https://github.com/sched-ext/scx
- 这个也是其中一个作者写的: https://news.ycombinator.com/item?id=39442400
先试试这个
https://github.com/parttimenerd/minimal-scheduler
[ 9178.179103] sched_ext: "minimal_scheduler" does not implement cgroup cpu.weight
[ 9178.180274] sched_ext: scx_bpf_consume() renamed to scx_bpf_dsq_move_to_local()
[ 9178.180275] sched_ext: scx_bpf_consume() renamed to scx_bpf_dsq_move_to_local()
[ 9178.180275] sched_ext: scx_bpf_consume() renamed to scx_bpf_dsq_move_to_local()
[ 9178.180275] sched_ext: scx_bpf_consume() renamed to scx_bpf_dsq_move_to_local()
[ 9178.180275] sched_ext: scx_bpf_consume() renamed to scx_bpf_dsq_move_to_local()
[ 9178.180275] sched_ext: scx_bpf_consume() renamed to scx_bpf_dsq_move_to_local()
[ 9178.180275] sched_ext: scx_bpf_consume() renamed to scx_bpf_dsq_move_to_local()
[ 9178.180275] sched_ext: scx_bpf_consume() renamed to scx_bpf_dsq_move_to_local()
[ 9178.180450] sched_ext: BPF scheduler "minimal_scheduler" enabled
似乎有两个警告，这个代码更新一下:
是不是使用了
测试
的确，在 start.sh 之后，可以观察到很多这个:
@[
scx_pick_idle_cpu+5
scx_select_cpu_dfl+445
select_task_rq_scx+300
try_to_wake_up+445
wake_up_q+78
futex_wake+345
do_futex+293
__x64_sys_futex+297
do_syscall_64+188
entry_SYSCALL_64_after_hwframe+119
]: 42973
@[
scx_select_cpu_dfl+5
select_task_rq_scx+300
try_to_wake_up+445
wake_up_q+78
futex_wake+345
do_futex+293
__x64_sys_futex+297
do_syscall_64+188
entry_SYSCALL_64_after_hwframe+119
]: 1618732
@[
scx_bpf_dsq_move_to_local+9
bpf_prog_3768393fc527e957_sched_dispatch+27
bpf__sched_ext_ops_dispatch+75
balance_one+334
balance_scx+53
prev_balance+67
__pick_next_task+107
__schedule+362
schedule+65
futex_wait_queue+101
__futex_wait+334
futex_wait+121
do_futex+203
__x64_sys_futex+297
do_syscall_64+188
entry_SYSCALL_64_after_hwframe+119
]: 1618967
关闭之后，就仅仅可以看到这个:
@[
scx_tick+9
sched_tick+221
update_process_times+150
tick_nohz_handler+143
__hrtimer_run_queues+133
hrtimer_interrupt+255
__sysvec_apic_timer_interrupt+82
sysvec_apic_timer_interrupt+110
asm_sysvec_apic_timer_interrupt+26
default_idle+15
default_idle_call+63
do_idle+463
cpu_startup_entry+41
start_secondary+286
common_startup_64+318
]: 7932
具体代码
- sched_fork
难道真的是这个影响的吗？
struct task_struct {
// ...
const struct sched_class *sched_class;
似乎是的，例如看这个例子:
例如 enqueue_task 中
void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
{
if (!(flags & ENQUEUE_NOCLOCK))
update_rq_clock(rq);
p->sched_class->enqueue_task(rq, p, flags);
/*
* Must be after ->enqueue_task() because ENQUEUE_DELAYED can clear
* ->sched_delayed.
