amdgpu-y += \
vi.o mxgpu_vi.o nbio_v6_1.o soc15.o emu_soc.o mxgpu_ai.o nbio_v7_0.o vega10_reg_init.o
+ # add DF block
+ amdgpu-y += \
+ df_v1_7.o
+
# add GMC block
amdgpu-y += \
gmc_v7_0.o \
amdgpu_amdkfd.o \
amdgpu_amdkfd_fence.o \
amdgpu_amdkfd_gpuvm.o \
- amdgpu_amdkfd_gfx_v8.o
+ amdgpu_amdkfd_gfx_v8.o \
+ amdgpu_amdkfd_gfx_v9.o
# add cgs
amdgpu-y += amdgpu_cgs.o
case CHIP_POLARIS11:
kfd2kgd = amdgpu_amdkfd_gfx_8_0_get_functions();
break;
+ case CHIP_VEGA10:
+ case CHIP_RAVEN:
+ kfd2kgd = amdgpu_amdkfd_gfx_9_0_get_functions();
+ break;
default:
dev_dbg(adev->dev, "kfd not supported on this ASIC\n");
return;
&gpu_resources.doorbell_physical_address,
&gpu_resources.doorbell_aperture_size,
&gpu_resources.doorbell_start_offset);
+ if (adev->asic_type >= CHIP_VEGA10) {
+ /* On SOC15 the BIF is involved in routing
+ * doorbells using the low 12 bits of the
+ * address. Communicate the assignments to
+ * KFD. KFD uses two doorbell pages per
+ * process in case of 64-bit doorbells so we
+ * can use each doorbell assignment twice.
+ */
+ gpu_resources.sdma_doorbell[0][0] =
+ AMDGPU_DOORBELL64_sDMA_ENGINE0;
+ gpu_resources.sdma_doorbell[0][1] =
+ AMDGPU_DOORBELL64_sDMA_ENGINE0 + 0x200;
+ gpu_resources.sdma_doorbell[1][0] =
+ AMDGPU_DOORBELL64_sDMA_ENGINE1;
+ gpu_resources.sdma_doorbell[1][1] =
+ AMDGPU_DOORBELL64_sDMA_ENGINE1 + 0x200;
+ /* Doorbells 0x0f0-0ff and 0x2f0-2ff are reserved for
+ * SDMA, IH and VCN. So don't use them for the CP.
+ */
+ gpu_resources.reserved_doorbell_mask = 0x1f0;
+ gpu_resources.reserved_doorbell_val = 0x0f0;
+ }
kgd2kfd->device_init(adev->kfd, &gpu_resources);
}
{
struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
struct amdgpu_bo *bo = NULL;
+ struct amdgpu_bo_param bp;
int r;
uint64_t gpu_addr_tmp = 0;
void *cpu_ptr_tmp = NULL;
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
- AMDGPU_GEM_CREATE_CPU_GTT_USWC, ttm_bo_type_kernel,
- NULL, &bo);
+ memset(&bp, 0, sizeof(bp));
+ bp.size = size;
+ bp.byte_align = PAGE_SIZE;
+ bp.domain = AMDGPU_GEM_DOMAIN_GTT;
+ bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC;
+ bp.type = ttm_bo_type_kernel;
+ bp.resv = NULL;
+ r = amdgpu_bo_create(adev, &bp, &bo);
if (r) {
dev_err(adev->dev,
"failed to allocate BO for amdkfd (%d)\n", r);
#define pr_fmt(fmt) "kfd2kgd: " fmt
#include <linux/list.h>
+#include <linux/sched/mm.h>
#include <drm/drmP.h>
#include "amdgpu_object.h"
#include "amdgpu_vm.h"
*/
#define VI_BO_SIZE_ALIGN (0x8000)
+/* BO flag to indicate a KFD userptr BO */
+#define AMDGPU_AMDKFD_USERPTR_BO (1ULL << 63)
+
+/* Userptr restore delay, just long enough to allow consecutive VM
+ * changes to accumulate
+ */
+#define AMDGPU_USERPTR_RESTORE_DELAY_MS 1
+
/* Impose limit on how much memory KFD can use */
static struct {
uint64_t max_system_mem_limit;
+ uint64_t max_userptr_mem_limit;
int64_t system_mem_used;
+ int64_t userptr_mem_used;
spinlock_t mem_limit_lock;
} kfd_mem_limit;
#define domain_string(domain) domain_bit_to_string[ffs(domain)-1]
+static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work);
static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
/* Set memory usage limits. Current, limits are
* System (kernel) memory - 3/8th System RAM
+ * Userptr memory - 3/4th System RAM
*/
void amdgpu_amdkfd_gpuvm_init_mem_limits(void)
{
spin_lock_init(&kfd_mem_limit.mem_limit_lock);
kfd_mem_limit.max_system_mem_limit = (mem >> 1) - (mem >> 3);
- pr_debug("Kernel memory limit %lluM\n",
- (kfd_mem_limit.max_system_mem_limit >> 20));
+ kfd_mem_limit.max_userptr_mem_limit = mem - (mem >> 2);
+ pr_debug("Kernel memory limit %lluM, userptr limit %lluM\n",
+ (kfd_mem_limit.max_system_mem_limit >> 20),
+ (kfd_mem_limit.max_userptr_mem_limit >> 20));
}
static int amdgpu_amdkfd_reserve_system_mem_limit(struct amdgpu_device *adev,
goto err_no_mem;
}
kfd_mem_limit.system_mem_used += (acc_size + size);
+ } else if (domain == AMDGPU_GEM_DOMAIN_CPU) {
+ if ((kfd_mem_limit.system_mem_used + acc_size >
+ kfd_mem_limit.max_system_mem_limit) ||
+ (kfd_mem_limit.userptr_mem_used + (size + acc_size) >
+ kfd_mem_limit.max_userptr_mem_limit)) {
+ ret = -ENOMEM;
+ goto err_no_mem;
+ }
+ kfd_mem_limit.system_mem_used += acc_size;
+ kfd_mem_limit.userptr_mem_used += size;
}
err_no_mem:
spin_unlock(&kfd_mem_limit.mem_limit_lock);
sizeof(struct amdgpu_bo));
spin_lock(&kfd_mem_limit.mem_limit_lock);
- if (domain == AMDGPU_GEM_DOMAIN_GTT)
+ if (domain == AMDGPU_GEM_DOMAIN_GTT) {
kfd_mem_limit.system_mem_used -= (acc_size + size);
+ } else if (domain == AMDGPU_GEM_DOMAIN_CPU) {
+ kfd_mem_limit.system_mem_used -= acc_size;
+ kfd_mem_limit.userptr_mem_used -= size;
+ }
WARN_ONCE(kfd_mem_limit.system_mem_used < 0,
"kfd system memory accounting unbalanced");
+ WARN_ONCE(kfd_mem_limit.userptr_mem_used < 0,
+ "kfd userptr memory accounting unbalanced");
spin_unlock(&kfd_mem_limit.mem_limit_lock);
}
{
spin_lock(&kfd_mem_limit.mem_limit_lock);
- if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GTT) {
+ if (bo->flags & AMDGPU_AMDKFD_USERPTR_BO) {
+ kfd_mem_limit.system_mem_used -= bo->tbo.acc_size;
+ kfd_mem_limit.userptr_mem_used -= amdgpu_bo_size(bo);
+ } else if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GTT) {
kfd_mem_limit.system_mem_used -=
(bo->tbo.acc_size + amdgpu_bo_size(bo));
}
WARN_ONCE(kfd_mem_limit.system_mem_used < 0,
"kfd system memory accounting unbalanced");
+ WARN_ONCE(kfd_mem_limit.userptr_mem_used < 0,
+ "kfd userptr memory accounting unbalanced");
spin_unlock(&kfd_mem_limit.mem_limit_lock);
}
}
static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem,
- struct amdkfd_process_info *process_info)
+ struct amdkfd_process_info *process_info,
+ bool userptr)
{
struct ttm_validate_buffer *entry = &mem->validate_list;
struct amdgpu_bo *bo = mem->bo;
entry->shared = true;
entry->bo = &bo->tbo;
mutex_lock(&process_info->lock);
- list_add_tail(&entry->head, &process_info->kfd_bo_list);
+ if (userptr)
+ list_add_tail(&entry->head, &process_info->userptr_valid_list);
+ else
+ list_add_tail(&entry->head, &process_info->kfd_bo_list);
mutex_unlock(&process_info->lock);
}
+/* Initializes user pages. It registers the MMU notifier and validates
+ * the userptr BO in the GTT domain.
+ *
+ * The BO must already be on the userptr_valid_list. Otherwise an
+ * eviction and restore may happen that leaves the new BO unmapped
+ * with the user mode queues running.
+ *
+ * Takes the process_info->lock to protect against concurrent restore
+ * workers.
+ *
+ * Returns 0 for success, negative errno for errors.
