mutex_unlock(&vm->mutex);
mutex_unlock(&rdev->vm_manager.lock);
return r;
diff --git a/drivers/gpu/drm/radeon/radeon_device.c
b/drivers/gpu/drm/radeon/radeon_device.c
index fa7841b..e58dbab 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -1198,7 +1198,6 @@ int radeon_device_init(struct radeon_device *rdev,
* Max GPUVM size for cayman and SI is 40 bits.
*/
rdev->vm_manager.max_pfn = 1 << 20;
- INIT_LIST_HEAD(&rdev->vm_manager.lru_vm);
/* Set asic functions */
r = radeon_asic_init(rdev);
diff --git a/drivers/gpu/drm/radeon/radeon_vm.c
b/drivers/gpu/drm/radeon/radeon_vm.c
index b50e50f..5441a6a 100644
--- a/drivers/gpu/drm/radeon/radeon_vm.c
+++ b/drivers/gpu/drm/radeon/radeon_vm.c
@@ -84,83 +84,19 @@ static unsigned radeon_vm_directory_size(struct
radeon_device *rdev)
*/
int radeon_vm_manager_init(struct radeon_device *rdev)
{
- struct radeon_vm *vm;
- struct radeon_bo_va *bo_va;
int r;
- unsigned size;
if (!rdev->vm_manager.enabled) {
- /* allocate enough for 2 full VM pts */
- size = radeon_vm_directory_size(rdev);
- size += rdev->vm_manager.max_pfn * 8;
- size *= 2;
- r = radeon_sa_bo_manager_init(rdev,
&rdev->vm_manager.sa_manager,
- RADEON_GPU_PAGE_ALIGN(size),
- RADEON_VM_PTB_ALIGN_SIZE,
- RADEON_GEM_DOMAIN_VRAM);
- if (r) {
- dev_err(rdev->dev, "failed to allocate vm bo (%dKB)\n",
- (rdev->vm_manager.max_pfn * 8) >> 10);
- return r;
- }
-
r = radeon_asic_vm_init(rdev);
if (r)
return r;
rdev->vm_manager.enabled = true;
-
- r = radeon_sa_bo_manager_start(rdev,
&rdev->vm_manager.sa_manager);
- if (r)
- return r;
- }
-
- /* restore page table */
- list_for_each_entry(vm, &rdev->vm_manager.lru_vm, list) {
- if (vm->page_tables == NULL)
- continue;
-
- list_for_each_entry(bo_va, &vm->va, vm_list) {
- bo_va->valid = false;
- }
}
return 0;
}
/**
- * radeon_vm_free_pt - free the page table for a specific vm
- *
- * @rdev: radeon_device pointer
- * @vm: vm to unbind
- *
- * Free the page table of a specific vm (cayman+).
- *
- * Global and local mutex must be lock!
- */
-static void radeon_vm_free_pt(struct radeon_device *rdev,
- struct radeon_vm *vm)
-{
- struct radeon_bo_va *bo_va;
- int i;
-
- if (!vm->page_tables)
- return;
-
- list_del_init(&vm->list);
- list_for_each_entry(bo_va, &vm->va, vm_list) {
- bo_va->valid = false;
- }
-
- if (vm->page_tables == NULL)
- return;
-
- for (i = 0; i < radeon_vm_num_pdes(rdev); i++)
- radeon_sa_bo_free(rdev, &vm->page_tables[i], vm->fence);
-
- kfree(vm->page_tables);
-}
-
-/**
* radeon_vm_manager_fini - tear down the vm manager
*
* @rdev: radeon_device pointer
@@ -169,105 +105,17 @@ static void radeon_vm_free_pt(struct radeon_device *rdev,
*/
void radeon_vm_manager_fini(struct radeon_device *rdev)
{
- struct radeon_vm *vm, *tmp;
int i;
if (!rdev->vm_manager.enabled)
return;
mutex_lock(&rdev->vm_manager.lock);
- /* free all allocated page tables */
- list_for_each_entry_safe(vm, tmp, &rdev->vm_manager.lru_vm, list) {
- mutex_lock(&vm->mutex);
- radeon_vm_free_pt(rdev, vm);
- mutex_unlock(&vm->mutex);
- }
- for (i = 0; i < RADEON_NUM_VM; ++i) {
+ for (i = 0; i < RADEON_NUM_VM; ++i)
radeon_fence_unref(&rdev->vm_manager.active[i]);
- }
radeon_asic_vm_fini(rdev);
- mutex_unlock(&rdev->vm_manager.lock);
-
- radeon_sa_bo_manager_suspend(rdev, &rdev->vm_manager.sa_manager);
- radeon_sa_bo_manager_fini(rdev, &rdev->vm_manager.sa_manager);
rdev->vm_manager.enabled = false;
-}
-
-/**
- * radeon_vm_evict - evict page table to make room for new one
- *
- * @rdev: radeon_device pointer
- * @vm: VM we want to allocate something for
- *
- * Evict a VM from the lru, making sure that it isn't @vm. (cayman+).
