[PATCH 01/14] drm/amdgpu: fix wrong release of vmid owner

List overview All Threads
Download

newer

older

[GIT PULL 2nd] drm-hisilicon-next...

[PATCH 1/2] drm/omapdrm: include...

Alex Deucher

4 May 2016 4 May '16

6:26 p.m.

From: Chunming Zhou David1.Zhou@amd.com

The release of the vmid owner was not handled correctly. We need to take the lock and walk the lru list.

Signed-off-by: Chunming Zhou David1.Zhou@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Monk Liu monk.liu@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 856116a..e06d066 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1454,6 +1454,7 @@ error_free_sched_entity: void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) { struct amdgpu_bo_va_mapping *mapping, *tmp; + struct amdgpu_vm_id *id, *id_tmp; int i;

amd_sched_entity_fini(vm->entity.sched, &vm->entity); @@ -1478,14 +1479,17 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) amdgpu_bo_unref(&vm->page_directory); fence_put(vm->page_directory_fence);

- for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { - struct amdgpu_vm_id *id = vm->ids[i]; - + mutex_lock(&adev->vm_manager.lock); + list_for_each_entry_safe(id, id_tmp, &adev->vm_manager.ids_lru, + list) { if (!id) continue; - - atomic_long_cmpxchg(&id->owner, (long)vm, 0); + if (atomic_long_read(&id->owner) == (long)vm) { + atomic_long_set(&id->owner, 0); + id->pd_gpu_addr = 0; + } } + mutex_unlock(&adev->vm_manager.lock); }

/**

-- 2.5.5

Show replies by date

Alex Deucher

4 May 4 May

6:26 p.m.

New subject: [PATCH 02/14] drm/amdgpu: add client id for every vm

From: Chunming Zhou David1.Zhou@amd.com

This adds a unique id for each vm client so we can properly track them.

Signed-off-by: Chunming Zhou David1.Zhou@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Monk Liu monk.liu@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 6 ++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 2 ++ 2 files changed, 8 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 1012bd3..7bea3e2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -363,6 +363,7 @@ struct amdgpu_fence_driver { /* some special values for the owner field */ #define AMDGPU_FENCE_OWNER_UNDEFINED ((void*)0ul) #define AMDGPU_FENCE_OWNER_VM ((void*)1ul) +#define AMDGPU_CLIENT_ID_RESERVED 2

#define AMDGPU_FENCE_FLAG_64BIT (1 << 0) #define AMDGPU_FENCE_FLAG_INT (1 << 1) @@ -885,6 +886,9 @@ struct amdgpu_vm {

/* Scheduler entity for page table updates */ struct amd_sched_entity entity; + + /* client id */ + u64 client_id; };

struct amdgpu_vm_id { @@ -924,6 +928,8 @@ struct amdgpu_vm_manager { struct amdgpu_ring *vm_pte_rings[AMDGPU_MAX_RINGS]; unsigned vm_pte_num_rings; atomic_t vm_pte_next_ring; + /* client id counter */ + atomic64_t client_counter; };

void amdgpu_vm_manager_init(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index e06d066..275378c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1386,6 +1386,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm) for (i = 0; i < AMDGPU_MAX_RINGS; ++i) vm->ids[i] = NULL; vm->va = RB_ROOT; + vm->client_id = atomic64_inc_return(&adev->vm_manager.client_counter); spin_lock_init(&vm->status_lock); INIT_LIST_HEAD(&vm->invalidated); INIT_LIST_HEAD(&vm->cleared); @@ -1514,6 +1515,7 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev) }

atomic_set(&adev->vm_manager.vm_pte_next_ring, 0); + atomic64_set(&adev->vm_manager.client_counter, AMDGPU_CLIENT_ID_RESERVED); }

/**

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 03/14] drm/amdgpu: make vmid owner be client_id

From: Chunming Zhou David1.Zhou@amd.com

Using the pointer is not adequate.

Signed-off-by: Chunming Zhou David1.Zhou@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Monk Liu monk.liu@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 275378c..2c3d955 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -185,7 +185,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, if (!id) continue;

- if (atomic_long_read(&id->owner) != (long)vm) + if (atomic_long_read(&id->owner) != vm->client_id) continue;

if (pd_addr != id->pd_gpu_addr) @@ -261,7 +261,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,

list_move_tail(&id->list, &adev->vm_manager.ids_lru); id->last_user = ring; - atomic_long_set(&id->owner, (long)vm); + atomic_long_set(&id->owner, vm->client_id); vm->ids[ring->idx] = id;

*vm_id = id - adev->vm_manager.ids; @@ -1485,7 +1485,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) list) { if (!id) continue; - if (atomic_long_read(&id->owner) == (long)vm) { + if (atomic_long_read(&id->owner) == vm->client_id) { atomic_long_set(&id->owner, 0); id->pd_gpu_addr = 0; }

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 04/14] drm/amdgpu: keep vm in job instead of ib (v2)

From: Monk Liu Monk.Liu@amd.com

ib.vm is a legacy way to get vm, after scheduler implemented vm should be get from job, and all ibs from one job share the same vm, no need to keep ib.vm just move vm field to job.

this patch as well add job as paramter to ib_schedule so it can get vm from job->vm.

v2: agd: sqaush in: drm/amdgpu: check if ring emit_vm_flush exists in vm flush

No vm flush on engines that don't support VM.

bug: https://bugs.freedesktop.org/show_bug.cgi?id=95195

Signed-off-by: Monk Liu Monk.Liu@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 6 +++--- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 3 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 16 ++++------------ drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 9 +++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 3 ++- drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 2 +- drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 2 +- drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 2 +- 12 files changed, 25 insertions(+), 30 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 7bea3e2..2ee99dc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -740,7 +740,6 @@ struct amdgpu_ib { uint64_t gpu_addr; uint32_t *ptr; struct amdgpu_user_fence *user; - struct amdgpu_vm *vm; unsigned vm_id; uint64_t vm_pd_addr; struct amdgpu_ctx *ctx; @@ -763,7 +762,7 @@ enum amdgpu_ring_type { extern const struct amd_sched_backend_ops amdgpu_sched_ops;

int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, - struct amdgpu_job **job); + struct amdgpu_job **job, struct amdgpu_vm *vm); int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, struct amdgpu_job **job);

@@ -1191,7 +1190,7 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, struct fence *f); int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, struct amdgpu_ib *ib, struct fence *last_vm_update, - struct fence **f); + struct amdgpu_job *job, struct fence **f); int amdgpu_ib_pool_init(struct amdgpu_device *adev); void amdgpu_ib_pool_fini(struct amdgpu_device *adev); int amdgpu_ib_ring_tests(struct amdgpu_device *adev); @@ -1247,6 +1246,7 @@ struct amdgpu_cs_parser { struct amdgpu_job { struct amd_sched_job base; struct amdgpu_device *adev; + struct amdgpu_vm *vm; struct amdgpu_ring *ring; struct amdgpu_sync sync; struct amdgpu_ib *ibs; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 2ebba29..1a06596 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -120,6 +120,7 @@ static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p, int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) { struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; union drm_amdgpu_cs *cs = data; uint64_t *chunk_array_user; uint64_t *chunk_array; @@ -214,7 +215,7 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) } }

- ret = amdgpu_job_alloc(p->adev, num_ibs, &p->job); + ret = amdgpu_job_alloc(p->adev, num_ibs, &p->job, vm); if (ret) goto free_all_kdata;

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 0129617..0ed6430 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -74,7 +74,6 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo); }

- ib->vm = vm; ib->vm_id = 0;

return 0; @@ -117,13 +116,13 @@ void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, struct fen */ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, struct amdgpu_ib *ibs, struct fence *last_vm_update, - struct fence **f) + struct amdgpu_job *job, struct fence **f) { struct amdgpu_device *adev = ring->adev; struct amdgpu_ib *ib = &ibs[0]; struct amdgpu_ctx *ctx, *old_ctx; - struct amdgpu_vm *vm; struct fence *hwf; + struct amdgpu_vm *vm = NULL; unsigned i, patch_offset = ~0;

int r = 0; @@ -132,7 +131,8 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, return -EINVAL;

ctx = ibs->ctx; - vm = ibs->vm; + if (job) /* for domain0 job like ring test, ibs->job is not assigned */ + vm = job->vm;

if (!ring->ready) { dev_err(adev->dev, "couldn't schedule ib\n"); @@ -174,14 +174,6 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, old_ctx = ring->current_ctx; for (i = 0; i < num_ibs; ++i) { ib = &ibs[i]; - - if (ib->ctx != ctx || ib->vm != vm) { - ring->current_ctx = old_ctx; - if (ib->vm_id) - amdgpu_vm_reset_id(adev, ib->vm_id); - amdgpu_ring_undo(ring); - return -EINVAL; - } amdgpu_ring_emit_ib(ring, ib); ring->current_ctx = ctx; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index 4eea2a1..917c6f3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -46,7 +46,7 @@ void amdgpu_job_timeout_func(struct work_struct *work) }

int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, - struct amdgpu_job **job) + struct amdgpu_job **job, struct amdgpu_vm *vm) { size_t size = sizeof(struct amdgpu_job);

@@ -60,6 +60,7 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, return -ENOMEM;

(*job)->adev = adev; + (*job)->vm = vm; (*job)->ibs = (void *)&(*job)[1]; (*job)->num_ibs = num_ibs; INIT_WORK(&(*job)->base.work_free_job, amdgpu_job_free_handler); @@ -74,7 +75,7 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, { int r;

