Update copyright dates
[src/app-framework-binder.git] / src / jobs.c
index 2910f0a..c2a2ec3 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2016, 2017 "IoT.bzh"
+ * Copyright (C) 2015-2020 "IoT.bzh"
  * Author José Bollo <jose.bollo@iot.bzh>
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
 #define _GNU_SOURCE
 
 #include <stdlib.h>
+#include <stdint.h>
 #include <unistd.h>
 #include <signal.h>
+#include <string.h>
 #include <time.h>
 #include <sys/syscall.h>
 #include <pthread.h>
 #include <errno.h>
 #include <assert.h>
+#include <sys/eventfd.h>
+
+#include <systemd/sd-event.h>
 
 #include "jobs.h"
+#include "evmgr.h"
 #include "sig-monitor.h"
 #include "verbose.h"
+#include "systemd.h"
 
-/** control of threads */
-struct thread
+#define EVENT_TIMEOUT_TOP      ((uint64_t)-1)
+#define EVENT_TIMEOUT_CHILD    ((uint64_t)10000)
+
+/** Internal shortcut for callback */
+typedef void (*job_cb_t)(int, void*);
+
+/** starting mode for jobs */
+enum start_mode
 {
-       struct thread *next; /**< next thread of the list */
-       pthread_t tid;     /**< the thread id */
-       unsigned stop: 1;  /**< stop request */
+       Start_Default,  /**< Start a thread if more than one jobs is pending */
+       Start_Urgent,   /**< Always start a thread */
+       Start_Lazy      /**< Never start a thread */
 };
 
-/* describes pending job */
+/** Description of a pending job */
 struct job
 {
-       struct job *next;   /* link to the next job enqueued */
-       void *group;        /* group of the request */
-       void (*callback)(int,void*,void*,void*);     /* processing callback */
-       void *arg1;         /* first arg */
-       void *arg2;         /* second arg */
-       void *arg3;         /* second arg */
-       int timeout;        /* timeout in second for processing the request */
-       int blocked;        /* is an other request blocking this one ? */
+       struct job *next;    /**< link to the next job enqueued */
+       const void *group;   /**< group of the request */
+       job_cb_t callback;   /**< processing callback */
+       void *arg;           /**< argument */
+       int timeout;         /**< timeout in second for processing the request */
+       unsigned blocked: 1; /**< is an other request blocking this one ? */
+       unsigned dropped: 1; /**< is removed ? */
+};
+
+/** Description of threads */
+struct thread
+{
+       struct thread *next;   /**< next thread of the list */
+       struct thread *upper;  /**< upper same thread */
+       struct thread *nholder;/**< next holder for evloop */
+       pthread_cond_t *cwhold;/**< condition wait for holding */
+       struct job *job;       /**< currently processed job */
+       pthread_t tid;         /**< the thread id */
+       volatile unsigned stop: 1;      /**< stop requested */
+       volatile unsigned waits: 1;     /**< is waiting? */
+       volatile unsigned leaved: 1;    /**< was leaved? */
+};
+
+/**
+ * Description of synchronous callback
+ */
+struct sync
+{
+       struct thread thread;   /**< thread loop data */
+       union {
+               void (*callback)(int, void*);   /**< the synchronous callback */
+               void (*enter)(int signum, void *closure, struct jobloop *jobloop);
+                               /**< the entering synchronous routine */
+       };
+       void *arg;              /**< the argument of the callback */
 };
 
 /* synchronisation of threads */
 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
 static pthread_cond_t  cond = PTHREAD_COND_INITIALIZER;
 
-/* count allowed, started and running threads */
-static int allowed = 0; /** allowed count of threads */
-static int started = 0; /** started count of threads */
-static int running = 0; /** running count of threads */
-static int remains = 0; /** remaining count of jobs that can be created */
+/* counts for threads */
+static int allowed_thread_count = 0; /** allowed count of threads */
+static int started_thread_count = 0; /** started count of threads */
+static int busy_thread_count = 0;    /** count of busy threads */
 