- * Returns 0 for success, -ENOMEM for failure.
- *
- * Global and local mutex must be locked!
- */
-static int radeon_vm_evict(struct radeon_device *rdev, struct radeon_vm *vm)
-{
- struct radeon_vm *vm_evict;
-
- if (list_empty(&rdev->vm_manager.lru_vm))
- return -ENOMEM;
-
- vm_evict = list_first_entry(&rdev->vm_manager.lru_vm,
- struct radeon_vm, list);
- if (vm_evict == vm)
- return -ENOMEM;
-
- mutex_lock(&vm_evict->mutex);
- radeon_vm_free_pt(rdev, vm_evict);
- mutex_unlock(&vm_evict->mutex);
- return 0;
-}
-
-/**
- * radeon_vm_alloc_pt - allocates a page table for a VM
- *
- * @rdev: radeon_device pointer
- * @vm: vm to bind
- *
- * Allocate a page table for the requested vm (cayman+).
- * Returns 0 for success, error for failure.
- *
- * Global and local mutex must be locked!
- */
-int radeon_vm_alloc_pt(struct radeon_device *rdev, struct radeon_vm *vm)
-{
- unsigned pts_size;
-
- if (vm == NULL)
- return -EINVAL;
-
- if (vm->page_tables != NULL)
- return 0;
-
- /* allocate page table array */
- pts_size = radeon_vm_num_pdes(rdev) * sizeof(struct radeon_sa_bo *);
- vm->page_tables = kzalloc(pts_size, GFP_KERNEL);
- if (vm->page_tables == NULL) {
- DRM_ERROR("Cannot allocate memory for page table array\n");
- return -ENOMEM;
- }
-
- return 0;
-}
-
-/**
- * radeon_vm_add_to_lru - add VMs page table to LRU list
- *
- * @rdev: radeon_device pointer
- * @vm: vm to add to LRU
- *
- * Add the allocated page table to the LRU list (cayman+).
- *
- * Global mutex must be locked!
- */
-void radeon_vm_add_to_lru(struct radeon_device *rdev, struct radeon_vm *vm)
-{
- list_del_init(&vm->list);
- list_add_tail(&vm->list, &rdev->vm_manager.lru_vm);
+ mutex_unlock(&rdev->vm_manager.lock);
}
/**
@@ -281,8 +129,10 @@ struct radeon_bo_list *radeon_vm_add_bos(struct
radeon_device *rdev,
struct list_head *head)
{
struct radeon_bo_list *list;
+ unsigned i, idx, size;
- list = kmalloc(sizeof(struct radeon_bo_list), GFP_KERNEL);
+ size = (radeon_vm_num_pdes(rdev) + 1) * sizeof(struct radeon_bo_list);
+ list = kmalloc(size, GFP_KERNEL);
if (!list)
return NULL;
@@ -294,6 +144,18 @@ struct radeon_bo_list *radeon_vm_add_bos(struct
radeon_device *rdev,
list[0].tv.bo = &vm->page_directory->tbo;
radeon_bo_list_add_object(&list[0], head);
+ for (i = 0, idx = 1; i < radeon_vm_num_pdes(rdev); i++) {
+ if (!vm->page_tables[i].bo)
+ continue;
+
+ list[idx].bo = vm->page_tables[i].bo;
+ list[idx].written = 1;
+ list[idx].domain = RADEON_GEM_DOMAIN_VRAM;
+ list[idx].alt_domain = RADEON_GEM_DOMAIN_VRAM;
+ list[idx].tv.bo = &list[idx].bo->tbo;
+ radeon_bo_list_add_object(&list[idx++], head);
+ }
+
return list;
}
@@ -475,6 +337,63 @@ struct radeon_bo_va *radeon_vm_bo_add(struct radeon_device