- r = amdgpu_job_alloc(adev, 1, job); + r = amdgpu_job_alloc(adev, 1, job, NULL); if (r) return r;

@@ -138,7 +139,7 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring, static struct fence *amdgpu_job_dependency(struct amd_sched_job *sched_job) { struct amdgpu_job *job = to_amdgpu_job(sched_job); - struct amdgpu_vm *vm = job->ibs->vm; + struct amdgpu_vm *vm = job->vm;

struct fence *fence = amdgpu_sync_get_fence(&job->sync);

@@ -186,7 +187,7 @@ static struct fence *amdgpu_job_run(struct amd_sched_job *sched_job)

trace_amdgpu_sched_run_job(job); r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, - job->sync.last_vm_update, &fence); + job->sync.last_vm_update, job, &fence); if (r) { DRM_ERROR("Error scheduling IBs (%d)\n", r); goto err; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index db86012..dcf91fb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -905,7 +905,7 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, ib->length_dw = 16;

if (direct) { - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, ib, NULL, NULL, &f); job->fence = f; if (r) goto err_free; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 80c1048..77c6c08 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -435,7 +435,7 @@ int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0;

- r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, ib, NULL, NULL, &f); job->fence = f; if (r) goto err; @@ -497,7 +497,7 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, ib->ptr[i] = 0x0;

if (direct) { - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, ib, NULL, NULL, &f); job->fence = f; if (r) goto err; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 2c3d955..692d0d02 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -303,7 +303,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, pd_addr != AMDGPU_VM_NO_FLUSH || gds_switch_needed)) amdgpu_ring_emit_pipeline_sync(ring);

- if (pd_addr != AMDGPU_VM_NO_FLUSH) { + if (ring->funcs->emit_vm_flush && + pd_addr != AMDGPU_VM_NO_FLUSH) { struct fence *fence;

trace_amdgpu_vm_flush(pd_addr, ring->idx, vm_id); diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index b7ed9d3..8d69c65 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -643,7 +643,7 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[3] = 1; ib.ptr[4] = 0xDEADBEEF; ib.length_dw = 5; - r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, NULL, &f); if (r) goto err1;

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 6686c9c..0310890 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -2136,7 +2136,7 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[2] = 0xDEADBEEF; ib.length_dw = 3;

- r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, NULL, &f); if (r) goto err2;

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 021c17e..a82945f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -800,7 +800,7 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[2] = 0xDEADBEEF; ib.length_dw = 3;

- r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, NULL, &f); if (r) goto err2;

@@ -1551,7 +1551,7 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) ib.ptr[ib.length_dw++] = EVENT_TYPE(7) | EVENT_INDEX(4);

/* shedule the ib on the ring */ - r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, NULL, &f); if (r) { DRM_ERROR("amdgpu: ib submit failed (%d).\n", r); goto fail; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index e6d3544..27ca46d 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -701,7 +701,7 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[7] = SDMA_PKT_HEADER_OP(SDMA_OP_NOP); ib.length_dw = 8;

- r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, NULL, &f); if (r) goto err1;

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 00b4370..278b1fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -925,7 +925,7 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[7] = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP); ib.length_dw = 8;

- r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, NULL, &f); if (r) goto err1;

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 05/14] drm/amdgpu: use fence_context to judge ctx switch

From: Monk Liu Monk.Liu@amd.com

use ctx pointer is not safe, cuz they are likely already be assigned to another ctx when doing comparing.

fence_context is always increasing and have rare chance to overback to used number for jobs that scheduled to ring continueonsly

Signed-off-by: Monk Liu Monk.Liu@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 8 ++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 14 +++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h | 2 +- drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 2 +- drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 11 +++++------ drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 11 +++++------ drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 2 +- drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c | 2 +- drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c | 2 +- drivers/gpu/drm/amd/scheduler/gpu_scheduler.c | 1 + drivers/gpu/drm/amd/scheduler/gpu_scheduler.h | 1 + 15 files changed, 31 insertions(+), 32 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 2ee99dc..9d54d76 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -283,7 +283,7 @@ struct amdgpu_ring_funcs { int (*parse_cs)(struct amdgpu_cs_parser *p, uint32_t ib_idx); /* command emit functions */ void (*emit_ib)(struct amdgpu_ring *ring, - struct amdgpu_ib *ib); + struct amdgpu_ib *ib, bool ctx_switch); void (*emit_fence)(struct amdgpu_ring *ring, uint64_t addr, uint64_t seq, unsigned flags); void (*emit_pipeline_sync)(struct amdgpu_ring *ring); @@ -742,7 +742,6 @@ struct amdgpu_ib { struct amdgpu_user_fence *user; unsigned vm_id; uint64_t vm_pd_addr; - struct amdgpu_ctx *ctx; uint32_t gds_base, gds_size; uint32_t gws_base, gws_size; uint32_t oa_base, oa_size; @@ -805,7 +804,7 @@ struct amdgpu_ring { unsigned wptr_offs; unsigned next_rptr_offs; unsigned fence_offs; - struct amdgpu_ctx *current_ctx; + uint64_t last_fence_context; enum amdgpu_ring_type type; char name[16]; unsigned cond_exe_offs; @@ -1253,6 +1252,7 @@ struct amdgpu_job { struct fence *fence; /* the hw fence */ uint32_t num_ibs; void *owner; + uint64_t fence_context; struct amdgpu_user_fence uf; }; #define to_amdgpu_job(sched_job) \ @@ -2219,7 +2219,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r)) #define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r)) #define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r)) -#define amdgpu_ring_emit_ib(r, ib) (r)->funcs->emit_ib((r), (ib)) +#define amdgpu_ring_emit_ib(r, ib, f) (r)->funcs->emit_ib((r), (ib), (f)) #define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r)) #define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr)) #define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 1a06596..8c3bf63 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -741,7 +741,6 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev,

ib->length_dw = chunk_ib->ib_bytes / 4; ib->flags = chunk_ib->flags; - ib->ctx = parser->ctx; j++; }

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 0ed6430..1693fc7e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -120,7 +120,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, { struct amdgpu_device *adev = ring->adev; struct amdgpu_ib *ib = &ibs[0]; - struct amdgpu_ctx *ctx, *old_ctx; + uint64_t fence_context = 0, old = ring->last_fence_context; struct fence *hwf; struct amdgpu_vm *vm = NULL; unsigned i, patch_offset = ~0; @@ -130,9 +130,10 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, if (num_ibs == 0) return -EINVAL;

- ctx = ibs->ctx; - if (job) /* for domain0 job like ring test, ibs->job is not assigned */ + if (job) {/* for domain0 job like ring test, ibs->job is not assigned */ vm = job->vm; + fence_context = job->fence_context; + }

if (!ring->ready) { dev_err(adev->dev, "couldn't schedule ib\n"); @@ -171,12 +172,11 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, /* always set cond_exec_polling to CONTINUE */ *ring->cond_exe_cpu_addr = 1;

- old_ctx = ring->current_ctx; for (i = 0; i < num_ibs; ++i) { ib = &ibs[i]; - amdgpu_ring_emit_ib(ring, ib); - ring->current_ctx = ctx; + amdgpu_ring_emit_ib(ring, ib, (i == 0 && old != fence_context)); } + ring->last_fence_context = fence_context;

if (vm) { if (ring->funcs->emit_hdp_invalidate) @@ -186,7 +186,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, r = amdgpu_fence_emit(ring, &hwf); if (r) { dev_err(adev->dev, "failed to emit fence (%d)\n", r); - ring->current_ctx = old_ctx; + ring->last_fence_context = old; if (ib->vm_id) amdgpu_vm_reset_id(adev, ib->vm_id); amdgpu_ring_undo(ring); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 77c6c08..1fabf15 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -761,7 +761,7 @@ out: * @ib: the IB to execute * */ -void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib) +void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib, bool ctx_switch) { amdgpu_ring_write(ring, VCE_CMD_IB); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h index ef99d23..40d0650 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h @@ -34,7 +34,7 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, bool direct, struct fence **fence); void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp); int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx); -void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib); +void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib, bool ctx_switch); void amdgpu_vce_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, unsigned flags); int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring); diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index 8d69c65..d54fb19 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -210,7 +210,7 @@ static void cik_sdma_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) * Schedule an IB in the DMA ring (CIK). */ static void cik_sdma_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { u32 extra_bits = ib->vm_id & 0xf; u32 next_rptr = ring->wptr + 5; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 0310890..9813819 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -2029,17 +2029,16 @@ static void gfx_v7_0_ring_emit_fence_compute(struct amdgpu_ring *ring, * on the gfx ring for execution by the GPU. */ static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { - bool need_ctx_switch = ring->current_ctx != ib->ctx; u32 header, control = 0; u32 next_rptr = ring->wptr + 5;

/* drop the CE preamble IB for the same context */ - if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !need_ctx_switch) + if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !ctx_switch) return;

- if (need_ctx_switch) + if (ctx_switch) next_rptr += 2;

next_rptr += 4; @@ -2050,7 +2049,7 @@ static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, amdgpu_ring_write(ring, next_rptr);

