Merge branch 'maint'

[paraslash.git] / sched.c

/* Copyright (C) 2006 Andre Noll <maan@tuebingen.mpg.de>, see file COPYING. */

/** \file sched.c Paraslash's scheduling functions. */

#include <regex.h>

#include "para.h"
#include "ipc.h"
#include "fd.h"
#include "list.h"
#include "sched.h"
#include "string.h"
#include "time.h"
#include "error.h"

/**
 * The possible states of a task.
 *
 * In addition to the states listed here, a task may also enter zombie state.
 * This happens when its ->post_monitor function returns negative, the ->status
 * field is then set to this return value. Such tasks are not scheduled any
 * more (i.e. ->pre_monitor() and ->post_monitor() are no longer called), but
 * they stay on the scheduler task list until \ref task_reap() or
 * \ref sched_shutdown() is called.
 */
enum task_status {
        /** Task has been reaped and may be removed from the task list. */
        TS_DEAD,
        /** Task is active. */
        TS_RUNNING,
};

struct task {
        /** A copy of the task name supplied when the task was registered. */
        char *name;
        /** Copied during task_register(). */
        struct task_info info;
        /* TS_RUNNING, TS_DEAD, or zombie (negative value). */
        int status;
        /** Position of the task in the task list of the scheduler. */
        struct list_head node;
        /** If less than zero, the task was notified by another task. */
        int notification;
};

static struct timeval now_struct;
const struct timeval *now = &now_struct;

static void sched_pre_monitor(struct sched *s)
{
        struct task *t, *tmp;

        list_for_each_entry_safe(t, tmp, &s->task_list, node) {
                if (t->status < 0)
                        continue;
                if (t->notification != 0)
                        sched_min_delay(s);
                if (t->info.pre_monitor)
                        t->info.pre_monitor(s, t->info.context);
        }
}

static void unlink_and_free_task(struct task *t)
{
        PARA_INFO_LOG("freeing task %s (%s)\n", t->name, t->status < 0?
                para_strerror(-t->status) :
                (t->status == TS_DEAD? "[dead]" : "[running]"));

        list_del(&t->node);
        free(t->name);
        free(t);
}

//#define SCHED_DEBUG 1
static inline void call_post_monitor(struct sched *s, struct task *t)
{
        int ret;

#ifndef SCHED_DEBUG
        ret = t->info.post_monitor(s, t->info.context);
#else
        struct timeval t1, t2, diff;
        unsigned long pst;

        clock_get_realtime(&t1);
        ret = t->info.post_monitor(s, t->info.context);
        clock_get_realtime(&t2);
        tv_diff(&t1, &t2, &diff);
        pst = tv2ms(&diff);
        if (pst > 50)
                PARA_WARNING_LOG("%s: post_monitor time: %lums\n",
                        t->name, pst);
#endif
        t->status = ret < 0? ret : TS_RUNNING;
}

static unsigned sched_post_monitor(struct sched *s)
{
        struct task *t, *tmp;
        unsigned num_running_tasks = 0;

        list_for_each_entry_safe(t, tmp, &s->task_list, node) {
                if (t->status == TS_DEAD) /* task has been reaped */
                        unlink_and_free_task(t);
                else if (t->status == TS_RUNNING) {
                        call_post_monitor(s, t); /* sets t->status */
                        t->notification = 0;
                        if (t->status == TS_RUNNING)
                                num_running_tasks++;
                }
        }
        return num_running_tasks;
}

/**
 * The core function of all paraslash programs.
 *
 * \param s Pointer to the scheduler struct.
 *
 * This function updates the global now pointer, calls all registered
 * pre_monitor hooks which may set the timeout and add any file descriptors to
 * the pollfd array. Next, it calls the poll function and makes the result
 * available to the registered tasks by calling their post_monitor hook.
 *
 * \return Zero if no more tasks are left in the task list, negative if the
 * poll function returned an error.
 *
 * \sa \ref now.
 */
int schedule(struct sched *s)
{
        int ret;
        unsigned num_running_tasks;

        if (!s->poll_function)
                s->poll_function = xpoll;
again:
        s->num_pfds = 0;
        if (s->pidx)
                memset(s->pidx, 0xff, s->pidx_array_len * sizeof(unsigned));
        s->timeout = s->default_timeout;
        clock_get_realtime(&now_struct);
        sched_pre_monitor(s);
        ret = s->poll_function(s->pfd, s->num_pfds, s->timeout);
        if (ret < 0)
                return ret;
        clock_get_realtime(&now_struct);
        num_running_tasks = sched_post_monitor(s);
        if (num_running_tasks == 0)
                return 0;
        goto again;
}

