[lttng-modules.git] / instrumentation / events / lttng-module / sched.h

/* SPDX-License-Identifier: GPL-2.0-only */
#undef TRACE_SYSTEM
#define TRACE_SYSTEM sched

#if !defined(LTTNG_TRACE_SCHED_H) || defined(TRACE_HEADER_MULTI_READ)
#define LTTNG_TRACE_SCHED_H

#include <probes/lttng-tracepoint-event.h>
#include <linux/sched.h>
#include <linux/pid_namespace.h>
#include <linux/binfmts.h>
#include <linux/version.h>
#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,9,0))
#include <linux/sched/rt.h>
#endif

#define LTTNG_MAX_PID_NS_LEVEL 32

#ifndef _TRACE_SCHED_DEF_
#define _TRACE_SCHED_DEF_

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,15,0))

static inline long __trace_sched_switch_state(bool preempt, struct task_struct *p)
{
        unsigned int state;

#ifdef CONFIG_SCHED_DEBUG
        BUG_ON(p != current);
#endif /* CONFIG_SCHED_DEBUG */

        /*
         * Preemption ignores task state, therefore preempted tasks are always
         * RUNNING (we will not have dequeued if state != RUNNING).
         */
        if (preempt)
                return TASK_REPORT_MAX;

        /*
         * task_state_index() uses fls() and returns a value from 0-8 range.
         * Decrement it by 1 (except TASK_RUNNING state i.e 0) before using
         * it for left shift operation to get the correct task->state
         * mapping.
         */
        state = task_state_index(p);

        return state ? (1 << (state - 1)) : state;
}

#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(4,14,0))

static inline long __trace_sched_switch_state(bool preempt, struct task_struct *p)
{
        unsigned int state;

#ifdef CONFIG_SCHED_DEBUG
        BUG_ON(p != current);
#endif /* CONFIG_SCHED_DEBUG */

        /*
         * Preemption ignores task state, therefore preempted tasks are always
         * RUNNING (we will not have dequeued if state != RUNNING).
         */
        if (preempt)
                return TASK_REPORT_MAX;

        /*
         * __get_task_state() uses fls() and returns a value from 0-8 range.
         * Decrement it by 1 (except TASK_RUNNING state i.e 0) before using
         * it for left shift operation to get the correct task->state
         * mapping.
         */
        state = __get_task_state(p);

        return state ? (1 << (state - 1)) : state;
}

#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(4,4,0))

static inline long __trace_sched_switch_state(bool preempt, struct task_struct *p)
{
#ifdef CONFIG_SCHED_DEBUG
	BUG_ON(p != current);
#endif /* CONFIG_SCHED_DEBUG */
	/*
	 * Preemption ignores task state, therefore preempted tasks are always RUNNING
	 * (we will not have dequeued if state != RUNNING).
	 */
	return preempt ? TASK_RUNNING | TASK_STATE_MAX : p->state;
}

#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,19,0))

static inline long __trace_sched_switch_state(struct task_struct *p)
{
	long state = p->state;

#ifdef CONFIG_PREEMPT
#ifdef CONFIG_SCHED_DEBUG
	BUG_ON(p != current);
#endif /* CONFIG_SCHED_DEBUG */
	/*
	 * For all intents and purposes a preempted task is a running task.
	 */
	if (preempt_count() & PREEMPT_ACTIVE)
		state = TASK_RUNNING | TASK_STATE_MAX;
#endif	/* CONFIG_PREEMPT */

	return state;
}

#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,13,0))

static inline long __trace_sched_switch_state(struct task_struct *p)
{
	long state = p->state;

#ifdef CONFIG_PREEMPT
	/*
	 * For all intents and purposes a preempted task is a running task.
	 */
	if (task_preempt_count(p) & PREEMPT_ACTIVE)
		state = TASK_RUNNING | TASK_STATE_MAX;
#endif

	return state;
}

#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0))

static inline long __trace_sched_switch_state(struct task_struct *p)
{
	long state = p->state;

#ifdef CONFIG_PREEMPT
	/*
	 * For all intents and purposes a preempted task is a running task.
	 */
	if (task_thread_info(p)->preempt_count & PREEMPT_ACTIVE)
		state = TASK_RUNNING | TASK_STATE_MAX;
#endif

	return state;
}

#else

static inline long __trace_sched_switch_state(struct task_struct *p)
{
	long state = p->state;

#ifdef CONFIG_PREEMPT
	/*
	 * For all intents and purposes a preempted task is a running task.
	 */
	if (task_thread_info(p)->preempt_count & PREEMPT_ACTIVE)
		state = TASK_RUNNING;
#endif

	return state;
}

#endif

#endif /* _TRACE_SCHED_DEF_ */

/*
 * Enumeration of the task state bitmask.
 * Only bit flags are enumerated here, not composition of states.
 */
LTTNG_TRACEPOINT_ENUM(task_state,
	TP_ENUM_VALUES(
		ctf_enum_value("TASK_RUNNING", TASK_RUNNING)
		ctf_enum_value("TASK_INTERRUPTIBLE", TASK_INTERRUPTIBLE)
		ctf_enum_value("TASK_UNINTERRUPTIBLE", TASK_UNINTERRUPTIBLE)
		ctf_enum_value("TASK_STOPPED", __TASK_STOPPED)
		ctf_enum_value("TASK_TRACED", __TASK_TRACED)
		ctf_enum_value("EXIT_DEAD", EXIT_DEAD)
		ctf_enum_value("EXIT_ZOMBIE", EXIT_ZOMBIE)

