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
|
From 2a34d5c262d7989b796dd9e0da1a3816a9a00732 Mon Sep 17 00:00:00 2001
From: hamadmarri <hamad.s.almarri@gmail.com>
Date: Sat, 23 Mar 2024 01:47:51 +0300
Subject: [PATCH 6/8] make quota for preempted task preserved so it doesn't
start over
---
include/linux/sched.h | 5 ++---
kernel/sched/bs.c | 34 ++++++++++++++++++++++------------
kernel/sched/core.c | 1 -
kernel/sched/debug.c | 1 -
4 files changed, 24 insertions(+), 17 deletions(-)
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 62e5be998..a68f67356 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -550,14 +550,13 @@ struct sched_statistics {
struct bs_node {
struct bs_node* next;
- u64 vruntime_start;
+ u64 c_vrt_start;
+ u64 r_vrt_start;
u64 vburst;
#ifdef CONFIG_SCHED_DEBUG
u64 prev_vburst;
#endif
u64 est;
- u32 alpha;
-
};
struct sched_entity {
diff --git a/kernel/sched/bs.c b/kernel/sched/bs.c
index 6fe2ba6e3..ca12d3aa5 100644
--- a/kernel/sched/bs.c
+++ b/kernel/sched/bs.c
@@ -16,6 +16,7 @@
unsigned int sysctl_sched_base_slice = 7000ULL;
unsigned int bs_shared_quota = 6600000ULL; // 6.6ms
+u32 alpha = 500U;
struct lb_env {
struct rq *src_rq;
@@ -39,7 +40,7 @@ struct global_candidate {
raw_spinlock_t lock;
};
-#define MAX_EST 0xFFFFFFFFFFFFFFFULL
+#define MAX_EST 0xFFFFFFFFFFFFFFFULL
struct global_candidate global_candidate = {0, 0, MAX_EST};
@@ -161,8 +162,10 @@ static void update_curr(struct cfs_rq *cfs_rq)
schedstat_add(cfs_rq->exec_clock, delta_exec);
calc = calc_delta_fair(delta_exec, curr);
- curr->vruntime += calc;
- curr->bs_node.vburst += calc;
+ curr->vruntime += calc;
+ curr->bs_node.vburst += calc;
+ curr->bs_node.c_vrt_start += calc;
+ curr->bs_node.r_vrt_start += calc;
#ifdef CONFIG_SCHED_DEBUG
curr->bs_node.prev_vburst = curr->bs_node.vburst;
#endif
@@ -184,6 +187,14 @@ static void update_curr_fair(struct rq *rq)
update_curr(cfs_rq_of(&rq->curr->se));
}
+/**
+ * Should `a` preempts `b`?
+ */
+static inline bool entity_before(struct bs_node *a, struct bs_node *b)
+{
+ return (s64)(a->est - b->est) < 0;
+}
+
static void __enqueue_entity(struct bs_node **q, struct bs_node *bsn)
{
struct bs_node *prev;
@@ -274,7 +285,6 @@ update_est_entity(struct sched_entity *se)
struct bs_node *bsn = &se->bs_node;
u64 vburst = bsn->vburst;
u64 prev_est = bsn->est;
- u32 alpha = bsn->alpha;
u64 next_est;
// alpha = update_alpha(bsn, vburst, prev_est, alpha);
@@ -359,6 +369,8 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
cfs_rq->idle_h_nr_running += idle_h_nr_running;
}
+ se->bs_node.r_vrt_start = 0;
+
update_candidate(cfs_rq);
add_nr_running(rq, 1);
@@ -431,12 +443,10 @@ int __entity_end_quota(struct cfs_rq *cfs_rq, struct sched_entity *curr)
unsigned int n = max(cfs_rq->nr_running, 1);
unsigned int quota;
struct bs_node *bs = &curr->bs_node;
- s64 vburst;
quota = max(bs_shared_quota / n, sysctl_sched_base_slice);
- vburst = curr->vruntime - bs->vruntime_start;
- return vburst >= (s64)quota;
+ return (s64)(bs->r_vrt_start - (u64)quota) >= 0;
}
static int entity_end_quota(struct cfs_rq *cfs_rq, struct sched_entity *curr)
@@ -452,11 +462,8 @@ static int entity_end_quota(struct cfs_rq *cfs_rq, struct sched_entity *curr)
static int entity_end_min_slice(struct sched_entity *curr)
{
struct bs_node *bs = &curr->bs_node;
- s64 vburst;
-
- vburst = curr->vruntime - bs->vruntime_start;
- return (vburst - (s64)sysctl_sched_base_slice) >= 0;
+ return (s64)(bs->c_vrt_start - (u64)sysctl_sched_base_slice) >= 0;
}
static void check_preempt_wakeup_fair(struct rq *rq, struct task_struct *p, int wake_flags)
@@ -510,7 +517,8 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
se->exec_start = rq_clock_task(rq_of(cfs_rq));
- se->bs_node.vruntime_start = se->vruntime;
+ se->bs_node.c_vrt_start = 0;
+
update_candidate(cfs_rq);
cfs_rq->local_cand_est = se->bs_node.est;
@@ -665,6 +673,8 @@ static void __enqueue_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *se
{
if (se->yielded || entity_end_quota(cfs_rq, se)) {
se->yielded = false;
+ se->bs_node.r_vrt_start = 0;
+
__enqueue_entity(&cfs_rq->q2_head, &se->bs_node);
} else {
__enqueue_entity(&cfs_rq->head, &se->bs_node);
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index e9ad2a7f5..f99b51020 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4517,7 +4517,6 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
p->se.bs_node.vburst = 0;
p->se.bs_node.est = (u64)bs_shared_quota;
- p->se.bs_node.alpha = 500U;
INIT_LIST_HEAD(&p->se.group_node);
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 2d8660b48..6ebd68faa 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -1001,7 +1001,6 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
PN(se.bs_node.vburst);
PN(se.bs_node.prev_vburst);
PN(se.bs_node.est);
- PN(se.bs_node.alpha);
nr_switches = p->nvcsw + p->nivcsw;
--
2.44.0
|