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
|
From bacbde41338ca493a4557fb33faf4ea4ab788bb5 Mon Sep 17 00:00:00 2001
From: Morten Rasmussen <morten.rasmussen@arm.com>
Date: Wed, 22 Jun 2016 14:51:17 +0100
Subject: [PATCH 44/92] arm64: Set SD_ASYM_CPUCAPACITY topology flag for
asymmetric cpu capacity systems
In addition to adjusting the cpu capacities asymmetric cpu capacity
systems also have to set the SD_ASYM_CPUCAPACITY topology flag to inform
the scheduler to look harder when load balancing. This patch sets the
topology flag on DIE level for such systems based on the cpu capacities
provided through DT or sysfs.
cc: Catalin Marinas <catalin.marinas@arm.com>
cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Morten Rasmussen <morten.rasmussen@arm.com>
(cherry picked from commit 8a8992543ec08e26f8a8ab3cdb4bb556283ce9ea)
Signed-off-by: Gaku Inami <gaku.inami.xw@bp.renesas.com>
---
arch/arm64/kernel/topology.c | 52 ++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 52 insertions(+)
diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c
index 280319b..eb2ad75 100644
--- a/arch/arm64/kernel/topology.c
+++ b/arch/arm64/kernel/topology.c
@@ -21,12 +21,15 @@
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/cpufreq.h>
+#include <linux/cpuset.h>
#include <asm/cputype.h>
#include <asm/topology.h>
static DEFINE_PER_CPU(unsigned long, cpu_scale) = SCHED_CAPACITY_SCALE;
static DEFINE_MUTEX(cpu_scale_mutex);
+static bool asym_cpucap;
+static bool update_flags;
unsigned long scale_cpu_capacity(struct sched_domain *sd, int cpu)
{
@@ -38,6 +41,14 @@ static void set_capacity_scale(unsigned int cpu, unsigned long capacity)
per_cpu(cpu_scale, cpu) = capacity;
}
+static void update_sched_flags(void)
+{
+ update_flags = true;
+ rebuild_sched_domains();
+ update_flags = false;
+ pr_debug("cpu_capacity: Rebuilt sched_domain hierarchy.\n");
+}
+
#ifdef CONFIG_PROC_SYSCTL
#include <asm/cpu.h>
#include <linux/string.h>
@@ -67,6 +78,7 @@ static ssize_t store_cpu_capacity(struct device *dev,
if (count) {
char *p = (char *) buf;
+ bool asym = false;
ret = kstrtoul(p, 0, &new_capacity);
if (ret)
@@ -77,6 +89,16 @@ static ssize_t store_cpu_capacity(struct device *dev,
mutex_lock(&cpu_scale_mutex);
for_each_cpu(i, &cpu_topology[this_cpu].core_sibling)
set_capacity_scale(i, new_capacity);
+
+ for_each_possible_cpu(i) {
+ if (per_cpu(cpu_scale, i) != new_capacity)
+ asym = true;
+ }
+
+ if (asym != asym_cpucap) {
+ asym_cpucap = asym;
+ update_sched_flags();
+ }
mutex_unlock(&cpu_scale_mutex);
}
@@ -153,6 +175,7 @@ static void normalize_cpu_capacity(void)
{
u64 capacity;
int cpu;
+ bool asym = false;
if (!raw_capacity || cap_parsing_failed)
return;
@@ -167,8 +190,13 @@ static void normalize_cpu_capacity(void)
set_capacity_scale(cpu, capacity);
pr_debug("cpu_capacity: CPU%d cpu_capacity=%lu\n",
cpu, arch_scale_cpu_capacity(NULL, cpu));
+ if (capacity < capacity_scale)
+ asym = true;
}
mutex_unlock(&cpu_scale_mutex);
+
+ if (asym != asym_cpucap)
+ asym_cpucap = asym;
}
#ifdef CONFIG_CPU_FREQ
@@ -186,6 +214,7 @@ static void normalize_cpu_capacity(void)
{
struct cpufreq_policy *policy = data;
int cpu;
+ bool asym;
if (cap_parsing_done)
return 0;
@@ -208,7 +237,10 @@ static void normalize_cpu_capacity(void)
}
if (cpumask_empty(cpus_to_visit)) {
if (!cap_parsing_failed) {
+ asym = asym_cpucap;
normalize_cpu_capacity();
+ if (asym != asym_cpucap)
+ update_sched_flags();
kfree(raw_capacity);
pr_debug("cpu_capacity: parsing done\n");
} else {
@@ -488,6 +520,11 @@ const struct cpumask *cpu_coregroup_mask(int cpu)
return &cpu_topology[cpu].core_sibling;
}
+static int cpu_cpu_flags(void)
+{
+ return asym_cpucap ? SD_ASYM_CPUCAPACITY : 0;
+}
+
static void update_siblings_masks(unsigned int cpuid)
{
struct cpu_topology *cpu_topo, *cpuid_topo = &cpu_topology[cpuid];
@@ -569,6 +606,19 @@ static void __init reset_cpu_topology(void)
}
}
+static struct sched_domain_topology_level arm64_topology[] = {
+#ifdef CONFIG_SCHED_MC
+ { cpu_coregroup_mask, cpu_core_flags, SD_INIT_NAME(MC) },
+#endif
+ { cpu_cpu_mask, cpu_cpu_flags, SD_INIT_NAME(DIE) },
+ { NULL, }
+};
+
+int arch_update_cpu_topology(void)
+{
+ return update_flags ? 1 : 0;
+}
+
void __init init_cpu_topology(void)
{
reset_cpu_topology();
@@ -579,4 +629,6 @@ void __init init_cpu_topology(void)
*/
if (of_have_populated_dt() && parse_dt_topology())
reset_cpu_topology();
+ else
+ set_sched_topology(arm64_topology);
}
--
1.9.1
|
