进程调度窥探，进程调度直观感受"/>
sched，进程调度窥探，进程调度直观感受
前言
进程调度是一个复杂工程，调度策略的好坏直接影响系统的体验。
在 LInux Kernel 中，使用的最多的就是完全公平调度（CFS）算法，今天我们 hack 内核，窥探它是如何调度一个个进程的。
挑选下一个任务
pick_next_task_fair()，挑选下一个要执行的任务
我们添加打印，直观感受下进程间的调度切换
kernel/sched/core.c
/*
* Pick up the highest-prio task:
*/
static inline struct task_struct *
__pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
{const struct sched_class *class;struct task_struct *p;// 添加static int count = 0;/** Optimization: we know that if all tasks are in the fair class we can* call that function directly, but only if the @prev task wasn't of a* higher scheduling class, because otherwise those lose the* opportunity to pull in more work from other CPUs.*/if (likely(prev->sched_class <= &fair_sched_class &&rq->nr_running == rq->cfs.h_nr_running)) {p = pick_next_task_fair(rq, prev, rf);if (unlikely(p == RETRY_TASK))goto restart;/* Assume the next prioritized class is idle_sched_class */if (!p) {put_prev_task(rq, prev);p = pick_next_task_idle(rq);}// 添加以下四行if (count < 10000) {printk("-pid: %3d, %16.16s, %d\n", p->pid, p->comm, count);count++;}return p;}restart:put_prev_task_balance(rq, prev, rf);for_each_class(class) {p = class->pick_next_task(rq);if (p)return p;}/* The idle class should always have a runnable task: */BUG();
}
打印
smp_twd: clock not found -2
Console: colour dummy device 80x30
Calibrating local timer... 93.37MHz.
Calibrating delay loop... 1581.05 BogoMIPS (lpj=7905280)
pid_max: default: 32768 minimum: 301
Mount-cache hash table entries: 1024 (order: 0, 4096 bytes, linear)
Mountpoint-cache hash table entries: 1024 (order: 0, 4096 bytes, linear)
CPU: Testing write buffer coherency: ok
CPU0: Spectre v2: using BPIALL workaround
-pid: 2, swapper/0, 0
-pid: 1, swapper/0, 1
CPU0: thread -1, cpu 0, socket 0, mpidr 80000000
-pid: 2, kthreadd, 2
-pid: 3, kthreadd, 3
-pid: 1, swapper/0, 4
-pid: 2, kthreadd, 5
-pid: 3, rcu_gp, 6
-pid: 4, kthreadd, 7
-pid: 1, swapper/0, 8
-pid: 2, kthreadd, 9
-pid: 4, rcu_par_gp, 10
-pid: 5, kthreadd, 11
-pid: 1, swapper/0, 12
-pid: 2, kthreadd, 13
-pid: 5, kworker/0:0, 14
-pid: 6, kthreadd, 15
-pid: 1, swapper/0, 16
-pid: 2, kthreadd, 17
-pid: 6, kworker/0:0H, 18
-pid: 7, kthreadd, 19
-pid: 1, swapper/0, 20
-pid: 2, kthreadd, 21
-pid: 7, kworker/u8:0, 22
-pid: 8, kthreadd, 23
-pid: 1, swapper/0, 24
Setting up static identity map for 0x60100000 - 0x60100060
-pid: 2, kthreadd, 25
-pid: 8, mm_percpu_wq, 26
-pid: 9, kthreadd, 27
-pid: 1, swapper/0, 28
-pid: 9, ksoftirqd/0, 29
-pid: 1, swapper/0, 30
rcu: Hierarchical SRCU implementation.
-pid: 2, kthreadd, 31
-pid: 9, ksoftirqd/0, 32
-pid: 10, kthreadd, 33
-pid: 1, swapper/0, 34
-pid: 2, kthreadd, 35
-pid: 10, rcu_sched, 36
-pid: 11, kthreadd, 37
-pid: 1, swapper/0, 38
-pid: 11, migration/0, 39
-pid: 1, swapper/0, 40
-pid: 12, kthreadd, 41
-pid: 1, swapper/0, 42
-pid: 12, cpuhp/0, 43
-pid: 1, swapper/0, 44
smp: Bringing up secondary CPUs ...
