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
208
209
210
211
212
213
214
215
216
217
218
219
220
221
|
From 943206ea979d3e38024df99792be02ae2f285473 Mon Sep 17 00:00:00 2001
From: Sharat Masetty <smasetty@codeaurora.org>
Date: Thu, 29 Nov 2018 15:35:20 +0530
Subject: [PATCH 1131/2940] drm/scheduler: Add
drm_sched_suspend/resume_timeout()
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This patch adds two new functions to help client drivers suspend and
resume the scheduler job timeout. This can be useful in cases where the
hardware has preemption support enabled. Using this, it is possible to have
the timeout active only for the ring which is active on the ringbuffer.
This patch also makes the job_list_lock IRQ safe.
Suggested-by: Christian Koenig <Christian.Koenig@amd.com>
Signed-off-by: Sharat Masetty <smasetty@codeaurora.org>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Chaudhary Amit Kumar <Chaudharyamit.Kumar@amd.com>
---
drivers/gpu/drm/scheduler/sched_main.c | 83 ++++++++++++++++++++++----
include/drm/gpu_scheduler.h | 4 ++
2 files changed, 77 insertions(+), 10 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index 0f9d473d572c..fef0282eb398 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -209,6 +209,62 @@ static void drm_sched_start_timeout(struct drm_gpu_scheduler *sched)
schedule_delayed_work(&sched->work_tdr, sched->timeout);
}
+/**
+ * drm_sched_suspend_timeout - Suspend scheduler job timeout
+ *
+ * @sched: scheduler instance for which to suspend the timeout
+ *
+ * Suspend the delayed work timeout for the scheduler. This is done by
+ * modifying the delayed work timeout to an arbitrary large value,
+ * MAX_SCHEDULE_TIMEOUT in this case. Note that this function can be
+ * called from an IRQ context.
+ *
+ * Returns the timeout remaining
+ *
+ */
+unsigned long drm_sched_suspend_timeout(struct drm_gpu_scheduler *sched)
+{
+ unsigned long sched_timeout, now = jiffies;
+
+ sched_timeout = sched->work_tdr.timer.expires;
+
+ /*
+ * Modify the timeout to an arbitrarily large value. This also prevents
+ * the timeout to be restarted when new submissions arrive
+ */
+ if (mod_delayed_work(system_wq, &sched->work_tdr, MAX_SCHEDULE_TIMEOUT)
+ && time_after(sched_timeout, now))
+ return sched_timeout - now;
+ else
+ return sched->timeout;
+}
+EXPORT_SYMBOL(drm_sched_suspend_timeout);
+
+/**
+ * drm_sched_resume_timeout - Resume scheduler job timeout
+ *
+ * @sched: scheduler instance for which to resume the timeout
+ * @remaining: remaining timeout
+ *
+ * Resume the delayed work timeout for the scheduler. Note that
+ * this function can be called from an IRQ context.
+ */
+void drm_sched_resume_timeout(struct drm_gpu_scheduler *sched,
+ unsigned long remaining)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&sched->job_list_lock, flags);
+
+ if (list_empty(&sched->ring_mirror_list))
+ cancel_delayed_work(&sched->work_tdr);
+ else
+ mod_delayed_work(system_wq, &sched->work_tdr, remaining);
+
+ spin_unlock_irqrestore(&sched->job_list_lock, flags);
+}
+EXPORT_SYMBOL(drm_sched_resume_timeout);
+
/* job_finish is called after hw fence signaled
*/
static void drm_sched_job_finish(struct work_struct *work)
@@ -216,6 +272,7 @@ static void drm_sched_job_finish(struct work_struct *work)
struct drm_sched_job *s_job = container_of(work, struct drm_sched_job,
finish_work);
struct drm_gpu_scheduler *sched = s_job->sched;
+ unsigned long flags;
/*
* Canceling the timeout without removing our job from the ring mirror
@@ -226,12 +283,12 @@ static void drm_sched_job_finish(struct work_struct *work)
*/
cancel_delayed_work_sync(&sched->work_tdr);
- spin_lock(&sched->job_list_lock);
+ spin_lock_irqsave(&sched->job_list_lock, flags);
/* remove job from ring_mirror_list */
