From 0e3f154a9eb9f7a5ec365c4586a606ba882a3f15 Mon Sep 17 00:00:00 2001
From: Chunming Zhou <david1.zhou@amd.com>
Date: Mon, 3 Aug 2015 13:11:04 +0800
Subject: drm/amdgpu: change uvd ib test to use kernel fence directly
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian K?nig <christian.koenig@amd.com>
Reviewed-by: Jammy Zhou <jammy.zhou@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_test.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_test.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
index df202999fbfe..567c5b3c6f3a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
@@ -244,17 +244,19 @@ static int amdgpu_test_create_and_emit_fence(struct amdgpu_device *adev,
int r;
if (ring == &adev->uvd.ring) {
+ struct fence *f = NULL;
r = amdgpu_uvd_get_create_msg(ring, handle, NULL);
if (r) {
DRM_ERROR("Failed to get dummy create msg\n");
return r;
}
- r = amdgpu_uvd_get_destroy_msg(ring, handle, fence);
+ r = amdgpu_uvd_get_destroy_msg(ring, handle, &f);
if (r) {
DRM_ERROR("Failed to get dummy destroy msg\n");
return r;
}
+ *fence = to_amdgpu_fence(f);
} else if (ring == &adev->vce.ring[0] ||
ring == &adev->vce.ring[1]) {
--
cgit v1.2.3-70-g09d2
From ed40bfb81af54fcdc1b96ffdeb14d69458234642 Mon Sep 17 00:00:00 2001
From: Chunming Zhou <david1.zhou@amd.com>
Date: Mon, 3 Aug 2015 13:28:16 +0800
Subject: drm/amdgpu: use kernel fence for vce ib test
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian K?nig <christian.koenig@amd.com>
Reviewed-by: Jammy Zhou <jammy.zhou@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_test.c | 5 +++--
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 14 +++++++-------
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h | 4 ++--
3 files changed, 12 insertions(+), 11 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_test.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
index 567c5b3c6f3a..251b1ebef722 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
@@ -260,18 +260,19 @@ static int amdgpu_test_create_and_emit_fence(struct amdgpu_device *adev,
} else if (ring == &adev->vce.ring[0] ||
ring == &adev->vce.ring[1]) {
+ struct fence *f = NULL;
r = amdgpu_vce_get_create_msg(ring, handle, NULL);
if (r) {
DRM_ERROR("Failed to get dummy create msg\n");
return r;
}
- r = amdgpu_vce_get_destroy_msg(ring, handle, fence);
+ r = amdgpu_vce_get_destroy_msg(ring, handle, &f);
if (r) {
DRM_ERROR("Failed to get dummy destroy msg\n");
return r;
}
-
+ *fence = to_amdgpu_fence(f);
} else {
r = amdgpu_ring_lock(ring, 64);
if (r) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
index 59acb0b990e1..94c40ca5ba74 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
@@ -358,7 +358,7 @@ static int amdgpu_vce_free_job(
* Open up a stream for HW test
*/
int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
- struct amdgpu_fence **fence)
+ struct fence **fence)
{
const unsigned ib_size_dw = 1024;
struct amdgpu_ib *ib = NULL;
@@ -412,7 +412,7 @@ int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
if (r)
goto err;
if (fence)
- *fence = amdgpu_fence_ref(ib->fence);
+ *fence = fence_get(&ib->fence->base);
if (amdgpu_enable_scheduler)
return 0;
err:
@@ -432,7 +432,7 @@ err:
* Close up a stream for HW test or if userspace failed to do so
*/
int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