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,9,0))
		ctf_enum_value("TASK_PARKED", TASK_PARKED)
#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,9,0)) */

		ctf_enum_value("TASK_DEAD", TASK_DEAD)
		ctf_enum_value("TASK_WAKEKILL", TASK_WAKEKILL)
		ctf_enum_value("TASK_WAKING", TASK_WAKING)

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,2,0))
		ctf_enum_value("TASK_NOLOAD", TASK_NOLOAD)
#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,2,0)) */

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,8,0))
		ctf_enum_value("TASK_NEW", TASK_NEW)
#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,8,0)) */

		ctf_enum_value("TASK_STATE_MAX", TASK_STATE_MAX)
	)
)

/*
 * Tracepoint for calling kthread_stop, performed to end a kthread:
 */
LTTNG_TRACEPOINT_EVENT(sched_kthread_stop,

	TP_PROTO(struct task_struct *t),

	TP_ARGS(t),

	TP_FIELDS(
		ctf_array_text(char, comm, t->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, t->pid)
	)
)

/*
 * Tracepoint for the return value of the kthread stopping:
 */
LTTNG_TRACEPOINT_EVENT(sched_kthread_stop_ret,

	TP_PROTO(int ret),

	TP_ARGS(ret),

	TP_FIELDS(
		ctf_integer(int, ret, ret)
	)
)

/*
 * Tracepoint for waking up a task:
 */
#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0) || \
	LTTNG_RT_KERNEL_RANGE(4,1,10,11, 4,2,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,18,27,26, 3,19,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,14,61,63, 3,15,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,12,54,73, 3,13,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,10,97,106, 3,11,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,4,110,139, 3,5,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,2,77,111, 3,3,0,0))
LTTNG_TRACEPOINT_EVENT_CLASS(sched_wakeup_template,

	TP_PROTO(struct task_struct *p),

	TP_ARGS(p),

	TP_FIELDS(
		ctf_array_text(char, comm, p->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, p->pid)
		ctf_integer(int, prio, p->prio - MAX_RT_PRIO)
		ctf_integer(int, target_cpu, task_cpu(p))
	)
)
#else /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0)) */
LTTNG_TRACEPOINT_EVENT_CLASS(sched_wakeup_template,

	TP_PROTO(struct task_struct *p, int success),

	TP_ARGS(p, success),

	TP_FIELDS(
		ctf_array_text(char, comm, p->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, p->pid)
		ctf_integer(int, prio, p->prio - MAX_RT_PRIO)
		ctf_integer(int, success, success)
		ctf_integer(int, target_cpu, task_cpu(p))
	)
)
#endif /* #else #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0)) */

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0) || \
	LTTNG_RT_KERNEL_RANGE(4,1,10,11, 4,2,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,18,27,26, 3,19,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,14,61,63, 3,15,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,12,54,73, 3,13,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,10,97,106, 3,11,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,4,110,139, 3,5,0,0) || \
	LTTNG_RT_KERNEL_RANGE(3,2,77,111, 3,3,0,0))

/*
 * Tracepoint called when waking a task; this tracepoint is guaranteed to be
 * called from the waking context.
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_waking,
	     TP_PROTO(struct task_struct *p),
	     TP_ARGS(p))

/*
 * Tracepoint called when the task is actually woken; p->state == TASK_RUNNNG.
 * It it not always called from the waking context.
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup,
	     TP_PROTO(struct task_struct *p),
	     TP_ARGS(p))

/*
 * Tracepoint for waking up a new task:
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup_new,
	     TP_PROTO(struct task_struct *p),
	     TP_ARGS(p))

#else

LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup,
	     TP_PROTO(struct task_struct *p, int success),
	     TP_ARGS(p, success))

/*
 * Tracepoint for waking up a new task:
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup_new,
	     TP_PROTO(struct task_struct *p, int success),
	     TP_ARGS(p, success))

#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0)) */

/*
 * Tracepoint for task switches, performed by the scheduler:
 */
LTTNG_TRACEPOINT_EVENT(sched_switch,

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,4,0))
	TP_PROTO(bool preempt,
		 struct task_struct *prev,
		 struct task_struct *next),

	TP_ARGS(preempt, prev, next),
#else
	TP_PROTO(struct task_struct *prev,
		 struct task_struct *next),

	TP_ARGS(prev, next),
#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,4,0)) */

	TP_FIELDS(
		ctf_array_text(char, prev_comm,	prev->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, prev_tid, prev->pid)
		ctf_integer(int, prev_prio, prev->prio - MAX_RT_PRIO)
#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,4,0))
		ctf_enum(task_state, long, prev_state, __trace_sched_switch_state(preempt, prev))
#else
		ctf_enum(task_state, long, prev_state, __trace_sched_switch_state(prev))
#endif
		ctf_array_text(char, next_comm, next->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, next_tid, next->pid)
		ctf_integer(int, next_prio, next->prio - MAX_RT_PRIO)
	)
)

/*
 * Tracepoint for a task being migrated:
 */
LTTNG_TRACEPOINT_EVENT(sched_migrate_task,

	TP_PROTO(struct task_struct *p, int dest_cpu),

	TP_ARGS(p, dest_cpu),

	TP_FIELDS(
		ctf_array_text(char, comm, p->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, p->pid)
		ctf_integer(int, prio, p->prio - MAX_RT_PRIO)
		ctf_integer(int, orig_cpu, task_cpu(p))
		ctf_integer(int, dest_cpu, dest_cpu)
	)
)