/* insert SWITCH_BUFFER packet before first IB in the ring frame */ - if (need_ctx_switch) { + if (ctx_switch) { amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0)); amdgpu_ring_write(ring, 0); } @@ -2073,7 +2072,7 @@ static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, }

static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { u32 header, control = 0; u32 next_rptr = ring->wptr + 5; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index a82945f..064f255 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -5644,17 +5644,16 @@ static void gfx_v8_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) }

static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { - bool need_ctx_switch = ring->current_ctx != ib->ctx; u32 header, control = 0; u32 next_rptr = ring->wptr + 5;

/* drop the CE preamble IB for the same context */ - if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !need_ctx_switch) + if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !ctx_switch) return;

- if (need_ctx_switch) + if (ctx_switch) next_rptr += 2;

next_rptr += 4; @@ -5665,7 +5664,7 @@ static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, amdgpu_ring_write(ring, next_rptr);

/* insert SWITCH_BUFFER packet before first IB in the ring frame */ - if (need_ctx_switch) { + if (ctx_switch) { amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0)); amdgpu_ring_write(ring, 0); } @@ -5688,7 +5687,7 @@ static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, }

static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { u32 header, control = 0; u32 next_rptr = ring->wptr + 5; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index 27ca46d..abe39024 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -242,7 +242,7 @@ static void sdma_v2_4_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) * Schedule an IB in the DMA ring (VI). */ static void sdma_v2_4_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { u32 vmid = ib->vm_id & 0xf; u32 next_rptr = ring->wptr + 5; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 278b1fe..e222762 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -400,7 +400,7 @@ static void sdma_v3_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) * Schedule an IB in the DMA ring (VI). */ static void sdma_v3_0_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { u32 vmid = ib->vm_id & 0xf; u32 next_rptr = ring->wptr + 5; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c index abd37a7..b9c2a49 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c @@ -489,7 +489,7 @@ static int uvd_v4_2_ring_test_ring(struct amdgpu_ring *ring) * Write ring commands to execute the indirect buffer */ static void uvd_v4_2_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { amdgpu_ring_write(ring, PACKET0(mmUVD_RBC_IB_BASE, 0)); amdgpu_ring_write(ring, ib->gpu_addr); diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c index 1c1a0e2c..465b77c 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c @@ -539,7 +539,7 @@ static int uvd_v5_0_ring_test_ring(struct amdgpu_ring *ring) * Write ring commands to execute the indirect buffer */ static void uvd_v5_0_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { amdgpu_ring_write(ring, PACKET0(mmUVD_LMI_RBC_IB_64BIT_BAR_LOW, 0)); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index d015cb0..c2f790b 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -631,7 +631,7 @@ static int uvd_v6_0_ring_test_ring(struct amdgpu_ring *ring) * Write ring commands to execute the indirect buffer */ static void uvd_v6_0_ring_emit_ib(struct amdgpu_ring *ring, - struct amdgpu_ib *ib) + struct amdgpu_ib *ib, bool ctx_switch) { amdgpu_ring_write(ring, PACKET0(mmUVD_LMI_RBC_IB_64BIT_BAR_LOW, 0)); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c index c16248c..a5d736c 100644 --- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c +++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c @@ -392,6 +392,7 @@ int amd_sched_job_init(struct amd_sched_job *job, kref_init(&job->refcount); job->sched = sched; job->s_entity = entity; + job->fence_context = entity->fence_context; job->s_fence = amd_sched_fence_create(entity, owner); if (!job->s_fence) return -ENOMEM; diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h index 169f70f..5561020 100644 --- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h +++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h @@ -87,6 +87,7 @@ struct amd_sched_job { struct work_struct work_free_job; struct list_head node; struct delayed_work work_tdr; + uint64_t fence_context; void (*timeout_callback) (struct work_struct *work); void (*free_callback)(struct kref *refcount); };

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 06/14] drm/amdgpu: add pipeline sync for compute job

From: Chunming Zhou David1.Zhou@amd.com

hardware ring is async processed, the job is executed in parallel. In some case, this will result vm fault, like jobs with different vmids.

This works around a CPC hw issue which will eventually be fixed in fw.

Signed-off-by: Chunming Zhou David1.Zhou@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Christian König christian.koenig@amd.com Reviewed-by: Monk Liu monk.liu@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 692d0d02..62ce725 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -300,7 +300,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, int r;

if (ring->funcs->emit_pipeline_sync && ( - pd_addr != AMDGPU_VM_NO_FLUSH || gds_switch_needed)) + pd_addr != AMDGPU_VM_NO_FLUSH || gds_switch_needed || + ring->type == AMDGPU_RING_TYPE_COMPUTE)) amdgpu_ring_emit_pipeline_sync(ring);

if (ring->funcs->emit_vm_flush &&

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 07/14] drm/amd/amdgpu: Enable CG for UVD6 on Carrizo

From: Tom St Denis tom.stdenis@amd.com

Tested via vdpau/mpv.

Signed-off-by: Tom St Denis tom.stdenis@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/vi.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 340a166..31dd630 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -1193,7 +1193,8 @@ static int vi_common_early_init(void *handle) adev->external_rev_id = adev->rev_id + 0x50; break; case CHIP_CARRIZO: - adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | + adev->cg_flags = AMD_CG_SUPPORT_UVD_MGCG | + AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS | AMD_CG_SUPPORT_GFX_RLC_LS | AMD_CG_SUPPORT_GFX_CP_LS |

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 08/14] drm/amdgpu: hdp flush&inval should always do

From: Monk Liu Monk.Liu@amd.com

This fixes Tonga vm-fault issue when running disaster (a multiple context GL heavy tests), We should always flush & invalidate hdp no matter vm used or not.

Signed-off-by: Monk Liu Monk.Liu@amd.com Reviewed-by: Christian König christian.koenig@amd.com Reviewed-by: Chunming Zhou David1.Zhou@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 1693fc7e..61ca7e1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -164,11 +164,11 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, amdgpu_ring_undo(ring); return r; } - - if (ring->funcs->emit_hdp_flush) - amdgpu_ring_emit_hdp_flush(ring); }

+ if (ring->funcs->emit_hdp_flush) + amdgpu_ring_emit_hdp_flush(ring); + /* always set cond_exec_polling to CONTINUE */ *ring->cond_exe_cpu_addr = 1;

@@ -178,10 +178,8 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, } ring->last_fence_context = fence_context;

- if (vm) { - if (ring->funcs->emit_hdp_invalidate) - amdgpu_ring_emit_hdp_invalidate(ring); - } + if (ring->funcs->emit_hdp_invalidate) + amdgpu_ring_emit_hdp_invalidate(ring);

r = amdgpu_fence_emit(ring, &hwf); if (r) {

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 09/14] drm/amdgpu: two minor 80 char fixes

From: Christian König christian.koenig@amd.com

Signed-off-by: Christian König christian.koenig@amd.com Reviewed-by: Tom St Denis tom.stdenis@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 10 ++++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 3 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 7 ++++--- 3 files changed, 12 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 9d54d76..8e3b14d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -508,9 +508,10 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, struct drm_file *file_priv); unsigned long amdgpu_gem_timeout(uint64_t timeout_ns); struct sg_table *amdgpu_gem_prime_get_sg_table(struct drm_gem_object *obj); -struct drm_gem_object *amdgpu_gem_prime_import_sg_table(struct drm_device *dev, - struct dma_buf_attachment *attach, - struct sg_table *sg); +struct drm_gem_object * +amdgpu_gem_prime_import_sg_table(struct drm_device *dev, + struct dma_buf_attachment *attach, + struct sg_table *sg); struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev, struct drm_gem_object *gobj, int flags); @@ -1186,7 +1187,8 @@ struct amdgpu_gfx {

int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned size, struct amdgpu_ib *ib); -void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, struct fence *f); +void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, + struct fence *f); int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, struct amdgpu_ib *ib, struct fence *last_vm_update, struct amdgpu_job *job, struct fence **f); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 61ca7e1..32dd199 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -88,7 +88,8 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, * * Free an IB (all asics). */ -void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, struct fence *f) +void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, + struct fence *f) { amdgpu_sa_bo_free(adev, &ib->sa_bo, f); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index be6388f..7700dc2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -57,9 +57,10 @@ void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr) ttm_bo_kunmap(&bo->dma_buf_vmap); }

-struct drm_gem_object *amdgpu_gem_prime_import_sg_table(struct drm_device *dev, - struct dma_buf_attachment *attach, - struct sg_table *sg) +struct drm_gem_object * +amdgpu_gem_prime_import_sg_table(struct drm_device *dev, + struct dma_buf_attachment *attach, + struct sg_table *sg) { struct reservation_object *resv = attach->dmabuf->resv; struct amdgpu_device *adev = dev->dev_private;

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 10/14] drm/amdgpu: make the VMID owner always 64bit

From: Christian König christian.koenig@amd.com

Otherwise we could (in theory) run into problems on 32bit systems.

