[linux-2.6-block.git] / samples / bpf / offwaketime.bpf.c

/* Copyright (c) 2016 Facebook
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of version 2 of the GNU General Public
 * License as published by the Free Software Foundation.
 */
#include "vmlinux.h"
#include <linux/version.h>
#include <bpf/bpf_helpers.h>
#include <bpf/bpf_tracing.h>
#include <bpf/bpf_core_read.h>

#ifndef PERF_MAX_STACK_DEPTH
#define PERF_MAX_STACK_DEPTH         127
#endif

#define MINBLOCK_US	1
#define MAX_ENTRIES	10000

struct key_t {
	char waker[TASK_COMM_LEN];
	char target[TASK_COMM_LEN];
	u32 wret;
	u32 tret;
};

struct {
	__uint(type, BPF_MAP_TYPE_HASH);
	__type(key, struct key_t);
	__type(value, u64);
	__uint(max_entries, MAX_ENTRIES);
} counts SEC(".maps");

struct {
	__uint(type, BPF_MAP_TYPE_HASH);
	__type(key, u32);
	__type(value, u64);
	__uint(max_entries, MAX_ENTRIES);
} start SEC(".maps");

struct wokeby_t {
	char name[TASK_COMM_LEN];
	u32 ret;
};

struct {
	__uint(type, BPF_MAP_TYPE_HASH);
	__type(key, u32);
	__type(value, struct wokeby_t);
	__uint(max_entries, MAX_ENTRIES);
} wokeby SEC(".maps");

struct {
	__uint(type, BPF_MAP_TYPE_STACK_TRACE);
	__uint(key_size, sizeof(u32));
	__uint(value_size, PERF_MAX_STACK_DEPTH * sizeof(u64));
	__uint(max_entries, MAX_ENTRIES);
} stackmap SEC(".maps");

#define STACKID_FLAGS (0 | BPF_F_FAST_STACK_CMP)

SEC("kprobe/try_to_wake_up")
int waker(struct pt_regs *ctx)
{
	struct task_struct *p = (void *)PT_REGS_PARM1_CORE(ctx);
	u32 pid = BPF_CORE_READ(p, pid);
	struct wokeby_t woke;

	bpf_get_current_comm(&woke.name, sizeof(woke.name));
	woke.ret = bpf_get_stackid(ctx, &stackmap, STACKID_FLAGS);

	bpf_map_update_elem(&wokeby, &pid, &woke, BPF_ANY);
	return 0;
}

static inline int update_counts(void *ctx, u32 pid, u64 delta)
{
	struct wokeby_t *woke;
	u64 zero = 0, *val;
	struct key_t key;

	__builtin_memset(&key.waker, 0, sizeof(key.waker));
	bpf_get_current_comm(&key.target, sizeof(key.target));
	key.tret = bpf_get_stackid(ctx, &stackmap, STACKID_FLAGS);
	key.wret = 0;

	woke = bpf_map_lookup_elem(&wokeby, &pid);
	if (woke) {
		key.wret = woke->ret;
		__builtin_memcpy(&key.waker, woke->name, sizeof(key.waker));
		bpf_map_delete_elem(&wokeby, &pid);
	}

	val = bpf_map_lookup_elem(&counts, &key);
	if (!val) {
		bpf_map_update_elem(&counts, &key, &zero, BPF_NOEXIST);
		val = bpf_map_lookup_elem(&counts, &key);
		if (!val)
			return 0;
	}
	(*val) += delta;
	return 0;
}

#if 1
/* taken from /sys/kernel/tracing/events/sched/sched_switch/format */
SEC("tracepoint/sched/sched_switch")
int oncpu(struct trace_event_raw_sched_switch *ctx)
{
	/* record previous thread sleep time */
	u32 pid = ctx->prev_pid;
#else
SEC("kprobe.multi/finish_task_switch*")
int oncpu(struct pt_regs *ctx)
{
	struct task_struct *p = (void *)PT_REGS_PARM1_CORE(ctx);
	/* record previous thread sleep time */
	u32 pid = BPF_CORE_READ(p, pid);
#endif
	u64 delta, ts, *tsp;

	ts = bpf_ktime_get_ns();
	bpf_map_update_elem(&start, &pid, &ts, BPF_ANY);