- struct amdgpu_fence **fence)
+ struct fence **fence)
{
const unsigned ib_size_dw = 1024;
struct amdgpu_ib *ib = NULL;
@@ -476,7 +476,7 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
if (r)
goto err;
if (fence)
- *fence = amdgpu_fence_ref(ib->fence);
+ *fence = fence_get(&ib->fence->base);
if (amdgpu_enable_scheduler)
return 0;
err:
@@ -827,7 +827,7 @@ int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring)
*/
int amdgpu_vce_ring_test_ib(struct amdgpu_ring *ring)
{
- struct amdgpu_fence *fence = NULL;
+ struct fence *fence = NULL;
int r;
r = amdgpu_vce_get_create_msg(ring, 1, NULL);
@@ -842,13 +842,13 @@ int amdgpu_vce_ring_test_ib(struct amdgpu_ring *ring)
goto error;
}
- r = amdgpu_fence_wait(fence, false);
+ r = fence_wait(fence, false);
if (r) {
DRM_ERROR("amdgpu: fence wait failed (%d).\n", r);
} else {
DRM_INFO("ib test on ring %d succeeded\n", ring->idx);
}
error:
- amdgpu_fence_unref(&fence);
+ fence_put(fence);
return r;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
index 7ccdb5927da5..ba2da8ee5906 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
@@ -29,9 +29,9 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev);
int amdgpu_vce_suspend(struct amdgpu_device *adev);
int amdgpu_vce_resume(struct amdgpu_device *adev);
int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
- struct amdgpu_fence **fence);
+ struct fence **fence);
int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
- struct amdgpu_fence **fence);
+ 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);
bool amdgpu_vce_ring_emit_semaphore(struct amdgpu_ring *ring,
--
cgit v1.2.3-70-g09d2
From 6863bc8bb3750049d1b63574ef29cbf94d52eb47 Mon Sep 17 00:00:00 2001
From: Chunming Zhou <david1.zhou@amd.com>
Date: Mon, 3 Aug 2015 16:19:29 +0800
Subject: drm/amdgpu: use kernel fence in amdgpu_test
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian K?nig <christian.koenig@amd.com>
Reviewed-by: Jammy Zhou <jammy.zhou@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_test.c | 46 +++++++++++++++-----------------
1 file changed, 22 insertions(+), 24 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_test.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
index 251b1ebef722..afb57de16521 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
@@ -238,49 +238,47 @@ void amdgpu_test_moves(struct amdgpu_device *adev)
static int amdgpu_test_create_and_emit_fence(struct amdgpu_device *adev,
struct amdgpu_ring *ring,
- struct amdgpu_fence **fence)
+ struct fence **fence)
{
uint32_t handle = ring->idx ^ 0xdeafbeef;
int r;
if (ring == &adev->uvd.ring) {
- struct fence *f = NULL;
r = amdgpu_uvd_get_create_msg(ring, handle, NULL);
if (r) {
DRM_ERROR("Failed to get dummy create msg\n");
return r;
}
- r = amdgpu_uvd_get_destroy_msg(ring, handle, &f);
+ r = amdgpu_uvd_get_destroy_msg(ring, handle, fence);
if (r) {
DRM_ERROR("Failed to get dummy destroy msg\n");
return r;
}
- *fence = to_amdgpu_fence(f);
} else if (ring == &adev->vce.ring[0] ||
ring == &adev->vce.ring[1]) {
- struct fence *f = NULL;
r = amdgpu_vce_get_create_msg(ring, handle, NULL);
if (r) {
DRM_ERROR("Failed to get dummy create msg\n");
return r;
}
- r = amdgpu_vce_get_destroy_msg(ring, handle, &f);
+ r = amdgpu_vce_get_destroy_msg(ring, handle, fence);
