Quelle  scx_central.c   Sprache: C

/* SPDX-License-Identifier: GPL-2.0 */
/*
 * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
 * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
 * Copyright (c) 2022 David Vernet <dvernet@meta.com>
 */
#define _GNU_SOURCE
#include <sched.h>
#include <stdio.h>
#include <unistd.h>
#include <inttypes.h>
#include <signal.h>
#include <assert.h>
#include <libgen.h>
#include <bpf/bpf.h>
#include <scx/common.h>
#include "scx_central.bpf.skel.h"

const char help_fmt[] =
"A central FIFO sched_ext scheduler.\n"
"\n"
"See the top-level comment in .bpf.c for more details.\n"
"\n"
"Usage: %s [-s SLICE_US] [-c CPU]\n"
"\n"
" -s SLICE_US Override slice duration\n"
" -c CPU Override the central CPU (default: 0)\n"
" -v Print libbpf debug messages\n"
" -h Display this help and exit\n";

static bool verbose;
static volatile int exit_req;

static int libbpf_print_fn(enum libbpf_print_level level, const char *format, va_list args)
{
 if (level == LIBBPF_DEBUG && !verbose)
  return 0;
 return vfprintf(stderr, format, args);
}

static void sigint_handler(int dummy)
{
 exit_req = 1;
}

int main(int argc, char **argv)
{
 struct scx_central *skel;
 struct bpf_link *link;
 __u64 seq = 0, ecode;
 __s32 opt;
 cpu_set_t *cpuset;

 libbpf_set_print(libbpf_print_fn);
 signal(SIGINT, sigint_handler);
 signal(SIGTERM, sigint_handler);
restart:
 skel = SCX_OPS_OPEN(central_ops, scx_central);

 skel->rodata->central_cpu = 0;
 skel->rodata->nr_cpu_ids = libbpf_num_possible_cpus();
 skel->rodata->slice_ns = __COMPAT_ENUM_OR_ZERO("scx_public_consts", "SCX_SLICE_DFL");

 assert(skel->rodata->nr_cpu_ids <= INT32_MAX);

 while ((opt = getopt(argc, argv, "s:c:pvh")) != -1) {
  switch (opt) {
  case 's':
   skel->rodata->slice_ns = strtoull(optarg, NULL, 0) * 1000;
   break;
  case 'c': {
   u32 central_cpu = strtoul(optarg, NULL, 0);
   if (central_cpu >= skel->rodata->nr_cpu_ids) {
    fprintf(stderr, "invalid central CPU id value, %u given (%u max)\n", central_cpu, skel->rodata->nr_cpu_ids);
    return -1;
   }
   skel->rodata->central_cpu = (s32)central_cpu;
   break;
  }
  case 'v':
   verbose = true;
   break;
  default:
   fprintf(stderr, help_fmt, basename(argv[0]));
   return opt != 'h';
  }
 }

 /* Resize arrays so their element count is equal to cpu count. */
 RESIZE_ARRAY(skel, data, cpu_gimme_task, skel->rodata->nr_cpu_ids);
 RESIZE_ARRAY(skel, data, cpu_started_at, skel->rodata->nr_cpu_ids);

 SCX_OPS_LOAD(skel, central_ops, scx_central, uei);

 /*
 * Affinitize the loading thread to the central CPU, as:
 * - That's where the BPF timer is first invoked in the BPF program.
 * - We probably don't want this user space component to take up a core
 *   from a task that would benefit from avoiding preemption on one of
 *   the tickless cores.
 *
 * Until BPF supports pinning the timer, it's not guaranteed that it
 * will always be invoked on the central CPU. In practice, this
 * suffices the majority of the time.
 */
 cpuset = CPU_ALLOC(skel->rodata->nr_cpu_ids);
 SCX_BUG_ON(!cpuset, "Failed to allocate cpuset");
 CPU_ZERO_S(CPU_ALLOC_SIZE(skel->rodata->nr_cpu_ids), cpuset);
 CPU_SET(skel->rodata->central_cpu, cpuset);
 SCX_BUG_ON(sched_setaffinity(0, sizeof(*cpuset), cpuset),
     "Failed to affinitize to central CPU %d (max %d)",
     skel->rodata->central_cpu, skel->rodata->nr_cpu_ids - 1);
 CPU_FREE(cpuset);

 link = SCX_OPS_ATTACH(skel, central_ops, scx_central);

 if (!skel->data->timer_pinned)
  printf("WARNING : BPF_F_TIMER_CPU_PIN not available, timer not pinned to central\n");

 while (!exit_req && !UEI_EXITED(skel, uei)) {
  printf("[SEQ %llu]\n", seq++);
  printf("total :%10" PRIu64 " local:%10" PRIu64 " queued:%10" PRIu64 " lost:%10" PRIu64 "\n",
         skel->bss->nr_total,
         skel->bss->nr_locals,
         skel->bss->nr_queued,
         skel->bss->nr_lost_pids);
  printf("timer :%10" PRIu64 " dispatch:%10" PRIu64 " mismatch:%10" PRIu64 " retry:%10" PRIu64 "\n",
         skel->bss->nr_timers,
         skel->bss->nr_dispatches,
         skel->bss->nr_mismatches,
         skel->bss->nr_retries);
  printf("overflow:%10" PRIu64 "\n",
         skel->bss->nr_overflows);
  fflush(stdout);
  sleep(1);
 }

 bpf_link__destroy(link);
 ecode = UEI_REPORT(skel, uei);
 scx_central__destroy(skel);

 if (UEI_ECODE_RESTART(ecode))
  goto restart;
 return 0;
}