Signed-off-by: Christian König christian.koenig@amd.com Reviewed-by: Chunming Zhou david1.zhou@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 8e3b14d..de5bfc7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -896,7 +896,7 @@ struct amdgpu_vm_id { struct amdgpu_sync active; struct fence *last_flush; struct amdgpu_ring *last_user; - atomic_long_t owner; + atomic64_t owner;

uint64_t pd_gpu_addr; /* last flushed PD/PT update */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 62ce725..cd57898 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -185,7 +185,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, if (!id) continue;

- if (atomic_long_read(&id->owner) != vm->client_id) + if (atomic64_read(&id->owner) != vm->client_id) continue;

if (pd_addr != id->pd_gpu_addr) @@ -261,7 +261,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,

list_move_tail(&id->list, &adev->vm_manager.ids_lru); id->last_user = ring; - atomic_long_set(&id->owner, vm->client_id); + atomic64_set(&id->owner, vm->client_id); vm->ids[ring->idx] = id;

*vm_id = id - adev->vm_manager.ids;

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 11/14] drm/amdgpu: remove owner cleanup v2

From: Christian König christian.koenig@amd.com

The client ID is now unique, so no need to resert the owner fields any more.

v2: remove unused variables as well

Signed-off-by: Christian König christian.koenig@amd.com Reviewed-by: Chunming Zhou david1.zhou@amd.com (v1) Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 13 ------------- 1 file changed, 13 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index cd57898..67f6c2e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1457,7 +1457,6 @@ error_free_sched_entity: void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) { struct amdgpu_bo_va_mapping *mapping, *tmp; - struct amdgpu_vm_id *id, *id_tmp; int i;

amd_sched_entity_fini(vm->entity.sched, &vm->entity); @@ -1481,18 +1480,6 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)

amdgpu_bo_unref(&vm->page_directory); fence_put(vm->page_directory_fence); - - mutex_lock(&adev->vm_manager.lock); - list_for_each_entry_safe(id, id_tmp, &adev->vm_manager.ids_lru, - list) { - if (!id) - continue; - if (atomic_long_read(&id->owner) == vm->client_id) { - atomic_long_set(&id->owner, 0); - id->pd_gpu_addr = 0; - } - } - mutex_unlock(&adev->vm_manager.lock); }

/**

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 12/14] drm/amdgpu: remove define for reserved client ID

From: Christian König christian.koenig@amd.com

Just set it to zero instead.

Signed-off-by: Christian König christian.koenig@amd.com Reviewed-by: Chunming Zhou david1.zhou@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 2 +- 2 files changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index de5bfc7..416e814 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -363,7 +363,6 @@ struct amdgpu_fence_driver { /* some special values for the owner field */ #define AMDGPU_FENCE_OWNER_UNDEFINED ((void*)0ul) #define AMDGPU_FENCE_OWNER_VM ((void*)1ul) -#define AMDGPU_CLIENT_ID_RESERVED 2

#define AMDGPU_FENCE_FLAG_64BIT (1 << 0) #define AMDGPU_FENCE_FLAG_INT (1 << 1) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 67f6c2e..ea708cb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1504,7 +1504,7 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev) }

atomic_set(&adev->vm_manager.vm_pte_next_ring, 0); - atomic64_set(&adev->vm_manager.client_counter, AMDGPU_CLIENT_ID_RESERVED); + atomic64_set(&adev->vm_manager.client_counter, 0); }

/**

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 13/14] drm/amd: cleanup remaining spaces and tabs v2

From: Christian König christian.koenig@amd.com

This is the result of running the following commands: find drivers/gpu/drm/amd/ -name "*.h" -exec sed -i 's/[ \t]+$//' {} ; find drivers/gpu/drm/amd/ -name "*.c" -exec sed -i 's/[ \t]+$//' {} ; find drivers/gpu/drm/amd/ -name "*.h" -exec sed -i 's/ +\t/\t/' {} ; find drivers/gpu/drm/amd/ -name "*.c" -exec sed -i 's/ +\t/\t/' {} ;

v2: drop changes to DAL and internal headers

Signed-off-by: Christian König christian.koenig@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 12 ++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c | 6 +++--- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h | 6 +++--- drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 6 +++--- drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 6 +++--- drivers/gpu/drm/amd/amdgpu/atom.h | 2 +- drivers/gpu/drm/amd/amdgpu/ci_dpm.c | 2 +- drivers/gpu/drm/amd/amdgpu/cik_ih.c | 2 +- drivers/gpu/drm/amd/amdgpu/cikd.h | 4 ++-- drivers/gpu/drm/amd/amdgpu/cz_ih.c | 2 +- drivers/gpu/drm/amd/amdgpu/cz_smumgr.h | 2 +- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/vce_v3_0.c | 6 +++--- drivers/gpu/drm/amd/amdgpu/vid.h | 2 +- drivers/gpu/drm/amd/powerplay/hwmgr/fiji_hwmgr.c | 6 +++--- drivers/gpu/drm/amd/powerplay/hwmgr/polaris10_hwmgr.c | 8 ++++---- drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.c | 4 ++-- drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.c | 8 ++++---- drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.h | 18 +++++++++--------- drivers/gpu/drm/amd/powerplay/inc/hwmgr.h | 2 +- drivers/gpu/drm/amd/scheduler/gpu_scheduler.h | 2 +- 23 files changed, 56 insertions(+), 56 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 416e814..19ac6c5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -369,7 +369,7 @@ struct amdgpu_fence_driver {

struct amdgpu_user_fence { /* write-back bo */ - struct amdgpu_bo *bo; + struct amdgpu_bo *bo; /* write-back address offset to bo start */ uint32_t offset; }; @@ -776,7 +776,7 @@ struct amdgpu_ring { struct amdgpu_device *adev; const struct amdgpu_ring_funcs *funcs; struct amdgpu_fence_driver fence_drv; - struct amd_gpu_scheduler sched; + struct amd_gpu_scheduler sched;

spinlock_t fence_lock; struct amdgpu_bo *ring_obj; @@ -1246,7 +1246,7 @@ struct amdgpu_cs_parser { struct amdgpu_job { struct amd_sched_job base; struct amdgpu_device *adev; - struct amdgpu_vm *vm; + struct amdgpu_vm *vm; struct amdgpu_ring *ring; struct amdgpu_sync sync; struct amdgpu_ib *ibs; @@ -1700,7 +1700,7 @@ struct amdgpu_sdma { struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES]; struct amdgpu_irq_src trap_irq; struct amdgpu_irq_src illegal_inst_irq; - int num_instances; + int num_instances; };

/* @@ -1954,11 +1954,11 @@ struct amdgpu_device { bool shutdown; bool need_dma32; bool accel_working; - struct work_struct reset_work; + struct work_struct reset_work; struct notifier_block acpi_nb; struct amdgpu_i2c_chan *i2c_bus[AMDGPU_MAX_I2C_BUS]; struct amdgpu_debugfs debugfs[AMDGPU_DEBUGFS_MAX_COMPONENTS]; - unsigned debugfs_count; + unsigned debugfs_count; #if defined(CONFIG_DEBUG_FS) struct dentry *debugfs_regs[AMDGPU_DEBUGFS_MAX_COMPONENTS]; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c index eacd810..35d0856 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c @@ -263,7 +263,7 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, for (i = 0; i < args->in.bo_number; ++i) { if (copy_from_user(&info[i], uptr, bytes)) goto error_free; - + uptr += args->in.bo_info_size; } } @@ -271,7 +271,7 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, switch (args->in.operation) { case AMDGPU_BO_LIST_OP_CREATE: r = amdgpu_bo_list_create(fpriv, &list, &handle); - if (r) + if (r) goto error_free;

r = amdgpu_bo_list_set(adev, filp, list, info, @@ -281,7 +281,7 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, goto error_free;

break; - + case AMDGPU_BO_LIST_OP_DESTROY: amdgpu_bo_list_destroy(fpriv, handle); handle = 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 8f037e5..17689f0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -354,7 +354,7 @@ static int amdgpu_doorbell_init(struct amdgpu_device *adev) adev->doorbell.base = pci_resource_start(adev->pdev, 2); adev->doorbell.size = pci_resource_len(adev->pdev, 2);

- adev->doorbell.num_doorbells = min_t(u32, adev->doorbell.size / sizeof(u32), + adev->doorbell.num_doorbells = min_t(u32, adev->doorbell.size / sizeof(u32), AMDGPU_DOORBELL_MAX_ASSIGNMENT+1); if (adev->doorbell.num_doorbells == 0) return -EINVAL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h index c3f4e85..503d540 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h @@ -43,7 +43,7 @@ struct amdgpu_ring; struct amdgpu_bo;

struct amdgpu_gds_asic_info { - uint32_t total_size; + uint32_t total_size; uint32_t gfx_partition_size; uint32_t cs_partition_size; }; @@ -52,8 +52,8 @@ struct amdgpu_gds { struct amdgpu_gds_asic_info mem; struct amdgpu_gds_asic_info gws; struct amdgpu_gds_asic_info oa; - /* At present, GDS, GWS and OA resources for gfx (graphics) - * is always pre-allocated and available for graphics operation. + /* At present, GDS, GWS and OA resources for gfx (graphics) + * is always pre-allocated and available for graphics operation. * Such resource is shared between all gfx clients. * TODO: move this operation to user space * */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index 81bd964..8a253aa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -530,7 +530,7 @@ struct amdgpu_framebuffer { ((em) == ATOM_ENCODER_MODE_DP_MST))