*rdev,
}
/**
+ * radeon_vm_clear_bo - initially clear the page dir/table
+ *
+ * @rdev: radeon_device pointer
+ * @bo: bo to clear
+ */
+static int radeon_vm_clear_bo(struct radeon_device *rdev,
+ struct radeon_bo *bo)
+{
+ struct ttm_validate_buffer tv;
+ struct ww_acquire_ctx ticket;
+ struct list_head head;
+ struct radeon_ib ib;
+ unsigned entries;
+ uint64_t addr;
+ int r;
+
+ memset(&tv, 0, sizeof(tv));
+ tv.bo = &bo->tbo;
+
+ INIT_LIST_HEAD(&head);
+ list_add(&tv.head, &head);
+
+ r = ttm_eu_reserve_buffers(&ticket, &head);
+ if (r)
+ return r;
+
+ r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false);
+ if (r)
+ goto error;
+
+ addr = radeon_bo_gpu_offset(bo);
+ entries = radeon_bo_size(bo) / 8;
+
+ r = radeon_ib_get(rdev, R600_RING_TYPE_DMA_INDEX, &ib,
+ NULL, entries * 2 + 64);
+ if (r)
+ goto error;
+
+ ib.length_dw = 0;
+
+ radeon_asic_vm_set_page(rdev, &ib, addr, 0, entries, 0, 0);
+
+ r = radeon_ib_schedule(rdev, &ib, NULL);
+ if (r)
+ goto error;
+
+ ttm_eu_fence_buffer_objects(&ticket, &head, ib.fence);
+ radeon_ib_free(rdev, &ib);
+
+ return 0;
+
+error:
+ ttm_eu_backoff_reservation(&ticket, &head);
+ return r;
+}
+
+/**
* radeon_vm_bo_set_addr - set bos virtual address inside a vm
*
* @rdev: radeon_device pointer
@@ -498,7 +417,8 @@ int radeon_vm_bo_set_addr(struct radeon_device *rdev,
struct radeon_vm *vm = bo_va->vm;
struct radeon_bo_va *tmp;
struct list_head *head;
- unsigned last_pfn;
+ unsigned last_pfn, pt_idx;
+ int r;
if (soffset) {
/* make sure object fit at this offset */
@@ -549,8 +469,36 @@ int radeon_vm_bo_set_addr(struct radeon_device *rdev,
bo_va->valid = false;
list_move(&bo_va->vm_list, head);
+ soffset = (soffset / RADEON_GPU_PAGE_SIZE) >> RADEON_VM_BLOCK_SIZE;
+ eoffset = (eoffset / RADEON_GPU_PAGE_SIZE) >> RADEON_VM_BLOCK_SIZE;
+
+ radeon_bo_unreserve(bo_va->bo);
+
+ /* walk over the address space and allocate the page tables */
+ for (pt_idx = soffset; pt_idx <= eoffset; ++pt_idx) {
+ if (vm->page_tables[pt_idx].bo)
+ continue;
+
+ vm->page_tables[pt_idx].addr = 0;
+ r = radeon_bo_create(rdev, RADEON_VM_PTE_COUNT * 8,
+ RADEON_GPU_PAGE_SIZE, false,
+ RADEON_GEM_DOMAIN_VRAM, NULL,
+ &vm->page_tables[pt_idx].bo);
+ if (r)
+ return r;
+
+ mutex_unlock(&vm->mutex);
+ r = radeon_vm_clear_bo(rdev, vm->page_tables[pt_idx].bo);
+ mutex_lock(&vm->mutex);
+ if (r) {
+ radeon_bo_unref(&vm->page_tables[pt_idx].bo);
+ vm->page_tables[pt_idx].bo = NULL;
+ return r;
+ }
+ }
+
mutex_unlock(&vm->mutex);
- return 0;
+ return radeon_bo_reserve(bo_va->bo, false);
}
/**
@@ -620,37 +568,21 @@ static int radeon_vm_update_pdes(struct radeon_device
*rdev,