-pid: 2, kthreadd, 45
-pid: 12, cpuhp/0, 46
-pid: 13, kthreadd, 47
-pid: 1, swapper/0, 48
-pid: 13, cpuhp/1, 49
-pid: 1, swapper/0, 50
-pid: 2, kthreadd, 51
-pid: 14, kthreadd, 52
-pid: 1, swapper/0, 53
-pid: 14, migration/1, 54
-pid: 1, swapper/0, 55
-pid: 2, kthreadd, 56
-pid: 15, kthreadd, 57
-pid: 1, swapper/0, 58
-pid: 15, ksoftirqd/1, 59
-pid: 1, swapper/0, 60
-pid: 2, kthreadd, 61
-pid: 10, rcu_sched, 62
-pid: 16, kthreadd, 63
-pid: 1, swapper/0, 64
-pid: 2, kthreadd, 65
-pid: 16, kworker/1:0, 66
-pid: 17, kthreadd, 67
-pid: 1, swapper/0, 68
-pid: 17, kworker/1:0H, 69
-pid: 0, swapper/0, 70
CPU1: thread -1, cpu 1, socket 0, mpidr 80000001
CPU1: Spectre v2: using BPIALL workaround
-pid: 1, swapper/0, 71
-pid: 13, cpuhp/1, 72
-pid: 0, swapper/0, 72
-pid: 16, kworker/1:0, 74
-pid: 1, swapper/0, 74
-pid: 15, ksoftirqd/1, 75
-pid: 17, kworker/1:0H, 76
-pid: 0, swapper/1, 77
-pid: 2, kthreadd, 79
-pid: 18, kthreadd, 80
-pid: 1, swapper/0, 81
-pid: 18, cpuhp/2, 82
-pid: 1, swapper/0, 83
-pid: 2, kthreadd, 84
-pid: 19, kthreadd, 85
-pid: 1, swapper/0, 86
-pid: 19, migration/2, 87
-pid: 1, swapper/0, 88
-pid: 2, kthreadd, 89
-pid: 20, kthreadd, 90
-pid: 1, swapper/0, 91
-pid: 20, ksoftirqd/2, 92
-pid: 1, swapper/0, 93
-pid: 2, kthreadd, 94
-pid: 21, kthreadd, 95
-pid: 1, swapper/0, 96
-pid: 2, kthreadd, 97
-pid: 21, kworker/2:0, 98
-pid: 22, kthreadd, 99
-pid: 1, swapper/0, 100
-pid: 10, rcu_sched, 101
-pid: 22, kworker/2:0H, 102
-pid: 0, swapper/0, 103
-pid: 10, rcu_sched, 104
-pid: 0, swapper/0, 105
-pid: 10, rcu_sched, 106
-pid: 0, swapper/0, 107
-pid: 10, rcu_sched, 108
-pid: 0, swapper/0, 109
CPU2: thread -1, cpu 2, socket 0, mpidr 80000002
CPU2: Spectre v2: using BPIALL workaround
-pid: 1, swapper/0, 110
-pid: 10, rcu_sched, 111
-pid: 18, cpuhp/2, 112
-pid: 0, swapper/0, 113
-pid: 20, ksoftirqd/2, 114
-pid: 1, swapper/0, 114
-pid: 2, kthreadd, 116
-pid: 21, kworker/2:0, 117
-pid: 22, kworker/2:0H, 118
-pid: 23, kthreadd, 119
-pid: 0, swapper/2, 119
-pid: 1, swapper/0, 120
-pid: 23, cpuhp/3, 122
-pid: 1, swapper/0, 123
-pid: 2, kthreadd, 124
-pid: 24, kthreadd, 125
-pid: 1, swapper/0, 126
-pid: 24, migration/3, 127
-pid: 1, swapper/0, 128
-pid: 2, kthreadd, 129
-pid: 25, kthreadd, 130
-pid: 1, swapper/0, 131
-pid: 25, ksoftirqd/3, 132
-pid: 1, swapper/0, 133
-pid: 2, kthreadd, 134
-pid: 26, kthreadd, 135
-pid: 1, swapper/0, 136
-pid: 2, kthreadd, 137
-pid: 26, kworker/3:0, 138
-pid: 27, kthreadd, 139
-pid: 1, swapper/0, 140
-pid: 27, kworker/3:0H, 141
-pid: 0, swapper/0, 142
CPU3: thread -1, cpu 3, socket 0, mpidr 80000003
CPU3: Spectre v2: using BPIALL workaround
-pid: 1, swapper/0, 143
-pid: 0, swapper/0, 144
-pid: 1, swapper/0, 145
-pid: 0, swapper/0, 146
-pid: 23, cpuhp/3, 147
-pid: 25, ksoftirqd/3, 148
-pid: 26, kworker/3:0, 149
-pid: 27, kworker/3:0H, 150
-pid: 0, swapper/3, 151
-pid: 1, swapper/0, 152
smp: Brought up 1 node, 4 CPUs
SMP: Total of 4 processors activated (5903.15 BogoMIPS).