+ */
+static int init_user_pages(struct kgd_mem *mem, struct mm_struct *mm,
+ uint64_t user_addr)
+{
+ struct amdkfd_process_info *process_info = mem->process_info;
+ struct amdgpu_bo *bo = mem->bo;
+ struct ttm_operation_ctx ctx = { true, false };
+ int ret = 0;
+
+ mutex_lock(&process_info->lock);
+
+ ret = amdgpu_ttm_tt_set_userptr(bo->tbo.ttm, user_addr, 0);
+ if (ret) {
+ pr_err("%s: Failed to set userptr: %d\n", __func__, ret);
+ goto out;
+ }
+
+ ret = amdgpu_mn_register(bo, user_addr);
+ if (ret) {
+ pr_err("%s: Failed to register MMU notifier: %d\n",
+ __func__, ret);
+ goto out;
+ }
+
+ /* If no restore worker is running concurrently, user_pages
+ * should not be allocated
+ */
+ WARN(mem->user_pages, "Leaking user_pages array");
+
+ mem->user_pages = kvmalloc_array(bo->tbo.ttm->num_pages,
+ sizeof(struct page *),
+ GFP_KERNEL | __GFP_ZERO);
+ if (!mem->user_pages) {
+ pr_err("%s: Failed to allocate pages array\n", __func__);
+ ret = -ENOMEM;
+ goto unregister_out;
+ }
+
+ ret = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm, mem->user_pages);
+ if (ret) {
+ pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
+ goto free_out;
+ }
+
+ amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, mem->user_pages);
+
+ ret = amdgpu_bo_reserve(bo, true);
+ if (ret) {
+ pr_err("%s: Failed to reserve BO\n", __func__);
+ goto release_out;
+ }
+ amdgpu_ttm_placement_from_domain(bo, mem->domain);
+ ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+ if (ret)
+ pr_err("%s: failed to validate BO\n", __func__);
+ amdgpu_bo_unreserve(bo);
+
+release_out:
+ if (ret)
+ release_pages(mem->user_pages, bo->tbo.ttm->num_pages);
+free_out:
+ kvfree(mem->user_pages);
+ mem->user_pages = NULL;
+unregister_out:
+ if (ret)
+ amdgpu_mn_unregister(bo);
+out:
+ mutex_unlock(&process_info->lock);
+ return ret;
+}
+
/* Reserving a BO and its page table BOs must happen atomically to
* avoid deadlocks. Some operations update multiple VMs at once. Track
* all the reservation info in a context structure. Optionally a sync
}
static int map_bo_to_gpuvm(struct amdgpu_device *adev,
- struct kfd_bo_va_list *entry, struct amdgpu_sync *sync)
+ struct kfd_bo_va_list *entry, struct amdgpu_sync *sync,
+ bool no_update_pte)
{
int ret;
return ret;
}
+ if (no_update_pte)
+ return 0;
+
ret = update_gpuvm_pte(adev, entry, sync);
if (ret) {
pr_err("update_gpuvm_pte() failed\n");
mutex_init(&info->lock);
INIT_LIST_HEAD(&info->vm_list_head);
INIT_LIST_HEAD(&info->kfd_bo_list);
+ INIT_LIST_HEAD(&info->userptr_valid_list);
+ INIT_LIST_HEAD(&info->userptr_inval_list);
info->eviction_fence =
amdgpu_amdkfd_fence_create(dma_fence_context_alloc(1),
goto create_evict_fence_fail;
}
+ info->pid = get_task_pid(current->group_leader, PIDTYPE_PID);
+ atomic_set(&info->evicted_bos, 0);
+ INIT_DELAYED_WORK(&info->restore_userptr_work,
+ amdgpu_amdkfd_restore_userptr_worker);
+
*process_info = info;
*ef = dma_fence_get(&info->eviction_fence->base);
}
dma_fence_put(*ef);
*ef = NULL;
*process_info = NULL;
+ put_pid(info->pid);
create_evict_fence_fail:
mutex_destroy(&info->lock);
kfree(info);
/* Release per-process resources when last compute VM is destroyed */
if (!process_info->n_vms) {
WARN_ON(!list_empty(&process_info->kfd_bo_list));
+ WARN_ON(!list_empty(&process_info->userptr_valid_list));
+ WARN_ON(!list_empty(&process_info->userptr_inval_list));
dma_fence_put(&process_info->eviction_fence->base);
+ cancel_delayed_work_sync(&process_info->restore_userptr_work);
+ put_pid(process_info->pid);
mutex_destroy(&process_info->lock);
kfree(process_info);
}
{
struct amdgpu_device *adev = get_amdgpu_device(kgd);
struct amdgpu_vm *avm = (struct amdgpu_vm *)vm;
+ uint64_t user_addr = 0;
struct amdgpu_bo *bo;
+ struct amdgpu_bo_param bp;
int byte_align;
- u32 alloc_domain;
+ u32 domain, alloc_domain;
u64 alloc_flags;
uint32_t mapping_flags;
int ret;
* Check on which domain to allocate BO
*/
if (flags & ALLOC_MEM_FLAGS_VRAM) {
- alloc_domain = AMDGPU_GEM_DOMAIN_VRAM;
+ domain = alloc_domain = AMDGPU_GEM_DOMAIN_VRAM;
alloc_flags = AMDGPU_GEM_CREATE_VRAM_CLEARED;
alloc_flags |= (flags & ALLOC_MEM_FLAGS_PUBLIC) ?
AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED :
AMDGPU_GEM_CREATE_NO_CPU_ACCESS;
} else if (flags & ALLOC_MEM_FLAGS_GTT) {
- alloc_domain = AMDGPU_GEM_DOMAIN_GTT;
+ domain = alloc_domain = AMDGPU_GEM_DOMAIN_GTT;
alloc_flags = 0;
+ } else if (flags & ALLOC_MEM_FLAGS_USERPTR) {
+ domain = AMDGPU_GEM_DOMAIN_GTT;
+ alloc_domain = AMDGPU_GEM_DOMAIN_CPU;
+ alloc_flags = 0;
+ if (!offset || !*offset)
+ return -EINVAL;
+ user_addr = *offset;
} else {
return -EINVAL;
}
pr_debug("\tcreate BO VA 0x%llx size 0x%llx domain %s\n",
va, size, domain_string(alloc_domain));
- ret = amdgpu_bo_create(adev, size, byte_align,
- alloc_domain, alloc_flags, ttm_bo_type_device, NULL, &bo);
+ memset(&bp, 0, sizeof(bp));
+ bp.size = size;
+ bp.byte_align = byte_align;
+ bp.domain = alloc_domain;
+ bp.flags = alloc_flags;
+ bp.type = ttm_bo_type_device;
+ bp.resv = NULL;
+ ret = amdgpu_bo_create(adev, &bp, &bo);
if (ret) {
pr_debug("Failed to create BO on domain %s. ret %d\n",
domain_string(alloc_domain), ret);
}
bo->kfd_bo = *mem;
(*mem)->bo = bo;
+ if (user_addr)
+ bo->flags |= AMDGPU_AMDKFD_USERPTR_BO;
(*mem)->va = va;
- (*mem)->domain = alloc_domain;
+ (*mem)->domain = domain;
(*mem)->mapped_to_gpu_memory = 0;
(*mem)->process_info = avm->process_info;
- add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info);
+ add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info, user_addr);
+
+ if (user_addr) {
+ ret = init_user_pages(*mem, current->mm, user_addr);
+ if (ret) {
+ mutex_lock(&avm->process_info->lock);
+ list_del(&(*mem)->validate_list.head);
+ mutex_unlock(&avm->process_info->lock);
+ goto allocate_init_user_pages_failed;
+ }
+ }
if (offset)
*offset = amdgpu_bo_mmap_offset(bo);
return 0;
+allocate_init_user_pages_failed:
+ amdgpu_bo_unref(&bo);
+ /* Don't unreserve system mem limit twice */
+ goto err_reserve_system_mem;
err_bo_create:
unreserve_system_mem_limit(adev, size, alloc_domain);
err_reserve_system_mem:
* be freed anyway
*/
+ /* No more MMU notifiers */
+ amdgpu_mn_unregister(mem->bo);
+
/* Make sure restore workers don't access the BO any more */
bo_list_entry = &mem->validate_list;
mutex_lock(&process_info->lock);
list_del(&bo_list_entry->head);
mutex_unlock(&process_info->lock);
+ /* Free user pages if necessary */
+ if (mem->user_pages) {
+ pr_debug("%s: Freeing user_pages array\n", __func__);
+ if (mem->user_pages[0])
+ release_pages(mem->user_pages,
+ mem->bo->tbo.ttm->num_pages);
+ kvfree(mem->user_pages);
+ }
+
ret = reserve_bo_and_cond_vms(mem, NULL, BO_VM_ALL, &ctx);
if (unlikely(ret))
return ret;
struct kfd_bo_va_list *bo_va_entry = NULL;
struct kfd_bo_va_list *bo_va_entry_aql = NULL;
unsigned long bo_size;
-
- /* Make sure restore is not running concurrently.
- */
- mutex_lock(&mem->process_info->lock);
-
- mutex_lock(&mem->lock);
+ bool is_invalid_userptr = false;
bo = mem->bo;
-
if (!bo) {
pr_err("Invalid BO when mapping memory to GPU\n");
- ret = -EINVAL;
- goto out;
+ return -EINVAL;
+ }
+
+ /* Make sure restore is not running concurrently. Since we
+ * don't map invalid userptr BOs, we rely on the next restore
+ * worker to do the mapping
+ */
+ mutex_lock(&mem->process_info->lock);
+
+ /* Lock mmap-sem. If we find an invalid userptr BO, we can be
+ * sure that the MMU notifier is no longer running
+ * concurrently and the queues are actually stopped
+ */
+ if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) {
+ down_write(¤t->mm->mmap_sem);
+ is_invalid_userptr = atomic_read(&mem->invalid);
+ up_write(¤t->mm->mmap_sem);
}
+ mutex_lock(&mem->lock);
+
domain = mem->domain;
bo_size = bo->tbo.mem.size;
if (unlikely(ret))
goto out;
+ /* Userptr can be marked as "not invalid", but not actually be
+ * validated yet (still in the system domain). In that case
+ * the queues are still stopped and we can leave mapping for
+ * the next restore worker
+ */
+ if (bo->tbo.mem.mem_type == TTM_PL_SYSTEM)
+ is_invalid_userptr = true;
+
if (check_if_add_bo_to_vm(avm, mem)) {
ret = add_bo_to_vm(adev, mem, avm, false,
&bo_va_entry);
goto add_bo_to_vm_failed;
}
- if (mem->mapped_to_gpu_memory == 0) {
+ if (mem->mapped_to_gpu_memory == 0 &&
+ !amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) {
/* Validate BO only once. The eviction fence gets added to BO
* the first time it is mapped. Validate will wait for all
* background evictions to complete.
entry->va, entry->va + bo_size,
entry);
- ret = map_bo_to_gpuvm(adev, entry, ctx.sync);
+ ret = map_bo_to_gpuvm(adev, entry, ctx.sync,
+ is_invalid_userptr);
if (ret) {
pr_err("Failed to map radeon bo to gpuvm\n");
goto map_bo_to_gpuvm_failed;
return ret;
}
+/* Evict a userptr BO by stopping the queues if necessary
+ *
+ * Runs in MMU notifier, may be in RECLAIM_FS context. This means it
+ * cannot do any memory allocations, and cannot take any locks that
+ * are held elsewhere while allocating memory. Therefore this is as
+ * simple as possible, using atomic counters.
+ *
+ * It doesn't do anything to the BO itself. The real work happens in
+ * restore, where we get updated page addresses. This function only
+ * ensures that GPU access to the BO is stopped.
+ */
+int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem,
+ struct mm_struct *mm)
+{
+ struct amdkfd_process_info *process_info = mem->process_info;
+ int invalid, evicted_bos;
+ int r = 0;
+
+ invalid = atomic_inc_return(&mem->invalid);
+ evicted_bos = atomic_inc_return(&process_info->evicted_bos);
+ if (evicted_bos == 1) {
+ /* First eviction, stop the queues */
+ r = kgd2kfd->quiesce_mm(mm);
+ if (r)
+ pr_err("Failed to quiesce KFD\n");
+ schedule_delayed_work(&process_info->restore_userptr_work,
+ msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
+ }
+
+ return r;
+}
+
+/* Update invalid userptr BOs
+ *
+ * Moves invalidated (evicted) userptr BOs from userptr_valid_list to
+ * userptr_inval_list and updates user pages for all BOs that have
+ * been invalidated since their last update.