LTTNG_TRACEPOINT_EVENT_CLASS(sched_process_template,

	TP_PROTO(struct task_struct *p),

	TP_ARGS(p),

	TP_FIELDS(
		ctf_array_text(char, comm, p->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, p->pid)
		ctf_integer(int, prio, p->prio - MAX_RT_PRIO)
	)
)

/*
 * Tracepoint for freeing a task:
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_process_template, sched_process_free,
	     TP_PROTO(struct task_struct *p),
	     TP_ARGS(p))


/*
 * Tracepoint for a task exiting:
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_process_template, sched_process_exit,
	     TP_PROTO(struct task_struct *p),
	     TP_ARGS(p))

/*
 * Tracepoint for waiting on task to unschedule:
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_process_template, sched_wait_task,
	TP_PROTO(struct task_struct *p),
	TP_ARGS(p))

/*
 * Tracepoint for a waiting task:
 */
LTTNG_TRACEPOINT_EVENT(sched_process_wait,

	TP_PROTO(struct pid *pid),

	TP_ARGS(pid),

	TP_FIELDS(
		ctf_array_text(char, comm, current->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, pid_nr(pid))
		ctf_integer(int, prio, current->prio - MAX_RT_PRIO)
	)
)

/*
 * Tracepoint for do_fork.
 * Saving both TID and PID information, especially for the child, allows
 * trace analyzers to distinguish between creation of a new process and
 * creation of a new thread. Newly created processes will have child_tid
 * == child_pid, while creation of a thread yields to child_tid !=
 * child_pid.
 */
LTTNG_TRACEPOINT_EVENT_CODE(sched_process_fork,

	TP_PROTO(struct task_struct *parent, struct task_struct *child),

	TP_ARGS(parent, child),

	TP_locvar(
		pid_t vtids[LTTNG_MAX_PID_NS_LEVEL];
		unsigned int ns_level;
	),

	TP_code_pre(
		if (child) {
			struct pid *child_pid;
			unsigned int i;

			child_pid = task_pid(child);
			tp_locvar->ns_level =
				min_t(unsigned int, child_pid->level + 1,
					LTTNG_MAX_PID_NS_LEVEL);
			for (i = 0; i < tp_locvar->ns_level; i++)
				tp_locvar->vtids[i] = child_pid->numbers[i].nr;
		}
	),

	TP_FIELDS(
		ctf_array_text(char, parent_comm, parent->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, parent_tid, parent->pid)
		ctf_integer(pid_t, parent_pid, parent->tgid)
#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,8,0))
		ctf_integer(unsigned int, parent_ns_inum,
			({
				unsigned int parent_ns_inum = 0;

				if (parent) {
					struct pid_namespace *pid_ns;

					pid_ns = task_active_pid_ns(parent);
					if (pid_ns)
						parent_ns_inum =
							pid_ns->ns.inum;
				}
				parent_ns_inum;
			}))
#endif
		ctf_array_text(char, child_comm, child->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, child_tid, child->pid)
		ctf_sequence(pid_t, vtids, tp_locvar->vtids, u8, tp_locvar->ns_level)
		ctf_integer(pid_t, child_pid, child->tgid)
#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,8,0))
		ctf_integer(unsigned int, child_ns_inum,
			({
				unsigned int child_ns_inum = 0;

				if (child) {
					struct pid_namespace *pid_ns;

					pid_ns = task_active_pid_ns(child);
					if (pid_ns)
						child_ns_inum =
							pid_ns->ns.inum;
				}
				child_ns_inum;
			}))
#endif
	),

	TP_code_post()
)

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,4,0))
/*
 * Tracepoint for exec:
 */
LTTNG_TRACEPOINT_EVENT(sched_process_exec,

	TP_PROTO(struct task_struct *p, pid_t old_pid,
		 struct linux_binprm *bprm),

	TP_ARGS(p, old_pid, bprm),

	TP_FIELDS(
		ctf_string(filename, bprm->filename)
		ctf_integer(pid_t, tid, p->pid)
		ctf_integer(pid_t, old_tid, old_pid)
	)
)
#endif

/*
 * XXX the below sched_stat tracepoints only apply to SCHED_OTHER/BATCH/IDLE
 *     adding sched_stat support to SCHED_FIFO/RR would be welcome.
 */
LTTNG_TRACEPOINT_EVENT_CLASS(sched_stat_template,

	TP_PROTO(struct task_struct *tsk, u64 delay),

	TP_ARGS(tsk, delay),

	TP_FIELDS(
		ctf_array_text(char, comm, tsk->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, tsk->pid)
		ctf_integer(u64, delay, delay)
	)
)


/*
 * Tracepoint for accounting wait time (time the task is runnable
 * but not actually running due to scheduler contention).
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_stat_template, sched_stat_wait,
	     TP_PROTO(struct task_struct *tsk, u64 delay),
	     TP_ARGS(tsk, delay))

/*
 * Tracepoint for accounting sleep time (time the task is not runnable,
 * including iowait, see below).
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_stat_template, sched_stat_sleep,
	     TP_PROTO(struct task_struct *tsk, u64 delay),
	     TP_ARGS(tsk, delay))

/*
 * Tracepoint for accounting iowait time (time the task is not runnable
 * due to waiting on IO to complete).
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_stat_template, sched_stat_iowait,
	     TP_PROTO(struct task_struct *tsk, u64 delay),
	     TP_ARGS(tsk, delay))