CPU: All CPU(s) started in SVC mode.
-pid: 10, rcu_sched, 153
-pid: 2, kthreadd, 154
-pid: 28, kthreadd, 155
-pid: 0, swapper/1, 156
-pid: 0, swapper/0, 156
-pid: 1, swapper/0, 158
-pid: 0, swapper/0, 159
-pid: 28, kdevtmpfs, 160
-pid: 0, swapper/1, 161
-pid: 1, swapper/0, 161
devtmpfs: initialized
-pid: 10, rcu_sched, 163
-pid: 0, swapper/2, 164
-pid: 2, kthreadd, 165
-pid: 0, swapper/0, 166
-pid: 29, kthreadd, 167
-pid: 0, swapper/3, 168
-pid: 1, swapper/0, 169
-pid: 29, inet_frag_wq, 170
-pid: 0, swapper/3, 171
VFP support v0.3: implementor 41 architecture 3 part 30 variant 9 rev 0
clocksource: jiffies: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 19112604462750000 ns
futex hash table entries: 1024 (order: 4, 65536 bytes, linear)
NET: Registered PF_NETLINK/PF_ROUTE protocol family
-pid: 16, kworker/1:0, 172
-pid: 5, kworker/0:0, 172
-pid: 26, kworker/3:0, 172
-pid: 2, kthreadd, 175
-pid: 0, swapper/1, 176
-pid: 0, swapper/3, 176
-pid: 30, kthreadd, 178
-pid: 0, swapper/0, 179
-pid: 32, kthreadd, 179
-pid: 31, kthreadd, 181
-pid: 0, swapper/2, 182
-pid: 5, kworker/0:0, 183
-pid: 16, kworker/1:0, 183
-pid: 26, kworker/3:0, 184
-pid: 30, kworker/1:1, 186
-pid: 32, kworker/3:1, 186
-pid: 31, kworker/0:1, 186
-pid: 1, swapper/0, 189
-pid: 0, swapper/1, 190
-pid: 0, swapper/3, 190
-pid: 31, kworker/0:1, 192
-pid: 30, kworker/1:1, 193
-pid: 32, kworker/3:1, 194
-pid: 21, kworker/2:0, 194
-pid: 1, swapper/0, 196
-pid: 0, swapper/3, 196
-pid: 0, swapper/2, 197
-pid: 2, kthreadd, 199
-pid: 1, swapper/0, 200
-pid: 33, kthreadd, 201
-pid: 21, kworker/2:0, 202
-pid: 0, swapper/1, 203
-pid: 33, kworker/2:1, 204
-pid: 0, swapper/0, 205
-pid: 0, swapper/2, 205
-pid: 1, swapper/0, 207
DMA: preallocated 256 KiB pool for atomic coherent allocations
-pid: 31, kworker/0:1, 208
-pid: 1, swapper/0, 209
cpuidle: using governor ladder
hw-breakpoint: debug architecture 0x4 unsupported.