*/
uclamp_rq_inc(rq, p);
psi_enqueue(p, flags);
if (!(flags & ENQUEUE_RESTORE))
sched_info_enqueue(rq, p);
if (sched_core_enabled(rq))
sched_core_enqueue(rq, p);
}
当被 wake up 的时候，总是需要先加入到队列中，然后从队列中离开:
@[
dequeue_entity+1
dequeue_entities+289
dequeue_task_fair+151
__schedule+1940
schedule+65
futex_wait_queue+101
__futex_wait+334
futex_wait+121
do_futex+203
__x64_sys_futex+297
do_syscall_64+188
entry_SYSCALL_64_after_hwframe+119
]: 279
在中断的时候加入到队列中:
@[
enqueue_task_fair+5
enqueue_task+55
ttwu_do_activate+111
sched_ttwu_pending+245
__flush_smp_call_function_queue+320
__sysvec_call_function_single+28
sysvec_call_function_single+110
asm_sysvec_call_function_single+26
default_idle+15
@[
enqueue_task_fair+5
enqueue_task+55
ttwu_do_activate+111
sched_ttwu_pending+245
__flush_smp_call_function_queue+320
__sysvec_call_function_single+28
sysvec_call_function_single+110
asm_sysvec_call_function_single+26
default_idle+15
default_idle_call+63
do_idle+463
cpu_startup_entry+41
start_secondary+286
common_startup_64+318
]: 1878
看看 sysfs 的结果
/sys/kernel/sched_ext🔒 🐕
🧀 tree
.
├── enable_seq
├── hotplug_seq
├── nr_rejected
├── root
│ └── ops
├── state
└── switch_all
2 directories, 6 files
tools/sched_ext/ 是做什么的?
kernel 中的 tools/sched_ext/ 做什么的?
当时使用了 ebpf 吗?
https://github.com/google/ghost-userspace
SCHED_EXT
https://mp.weixin.qq.com/s/89PuLJDE4aE1c3cWG6ZL8g
- include/linux/sched/ext.h // ext 调度类的ops接口声明以及调度类核心结构体实现
- kernel/sched/ext.h // ext 调度类的普通功能接口及标志位声明
-
kernel/sched/ext.c // ext 调度类的核心接口实现
- tools/sched_ext/scx_central.c // central scheduler 的加载注册及监控部分实现
- tools/sched_ext/scx_central.bpf.c // central scheduler 的具体策略实现
获取一点代码的基本感觉
基本的感觉是在这里的:
DEFINE_SCHED_CLASS : 所有的都是有的
DEFINE_SCHED_CLASS(ext) = {
.enqueue_task = enqueue_task_scx,
.dequeue_task = dequeue_task_scx,
.yield_task = yield_task_scx,
.yield_to_task = yield_to_task_scx,
.wakeup_preempt = wakeup_preempt_scx,
.balance = balance_scx,
.pick_task = pick_task_scx,
.put_prev_task = put_prev_task_scx,
.set_next_task = set_next_task_scx,
#ifdef CONFIG_SMP
.select_task_rq = select_task_rq_scx,
.task_woken = task_woken_scx,
.set_cpus_allowed = set_cpus_allowed_scx,
.rq_online = rq_online_scx,
.rq_offline = rq_offline_scx,
#endif
.task_tick = task_tick_scx,
.switching_to = switching_to_scx,
.switched_from = switched_from_scx,
.switched_to = switched_to_scx,
.reweight_task = reweight_task_scx,
.prio_changed = prio_changed_scx,
.update_curr = update_curr_scx,
#ifdef CONFIG_UCLAMP_TASK
.uclamp_enabled = 1,
#endif
};
sched_ext_ops : 对于用户提供的 scheduler
static struct sched_ext_ops __bpf_ops_sched_ext_ops = {
.select_cpu = sched_ext_ops__select_cpu,
.enqueue = sched_ext_ops__enqueue,
.dequeue = sched_ext_ops__dequeue,
.dispatch = sched_ext_ops__dispatch,
.tick = sched_ext_ops__tick,
.runnable = sched_ext_ops__runnable,
.running = sched_ext_ops__running,
.stopping = sched_ext_ops__stopping,
.quiescent = sched_ext_ops__quiescent,