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,3,0))
/*
 * Tracepoint for accounting blocked time (time the task is in uninterruptible).
 */
LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_stat_template, sched_stat_blocked,
	     TP_PROTO(struct task_struct *tsk, u64 delay),
	     TP_ARGS(tsk, delay))
#endif

/*
 * Tracepoint for accounting runtime (time the task is executing
 * on a CPU).
 */
LTTNG_TRACEPOINT_EVENT(sched_stat_runtime,

	TP_PROTO(struct task_struct *tsk, u64 runtime, u64 vruntime),

	TP_ARGS(tsk, runtime, vruntime),

	TP_FIELDS(
		ctf_array_text(char, comm, tsk->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, tsk->pid)
		ctf_integer(u64, runtime, runtime)
		ctf_integer(u64, vruntime, vruntime)
	)
)

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,12,0) || \
	LTTNG_RT_KERNEL_RANGE(4,9,27,18, 4,10,0,0) || \
	LTTNG_RT_KERNEL_RANGE(4,11,5,1, 4,12,0,0))
/*
 * Tracepoint for showing priority inheritance modifying a tasks
 * priority.
 */
LTTNG_TRACEPOINT_EVENT(sched_pi_setprio,

	TP_PROTO(struct task_struct *tsk, struct task_struct *pi_task),

	TP_ARGS(tsk, pi_task),

	TP_FIELDS(
		ctf_array_text(char, comm, tsk->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, tsk->pid)
		ctf_integer(int, oldprio, tsk->prio - MAX_RT_PRIO)
		ctf_integer(int, newprio, pi_task ? pi_task->prio - MAX_RT_PRIO : tsk->prio - MAX_RT_PRIO)
	)
)
#else
/*
 * Tracepoint for showing priority inheritance modifying a tasks
 * priority.
 */
LTTNG_TRACEPOINT_EVENT(sched_pi_setprio,

	TP_PROTO(struct task_struct *tsk, int newprio),

	TP_ARGS(tsk, newprio),

	TP_FIELDS(
		ctf_array_text(char, comm, tsk->comm, TASK_COMM_LEN)
		ctf_integer(pid_t, tid, tsk->pid)
		ctf_integer(int, oldprio, tsk->prio - MAX_RT_PRIO)
		ctf_integer(int, newprio, newprio - MAX_RT_PRIO)
	)
)
#endif

