klockstat.c

// SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause)
/* Copyright (c) 2021 Google LLC.
 *
 * Based on klockstat from BCC by Jiri Olsa and others
 * 2021-10-26   Barret Rhoden   Created this.
 */
/* Differences from BCC python tool:
 * - can specify a lock by ksym name, using '-L'
 * - tracks whichever task had the max time for acquire and hold, outputted
 *     when '-s' > 1 (otherwise it's cluttered).
 * - does not reset stats each interval by default. Can request with -R.
 */
#ifndef _GNU_SOURCE
#define _GNU_SOURCE
#endif
#include <argp.h>
#include <errno.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <unistd.h>
#include <sys/param.h>

#include <bpf/libbpf.h>
#include <bpf/bpf.h>
#include "klockstat.h"
#include "klockstat.skel.h"
#include "trace_helpers.h"

#define warn(...) fprintf(stderr, __VA_ARGS__)

enum {
	SORT_ACQ_MAX,
	SORT_ACQ_COUNT,
	SORT_ACQ_TOTAL,
	SORT_HLD_MAX,
	SORT_HLD_COUNT,
	SORT_HLD_TOTAL,
};

static struct prog_env {
	pid_t pid;
	pid_t tid;
	char *caller;
	char *lock_name;
	unsigned int nr_locks;
	unsigned int nr_stack_entries;
	unsigned int sort_acq;
	unsigned int sort_hld;
	unsigned int duration;
	unsigned int interval;
	unsigned int iterations;
	bool reset;
	bool timestamp;
	bool verbose;
} env = {
	.nr_locks = 99999999,
	.nr_stack_entries = 1,
	.sort_acq = SORT_ACQ_MAX,
	.sort_hld = SORT_HLD_MAX,
	.interval = 99999999,
	.iterations = 99999999,
};

const char *argp_program_version = "klockstat 0.2";
const char *argp_program_bug_address =
	"https://github.com/iovisor/bcc/tree/master/libbpf-tools";
static const char args_doc[] = "FUNCTION";
static const char program_doc[] =
"Trace mutex/sem lock acquisition and hold times, in nsec\n"
"\n"
"Usage: klockstat [-hRTv] [-p PID] [-t TID] [-c FUNC] [-L LOCK] [-n NR_LOCKS]\n"
"                 [-s NR_STACKS] [-S SORT] [-d DURATION] [-i INTERVAL]\n"
"\v"
"Examples:\n"
"  klockstat                     # trace system wide until ctrl-c\n"
"  klockstat -d 5                # trace for 5 seconds\n"
"  klockstat -i 5                # print stats every 5 seconds\n"
"  klockstat -p 181              # trace process 181 only\n"
"  klockstat -t 181              # trace thread 181 only\n"
"  klockstat -c pipe_            # print only for lock callers with 'pipe_'\n"
"                                # prefix\n"
"  klockstat -L cgroup_mutex     # trace the cgroup_mutex lock only\n"
"  klockstat -S acq_count        # sort lock acquired results by acquire count\n"
"  klockstat -S hld_total        # sort lock held results by total held time\n"
"  klockstat -S acq_count,hld_total  # combination of above\n"
"  klockstat -n 3                # display top 3 locks\n"
"  klockstat -s 6                # display 6 stack entries per lock\n"
;

static const struct argp_option opts[] = {
	{ "pid", 'p', "PID", 0, "Filter by process ID" },
	{ "tid", 't', "TID", 0, "Filter by thread ID" },
	{ 0, 0, 0, 0, "" },
	{ "caller", 'c', "FUNC", 0, "Filter by caller string prefix" },
	{ "lock", 'L', "LOCK", 0, "Filter by specific ksym lock name" },
	{ 0, 0, 0, 0, "" },
	{ "locks", 'n', "NR_LOCKS", 0, "Number of locks to print" },
	{ "stacks", 's', "NR_STACKS", 0, "Number of stack entries to print per lock" },
	{ "sort", 'S', "SORT", 0, "Sort by field:\n  acq_[max|total|count]\n  hld_[max|total|count]" },
	{ 0, 0, 0, 0, "" },
	{ "duration", 'd', "SECONDS", 0, "Duration to trace" },
	{ "interval", 'i', "SECONDS", 0, "Print interval" },
	{ "reset", 'R', NULL, 0, "Reset stats each interval" },
	{ "timestamp", 'T', NULL, 0, "Print timestamp" },
	{ "verbose", 'v', NULL, 0, "Verbose debug output" },