.yield = sched_ext_ops__yield,
.core_sched_before = sched_ext_ops__core_sched_before,
.set_weight = sched_ext_ops__set_weight,
.set_cpumask = sched_ext_ops__set_cpumask,
.update_idle = sched_ext_ops__update_idle,
.cpu_acquire = sched_ext_ops__cpu_acquire,
.cpu_release = sched_ext_ops__cpu_release,
.init_task = sched_ext_ops__init_task,
.exit_task = sched_ext_ops__exit_task,
.enable = sched_ext_ops__enable,
.disable = sched_ext_ops__disable,
#ifdef CONFIG_EXT_GROUP_SCHED
.cgroup_init = sched_ext_ops__cgroup_init,
.cgroup_exit = sched_ext_ops__cgroup_exit,
.cgroup_prep_move = sched_ext_ops__cgroup_prep_move,
.cgroup_move = sched_ext_ops__cgroup_move,
.cgroup_cancel_move = sched_ext_ops__cgroup_cancel_move,
.cgroup_set_weight = sched_ext_ops__cgroup_set_weight,
#endif
.cpu_online = sched_ext_ops__cpu_online,
.cpu_offline = sched_ext_ops__cpu_offline,
.init = sched_ext_ops__init,
.exit = sched_ext_ops__exit,
.dump = sched_ext_ops__dump,
.dump_cpu = sched_ext_ops__dump_cpu,
.dump_task = sched_ext_ops__dump_task,
};
为什么有 kthread 程序在不断的调整自己的 level
#0 __sched_setscheduler (p=p@entry=0xffff888004393280,
attr=0xffffc9000013beb8, user=user@entry=false, pi=pi@entry=true)
at kernel/sched/syscalls.c:529
#1 0xffffffff8112d92b in _sched_setscheduler (
param=0xffffffff8220eb00 <param>, check=false,
policy=<optimized out>, p=0xffff888004393280)
at kernel/sched/syscalls.c:788
#2 sched_setscheduler_nocheck (p=p@entry=0xffff888004393280,
policy=policy@entry=0,
param=param@entry=0xffffffff8220eb00 <param>)
at kernel/sched/syscalls.c:835
#3 0xffffffff810dacb1 in kthread (_create=0xffff888004255f80)
at ./arch/x86/include/asm/current.h:47
#4 0xffffffff810487b1 in ret_from_fork (prev=<optimized out>,
regs=0xffffc9000013bf58, fn=0xffffffff810dac50 <kthread>,
fn_arg=0xffff888004255f80) at arch/x86/kernel/process.c:147
#5 0xffffffff810036ea in ret_from_fork_asm ()
at arch/x86/entry/entry_64.S:244
__sched_setscheduler 中在不断在判断 user 这个变量。
此外，如何理解: __sched_setscheduler 中 man sched_setscheduler(2)
/*
* Valid priorities for SCHED_FIFO and SCHED_RR are
* 1..MAX_RT_PRIO-1, valid priority for SCHED_NORMAL,
* SCHED_BATCH and SCHED_IDLE is 0.
*/
if (attr->sched_priority > MAX_RT_PRIO-1)
return -EINVAL;
if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
(rt_policy(policy) != (attr->sched_priority != 0)))
return -EINVAL;
Currently, Linux supports the following "normal" (i.e., non-
real-time) scheduling policies as values that may be specified
in policy:
SCHED_OTHER the standard round-robin time-sharing policy;
SCHED_BATCH for "batch" style execution of processes; and
SCHED_IDLE for running very low priority background jobs.
For each of the above policies, param->sched_priority must be 0.
看看
https://mp.weixin.qq.com/s/d043Be7vfXaq3HFNxUdfLw
https://mp.weixin.qq.com/s/B6zeBR-h-vLgTBe5CwGTSg
https://mp.weixin.qq.com/s/SE1W3IbYP8jeNX5rUDCaGg
https://mp.weixin.qq.com/s/WAPL5h4pigZ3m0-QMi-c0A
本站所有文章转发 CSDN 将按侵权追究法律责任，其它情况随意。