 /* list of threads */
 static struct thread *threads;
+static _Thread_local struct thread *current_thread;
+
+/* counts for jobs */
+static int remaining_job_count = 0;  /** count of job that can be created */
+static int allowed_job_count = 0;    /** allowed count of pending jobs */
 
 /* queue of pending jobs */
-static struct job *first_job;
-static struct job *first_evloop;
-static struct job *free_jobs;
+static struct job *first_pending_job;
+static struct job *first_free_job;
+
+/* event loop */
+static struct evmgr *evmgr;
+
+static void (*exit_handler)();
 
 /**
- * Adds the 'job' at the end of the list of jobs, marking it
+ * Create a new job with the given parameters
+ * @param group    the group of the job
+ * @param timeout  the timeout of the job (0 if none)
+ * @param callback the function that achieves the job
+ * @param arg      the argument of the callback
+ * @return the created job unblock or NULL when no more memory
+ */
+static struct job *job_create(
+               const void *group,
+               int timeout,
+               job_cb_t callback,
+               void *arg)
+{
+       struct job *job;
+
+       /* try recyle existing job */
+       job = first_free_job;
+       if (job)
+               first_free_job = job->next;
+       else {
+               /* allocation without blocking */
+               pthread_mutex_unlock(&mutex);
+               job = malloc(sizeof *job);
+               pthread_mutex_lock(&mutex);
+               if (!job) {
+                       ERROR("out of memory");
+                       errno = ENOMEM;
+                       goto end;
+               }
+       }
+       /* initialises the job */
+       job->group = group;
+       job->timeout = timeout;
+       job->callback = callback;
+       job->arg = arg;
+       job->blocked = 0;
+       job->dropped = 0;
+end:
+       return job;
+}
+
+/**
+ * Adds 'job' at the end of the list of jobs, marking it
  * as blocked if an other job with the same group is pending.
  * @param job the job to add
  */
-static inline void job_add(struct job *job)
+static void job_add(struct job *job)
 {
-       void *group;
+       const void *group;
        struct job *ijob, **pjob;
 
-       pjob = &first_job;
-       ijob = first_job;
-       group = job->group ? : (void*)(intptr_t)1;
+       /* prepare to add */
+       group = job->group;
+       job->next = NULL;
+
+       /* search end and blockers */
+       pjob = &first_pending_job;
+       ijob = first_pending_job;
        while (ijob) {
-               if (ijob->group == group)
+               if (group && ijob->group == group)
                        job->blocked = 1;
                pjob = &ijob->next;
                ijob = ijob->next;
        }
-       job->next = NULL;
+
+       /* queue the jobs */
        *pjob = job;
-       remains--;
+       remaining_job_count--;
 }
 
 /**
  * Get the next job to process or NULL if none.
- * The returned job if any is removed from the list of
- * jobs.
- * @return the job to process
+ * @return the first job that isn't blocked or NULL
  */
 static inline struct job *job_get()
 {
-       struct job *job, **pjob;
-
-       pjob = &first_job;
-       job = first_job;
-       while (job && job->blocked) {
-               pjob = &job->next;
+       struct job *job = first_pending_job;
+       while (job && job->blocked)
                job = job->next;
-       }
-       if (job) {
-               *pjob = job->next;
-               remains++;
-       }
+       if (job)
+               remaining_job_count++;
        return job;
 }
 
 /**
- * Unblock the first pending job of a group (if any)
- * @param group the group to unblock
+ * Releases the processed 'job': removes it
+ * from the list of jobs and unblock the first
+ * pending job of the same group if any.
+ * @param job the job to release
  */
-static inline void job_unblock(void *group)
+static inline void job_release(struct job *job)
 {
-       struct job *job;
+       struct job *ijob, **pjob;
+       const void *group;
 
