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
|
// SPDX-License-Identifier: GPL-2.0
/* Copyright (c) 2025 Meta Platforms, Inc. and affiliates. */
#include <linux/sched.h>
#include <linux/smp.h>
#include <linux/delay.h>
#include <linux/module.h>
#include <linux/prandom.h>
#include <asm/rqspinlock.h>
#include <linux/perf_event.h>
#include <linux/kthread.h>
#include <linux/atomic.h>
#include <linux/slab.h>
static struct perf_event_attr hw_attr = {
.type = PERF_TYPE_HARDWARE,
.config = PERF_COUNT_HW_CPU_CYCLES,
.size = sizeof(struct perf_event_attr),
.pinned = 1,
.disabled = 1,
.sample_period = 100000,
};
static rqspinlock_t lock_a;
static rqspinlock_t lock_b;
static struct perf_event **rqsl_evts;
static int rqsl_nevts;
static bool test_ab = false;
module_param(test_ab, bool, 0644);
MODULE_PARM_DESC(test_ab, "Test ABBA situations instead of AA situations");
static struct task_struct **rqsl_threads;
static int rqsl_nthreads;
static atomic_t rqsl_ready_cpus = ATOMIC_INIT(0);
static int pause = 0;
static bool nmi_locks_a(int cpu)
{
return (cpu & 1) && test_ab;
}
static int rqspinlock_worker_fn(void *arg)
{
int cpu = smp_processor_id();
unsigned long flags;
int ret;
if (cpu) {
atomic_inc(&rqsl_ready_cpus);
while (!kthread_should_stop()) {
if (READ_ONCE(pause)) {
msleep(1000);
continue;
}
if (nmi_locks_a(cpu))
ret = raw_res_spin_lock_irqsave(&lock_b, flags);
else
ret = raw_res_spin_lock_irqsave(&lock_a, flags);
mdelay(20);
if (nmi_locks_a(cpu) && !ret)
raw_res_spin_unlock_irqrestore(&lock_b, flags);
else if (!ret)
raw_res_spin_unlock_irqrestore(&lock_a, flags);
cpu_relax();
}
return 0;
}
while (!kthread_should_stop()) {
int expected = rqsl_nthreads > 0 ? rqsl_nthreads - 1 : 0;
int ready = atomic_read(&rqsl_ready_cpus);
if (ready == expected && !READ_ONCE(pause)) {
for (int i = 0; i < rqsl_nevts; i++)
perf_event_enable(rqsl_evts[i]);
pr_err("Waiting 5 secs to pause the test\n");
msleep(1000 * 5);
WRITE_ONCE(pause, 1);
pr_err("Paused the test\n");
} else {
msleep(1000);
cpu_relax();
}
}
return 0;
}
static void nmi_cb(struct perf_event *event, struct perf_sample_data *data,
struct pt_regs *regs)
{
int cpu = smp_processor_id();
unsigned long flags;
int ret;
if (!cpu || READ_ONCE(pause))
return;
if (nmi_locks_a(cpu))
ret = raw_res_spin_lock_irqsave(&lock_a, flags);
else
ret = raw_res_spin_lock_irqsave(test_ab ? &lock_b : &lock_a, flags);
mdelay(10);
if (nmi_locks_a(cpu) && !ret)
raw_res_spin_unlock_irqrestore(&lock_a, flags);
else if (!ret)
raw_res_spin_unlock_irqrestore(test_ab ? &lock_b : &lock_a, flags);
}
static void free_rqsl_threads(void)
{
int i;
if (rqsl_threads) {
for_each_online_cpu(i) {
if (rqsl_threads[i])
kthread_stop(rqsl_threads[i]);
}
kfree(rqsl_threads);
}
}
static void free_rqsl_evts(void)
{
int i;
if (rqsl_evts) {
for (i = 0; i < rqsl_nevts; i++) {
if (rqsl_evts[i])
perf_event_release_kernel(rqsl_evts[i]);
}
kfree(rqsl_evts);
}
}
static int bpf_test_rqspinlock_init(void)
{
int i, ret;
int ncpus = num_online_cpus();
pr_err("Mode = %s\n", test_ab ? "ABBA" : "AA");
if (ncpus < 3)
return -ENOTSUPP;
raw_res_spin_lock_init(&lock_a);
raw_res_spin_lock_init(&lock_b);
rqsl_evts = kcalloc(ncpus - 1, sizeof(*rqsl_evts), GFP_KERNEL);
if (!rqsl_evts)
return -ENOMEM;
rqsl_nevts = ncpus - 1;
for (i = 1; i < ncpus; i++) {
struct perf_event *e;
e = perf_event_create_kernel_counter(&hw_attr, i, NULL, nmi_cb, NULL);
if (IS_ERR(e)) {
ret = PTR_ERR(e);
goto err_perf_events;
}
rqsl_evts[i - 1] = e;
}
rqsl_threads = kcalloc(ncpus, sizeof(*rqsl_threads), GFP_KERNEL);
if (!rqsl_threads) {
ret = -ENOMEM;
goto err_perf_events;
}
rqsl_nthreads = ncpus;
for_each_online_cpu(i) {
struct task_struct *t;
t = kthread_create(rqspinlock_worker_fn, NULL, "rqsl_w/%d", i);
if (IS_ERR(t)) {
ret = PTR_ERR(t);
goto err_threads_create;
}
kthread_bind(t, i);
rqsl_threads[i] = t;
wake_up_process(t);
}
return 0;
err_threads_create:
free_rqsl_threads();
err_perf_events:
free_rqsl_evts();
return ret;
}
module_init(bpf_test_rqspinlock_init);
static void bpf_test_rqspinlock_exit(void)
{
free_rqsl_threads();
free_rqsl_evts();
}
module_exit(bpf_test_rqspinlock_exit);
MODULE_AUTHOR("Kumar Kartikeya Dwivedi");
MODULE_DESCRIPTION("BPF rqspinlock stress test module");
MODULE_LICENSE("GPL");
|