+ */
+static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
+ struct mm_struct *mm)
+{
+ struct kgd_mem *mem, *tmp_mem;
+ struct amdgpu_bo *bo;
+ struct ttm_operation_ctx ctx = { false, false };
+ int invalid, ret;
+
+ /* Move all invalidated BOs to the userptr_inval_list and
+ * release their user pages by migration to the CPU domain
+ */
+ list_for_each_entry_safe(mem, tmp_mem,
+ &process_info->userptr_valid_list,
+ validate_list.head) {
+ if (!atomic_read(&mem->invalid))
+ continue; /* BO is still valid */
+
+ bo = mem->bo;
+
+ if (amdgpu_bo_reserve(bo, true))
+ return -EAGAIN;
+ amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
+ ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+ amdgpu_bo_unreserve(bo);
+ if (ret) {
+ pr_err("%s: Failed to invalidate userptr BO\n",
+ __func__);
+ return -EAGAIN;
+ }
+
+ list_move_tail(&mem->validate_list.head,
+ &process_info->userptr_inval_list);
+ }
+
+ if (list_empty(&process_info->userptr_inval_list))
+ return 0; /* All evicted userptr BOs were freed */
+
+ /* Go through userptr_inval_list and update any invalid user_pages */
+ list_for_each_entry(mem, &process_info->userptr_inval_list,
+ validate_list.head) {
+ invalid = atomic_read(&mem->invalid);
+ if (!invalid)
+ /* BO hasn't been invalidated since the last
+ * revalidation attempt. Keep its BO list.
+ */
+ continue;
+
+ bo = mem->bo;
+
+ if (!mem->user_pages) {
+ mem->user_pages =
+ kvmalloc_array(bo->tbo.ttm->num_pages,
+ sizeof(struct page *),
+ GFP_KERNEL | __GFP_ZERO);
+ if (!mem->user_pages) {
+ pr_err("%s: Failed to allocate pages array\n",
+ __func__);
+ return -ENOMEM;
+ }
+ } else if (mem->user_pages[0]) {
+ release_pages(mem->user_pages, bo->tbo.ttm->num_pages);
+ }
+
+ /* Get updated user pages */
+ ret = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm,
+ mem->user_pages);
+ if (ret) {
+ mem->user_pages[0] = NULL;
+ pr_info("%s: Failed to get user pages: %d\n",
+ __func__, ret);
+ /* Pretend it succeeded. It will fail later
+ * with a VM fault if the GPU tries to access
+ * it. Better than hanging indefinitely with
+ * stalled user mode queues.
+ */
+ }
+
+ /* Mark the BO as valid unless it was invalidated
+ * again concurrently
+ */
+ if (atomic_cmpxchg(&mem->invalid, invalid, 0) != invalid)
+ return -EAGAIN;
+ }
+
+ return 0;
+}
+
+/* Validate invalid userptr BOs
+ *
+ * Validates BOs on the userptr_inval_list, and moves them back to the
+ * userptr_valid_list. Also updates GPUVM page tables with new page
+ * addresses and waits for the page table updates to complete.
+ */
+static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
+{
+ struct amdgpu_bo_list_entry *pd_bo_list_entries;
+ struct list_head resv_list, duplicates;
+ struct ww_acquire_ctx ticket;
+ struct amdgpu_sync sync;
+
+ struct amdgpu_vm *peer_vm;
+ struct kgd_mem *mem, *tmp_mem;
+ struct amdgpu_bo *bo;
+ struct ttm_operation_ctx ctx = { false, false };
+ int i, ret;
+
+ pd_bo_list_entries = kcalloc(process_info->n_vms,
+ sizeof(struct amdgpu_bo_list_entry),
+ GFP_KERNEL);
+ if (!pd_bo_list_entries) {
+ pr_err("%s: Failed to allocate PD BO list entries\n", __func__);
+ return -ENOMEM;
+ }
+
+ INIT_LIST_HEAD(&resv_list);
+ INIT_LIST_HEAD(&duplicates);
+
+ /* Get all the page directory BOs that need to be reserved */
+ i = 0;
+ list_for_each_entry(peer_vm, &process_info->vm_list_head,
+ vm_list_node)
+ amdgpu_vm_get_pd_bo(peer_vm, &resv_list,
+ &pd_bo_list_entries[i++]);
+ /* Add the userptr_inval_list entries to resv_list */
+ list_for_each_entry(mem, &process_info->userptr_inval_list,
+ validate_list.head) {
+ list_add_tail(&mem->resv_list.head, &resv_list);
+ mem->resv_list.bo = mem->validate_list.bo;
+ mem->resv_list.shared = mem->validate_list.shared;
+ }
+
+ /* Reserve all BOs and page tables for validation */
+ ret = ttm_eu_reserve_buffers(&ticket, &resv_list, false, &duplicates);
+ WARN(!list_empty(&duplicates), "Duplicates should be empty");
+ if (ret)
+ goto out;
+
+ amdgpu_sync_create(&sync);
+
+ /* Avoid triggering eviction fences when unmapping invalid
+ * userptr BOs (waits for all fences, doesn't use
+ * FENCE_OWNER_VM)
+ */
+ list_for_each_entry(peer_vm, &process_info->vm_list_head,
+ vm_list_node)
+ amdgpu_amdkfd_remove_eviction_fence(peer_vm->root.base.bo,
+ process_info->eviction_fence,
+ NULL, NULL);
+
+ ret = process_validate_vms(process_info);
+ if (ret)
+ goto unreserve_out;
+
+ /* Validate BOs and update GPUVM page tables */
+ list_for_each_entry_safe(mem, tmp_mem,
+ &process_info->userptr_inval_list,
+ validate_list.head) {
+ struct kfd_bo_va_list *bo_va_entry;
+
+ bo = mem->bo;
+
+ /* Copy pages array and validate the BO if we got user pages */
+ if (mem->user_pages[0]) {
+ amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm,
+ mem->user_pages);
+ amdgpu_ttm_placement_from_domain(bo, mem->domain);
+ ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+ if (ret) {
+ pr_err("%s: failed to validate BO\n", __func__);
+ goto unreserve_out;
+ }
+ }
+
+ /* Validate succeeded, now the BO owns the pages, free
+ * our copy of the pointer array. Put this BO back on
+ * the userptr_valid_list. If we need to revalidate
+ * it, we need to start from scratch.
+ */
+ kvfree(mem->user_pages);
+ mem->user_pages = NULL;
+ list_move_tail(&mem->validate_list.head,
+ &process_info->userptr_valid_list);
+
+ /* Update mapping. If the BO was not validated
+ * (because we couldn't get user pages), this will
+ * clear the page table entries, which will result in
+ * VM faults if the GPU tries to access the invalid
+ * memory.
+ */
+ list_for_each_entry(bo_va_entry, &mem->bo_va_list, bo_list) {
+ if (!bo_va_entry->is_mapped)
+ continue;
+
+ ret = update_gpuvm_pte((struct amdgpu_device *)
+ bo_va_entry->kgd_dev,
+ bo_va_entry, &sync);
+ if (ret) {
+ pr_err("%s: update PTE failed\n", __func__);
+ /* make sure this gets validated again */
+ atomic_inc(&mem->invalid);
+ goto unreserve_out;
+ }
+ }
+ }
+
+ /* Update page directories */
+ ret = process_update_pds(process_info, &sync);
+
+unreserve_out:
+ list_for_each_entry(peer_vm, &process_info->vm_list_head,
+ vm_list_node)
+ amdgpu_bo_fence(peer_vm->root.base.bo,
+ &process_info->eviction_fence->base, true);
+ ttm_eu_backoff_reservation(&ticket, &resv_list);
+ amdgpu_sync_wait(&sync, false);
+ amdgpu_sync_free(&sync);
+out:
+ kfree(pd_bo_list_entries);
+
+ return ret;
+}
+
+/* Worker callback to restore evicted userptr BOs
+ *
+ * Tries to update and validate all userptr BOs. If successful and no
+ * concurrent evictions happened, the queues are restarted. Otherwise,
+ * reschedule for another attempt later.
+ */
+static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work)
+{
+ struct delayed_work *dwork = to_delayed_work(work);
+ struct amdkfd_process_info *process_info =
+ container_of(dwork, struct amdkfd_process_info,
+ restore_userptr_work);
+ struct task_struct *usertask;
+ struct mm_struct *mm;
+ int evicted_bos;
+
+ evicted_bos = atomic_read(&process_info->evicted_bos);
+ if (!evicted_bos)
+ return;
+
+ /* Reference task and mm in case of concurrent process termination */
+ usertask = get_pid_task(process_info->pid, PIDTYPE_PID);
+ if (!usertask)
+ return;
+ mm = get_task_mm(usertask);
+ if (!mm) {
+ put_task_struct(usertask);
+ return;
+ }
+
+ mutex_lock(&process_info->lock);
+
+ if (update_invalid_user_pages(process_info, mm))
+ goto unlock_out;
+ /* userptr_inval_list can be empty if all evicted userptr BOs
+ * have been freed. In that case there is nothing to validate
+ * and we can just restart the queues.
+ */
+ if (!list_empty(&process_info->userptr_inval_list)) {
+ if (atomic_read(&process_info->evicted_bos) != evicted_bos)
+ goto unlock_out; /* Concurrent eviction, try again */
+
+ if (validate_invalid_user_pages(process_info))
+ goto unlock_out;
+ }
+ /* Final check for concurrent evicton and atomic update. If
+ * another eviction happens after successful update, it will
+ * be a first eviction that calls quiesce_mm. The eviction
+ * reference counting inside KFD will handle this case.
+ */
+ if (atomic_cmpxchg(&process_info->evicted_bos, evicted_bos, 0) !=
+ evicted_bos)
+ goto unlock_out;
+ evicted_bos = 0;
+ if (kgd2kfd->resume_mm(mm)) {
+ pr_err("%s: Failed to resume KFD\n", __func__);
+ /* No recovery from this failure. Probably the CP is
+ * hanging. No point trying again.