list_del(&s_job->node);
/* queue TDR for next job */
drm_sched_start_timeout(sched);
- spin_unlock(&sched->job_list_lock);
+ spin_unlock_irqrestore(&sched->job_list_lock, flags);
sched->ops->free_job(s_job);
}
@@ -247,20 +304,22 @@ static void drm_sched_job_finish_cb(struct dma_fence *f,
static void drm_sched_job_begin(struct drm_sched_job *s_job)
{
struct drm_gpu_scheduler *sched = s_job->sched;
+ unsigned long flags;
dma_fence_add_callback(&s_job->s_fence->finished, &s_job->finish_cb,
drm_sched_job_finish_cb);
- spin_lock(&sched->job_list_lock);
+ spin_lock_irqsave(&sched->job_list_lock, flags);
list_add_tail(&s_job->node, &sched->ring_mirror_list);
drm_sched_start_timeout(sched);
- spin_unlock(&sched->job_list_lock);
+ spin_unlock_irqrestore(&sched->job_list_lock, flags);
}
static void drm_sched_job_timedout(struct work_struct *work)
{
struct drm_gpu_scheduler *sched;
struct drm_sched_job *job;
+ unsigned long flags;
sched = container_of(work, struct drm_gpu_scheduler, work_tdr.work);
job = list_first_entry_or_null(&sched->ring_mirror_list,
@@ -269,7 +328,9 @@ static void drm_sched_job_timedout(struct work_struct *work)
if (job)
job->sched->ops->timedout_job(job);
+ spin_lock_irqsave(&sched->job_list_lock, flags);
drm_sched_start_timeout(sched);
+ spin_unlock_irqrestore(&sched->job_list_lock, flags);
}
/**
@@ -283,9 +344,10 @@ void drm_sched_hw_job_reset(struct drm_gpu_scheduler *sched, struct drm_sched_jo
{
struct drm_sched_job *s_job;
struct drm_sched_entity *entity, *tmp;
+ unsigned long flags;
int i;
- spin_lock(&sched->job_list_lock);
+ spin_lock_irqsave(&sched->job_list_lock, flags);
list_for_each_entry_reverse(s_job, &sched->ring_mirror_list, node) {
if (s_job->s_fence->parent &&
dma_fence_remove_callback(s_job->s_fence->parent,
@@ -295,7 +357,7 @@ void drm_sched_hw_job_reset(struct drm_gpu_scheduler *sched, struct drm_sched_jo
atomic_dec(&sched->hw_rq_count);
}
}
- spin_unlock(&sched->job_list_lock);
+ spin_unlock_irqrestore(&sched->job_list_lock, flags);
if (bad && bad->s_priority != DRM_SCHED_PRIORITY_KERNEL) {
atomic_inc(&bad->karma);
@@ -333,9 +395,10 @@ void drm_sched_job_recovery(struct drm_gpu_scheduler *sched)
{
struct drm_sched_job *s_job, *tmp;
bool found_guilty = false;
+ unsigned long flags;
int r;
- spin_lock(&sched->job_list_lock);
+ spin_lock_irqsave(&sched->job_list_lock, flags);
list_for_each_entry_safe(s_job, tmp, &sched->ring_mirror_list, node) {
struct drm_sched_fence *s_fence = s_job->s_fence;
struct dma_fence *fence;
@@ -349,7 +412,7 @@ void drm_sched_job_recovery(struct drm_gpu_scheduler *sched)
if (found_guilty && s_job->s_fence->scheduled.context == guilty_context)
dma_fence_set_error(&s_fence->finished, -ECANCELED);
- spin_unlock(&sched->job_list_lock);
+ spin_unlock_irqrestore(&sched->job_list_lock, flags);
fence = sched->ops->run_job(s_job);
atomic_inc(&sched->hw_rq_count);
@@ -366,10 +429,10 @@ void drm_sched_job_recovery(struct drm_gpu_scheduler *sched)
} else {
drm_sched_process_job(NULL, &s_fence->cb);
}
- spin_lock(&sched->job_list_lock);
+ spin_lock_irqsave(&sched->job_list_lock, flags);
}
drm_sched_start_timeout(sched);
- spin_unlock(&sched->job_list_lock);
+ spin_unlock_irqrestore(&sched->job_list_lock, flags);
}
EXPORT_SYMBOL(drm_sched_job_recovery);
diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
index dfc0baa795d8..046620acd6de 100644
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -344,4 +344,8 @@ bool drm_sched_dependency_optimized(struct dma_fence* fence,
struct drm_sched_entity *entity);
void drm_sched_job_kickout(struct drm_sched_job *s_job);
+unsigned long drm_sched_suspend_timeout(struct drm_gpu_scheduler *sched);
+void drm_sched_resume_timeout(struct drm_gpu_scheduler *sched,
+ unsigned long remaining);
+
#endif
--
2.17.1
|