-       job = first_job;
-       while (job) {
-               if (job->group == group) {
-                       job->blocked = 0;
-                       break;
-               }
-               job = job->next;
+       /* first unqueue the job */
+       pjob = &first_pending_job;
+       ijob = first_pending_job;
+       while (ijob != job) {
+               pjob = &ijob->next;
+               ijob = ijob->next;
+       }
+       *pjob = job->next;
+
+       /* then unblock jobs of the same group */
+       group = job->group;
+       if (group) {
+               ijob = job->next;
+               while (ijob && ijob->group != group)
+                       ijob = ijob->next;
+               if (ijob)
+                       ijob->blocked = 0;
        }
+
+       /* recycle the job */
+       job->next = first_free_job;
+       first_free_job = job;
 }
 
-static struct job *job_create(
-               void *group,
-               int timeout,
-               void (*callback)(int, void*, void *, void*),
-               void *arg1,
-               void *arg2,
-               void *arg3)
+/**
+ * Monitored cancel callback for a job.
+ * This function is called by the monitor
+ * to cancel the job when the safe environment
+ * is set.
+ * @param signum 0 on normal flow or the number
+ *               of the signal that interrupted the normal
+ *               flow, isn't used
+ * @param arg    the job to run
+ */
+__attribute__((unused))
+static void job_cancel(int signum, void *arg)
 {
-       struct job *job;
+       struct job *job = arg;
+       job->callback(SIGABRT, job->arg);
+}
 
-       /* allocates the job */
-       job = free_jobs;
-       if (job)
-               free_jobs = job->next;
-       else {
-               pthread_mutex_unlock(&mutex);
-               job = malloc(sizeof *job);
-               pthread_mutex_lock(&mutex);
-               if (!job) {
-                       errno = -ENOMEM;
-                       goto end;
+/**
+ * wakeup the event loop if needed by sending
+ * an event.
+ */
+static void evloop_wakeup()
+{
+       if (evmgr)
+               evmgr_wakeup(evmgr);
+}
+
+/**
+ * Release the currently held event loop
+ */
+static void evloop_release()
+{
+       struct thread *nh, *ct = current_thread;
+
+       if (ct && evmgr && evmgr_release_if(evmgr, ct)) {
+               nh = ct->nholder;
+               ct->nholder = 0;
+               if (nh) {
+                       evmgr_try_hold(evmgr, nh);
+                       pthread_cond_signal(nh->cwhold);
                }
        }
-       job->group = group;
-       job->timeout = timeout;
-       job->callback = callback;
-       job->arg1 = arg1;
-       job->arg2 = arg2;
-       job->arg3 = arg3;
-       job->blocked = 0;
-end:
-       return job;
 }
 
-static inline void job_destroy(struct job *job)
+/**
+ * get the eventloop for the current thread
+ */
+static int evloop_get()
 {
-       job->next = free_jobs;
-       free_jobs = job;
+       return evmgr && evmgr_try_hold(evmgr, current_thread);
 }
 
-static inline void job_release(struct job *job)
+/**
+ * acquire the eventloop for the current thread
+ */
+static void evloop_acquire()
 {
-       if (job->group)
-               job_unblock(job->group);
-       job_destroy(job);
+       struct thread *pwait, *ct;
+       pthread_cond_t cond;
+
+       /* try to get the evloop */
+       if (!evloop_get()) {
+               /* failed, init waiting state */
+               ct = current_thread;
+               ct->nholder = NULL;
+               ct->cwhold = &cond;
+               pthread_cond_init(&cond, NULL);
+
+               /* queue current thread in holder list */
+               pwait = evmgr_holder(evmgr);
+               while (pwait->nholder)
+                       pwait = pwait->nholder;
+               pwait->nholder = ct;
+
+               /* wake up the evloop */
+               evloop_wakeup();
+
+               /* wait to acquire the evloop */
+               pthread_cond_wait(&cond, &mutex);
+               pthread_cond_destroy(&cond);
+       }
 }
 