+ */
+ }
+unlock_out:
+ mutex_unlock(&process_info->lock);
+ mmput(mm);
+ put_task_struct(usertask);
+
+ /* If validation failed, reschedule another attempt */
+ if (evicted_bos)
+ schedule_delayed_work(&process_info->restore_userptr_work,
+ msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
+}
+
/** amdgpu_amdkfd_gpuvm_restore_process_bos - Restore all BOs for the given
* KFD process identified by process_info
*
p->bytes_moved += ctx.bytes_moved;
if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size &&
- bo->tbo.mem.mem_type == TTM_PL_VRAM &&
- bo->tbo.mem.start < adev->gmc.visible_vram_size >> PAGE_SHIFT)
+ amdgpu_bo_in_cpu_visible_vram(bo))
p->bytes_moved_vis += ctx.bytes_moved;
if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) {
struct amdgpu_bo_list_entry *candidate = p->evictable;
struct amdgpu_bo *bo = candidate->robj;
struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
- u64 initial_bytes_moved, bytes_moved;
bool update_bytes_moved_vis;
uint32_t other;
continue;
/* Good we can try to move this BO somewhere else */
- amdgpu_ttm_placement_from_domain(bo, other);
update_bytes_moved_vis =
adev->gmc.visible_vram_size < adev->gmc.real_vram_size &&
- bo->tbo.mem.mem_type == TTM_PL_VRAM &&
- bo->tbo.mem.start < adev->gmc.visible_vram_size >> PAGE_SHIFT;
- initial_bytes_moved = atomic64_read(&adev->num_bytes_moved);
+ amdgpu_bo_in_cpu_visible_vram(bo);
+ amdgpu_ttm_placement_from_domain(bo, other);
r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
- bytes_moved = atomic64_read(&adev->num_bytes_moved) -
- initial_bytes_moved;
- p->bytes_moved += bytes_moved;
+ p->bytes_moved += ctx.bytes_moved;
if (update_bytes_moved_vis)
- p->bytes_moved_vis += bytes_moved;
+ p->bytes_moved_vis += ctx.bytes_moved;
if (unlikely(r))
break;
if (p->bo_list) {
amdgpu_bo_list_get_list(p->bo_list, &p->validated);
if (p->bo_list->first_userptr != p->bo_list->num_entries)
- p->mn = amdgpu_mn_get(p->adev);
+ p->mn = amdgpu_mn_get(p->adev, AMDGPU_MN_TYPE_GFX);
}
INIT_LIST_HEAD(&duplicates);
ring = adev->mman.buffer_funcs_ring;
rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL];
r = drm_sched_entity_init(&ring->sched, &adev->mman.entity,
- rq, amdgpu_sched_jobs, NULL);
+ rq, NULL);
if (r) {
DRM_ERROR("Failed setting up TTM BO move run queue.\n");
goto error_entity;
if (!adev->mman.buffer_funcs_enabled) {
amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU);
} else if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size &&
- !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)) {
- unsigned fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
- struct drm_mm_node *node = bo->mem.mm_node;
- unsigned long pages_left;
-
- for (pages_left = bo->mem.num_pages;
- pages_left;
- pages_left -= node->size, node++) {
- if (node->start < fpfn)
- break;
- }
-
- if (!pages_left)
- goto gtt;
+ !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) &&
+ amdgpu_bo_in_cpu_visible_vram(abo)) {
/* Try evicting to the CPU inaccessible part of VRAM
* first, but only set GTT as busy placement, so this
*/
amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM |
AMDGPU_GEM_DOMAIN_GTT);
- abo->placements[0].fpfn = fpfn;
+ abo->placements[0].fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
abo->placements[0].lpfn = 0;
abo->placement.busy_placement = &abo->placements[1];
abo->placement.num_busy_placement = 1;
} else {
- gtt:
amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT);
}
break;
struct ttm_dma_tt ttm;
u64 offset;
uint64_t userptr;
- struct mm_struct *usermm;
+ struct task_struct *usertask;
uint32_t userflags;
spinlock_t guptasklock;
struct list_head guptasks;
int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
{
struct amdgpu_ttm_tt *gtt = (void *)ttm;
+ struct mm_struct *mm = gtt->usertask->mm;
unsigned int flags = 0;
unsigned pinned = 0;
int r;
+ if (!mm) /* Happens during process shutdown */
+ return -ESRCH;
+
if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY))
flags |= FOLL_WRITE;
- down_read(¤t->mm->mmap_sem);
+ down_read(&mm->mmap_sem);
if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) {
/* check that we only use anonymous memory
unsigned long end = gtt->userptr + ttm->num_pages * PAGE_SIZE;
struct vm_area_struct *vma;
- vma = find_vma(gtt->usermm, gtt->userptr);
+ vma = find_vma(mm, gtt->userptr);
if (!vma || vma->vm_file || vma->vm_end < end) {
- up_read(¤t->mm->mmap_sem);
+ up_read(&mm->mmap_sem);
return -EPERM;
}
}
list_add(&guptask.list, >t->guptasks);
spin_unlock(>t->guptasklock);
- r = get_user_pages(userptr, num_pages, flags, p, NULL);
+ if (mm == current->mm)
+ r = get_user_pages(userptr, num_pages, flags, p, NULL);
+ else
+ r = get_user_pages_remote(gtt->usertask,
+ mm, userptr, num_pages,
+ flags, p, NULL, NULL);
spin_lock(>t->guptasklock);
list_del(&guptask.list);
} while (pinned < ttm->num_pages);
- up_read(¤t->mm->mmap_sem);
+ up_read(&mm->mmap_sem);
return 0;
release_pages:
release_pages(pages, pinned);
- up_read(¤t->mm->mmap_sem);
+ up_read(&mm->mmap_sem);
return r;
}
sg_free_table(ttm->sg);
}
+ int amdgpu_ttm_gart_bind(struct amdgpu_device *adev,
+ struct ttm_buffer_object *tbo,
+ uint64_t flags)
+ {
+ struct amdgpu_bo *abo = ttm_to_amdgpu_bo(tbo);
+ struct ttm_tt *ttm = tbo->ttm;
+ struct amdgpu_ttm_tt *gtt = (void *)ttm;
+ int r;
+
+ if (abo->flags & AMDGPU_GEM_CREATE_MQD_GFX9) {
+ uint64_t page_idx = 1;
+
+ r = amdgpu_gart_bind(adev, gtt->offset, page_idx,
+ ttm->pages, gtt->ttm.dma_address, flags);
+ if (r)
+ goto gart_bind_fail;
+
+ /* Patch mtype of the second part BO */
+ flags &= ~AMDGPU_PTE_MTYPE_MASK;
+ flags |= AMDGPU_PTE_MTYPE(AMDGPU_MTYPE_NC);
+
+ r = amdgpu_gart_bind(adev,
+ gtt->offset + (page_idx << PAGE_SHIFT),
+ ttm->num_pages - page_idx,
+ &ttm->pages[page_idx],
+ &(gtt->ttm.dma_address[page_idx]), flags);
+ } else {
+ r = amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages,
+ ttm->pages, gtt->ttm.dma_address, flags);
+ }
+
+ gart_bind_fail:
+ if (r)
+ DRM_ERROR("failed to bind %lu pages at 0x%08llX\n",
+ ttm->num_pages, gtt->offset);
+
+ return r;
+ }
+
static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm,
struct ttm_mem_reg *bo_mem)
{
flags = amdgpu_ttm_tt_pte_flags(adev, bo->ttm, &tmp);
gtt->offset = (u64)tmp.start << PAGE_SHIFT;
- r = amdgpu_gart_bind(adev, gtt->offset, bo->ttm->num_pages,
- bo->ttm->pages, gtt->ttm.dma_address, flags);
+ r = amdgpu_ttm_gart_bind(adev, bo, flags);
if (unlikely(r)) {
ttm_bo_mem_put(bo, &tmp);
return r;
int amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo)
{
struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev);
- struct amdgpu_ttm_tt *gtt = (void *)tbo->ttm;
uint64_t flags;
int r;
- if (!gtt)
+ if (!tbo->ttm)
return 0;
- flags = amdgpu_ttm_tt_pte_flags(adev, >t->ttm.ttm, &tbo->mem);
- r = amdgpu_gart_bind(adev, gtt->offset, gtt->ttm.ttm.num_pages,
- gtt->ttm.ttm.pages, gtt->ttm.dma_address, flags);
- if (r)
- DRM_ERROR("failed to bind %lu pages at 0x%08llX\n",
- gtt->ttm.ttm.num_pages, gtt->offset);
+ flags = amdgpu_ttm_tt_pte_flags(adev, tbo->ttm, &tbo->mem);
+ r = amdgpu_ttm_gart_bind(adev, tbo, flags);
+
return r;
}
{
struct amdgpu_ttm_tt *gtt = (void *)ttm;
+ if (gtt->usertask)
+ put_task_struct(gtt->usertask);
+
ttm_dma_tt_fini(>t->ttm);
kfree(gtt);
}
return -EINVAL;
gtt->userptr = addr;
- gtt->usermm = current->mm;
gtt->userflags = flags;
+
+ if (gtt->usertask)
+ put_task_struct(gtt->usertask);
+ gtt->usertask = current->group_leader;
+ get_task_struct(gtt->usertask);
+
spin_lock_init(>t->guptasklock);
INIT_LIST_HEAD(>t->guptasks);
atomic_set(>t->mmu_invalidations, 0);
if (gtt == NULL)
return NULL;
- return gtt->usermm;
+ if (gtt->usertask == NULL)
+ return NULL;
+
+ return gtt->usertask->mm;
}
bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev)
{
struct ttm_operation_ctx ctx = { false, false };
+ struct amdgpu_bo_param bp;
int r = 0;
int i;
u64 vram_size = adev->gmc.visible_vram_size;
u64 size = adev->fw_vram_usage.size;
struct amdgpu_bo *bo;
+ memset(&bp, 0, sizeof(bp));
+ bp.size = adev->fw_vram_usage.size;
+ bp.byte_align = PAGE_SIZE;
+ bp.domain = AMDGPU_GEM_DOMAIN_VRAM;
+ bp.flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
+ AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+ bp.type = ttm_bo_type_kernel;
+ bp.resv = NULL;
adev->fw_vram_usage.va = NULL;
adev->fw_vram_usage.reserved_bo = NULL;
if (adev->fw_vram_usage.size > 0 &&
adev->fw_vram_usage.size <= vram_size) {
- r = amdgpu_bo_create(adev, adev->fw_vram_usage.size, PAGE_SIZE,
- AMDGPU_GEM_DOMAIN_VRAM,
- AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
- AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
- ttm_bo_type_kernel, NULL,
+ r = amdgpu_bo_create(adev, &bp,
&adev->fw_vram_usage.reserved_bo);
if (r)
goto error_create;
return r;
}
- 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;
+ if (adev->gmc.stolen_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->gmc.real_vram_size / (1024 * 1024)));
return 0;
}
+ void amdgpu_ttm_late_init(struct amdgpu_device *adev)
+ {
+ amdgpu_bo_free_kernel(&adev->stolen_vga_memory, NULL, NULL);
+ }
+
void amdgpu_ttm_fini(struct amdgpu_device *adev)
{
if (!adev->mman.initialized)
return;
amdgpu_ttm_debugfs_fini(adev);
- amdgpu_bo_free_kernel(&adev->stolen_vga_memory, NULL, NULL);
amdgpu_ttm_fw_reserve_vram_fini(adev);
if (adev->mman.aper_base_kaddr)
iounmap(adev->mman.aper_base_kaddr);
le32_to_cpu(rlc_hdr->reg_list_format_separate_array_offset_bytes));
DRM_DEBUG("reg_list_separate_size_bytes: %u\n",
le32_to_cpu(rlc_hdr->reg_list_separate_size_bytes));
- DRM_DEBUG("reg_list_separate_size_bytes: %u\n",
- le32_to_cpu(rlc_hdr->reg_list_separate_size_bytes));
+ DRM_DEBUG("reg_list_separate_array_offset_bytes: %u\n",
+ le32_to_cpu(rlc_hdr->reg_list_separate_array_offset_bytes));
+ if (version_minor == 1) {
+ const struct rlc_firmware_header_v2_1 *v2_1 =
+ container_of(rlc_hdr, struct rlc_firmware_header_v2_1, v2_0);
+ DRM_DEBUG("reg_list_format_direct_reg_list_length: %u\n",
+ le32_to_cpu(v2_1->reg_list_format_direct_reg_list_length));
+ DRM_DEBUG("save_restore_list_cntl_ucode_ver: %u\n",
+ le32_to_cpu(v2_1->save_restore_list_cntl_ucode_ver));
+ DRM_DEBUG("save_restore_list_cntl_feature_ver: %u\n",