if (r) {
DRM_ERROR("Failed to get dummy destroy msg\n");
return r;
}
- *fence = to_amdgpu_fence(f);
} else {
+ struct amdgpu_fence *a_fence = NULL;
r = amdgpu_ring_lock(ring, 64);
if (r) {
DRM_ERROR("Failed to lock ring A %d\n", ring->idx);
return r;
}
- amdgpu_fence_emit(ring, AMDGPU_FENCE_OWNER_UNDEFINED, fence);
+ amdgpu_fence_emit(ring, AMDGPU_FENCE_OWNER_UNDEFINED, &a_fence);
amdgpu_ring_unlock_commit(ring);
+ *fence = &a_fence->base;
}
return 0;
}
@@ -289,7 +287,7 @@ void amdgpu_test_ring_sync(struct amdgpu_device *adev,
struct amdgpu_ring *ringA,
struct amdgpu_ring *ringB)
{
- struct amdgpu_fence *fence1 = NULL, *fence2 = NULL;
+ struct fence *fence1 = NULL, *fence2 = NULL;
struct amdgpu_semaphore *semaphore = NULL;
int r;
@@ -325,7 +323,7 @@ void amdgpu_test_ring_sync(struct amdgpu_device *adev,
mdelay(1000);
- if (amdgpu_fence_signaled(fence1)) {
+ if (fence_is_signaled(fence1)) {
DRM_ERROR("Fence 1 signaled without waiting for semaphore.\n");
goto out_cleanup;
}
@@ -338,7 +336,7 @@ void amdgpu_test_ring_sync(struct amdgpu_device *adev,
amdgpu_semaphore_emit_signal(ringB, semaphore);
amdgpu_ring_unlock_commit(ringB);
- r = amdgpu_fence_wait(fence1, false);
+ r = fence_wait(fence1, false);
if (r) {
DRM_ERROR("Failed to wait for sync fence 1\n");
goto out_cleanup;
@@ -346,7 +344,7 @@ void amdgpu_test_ring_sync(struct amdgpu_device *adev,
mdelay(1000);
- if (amdgpu_fence_signaled(fence2)) {
+ if (fence_is_signaled(fence2)) {
DRM_ERROR("Fence 2 signaled without waiting for semaphore.\n");
goto out_cleanup;
}
@@ -359,7 +357,7 @@ void amdgpu_test_ring_sync(struct amdgpu_device *adev,
amdgpu_semaphore_emit_signal(ringB, semaphore);
amdgpu_ring_unlock_commit(ringB);
- r = amdgpu_fence_wait(fence2, false);
+ r = fence_wait(fence2, false);
if (r) {
DRM_ERROR("Failed to wait for sync fence 1\n");
goto out_cleanup;
@@ -369,10 +367,10 @@ out_cleanup:
amdgpu_semaphore_free(adev, &semaphore, NULL);
if (fence1)
- amdgpu_fence_unref(&fence1);
+ fence_put(fence1);
if (fence2)
- amdgpu_fence_unref(&fence2);
+ fence_put(fence2);
if (r)
printk(KERN_WARNING "Error while testing ring sync (%d).\n", r);
@@ -383,7 +381,7 @@ static void amdgpu_test_ring_sync2(struct amdgpu_device *adev,
struct amdgpu_ring *ringB,
struct amdgpu_ring *ringC)
{
- struct amdgpu_fence *fenceA = NULL, *fenceB = NULL;
+ struct fence *fenceA = NULL, *fenceB = NULL;
struct amdgpu_semaphore *semaphore = NULL;
bool sigA, sigB;
int i, r;
@@ -419,11 +417,11 @@ static void amdgpu_test_ring_sync2(struct amdgpu_device *adev,
mdelay(1000);
- if (amdgpu_fence_signaled(fenceA)) {
+ if (fence_is_signaled(fenceA)) {
DRM_ERROR("Fence A signaled without waiting for semaphore.\n");
goto out_cleanup;
}
- if (amdgpu_fence_signaled(fenceB)) {
+ if (fence_is_signaled(fenceB)) {
DRM_ERROR("Fence B signaled without waiting for semaphore.\n");
goto out_cleanup;
}
@@ -438,8 +436,8 @@ static void amdgpu_test_ring_sync2(struct amdgpu_device *adev,
for (i = 0; i < 30; ++i) {
mdelay(100);
- sigA = amdgpu_fence_signaled(fenceA);
- sigB = amdgpu_fence_signaled(fenceB);
+ sigA = fence_is_signaled(fenceA);
+ sigB = fence_is_signaled(fenceB);
if (sigA || sigB)
break;
}
@@ -464,12 +462,12 @@ static void amdgpu_test_ring_sync2(struct amdgpu_device *adev,
mdelay(1000);
- r = amdgpu_fence_wait(fenceA, false);