#endif /* LTTNG_TRACE_SCHED_H */

/* This part must be outside protection */
#include <probes/define_trace.h>
Commit	Line	Data
	1	/* SPDX-License-Identifier: GPL-2.0-only */
	2	#undef TRACE_SYSTEM
	3	#define TRACE_SYSTEM sched
	4
	5	#if !defined(LTTNG_TRACE_SCHED_H) \|\| defined(TRACE_HEADER_MULTI_READ)
	6	#define LTTNG_TRACE_SCHED_H
	7
	8	#include <probes/lttng-tracepoint-event.h>
	9	#include <linux/sched.h>
	10	#include <linux/pid_namespace.h>
	11	#include <linux/binfmts.h>
	12	#include <linux/version.h>
	13	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,9,0))
	14	#include <linux/sched/rt.h>
	15	#endif
	16
	17	#define LTTNG_MAX_PID_NS_LEVEL 32
	18
	19	#ifndef _TRACE_SCHED_DEF_
	20	#define _TRACE_SCHED_DEF_
	21
	22	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,15,0))
	23
	24	static inline long __trace_sched_switch_state(bool preempt, struct task_struct *p)
	25	{
	26	unsigned int state;
	27
	28	#ifdef CONFIG_SCHED_DEBUG
	29	BUG_ON(p != current);
	30	#endif /* CONFIG_SCHED_DEBUG */
	31
	32	/*
	33	* Preemption ignores task state, therefore preempted tasks are always
	34	* RUNNING (we will not have dequeued if state != RUNNING).
	35	*/
	36	if (preempt)
	37	return TASK_REPORT_MAX;
	38
	39	/*
	40	* task_state_index() uses fls() and returns a value from 0-8 range.
	41	* Decrement it by 1 (except TASK_RUNNING state i.e 0) before using
	42	* it for left shift operation to get the correct task->state
	43	* mapping.
	44	*/
	45	state = task_state_index(p);
	46
	47	return state ? (1 << (state - 1)) : state;
	48	}
	49
	50	#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(4,14,0))
	51
	52	static inline long __trace_sched_switch_state(bool preempt, struct task_struct *p)
	53	{
	54	unsigned int state;
	55
	56	#ifdef CONFIG_SCHED_DEBUG
	57	BUG_ON(p != current);
	58	#endif /* CONFIG_SCHED_DEBUG */
	59
	60	/*
	61	* Preemption ignores task state, therefore preempted tasks are always
	62	* RUNNING (we will not have dequeued if state != RUNNING).
	63	*/
	64	if (preempt)
	65	return TASK_REPORT_MAX;
	66
	67	/*
	68	* __get_task_state() uses fls() and returns a value from 0-8 range.
	69	* Decrement it by 1 (except TASK_RUNNING state i.e 0) before using
	70	* it for left shift operation to get the correct task->state
	71	* mapping.
	72	*/
	73	state = __get_task_state(p);
	74
	75	return state ? (1 << (state - 1)) : state;
	76	}
	77
	78	#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(4,4,0))
	79
	80	static inline long __trace_sched_switch_state(bool preempt, struct task_struct *p)
	81	{
	82	#ifdef CONFIG_SCHED_DEBUG
	83	BUG_ON(p != current);
	84	#endif /* CONFIG_SCHED_DEBUG */
	85	/*
	86	* Preemption ignores task state, therefore preempted tasks are always RUNNING
	87	* (we will not have dequeued if state != RUNNING).
	88	*/
	89	return preempt ? TASK_RUNNING \| TASK_STATE_MAX : p->state;
	90	}
	91
	92	#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,19,0))
	93
	94	static inline long __trace_sched_switch_state(struct task_struct *p)
	95	{
	96	long state = p->state;
	97
	98	#ifdef CONFIG_PREEMPT
	99	#ifdef CONFIG_SCHED_DEBUG
	100	BUG_ON(p != current);
	101	#endif /* CONFIG_SCHED_DEBUG */
	102	/*
	103	* For all intents and purposes a preempted task is a running task.
	104	*/
	105	if (preempt_count() & PREEMPT_ACTIVE)
	106	state = TASK_RUNNING \| TASK_STATE_MAX;
	107	#endif /* CONFIG_PREEMPT */
	108
	109	return state;
	110	}
	111
	112	#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,13,0))
	113
	114	static inline long __trace_sched_switch_state(struct task_struct *p)
	115	{
	116	long state = p->state;
	117
	118	#ifdef CONFIG_PREEMPT
	119	/*
	120	* For all intents and purposes a preempted task is a running task.
	121	*/
	122	if (task_preempt_count(p) & PREEMPT_ACTIVE)
	123	state = TASK_RUNNING \| TASK_STATE_MAX;
	124	#endif
	125
	126	return state;
	127	}
	128
	129	#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0))
	130
	131	static inline long __trace_sched_switch_state(struct task_struct *p)
	132	{
	133	long state = p->state;
	134
	135	#ifdef CONFIG_PREEMPT
	136	/*
	137	* For all intents and purposes a preempted task is a running task.
	138	*/
	139	if (task_thread_info(p)->preempt_count & PREEMPT_ACTIVE)
	140	state = TASK_RUNNING \| TASK_STATE_MAX;
	141	#endif
	142
	143	return state;
	144	}
	145
	146	#else
	147
	148	static inline long __trace_sched_switch_state(struct task_struct *p)
	149	{
	150	long state = p->state;
	151
	152	#ifdef CONFIG_PREEMPT
	153	/*
	154	* For all intents and purposes a preempted task is a running task.
	155	*/
	156	if (task_thread_info(p)->preempt_count & PREEMPT_ACTIVE)
	157	state = TASK_RUNNING;
	158	#endif
	159
	160	return state;
	161	}
	162
	163	#endif
	164
	165	#endif /* _TRACE_SCHED_DEF_ */