+ le32_to_cpu(v2_1->save_restore_list_cntl_feature_ver));
+ DRM_DEBUG("save_restore_list_cntl_size_bytes %u\n",
+ le32_to_cpu(v2_1->save_restore_list_cntl_size_bytes));
+ DRM_DEBUG("save_restore_list_cntl_offset_bytes: %u\n",
+ le32_to_cpu(v2_1->save_restore_list_cntl_offset_bytes));
+ DRM_DEBUG("save_restore_list_gpm_ucode_ver: %u\n",
+ le32_to_cpu(v2_1->save_restore_list_gpm_ucode_ver));
+ DRM_DEBUG("save_restore_list_gpm_feature_ver: %u\n",
+ le32_to_cpu(v2_1->save_restore_list_gpm_feature_ver));
+ DRM_DEBUG("save_restore_list_gpm_size_bytes %u\n",
+ le32_to_cpu(v2_1->save_restore_list_gpm_size_bytes));
+ DRM_DEBUG("save_restore_list_gpm_offset_bytes: %u\n",
+ le32_to_cpu(v2_1->save_restore_list_gpm_offset_bytes));
+ DRM_DEBUG("save_restore_list_srm_ucode_ver: %u\n",
+ le32_to_cpu(v2_1->save_restore_list_srm_ucode_ver));
+ DRM_DEBUG("save_restore_list_srm_feature_ver: %u\n",
+ le32_to_cpu(v2_1->save_restore_list_srm_feature_ver));
+ DRM_DEBUG("save_restore_list_srm_size_bytes %u\n",
+ le32_to_cpu(v2_1->save_restore_list_srm_size_bytes));
+ DRM_DEBUG("save_restore_list_srm_offset_bytes: %u\n",
+ le32_to_cpu(v2_1->save_restore_list_srm_offset_bytes));
+ }
} else {
DRM_ERROR("Unknown RLC ucode version: %u.%u\n", version_major, version_minor);
}
case CHIP_POLARIS10:
case CHIP_POLARIS11:
case CHIP_POLARIS12:
+ case CHIP_VEGAM:
if (!load_type)
return AMDGPU_FW_LOAD_DIRECT;
else
else
return AMDGPU_FW_LOAD_PSP;
default:
- DRM_ERROR("Unknow firmware load type\n");
+ DRM_ERROR("Unknown firmware load type\n");
}
return AMDGPU_FW_LOAD_DIRECT;
(ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC1 &&
ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC2 &&
ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC1_JT &&
- ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC2_JT)) {
+ ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC2_JT &&
+ ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL &&
+ ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM &&
+ ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM)) {
ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes);
memcpy(ucode->kaddr, (void *)((uint8_t *)ucode->fw->data +
le32_to_cpu(header->ucode_array_offset_bytes) +
le32_to_cpu(cp_hdr->jt_offset) * 4),
ucode->ucode_size);
+ } else if (ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL) {
+ ucode->ucode_size = adev->gfx.rlc.save_restore_list_cntl_size_bytes;
+ memcpy(ucode->kaddr, adev->gfx.rlc.save_restore_list_cntl,
+ ucode->ucode_size);
+ } else if (ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM) {
+ ucode->ucode_size = adev->gfx.rlc.save_restore_list_gpm_size_bytes;
+ memcpy(ucode->kaddr, adev->gfx.rlc.save_restore_list_gpm,
+ ucode->ucode_size);
+ } else if (ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM) {
+ ucode->ucode_size = adev->gfx.rlc.save_restore_list_srm_size_bytes;
+ memcpy(ucode->kaddr, adev->gfx.rlc.save_restore_list_srm,
+ ucode->ucode_size);
}
return 0;
#define GFX9_MEC_HPD_SIZE 2048
#define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L
#define RLC_SAVE_RESTORE_ADDR_STARTING_OFFSET 0x00000000L
- #define GFX9_RLC_FORMAT_DIRECT_REG_LIST_LENGTH 34
#define mmPWR_MISC_CNTL_STATUS 0x0183
#define mmPWR_MISC_CNTL_STATUS_BASE_IDX 0
SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x01bd9f33, 0x01000000)
};
+ static const u32 GFX_RLC_SRM_INDEX_CNTL_ADDR_OFFSETS[] =
+ {
+ mmRLC_SRM_INDEX_CNTL_ADDR_0 - mmRLC_SRM_INDEX_CNTL_ADDR_0,
+ mmRLC_SRM_INDEX_CNTL_ADDR_1 - mmRLC_SRM_INDEX_CNTL_ADDR_0,
+ mmRLC_SRM_INDEX_CNTL_ADDR_2 - mmRLC_SRM_INDEX_CNTL_ADDR_0,
+ mmRLC_SRM_INDEX_CNTL_ADDR_3 - mmRLC_SRM_INDEX_CNTL_ADDR_0,
+ mmRLC_SRM_INDEX_CNTL_ADDR_4 - mmRLC_SRM_INDEX_CNTL_ADDR_0,
+ mmRLC_SRM_INDEX_CNTL_ADDR_5 - mmRLC_SRM_INDEX_CNTL_ADDR_0,
+ mmRLC_SRM_INDEX_CNTL_ADDR_6 - mmRLC_SRM_INDEX_CNTL_ADDR_0,
+ mmRLC_SRM_INDEX_CNTL_ADDR_7 - mmRLC_SRM_INDEX_CNTL_ADDR_0,
+ };
+
+ static const u32 GFX_RLC_SRM_INDEX_CNTL_DATA_OFFSETS[] =
+ {
+ mmRLC_SRM_INDEX_CNTL_DATA_0 - mmRLC_SRM_INDEX_CNTL_DATA_0,
+ mmRLC_SRM_INDEX_CNTL_DATA_1 - mmRLC_SRM_INDEX_CNTL_DATA_0,
+ mmRLC_SRM_INDEX_CNTL_DATA_2 - mmRLC_SRM_INDEX_CNTL_DATA_0,
+ mmRLC_SRM_INDEX_CNTL_DATA_3 - mmRLC_SRM_INDEX_CNTL_DATA_0,
+ mmRLC_SRM_INDEX_CNTL_DATA_4 - mmRLC_SRM_INDEX_CNTL_DATA_0,
+ mmRLC_SRM_INDEX_CNTL_DATA_5 - mmRLC_SRM_INDEX_CNTL_DATA_0,
+ mmRLC_SRM_INDEX_CNTL_DATA_6 - mmRLC_SRM_INDEX_CNTL_DATA_0,
+ mmRLC_SRM_INDEX_CNTL_DATA_7 - mmRLC_SRM_INDEX_CNTL_DATA_0,
+ };
+
#define VEGA10_GB_ADDR_CONFIG_GOLDEN 0x2a114042
#define VEGA12_GB_ADDR_CONFIG_GOLDEN 0x24104041
#define RAVEN_GB_ADDR_CONFIG_GOLDEN 0x24000042
kfree(adev->gfx.rlc.register_list_format);
}
+ static void gfx_v9_0_init_rlc_ext_microcode(struct amdgpu_device *adev)
+ {
+ const struct rlc_firmware_header_v2_1 *rlc_hdr;
+
+ rlc_hdr = (const struct rlc_firmware_header_v2_1 *)adev->gfx.rlc_fw->data;
+ adev->gfx.rlc_srlc_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_ucode_ver);
+ adev->gfx.rlc_srlc_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_feature_ver);
+ adev->gfx.rlc.save_restore_list_cntl_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_cntl_size_bytes);
+ adev->gfx.rlc.save_restore_list_cntl = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_cntl_offset_bytes);
+ adev->gfx.rlc_srlg_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_ucode_ver);
+ adev->gfx.rlc_srlg_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_feature_ver);
+ adev->gfx.rlc.save_restore_list_gpm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_gpm_size_bytes);
+ adev->gfx.rlc.save_restore_list_gpm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_gpm_offset_bytes);
+ adev->gfx.rlc_srls_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_ucode_ver);
+ adev->gfx.rlc_srls_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_feature_ver);
+ adev->gfx.rlc.save_restore_list_srm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_srm_size_bytes);
+ adev->gfx.rlc.save_restore_list_srm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_srm_offset_bytes);
+ adev->gfx.rlc.reg_list_format_direct_reg_list_length =
+ le32_to_cpu(rlc_hdr->reg_list_format_direct_reg_list_length);
+ }
+
static int gfx_v9_0_init_microcode(struct amdgpu_device *adev)
{
const char *chip_name;
const struct rlc_firmware_header_v2_0 *rlc_hdr;
unsigned int *tmp = NULL;
unsigned int i = 0;
+ uint16_t version_major;
+ uint16_t version_minor;
DRM_DEBUG("\n");
goto out;
err = amdgpu_ucode_validate(adev->gfx.rlc_fw);
rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+
+ version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
+ version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
+ if (version_major == 2 && version_minor == 1)
+ adev->gfx.rlc.is_rlc_v2_1 = true;
+
adev->gfx.rlc_fw_version = le32_to_cpu(rlc_hdr->header.ucode_version);
adev->gfx.rlc_feature_version = le32_to_cpu(rlc_hdr->ucode_feature_version);
adev->gfx.rlc.save_and_restore_offset =
for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++)
adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]);
+ if (adev->gfx.rlc.is_rlc_v2_1)
+ gfx_v9_0_init_rlc_ext_microcode(adev);
+
snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
if (err)
adev->firmware.fw_size +=
ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE);
+ if (adev->gfx.rlc.is_rlc_v2_1) {
+ info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL];
+ info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL;
+ info->fw = adev->gfx.rlc_fw;
+ adev->firmware.fw_size +=
+ ALIGN(adev->gfx.rlc.save_restore_list_cntl_size_bytes, PAGE_SIZE);
+
+ info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM];
+ info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM;
+ info->fw = adev->gfx.rlc_fw;
+ adev->firmware.fw_size +=
+ ALIGN(adev->gfx.rlc.save_restore_list_gpm_size_bytes, PAGE_SIZE);
+
+ info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM];
+ info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM;
+ info->fw = adev->gfx.rlc_fw;
+ adev->firmware.fw_size +=
+ ALIGN(adev->gfx.rlc.save_restore_list_srm_size_bytes, PAGE_SIZE);
+ }
+
info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1];
info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1;
info->fw = adev->gfx.mec_fw;
gfx_v9_0_setup_rb(adev);
gfx_v9_0_get_cu_info(adev, &adev->gfx.cu_info);
+ adev->gfx.config.db_debug2 = RREG32_SOC15(GC, 0, mmDB_DEBUG2);
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
SH_MEM_ALIGNMENT_MODE_UNALIGNED);
WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, tmp);
- tmp = adev->gmc.shared_aperture_start >> 48;
+ tmp = REG_SET_FIELD(0, SH_MEM_BASES, PRIVATE_BASE,
+ (adev->gmc.private_aperture_start >> 48));
+ tmp = REG_SET_FIELD(tmp, SH_MEM_BASES, SHARED_BASE,
+ (adev->gmc.shared_aperture_start >> 48));
WREG32_SOC15(GC, 0, mmSH_MEM_BASES, tmp);
}
}
adev->gfx.rlc.clear_state_size);
}
- static void gfx_v9_0_parse_ind_reg_list(int *register_list_format,
+ static void gfx_v9_1_parse_ind_reg_list(int *register_list_format,
int indirect_offset,
int list_size,
int *unique_indirect_regs,
int *unique_indirect_reg_count,
- int max_indirect_reg_count,
int *indirect_start_offsets,
- int *indirect_start_offsets_count,
- int max_indirect_start_offsets_count)
+ int *indirect_start_offsets_count)
{
int idx;
- bool new_entry = true;
for (; indirect_offset < list_size; indirect_offset++) {
+ indirect_start_offsets[*indirect_start_offsets_count] = indirect_offset;
+ *indirect_start_offsets_count = *indirect_start_offsets_count + 1;
- if (new_entry) {
- new_entry = false;
- indirect_start_offsets[*indirect_start_offsets_count] = indirect_offset;
- *indirect_start_offsets_count = *indirect_start_offsets_count + 1;
- BUG_ON(*indirect_start_offsets_count >= max_indirect_start_offsets_count);
- }
+ while (register_list_format[indirect_offset] != 0xFFFFFFFF) {
+ indirect_offset += 2;
- if (register_list_format[indirect_offset] == 0xFFFFFFFF) {
- new_entry = true;
- continue;
- }
+ /* look for the matching indice */
+ for (idx = 0; idx < *unique_indirect_reg_count; idx++) {
+ if (unique_indirect_regs[idx] ==
+ register_list_format[indirect_offset] ||
+ !unique_indirect_regs[idx])
