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
194
195
196
197
198
199
200
201
202
203
204
205
206
207
|
From 03d207575c7c3c6dfe7bf00fcb9c2129a562d009 Mon Sep 17 00:00:00 2001
From: Monk Liu <Monk.Liu@amd.com>
Date: Fri, 4 Mar 2016 18:51:02 +0800
Subject: [PATCH 0450/1110] drm/amdgpu: rework TDR in scheduler (v2)
Add two callbacks to scheduler to maintain jobs, and invoked for
job timeout calculations. Now TDR measures time gap from
job is processed by hw.
v2:
fix typo
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@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 | 1 +
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 1 +
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 16 +++++++++++-
drivers/gpu/drm/amd/scheduler/gpu_scheduler.c | 37 +++++++++++++++++++++++++++
drivers/gpu/drm/amd/scheduler/gpu_scheduler.h | 7 +++++
drivers/gpu/drm/amd/scheduler/sched_fence.c | 1 +
6 files changed, 62 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 890844b..2474405 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -753,6 +753,7 @@ void amdgpu_job_free(struct amdgpu_job *job);
int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
struct amd_sched_entity *entity, void *owner,
struct fence **f);
+void amdgpu_job_timeout_func(struct work_struct *work);
struct amdgpu_ring {
struct amdgpu_device *adev;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index ed6d8b3..df923cf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -923,6 +923,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
r = amd_sched_job_init(&job->base, &ring->sched,
&p->ctx->rings[ring->idx].entity,
+ amdgpu_job_timeout_func,
p->filp, &fence);
if (r) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index e593ed2..d00335a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -34,6 +34,15 @@ static void amdgpu_job_free_handler(struct work_struct *ws)
kfree(job);
}
+void amdgpu_job_timeout_func(struct work_struct *work)
+{
+ struct amdgpu_job *job = container_of(work, struct amdgpu_job, base.work_tdr.work);
+ DRM_ERROR("ring %s timeout, last signaled seq=%u, last emitted seq=%u\n",
+ job->base.sched->name,
+ (uint32_t)atomic_read(&job->ring->fence_drv.last_seq),
+ job->ring->fence_drv.sync_seq);
+}
+
int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs,
struct amdgpu_job **job)
{
@@ -102,7 +111,10 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
if (!f)
return -EINVAL;
- r = amd_sched_job_init(&job->base, &ring->sched, entity, owner, &fence);
+ r = amd_sched_job_init(&job->base, &ring->sched,
+ entity, owner,
+ amdgpu_job_timeout_func,
+ &fence);
if (r)
return r;
@@ -179,6 +191,8 @@ err:
struct amd_sched_backend_ops amdgpu_sched_ops = {
.dependency = amdgpu_job_dependency,
.run_job = amdgpu_job_run,
+ .begin_job = amd_sched_job_begin,
+ .finish_job = amd_sched_job_finish,
};
diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
index 9a9fffd..b7e8071 100644
--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
+++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
@@ -324,6 +324,40 @@ static void amd_sched_free_job(struct fence *f, struct fence_cb *cb) {
schedule_work(&job->work_free_job);
}
+/* job_finish is called after hw fence signaled, and
+ * the job had already been deleted from ring_mirror_list
+ */
+void amd_sched_job_finish(struct amd_sched_job *s_job)
+{
+ struct amd_sched_job *next;
+ struct amd_gpu_scheduler *sched = s_job->sched;
+
+ if (sched->timeout != MAX_SCHEDULE_TIMEOUT) {