/**
 * Obtain the error status of a task and deallocate its resources.
 *
 * \param tptr Identifies the task to reap.
 *
 * This function is similar to wait(2) in that it returns information about a
 * terminated task which allows releasing the resources associated with the
 * task. Until this function is called, the terminated task remains in a zombie
 * state.
 *
 * \return If \a tptr is \p NULL, or \a *tptr is \p NULL, the function does
 * nothing and returns zero. Otherwise, it is checked whether the task
 * identified by \a tptr is still running. If it is, the function returns zero
 * and again, no action is taken. Otherwise the (negative) error code of the
 * terminated task is returned and \a *tptr is set to \p NULL. The task will
 * then be removed removed from the scheduler task list.
 *
 * \sa \ref sched_shutdown(), wait(2).
 */
int task_reap(struct task **tptr)
{
        struct task *t;
        int ret;

        if (!tptr)
                return 0;
        t = *tptr;
        if (!t)
                return 0;
        if (t->status >= 0)
                return 0;
        ret = t->status;
        /*
         * With list_for_each_entry_safe() it is only safe to remove the
         * _current_ list item. Since we are being called from the loop in
         * schedule() via some task's ->post_monitor() function, freeing the
         * given task here would result in use-after-free bugs in schedule().
         * So we only set the task status to TS_DEAD which tells schedule() to
         * free the task in the next iteration of its loop.
         */
        t->status = TS_DEAD;

        *tptr = NULL;
        return ret;
}

/**
 * Deallocate all resources of all tasks of a scheduler instance.
 *
 * \param s The scheduler instance.
 *
 * This should only be called after \ref schedule() has returned.
 */
void sched_shutdown(struct sched *s)
{
        struct task *t, *tmp;

        list_for_each_entry_safe(t, tmp, &s->task_list, node) {
                if (t->status == TS_RUNNING)
                        /* The task list should contain only terminated tasks. */
                        PARA_WARNING_LOG("shutting down running task %s\n",
                                t->name);
                unlink_and_free_task(t);
        }
        free(s->pfd);
        free(s->pidx);
}

/**
 * Add a task to the scheduler task list.
 *
 * \param info Task information supplied by the caller.
 * \param s The scheduler instance.
 *
 * \return A pointer to a newly allocated task structure. It will be
 * freed by sched_shutdown().
 */
struct task *task_register(struct task_info *info, struct sched *s)
{
        struct task *t = alloc(sizeof(*t));

        assert(info->post_monitor);

        if (!s->task_list.next)
                init_list_head(&s->task_list);

        t->info = *info;
        t->name = para_strdup(info->name);
        t->notification = 0;
        t->status = TS_RUNNING;
        list_add_tail(&t->node, &s->task_list);
        return t;
}

/**
 * Get the list of all registered tasks.
 *
 * \param s The scheduler instance to get the task list from.
 *
 * \return The task list.
 *
 * Each entry of the list contains an identifier which is simply a hex number.
 * The result is dynamically allocated and must be freed by the caller.
 */
char *get_task_list(struct sched *s)
{
        struct task *t, *tmp;
        char *msg = NULL;

        list_for_each_entry_safe(t, tmp, &s->task_list, node) {
                char *tmp_msg;
                tmp_msg = make_message("%s%p\t%s\t%s\n", msg? msg : "", t,
                        t->status == TS_DEAD? "dead" :
                                (t->status == TS_RUNNING? "running" : "zombie"),
                        t->name);
                free(msg);
                msg = tmp_msg;
        }
        return msg;
}

/**
 * Set the notification value of a task.
 *
 * \param t The task to notify.
 * \param err A positive error code.
 *
 * Tasks which honor notifications are supposed to call \ref
 * task_get_notification() in their post_monitor function and act on the
 * returned notification value.
 *
 * If the scheduler detects during its pre_monitor loop that at least one task
 * has been notified, the loop terminates, and the post_monitor methods of all
 * taks are immediately called again.
 *
 * The notification for a task is reset after the call to its post_monitor
 * method.
 *
 * \sa \ref task_get_notification().
 */
void task_notify(struct task *t, int err)
{
        assert(err > 0);
        if (t->notification == -err) /* ignore subsequent notifications */
                return;
        PARA_INFO_LOG("notifying task %s: %s\n", t->name, para_strerror(err));
        t->notification = -err;
}