+ break;
+ }
- indirect_offset += 2;
+ BUG_ON(idx >= *unique_indirect_reg_count);
- /* look for the matching indice */
- for (idx = 0; idx < *unique_indirect_reg_count; idx++) {
- if (unique_indirect_regs[idx] ==
- register_list_format[indirect_offset])
- break;
- }
+ if (!unique_indirect_regs[idx])
+ unique_indirect_regs[idx] = register_list_format[indirect_offset];
- if (idx >= *unique_indirect_reg_count) {
- unique_indirect_regs[*unique_indirect_reg_count] =
- register_list_format[indirect_offset];
- idx = *unique_indirect_reg_count;
- *unique_indirect_reg_count = *unique_indirect_reg_count + 1;
- BUG_ON(*unique_indirect_reg_count >= max_indirect_reg_count);
+ indirect_offset++;
}
-
- register_list_format[indirect_offset] = idx;
}
}
- static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev)
+ static int gfx_v9_1_init_rlc_save_restore_list(struct amdgpu_device *adev)
{
int unique_indirect_regs[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0};
int unique_indirect_reg_count = 0;
int indirect_start_offsets_count = 0;
int list_size = 0;
- int i = 0;
+ int i = 0, j = 0;
u32 tmp = 0;
u32 *register_list_format =
adev->gfx.rlc.reg_list_format_size_bytes);
/* setup unique_indirect_regs array and indirect_start_offsets array */
- gfx_v9_0_parse_ind_reg_list(register_list_format,
- GFX9_RLC_FORMAT_DIRECT_REG_LIST_LENGTH,
- adev->gfx.rlc.reg_list_format_size_bytes >> 2,
- unique_indirect_regs,
- &unique_indirect_reg_count,
- ARRAY_SIZE(unique_indirect_regs),
- indirect_start_offsets,
- &indirect_start_offsets_count,
- ARRAY_SIZE(indirect_start_offsets));
+ unique_indirect_reg_count = ARRAY_SIZE(unique_indirect_regs);
+ gfx_v9_1_parse_ind_reg_list(register_list_format,
+ adev->gfx.rlc.reg_list_format_direct_reg_list_length,
+ adev->gfx.rlc.reg_list_format_size_bytes >> 2,
+ unique_indirect_regs,
+ &unique_indirect_reg_count,
+ indirect_start_offsets,
+ &indirect_start_offsets_count);
/* enable auto inc in case it is disabled */
tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL));
WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_ARAM_DATA),
adev->gfx.rlc.register_restore[i]);
- /* load direct register */
- WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_ARAM_ADDR), 0);
- for (i = 0; i < adev->gfx.rlc.reg_list_size_bytes >> 2; i++)
- WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_ARAM_DATA),
- adev->gfx.rlc.register_restore[i]);
-
/* load indirect register */
WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_ADDR),
adev->gfx.rlc.reg_list_format_start);
- for (i = 0; i < adev->gfx.rlc.reg_list_format_size_bytes >> 2; i++)
+
+ /* direct register portion */
+ for (i = 0; i < adev->gfx.rlc.reg_list_format_direct_reg_list_length; i++)
WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_DATA),
register_list_format[i]);
+ /* indirect register portion */
+ while (i < (adev->gfx.rlc.reg_list_format_size_bytes >> 2)) {
+ if (register_list_format[i] == 0xFFFFFFFF) {
+ WREG32_SOC15(GC, 0, mmRLC_GPM_SCRATCH_DATA, register_list_format[i++]);
+ continue;
+ }
+
+ WREG32_SOC15(GC, 0, mmRLC_GPM_SCRATCH_DATA, register_list_format[i++]);
+ WREG32_SOC15(GC, 0, mmRLC_GPM_SCRATCH_DATA, register_list_format[i++]);
+
+ for (j = 0; j < unique_indirect_reg_count; j++) {
+ if (register_list_format[i] == unique_indirect_regs[j]) {
+ WREG32_SOC15(GC, 0, mmRLC_GPM_SCRATCH_DATA, j);
+ break;
+ }
+ }
+
+ BUG_ON(j >= unique_indirect_reg_count);
+
+ i++;
+ }
+
/* set save/restore list size */
list_size = adev->gfx.rlc.reg_list_size_bytes >> 2;
list_size = list_size >> 1;
adev->gfx.rlc.starting_offsets_start);
for (i = 0; i < ARRAY_SIZE(indirect_start_offsets); i++)
WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_DATA),
- indirect_start_offsets[i]);
+ indirect_start_offsets[i]);
/* load unique indirect regs*/
for (i = 0; i < ARRAY_SIZE(unique_indirect_regs); i++) {
- WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_ADDR_0) + i,
- unique_indirect_regs[i] & 0x3FFFF);
- WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_DATA_0) + i,
- unique_indirect_regs[i] >> 20);
+ if (unique_indirect_regs[i] != 0) {
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_ADDR_0)
+ + GFX_RLC_SRM_INDEX_CNTL_ADDR_OFFSETS[i],
+ unique_indirect_regs[i] & 0x3FFFF);
+
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_DATA_0)
+ + GFX_RLC_SRM_INDEX_CNTL_DATA_OFFSETS[i],
+ unique_indirect_regs[i] >> 20);
+ }
}
kfree(register_list_format);
static void gfx_v9_0_init_pg(struct amdgpu_device *adev)
{
+ if (!adev->gfx.rlc.is_rlc_v2_1)
+ return;
+
if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_PG |
AMD_PG_SUPPORT_GFX_SMG |
AMD_PG_SUPPORT_GFX_DMG |
AMD_PG_SUPPORT_GDS |
AMD_PG_SUPPORT_RLC_SMU_HS)) {
gfx_v9_0_init_csb(adev);
- gfx_v9_0_init_rlc_save_restore_list(adev);
+ gfx_v9_1_init_rlc_save_restore_list(adev);
gfx_v9_0_enable_save_restore_machine(adev);
- if (adev->asic_type == CHIP_RAVEN) {
- WREG32(mmRLC_JUMP_TABLE_RESTORE,
- adev->gfx.rlc.cp_table_gpu_addr >> 8);
- gfx_v9_0_init_gfx_power_gating(adev);
-
- if (adev->pg_flags & AMD_PG_SUPPORT_RLC_SMU_HS) {
- gfx_v9_0_enable_sck_slow_down_on_power_up(adev, true);
- gfx_v9_0_enable_sck_slow_down_on_power_down(adev, true);
- } else {
- gfx_v9_0_enable_sck_slow_down_on_power_up(adev, false);
- gfx_v9_0_enable_sck_slow_down_on_power_down(adev, false);
- }
-
- if (adev->pg_flags & AMD_PG_SUPPORT_CP)
- gfx_v9_0_enable_cp_power_gating(adev, true);
- else
- gfx_v9_0_enable_cp_power_gating(adev, false);
- }
+ WREG32(mmRLC_JUMP_TABLE_RESTORE,
+ adev->gfx.rlc.cp_table_gpu_addr >> 8);
+ gfx_v9_0_init_gfx_power_gating(adev);
}
}
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
int i;
+ amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_GFX,
+ AMD_PG_STATE_UNGATE);
+
amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
if (r)
return r;
+ r = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_GFX,
+ AMD_PG_STATE_GATE);
+ if (r)
+ return r;
+
return 0;
}
static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev,
bool enable)
{
- /* TODO: double check if we need to perform under safe mdoe */
- /* gfx_v9_0_enter_rlc_safe_mode(adev); */
+ gfx_v9_0_enter_rlc_safe_mode(adev);
if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) && enable) {
gfx_v9_0_enable_gfx_cg_power_gating(adev, true);
gfx_v9_0_enable_gfx_pipeline_powergating(adev, false);
}
- /* gfx_v9_0_exit_rlc_safe_mode(adev); */
+ gfx_v9_0_exit_rlc_safe_mode(adev);
}
static void gfx_v9_0_update_gfx_mg_power_gating(struct amdgpu_device *adev,
}
amdgpu_ring_write(ring, header);
- BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
+ BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
amdgpu_ring_write(ring,
#ifdef __BIG_ENDIAN
(2 << 0) |
{
bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
+ bool writeback = flags & AMDGPU_FENCE_FLAG_TC_WB_ONLY;
/* RELEASE_MEM - flush caches, send int */
amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6));
- amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
- EOP_TC_ACTION_EN |
- EOP_TC_WB_ACTION_EN |
- EOP_TC_MD_ACTION_EN |
+ amdgpu_ring_write(ring, ((writeback ? (EOP_TC_WB_ACTION_EN |
+ EOP_TC_NC_ACTION_EN) :
+ (EOP_TCL1_ACTION_EN |
+ EOP_TC_ACTION_EN |
+ EOP_TC_WB_ACTION_EN |
+ EOP_TC_MD_ACTION_EN)) |
EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
EVENT_INDEX(5)));
amdgpu_ring_write(ring, DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0));
gfx_v9_0_wait_reg_mem(ring, 0, 0, 0, reg, 0, val, mask, 0x20);
}
+ static void gfx_v9_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
+ uint32_t reg0, uint32_t reg1,
+ uint32_t ref, uint32_t mask)
+ {
+ int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
+
+ if (amdgpu_sriov_vf(ring->adev))
+ gfx_v9_0_wait_reg_mem(ring, usepfp, 0, 1, reg0, reg1,
+ ref, mask, 0x20);
+ else
+ amdgpu_ring_emit_reg_write_reg_wait_helper(ring, reg0, reg1,
+ ref, mask);
+ }
+
static void gfx_v9_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
enum amdgpu_interrupt_state state)
{
.emit_tmz = gfx_v9_0_ring_emit_tmz,
.emit_wreg = gfx_v9_0_ring_emit_wreg,
.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
+ .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
};
static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
.set_priority = gfx_v9_0_ring_set_priority_compute,
.emit_wreg = gfx_v9_0_ring_emit_wreg,
.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
+ .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
};
static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
.emit_rreg = gfx_v9_0_ring_emit_rreg,
.emit_wreg = gfx_v9_0_ring_emit_wreg,
.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
+ .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
};
static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev)
cu_info->number = active_cu_number;
cu_info->ao_cu_mask = ao_cu_mask;
+ cu_info->simd_per_cu = NUM_SIMD_PER_CU;
return 0;
}
#define EOP_TC_WB_ACTION_EN (1 << 15) /* L2 */
#define EOP_TCL1_ACTION_EN (1 << 16)
#define EOP_TC_ACTION_EN (1 << 17) /* L2 */
+ #define EOP_TC_NC_ACTION_EN (1 << 19)
#define EOP_TC_MD_ACTION_EN (1 << 21) /* L2 metadata */
#define DATA_SEL(x) ((x) << 29)
* x=1: tmz_end
*/
+#define PACKET3_INVALIDATE_TLBS 0x98
+# define PACKET3_INVALIDATE_TLBS_DST_SEL(x) ((x) << 0)
+# define PACKET3_INVALIDATE_TLBS_ALL_HUB(x) ((x) << 4)
+# define PACKET3_INVALIDATE_TLBS_PASID(x) ((x) << 5)
+# define PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(x) ((x) << 29)
#define PACKET3_SET_RESOURCES 0xA0
/* 1. header
* 2. CONTROL
.end_use = amdgpu_vce_ring_end_use,
.emit_wreg = vce_v4_0_emit_wreg,
.emit_reg_wait = vce_v4_0_emit_reg_wait,
+ .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
};
static void vce_v4_0_set_ring_funcs(struct amdgpu_device *adev)
******************************************************************************/
struct dc *dc_create(const struct dc_init_data *init_params)
- {
+{
struct dc *dc = kzalloc(sizeof(*dc), GFP_KERNEL);
unsigned int full_pipe_count;
return true;
}
- /*
- * TODO this whole function needs to go
- *
- * dc_surface_update is needlessly complex. See if we can just replace this
- * with a dc_plane_state and follow the atomic model a bit more closely here.