	{ NULL, 'h', NULL, OPTION_HIDDEN, "Show the full help" },
	{},
};

static bool parse_one_sort(struct prog_env *env, const char *sort)
{
	const char *field = sort + 4;

	if (!strncmp(sort, "acq_", 4)) {
		if (!strcmp(field, "max")) {
			env->sort_acq = SORT_ACQ_MAX;
			return true;
		} else if (!strcmp(field, "total")) {
			env->sort_acq = SORT_ACQ_TOTAL;
			return true;
		} else if (!strcmp(field, "count")) {
			env->sort_acq = SORT_ACQ_COUNT;
			return true;
		}
	} else if (!strncmp(sort, "hld_", 4)) {
		if (!strcmp(field, "max")) {
			env->sort_hld = SORT_HLD_MAX;
			return true;
		} else if (!strcmp(field, "total")) {
			env->sort_hld = SORT_HLD_TOTAL;
			return true;
		} else if (!strcmp(field, "count")) {
			env->sort_hld = SORT_HLD_COUNT;
			return true;
		}
	}

	return false;
}

static bool parse_sorts(struct prog_env *env, char *arg)
{
	char *comma = strchr(arg, ',');

	if (comma) {
		*comma = '\0';
		comma++;
		if (!parse_one_sort(env, comma))
			return false;
	}
	return parse_one_sort(env, arg);
}

static error_t parse_arg(int key, char *arg, struct argp_state *state)
{
	struct prog_env *env = state->input;
	long duration, interval;

	switch (key) {
	case 'p':
		errno = 0;
		env->pid = strtol(arg, NULL, 10);
		if (errno || env->pid <= 0) {
			warn("Invalid PID: %s\n", arg);
			argp_usage(state);
		}
		break;
	case 't':
		errno = 0;
		env->tid = strtol(arg, NULL, 10);
		if (errno || env->tid <= 0) {
			warn("Invalid TID: %s\n", arg);
			argp_usage(state);
		}
		break;
	case 'c':
		env->caller = arg;
		break;
	case 'L':
		env->lock_name = arg;
		break;
	case 'n':
		errno = 0;
		env->nr_locks = strtol(arg, NULL, 10);
		if (errno || env->nr_locks <= 0) {
			warn("Invalid NR_LOCKS: %s\n", arg);
			argp_usage(state);
		}
		break;
	case 's':
		errno = 0;
		env->nr_stack_entries = strtol(arg, NULL, 10);
		if (errno || env->nr_stack_entries <= 0) {
			warn("Invalid NR_STACKS: %s\n", arg);
			argp_usage(state);
		}
		break;
	case 'S':
		if (!parse_sorts(env, arg)) {
			warn("Bad sort string: %s\n", arg);
			argp_usage(state);
		}
		break;
	case 'd':
		errno = 0;
		duration = strtol(arg, NULL, 10);
		if (errno || duration <= 0) {
			warn("Invalid duration: %s\n", arg);
			argp_usage(state);
		}
		env->duration = duration;
		break;
	case 'i':
		errno = 0;
		interval = strtol(arg, NULL, 10);
		if (errno || interval <= 0) {
			warn("Invalid interval: %s\n", arg);
			argp_usage(state);
		}
		env->interval = interval;
		break;
	case 'R':
		env->reset = true;
		break;
	case 'T':
		env->timestamp = true;
		break;
	case 'h':
		argp_state_help(state, stderr, ARGP_HELP_STD_HELP);
		break;
	case 'v':
		env->verbose = true;
		break;
	case ARGP_KEY_END:
		if (env->duration) {
			if (env->interval > env->duration)
				env->interval = env->duration;
			env->iterations = env->duration / env->interval;
		}
		break;
	default:
		return ARGP_ERR_UNKNOWN;
	}
	return 0;
}