+ r = fence_wait(fenceA, false);
if (r) {
DRM_ERROR("Failed to wait for sync fence A\n");
goto out_cleanup;
}
- r = amdgpu_fence_wait(fenceB, false);
+ r = fence_wait(fenceB, false);
if (r) {
DRM_ERROR("Failed to wait for sync fence B\n");
goto out_cleanup;
@@ -479,10 +477,10 @@ out_cleanup:
amdgpu_semaphore_free(adev, &semaphore, NULL);
if (fenceA)
- amdgpu_fence_unref(&fenceA);
+ fence_put(fenceA);
if (fenceB)
- amdgpu_fence_unref(&fenceB);
+ fence_put(fenceB);
if (r)
printk(KERN_WARNING "Error while testing ring sync (%d).\n", r);
--
cgit v1.2.3-70-g09d2
From 02bc0650bc930105346773253e5bf33c2917c601 Mon Sep 17 00:00:00 2001
From: Christian König <christian.koenig@amd.com>
Date: Fri, 7 Aug 2015 13:53:36 +0200
Subject: drm/amdgpu: remove amdgpu_fence_wait
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
It was just a wrapper for fence_wait anyway.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 -
drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_display.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c | 22 ----------------------
drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c | 6 +++---
drivers/gpu/drm/amd/amdgpu/amdgpu_test.c | 4 ++--
6 files changed, 7 insertions(+), 30 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_test.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index abf1650f6936..7a3a00fd339c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -439,7 +439,6 @@ int amdgpu_fence_wait_next(struct amdgpu_ring *ring);
int amdgpu_fence_wait_empty(struct amdgpu_ring *ring);
unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring);
-int amdgpu_fence_wait(struct amdgpu_fence *fence, bool interruptible);
signed long amdgpu_fence_wait_any(struct amdgpu_device *adev,
struct amdgpu_fence **fences,
bool intr, long t);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
index 2742b9a35cbc..759482e4300d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
@@ -42,7 +42,7 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size,
r = amdgpu_copy_buffer(ring, saddr, daddr, size, NULL, &fence);
if (r)
goto exit_do_move;
- r = amdgpu_fence_wait(fence, false);
+ r = fence_wait(&fence->base, false);
if (r)
goto exit_do_move;
amdgpu_fence_unref(&fence);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
index b16b9256883e..e12931ea66e4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
@@ -52,7 +52,7 @@ static void amdgpu_flip_work_func(struct work_struct *__work)
if (work->fence) {
fence = to_amdgpu_fence(work->fence);
if (fence) {
- r = amdgpu_fence_wait(fence, false);
+ r = fence_wait(&fence->base, false);
if (r == -EDEADLK) {
up_read(&adev->exclusive_lock);
r = amdgpu_gpu_reset(adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
index 5a2a15cf1a7c..1097259df3c2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
@@ -497,28 +497,6 @@ static long amdgpu_fence_ring_wait_seq_timeout(struct amdgpu_ring *ring, uint64_
}
}
-
-/**
- * amdgpu_fence_wait - wait for a fence to signal
- *
- * @fence: amdgpu fence object
- * @intr: use interruptable sleep
- *
- * Wait for the requested fence to signal (all asics).
- * @intr selects whether to use interruptable (true) or non-interruptable
- * (false) sleep when waiting for the fence.
- * Returns 0 if the fence has passed, error for all other cases.
