1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
|
From 26ec1581529fbccb5bfbf5ec61d2711dd9c38622 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Christian=20K=C3=B6nig?= <christian.koenig@amd.com>
Date: Mon, 18 Jan 2016 17:01:42 +0100
Subject: [PATCH 0238/1110] drm/amdgpu: merge vm_grab_id and vm_fence v2
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
No need for an extra function any more.
v2: comment cleanups
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Kalyan Alle <kalyan.alle@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 5 +--
drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c | 13 +++----
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 57 ++++++++++++++-----------------
3 files changed, 30 insertions(+), 45 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index f970afa..64784e4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -924,13 +924,10 @@ void amdgpu_vm_get_pt_bos(struct amdgpu_vm *vm, struct list_head *duplicates);
void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev,
struct amdgpu_vm *vm);
int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
- struct amdgpu_sync *sync);
+ struct amdgpu_sync *sync, struct fence *fence);
void amdgpu_vm_flush(struct amdgpu_ring *ring,
struct amdgpu_vm *vm,
struct fence *updates);
-void amdgpu_vm_fence(struct amdgpu_device *adev,
- struct amdgpu_vm *vm,
- struct fence *fence);
uint64_t amdgpu_vm_map_gart(struct amdgpu_device *adev, uint64_t addr);
int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
struct amdgpu_vm *vm);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
index b22a95f..76a1f82 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
@@ -38,19 +38,14 @@ static struct fence *amdgpu_sched_dependency(struct amd_sched_job *sched_job)
if (fence == NULL && vm && !job->ibs->grabbed_vmid) {
struct amdgpu_ring *ring = job->ibs->ring;
- struct amdgpu_device *adev = ring->adev;
int r;
- mutex_lock(&adev->vm_manager.lock);
- r = amdgpu_vm_grab_id(vm, ring, sync);
- if (r) {
+ r = amdgpu_vm_grab_id(vm, ring, sync,
+ &job->base.s_fence->base);
+ if (r)
DRM_ERROR("Error getting VM ID (%d)\n", r);
- } else {
- fence = &job->base.s_fence->base;
- amdgpu_vm_fence(ring->adev, vm, fence);
+ else
job->ibs->grabbed_vmid = true;
- }
- mutex_unlock(&adev->vm_manager.lock);
fence = amdgpu_sync_get_fence(sync);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index e688fcf..38ab4a5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -159,13 +159,14 @@ void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev,
* @vm: vm to allocate id for
* @ring: ring we want to submit job to
* @sync: sync object where we add dependencies
+ * @fence: fence protecting ID from reuse
*
* Allocate an id for the vm, adding fences to the sync obj as necessary.
*
* Global mutex must be locked!
*/
int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
- struct amdgpu_sync *sync)
+ struct amdgpu_sync *sync, struct fence *fence)
{
struct fence *best[AMDGPU_MAX_RINGS] = {};
struct amdgpu_vm_id *vm_id = &vm->ids[ring->idx];
@@ -174,6 +175,8 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
unsigned choices[2] = {};
unsigned i;
+ mutex_lock(&adev->vm_manager.lock);
+
/* check if the id is still valid */
if (vm_id->id) {
unsigned id = vm_id->id;
@@ -182,6 +185,9 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
owner = atomic_long_read(&adev->vm_manager.ids[id].owner);
if (owner == (long)vm) {
trace_amdgpu_vm_grab_id(vm, vm_id->id, ring->idx);
+ fence_put(adev->vm_manager.ids[id].active);
+ adev->vm_manager.ids[id].active = fence_get(fence);
+ mutex_unlock(&adev->vm_manager.lock);
return 0;
}
}
@@ -198,6 +204,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
/* found a free one */
vm_id->id = i;
trace_amdgpu_vm_grab_id(vm, i, ring->idx);
+ mutex_unlock(&adev->vm_manager.lock);
return 0;
}
@@ -210,19 +217,29 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
}
for (i = 0; i < 2; ++i) {
- if (choices[i]) {
- struct fence *fence;
+ struct fence *active;
+ int r;
- fence = adev->vm_manager.ids[choices[i]].active;
- vm_id->id = choices[i];
+ if (!choices[i])
+ continue;
- trace_amdgpu_vm_grab_id(vm, choices[i], ring->idx);
- return amdgpu_sync_fence(ring->adev, sync, fence);
- }
+ vm_id->id = choices[i];
+ active = adev->vm_manager.ids[vm_id->id].active;
+ r = amdgpu_sync_fence(ring->adev, sync, active);
+
+ trace_amdgpu_vm_grab_id(vm, choices[i], ring->idx);
+ atomic_long_set(&adev->vm_manager.ids[vm_id->id].owner, (long)vm);
+
+ fence_put(adev->vm_manager.ids[vm_id->id].active);
+ adev->vm_manager.ids[vm_id->id].active = fence_get(fence);
+
+ mutex_unlock(&adev->vm_manager.lock);
+ return r;
}
/* should never happen */
BUG();
+ mutex_unlock(&adev->vm_manager.lock);
return -EINVAL;
}
@@ -265,30 +282,6 @@ void amdgpu_vm_flush(struct amdgpu_ring *ring,
}
/**
- * amdgpu_vm_fence - remember fence for vm
- *
- * @adev: amdgpu_device pointer
- * @vm: vm we want to fence
- * @fence: fence to remember
- *
- * Fence the vm (cayman+).
- * Set the fence used to protect page table and id.
- *
- * Global and local mutex must be locked!
- */
-void amdgpu_vm_fence(struct amdgpu_device *adev,
- struct amdgpu_vm *vm,
- struct fence *fence)
-{
- struct amdgpu_ring *ring = amdgpu_ring_from_fence(fence);
- unsigned vm_id = vm->ids[ring->idx].id;
-
- fence_put(adev->vm_manager.ids[vm_id].active);
- adev->vm_manager.ids[vm_id].active = fence_get(fence);
- atomic_long_set(&adev->vm_manager.ids[vm_id].owner, (long)vm);
-}
-
-/**
* amdgpu_vm_bo_find - find the bo_va for a specific vm & bo
*
* @vm: requested vm
--
2.7.4
|