-/** monitored call to the job */
-static void job_call(int signum, void *arg)
+/**
+ * Enter the thread
+ * @param me the description of the thread to enter
+ */
+static void thread_enter(volatile struct thread *me)
 {
-       struct job *job = arg;
-       job->callback(signum, job->arg1, job->arg2, job->arg3);
+       evloop_release();
+       /* initialize description of itself and link it in the list */
+       me->tid = pthread_self();
+       me->stop = 0;
+       me->waits = 0;
+       me->leaved = 0;
+       me->nholder = 0;
+       me->upper = current_thread;
+       me->next = threads;
+       threads = (struct thread*)me;
+       current_thread = (struct thread*)me;
 }
 
-/** monitored cancel of the job */
-static void job_cancel(int signum, void *arg)
+/**
+ * leave the thread
+ * @param me the description of the thread to leave
+ */
+static void thread_leave()
 {
-       job_call(SIGABRT, arg);
+       struct thread **prv, *me;
+
+       /* unlink the current thread and cleanup */
+       me = current_thread;
+       prv = &threads;
+       while (*prv != me)
+               prv = &(*prv)->next;
+       *prv = me->next;
+
+       current_thread = me->upper;
 }
 
-/* main loop of processing threads */
-static void *thread_main_loop(void *data)
+/**
+ * Main processing loop of internal threads with processing jobs.
+ * The loop must be called with the mutex locked
+ * and it returns with the mutex locked.
+ * @param me the description of the thread to use
+ * TODO: how are timeout handled when reentering?
+ */
+static void thread_run_internal(volatile struct thread *me)
 {
-       struct thread me, **prv;
        struct job *job;
 
-       /* init */
-       me.tid = pthread_self();
-       me.stop = 0;
-       sig_monitor_init_timeouts();
-
-       /* chain in */
-       pthread_mutex_lock(&mutex);
-       me.next = threads;
-       threads = &me;
+       /* enter thread */
+       thread_enter(me);
 
        /* loop until stopped */
-       running++;
-       while (!me.stop) {
+       while (!me->stop) {
+               /* release the current event loop */
+               evloop_release();
+
                /* get a job */
                job = job_get();
-               if (!job && first_job && running == 0) {
-                       /* sad situation!! should not happen */
-                       ERROR("threads are blocked!");
-                       job = first_job;
-                       first_job = job->next;
-               }
                if (job) {
+                       /* prepare running the job */
+                       job->blocked = 1; /* mark job as blocked */
+                       me->job = job; /* record the job (only for terminate) */
+
                        /* run the job */
                        pthread_mutex_unlock(&mutex);
-                       sig_monitor(job->timeout, job_call, job);
+                       sig_monitor(job->timeout, job->callback, job->arg);
                        pthread_mutex_lock(&mutex);
+
+                       /* release the run job */
                        job_release(job);
-               } else {
-                       /* no job, check evloop */
-                       job = first_evloop;
-                       if (job) {
-                               /* evloop */
-                               first_evloop = job->next;
-                               pthread_mutex_unlock(&mutex);
-                               sig_monitor(job->timeout, job_call, job);
-                               pthread_mutex_lock(&mutex);
-                               job->next = first_evloop;
-                               first_evloop = job;
-                       } else {
-                               /* no job and not evloop */
-                               running--;
-                               pthread_cond_wait(&cond, &mutex);
-                               running++;
+               /* no job, check event loop wait */
+               } else if (evloop_get()) {
+                       if (!evmgr_can_run(evmgr)) {
+                               /* busy ? */
+                               CRITICAL("Can't enter dispatch while in dispatch!");
+                               abort();
                        }
+                       /* run the events */
+                       evmgr_prepare_run(evmgr);
+                       pthread_mutex_unlock(&mutex);
+                       sig_monitor(0, (void(*)(int,void*))evmgr_job_run, evmgr);
+                       pthread_mutex_lock(&mutex);
+               } else {
+                       /* no job and no event loop */
+                       busy_thread_count--;
+                       if (!busy_thread_count)
+                               ERROR("Entering job deep sleep! Check your bindings.");
+                       me->waits = 1;
+                       pthread_cond_wait(&cond, &mutex);
+                       me->waits = 0;
+                       busy_thread_count++;
                }
        }
-       running--;
+       /* cleanup */
+       evloop_release();
+       thread_leave();
+}
 