	166
	167	/*
	168	* Enumeration of the task state bitmask.
	169	* Only bit flags are enumerated here, not composition of states.
	170	*/
	171	LTTNG_TRACEPOINT_ENUM(task_state,
	172	TP_ENUM_VALUES(
	173	ctf_enum_value("TASK_RUNNING", TASK_RUNNING)
	174	ctf_enum_value("TASK_INTERRUPTIBLE", TASK_INTERRUPTIBLE)
	175	ctf_enum_value("TASK_UNINTERRUPTIBLE", TASK_UNINTERRUPTIBLE)
	176	ctf_enum_value("TASK_STOPPED", __TASK_STOPPED)
	177	ctf_enum_value("TASK_TRACED", __TASK_TRACED)
	178	ctf_enum_value("EXIT_DEAD", EXIT_DEAD)
	179	ctf_enum_value("EXIT_ZOMBIE", EXIT_ZOMBIE)
	180
	181	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,9,0))
	182	ctf_enum_value("TASK_PARKED", TASK_PARKED)
	183	#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,9,0)) */
	184
	185	ctf_enum_value("TASK_DEAD", TASK_DEAD)
	186	ctf_enum_value("TASK_WAKEKILL", TASK_WAKEKILL)
	187	ctf_enum_value("TASK_WAKING", TASK_WAKING)
	188
	189	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,2,0))
	190	ctf_enum_value("TASK_NOLOAD", TASK_NOLOAD)
	191	#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,2,0)) */
	192
	193	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,8,0))
	194	ctf_enum_value("TASK_NEW", TASK_NEW)
	195	#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,8,0)) */
	196
	197	ctf_enum_value("TASK_STATE_MAX", TASK_STATE_MAX)
	198	)
	199	)
	200
	201	/*
	202	* Tracepoint for calling kthread_stop, performed to end a kthread:
	203	*/
	204	LTTNG_TRACEPOINT_EVENT(sched_kthread_stop,
	205
	206	TP_PROTO(struct task_struct *t),
	207
	208	TP_ARGS(t),
	209
	210	TP_FIELDS(
	211	ctf_array_text(char, comm, t->comm, TASK_COMM_LEN)
	212	ctf_integer(pid_t, tid, t->pid)
	213	)
	214	)
	215
	216	/*
	217	* Tracepoint for the return value of the kthread stopping:
	218	*/
	219	LTTNG_TRACEPOINT_EVENT(sched_kthread_stop_ret,
	220
	221	TP_PROTO(int ret),
	222
	223	TP_ARGS(ret),
	224
	225	TP_FIELDS(
	226	ctf_integer(int, ret, ret)
	227	)
	228	)
	229
	230	/*
	231	* Tracepoint for waking up a task:
	232	*/
	233	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0) \|\| \
	234	LTTNG_RT_KERNEL_RANGE(4,1,10,11, 4,2,0,0) \|\| \
	235	LTTNG_RT_KERNEL_RANGE(3,18,27,26, 3,19,0,0) \|\| \
	236	LTTNG_RT_KERNEL_RANGE(3,14,61,63, 3,15,0,0) \|\| \
	237	LTTNG_RT_KERNEL_RANGE(3,12,54,73, 3,13,0,0) \|\| \
	238	LTTNG_RT_KERNEL_RANGE(3,10,97,106, 3,11,0,0) \|\| \
	239	LTTNG_RT_KERNEL_RANGE(3,4,110,139, 3,5,0,0) \|\| \
	240	LTTNG_RT_KERNEL_RANGE(3,2,77,111, 3,3,0,0))
	241	LTTNG_TRACEPOINT_EVENT_CLASS(sched_wakeup_template,
	242
	243	TP_PROTO(struct task_struct *p),
	244
	245	TP_ARGS(p),
	246
	247	TP_FIELDS(
	248	ctf_array_text(char, comm, p->comm, TASK_COMM_LEN)
	249	ctf_integer(pid_t, tid, p->pid)
	250	ctf_integer(int, prio, p->prio - MAX_RT_PRIO)
	251	ctf_integer(int, target_cpu, task_cpu(p))
	252	)
	253	)
	254	#else /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0)) */
	255	LTTNG_TRACEPOINT_EVENT_CLASS(sched_wakeup_template,
	256
	257	TP_PROTO(struct task_struct *p, int success),
	258
	259	TP_ARGS(p, success),
	260
	261	TP_FIELDS(
	262	ctf_array_text(char, comm, p->comm, TASK_COMM_LEN)
	263	ctf_integer(pid_t, tid, p->pid)
	264	ctf_integer(int, prio, p->prio - MAX_RT_PRIO)
	265	ctf_integer(int, success, success)
	266	ctf_integer(int, target_cpu, task_cpu(p))
	267	)
	268	)
	269	#endif /* #else #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0)) */
	270
	271	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0) \|\| \
	272	LTTNG_RT_KERNEL_RANGE(4,1,10,11, 4,2,0,0) \|\| \
	273	LTTNG_RT_KERNEL_RANGE(3,18,27,26, 3,19,0,0) \|\| \
	274	LTTNG_RT_KERNEL_RANGE(3,14,61,63, 3,15,0,0) \|\| \
	275	LTTNG_RT_KERNEL_RANGE(3,12,54,73, 3,13,0,0) \|\| \
	276	LTTNG_RT_KERNEL_RANGE(3,10,97,106, 3,11,0,0) \|\| \
	277	LTTNG_RT_KERNEL_RANGE(3,4,110,139, 3,5,0,0) \|\| \
	278	LTTNG_RT_KERNEL_RANGE(3,2,77,111, 3,3,0,0))
	279
	280	/*
	281	* Tracepoint called when waking a task; this tracepoint is guaranteed to be
	282	* called from the waking context.
	283	*/
	284	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_waking,
	285	TP_PROTO(struct task_struct *p),
	286	TP_ARGS(p))
	287
	288	/*
	289	* Tracepoint called when the task is actually woken; p->state == TASK_RUNNNG.
	290	* It it not always called from the waking context.
	291	*/
	292	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup,
	293	TP_PROTO(struct task_struct *p),
	294	TP_ARGS(p))
	295
	296	/*
	297	* Tracepoint for waking up a new task:
	298	*/
	299	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup_new,
	300	TP_PROTO(struct task_struct *p),
	301	TP_ARGS(p))
	302
	303	#else
	304
	305	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup,
	306	TP_PROTO(struct task_struct *p, int success),
	307	TP_ARGS(p, success))
	308
	309	/*