+ cancel_delayed_work(&s_job->work_tdr); /*TODO: how to deal the case that tdr is running */
+
+ /* queue TDR for next job */
+ next = list_first_entry_or_null(&sched->ring_mirror_list,
+ struct amd_sched_job, node);
+
+ if (next) {
+ INIT_DELAYED_WORK(&next->work_tdr, s_job->timeout_callback);
+ schedule_delayed_work(&next->work_tdr, sched->timeout);
+ }
+ }
+}
+
+void amd_sched_job_begin(struct amd_sched_job *s_job)
+{
+ struct amd_gpu_scheduler *sched = s_job->sched;
+
+ if (sched->timeout != MAX_SCHEDULE_TIMEOUT &&
+ list_first_entry_or_null(&sched->ring_mirror_list, struct amd_sched_job, node) == s_job)
+ {
+ INIT_DELAYED_WORK(&s_job->work_tdr, s_job->timeout_callback);
+ schedule_delayed_work(&s_job->work_tdr, sched->timeout);
+ }
+}
+
/**
* Submit a job to the job queue
*
@@ -347,6 +381,7 @@ void amd_sched_entity_push_job(struct amd_sched_job *sched_job)
int amd_sched_job_init(struct amd_sched_job *job,
struct amd_gpu_scheduler *sched,
struct amd_sched_entity *entity,
+ void (*timeout_cb)(struct work_struct *work),
void *owner, struct fence **fence)
{
INIT_LIST_HEAD(&job->node);
@@ -357,6 +392,7 @@ int amd_sched_job_init(struct amd_sched_job *job,
return -ENOMEM;
job->s_fence->s_job = job;
+ job->timeout_callback = timeout_cb;
if (fence)
*fence = &job->s_fence->base;
@@ -415,6 +451,7 @@ static void amd_sched_process_job(struct fence *f, struct fence_cb *cb)
/* remove job from ring_mirror_list */
spin_lock_irqsave(&sched->job_list_lock, flags);
list_del_init(&s_fence->s_job->node);
+ sched->ops->finish_job(s_fence->s_job);
spin_unlock_irqrestore(&sched->job_list_lock, flags);
amd_sched_fence_signal(s_fence);
diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
index b26148d..a5700ad 100644
--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
+++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
@@ -85,6 +85,8 @@ struct amd_sched_job {
struct fence_cb cb_free_job;
struct work_struct work_free_job;
struct list_head node;
+ struct delayed_work work_tdr;
+ void (*timeout_callback) (struct work_struct *work);
};
extern const struct fence_ops amd_sched_fence_ops;
@@ -105,6 +107,8 @@ static inline struct amd_sched_fence *to_amd_sched_fence(struct fence *f)
struct amd_sched_backend_ops {
struct fence *(*dependency)(struct amd_sched_job *sched_job);
struct fence *(*run_job)(struct amd_sched_job *sched_job);
+ void (*begin_job)(struct amd_sched_job *sched_job);
+ void (*finish_job)(struct amd_sched_job *sched_job);
};
enum amd_sched_priority {
@@ -150,7 +154,10 @@ void amd_sched_fence_signal(struct amd_sched_fence *fence);
int amd_sched_job_init(struct amd_sched_job *job,
struct amd_gpu_scheduler *sched,
struct amd_sched_entity *entity,
+ void (*timeout_cb)(struct work_struct *work),
void *owner, struct fence **fence);
void amd_sched_job_pre_schedule(struct amd_gpu_scheduler *sched ,
struct amd_sched_job *s_job);
+void amd_sched_job_finish(struct amd_sched_job *s_job);
+void amd_sched_job_begin(struct amd_sched_job *s_job);
#endif
diff --git a/drivers/gpu/drm/amd/scheduler/sched_fence.c b/drivers/gpu/drm/amd/scheduler/sched_fence.c
index 33ddd38..2a732c4 100644
--- a/drivers/gpu/drm/amd/scheduler/sched_fence.c
+++ b/drivers/gpu/drm/amd/scheduler/sched_fence.c
@@ -63,6 +63,7 @@ void amd_sched_job_pre_schedule(struct amd_gpu_scheduler *sched ,
unsigned long flags;
spin_lock_irqsave(&sched->job_list_lock, flags);
list_add_tail(&s_job->node, &sched->ring_mirror_list);
+ sched->ops->begin_job(s_job);
spin_unlock_irqrestore(&sched->job_list_lock, flags);
}
--
2.7.4
|