/* Driver internal use only flags of amdgpu_get_crtc_scanoutpos() */ -#define USE_REAL_VBLANKSTART (1 << 30) +#define USE_REAL_VBLANKSTART (1 << 30) #define GET_DISTANCE_TO_VBLANKSTART (1 << 31)

void amdgpu_link_encoder_connector(struct drm_device *dev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index dcf91fb..e2c31da 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -45,9 +45,9 @@ /* Firmware Names */ #ifdef CONFIG_DRM_AMDGPU_CIK #define FIRMWARE_BONAIRE "radeon/bonaire_uvd.bin" -#define FIRMWARE_KABINI "radeon/kabini_uvd.bin" -#define FIRMWARE_KAVERI "radeon/kaveri_uvd.bin" -#define FIRMWARE_HAWAII "radeon/hawaii_uvd.bin" +#define FIRMWARE_KABINI "radeon/kabini_uvd.bin" +#define FIRMWARE_KAVERI "radeon/kaveri_uvd.bin" +#define FIRMWARE_HAWAII "radeon/hawaii_uvd.bin" #define FIRMWARE_MULLINS "radeon/mullins_uvd.bin" #endif #define FIRMWARE_TONGA "amdgpu/tonga_uvd.bin" diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 1fabf15..62cf732 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -41,9 +41,9 @@ /* Firmware Names */ #ifdef CONFIG_DRM_AMDGPU_CIK #define FIRMWARE_BONAIRE "radeon/bonaire_vce.bin" -#define FIRMWARE_KABINI "radeon/kabini_vce.bin" -#define FIRMWARE_KAVERI "radeon/kaveri_vce.bin" -#define FIRMWARE_HAWAII "radeon/hawaii_vce.bin" +#define FIRMWARE_KABINI "radeon/kabini_vce.bin" +#define FIRMWARE_KAVERI "radeon/kaveri_vce.bin" +#define FIRMWARE_HAWAII "radeon/hawaii_vce.bin" #define FIRMWARE_MULLINS "radeon/mullins_vce.bin" #endif #define FIRMWARE_TONGA "amdgpu/tonga_vce.bin" diff --git a/drivers/gpu/drm/amd/amdgpu/atom.h b/drivers/gpu/drm/amd/amdgpu/atom.h index fece8f4..49daf6d 100644 --- a/drivers/gpu/drm/amd/amdgpu/atom.h +++ b/drivers/gpu/drm/amd/amdgpu/atom.h @@ -92,7 +92,7 @@ #define ATOM_WS_AND_MASK 0x45 #define ATOM_WS_FB_WINDOW 0x46 #define ATOM_WS_ATTRIBUTES 0x47 -#define ATOM_WS_REGPTR 0x48 +#define ATOM_WS_REGPTR 0x48

#define ATOM_IIO_NOP 0 #define ATOM_IIO_START 1 diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c index 90f83b2..2f24797 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c @@ -6363,7 +6363,7 @@ static int ci_dpm_set_interrupt_state(struct amdgpu_device *adev, }

static int ci_dpm_process_interrupt(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, + struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { bool queue_thermal = false; diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c index f2f14fe..7e750a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c @@ -243,7 +243,7 @@ static void cik_ih_decode_iv(struct amdgpu_device *adev, /* wptr/rptr are in bytes! */ u32 ring_index = adev->irq.ih.rptr >> 2; uint32_t dw[4]; - + dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]); dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); diff --git a/drivers/gpu/drm/amd/amdgpu/cikd.h b/drivers/gpu/drm/amd/amdgpu/cikd.h index 60d4493..c4f6f00 100644 --- a/drivers/gpu/drm/amd/amdgpu/cikd.h +++ b/drivers/gpu/drm/amd/amdgpu/cikd.h @@ -190,8 +190,8 @@ # define MACRO_TILE_ASPECT(x) ((x) << 4) # define NUM_BANKS(x) ((x) << 6)

-#define MSG_ENTER_RLC_SAFE_MODE 1 -#define MSG_EXIT_RLC_SAFE_MODE 0 +#define MSG_ENTER_RLC_SAFE_MODE 1 +#define MSG_EXIT_RLC_SAFE_MODE 0

/* * PM4 diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c index 23bd912..874b928 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c @@ -222,7 +222,7 @@ static void cz_ih_decode_iv(struct amdgpu_device *adev, /* wptr/rptr are in bytes! */ u32 ring_index = adev->irq.ih.rptr >> 2; uint32_t dw[4]; - + dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]); dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); diff --git a/drivers/gpu/drm/amd/amdgpu/cz_smumgr.h b/drivers/gpu/drm/amd/amdgpu/cz_smumgr.h index 924d355..026342f 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_smumgr.h +++ b/drivers/gpu/drm/amd/amdgpu/cz_smumgr.h @@ -77,7 +77,7 @@ struct cz_smu_private_data { uint8_t driver_buffer_length; uint8_t scratch_buffer_length; uint16_t toc_entry_used_count; - uint16_t toc_entry_initialize_index; + uint16_t toc_entry_initialize_index; uint16_t toc_entry_power_profiling_index; uint16_t toc_entry_aram; uint16_t toc_entry_ih_register_restore_task_index; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 064f255..2e1ed18 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -603,7 +603,7 @@ static const u32 stoney_golden_settings_a11[] = mmPA_SC_LINE_STIPPLE_STATE, 0x0000ff0f, 0x00000000, mmRLC_CGCG_CGLS_CTRL, 0x00000003, 0x0001003c, mmTA_CNTL_AUX, 0x000f000f, 0x000b0000, - mmTCC_CTRL, 0x00100000, 0xf31fff7f, + mmTCC_CTRL, 0x00100000, 0xf31fff7f, mmTCC_EXE_DISABLE, 0x00000002, 0x00000002, mmTCP_ADDR_CONFIG, 0x0000000f, 0x000000f1, mmTCP_CHAN_STEER_LO, 0xffffffff, 0x10101010, diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index e1d6ae7..55b35da 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -40,9 +40,9 @@

#define GRBM_GFX_INDEX__VCE_INSTANCE__SHIFT 0x04 #define GRBM_GFX_INDEX__VCE_INSTANCE_MASK 0x10 -#define mmVCE_LMI_VCPU_CACHE_40BIT_BAR0 0x8616 -#define mmVCE_LMI_VCPU_CACHE_40BIT_BAR1 0x8617 -#define mmVCE_LMI_VCPU_CACHE_40BIT_BAR2 0x8618 +#define mmVCE_LMI_VCPU_CACHE_40BIT_BAR0 0x8616 +#define mmVCE_LMI_VCPU_CACHE_40BIT_BAR1 0x8617 +#define mmVCE_LMI_VCPU_CACHE_40BIT_BAR2 0x8618

#define VCE_V3_0_FW_SIZE (384 * 1024) #define VCE_V3_0_STACK_SIZE (64 * 1024) diff --git a/drivers/gpu/drm/amd/amdgpu/vid.h b/drivers/gpu/drm/amd/amdgpu/vid.h index ace4997..3bf7172 100644 --- a/drivers/gpu/drm/amd/amdgpu/vid.h +++ b/drivers/gpu/drm/amd/amdgpu/vid.h @@ -365,7 +365,7 @@ #define VCE_CMD_IB 0x00000002 #define VCE_CMD_FENCE 0x00000003 #define VCE_CMD_TRAP 0x00000004 -#define VCE_CMD_IB_AUTO 0x00000005 +#define VCE_CMD_IB_AUTO 0x00000005 #define VCE_CMD_SEMAPHORE 0x00000006

#endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_hwmgr.c index 55e877c..d05a5e0 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/fiji_hwmgr.c @@ -465,14 +465,14 @@ static int fiji_set_private_data_based_on_pptable(struct pp_hwmgr *hwmgr) table_info->vdd_dep_on_mclk;

PP_ASSERT_WITH_CODE(allowed_sclk_vdd_table != NULL, - "VDD dependency on SCLK table is missing. \ + "VDD dependency on SCLK table is missing. \ This table is mandatory", return -EINVAL); PP_ASSERT_WITH_CODE(allowed_sclk_vdd_table->count >= 1, - "VDD dependency on SCLK table has to have is missing. \ + "VDD dependency on SCLK table has to have is missing. \ This table is mandatory", return -EINVAL);

PP_ASSERT_WITH_CODE(allowed_mclk_vdd_table != NULL, - "VDD dependency on MCLK table is missing. \ + "VDD dependency on MCLK table is missing. \ This table is mandatory", return -EINVAL); PP_ASSERT_WITH_CODE(allowed_mclk_vdd_table->count >= 1, "VDD dependency on MCLK table has to have is missing. \ diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/polaris10_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/polaris10_hwmgr.c index 010199f..dbdcc68 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/polaris10_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/polaris10_hwmgr.c @@ -2900,14 +2900,14 @@ static int polaris10_set_private_data_based_on_pptable(struct pp_hwmgr *hwmgr) table_info->vdd_dep_on_mclk;

/* - * PPMCME_FirmwareDescriptorEntry *pfd = NULL; + * PPMCME_FirmwareDescriptorEntry *pfd = NULL; pfd = &tonga_mcmeFirmware; if (0 == PHM_READ_FIELD(hwmgr->device, MC_SEQ_SUP_CNTL, RUN)) polaris10_load_mc_microcode(hwmgr, pfd->dpmThreshold, diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.c b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.c index 8ba3ad5..da9f5f1 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.c @@ -1041,10 +1041,10 @@ int atomctrl_calculate_voltage_evv_on_sclk( }