uint64_t last_pde = ~0, last_pt = ~0;
unsigned count = 0;
uint64_t pt_idx;
- int r;
start = (start / RADEON_GPU_PAGE_SIZE) >> RADEON_VM_BLOCK_SIZE;
end = (end / RADEON_GPU_PAGE_SIZE) >> RADEON_VM_BLOCK_SIZE;
/* walk over the address space and update the page directory */
for (pt_idx = start; pt_idx <= end; ++pt_idx) {
- uint64_t pde, pt;
+ struct radeon_bo *bo = vm->page_tables[pt_idx].bo;
+ uint64_t pde, pt = radeon_bo_gpu_offset(bo);
- if (vm->page_tables[pt_idx])
+ if (vm->page_tables[pt_idx].addr == pt)
continue;
-
-retry:
- r = radeon_sa_bo_new(rdev, &rdev->vm_manager.sa_manager,
- &vm->page_tables[pt_idx],
- RADEON_VM_PTE_COUNT * 8,
- RADEON_GPU_PAGE_SIZE, false);
-
- if (r == -ENOMEM) {
- r = radeon_vm_evict(rdev, vm);
- if (r)
- return r;
- goto retry;
- } else if (r) {
- return r;
- }
+ vm->page_tables[pt_idx].addr = pt;
pde = radeon_bo_gpu_offset(vm->page_directory) + pt_idx * 8;
- pt = radeon_sa_bo_gpu_addr(vm->page_tables[pt_idx]);
-
if (((last_pde + 8 * count) != pde) ||
((last_pt + incr * count) != pt)) {
@@ -658,10 +590,6 @@ retry:
radeon_asic_vm_set_page(rdev, ib, last_pde,
last_pt, count, incr,
R600_PTE_VALID);
-
- count *= RADEON_VM_PTE_COUNT;
- radeon_asic_vm_set_page(rdev, ib, last_pt, 0,
- count, 0, 0);
}
count = 1;
@@ -672,15 +600,10 @@ retry:
}
}
- if (count) {
+ if (count)
radeon_asic_vm_set_page(rdev, ib, last_pde, last_pt, count,
incr, R600_PTE_VALID);
- count *= RADEON_VM_PTE_COUNT;
- radeon_asic_vm_set_page(rdev, ib, last_pt, 0,
- count, 0, 0);
- }
-
return 0;
}
@@ -724,7 +647,7 @@ static void radeon_vm_update_ptes(struct radeon_device
*rdev,
else
nptes = RADEON_VM_PTE_COUNT - (addr & mask);
- pte = radeon_sa_bo_gpu_addr(vm->page_tables[pt_idx]);
+ pte = radeon_bo_gpu_offset(vm->page_tables[pt_idx].bo);
pte += (addr & mask) * 8;
if ((last_pte + 8 * count) != pte) {
@@ -778,10 +701,6 @@ int radeon_vm_bo_update(struct radeon_device *rdev,
uint64_t addr;
int r;
- /* nothing to do if vm isn't bound */
- if (vm->page_tables == NULL)
- return 0;
-
bo_va = radeon_vm_bo_find(vm, bo);
if (bo_va == NULL) {
dev_err(rdev->dev, "bo %p not in vm %p\n", bo, vm);
@@ -938,11 +857,7 @@ void radeon_vm_bo_invalidate(struct radeon_device *rdev,
*/
int radeon_vm_init(struct radeon_device *rdev, struct radeon_vm *vm)
{
- unsigned pd_size, pd_entries;
- struct ttm_validate_buffer tv;
- struct ww_acquire_ctx ticket;
- struct list_head head;
- struct radeon_ib ib;
+ unsigned pd_size, pd_entries, pts_size;
int r;
vm->id = 0;
@@ -950,70 +865,33 @@ int radeon_vm_init(struct radeon_device *rdev, struct
radeon_vm *vm)
vm->last_flush = NULL;
vm->last_id_use = NULL;
mutex_init(&vm->mutex);
- INIT_LIST_HEAD(&vm->list);
INIT_LIST_HEAD(&vm->va);