-       /* chain out */
-       prv = &threads;
-       while (*prv != &me)
-               prv = &(*prv)->next;
-       *prv = me.next;
-       pthread_mutex_unlock(&mutex);
+/**
+ * Main processing loop of external threads.
+ * The loop must be called with the mutex locked
+ * and it returns with the mutex locked.
+ * @param me the description of the thread to use
+ */
+static void thread_run_external(volatile struct thread *me)
+{
+       /* enter thread */
+       thread_enter(me);
+
+       /* loop until stopped */
+       me->waits = 1;
+       while (!me->stop)
+               pthread_cond_wait(&cond, &mutex);
+       me->waits = 0;
+       thread_leave();
+}
 
-       /* uninit and terminate */
+/**
+ * Root for created threads.
+ */
+static void thread_main()
+{
+       struct thread me;
+
+       busy_thread_count++;
+       started_thread_count++;
+       sig_monitor_init_timeouts();
+       thread_run_internal(&me);
        sig_monitor_clean_timeouts();
+       started_thread_count--;
+       busy_thread_count--;
+}
+
+/**
+ * Entry point for created threads.
+ * @param data not used
+ * @return NULL
+ */
+static void *thread_starter(void *data)
+{
+       pthread_mutex_lock(&mutex);
+       thread_main();
+       pthread_mutex_unlock(&mutex);
        return NULL;
 }
 
-/* start a new thread */
+/**
+ * Starts a new thread
+ * @return 0 in case of success or -1 in case of error
+ */
 static int start_one_thread()
 {
        pthread_t tid;
        int rc;
 
-       assert(started < allowed);
-
-       started++;
-       rc = pthread_create(&tid, NULL, thread_main_loop, NULL);
+       rc = pthread_create(&tid, NULL, thread_starter, NULL);
        if (rc != 0) {
-               started--;
-               errno = rc;
+               /* errno = rc; */
                WARNING("not able to start thread: %m");
                rc = -1;
        }
        return rc;
 }
 
-int jobs_queue(
-               void *group,
+/**
+ * Queues a new asynchronous job represented by 'callback' and 'arg'
+ * for the 'group' and the 'timeout'.
+ * Jobs are queued FIFO and are possibly executed in parallel
+ * concurrently except for job of the same group that are
+ * executed sequentially in FIFO order.
+ * @param group    The group of the job or NULL when no group.
+ * @param timeout  The maximum execution time in seconds of the job
+ *                 or 0 for unlimited time.
+ * @param callback The function to execute for achieving the job.
+ *                 Its first parameter is either 0 on normal flow
+ *                 or the signal number that broke the normal flow.
+ *                 The remaining parameter is the parameter 'arg1'
+ *                 given here.
+ * @param arg      The second argument for 'callback'
+ * @param start    The start mode for threads
+ * @return 0 in case of success or -1 in case of error
+ */
+static int queue_job_internal(
+               const void *group,
                int timeout,
                void (*callback)(int, void*),
-               void *arg)
-{
-       return jobs_queue3(group, timeout, (void(*)(int,void*,void*,void*))callback, arg, NULL, NULL);
-}
-
-int jobs_queue2(
-               void *group,
-               int timeout,
-               void (*callback)(int, void*, void*),
-               void *arg1,
-               void *arg2)
+               void *arg,
+               enum start_mode start_mode)
 {
-       return jobs_queue3(group, timeout, (void(*)(int,void*,void*,void*))callback, arg1, arg2, NULL);
-}
-
-/* queue the job to the 'callback' using a separate thread if available */
-int jobs_queue3(
-               void *group,
-               int timeout,
-               void (*callback)(int, void*, void *, void*),
-               void *arg1,
-               void *arg2,
-               void *arg3)
-{
-       const char *info;
        struct job *job;
-       int rc;
+       int rc, busy;
 