	/* calculate current thread's delta time */
	pid = bpf_get_current_pid_tgid();
	tsp = bpf_map_lookup_elem(&start, &pid);
	if (!tsp)
		/* missed start or filtered */
		return 0;

	delta = bpf_ktime_get_ns() - *tsp;
	bpf_map_delete_elem(&start, &pid);
	delta = delta / 1000;
	if (delta < MINBLOCK_US)
		return 0;

	return update_counts(ctx, pid, delta);
}
char _license[] SEC("license") = "GPL";
u32 _version SEC("version") = LINUX_VERSION_CODE;
Commit	Line	Data
a6ffe7b9 AS	1	/* Copyright (c) 2016 Facebook
	2	*
	3	* This program is free software; you can redistribute it and/or
	4	* modify it under the terms of version 2 of the GNU General Public
	5	* License as published by the Free Software Foundation.
	6	*/
e7e6c774	7	#include "vmlinux.h"
a6ffe7b9	8	#include <linux/version.h>
03421a92 YS	9	#include <bpf/bpf_helpers.h>
03421a92 YS	10	#include <bpf/bpf_tracing.h>
11430421	11	#include <bpf/bpf_core_read.h>
a6ffe7b9	12
e7e6c774 DL	13	#ifndef PERF_MAX_STACK_DEPTH
	14	#define PERF_MAX_STACK_DEPTH 127
	15	#endif
	16
a6ffe7b9	17	#define MINBLOCK_US 1
d1bf7c4d	18	#define MAX_ENTRIES 10000
a6ffe7b9 AS	19
	20	struct key_t {
	21	char waker[TASK_COMM_LEN];
	22	char target[TASK_COMM_LEN];
	23	u32 wret;
	24	u32 tret;
	25	};
	26
f0c328f8 DL	27	struct {
	28	__uint(type, BPF_MAP_TYPE_HASH);
	29	__type(key, struct key_t);
	30	__type(value, u64);
d1bf7c4d	31	__uint(max_entries, MAX_ENTRIES);
f0c328f8	32	} counts SEC(".maps");
a6ffe7b9	33
f0c328f8 DL	34	struct {
	35	__uint(type, BPF_MAP_TYPE_HASH);
	36	__type(key, u32);
	37	__type(value, u64);
d1bf7c4d	38	__uint(max_entries, MAX_ENTRIES);
f0c328f8	39	} start SEC(".maps");
a6ffe7b9 AS	40
	41	struct wokeby_t {
	42	char name[TASK_COMM_LEN];
	43	u32 ret;
	44	};
	45
f0c328f8 DL	46	struct {
	47	__uint(type, BPF_MAP_TYPE_HASH);
	48	__type(key, u32);
	49	__type(value, struct wokeby_t);
d1bf7c4d	50	__uint(max_entries, MAX_ENTRIES);
f0c328f8	51	} wokeby SEC(".maps");
a6ffe7b9	52
f0c328f8 DL	53	struct {
	54	__uint(type, BPF_MAP_TYPE_STACK_TRACE);
	55	__uint(key_size, sizeof(u32));
	56	__uint(value_size, PERF_MAX_STACK_DEPTH * sizeof(u64));
d1bf7c4d	57	__uint(max_entries, MAX_ENTRIES);
f0c328f8	58	} stackmap SEC(".maps");
a6ffe7b9 AS	59
	60	#define STACKID_FLAGS (0 \| BPF_F_FAST_STACK_CMP)
	61
	62	SEC("kprobe/try_to_wake_up")
	63	int waker(struct pt_regs *ctx)
	64	{
11430421 DL	65	struct task_struct p = (void )PT_REGS_PARM1_CORE(ctx);
11430421 DL	66	u32 pid = BPF_CORE_READ(p, pid);