- */
-int amdgpu_fence_wait(struct amdgpu_fence *fence, bool intr)
-{
- long r;
-
- r = fence_wait_timeout(&fence->base, intr, MAX_SCHEDULE_TIMEOUT);
- if (r < 0)
- return r;
- return 0;
-}
-
/**
* amdgpu_fence_wait_next - wait for the next fence to signal
*
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
index 2c42f50912b9..7cb711fc1ee2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
@@ -180,7 +180,7 @@ int amdgpu_sync_rings(struct amdgpu_sync *sync,
if (amdgpu_enable_scheduler || (count >= AMDGPU_NUM_SYNCS)) {
/* not enough room, wait manually */
- r = amdgpu_fence_wait(fence, false);
+ r = fence_wait(&fence->base, false);
if (r)
return r;
continue;
@@ -200,7 +200,7 @@ int amdgpu_sync_rings(struct amdgpu_sync *sync,
if (!amdgpu_semaphore_emit_signal(other, semaphore)) {
/* signaling wasn't successful wait manually */
amdgpu_ring_undo(other);
- r = amdgpu_fence_wait(fence, false);
+ r = fence_wait(&fence->base, false);
if (r)
return r;
continue;
@@ -210,7 +210,7 @@ int amdgpu_sync_rings(struct amdgpu_sync *sync,
if (!amdgpu_semaphore_emit_wait(ring, semaphore)) {
/* waiting wasn't successful wait manually */
amdgpu_ring_undo(other);
- r = amdgpu_fence_wait(fence, false);
+ r = fence_wait(&fence->base, false);
if (r)
return r;
continue;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
index afb57de16521..962dd5552137 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
@@ -116,7 +116,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
goto out_lclean_unpin;
}
- r = amdgpu_fence_wait(fence, false);
+ r = fence_wait(&fence->base, false);
if (r) {
DRM_ERROR("Failed to wait for GTT->VRAM fence %d\n", i);
goto out_lclean_unpin;
@@ -161,7 +161,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
goto out_lclean_unpin;
}
- r = amdgpu_fence_wait(fence, false);
+ r = fence_wait(&fence->base, false);
if (r) {
DRM_ERROR("Failed to wait for VRAM->GTT fence %d\n", i);
goto out_lclean_unpin;
--
cgit v1.2.3-70-g09d2
From c7ae72c01be10f539f385f624713f8ba0aa11a8f Mon Sep 17 00:00:00 2001
From: Chunming Zhou <david1.zhou@amd.com>
Date: Tue, 25 Aug 2015 17:23:45 +0800
Subject: drm/amdgpu: use IB for copy buffer of eviction
This aids handling buffers moves with the scheduler.
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian K?nig <christian.koenig@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 8 +--
drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c | 8 +--
drivers/gpu/drm/amd/amdgpu/amdgpu_test.c | 12 ++---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 77 ++++++++++++++-------------
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 3 +-
drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 16 +++---
drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 18 +++----
drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 18 +++----
8 files changed, 81 insertions(+), 79 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_test.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index c9160430b5ac..aa2dcf578dd6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -247,7 +247,7 @@ struct amdgpu_buffer_funcs {
unsigned copy_num_dw;
/* used for buffer migration */
- void (*emit_copy_buffer)(struct amdgpu_ring *ring,
+ void (*emit_copy_buffer)(struct amdgpu_ib *ib,
/* src addr in bytes */
uint64_t src_offset,
/* dst addr in bytes */
@@ -518,7 +518,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring,
uint64_t dst_offset,
uint32_t byte_count,
struct reservation_object *resv,
- struct amdgpu_fence **fence);
+ struct fence **fence);
int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma);
struct amdgpu_bo_list_entry {
@@ -2247,7 +2247,7 @@ static inline void amdgpu_ring_write(struct amdgpu_ring *ring, uint32_t v)