/** atomctrl_get_voltage_evv_on_sclk gets voltage via call to ATOM COMMAND table. - * @param hwmgr input: pointer to hwManager + * @param hwmgr input: pointer to hwManager * @param voltage_type input: type of EVV voltage VDDC or VDDGFX * @param sclk input: in 10Khz unit. DPM state SCLK frequency - * which is define in PPTable SCLK/VDDC dependence + * which is define in PPTable SCLK/VDDC dependence * table associated with this virtual_voltage_Id * @param virtual_voltage_Id input: voltage id which match per voltage DPM state: 0xff01, 0xff02.. 0xff08 * @param voltage output: real voltage level in unit of mv diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.c index 670b628..d79af48 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.c @@ -2683,7 +2683,7 @@ static int tonga_populate_all_memory_levels(struct pp_hwmgr *hwmgr) struct TONGA_DLL_SPEED_SETTING { uint16_t Min; /* Minimum Data Rate*/ uint16_t Max; /* Maximum Data Rate*/ - uint32_t dll_speed; /* The desired DLL_SPEED setting*/ + uint32_t dll_speed; /* The desired DLL_SPEED setting*/ };

static int tonga_populate_clock_stretcher_data_table(struct pp_hwmgr *hwmgr) @@ -3316,14 +3316,14 @@ static int tonga_set_private_var_based_on_pptale(struct pp_hwmgr *hwmgr) pptable_info->vdd_dep_on_mclk;

PP_ASSERT_WITH_CODE(allowed_sclk_vdd_table != NULL, - "VDD dependency on SCLK table is missing. \ + "VDD dependency on SCLK table is missing. \ This table is mandatory", return -1); PP_ASSERT_WITH_CODE(allowed_sclk_vdd_table->count >= 1, - "VDD dependency on SCLK table has to have is missing. \ + "VDD dependency on SCLK table has to have is missing. \ This table is mandatory", return -1);

PP_ASSERT_WITH_CODE(allowed_mclk_vdd_table != NULL, - "VDD dependency on MCLK table is missing. \ + "VDD dependency on MCLK table is missing. \ This table is mandatory", return -1); PP_ASSERT_WITH_CODE(allowed_mclk_vdd_table->count >= 1, "VDD dependency on MCLK table has to have is missing. \ diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.h b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.h index c6a6b40..573cd39 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/tonga_hwmgr.h @@ -74,7 +74,7 @@ struct tonga_power_state { };

struct _phw_tonga_dpm_level { - bool enabled; + bool enabled; uint32_t value; uint32_t param1; }; @@ -237,20 +237,20 @@ struct tonga_hwmgr { irq_handler_func_t ctf_callback; void *ctf_context;

- phw_tonga_clock_registers clock_registers; + phw_tonga_clock_registers clock_registers; phw_tonga_voltage_smio_registers voltage_smio_registers;

- bool is_memory_GDDR5; + bool is_memory_GDDR5; uint16_t acpi_vddc; - bool pspp_notify_required; /* Flag to indicate if PSPP notification to SBIOS is required */ + bool pspp_notify_required; /* Flag to indicate if PSPP notification to SBIOS is required */ uint16_t force_pcie_gen; /* The forced PCI-E speed if not 0xffff */ uint16_t acpi_pcie_gen; /* The PCI-E speed at ACPI time */ uint32_t pcie_gen_cap; /* The PCI-E speed capabilities bitmap from CAIL */ uint32_t pcie_lane_cap; /* The PCI-E lane capabilities bitmap from CAIL */ uint32_t pcie_spc_cap; /* Symbol Per Clock Capabilities from registry */ - phw_tonga_leakage_voltage vddc_leakage; /* The Leakage VDDC supported (based on leakage ID).*/ - phw_tonga_leakage_voltage vddcgfx_leakage; /* The Leakage VDDC supported (based on leakage ID). */ - phw_tonga_leakage_voltage vddci_leakage; /* The Leakage VDDCI supported (based on leakage ID). */ + phw_tonga_leakage_voltage vddc_leakage; /* The Leakage VDDC supported (based on leakage ID).*/ + phw_tonga_leakage_voltage vddcgfx_leakage; /* The Leakage VDDC supported (based on leakage ID). */ + phw_tonga_leakage_voltage vddci_leakage; /* The Leakage VDDCI supported (based on leakage ID). */

uint32_t mvdd_control; uint32_t vddc_mask_low; @@ -263,8 +263,8 @@ struct tonga_hwmgr { uint32_t mclk_stutter_mode_threshold; uint32_t mclk_edc_enable_threshold; uint32_t mclk_edc_wr_enable_threshold; - bool is_uvd_enabled; - bool is_xdma_enabled; + bool is_uvd_enabled; + bool is_xdma_enabled; phw_tonga_vbios_boot_state vbios_boot_state;

bool battery_state; diff --git a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h index c96e5b1..fd4ce7a 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h @@ -500,7 +500,7 @@ struct phm_dynamic_state_info { struct phm_ppm_table *ppm_parameter_table; struct phm_cac_tdp_table *cac_dtp_table; struct phm_clock_voltage_dependency_table *vdd_gfx_dependency_on_sclk; - struct phm_vq_budgeting_table *vq_budgeting_table; + struct phm_vq_budgeting_table *vq_budgeting_table; };

struct pp_fan_info { diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h index 5561020..178c475 100644 --- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h +++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h @@ -74,7 +74,7 @@ struct amd_sched_fence { struct amd_gpu_scheduler *sched; spinlock_t lock; void *owner; - struct amd_sched_job *s_job; + struct amd_sched_job *s_job; };

struct amd_sched_job {

-- 2.5.5

Alex Deucher

6:26 p.m.

New subject: [PATCH 14/14] drm/amdgpu: fetch cu_info once at init

Fetch this info once at init and just store the results for future requests.

Reviewed-by: Christian König christian.koenig@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 20 +++++++++----------- drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c | 4 +--- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 9 ++++----- drivers/gpu/drm/amd/amdgpu/cik.c | 1 - drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 23 +++++++---------------- drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h | 1 - drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 11 ++++------- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h | 1 - drivers/gpu/drm/amd/amdgpu/vi.c | 1 - 9 files changed, 25 insertions(+), 46 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 19ac6c5..f549474 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1147,6 +1147,12 @@ struct amdgpu_gca_config { uint32_t macrotile_mode_array[16]; };

+struct amdgpu_cu_info { + uint32_t number; /* total active CU number */ + uint32_t ao_cu_mask; + uint32_t bitmap[4][4]; +}; + struct amdgpu_gfx { struct mutex gpu_clock_mutex; struct amdgpu_gca_config config; @@ -1179,9 +1185,10 @@ struct amdgpu_gfx { struct amdgpu_irq_src priv_reg_irq; struct amdgpu_irq_src priv_inst_irq; /* gfx status */ - uint32_t gfx_current_status; + uint32_t gfx_current_status; /* ce ram size*/ - unsigned ce_ram_size; + unsigned ce_ram_size; + struct amdgpu_cu_info cu_info; };

int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, @@ -1793,13 +1800,6 @@ struct amdgpu_allowed_register_entry { bool grbm_indexed; };

-struct amdgpu_cu_info { - uint32_t number; /* total active CU number */ - uint32_t ao_cu_mask; - uint32_t bitmap[4][4]; -}; - - /* * ASIC specific functions. */ @@ -1817,7 +1817,6 @@ struct amdgpu_asic_funcs { u32 (*get_xclk)(struct amdgpu_device *adev); /* get the gpu clock counter */ uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev); - int (*get_cu_info)(struct amdgpu_device *adev, struct amdgpu_cu_info *info); /* MM block clocks */ int (*set_uvd_clocks)(struct amdgpu_device *adev, u32 vclk, u32 dclk); int (*set_vce_clocks)(struct amdgpu_device *adev, u32 evclk, u32 ecclk); @@ -2208,7 +2207,6 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_asic_read_disabled_bios(adev) (adev)->asic_funcs->read_disabled_bios((adev)) #define amdgpu_asic_read_bios_from_rom(adev, b, l) (adev)->asic_funcs->read_bios_from_rom((adev), (b), (l)) #define amdgpu_asic_read_register(adev, se, sh, offset, v)((adev)->asic_funcs->read_register((adev), (se), (sh), (offset), (v))) -#define amdgpu_asic_get_cu_info(adev, info) (adev)->asic_funcs->get_cu_info((adev), (info)) #define amdgpu_gart_flush_gpu_tlb(adev, vmid) (adev)->gart.gart_funcs->flush_gpu_tlb((adev), (vmid)) #define amdgpu_gart_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gart.gart_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags)) #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count))) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index 490464e..199f76b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -793,7 +793,6 @@ static int amdgpu_cgs_query_system_info(struct cgs_device *cgs_device, struct cgs_system_info *sys_info) { CGS_FUNC_ADEV; - struct amdgpu_cu_info cu_info;

if (NULL == sys_info) return -ENODEV; @@ -818,8 +817,7 @@ static int amdgpu_cgs_query_system_info(struct cgs_device *cgs_device, sys_info->value = adev->pg_flags; break; case CGS_SYSTEM_INFO_GFX_CU_INFO: - amdgpu_asic_get_cu_info(adev, &cu_info); - sys_info->value = cu_info.number; + sys_info->value = adev->gfx.cu_info.number; break; default: return -ENODEV; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 4ac83c8..c63866e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -427,7 +427,6 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file } case AMDGPU_INFO_DEV_INFO: { struct drm_amdgpu_info_device dev_info = {}; - struct amdgpu_cu_info cu_info;

dev_info.device_id = dev->pdev->device; dev_info.chip_rev = adev->rev_id; @@ -461,11 +460,11 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file AMDGPU_GPU_PAGE_SIZE; dev_info.gart_page_size = AMDGPU_GPU_PAGE_SIZE;