pd_size = radeon_vm_directory_size(rdev);
pd_entries = radeon_vm_num_pdes(rdev);
+ /* allocate page table array */
+ pts_size = pd_entries * sizeof(struct radeon_vm_pt);
+ vm->page_tables = kzalloc(pts_size, GFP_KERNEL);
+ if (vm->page_tables == NULL) {
+ DRM_ERROR("Cannot allocate memory for page table array\n");
+ return -ENOMEM;
+ }
+
r = radeon_bo_create(rdev, pd_size, RADEON_VM_PTB_ALIGN_SIZE, false,
RADEON_GEM_DOMAIN_VRAM, NULL,
&vm->page_directory);
if (r)
return r;
- /* Initially clear the page directory */
-
- memset(&tv, 0, sizeof(tv));
- tv.bo = &vm->page_directory->tbo;
-
- INIT_LIST_HEAD(&head);
- list_add(&tv.head, &head);
-
- r = ttm_eu_reserve_buffers(&ticket, &head);
- if (r)
- goto error_free;
-
- r = ttm_bo_validate(&vm->page_directory->tbo,
- &vm->page_directory->placement,
- true, false);
- if (r)
- goto error_unreserve;
-
- vm->pd_gpu_addr = radeon_bo_gpu_offset(vm->page_directory);
-
- r = radeon_ib_get(rdev, R600_RING_TYPE_DMA_INDEX, &ib,
- NULL, pd_entries * 2 + 64);
- if (r)
- goto error_unreserve;
-
- ib.length_dw = 0;
-
- radeon_asic_vm_set_page(rdev, &ib, vm->pd_gpu_addr,
- 0, pd_entries, 0, 0);
-
- radeon_semaphore_sync_to(ib.semaphore, vm->last_id_use);
- r = radeon_ib_schedule(rdev, &ib, NULL);
- if (r)
- goto error_unreserve;
-
- ttm_eu_fence_buffer_objects(&ticket, &head, ib.fence);
- radeon_fence_unref(&vm->fence);
- vm->fence = radeon_fence_ref(ib.fence);
- radeon_fence_unref(&vm->last_flush);
-
- radeon_ib_free(rdev, &ib);
+ r = radeon_vm_clear_bo(rdev, vm->page_directory);
+ if (r) {
+ radeon_bo_unref(&vm->page_directory);
+ vm->page_directory = NULL;
+ return r;
+ }
return 0;
-
-error_unreserve:
- ttm_eu_backoff_reservation(&ticket, &head);
-
-error_free:
- radeon_bo_unref(&vm->page_directory);
- vm->page_directory = NULL;
- return r;
-
}
/**
@@ -1028,12 +906,9 @@ error_free:
void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm)
{
struct radeon_bo_va *bo_va, *tmp;
- int r;
+ int i, r;
- mutex_lock(&rdev->vm_manager.lock);
mutex_lock(&vm->mutex);
- radeon_vm_free_pt(rdev, vm);
- mutex_unlock(&rdev->vm_manager.lock);
if (!list_empty(&vm->va)) {
dev_err(rdev->dev, "still active bo inside vm\n");
@@ -1047,10 +922,16 @@ void radeon_vm_fini(struct radeon_device *rdev, struct
radeon_vm *vm)
kfree(bo_va);
}
}
+
+ for (i = 0; i < radeon_vm_num_pdes(rdev); i++)
+ radeon_bo_unref(&vm->page_tables[i].bo);
+ kfree(vm->page_tables);
+
+ radeon_bo_unref(&vm->page_directory);
+
radeon_fence_unref(&vm->fence);
radeon_fence_unref(&vm->last_flush);
radeon_fence_unref(&vm->last_id_use);
- mutex_unlock(&vm->mutex);
- radeon_bo_unref(&vm->page_directory);
+ mutex_unlock(&vm->mutex);
}
--
1.8.3.2