struct stack_stat {
	uint32_t stack_id;
	struct lock_stat ls;
	uint64_t bt[PERF_MAX_STACK_DEPTH];
};

static bool caller_is_traced(struct ksyms *ksyms, uint64_t caller_pc)
{
	const struct ksym *ksym;

	if (!env.caller)
		return true;
	ksym = ksyms__map_addr(ksyms, caller_pc);
	if (!ksym)
		return true;
	return strncmp(env.caller, ksym->name, strlen(env.caller)) == 0;
}

static int larger_first(uint64_t x, uint64_t y)
{
	if (x > y)
		return -1;
	if (x == y)
		return 0;
	return 1;
}

static int sort_by_acq(const void *x, const void *y)
{
	struct stack_stat *ss_x = *(struct stack_stat**)x;
	struct stack_stat *ss_y = *(struct stack_stat**)y;

	switch (env.sort_acq) {
	case SORT_ACQ_MAX:
		return larger_first(ss_x->ls.acq_max_time,
				    ss_y->ls.acq_max_time);
	case SORT_ACQ_COUNT:
		return larger_first(ss_x->ls.acq_count,
				    ss_y->ls.acq_count);
	case SORT_ACQ_TOTAL:
		return larger_first(ss_x->ls.acq_total_time,
				    ss_y->ls.acq_total_time);
	}

	warn("bad sort_acq %d\n", env.sort_acq);
	return -1;
}

static int sort_by_hld(const void *x, const void *y)
{
	struct stack_stat *ss_x = *(struct stack_stat**)x;
	struct stack_stat *ss_y = *(struct stack_stat**)y;

	switch (env.sort_hld) {
	case SORT_HLD_MAX:
		return larger_first(ss_x->ls.hld_max_time,
				    ss_y->ls.hld_max_time);
	case SORT_HLD_COUNT:
		return larger_first(ss_x->ls.hld_count,
				    ss_y->ls.hld_count);
	case SORT_HLD_TOTAL:
		return larger_first(ss_x->ls.hld_total_time,
				    ss_y->ls.hld_total_time);
	}

	warn("bad sort_hld %d\n", env.sort_hld);
	return -1;
}

static char *symname(struct ksyms *ksyms, uint64_t pc, char *buf, size_t n)
{
	const struct ksym *ksym = ksyms__map_addr(ksyms, pc);

	if (!ksym)
		return "Unknown";
	snprintf(buf, n, "%s+0x%lx", ksym->name, pc - ksym->addr);
	return buf;
}

static void print_acq_header(void)
{
	printf("\n                               Caller  Avg Wait    Count   Max Wait   Total Wait\n");
}

static void print_acq_stat(struct ksyms *ksyms, struct stack_stat *ss,
			   int nr_stack_entries)
{
	char buf[40];
	int i;

	printf("%37s %9llu %8llu %10llu %12llu\n",
	       symname(ksyms, ss->bt[0], buf, sizeof(buf)),
	       ss->ls.acq_total_time / ss->ls.acq_count,
	       ss->ls.acq_count,
	       ss->ls.acq_max_time,
	       ss->ls.acq_total_time);
	for (i = 1; i < nr_stack_entries; i++) {
		if (!ss->bt[i])
			break;
		printf("%37s\n", symname(ksyms, ss->bt[i], buf, sizeof(buf)));
	}
	if (nr_stack_entries > 1)
		printf("                              Max PID %llu, COMM %s\n",
		       ss->ls.acq_max_id >> 32,
		       ss->ls.acq_max_comm);
}

static void print_hld_header(void)
{
	printf("\n                               Caller  Avg Hold    Count   Max Hold   Total Hold\n");
}

