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
189
190
191
192
193
|
From a7a576dde42ff59b384fec823e21a7fd40aaa11c Mon Sep 17 00:00:00 2001
From: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Date: Tue, 24 Oct 2017 13:30:16 -0400
Subject: [PATCH 2067/4131] drm/amdgpu: Remove job->s_entity to avoid keeping
reference to stale pointer.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 7 ++++---
drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h | 9 ++++-----
drivers/gpu/drm/amd/scheduler/gpu_scheduler.c | 19 +++++++++----------
drivers/gpu/drm/amd/scheduler/gpu_scheduler.h | 7 ++++---
5 files changed, 22 insertions(+), 22 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index ed72bc9..0d32ada 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -1250,7 +1250,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
amdgpu_ring_priority_get(job->ring, job->base.s_priority);
trace_amdgpu_cs_ioctl(job);
- amd_sched_entity_push_job(&job->base);
+ amd_sched_entity_push_job(&job->base, entity);
ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence);
amdgpu_mn_unlock(p->mn);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index a58e3c5..f60662e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -142,12 +142,13 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
*f = dma_fence_get(&job->base.s_fence->finished);
amdgpu_job_free_resources(job);
amdgpu_ring_priority_get(job->ring, job->base.s_priority);
- amd_sched_entity_push_job(&job->base);
+ amd_sched_entity_push_job(&job->base, entity);
return 0;
}
-static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job)
+static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job,
+ struct amd_sched_entity *s_entity)
{
struct amdgpu_job *job = to_amdgpu_job(sched_job);
struct amdgpu_vm *vm = job->vm;
@@ -155,7 +156,7 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job)
struct dma_fence *fence = amdgpu_sync_get_fence(&job->dep_sync);
int r;
- if (amd_sched_dependency_optimized(fence, sched_job->s_entity)) {
+ if (amd_sched_dependency_optimized(fence, s_entity)) {
r = amdgpu_sync_fence(job->adev, &job->sched_sync, fence);
if (r)
DRM_ERROR("Error adding fence to sync (%d)\n", r);
diff --git a/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h b/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h
index 705380e..eebe323 100644
--- a/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h
+++ b/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h
@@ -13,8 +13,8 @@
#define TRACE_INCLUDE_FILE gpu_sched_trace
TRACE_EVENT(amd_sched_job,
- TP_PROTO(struct amd_sched_job *sched_job),
- TP_ARGS(sched_job),
+ TP_PROTO(struct amd_sched_job *sched_job, struct amd_sched_entity *entity),
+ TP_ARGS(sched_job, entity),
TP_STRUCT__entry(
__field(struct amd_sched_entity *, entity)
__field(struct dma_fence *, fence)
@@ -25,12 +25,11 @@ TRACE_EVENT(amd_sched_job,
),
TP_fast_assign(
- __entry->entity = sched_job->s_entity;
+ __entry->entity = entity;
__entry->id = sched_job->id;
__entry->fence = &sched_job->s_fence->finished;
__entry->name = sched_job->sched->name;
- __entry->job_count = spsc_queue_count(
- &sched_job->s_entity->job_queue);
+ __entry->job_count = spsc_queue_count(&entity->job_queue);
__entry->hw_job_count = atomic_read(
&sched_job->sched->hw_rq_count);
),
diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
index 9cbeade..903ef8b 100644
--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
+++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
@@ -340,11 +340,10 @@ amd_sched_entity_pop_job(struct amd_sched_entity *entity)
if (!sched_job)
return NULL;
- while ((entity->dependency = sched->ops->dependency(sched_job)))
+ while ((entity->dependency = sched->ops->dependency(sched_job, entity)))
if (amd_sched_entity_add_dependency_cb(entity))
return NULL;
- sched_job->s_entity = NULL;
spsc_queue_pop(&entity->job_queue);
return sched_job;
}
@@ -356,13 +355,13 @@ amd_sched_entity_pop_job(struct amd_sched_entity *entity)
*
* Returns 0 for success, negative error code otherwise.
*/
-void amd_sched_entity_push_job(struct amd_sched_job *sched_job)
+void amd_sched_entity_push_job(struct amd_sched_job *sched_job,
+ struct amd_sched_entity *entity)
{
struct amd_gpu_scheduler *sched = sched_job->sched;
- struct amd_sched_entity *entity = sched_job->s_entity;
bool first = false;
- trace_amd_sched_job(sched_job);
+ trace_amd_sched_job(sched_job, entity);
spin_lock(&entity->queue_lock);
first = spsc_queue_push(&entity->job_queue, &sched_job->queue_node);
@@ -441,11 +440,12 @@ static void amd_sched_job_timedout(struct work_struct *work)
job->sched->ops->timedout_job(job);
}
-static void amd_sched_set_guilty(struct amd_sched_job *s_job)
+static void amd_sched_set_guilty(struct amd_sched_job *s_job,
+ struct amd_sched_entity *s_entity)
{
if (atomic_inc_return(&s_job->karma) > s_job->sched->hang_limit)
- if (s_job->s_entity->guilty)
- atomic_set(s_job->s_entity->guilty, 1);
+ if (s_entity->guilty)
+ atomic_set(s_entity->guilty, 1);
}
void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched, struct amd_sched_job *bad)
@@ -476,7 +476,7 @@ void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched, struct amd_sched_jo
list_for_each_entry_safe(entity, tmp, &rq->entities, list) {
if (bad->s_fence->scheduled.context == entity->fence_context) {
found = true;
- amd_sched_set_guilty(bad);
+ amd_sched_set_guilty(bad, entity);
break;
}
}
@@ -540,7 +540,6 @@ int amd_sched_job_init(struct amd_sched_job *job,
void *owner)
{
job->sched = sched;
- job->s_entity = entity;
job->s_priority = entity->rq - sched->sched_rq;
job->s_fence = amd_sched_fence_create(entity, owner);
if (!job->s_fence)
diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
index 4fbbbc8..13c8e87 100644
--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
+++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
@@ -90,7 +90,6 @@ struct amd_sched_fence {
struct amd_sched_job {
struct spsc_node queue_node;
struct amd_gpu_scheduler *sched;
- struct amd_sched_entity *s_entity;
struct amd_sched_fence *s_fence;
struct dma_fence_cb finish_cb;
struct work_struct finish_work;
@@ -124,7 +123,8 @@ static inline bool amd_sched_invalidate_job(struct amd_sched_job *s_job, int thr
* these functions should be implemented in driver side
*/
struct amd_sched_backend_ops {
- struct dma_fence *(*dependency)(struct amd_sched_job *sched_job);
+ struct dma_fence *(*dependency)(struct amd_sched_job *sched_job,
+ struct amd_sched_entity *s_entity);
struct dma_fence *(*run_job)(struct amd_sched_job *sched_job);
void (*timedout_job)(struct amd_sched_job *sched_job);
void (*free_job)(struct amd_sched_job *sched_job);
@@ -160,7 +160,8 @@ int amd_sched_entity_init(struct amd_gpu_scheduler *sched,
uint32_t jobs, atomic_t* guilty);
void amd_sched_entity_fini(struct amd_gpu_scheduler *sched,
struct amd_sched_entity *entity);
-void amd_sched_entity_push_job(struct amd_sched_job *sched_job);
+void amd_sched_entity_push_job(struct amd_sched_job *sched_job,
+ struct amd_sched_entity *entity);
void amd_sched_entity_set_rq(struct amd_sched_entity *entity,
struct amd_sched_rq *rq);
--
2.7.4
|