#define amdgpu_display_add_connector(adev, ci, sd, ct, ib, coi, h, r) (adev)->mode_info.funcs->add_connector((adev), (ci), (sd), (ct), (ib), (coi), (h), (r))
#define amdgpu_display_stop_mc_access(adev, s) (adev)->mode_info.funcs->stop_mc_access((adev), (s))
#define amdgpu_display_resume_mc_access(adev, s) (adev)->mode_info.funcs->resume_mc_access((adev), (s))
-#define amdgpu_emit_copy_buffer(adev, r, s, d, b) (adev)->mman.buffer_funcs->emit_copy_buffer((r), (s), (d), (b))
+#define amdgpu_emit_copy_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_copy_buffer((ib), (s), (d), (b))
#define amdgpu_emit_fill_buffer(adev, r, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((r), (s), (d), (b))
#define amdgpu_dpm_get_temperature(adev) (adev)->pm.funcs->get_temperature((adev))
#define amdgpu_dpm_pre_set_power_state(adev) (adev)->pm.funcs->pre_set_power_state((adev))
@@ -2379,7 +2379,7 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
uint64_t addr);
void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
struct amdgpu_bo_va *bo_va);
-
+int amdgpu_vm_free_job(struct amdgpu_job *job);
/*
* functions used by amdgpu_encoder.c
*/
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
index 759482e4300d..98d59ee640ce 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
@@ -33,7 +33,7 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size,
{
unsigned long start_jiffies;
unsigned long end_jiffies;
- struct amdgpu_fence *fence = NULL;
+ struct fence *fence = NULL;
int i, r;
start_jiffies = jiffies;
@@ -42,17 +42,17 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size,
r = amdgpu_copy_buffer(ring, saddr, daddr, size, NULL, &fence);
if (r)
goto exit_do_move;
- r = fence_wait(&fence->base, false);
+ r = fence_wait(fence, false);
if (r)
goto exit_do_move;
- amdgpu_fence_unref(&fence);
+ fence_put(fence);
}
end_jiffies = jiffies;
r = jiffies_to_msecs(end_jiffies - start_jiffies);
exit_do_move:
if (fence)
- amdgpu_fence_unref(&fence);
+ fence_put(fence);
return r;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
index 962dd5552137..f80b1a43be8a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
@@ -77,7 +77,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
void *gtt_map, *vram_map;
void **gtt_start, **gtt_end;
void **vram_start, **vram_end;
- struct amdgpu_fence *fence = NULL;
+ struct fence *fence = NULL;
r = amdgpu_bo_create(adev, size, PAGE_SIZE, true,
AMDGPU_GEM_DOMAIN_GTT, 0, NULL, gtt_obj + i);
@@ -116,13 +116,13 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
goto out_lclean_unpin;
}
- r = fence_wait(&fence->base, false);
+ r = fence_wait(fence, false);
if (r) {
DRM_ERROR("Failed to wait for GTT->VRAM fence %d\n", i);
goto out_lclean_unpin;
}
- amdgpu_fence_unref(&fence);
+ fence_put(fence);
r = amdgpu_bo_kmap(vram_obj, &vram_map);
if (r) {
@@ -161,13 +161,13 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
goto out_lclean_unpin;
}
- r = fence_wait(&fence->base, false);
+ r = fence_wait(fence, false);
if (r) {
DRM_ERROR("Failed to wait for VRAM->GTT fence %d\n", i);
goto out_lclean_unpin;
}
- amdgpu_fence_unref(&fence);
+ fence_put(fence);
r = amdgpu_bo_kmap(gtt_obj[i], >t_map);
if (r) {
@@ -214,7 +214,7 @@ out_lclean:
amdgpu_bo_unref(>t_obj[i]);
}
if (fence)
- amdgpu_fence_unref(&fence);
+ fence_put(fence);
break;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 4cb81320b045..399143541d8a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -228,7 +228,7 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo,
struct amdgpu_device *adev;
struct amdgpu_ring *ring;
uint64_t old_start, new_start;
- struct amdgpu_fence *fence;
+ struct fence *fence;