- amdgpu_asic_get_cu_info(adev, &cu_info); - dev_info.cu_active_number = cu_info.number; - dev_info.cu_ao_mask = cu_info.ao_cu_mask; + dev_info.cu_active_number = adev->gfx.cu_info.number; + dev_info.cu_ao_mask = adev->gfx.cu_info.ao_cu_mask; dev_info.ce_ram_size = adev->gfx.ce_ram_size; - memcpy(&dev_info.cu_bitmap[0], &cu_info.bitmap[0], sizeof(cu_info.bitmap)); + memcpy(&dev_info.cu_bitmap[0], &adev->gfx.cu_info.bitmap[0], + sizeof(adev->gfx.cu_info.bitmap)); dev_info.vram_type = adev->mc.vram_type; dev_info.vram_bit_width = adev->mc.vram_width; dev_info.vce_harvest_config = adev->vce.harvest_config; diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c index c6127d6..68e5698 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik.c +++ b/drivers/gpu/drm/amd/amdgpu/cik.c @@ -2007,7 +2007,6 @@ static const struct amdgpu_asic_funcs cik_asic_funcs = .get_xclk = &cik_get_xclk, .set_uvd_clocks = &cik_set_uvd_clocks, .set_vce_clocks = &cik_set_vce_clocks, - .get_cu_info = &gfx_v7_0_get_cu_info, /* these should be moved to their own ip modules */ .get_gpu_clock_counter = &gfx_v7_0_get_gpu_clock_counter, .wait_for_mc_idle = &gmc_v7_0_mc_wait_for_idle, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 9813819..13d3e96 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -53,7 +53,6 @@ static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev); static void gfx_v7_0_set_irq_funcs(struct amdgpu_device *adev); static void gfx_v7_0_set_gds_init(struct amdgpu_device *adev); -int gfx_v7_0_get_cu_info(struct amdgpu_device *, struct amdgpu_cu_info *);

MODULE_FIRMWARE("radeon/bonaire_pfp.bin"); MODULE_FIRMWARE("radeon/bonaire_me.bin"); @@ -882,6 +881,7 @@ static u32 gfx_v7_0_get_csb_size(struct amdgpu_device *adev); static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, volatile u32 *buffer); static void gfx_v7_0_init_cp_pg_table(struct amdgpu_device *adev); static void gfx_v7_0_init_pg(struct amdgpu_device *adev); +static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev);

/* * Core functions @@ -1718,6 +1718,7 @@ static void gfx_v7_0_gpu_init(struct amdgpu_device *adev) gfx_v7_0_tiling_mode_table_init(adev);

gfx_v7_0_setup_rb(adev); + gfx_v7_0_get_cu_info(adev);

/* set HW defaults for 3D engine */ WREG32(mmCP_MEQ_THRESHOLDS, @@ -3868,18 +3869,13 @@ static u32 gfx_v7_0_get_cu_active_bitmap(struct amdgpu_device *adev)

static void gfx_v7_0_init_ao_cu_mask(struct amdgpu_device *adev) { - uint32_t tmp, active_cu_number; - struct amdgpu_cu_info cu_info; - - gfx_v7_0_get_cu_info(adev, &cu_info); - tmp = cu_info.ao_cu_mask; - active_cu_number = cu_info.number; + u32 tmp;

- WREG32(mmRLC_PG_ALWAYS_ON_CU_MASK, tmp); + WREG32(mmRLC_PG_ALWAYS_ON_CU_MASK, adev->gfx.cu_info.ao_cu_mask);

tmp = RREG32(mmRLC_MAX_PG_CU); tmp &= ~RLC_MAX_PG_CU__MAX_POWERED_UP_CU_MASK; - tmp |= (active_cu_number << RLC_MAX_PG_CU__MAX_POWERED_UP_CU__SHIFT); + tmp |= (adev->gfx.cu_info.number << RLC_MAX_PG_CU__MAX_POWERED_UP_CU__SHIFT); WREG32(mmRLC_MAX_PG_CU, tmp); }

@@ -5014,14 +5010,11 @@ static void gfx_v7_0_set_gds_init(struct amdgpu_device *adev) }

-int gfx_v7_0_get_cu_info(struct amdgpu_device *adev, - struct amdgpu_cu_info *cu_info) +static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev) { int i, j, k, counter, active_cu_number = 0; u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0; - - if (!adev || !cu_info) - return -EINVAL; + struct amdgpu_cu_info *cu_info = &adev->gfx.cu_info;

memset(cu_info, 0, sizeof(*cu_info));

@@ -5052,6 +5045,4 @@ int gfx_v7_0_get_cu_info(struct amdgpu_device *adev,

cu_info->number = active_cu_number; cu_info->ao_cu_mask = ao_cu_mask; - - return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h index c04bfba..e747aa9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h @@ -32,6 +32,5 @@ void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev); void gfx_v7_0_rlc_stop(struct amdgpu_device *adev); uint64_t gfx_v7_0_get_gpu_clock_counter(struct amdgpu_device *adev); void gfx_v7_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num); -int gfx_v7_0_get_cu_info(struct amdgpu_device *adev, struct amdgpu_cu_info *cu_info);

#endif diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 2e1ed18..feda7f7 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -636,6 +636,7 @@ static void gfx_v8_0_set_irq_funcs(struct amdgpu_device *adev); static void gfx_v8_0_set_gds_init(struct amdgpu_device *adev); static void gfx_v8_0_set_rlc_funcs(struct amdgpu_device *adev); static u32 gfx_v8_0_get_csb_size(struct amdgpu_device *adev); +static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev);

static void gfx_v8_0_init_golden_registers(struct amdgpu_device *adev) { @@ -3431,6 +3432,7 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) gfx_v8_0_tiling_mode_table_init(adev);

gfx_v8_0_setup_rb(adev); + gfx_v8_0_get_cu_info(adev);

/* XXX SH_MEM regs */ /* where to put LDS, scratch, GPUVM in FSA64 space */ @@ -6211,14 +6213,11 @@ static u32 gfx_v8_0_get_cu_active_bitmap(struct amdgpu_device *adev) return (~data) & mask; }

-int gfx_v8_0_get_cu_info(struct amdgpu_device *adev, - struct amdgpu_cu_info *cu_info) +static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev) { int i, j, k, counter, active_cu_number = 0; u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0; - - if (!adev || !cu_info) - return -EINVAL; + struct amdgpu_cu_info *cu_info = &adev->gfx.cu_info;

memset(cu_info, 0, sizeof(*cu_info));

@@ -6249,6 +6248,4 @@ int gfx_v8_0_get_cu_info(struct amdgpu_device *adev,

cu_info->number = active_cu_number; cu_info->ao_cu_mask = ao_cu_mask; - - return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h index 021e051..16a49f5 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h @@ -28,6 +28,5 @@ extern const struct amd_ip_funcs gfx_v8_0_ip_funcs;

uint64_t gfx_v8_0_get_gpu_clock_counter(struct amdgpu_device *adev); void gfx_v8_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num); -int gfx_v8_0_get_cu_info(struct amdgpu_device *adev, struct amdgpu_cu_info *cu_info);

#endif diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 31dd630..b7da094 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -1118,7 +1118,6 @@ static const struct amdgpu_asic_funcs vi_asic_funcs = .get_xclk = &vi_get_xclk, .set_uvd_clocks = &vi_set_uvd_clocks, .set_vce_clocks = &vi_set_vce_clocks, - .get_cu_info = &gfx_v8_0_get_cu_info, /* these should be moved to their own ip modules */ .get_gpu_clock_counter = &gfx_v8_0_get_gpu_clock_counter, .wait_for_mc_idle = &gmc_v8_0_mc_wait_for_idle,

-- 2.5.5

Daniel Vetter

9 May 9 May

8:17 a.m.

On Wed, May 04, 2016 at 02:26:42PM -0400, Alex Deucher wrote:

...

From: Chunming Zhou David1.Zhou@amd.com

The release of the vmid owner was not handled correctly. We need to take the lock and walk the lru list.

Signed-off-by: Chunming Zhou David1.Zhou@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Monk Liu monk.liu@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com

I know that it's super hard to get former proprietary driver teams to stick their heads out on a public mailing lists. But imo being steward for them is totally the worst case option you can pick long term. It means you keep all the frustration of them not being fully in control (because sometimes other people from outside the company jump in), never learning how to driver the process themselves. And from the community pov it just looks like code-drop over the wall. In my experience (I've been trying to pull this off in public for almost 4 years now) trying to make exceptions to get folks started just doesn't help anyone.

Imo contributors need to fence for their patches themselves (with you helping behind the scenes ofc) from the start.

Cheers, Daniel

...

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 856116a..e06d066 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1454,6 +1454,7 @@ error_free_sched_entity: void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) { struct amdgpu_bo_va_mapping *mapping, *tmp;

struct amdgpu_vm_id *id, *id_tmp; int i;

amd_sched_entity_fini(vm->entity.sched, &vm->entity);

@@ -1478,14 +1479,17 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) amdgpu_bo_unref(&vm->page_directory); fence_put(vm->page_directory_fence);
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
struct amdgpu_vm_id *id = vm->ids[i];
mutex_lock(&adev->vm_manager.lock);

list_for_each_entry_safe(id, id_tmp, &adev->vm_manager.ids_lru,
		 list) {
if (!id) continue;
atomic_long_cmpxchg(&id->owner, (long)vm, 0);
if (atomic_long_read(&id->owner) == (long)vm) {
	atomic_long_set(&id->owner, 0);
	id->pd_gpu_addr = 0;
}
}
mutex_unlock(&adev->vm_manager.lock);
}

/**

2.5.5

dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel

-- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch

Dave Airlie

10 May 10 May

5:05 a.m.