-       pthread_mutex_lock(&mutex);
+       /* check availability */
+       if (remaining_job_count <= 0) {
+               ERROR("can't process job with threads: too many jobs");
+               errno = EBUSY;
+               goto error;
+       }
 
        /* allocates the job */
-       job = job_create(group, timeout, callback, arg1, arg2, arg3);
-       if (!job) {
-               errno = ENOMEM;
-               info = "out of memory";
+       job = job_create(group, timeout, callback, arg);
+       if (!job)
                goto error;
-       }
 
        /* start a thread if needed */
-       if (remains == 0) {
-               errno = EBUSY;
-               info = "too many jobs";
-               goto error2;
-       }
-       if (started == running && started < allowed) {
+       busy = busy_thread_count == started_thread_count;
+       if (start_mode != Start_Lazy
+        && busy
+        && (start_mode == Start_Urgent || remaining_job_count + started_thread_count < allowed_job_count)
+        && started_thread_count < allowed_thread_count) {
+               /* all threads are busy and a new can be started */
                rc = start_one_thread();
-               if (rc < 0 && started == 0) {
-                       /* failed to start threading */
-                       info = "can't start first thread";
+               if (rc < 0 && started_thread_count == 0) {
+                       ERROR("can't start initial thread: %m");
                        goto error2;
                }
+               busy = 0;
        }
 
        /* queues the job */
        job_add(job);
-       pthread_mutex_unlock(&mutex);
 
-       /* signal an existing job */
+       /* wakeup an evloop if needed */
+       if (busy)
+               evloop_wakeup();
+
        pthread_cond_signal(&cond);
        return 0;
 
 error2:
-       job_destroy(job);
+       job->next = first_free_job;
+       first_free_job = job;
 error:
-       ERROR("can't process job with threads: %s, %m", info);
-       pthread_mutex_unlock(&mutex);
        return -1;
 }
 
-/* initialise the threads */
-int jobs_init(int allowed_count, int start_count, int waiter_count)
+/**
+ * Queues a new asynchronous job represented by 'callback' and 'arg'
+ * for the 'group' and the 'timeout'.
+ * Jobs are queued FIFO and are possibly executed in parallel
+ * concurrently except for job of the same group that are
+ * executed sequentially in FIFO order.
+ * @param group    The group of the job or NULL when no group.
+ * @param timeout  The maximum execution time in seconds of the job
+ *                 or 0 for unlimited time.
+ * @param callback The function to execute for achieving the job.
+ *                 Its first parameter is either 0 on normal flow
+ *                 or the signal number that broke the normal flow.
+ *                 The remaining parameter is the parameter 'arg1'
+ *                 given here.
+ * @param arg      The second argument for 'callback'
+ * @param start    The start mode for threads
+ * @return 0 in case of success or -1 in case of error
+ */
+static int queue_job(
+               const void *group,
+               int timeout,
+               void (*callback)(int, void*),
+               void *arg,
+               enum start_mode start_mode)
 {
-       /* records the allowed count */
-       allowed = allowed_count;
-       started = 0;
-       running = 0;
-       remains = waiter_count;
+       int rc;
 
-       /* start at least one thread */
        pthread_mutex_lock(&mutex);
-       while (started < start_count && start_one_thread() == 0);
+       rc = queue_job_internal(group, timeout, callback, arg, start_mode);
        pthread_mutex_unlock(&mutex);
+       return rc;
 