int r;
adev = amdgpu_get_adev(bo->bdev);
@@ -269,9 +269,9 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo,
new_mem->num_pages * PAGE_SIZE, /* bytes */
bo->resv, &fence);
/* FIXME: handle copy error */
- r = ttm_bo_move_accel_cleanup(bo, &fence->base,
+ r = ttm_bo_move_accel_cleanup(bo, fence,
evict, no_wait_gpu, new_mem);
- amdgpu_fence_unref(&fence);
+ fence_put(fence);
return r;
}
@@ -987,52 +987,48 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring,
uint64_t dst_offset,
uint32_t byte_count,
struct reservation_object *resv,
- struct amdgpu_fence **fence)
+ struct fence **fence)
{
struct amdgpu_device *adev = ring->adev;
- struct amdgpu_sync sync;
uint32_t max_bytes;
unsigned num_loops, num_dw;
+ struct amdgpu_ib *ib;
unsigned i;
int r;
- /* sync other rings */
- amdgpu_sync_create(&sync);
- if (resv) {
- r = amdgpu_sync_resv(adev, &sync, resv, false);
- if (r) {
- DRM_ERROR("sync failed (%d).\n", r);
- amdgpu_sync_free(adev, &sync, NULL);
- return r;
- }
- }
-
max_bytes = adev->mman.buffer_funcs->copy_max_bytes;
num_loops = DIV_ROUND_UP(byte_count, max_bytes);
num_dw = num_loops * adev->mman.buffer_funcs->copy_num_dw;
- /* for fence and sync */
- num_dw += 64 + AMDGPU_NUM_SYNCS * 8;
+ /* for IB padding */
+ while (num_dw & 0x7)
+ num_dw++;
+
+ ib = kzalloc(sizeof(struct amdgpu_ib), GFP_KERNEL);
+ if (!ib)
+ return -ENOMEM;
- r = amdgpu_sync_wait(&sync);
+ r = amdgpu_ib_get(ring, NULL, num_dw * 4, ib);
if (r) {
- DRM_ERROR("sync wait failed (%d).\n", r);
- amdgpu_sync_free(adev, &sync, NULL);
+ kfree(ib);
return r;
}
- r = amdgpu_ring_lock(ring, num_dw);
- if (r) {
- DRM_ERROR("ring lock failed (%d).\n", r);
- amdgpu_sync_free(adev, &sync, NULL);
- return r;
+ ib->length_dw = 0;
+
+ if (resv) {
+ r = amdgpu_sync_resv(adev, &ib->sync, resv,
+ AMDGPU_FENCE_OWNER_UNDEFINED);
+ if (r) {
+ DRM_ERROR("sync failed (%d).\n", r);
+ goto error_free;
+ }
}
- amdgpu_sync_rings(&sync, ring);
for (i = 0; i < num_loops; i++) {
uint32_t cur_size_in_bytes = min(byte_count, max_bytes);
- amdgpu_emit_copy_buffer(adev, ring, src_offset, dst_offset,
+ amdgpu_emit_copy_buffer(adev, ib, src_offset, dst_offset,
cur_size_in_bytes);
src_offset += cur_size_in_bytes;
@@ -1040,17 +1036,24 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring,
byte_count -= cur_size_in_bytes;
}
- r = amdgpu_fence_emit(ring, AMDGPU_FENCE_OWNER_MOVE, fence);
- if (r) {
- amdgpu_ring_unlock_undo(ring);
- amdgpu_sync_free(adev, &sync, NULL);
- return r;
- }
-
- amdgpu_ring_unlock_commit(ring);
- amdgpu_sync_free(adev, &sync, &(*fence)->base);
+ amdgpu_vm_pad_ib(adev, ib);
+ WARN_ON(ib->length_dw > num_dw);
+ r = amdgpu_sched_ib_submit_kernel_helper(adev, ring, ib, 1,
+ &amdgpu_vm_free_job,
+ AMDGPU_FENCE_OWNER_MOVE,
+ fence);
+ if (r)
+ goto error_free;
+ if (!amdgpu_enable_scheduler) {
+ amdgpu_ib_free(adev, ib);
+ kfree(ib);
+ }
return 0;
+error_free:
+ amdgpu_ib_free(adev, ib);
+ kfree(ib);
+ return r;
}
#if defined(CONFIG_DEBUG_FS)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 677266b90c74..83b7ce6f5f72 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -316,8 +316,7 @@ static void amdgpu_vm_update_pages(struct amdgpu_device *adev,
}
}
-static int amdgpu_vm_free_job(
- struct amdgpu_job *sched_job)
+int amdgpu_vm_free_job(struct amdgpu_job *sched_job)
{
int i;
for (i = 0; i < sched_job->num_ibs; i++)
diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
index af526557ac66..3920c1e346f8 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
@@ -1339,18 +1339,18 @@ static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev)
* Used by the amdgpu ttm implementation to move pages if
* registered as the asic copy callback.