On 9 May 2016 at 18:17, Daniel Vetter daniel@ffwll.ch wrote:

...

On Wed, May 04, 2016 at 02:26:42PM -0400, Alex Deucher wrote:

...
From: Chunming Zhou David1.Zhou@amd.com

The release of the vmid owner was not handled correctly. We need to take the lock and walk the lru list.

Signed-off-by: Chunming Zhou David1.Zhou@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Monk Liu monk.liu@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com

I know that it's super hard to get former proprietary driver teams to stick their heads out on a public mailing lists. But imo being steward for them is totally the worst case option you can pick long term. It means you keep all the frustration of them not being fully in control (because sometimes other people from outside the company jump in), never learning how to driver the process themselves. And from the community pov it just looks like code-drop over the wall. In my experience (I've been trying to pull this off in public for almost 4 years now) trying to make exceptions to get folks started just doesn't help anyone.

Imo contributors need to fence for their patches themselves (with you helping behind the scenes ofc) from the start.

I'd prefer this as well, I'd also prefer at least people who do develop upstream like Christian be set free to do so again, having patches spring on the lists fully formed isn't really great.

It would be also nice if there was more external discussion around design decisions etc, get the internal patch debate onto the mailing list.

Because at this rate I've no idea about most of the design internals of the VM stuff, and I really think you guys can do a lot better.

Maybe start setup another mailing list like intel-gfx for this, so it's a bit more private than dri-devel, but what is happening at the moment is worse than what used to happen.

Dave.

Christian König

8:21 a.m.

Am 10.05.2016 um 07:05 schrieb Dave Airlie:

...

On 9 May 2016 at 18:17, Daniel Vetter daniel@ffwll.ch wrote:

...
On Wed, May 04, 2016 at 02:26:42PM -0400, Alex Deucher wrote:

...
From: Chunming Zhou David1.Zhou@amd.com

The release of the vmid owner was not handled correctly. We need to take the lock and walk the lru list.

Signed-off-by: Chunming Zhou David1.Zhou@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Monk Liu monk.liu@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com

I know that it's super hard to get former proprietary driver teams to stick their heads out on a public mailing lists. But imo being steward for them is totally the worst case option you can pick long term. It means you keep all the frustration of them not being fully in control (because sometimes other people from outside the company jump in), never learning how to driver the process themselves. And from the community pov it just looks like code-drop over the wall. In my experience (I've been trying to pull this off in public for almost 4 years now) trying to make exceptions to get folks started just doesn't help anyone.

Imo contributors need to fence for their patches themselves (with you helping behind the scenes ofc) from the start.

I'd prefer this as well, I'd also prefer at least people who do develop upstream like Christian be set free to do so again, having patches spring on the lists fully formed isn't really great.

Yeah, completely agree.

I actually tried to work on the public list again. Especially since I'm working on TTM improvements that would make things much easier.

The problem is that then internal people started to complain that some patches only got reviewed and merged upstream, but not internally.

...

It would be also nice if there was more external discussion around design decisions etc, get the internal patch debate onto the mailing list.

Yeah, again completely agree. Alex and I already spend a lot of effort trying to explain the difference between releasing code to the public and making code open source.

...

Because at this rate I've no idea about most of the design internals of the VM stuff, and I really think you guys can do a lot better.

Maybe start setup another mailing list like intel-gfx for this, so it's a bit more private than dri-devel, but what is happening at the moment is worse than what used to happen.

Yeah, that sounds like a good idea to me.

Regards, Christian.

...

Dave. _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel

Daniel Vetter

11 May 11 May

7:46 a.m.

On Tue, May 10, 2016 at 10:21:53AM +0200, Christian König wrote:

...

Am 10.05.2016 um 07:05 schrieb Dave Airlie:

...
On 9 May 2016 at 18:17, Daniel Vetter daniel@ffwll.ch wrote:

...
On Wed, May 04, 2016 at 02:26:42PM -0400, Alex Deucher wrote:

...
From: Chunming Zhou David1.Zhou@amd.com

The release of the vmid owner was not handled correctly. We need to take the lock and walk the lru list.

Signed-off-by: Chunming Zhou David1.Zhou@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Monk Liu monk.liu@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com

I know that it's super hard to get former proprietary driver teams to stick their heads out on a public mailing lists. But imo being steward for them is totally the worst case option you can pick long term. It means you keep all the frustration of them not being fully in control (because sometimes other people from outside the company jump in), never learning how to driver the process themselves. And from the community pov it just looks like code-drop over the wall. In my experience (I've been trying to pull this off in public for almost 4 years now) trying to make exceptions to get folks started just doesn't help anyone.

Imo contributors need to fence for their patches themselves (with you helping behind the scenes ofc) from the start.

I'd prefer this as well, I'd also prefer at least people who do develop upstream like Christian be set free to do so again, having patches spring on the lists fully formed isn't really great.

Yeah, completely agree.

I actually tried to work on the public list again. Especially since I'm working on TTM improvements that would make things much easier.

The problem is that then internal people started to complain that some patches only got reviewed and merged upstream, but not internally.

Here at intel there's no internal review for patches, except things which are embargoed. And even there the internal review is very often just a "you can split out this refactoring step, please submit that part to upstream directly".

...

...
It would be also nice if there was more external discussion around design decisions etc, get the internal patch debate onto the mailing list.

Yeah, again completely agree. Alex and I already spend a lot of effort trying to explain the difference between releasing code to the public and making code open source.

...
Because at this rate I've no idea about most of the design internals of the VM stuff, and I really think you guys can do a lot better.

Maybe start setup another mailing list like intel-gfx for this, so it's a bit more private than dri-devel, but what is happening at the moment is worse than what used to happen.

Yeah, that sounds like a good idea to me.

One thing I do to force the isse, and it's a bit an asshole move, is to just not merge or review anything posted to internal lists that should go upstream. At least with repeat offenders, new people get 2-3 reminders how it's supposed to work. But once people realize that they're just wasting their own time, they tend to learn much faster ;-)

The other thing I learned is that you'll not win any popularity contest as upstream maintainer trying to get a proprietary team to open up :(

Good luck!

Cheers, Daniel

-- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch

Dave Airlie

7:48 a.m.

On 11 May 2016 at 17:46, Daniel Vetter daniel@ffwll.ch wrote:

...

On Tue, May 10, 2016 at 10:21:53AM +0200, Christian König wrote:

...
Am 10.05.2016 um 07:05 schrieb Dave Airlie:

...
On 9 May 2016 at 18:17, Daniel Vetter daniel@ffwll.ch wrote:

...
On Wed, May 04, 2016 at 02:26:42PM -0400, Alex Deucher wrote:

...
From: Chunming Zhou David1.Zhou@amd.com

The release of the vmid owner was not handled correctly. We need to take the lock and walk the lru list.

Signed-off-by: Chunming Zhou David1.Zhou@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com Reviewed-by: Monk Liu monk.liu@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com

I know that it's super hard to get former proprietary driver teams to stick their heads out on a public mailing lists. But imo being steward for them is totally the worst case option you can pick long term. It means you keep all the frustration of them not being fully in control (because sometimes other people from outside the company jump in), never learning how to driver the process themselves. And from the community pov it just looks like code-drop over the wall. In my experience (I've been trying to pull this off in public for almost 4 years now) trying to make exceptions to get folks started just doesn't help anyone.

Imo contributors need to fence for their patches themselves (with you helping behind the scenes ofc) from the start.

I'd prefer this as well, I'd also prefer at least people who do develop upstream like Christian be set free to do so again, having patches spring on the lists fully formed isn't really great.

Yeah, completely agree.

I actually tried to work on the public list again. Especially since I'm working on TTM improvements that would make things much easier.

The problem is that then internal people started to complain that some patches only got reviewed and merged upstream, but not internally.

Here at intel there's no internal review for patches, except things which are embargoed. And even there the internal review is very often just a "you can split out this refactoring step, please submit that part to upstream directly".

...
...
It would be also nice if there was more external discussion around design decisions etc, get the internal patch debate onto the mailing list.

Yeah, again completely agree. Alex and I already spend a lot of effort trying to explain the difference between releasing code to the public and making code open source.

...
Because at this rate I've no idea about most of the design internals of the VM stuff, and I really think you guys can do a lot better.

Maybe start setup another mailing list like intel-gfx for this, so it's a bit more private than dri-devel, but what is happening at the moment is worse than what used to happen.

Yeah, that sounds like a good idea to me.

One thing I do to force the isse, and it's a bit an asshole move, is to just not merge or review anything posted to internal lists that should go upstream. At least with repeat offenders, new people get 2-3 reminders how it's supposed to work. But once people realize that they're just wasting their own time, they tend to learn much faster ;-)

The other thing I learned is that you'll not win any popularity contest as upstream maintainer trying to get a proprietary team to open up :(

My feeling is things haven't gotten any better, in fact they may have gotten worse, if Christian is having such problems.

At this point I'm guessing you are going to have to use a stick rather than carrots unfortunately.

Nobody seems like they want to work in the open without encouragment.

Dave.

3289

Age (days ago)

3296

Last active (days ago)

dri-devel@lists.freedesktop.org

18 comments

4 participants

tags (0)

participants (4)

Alex Deucher
Christian König
Daniel Vetter
Dave Airlie