- */
- bool dc_commit_planes_to_stream(
- struct dc *dc,
- struct dc_plane_state **plane_states,
- uint8_t new_plane_count,
- struct dc_stream_state *dc_stream,
- struct dc_state *state)
- {
- /* no need to dynamically allocate this. it's pretty small */
- struct dc_surface_update updates[MAX_SURFACES];
- struct dc_flip_addrs *flip_addr;
- struct dc_plane_info *plane_info;
- struct dc_scaling_info *scaling_info;
- int i;
- struct dc_stream_update *stream_update =
- kzalloc(sizeof(struct dc_stream_update), GFP_KERNEL);
-
- if (!stream_update) {
- BREAK_TO_DEBUGGER();
- return false;
- }
-
- flip_addr = kcalloc(MAX_SURFACES, sizeof(struct dc_flip_addrs),
- GFP_KERNEL);
- plane_info = kcalloc(MAX_SURFACES, sizeof(struct dc_plane_info),
- GFP_KERNEL);
- scaling_info = kcalloc(MAX_SURFACES, sizeof(struct dc_scaling_info),
- GFP_KERNEL);
-
- if (!flip_addr || !plane_info || !scaling_info) {
- kfree(flip_addr);
- kfree(plane_info);
- kfree(scaling_info);
- kfree(stream_update);
- return false;
- }
-
- memset(updates, 0, sizeof(updates));
-
- stream_update->src = dc_stream->src;
- stream_update->dst = dc_stream->dst;
- stream_update->out_transfer_func = dc_stream->out_transfer_func;
-
- for (i = 0; i < new_plane_count; i++) {
- updates[i].surface = plane_states[i];
- updates[i].gamma =
- (struct dc_gamma *)plane_states[i]->gamma_correction;
- updates[i].in_transfer_func = plane_states[i]->in_transfer_func;
- flip_addr[i].address = plane_states[i]->address;
- flip_addr[i].flip_immediate = plane_states[i]->flip_immediate;
- plane_info[i].color_space = plane_states[i]->color_space;
- plane_info[i].input_tf = plane_states[i]->input_tf;
- plane_info[i].format = plane_states[i]->format;
- plane_info[i].plane_size = plane_states[i]->plane_size;
- plane_info[i].rotation = plane_states[i]->rotation;
- plane_info[i].horizontal_mirror = plane_states[i]->horizontal_mirror;
- plane_info[i].stereo_format = plane_states[i]->stereo_format;
- plane_info[i].tiling_info = plane_states[i]->tiling_info;
- plane_info[i].visible = plane_states[i]->visible;
- plane_info[i].per_pixel_alpha = plane_states[i]->per_pixel_alpha;
- plane_info[i].dcc = plane_states[i]->dcc;
- scaling_info[i].scaling_quality = plane_states[i]->scaling_quality;
- scaling_info[i].src_rect = plane_states[i]->src_rect;
- scaling_info[i].dst_rect = plane_states[i]->dst_rect;
- scaling_info[i].clip_rect = plane_states[i]->clip_rect;
-
- updates[i].flip_addr = &flip_addr[i];
- updates[i].plane_info = &plane_info[i];
- updates[i].scaling_info = &scaling_info[i];
- }
-
- dc_commit_updates_for_stream(
- dc,
- updates,
- new_plane_count,
- dc_stream, stream_update, plane_states, state);
-
- kfree(flip_addr);
- kfree(plane_info);
- kfree(scaling_info);
- kfree(stream_update);
- return true;
- }
-
struct dc_state *dc_create_state(void)
{
struct dc_state *context = kzalloc(sizeof(struct dc_state),
if (u->plane_info->color_space != u->surface->color_space)
update_flags->bits.color_space_change = 1;
- if (u->plane_info->input_tf != u->surface->input_tf)
- update_flags->bits.input_tf_change = 1;
-
if (u->plane_info->horizontal_mirror != u->surface->horizontal_mirror)
update_flags->bits.horizontal_mirror_change = 1;
if (u->input_csc_color_matrix)
update_flags->bits.input_csc_change = 1;
- if (update_flags->bits.in_transfer_func_change
- || update_flags->bits.input_csc_change) {
+ if (u->coeff_reduction_factor)
+ update_flags->bits.coeff_reduction_change = 1;
+
+ if (update_flags->bits.in_transfer_func_change) {
type = UPDATE_TYPE_MED;
elevate_update_type(&overall_type, type);
}
+ if (update_flags->bits.input_csc_change
+ || update_flags->bits.coeff_reduction_change) {
+ type = UPDATE_TYPE_FULL;
+ elevate_update_type(&overall_type, type);
+ }
+
return overall_type;
}
type = check_update_surfaces_for_stream(dc, updates, surface_count, stream_update, stream_status);
if (type == UPDATE_TYPE_FULL)
for (i = 0; i < surface_count; i++)
- updates[i].surface->update_flags.bits.full_update = 1;
+ updates[i].surface->update_flags.raw = 0xFFFFFFFF;
return type;
}
pipe_ctx->stream_res.abm->funcs->set_abm_level(
pipe_ctx->stream_res.abm, stream->abm_level);
}
+
+ if (stream_update && stream_update->periodic_fn_vsync_delta &&
+ pipe_ctx->stream_res.tg->funcs->program_vline_interrupt)
+ pipe_ctx->stream_res.tg->funcs->program_vline_interrupt(
+ pipe_ctx->stream_res.tg, &pipe_ctx->stream->timing,
+ pipe_ctx->stream->periodic_fn_vsync_delta);
}
}
--- /dev/null
- 32, NULL);
+// SPDX-License-Identifier: GPL-2.0+
+/* Copyright (C) 2014-2018 Broadcom */
+
+/**
+ * DOC: Broadcom V3D Graphics Driver
+ *
+ * This driver supports the Broadcom V3D 3.3 and 4.1 OpenGL ES GPUs.
+ * For V3D 2.x support, see the VC4 driver.
+ *
+ * Currently only single-core rendering using the binner and renderer
+ * is supported. The TFU (texture formatting unit) and V3D 4.x's CSD
+ * (compute shader dispatch) are not yet supported.
+ */
+
+#include <linux/clk.h>
+#include <linux/device.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/of_platform.h>
+#include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
+#include <drm/drm_fb_cma_helper.h>
+#include <drm/drm_fb_helper.h>
+
+#include "uapi/drm/v3d_drm.h"
+#include "v3d_drv.h"
+#include "v3d_regs.h"
+
+#define DRIVER_NAME "v3d"
+#define DRIVER_DESC "Broadcom V3D graphics"
+#define DRIVER_DATE "20180419"
+#define DRIVER_MAJOR 1
+#define DRIVER_MINOR 0
+#define DRIVER_PATCHLEVEL 0
+
+#ifdef CONFIG_PM
+static int v3d_runtime_suspend(struct device *dev)
+{
+ struct drm_device *drm = dev_get_drvdata(dev);
+ struct v3d_dev *v3d = to_v3d_dev(drm);
+
+ v3d_irq_disable(v3d);
+
+ clk_disable_unprepare(v3d->clk);
+
+ return 0;
+}
+
+static int v3d_runtime_resume(struct device *dev)
+{
+ struct drm_device *drm = dev_get_drvdata(dev);
+ struct v3d_dev *v3d = to_v3d_dev(drm);
+ int ret;
+
+ ret = clk_prepare_enable(v3d->clk);
+ if (ret != 0)
+ return ret;
+
+ /* XXX: VPM base */
+
+ v3d_mmu_set_page_table(v3d);
+ v3d_irq_enable(v3d);
+
+ return 0;
+}
+#endif
+
+static const struct dev_pm_ops v3d_v3d_pm_ops = {
+ SET_RUNTIME_PM_OPS(v3d_runtime_suspend, v3d_runtime_resume, NULL)
+};
+
+static int v3d_get_param_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file_priv)
+{
+ struct v3d_dev *v3d = to_v3d_dev(dev);
+ struct drm_v3d_get_param *args = data;
+ int ret;
+ static const u32 reg_map[] = {
+ [DRM_V3D_PARAM_V3D_UIFCFG] = V3D_HUB_UIFCFG,
+ [DRM_V3D_PARAM_V3D_HUB_IDENT1] = V3D_HUB_IDENT1,
+ [DRM_V3D_PARAM_V3D_HUB_IDENT2] = V3D_HUB_IDENT2,
+ [DRM_V3D_PARAM_V3D_HUB_IDENT3] = V3D_HUB_IDENT3,
+ [DRM_V3D_PARAM_V3D_CORE0_IDENT0] = V3D_CTL_IDENT0,
+ [DRM_V3D_PARAM_V3D_CORE0_IDENT1] = V3D_CTL_IDENT1,
+ [DRM_V3D_PARAM_V3D_CORE0_IDENT2] = V3D_CTL_IDENT2,
+ };
+
+ if (args->pad != 0)
+ return -EINVAL;
+
+ /* Note that DRM_V3D_PARAM_V3D_CORE0_IDENT0 is 0, so we need
+ * to explicitly allow it in the "the register in our
+ * parameter map" check.