*/
-static void cik_sdma_emit_copy_buffer(struct amdgpu_ring *ring,
+static void cik_sdma_emit_copy_buffer(struct amdgpu_ib *ib,
uint64_t src_offset,
uint64_t dst_offset,
uint32_t byte_count)
{
- amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0));
- amdgpu_ring_write(ring, byte_count);
- amdgpu_ring_write(ring, 0); /* src/dst endian swap */
- amdgpu_ring_write(ring, lower_32_bits(src_offset));
- amdgpu_ring_write(ring, upper_32_bits(src_offset));
- amdgpu_ring_write(ring, lower_32_bits(dst_offset));
- amdgpu_ring_write(ring, upper_32_bits(dst_offset));
+ ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0);
+ ib->ptr[ib->length_dw++] = byte_count;
+ ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
+ ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
+ ib->ptr[ib->length_dw++] = upper_32_bits(src_offset);
+ ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset);
+ ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset);
}
/**
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
index 2b0e89e05b1d..715e02d3bfba 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
@@ -1350,19 +1350,19 @@ static void sdma_v2_4_set_irq_funcs(struct amdgpu_device *adev)
* Used by the amdgpu ttm implementation to move pages if
* registered as the asic copy callback.
*/
-static void sdma_v2_4_emit_copy_buffer(struct amdgpu_ring *ring,
+static void sdma_v2_4_emit_copy_buffer(struct amdgpu_ib *ib,
uint64_t src_offset,
uint64_t dst_offset,
uint32_t byte_count)
{
- amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
- SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR));
- amdgpu_ring_write(ring, byte_count);
- amdgpu_ring_write(ring, 0); /* src/dst endian swap */
- amdgpu_ring_write(ring, lower_32_bits(src_offset));
- amdgpu_ring_write(ring, upper_32_bits(src_offset));
- amdgpu_ring_write(ring, lower_32_bits(dst_offset));
- amdgpu_ring_write(ring, upper_32_bits(dst_offset));
+ ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
+ SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR);
+ ib->ptr[ib->length_dw++] = byte_count;
+ ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
+ ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
+ ib->ptr[ib->length_dw++] = upper_32_bits(src_offset);
+ ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset);
+ ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset);
}
/**
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
index 6f1df03ebc70..67128c8e78b8 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
@@ -1474,19 +1474,19 @@ static void sdma_v3_0_set_irq_funcs(struct amdgpu_device *adev)
* Used by the amdgpu ttm implementation to move pages if
* registered as the asic copy callback.
*/
-static void sdma_v3_0_emit_copy_buffer(struct amdgpu_ring *ring,
+static void sdma_v3_0_emit_copy_buffer(struct amdgpu_ib *ib,
uint64_t src_offset,
uint64_t dst_offset,
uint32_t byte_count)
{
- amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
- SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR));
- amdgpu_ring_write(ring, byte_count);
- amdgpu_ring_write(ring, 0); /* src/dst endian swap */
- amdgpu_ring_write(ring, lower_32_bits(src_offset));
- amdgpu_ring_write(ring, upper_32_bits(src_offset));
- amdgpu_ring_write(ring, lower_32_bits(dst_offset));
- amdgpu_ring_write(ring, upper_32_bits(dst_offset));
+ ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
+ SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR);
+ ib->ptr[ib->length_dw++] = byte_count;
+ ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
+ ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
+ ib->ptr[ib->length_dw++] = upper_32_bits(src_offset);
+ ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset);
+ ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset);
}
/**
--
cgit v1.2.3-70-g09d2