static void print_hld_stat(struct ksyms *ksyms, struct stack_stat *ss,
			   int nr_stack_entries)
{
	char buf[40];
	int i;

	printf("%37s %9llu %8llu %10llu %12llu\n",
	       symname(ksyms, ss->bt[0], buf, sizeof(buf)),
	       ss->ls.hld_total_time / ss->ls.hld_count,
	       ss->ls.hld_count,
	       ss->ls.hld_max_time,
	       ss->ls.hld_total_time);
	for (i = 1; i < nr_stack_entries; i++) {
		if (!ss->bt[i])
			break;
		printf("%37s\n", symname(ksyms, ss->bt[i], buf, sizeof(buf)));
	}
	if (nr_stack_entries > 1)
		printf("                              Max PID %llu, COMM %s\n",
		       ss->ls.hld_max_id >> 32,
		       ss->ls.hld_max_comm);
}

static int print_stats(struct ksyms *ksyms, int stack_map, int stat_map)
{
	struct stack_stat **stats, *ss;
	size_t stat_idx = 0;
	size_t stats_sz = 1;
	uint32_t lookup_key = 0;
	uint32_t stack_id;
	int ret, i;

	stats = calloc(stats_sz, sizeof(void *));
	if (!stats) {
		warn("Out of memory\n");
		return -1;
	}

	while (bpf_map_get_next_key(stat_map, &lookup_key, &stack_id) == 0) {
		if (stat_idx == stats_sz) {
			stats_sz *= 2;
			stats = reallocarray(stats, stats_sz, sizeof(void *));
			if (!stats) {
				warn("Out of memory\n");
				return -1;
			}
		}
		ss = malloc(sizeof(struct stack_stat));
		if (!ss) {
			warn("Out of memory\n");
			return -1;
		}
		ss->stack_id = stack_id;
		if (env.reset) {
			ret = bpf_map_lookup_and_delete_elem(stat_map,
							     &stack_id,
							     &ss->ls);
			lookup_key = 0;
		} else {
			ret = bpf_map_lookup_elem(stat_map, &stack_id, &ss->ls);
			lookup_key = stack_id;
		}
		if (ret) {
			free(ss);
			continue;
		}
		if (bpf_map_lookup_elem(stack_map, &stack_id, &ss->bt)) {
			/* Can still report the results without a backtrace. */
			warn("failed to lookup stack_id %u\n", stack_id);
		}
		if (!caller_is_traced(ksyms, ss->bt[0])) {
			free(ss);
			continue;
		}
		stats[stat_idx++] = ss;
	}

	qsort(stats, stat_idx, sizeof(void*), sort_by_acq);
	for (i = 0; i < MIN(env.nr_locks, stat_idx); i++) {
		if (i == 0 || env.nr_stack_entries > 1)
			print_acq_header();
		print_acq_stat(ksyms, stats[i],
			       MIN(env.nr_stack_entries, PERF_MAX_STACK_DEPTH));
	}

	qsort(stats, stat_idx, sizeof(void*), sort_by_hld);
	for (i = 0; i < MIN(env.nr_locks, stat_idx); i++) {
		if (i == 0 || env.nr_stack_entries > 1)
			print_hld_header();
		print_hld_stat(ksyms, stats[i],
			       MIN(env.nr_stack_entries, PERF_MAX_STACK_DEPTH));
	}

	for (i = 0; i < stat_idx; i++)
		free(stats[i]);
	free(stats);

	return 0;
}

static void *get_lock_addr(struct ksyms *ksyms, const char *lock_name)
{
	const struct ksym *ksym = ksyms__get_symbol(ksyms, lock_name);

	return ksym ? (void*)ksym->addr : NULL;
}

static volatile bool exiting;

static void sig_hand(int signr)
{
	exiting = true;
}

static struct sigaction sigact = {.sa_handler = sig_hand};