-       /* end */
-       return -(started != start_count);
 }
 
-/* terminate all the threads and all pending requests */
-void jobs_terminate(int wait)
+/**
+ * Queues a new asynchronous job represented by 'callback' and 'arg'
+ * for the 'group' and the 'timeout'.
+ * Jobs are queued FIFO and are possibly executed in parallel
+ * concurrently except for job of the same group that are
+ * executed sequentially in FIFO order.
+ * @param group    The group of the job or NULL when no group.
+ * @param timeout  The maximum execution time in seconds of the job
+ *                 or 0 for unlimited time.
+ * @param callback The function to execute for achieving the job.
+ *                 Its first parameter is either 0 on normal flow
+ *                 or the signal number that broke the normal flow.
+ *                 The remaining parameter is the parameter 'arg1'
+ *                 given here.
+ * @param arg      The second argument for 'callback'
+ * @return 0 in case of success or -1 in case of error
+ */
+int jobs_queue(
+               const void *group,
+               int timeout,
+               void (*callback)(int, void*),
+               void *arg)
 {
-       struct job *job;
-       pthread_t me, other;
-       struct thread *t;
+       return queue_job(group, timeout, callback, arg, Start_Default);
+}
 
-       /* how am i? */
-       me = pthread_self();
+/**
+ * Queues lazyly a new asynchronous job represented by 'callback' and 'arg'
+ * for the 'group' and the 'timeout'.
+ * Jobs are queued FIFO and are possibly executed in parallel
+ * concurrently except for job of the same group that are
+ * executed sequentially in FIFO order.
+ * @param group    The group of the job or NULL when no group.
+ * @param timeout  The maximum execution time in seconds of the job
+ *                 or 0 for unlimited time.
+ * @param callback The function to execute for achieving the job.
+ *                 Its first parameter is either 0 on normal flow
+ *                 or the signal number that broke the normal flow.
+ *                 The remaining parameter is the parameter 'arg1'
+ *                 given here.
+ * @param arg      The second argument for 'callback'
+ * @return 0 in case of success or -1 in case of error
+ */
+int jobs_queue_lazy(
+               const void *group,
+               int timeout,
+               void (*callback)(int, void*),
+               void *arg)
+{
+       return queue_job(group, timeout, callback, arg, Start_Lazy);
+}
+
+/**
+ * Queues urgently a new asynchronous job represented by 'callback' and 'arg'
+ * for the 'group' and the 'timeout'.
+ * Jobs are queued FIFO and are possibly executed in parallel
+ * concurrently except for job of the same group that are
+ * executed sequentially in FIFO order.
+ * @param group    The group of the job or NULL when no group.
+ * @param timeout  The maximum execution time in seconds of the job
+ *                 or 0 for unlimited time.
+ * @param callback The function to execute for achieving the job.
+ *                 Its first parameter is either 0 on normal flow
+ *                 or the signal number that broke the normal flow.
+ *                 The remaining parameter is the parameter 'arg1'
+ *                 given here.
+ * @param arg      The second argument for 'callback'
+ * @return 0 in case of success or -1 in case of error
+ */
+int jobs_queue_urgent(
+               const void *group,
+               int timeout,
+               void (*callback)(int, void*),
+               void *arg)
+{
+       return queue_job(group, timeout, callback, arg, Start_Urgent);
+}
+
+/**
+ * Internal helper function for 'jobs_enter'.
+ * @see jobs_enter, jobs_leave
+ */
+static void enter_cb(int signum, void *closure)
+{
+       struct sync *sync = closure;
+       sync->enter(signum, sync->arg, (void*)&sync->thread);
+}
+
+/**
+ * Internal helper function for 'jobs_call'.
+ * @see jobs_call
+ */
+static void call_cb(int signum, void *closure)
+{
+       struct sync *sync = closure;
+       sync->callback(signum, sync->arg);
+       jobs_leave((void*)&sync->thread);
+}
+
+/**
+ * Internal helper for synchronous jobs. It enters
+ * a new thread loop for evaluating the given job
+ * as recorded by the couple 'sync_cb' and 'sync'.
+ * @see jobs_call, jobs_enter, jobs_leave
+ */
+static int do_sync(
+               const void *group,
+               int timeout,
+               void (*sync_cb)(int signum, void *closure),
+               struct sync *sync
+)
+{
+       int rc;
 