Serial: AMBA PL011 UART driver
irq: type mismatch, failed to map hwirq-75 for interrupt-controller@1e001000!
-pid: 10, rcu_sched, 210
-pid: 0, swapper/2, 211
-pid: 31, kworker/0:1, 212
-pid: 1, swapper/0, 213
-pid: 2, kthreadd, 214
-pid: 0, swapper/0, 215
-pid: 34, kthreadd, 216
-pid: 0, swapper/2, 217
-pid: 1, swapper/0, 218
-pid: 2, kthreadd, 219
-pid: 34, khungtaskd, 220
-pid: 0, swapper/0, 221
-pid: 0, swapper/2, 222
-pid: 35, kthreadd, 223
-pid: 0, swapper/1, 224
-pid: 1, swapper/0, 225
-pid: 35, oom_reaper, 226
-pid: 2, kthreadd, 227
-pid: 0, swapper/1, 228
-pid: 0, swapper/0, 229
-pid: 36, kthreadd, 230
-pid: 0, swapper/2, 231
-pid: 1, swapper/0, 232
-pid: 36, writeback, 233
-pid: 0, swapper/2, 234
-pid: 2, kthreadd, 235
-pid: 0, swapper/0, 236
-pid: 37, kthreadd, 237
-pid: 0, swapper/3, 238
-pid: 1, swapper/0, 239
-pid: 37, kcompactd0, 240
-pid: 0, swapper/3, 241
-pid: 10, rcu_sched, 242
-pid: 0, swapper/2, 243
-pid: 2, kthreadd, 244
-pid: 0, swapper/0, 245
-pid: 38, kthreadd, 246
-pid: 0, swapper/3, 247
-pid: 1, swapper/0, 248
-pid: 38, kblockd, 249
-pid: 0, swapper/3, 250
SCSI subsystem initialized
-pid: 2, kthreadd, 251
-pid: 0, swapper/0, 252
-pid: 39, kthreadd, 253
-pid: 0, swapper/3, 254
-pid: 1, swapper/0, 255
-pid: 39, ata_sff, 256
-pid: 0, swapper/3, 257
usbcore: registered new interface driver usbfs
usbcore: registered new interface driver hub
usbcore: registered new device driver usb
-pid: 10, rcu_sched, 258
-pid: 0, swapper/2, 259
pps_core: LinuxPPS API ver. 1 registered
pps_core: Software ver. 5.3.6 - Copyright 2005-2007 Rodolfo Giometti <giometti@linux.it>
PTP clock support registered
Advanced Linux Sound Architecture Driver Initialized.
-pid: 1, swapper/0, 260
clocksource: Switched to clocksource arm,sp804
-pid: 10, rcu_sched, 261
-pid: 0, swapper/2, 262
-pid: 7, kworker/u8:0, 263
-pid: 2, kthreadd, 264
-pid: 28, kdevtmpfs, 265
-pid: 0, swapper/0, 266
-pid: 40, kthreadd, 267
-pid: 0, swapper/3, 268
-pid: 7, kworker/u8:0, 269
-pid: 40, kworker/u8:1, 270
-pid: 0, swapper/3, 271
-pid: 0, swapper/1, 272
-pid: 1, swapper/0, 273
-pid: 0, swapper/3, 274
-pid: 28, kdevtmpfs, 275
-pid: 0, swapper/1, 276
-pid: 1, swapper/0, 277
-pid: 0, swapper/3, 278
-pid: 28, kdevtmpfs, 279
-pid: 0, swapper/1, 280
-pid: 1, swapper/0, 281
-pid: 0, swapper/3, 282
-pid: 28, kdevtmpfs, 283
-pid: 0, swapper/1, 284
-pid: 1, swapper/0, 285
-pid: 0, swapper/3, 286
-pid: 28, kdevtmpfs, 287
-pid: 0, swapper/1, 288
