[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>

#ifndef PERF_MAX_STACK_DEPTH
#define PERF_MAX_STACK_DEPTH         127
#endif

#define _(P)                                                                   \
	({                                                                     \
		typeof(P) val;                                                 \
		bpf_probe_read_kernel(&val, sizeof(val), &(P));                \
		val;                                                           \
	})

#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(ctx);
	struct wokeby_t woke;
	u32 pid;

	pid = _(p->pid);

	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/finish_task_switch")
int oncpu(struct pt_regs *ctx)
{
	struct task_struct *p = (void *) PT_REGS_PARM1(ctx);
	/* record previous thread sleep time */
	u32 pid = _(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>
a6ffe7b9	11
e7e6c774 DL	12	#ifndef PERF_MAX_STACK_DEPTH
	13	#define PERF_MAX_STACK_DEPTH 127
	14	#endif
	15
e4d9c232 IL	16	#define _(P) \
	17	({ \
	18	typeof(P) val; \
	19	bpf_probe_read_kernel(&val, sizeof(val), &(P)); \
	20	val; \
	21	})
a6ffe7b9 AS	22
a6ffe7b9 AS	23	#define MINBLOCK_US 1
d1bf7c4d	24	#define MAX_ENTRIES 10000
a6ffe7b9 AS	25
	26	struct key_t {
	27	char waker[TASK_COMM_LEN];
	28	char target[TASK_COMM_LEN];
	29	u32 wret;
	30	u32 tret;
	31	};
	32
f0c328f8 DL	33	struct {
	34	__uint(type, BPF_MAP_TYPE_HASH);
	35	__type(key, struct key_t);
	36	__type(value, u64);
d1bf7c4d	37	__uint(max_entries, MAX_ENTRIES);
f0c328f8	38	} counts SEC(".maps");
a6ffe7b9	39
f0c328f8 DL	40	struct {
	41	__uint(type, BPF_MAP_TYPE_HASH);
	42	__type(key, u32);
	43	__type(value, u64);
d1bf7c4d	44	__uint(max_entries, MAX_ENTRIES);
f0c328f8	45	} start SEC(".maps");
a6ffe7b9 AS	46
	47	struct wokeby_t {
	48	char name[TASK_COMM_LEN];
	49	u32 ret;
	50	};
	51
f0c328f8 DL	52	struct {
	53	__uint(type, BPF_MAP_TYPE_HASH);
	54	__type(key, u32);
	55	__type(value, struct wokeby_t);
d1bf7c4d	56	__uint(max_entries, MAX_ENTRIES);
f0c328f8	57	} wokeby SEC(".maps");
a6ffe7b9	58
f0c328f8 DL	59	struct {
	60	__uint(type, BPF_MAP_TYPE_STACK_TRACE);
	61	__uint(key_size, sizeof(u32));
	62	__uint(value_size, PERF_MAX_STACK_DEPTH * sizeof(u64));
d1bf7c4d	63	__uint(max_entries, MAX_ENTRIES);
f0c328f8	64	} stackmap SEC(".maps");
a6ffe7b9 AS	65
	66	#define STACKID_FLAGS (0 \| BPF_F_FAST_STACK_CMP)
	67
	68	SEC("kprobe/try_to_wake_up")
	69	int waker(struct pt_regs *ctx)
	70	{
	71	struct task_struct p = (void ) PT_REGS_PARM1(ctx);
02413cab	72	struct wokeby_t woke;
a6ffe7b9 AS	73	u32 pid;
	74
	75	pid = _(p->pid);
	76
	77	bpf_get_current_comm(&woke.name, sizeof(woke.name));
	78	woke.ret = bpf_get_stackid(ctx, &stackmap, STACKID_FLAGS);
	79
	80	bpf_map_update_elem(&wokeby, &pid, &woke, BPF_ANY);
	81	return 0;
	82	}
	83
3c9b1644	84	static inline int update_counts(void *ctx, u32 pid, u64 delta)
a6ffe7b9	85	{
a6ffe7b9 AS	86	struct wokeby_t *woke;
a6ffe7b9 AS	87	u64 zero = 0, *val;
02413cab	88	struct key_t key;
a6ffe7b9	89
02413cab	90	__builtin_memset(&key.waker, 0, sizeof(key.waker));
a6ffe7b9 AS	91	bpf_get_current_comm(&key.target, sizeof(key.target));
a6ffe7b9 AS	92	key.tret = bpf_get_stackid(ctx, &stackmap, STACKID_FLAGS);
02413cab	93	key.wret = 0;
a6ffe7b9 AS	94
	95	woke = bpf_map_lookup_elem(&wokeby, &pid);
	96	if (woke) {
	97	key.wret = woke->ret;
02413cab	98	__builtin_memcpy(&key.waker, woke->name, sizeof(key.waker));
a6ffe7b9 AS	99	bpf_map_delete_elem(&wokeby, &pid);
	100	}
	101
	102	val = bpf_map_lookup_elem(&counts, &key);
	103	if (!val) {
	104	bpf_map_update_elem(&counts, &key, &zero, BPF_NOEXIST);
	105	val = bpf_map_lookup_elem(&counts, &key);
	106	if (!val)
	107	return 0;
	108	}
	109	(*val) += delta;
	110	return 0;
	111	}
	112
3c9b1644	113	#if 1
27d7fdf0	114	/* taken from /sys/kernel/tracing/events/sched/sched_switch/format */
3c9b1644	115	SEC("tracepoint/sched/sched_switch")
e7e6c774	116	int oncpu(struct trace_event_raw_sched_switch *ctx)
3c9b1644 AS	117	{
	118	/* record previous thread sleep time */
	119	u32 pid = ctx->prev_pid;
	120	#else
a6ffe7b9 AS	121	SEC("kprobe/finish_task_switch")
	122	int oncpu(struct pt_regs *ctx)
	123	{
	124	struct task_struct p = (void ) PT_REGS_PARM1(ctx);
3c9b1644 AS	125	/* record previous thread sleep time */
	126	u32 pid = _(p->pid);
	127	#endif
a6ffe7b9	128	u64 delta, ts, *tsp;
a6ffe7b9	129
a6ffe7b9 AS	130	ts = bpf_ktime_get_ns();
	131	bpf_map_update_elem(&start, &pid, &ts, BPF_ANY);
	132
	133	/* calculate current thread's delta time */
	134	pid = bpf_get_current_pid_tgid();
	135	tsp = bpf_map_lookup_elem(&start, &pid);
	136	if (!tsp)
	137	/* missed start or filtered */
	138	return 0;
	139
	140	delta = bpf_ktime_get_ns() - *tsp;
	141	bpf_map_delete_elem(&start, &pid);
	142	delta = delta / 1000;
	143	if (delta < MINBLOCK_US)
	144	return 0;
	145
	146	return update_counts(ctx, pid, delta);
	147	}
	148	char _license[] SEC("license") = "GPL";
	149	u32 _version SEC("version") = LINUX_VERSION_CODE;