02413cab	67	struct wokeby_t woke;
a6ffe7b9 AS	68
	69	bpf_get_current_comm(&woke.name, sizeof(woke.name));
	70	woke.ret = bpf_get_stackid(ctx, &stackmap, STACKID_FLAGS);
	71
	72	bpf_map_update_elem(&wokeby, &pid, &woke, BPF_ANY);
	73	return 0;
	74	}
	75
3c9b1644	76	static inline int update_counts(void *ctx, u32 pid, u64 delta)
a6ffe7b9	77	{
a6ffe7b9 AS	78	struct wokeby_t *woke;
a6ffe7b9 AS	79	u64 zero = 0, *val;
02413cab	80	struct key_t key;
a6ffe7b9	81
02413cab	82	__builtin_memset(&key.waker, 0, sizeof(key.waker));
a6ffe7b9 AS	83	bpf_get_current_comm(&key.target, sizeof(key.target));
a6ffe7b9 AS	84	key.tret = bpf_get_stackid(ctx, &stackmap, STACKID_FLAGS);
02413cab	85	key.wret = 0;
a6ffe7b9 AS	86
	87	woke = bpf_map_lookup_elem(&wokeby, &pid);
	88	if (woke) {
	89	key.wret = woke->ret;
02413cab	90	__builtin_memcpy(&key.waker, woke->name, sizeof(key.waker));
a6ffe7b9 AS	91	bpf_map_delete_elem(&wokeby, &pid);
	92	}
	93
	94	val = bpf_map_lookup_elem(&counts, &key);
	95	if (!val) {
	96	bpf_map_update_elem(&counts, &key, &zero, BPF_NOEXIST);
	97	val = bpf_map_lookup_elem(&counts, &key);
	98	if (!val)
	99	return 0;
	100	}
	101	(*val) += delta;
	102	return 0;
	103	}
	104
3c9b1644	105	#if 1
27d7fdf0	106	/* taken from /sys/kernel/tracing/events/sched/sched_switch/format */
3c9b1644	107	SEC("tracepoint/sched/sched_switch")
e7e6c774	108	int oncpu(struct trace_event_raw_sched_switch *ctx)
3c9b1644 AS	109	{
	110	/* record previous thread sleep time */
	111	u32 pid = ctx->prev_pid;
	112	#else
02dabc24	113	SEC("kprobe.multi/finish_task_switch*")
a6ffe7b9 AS	114	int oncpu(struct pt_regs *ctx)
a6ffe7b9 AS	115	{
11430421	116	struct task_struct p = (void )PT_REGS_PARM1_CORE(ctx);
3c9b1644	117	/* record previous thread sleep time */
11430421	118	u32 pid = BPF_CORE_READ(p, pid);
3c9b1644	119	#endif
a6ffe7b9	120	u64 delta, ts, *tsp;
a6ffe7b9	121
a6ffe7b9 AS	122	ts = bpf_ktime_get_ns();
	123	bpf_map_update_elem(&start, &pid, &ts, BPF_ANY);
	124
	125	/* calculate current thread's delta time */
	126	pid = bpf_get_current_pid_tgid();
	127	tsp = bpf_map_lookup_elem(&start, &pid);
	128	if (!tsp)
	129	/* missed start or filtered */
	130	return 0;
	131
	132	delta = bpf_ktime_get_ns() - *tsp;
	133	bpf_map_delete_elem(&start, &pid);
	134	delta = delta / 1000;
	135	if (delta < MINBLOCK_US)
	136	return 0;
	137
	138	return update_counts(ctx, pid, delta);
	139	}
	140	char _license[] SEC("license") = "GPL";
	141	u32 _version SEC("version") = LINUX_VERSION_CODE;