/**
 * Return the notification value of a task.
 *
 * \param t The task to get the notification value from.
 *
 * \return The notification value. If this is negative, the task has been
 * notified by another task. Tasks are supposed to check for notifications by
 * calling this function from their post_monitor method.
 *
 * \sa \ref task_notify().
 */
int task_get_notification(const struct task *t)
{
        return t->notification;
}

/**
 * Return the status value of a task.
 *
 * \param t The task to get the status value from.
 *
 * \return Zero if task does not exist, one if task is running, negative error
 * code if task has terminated.
 */
int task_status(const struct task *t)
{
        if (!t)
                return 0;
        if (t->status == TS_DEAD) /* pretend dead tasks don't exist */
                return 0;
        if (t->status == TS_RUNNING)
                return 1;
        return t->status;
}

/**
 * Set the notification value of all tasks of a scheduler instance.
 *
 * \param s The scheduler instance whose tasks should be notified.
 * \param err A positive error code.
 *
 * This simply iterates over all existing tasks of \a s and sets each
 * task's notification value to \p -err.
 */
void task_notify_all(struct sched *s, int err)
{
        struct task *t;

        list_for_each_entry(t, &s->task_list, node)
                task_notify(t, err);
}

/**
 * Set the I/O timeout to the minimal possible value.
 *
 * \param s Pointer to the scheduler struct.
 *
 * This causes the next poll() call to return immediately.
 */
void sched_min_delay(struct sched *s)
{
        s->timeout = 0;
}

/**
 * Impose an upper bound for the I/O timeout.
 *
 * \param to Maximal allowed timeout.
 * \param s Pointer to the scheduler struct.
 *
 * If the current I/O timeout is already smaller than to, this function does
 * nothing. Otherwise the timeout is set to the given value.
 *
 * \sa \ref sched_request_timeout_ms().
 */
void sched_request_timeout(struct timeval *to, struct sched *s)
{
        long unsigned ms = tv2ms(to);
        if (s->timeout > ms)
                s->timeout = ms;
}

/**
 * Bound the I/O timeout to at most the given amount of milliseconds.
 *
 * \param ms The maximal allowed timeout in milliseconds.
 * \param s Pointer to the scheduler struct.
 *
 * Like \ref sched_request_timeout() this imposes an upper bound on the I/O
 * timeout.
 */
void sched_request_timeout_ms(long unsigned ms, struct sched *s)
{
        struct timeval tv;
        ms2tv(ms, &tv);
        sched_request_timeout(&tv, s);
}

/**
 * Bound the I/O timeout by an absolute time in the future.
 *
 * \param barrier Defines the upper bound for the timeout.
 * \param s Pointer to the scheduler struct.
 *
 * \return If the barrier is in the past, this function does nothing and
 * returns zero. Otherwise it returns one.
 *
 * \sa \ref sched_request_barrier_or_min_delay().
 */
int sched_request_barrier(struct timeval *barrier, struct sched *s)
{
        struct timeval diff;

        if (tv_diff(now, barrier, &diff) > 0)
                return 0;
        sched_request_timeout(&diff, s);
        return 1;
}

/**
 * Bound the I/O timeout or request a minimal delay.
 *
 * \param barrier Absolute time as in \ref sched_request_barrier().
 * \param s Pointer to the scheduler struct.
 *
 * \return If the barrier is in the past, this function requests a minimal
 * timeout and returns zero. Otherwise it returns one.
 *
 * \sa \ref sched_min_delay(), \ref sched_request_barrier().
 */
int sched_request_barrier_or_min_delay(struct timeval *barrier, struct sched *s)
{
        struct timeval diff;

        if (tv_diff(now, barrier, &diff) > 0) {
                sched_min_delay(s);
                return 0;
        }
        sched_request_timeout(&diff, s);
        return 1;
}