+ */
+ if (args->param < ARRAY_SIZE(reg_map) &&
+ (reg_map[args->param] ||
+ args->param == DRM_V3D_PARAM_V3D_CORE0_IDENT0)) {
+ u32 offset = reg_map[args->param];
+
+ if (args->value != 0)
+ return -EINVAL;
+
+ ret = pm_runtime_get_sync(v3d->dev);
+ if (args->param >= DRM_V3D_PARAM_V3D_CORE0_IDENT0 &&
+ args->param <= DRM_V3D_PARAM_V3D_CORE0_IDENT2) {
+ args->value = V3D_CORE_READ(0, offset);
+ } else {
+ args->value = V3D_READ(offset);
+ }
+ pm_runtime_mark_last_busy(v3d->dev);
+ pm_runtime_put_autosuspend(v3d->dev);
+ return 0;
+ }
+
+ /* Any params that aren't just register reads would go here. */
+
+ DRM_DEBUG("Unknown parameter %d\n", args->param);
+ return -EINVAL;
+}
+
+static int
+v3d_open(struct drm_device *dev, struct drm_file *file)
+{
+ struct v3d_dev *v3d = to_v3d_dev(dev);
+ struct v3d_file_priv *v3d_priv;
+ int i;
+
+ v3d_priv = kzalloc(sizeof(*v3d_priv), GFP_KERNEL);
+ if (!v3d_priv)
+ return -ENOMEM;
+
+ v3d_priv->v3d = v3d;
+
+ for (i = 0; i < V3D_MAX_QUEUES; i++) {
+ drm_sched_entity_init(&v3d->queue[i].sched,
+ &v3d_priv->sched_entity[i],
+ &v3d->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL],
++ NULL);
+ }
+
+ file->driver_priv = v3d_priv;
+
+ return 0;
+}
+
+static void
+v3d_postclose(struct drm_device *dev, struct drm_file *file)
+{
+ struct v3d_dev *v3d = to_v3d_dev(dev);
+ struct v3d_file_priv *v3d_priv = file->driver_priv;
+ enum v3d_queue q;
+
+ for (q = 0; q < V3D_MAX_QUEUES; q++) {
+ drm_sched_entity_fini(&v3d->queue[q].sched,
+ &v3d_priv->sched_entity[q]);
+ }
+
+ kfree(v3d_priv);
+}
+
+static const struct file_operations v3d_drm_fops = {
+ .owner = THIS_MODULE,
+ .open = drm_open,
+ .release = drm_release,
+ .unlocked_ioctl = drm_ioctl,
+ .mmap = v3d_mmap,
+ .poll = drm_poll,
+ .read = drm_read,
+ .compat_ioctl = drm_compat_ioctl,
+ .llseek = noop_llseek,
+};
+
+/* DRM_AUTH is required on SUBMIT_CL for now, while we don't have GMP
+ * protection between clients. Note that render nodes would be be
+ * able to submit CLs that could access BOs from clients authenticated
+ * with the master node.
+ */
+static const struct drm_ioctl_desc v3d_drm_ioctls[] = {
+ DRM_IOCTL_DEF_DRV(V3D_SUBMIT_CL, v3d_submit_cl_ioctl, DRM_RENDER_ALLOW | DRM_AUTH),
+ DRM_IOCTL_DEF_DRV(V3D_WAIT_BO, v3d_wait_bo_ioctl, DRM_RENDER_ALLOW),
+ DRM_IOCTL_DEF_DRV(V3D_CREATE_BO, v3d_create_bo_ioctl, DRM_RENDER_ALLOW),
+ DRM_IOCTL_DEF_DRV(V3D_MMAP_BO, v3d_mmap_bo_ioctl, DRM_RENDER_ALLOW),
+ DRM_IOCTL_DEF_DRV(V3D_GET_PARAM, v3d_get_param_ioctl, DRM_RENDER_ALLOW),
+ DRM_IOCTL_DEF_DRV(V3D_GET_BO_OFFSET, v3d_get_bo_offset_ioctl, DRM_RENDER_ALLOW),
+};
+
+static const struct vm_operations_struct v3d_vm_ops = {
+ .fault = v3d_gem_fault,
+ .open = drm_gem_vm_open,
+ .close = drm_gem_vm_close,
+};
+
+static struct drm_driver v3d_drm_driver = {
+ .driver_features = (DRIVER_GEM |
+ DRIVER_RENDER |
+ DRIVER_PRIME |
+ DRIVER_SYNCOBJ),
+
+ .open = v3d_open,
+ .postclose = v3d_postclose,
+
+#if defined(CONFIG_DEBUG_FS)
+ .debugfs_init = v3d_debugfs_init,
+#endif
+
+ .gem_free_object_unlocked = v3d_free_object,
+ .gem_vm_ops = &v3d_vm_ops,
+
+ .prime_handle_to_fd = drm_gem_prime_handle_to_fd,
+ .prime_fd_to_handle = drm_gem_prime_fd_to_handle,
+ .gem_prime_import = drm_gem_prime_import,
+ .gem_prime_export = drm_gem_prime_export,
+ .gem_prime_res_obj = v3d_prime_res_obj,
+ .gem_prime_get_sg_table = v3d_prime_get_sg_table,
+ .gem_prime_import_sg_table = v3d_prime_import_sg_table,
+ .gem_prime_mmap = v3d_prime_mmap,
+
+ .ioctls = v3d_drm_ioctls,
+ .num_ioctls = ARRAY_SIZE(v3d_drm_ioctls),
+ .fops = &v3d_drm_fops,
+
+ .name = DRIVER_NAME,
+ .desc = DRIVER_DESC,
+ .date = DRIVER_DATE,
+ .major = DRIVER_MAJOR,
+ .minor = DRIVER_MINOR,
+ .patchlevel = DRIVER_PATCHLEVEL,
+};
+
+static const struct of_device_id v3d_of_match[] = {
+ { .compatible = "brcm,7268-v3d" },
+ { .compatible = "brcm,7278-v3d" },
+ {},
+};
+MODULE_DEVICE_TABLE(of, v3d_of_match);
+
+static int
+map_regs(struct v3d_dev *v3d, void __iomem **regs, const char *name)
+{
+ struct resource *res =
+ platform_get_resource_byname(v3d->pdev, IORESOURCE_MEM, name);
+
+ *regs = devm_ioremap_resource(v3d->dev, res);
+ return PTR_ERR_OR_ZERO(*regs);
+}
+
+static int v3d_platform_drm_probe(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ struct drm_device *drm;
+ struct v3d_dev *v3d;
+ int ret;
+ u32 ident1;
+
+ dev->coherent_dma_mask = DMA_BIT_MASK(36);
+
+ v3d = kzalloc(sizeof(*v3d), GFP_KERNEL);
+ if (!v3d)
+ return -ENOMEM;
+ v3d->dev = dev;
+ v3d->pdev = pdev;
+ drm = &v3d->drm;
+
+ ret = map_regs(v3d, &v3d->bridge_regs, "bridge");
+ if (ret)
+ goto dev_free;
+
+ ret = map_regs(v3d, &v3d->hub_regs, "hub");
+ if (ret)
+ goto dev_free;
+
+ ret = map_regs(v3d, &v3d->core_regs[0], "core0");
+ if (ret)
+ goto dev_free;
+
+ ident1 = V3D_READ(V3D_HUB_IDENT1);
+ v3d->ver = (V3D_GET_FIELD(ident1, V3D_HUB_IDENT1_TVER) * 10 +
+ V3D_GET_FIELD(ident1, V3D_HUB_IDENT1_REV));
+ v3d->cores = V3D_GET_FIELD(ident1, V3D_HUB_IDENT1_NCORES);
+ WARN_ON(v3d->cores > 1); /* multicore not yet implemented */
+
+ if (v3d->ver < 41) {
+ ret = map_regs(v3d, &v3d->gca_regs, "gca");
+ if (ret)
+ goto dev_free;
+ }
+
+ v3d->mmu_scratch = dma_alloc_wc(dev, 4096, &v3d->mmu_scratch_paddr,
+ GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO);
+ if (!v3d->mmu_scratch) {
+ dev_err(dev, "Failed to allocate MMU scratch page\n");
+ ret = -ENOMEM;
+ goto dev_free;
+ }
+
+ pm_runtime_use_autosuspend(dev);
+ pm_runtime_set_autosuspend_delay(dev, 50);
+ pm_runtime_enable(dev);
+
+ ret = drm_dev_init(&v3d->drm, &v3d_drm_driver, dev);
+ if (ret)
+ goto dma_free;
+
+ platform_set_drvdata(pdev, drm);
+ drm->dev_private = v3d;
+
+ ret = v3d_gem_init(drm);
+ if (ret)
+ goto dev_destroy;
+
+ v3d_irq_init(v3d);
+
+ ret = drm_dev_register(drm, 0);
+ if (ret)
+ goto gem_destroy;
+
+ return 0;
+
+gem_destroy:
+ v3d_gem_destroy(drm);
+dev_destroy:
+ drm_dev_put(drm);
+dma_free:
+ dma_free_wc(dev, 4096, v3d->mmu_scratch, v3d->mmu_scratch_paddr);
+dev_free:
+ kfree(v3d);
+ return ret;
+}
+
+static int v3d_platform_drm_remove(struct platform_device *pdev)
+{
+ struct drm_device *drm = platform_get_drvdata(pdev);
+ struct v3d_dev *v3d = to_v3d_dev(drm);
+
+ drm_dev_unregister(drm);
+
+ v3d_gem_destroy(drm);
+
+ drm_dev_put(drm);
+
+ dma_free_wc(v3d->dev, 4096, v3d->mmu_scratch, v3d->mmu_scratch_paddr);
+
+ return 0;
+}
+
+static struct platform_driver v3d_platform_driver = {
+ .probe = v3d_platform_drm_probe,
+ .remove = v3d_platform_drm_remove,
+ .driver = {
+ .name = "v3d",
+ .of_match_table = v3d_of_match,
+ },
+};
+
+static int __init v3d_drm_register(void)
+{
+ return platform_driver_register(&v3d_platform_driver);
+}
+
+static void __exit v3d_drm_unregister(void)
+{
+ platform_driver_unregister(&v3d_platform_driver);
+}
+
+module_init(v3d_drm_register);
+module_exit(v3d_drm_unregister);
+
+MODULE_ALIAS("platform:v3d-drm");
+MODULE_DESCRIPTION("Broadcom V3D DRM Driver");
+MODULE_AUTHOR("Eric Anholt <eric@anholt.net>");
+MODULE_LICENSE("GPL v2");