-pid: 1, swapper/0, 289
-pid: 0, swapper/3, 290
-pid: 28, kdevtmpfs, 291
-pid: 0, swapper/1, 292
-pid: 1, swapper/0, 293
-pid: 0, swapper/3, 294
-pid: 28, kdevtmpfs, 295
-pid: 0, swapper/1, 296
-pid: 1, swapper/0, 297
-pid: 0, swapper/3, 298
-pid: 28, kdevtmpfs, 299
-pid: 0, swapper/1, 300
-pid: 1, swapper/0, 301
-pid: 0, swapper/3, 302
-pid: 28, kdevtmpfs, 303
-pid: 0, swapper/1, 304
-pid: 1, swapper/0, 305
-pid: 0, swapper/3, 306
-pid: 28, kdevtmpfs, 307
-pid: 0, swapper/1, 308
-pid: 1, swapper/0, 309
-pid: 0, swapper/3, 310
-pid: 28, kdevtmpfs, 311
# ps
PID USER COMMAND1 root init2 root [kthreadd]3 root [rcu_gp]4 root [rcu_par_gp]5 root [kworker/0:0-mm_]6 root [kworker/0:0H-mm]7 root [kworker/u8:0-ev]8 root [mm_percpu_wq]9 root [ksoftirqd/0]10 root [rcu_sched]11 root [migration/0]12 root [cpuhp/0]13 root [cpuhp/1]14 root [migration/1]15 root [ksoftirqd/1]16 root [kworker/1:0-rcu]17 root [kworker/1:0H-kb]18 root [cpuhp/2]19 root [migration/2]20 root [ksoftirqd/2]21 root [kworker/2:0-mm_]22 root [kworker/2:0H-kb]23 root [cpuhp/3]24 root [migration/3]25 root [ksoftirqd/3]26 root [kworker/3:0-rcu]27 root [kworker/3:0H-ev]28 root [kdevtmpfs]29 root [inet_frag_wq]30 root [kworker/1:1-pm]31 root [kworker/0:1-mm_]32 root [kworker/3:1-eve]33 root [kworker/2:1-mm_]34 root [khungtaskd]35 root [oom_reaper]36 root [writeback]37 root [kcompactd0]38 root [kblockd]39 root [ata_sff]40 root [kworker/u8:1-ev]41 root [rpciod]42 root [kworker/3:1H-kb]43 root [kworker/u9:0]44 root [xprtiod]45 root [kswapd0]46 root [nfsiod]47 root [kworker/3:2-eve]48 root [kworker/u8:2-ev]49 root [kworker/1:2-mm_]53 root [irq/33-mmci-pl1]54 root [kworker/1:3-eve]55 root [mmc_complete]56 root [kworker/1:1H-kb]57 root [kworker/0:1H-mm]58 root [card0-crtc0]59 root [ext4-rsv-conver]61 root [kworker/3:2H]74 root [kworker/2:1H-kb]78 root /sbin/syslogd -n82 root /sbin/klogd -n134 root udhcpc -t1 -A3 -b -R -O search -O staticroutes -p /var/run/udhcp136 root -sh138 root ps前几十次调度，都是在 20 以内的进程间切换，这个时候正处于内核线程创建阶段，进程逐渐增加
275 次之后，基本在 0 1 28 号进程之间切换，0 空闲，1 swapper/0 (init 进程的前身)，28 是 inet_frag_wq 网络相关
等到内核加载完毕，启动 init 进程后，1 号进程就从 swapper/0 变成 init 了
-pid: 58, kworker/0:1H, 1412
-pid: 0, swapper/0, 1413
-pid: 1, swapper/0, 1414
-pid: 57, kworker/1:1H, 1415
-pid: 0, swapper/1, 1416
-pid: 58, kworker/0:1H, 1417
-pid: 0, swapper/0, 1418
-pid: 1, swapper/0, 1419
-pid: 57, kworker/1:1H, 1420
-pid: 0, swapper/1, 1421
-pid: 37, kcompactd0, 1422
-pid: 0, swapper/2, 1423