	310	* Tracepoint for waking up a new task:
	311	*/
	312	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup_new,
	313	TP_PROTO(struct task_struct *p, int success),
	314	TP_ARGS(p, success))
	315
	316	#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,3,0)) */
	317
	318	/*
	319	* Tracepoint for task switches, performed by the scheduler:
	320	*/
	321	LTTNG_TRACEPOINT_EVENT(sched_switch,
	322
	323	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,4,0))
	324	TP_PROTO(bool preempt,
	325	struct task_struct *prev,
	326	struct task_struct *next),
	327
	328	TP_ARGS(preempt, prev, next),
	329	#else
	330	TP_PROTO(struct task_struct *prev,
	331	struct task_struct *next),
	332
	333	TP_ARGS(prev, next),
	334	#endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,4,0)) */
	335
	336	TP_FIELDS(
	337	ctf_array_text(char, prev_comm, prev->comm, TASK_COMM_LEN)
	338	ctf_integer(pid_t, prev_tid, prev->pid)
	339	ctf_integer(int, prev_prio, prev->prio - MAX_RT_PRIO)
	340	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,4,0))
	341	ctf_enum(task_state, long, prev_state, __trace_sched_switch_state(preempt, prev))
	342	#else
	343	ctf_enum(task_state, long, prev_state, __trace_sched_switch_state(prev))
	344	#endif
	345	ctf_array_text(char, next_comm, next->comm, TASK_COMM_LEN)
	346	ctf_integer(pid_t, next_tid, next->pid)
	347	ctf_integer(int, next_prio, next->prio - MAX_RT_PRIO)
	348	)
	349	)
	350
	351	/*
	352	* Tracepoint for a task being migrated:
	353	*/
	354	LTTNG_TRACEPOINT_EVENT(sched_migrate_task,
	355
	356	TP_PROTO(struct task_struct *p, int dest_cpu),
	357
	358	TP_ARGS(p, dest_cpu),
	359
	360	TP_FIELDS(
	361	ctf_array_text(char, comm, p->comm, TASK_COMM_LEN)
	362	ctf_integer(pid_t, tid, p->pid)
	363	ctf_integer(int, prio, p->prio - MAX_RT_PRIO)
	364	ctf_integer(int, orig_cpu, task_cpu(p))
	365	ctf_integer(int, dest_cpu, dest_cpu)
	366	)
	367	)
	368
	369	LTTNG_TRACEPOINT_EVENT_CLASS(sched_process_template,
	370
	371	TP_PROTO(struct task_struct *p),
	372
	373	TP_ARGS(p),
	374
	375	TP_FIELDS(
	376	ctf_array_text(char, comm, p->comm, TASK_COMM_LEN)
	377	ctf_integer(pid_t, tid, p->pid)
	378	ctf_integer(int, prio, p->prio - MAX_RT_PRIO)
	379	)
	380	)
	381
	382	/*
	383	* Tracepoint for freeing a task:
	384	*/
	385	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_process_template, sched_process_free,
	386	TP_PROTO(struct task_struct *p),
	387	TP_ARGS(p))
	388
	389
	390	/*
	391	* Tracepoint for a task exiting:
	392	*/
	393	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_process_template, sched_process_exit,
	394	TP_PROTO(struct task_struct *p),
	395	TP_ARGS(p))
	396
	397	/*
	398	* Tracepoint for waiting on task to unschedule:
	399	*/
	400	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_process_template, sched_wait_task,
	401	TP_PROTO(struct task_struct *p),
	402	TP_ARGS(p))
	403
	404	/*
	405	* Tracepoint for a waiting task:
	406	*/
	407	LTTNG_TRACEPOINT_EVENT(sched_process_wait,
	408
	409	TP_PROTO(struct pid *pid),
	410
	411	TP_ARGS(pid),
	412
	413	TP_FIELDS(
	414	ctf_array_text(char, comm, current->comm, TASK_COMM_LEN)
	415	ctf_integer(pid_t, tid, pid_nr(pid))
	416	ctf_integer(int, prio, current->prio - MAX_RT_PRIO)
	417	)
	418	)
	419
	420	/*
	421	* Tracepoint for do_fork.
	422	* Saving both TID and PID information, especially for the child, allows
	423	* trace analyzers to distinguish between creation of a new process and
	424	* creation of a new thread. Newly created processes will have child_tid
	425	* == child_pid, while creation of a thread yields to child_tid !=
	426	* child_pid.
	427	*/
	428	LTTNG_TRACEPOINT_EVENT_CODE(sched_process_fork,
	429
	430	TP_PROTO(struct task_struct parent, struct task_struct child),
	431
	432	TP_ARGS(parent, child),
	433
	434	TP_locvar(
	435	pid_t vtids[LTTNG_MAX_PID_NS_LEVEL];
	436	unsigned int ns_level;
	437	),
	438
	439	TP_code_pre(
	440	if (child) {
	441	struct pid *child_pid;
	442	unsigned int i;
	443
	444	child_pid = task_pid(child);
	445	tp_locvar->ns_level =
	446	min_t(unsigned int, child_pid->level + 1,
	447	LTTNG_MAX_PID_NS_LEVEL);
	448	for (i = 0; i < tp_locvar->ns_level; i++)
	449	tp_locvar->vtids[i] = child_pid->numbers[i].nr;
	450	}
	451	),
	452
	453	TP_FIELDS(
	454	ctf_array_text(char, parent_comm, parent->comm, TASK_COMM_LEN)
	455	ctf_integer(pid_t, parent_tid, parent->pid)
	456	ctf_integer(pid_t, parent_pid, parent->tgid)
	457	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,8,0))
	458	ctf_integer(unsigned int, parent_ns_inum,
	459	({
	460	unsigned int parent_ns_inum = 0;
	461
	462	if (parent) {
	463	struct pid_namespace *pid_ns;
	464
	465	pid_ns = task_active_pid_ns(parent);
	466	if (pid_ns)
	467	parent_ns_inum =
	468	pid_ns->ns.inum;
	469	}
	470	parent_ns_inum;
	471	}))