static void add_pollfd(int fd, struct sched *s, short events)
{
        assert(fd >= 0);
#if 0
        {
                int flags = fcntl(fd, F_GETFL);
                if (!(flags & O_NONBLOCK)) {
                        PARA_EMERG_LOG("fd %d is a blocking file descriptor\n", fd);
                        exit(EXIT_FAILURE);
                }
        }
#endif
        if (s->pidx_array_len > fd) { /* is fd already registered? */
                if (s->pidx[fd] < s->pfd_array_len) { /* yes, it is */
                        assert(s->pfd[s->pidx[fd]].fd == fd);
                        s->pfd[s->pidx[fd]].events |= events;
                        return;
                }
        } else { /* need to extend the index array */
                unsigned old_len = s->pidx_array_len;
                while (s->pidx_array_len <= fd)
                        s->pidx_array_len = s->pidx_array_len * 2 + 1;
                PARA_INFO_LOG("pidx array len: %u\n", s->pidx_array_len);
                s->pidx = para_realloc(s->pidx,
                        s->pidx_array_len * sizeof(unsigned));
                memset(s->pidx + old_len, 0xff,
                        (s->pidx_array_len - old_len) * sizeof(unsigned));
        }
        /*
         * The given fd is not part of the pfd array yet. Initialize pidx[fd]
         * to point at the next unused slot of this array and initialize the
         * slot.
         */
        s->pidx[fd] = s->num_pfds;
        if (s->pfd_array_len <= s->num_pfds) {
                unsigned old_len = s->pfd_array_len;
                s->pfd_array_len = old_len * 2 + 1;
                PARA_INFO_LOG("pfd array len: %u\n", s->pfd_array_len);
                s->pfd = para_realloc(s->pfd,
                        s->pfd_array_len * sizeof(struct pollfd));
                memset(s->pfd + old_len, 0,
                        (s->pfd_array_len - old_len) * sizeof(struct pollfd));
        }
        s->pfd[s->num_pfds].fd = fd;
        s->pfd[s->num_pfds].events = events;
        s->pfd[s->num_pfds].revents = 0;
        s->num_pfds++;
}

/**
 * Instruct the scheduler to monitor an fd for readiness for reading.
 *
 * \param fd The file descriptor.
 * \param s The scheduler.
 *
 * \sa \ref sched_monitor_writefd().
 */
void sched_monitor_readfd(int fd, struct sched *s)
{
        add_pollfd(fd, s, POLLIN);
}

/**
 * Instruct the scheduler to monitor an fd for readiness for writing.
 *
 * \param fd The file descriptor.
 * \param s The scheduler.
 *
 * \sa \ref sched_monitor_readfd().
 */
void sched_monitor_writefd(int fd, struct sched *s)
{
        add_pollfd(fd, s, POLLOUT);
}

static int get_revents(int fd, const struct sched *s)
{
        if (fd < 0)
                return 0;
        if (fd >= s->pidx_array_len)
                return 0;
        if (s->pidx[fd] >= s->num_pfds)
                return 0;
        if (s->pfd[s->pidx[fd]].fd != fd)
                return 0;
        assert((s->pfd[s->pidx[fd]].revents & POLLNVAL) == 0);
        return s->pfd[s->pidx[fd]].revents;
}

/**
 * Check whether there is data to read on the given fd.
 *
 * To be called from the ->post_monitor() method of a task.
 *
 * \param fd Should have been monitored with \ref sched_monitor_readfd().
 * \param s The scheduler instance.
 *
 * \return True if the file descriptor is ready for reading, false otherwise.
 * If fd is negative, or has not been monitored in the current iteration of the
 * scheduler's main loop, the function also returns false.
 *
 * \sa \ref sched_write_ok().
 */
bool sched_read_ok(int fd, const struct sched *s)
{
        return get_revents(fd, s) & (POLLIN | POLLERR | POLLHUP);
}

/**
 * Check whether writing is possible (i.e., does not block).
 *
 * \param fd Should have been monitored with \ref sched_monitor_writefd().
 * \param s The scheduler instance.
 *
 * \return True if the file descriptor is ready for writing, false otherwise.
 * The comment in \ref sched_read_ok() about invalid file descriptors applies
 * to this function as well.
 */
bool sched_write_ok(int fd, const struct sched *s)
{
        return get_revents(fd, s) & (POLLOUT | POLLERR | POLLHUP);
}