-pid: 57, kworker/1:1H, 1424
-pid: 15, ksoftirqd/1, 1425
-pid: 1, swapper/0, 1426
-pid: 57, kworker/1:1H, 1427
-pid: 1, swapper/0, 1428 //
-pid: 57, kworker/1:1H, 1429
-pid: 15, ksoftirqd/1, 1430
-pid: 1, init, 1431 // 1 号进程由 swapper/0 变为 init
-pid: 9, ksoftirqd/0, 1432
-pid: 58, kworker/0:1H, 1433
-pid: 0, swapper/0, 1434
-pid: 57, kworker/1:1H, 1435
-pid: 0, swapper/1, 1436
-pid: 58, kworker/0:1H, 1437
-pid: 0, swapper/0, 1438
-pid: 1, init, 1439
-pid: 57, kworker/1:1H, 1440
-pid: 0, swapper/1, 1441
-pid: 58, kworker/0:1H, 1442
-pid: 0, swapper/0, 1443
-pid: 1, init, 1444
看到 S02sysctl 之类的，就知道是在执行 /etcinit.d/ 下的各个启动脚本
-pid: 87, S02sysctl, 2303
-pid: 0, swapper/1, 2304
-pid: 88, S02sysctl, 2305
-pid: 81, syslogd, 2306
-pid: 88, S02sysctl, 2307
-pid: 89, S02sysctl, 2308
-pid: 81, syslogd, 2309
-pid: 37, kcompactd0, 2310
-pid: 10, rcu_sched, 2311
-pid: 88, S02sysctl, 2311
-pid: 85, klogd, 2312
-pid: 81, syslogd, 2314
-pid: 20, ksoftirqd/2, 2314
-pid: 85, klogd, 2316
-pid: 88, S02sysctl, 2316
-pid: 81, syslogd, 2318
-pid: 87, S02sysctl, 2319
-pid: 88, S02sysctl, 2319
-pid: 81, syslogd, 2321
-pid: 89, S02sysctl, 2322
-pid: 88, S02sysctl, 2322
-pid: 90, S02sysctl, 2323
-pid: 10, rcu_sched, 2323
-pid: 85, klogd, 2324
-pid: 81, syslogd, 2326
-pid: 88, S02sysctl, 2327
-pid: 81, syslogd, 2328
-pid: 91, S02sysctl, 2328
-pid: 10, rcu_sched, 2330
-pid: 85, klogd, 2332
-pid: 20, ksoftirqd/2, 2333
-pid: 88, S02sysctl, 2333
-pid: 85, klogd, 2334
-pid: 30, kworker/0:1, 2334
-pid: 89, xargs, 2337
-pid: 91, S02sysctl, 2338
-pid: 81, syslogd, 2339
-pid: 0, swapper/3, 2340
再往后，进程调度基本定格在 0 idle 进程和 syslogd 进程之间
-pid: 81, syslogd, 3145
-pid: 85, klogd, 3147
-pid: 0, swapper/1, 3148
-pid: 81, syslogd, 3149
-pid: 0, swapper/1, 3150
-pid: 81, syslogd, 3151
-pid: 0, swapper/1, 3152
-pid: 81, syslogd, 3153
-pid: 0, swapper/1, 3154
-pid: 81, syslogd, 3155
-pid: 0, swapper/1, 3156
-pid: 81, syslogd, 3157
-pid: 0, swapper/1, 3158
-pid: 81, syslogd, 3159
-pid: 10, rcu_sched, 3160
-pid: 0, swapper/0, 3161
-pid: 0, swapper/1, 3162
-pid: 81, syslogd, 3163
-pid: 20, ksoftirqd/2, 3164
-pid: 85, klogd, 3165
-pid: 0, swapper/1, 3165
-pid: 81, syslogd, 3167
-pid: 0, swapper/1, 3168
-pid: 42, kworker/3:1H, 3168
-pid: 81, syslogd, 3170
-pid: 0, swapper/3, 3171
-pid: 58, kworker/0:1H, 3172
-pid: 0, swapper/0, 3173
-pid: 0, swapper/1, 3174
-pid: 107, ifup, 3174
-pid: 81, syslogd, 3176
-pid: 0, swapper/1, 3177
更多推荐
sched,进程调度窥探,进程调度直观感受
发布评论