	472	#endif
	473	ctf_array_text(char, child_comm, child->comm, TASK_COMM_LEN)
	474	ctf_integer(pid_t, child_tid, child->pid)
	475	ctf_sequence(pid_t, vtids, tp_locvar->vtids, u8, tp_locvar->ns_level)
	476	ctf_integer(pid_t, child_pid, child->tgid)
	477	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,8,0))
	478	ctf_integer(unsigned int, child_ns_inum,
	479	({
	480	unsigned int child_ns_inum = 0;
	481
	482	if (child) {
	483	struct pid_namespace *pid_ns;
	484
	485	pid_ns = task_active_pid_ns(child);
	486	if (pid_ns)
	487	child_ns_inum =
	488	pid_ns->ns.inum;
	489	}
	490	child_ns_inum;
	491	}))
	492	#endif
	493	),
	494
	495	TP_code_post()
	496	)
	497
	498	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,4,0))
	499	/*
	500	* Tracepoint for exec:
	501	*/
	502	LTTNG_TRACEPOINT_EVENT(sched_process_exec,
	503
	504	TP_PROTO(struct task_struct *p, pid_t old_pid,
	505	struct linux_binprm *bprm),
	506
	507	TP_ARGS(p, old_pid, bprm),
	508
	509	TP_FIELDS(
	510	ctf_string(filename, bprm->filename)
	511	ctf_integer(pid_t, tid, p->pid)
	512	ctf_integer(pid_t, old_tid, old_pid)
	513	)
	514	)
	515	#endif
	516
	517	/*
	518	* XXX the below sched_stat tracepoints only apply to SCHED_OTHER/BATCH/IDLE
	519	* adding sched_stat support to SCHED_FIFO/RR would be welcome.
	520	*/
	521	LTTNG_TRACEPOINT_EVENT_CLASS(sched_stat_template,
	522
	523	TP_PROTO(struct task_struct *tsk, u64 delay),
	524
	525	TP_ARGS(tsk, delay),
	526
	527	TP_FIELDS(
	528	ctf_array_text(char, comm, tsk->comm, TASK_COMM_LEN)
	529	ctf_integer(pid_t, tid, tsk->pid)
	530	ctf_integer(u64, delay, delay)
	531	)
	532	)
	533
	534
	535	/*
	536	* Tracepoint for accounting wait time (time the task is runnable
	537	* but not actually running due to scheduler contention).
	538	*/
	539	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_stat_template, sched_stat_wait,
	540	TP_PROTO(struct task_struct *tsk, u64 delay),
	541	TP_ARGS(tsk, delay))
	542
	543	/*
	544	* Tracepoint for accounting sleep time (time the task is not runnable,
	545	* including iowait, see below).
	546	*/
	547	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_stat_template, sched_stat_sleep,
	548	TP_PROTO(struct task_struct *tsk, u64 delay),
	549	TP_ARGS(tsk, delay))
	550
	551	/*
	552	* Tracepoint for accounting iowait time (time the task is not runnable
	553	* due to waiting on IO to complete).
	554	*/
	555	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_stat_template, sched_stat_iowait,
	556	TP_PROTO(struct task_struct *tsk, u64 delay),
	557	TP_ARGS(tsk, delay))
	558
	559	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,3,0))
	560	/*
	561	* Tracepoint for accounting blocked time (time the task is in uninterruptible).
	562	*/
	563	LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_stat_template, sched_stat_blocked,
	564	TP_PROTO(struct task_struct *tsk, u64 delay),
	565	TP_ARGS(tsk, delay))
	566	#endif
	567
	568	/*
	569	* Tracepoint for accounting runtime (time the task is executing
	570	* on a CPU).
	571	*/
	572	LTTNG_TRACEPOINT_EVENT(sched_stat_runtime,
	573
	574	TP_PROTO(struct task_struct *tsk, u64 runtime, u64 vruntime),
	575
	576	TP_ARGS(tsk, runtime, vruntime),
	577
	578	TP_FIELDS(
	579	ctf_array_text(char, comm, tsk->comm, TASK_COMM_LEN)
	580	ctf_integer(pid_t, tid, tsk->pid)
	581	ctf_integer(u64, runtime, runtime)
	582	ctf_integer(u64, vruntime, vruntime)
	583	)
	584	)
	585
	586	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,12,0) \|\| \
	587	LTTNG_RT_KERNEL_RANGE(4,9,27,18, 4,10,0,0) \|\| \
	588	LTTNG_RT_KERNEL_RANGE(4,11,5,1, 4,12,0,0))
	589	/*
	590	* Tracepoint for showing priority inheritance modifying a tasks
	591	* priority.
	592	*/
	593	LTTNG_TRACEPOINT_EVENT(sched_pi_setprio,
	594
	595	TP_PROTO(struct task_struct tsk, struct task_struct pi_task),
	596
	597	TP_ARGS(tsk, pi_task),
	598
	599	TP_FIELDS(
	600	ctf_array_text(char, comm, tsk->comm, TASK_COMM_LEN)
	601	ctf_integer(pid_t, tid, tsk->pid)
	602	ctf_integer(int, oldprio, tsk->prio - MAX_RT_PRIO)
	603	ctf_integer(int, newprio, pi_task ? pi_task->prio - MAX_RT_PRIO : tsk->prio - MAX_RT_PRIO)
	604	)
	605	)
	606	#else
	607	/*
	608	* Tracepoint for showing priority inheritance modifying a tasks
	609	* priority.
	610	*/
	611	LTTNG_TRACEPOINT_EVENT(sched_pi_setprio,
	612
	613	TP_PROTO(struct task_struct *tsk, int newprio),
	614
	615	TP_ARGS(tsk, newprio),
	616
	617	TP_FIELDS(
	618	ctf_array_text(char, comm, tsk->comm, TASK_COMM_LEN)
	619	ctf_integer(pid_t, tid, tsk->pid)
	620	ctf_integer(int, oldprio, tsk->prio - MAX_RT_PRIO)
	621	ctf_integer(int, newprio, newprio - MAX_RT_PRIO)
	622	)
	623	)
	624	#endif
	625
	626	#endif /* LTTNG_TRACE_SCHED_H */
	627
	628	/* This part must be outside protection */
	629	#include <probes/define_trace.h>