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

int main(int argc, char **argv)
{
	static const struct argp argp = {
		.options = opts,
		.parser = parse_arg,
		.args_doc = args_doc,
		.doc = program_doc,
	};
	struct klockstat_bpf *obj = NULL;
	struct ksyms *ksyms = NULL;
	int i, err;
	struct tm *tm;
	char ts[32];
	time_t t;
	void *lock_addr = NULL;

	err = argp_parse(&argp, argc, argv, 0, NULL, &env);
	if (err)
		return err;

	sigaction(SIGINT, &sigact, 0);

	libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
	libbpf_set_print(libbpf_print_fn);

	ksyms = ksyms__load();
	if (!ksyms) {
		warn("failed to load kallsyms\n");
		err = 1;
		goto cleanup;
	}
	if (env.lock_name) {
		lock_addr = get_lock_addr(ksyms, env.lock_name);
		if (!lock_addr) {
			warn("failed to find lock %s\n", env.lock_name);
			err = 1;
			goto cleanup;
		}
	}

	obj = klockstat_bpf__open();
	if (!obj) {
		warn("failed to open BPF object\n");
		err = 1;
		goto cleanup;
	}

	obj->rodata->targ_tgid = env.pid;
	obj->rodata->targ_pid = env.tid;
	obj->rodata->targ_lock = lock_addr;

	if (fentry_can_attach("mutex_lock_nested", NULL)) {
		bpf_program__set_attach_target(obj->progs.mutex_lock, 0,
					       "mutex_lock_nested");
		bpf_program__set_attach_target(obj->progs.mutex_lock_exit, 0,
					       "mutex_lock_nested");
		bpf_program__set_attach_target(obj->progs.mutex_lock_interruptible, 0,
					       "mutex_lock_interruptible_nested");
		bpf_program__set_attach_target(obj->progs.mutex_lock_interruptible_exit, 0,
					       "mutex_lock_interruptible_nested");
		bpf_program__set_attach_target(obj->progs.mutex_lock_killable, 0,
					       "mutex_lock_killable_nested");
		bpf_program__set_attach_target(obj->progs.mutex_lock_killable_exit, 0,
					       "mutex_lock_killable_nested");
	}

	if (fentry_can_attach("down_read_nested", NULL)) {
		bpf_program__set_attach_target(obj->progs.down_read, 0,
					       "down_read_nested");
		bpf_program__set_attach_target(obj->progs.down_read_exit, 0,
					       "down_read_nested");
		bpf_program__set_attach_target(obj->progs.down_read_killable, 0,
					       "down_read_killable_nested");
		bpf_program__set_attach_target(obj->progs.down_read_killable_exit, 0,
					       "down_read_killable_nested");

		bpf_program__set_attach_target(obj->progs.down_write, 0,
					       "down_write_nested");
		bpf_program__set_attach_target(obj->progs.down_write_exit, 0,
					       "down_write_nested");
		bpf_program__set_attach_target(obj->progs.down_write_killable, 0,
					       "down_write_killable_nested");
		bpf_program__set_attach_target(obj->progs.down_write_killable_exit, 0,
					       "down_write_killable_nested");
	}

	err = klockstat_bpf__load(obj);
	if (err) {
		warn("failed to load BPF object\n");
		return 1;
	}
	err = klockstat_bpf__attach(obj);
	if (err) {
		warn("failed to attach BPF object\n");
		goto cleanup;
	}

	printf("Tracing mutex/sem lock events...  Hit Ctrl-C to end\n");

	for (i = 0; i < env.iterations && !exiting; i++) {
		sleep(env.interval);

		printf("\n");
		if (env.timestamp) {
			time(&t);
			tm = localtime(&t);
			strftime(ts, sizeof(ts), "%H:%M:%S", tm);
			printf("%-8s\n", ts);
		}

		if (print_stats(ksyms, bpf_map__fd(obj->maps.stack_map),
				bpf_map__fd(obj->maps.stat_map))) {
			warn("print_stats error, aborting.\n");
			break;
		}
	}

	printf("Exiting trace of mutex/sem locks\n");

cleanup:
	klockstat_bpf__destroy(obj);
	ksyms__free(ksyms);

	return err != 0;
}