-       /* request all threads to stop */
        pthread_mutex_lock(&mutex);
-       allowed = 0;
-       for(;;) {
-               /* search the next thread to stop */
-               t = threads;
-               while (t && pthread_equal(t->tid, me))
-                       t = t->next;
-               if (!t)
-                       break;
-               /* stop it */
-               other = t->tid;
+
+       rc = queue_job_internal(group, timeout, sync_cb, sync, Start_Default);
+       if (rc == 0) {
+               /* run until stopped */
+               if (current_thread)
+                       thread_run_internal(&sync->thread);
+               else
+                       thread_run_external(&sync->thread);
+               if (!sync->thread.leaved) {
+                       errno = EINTR;
+                       rc = -1;
+               }
+       }
+       pthread_mutex_unlock(&mutex);
+       return rc;
+}
+
+/**
+ * Enter a synchronisation point: activates the job given by 'callback'
+ * and 'closure' using 'group' and 'timeout' to control sequencing and
+ * execution time.
+ * @param group the group for sequencing jobs
+ * @param timeout the time in seconds allocated to the job
+ * @param callback the callback that will handle the job.
+ *                 it receives 3 parameters: 'signum' that will be 0
+ *                 on normal flow or the catched signal number in case
+ *                 of interrupted flow, the context 'closure' as given and
+ *                 a 'jobloop' reference that must be used when the job is
+ *                 terminated to unlock the current execution flow.
+ * @param closure the argument to the callback
+ * @return 0 on success or -1 in case of error
+ */
+int jobs_enter(
+               const void *group,
+               int timeout,
+               void (*callback)(int signum, void *closure, struct jobloop *jobloop),
+               void *closure
+)
+{
+       struct sync sync;
+
+       sync.enter = callback;
+       sync.arg = closure;
+       return do_sync(group, timeout, enter_cb, &sync);
+}
+
+/**
+ * Unlocks the execution flow designed by 'jobloop'.
+ * @param jobloop indication of the flow to unlock
+ * @return 0 in case of success of -1 on error
+ */
+int jobs_leave(struct jobloop *jobloop)
+{
+       struct thread *t;
+
+       pthread_mutex_lock(&mutex);
+       t = threads;
+       while (t && t != (struct thread*)jobloop)
+               t = t->next;
+       if (!t) {
+               errno = EINVAL;
+       } else {
+               t->leaved = 1;
                t->stop = 1;
-               pthread_mutex_unlock(&mutex);
-               pthread_cond_broadcast(&cond);
-               pthread_join(other, NULL);
-               pthread_mutex_lock(&mutex);
+               if (t->waits)
+                       pthread_cond_broadcast(&cond);
+               else
+                       evloop_wakeup();
+       }
+       pthread_mutex_unlock(&mutex);
+       return -!t;
+}
+
+/**
+ * Calls synchronously the job represented by 'callback' and 'arg1'
+ * for the 'group' and the 'timeout' and waits for its completion.
+ * @param group    The group of the job or NULL when no group.
+ * @param timeout  The maximum execution time in seconds of the job
+ *                 or 0 for unlimited time.
+ * @param callback The function to execute for achieving the job.
+ *                 Its first parameter is either 0 on normal flow
+ *                 or the signal number that broke the normal flow.
+ *                 The remaining parameter is the parameter 'arg1'
+ *                 given here.
+ * @param arg      The second argument for 'callback'
+ * @return 0 in case of success or -1 in case of error
+ */
+int jobs_call(
+               const void *group,
+               int timeout,
+               void (*callback)(int, void*),
+               void *arg)
+{
+       struct sync sync;
+
+       sync.callback = callback;
+       sync.arg = arg;
+
+       return do_sync(group, timeout, call_cb, &sync);
+}
+
+/**
+ * Ensure that the current running thread can control the event loop.
+ */
+void jobs_acquire_event_manager()
+{
+       struct thread lt;
+
+       /* ensure an existing thread environment */
+       if (!current_thread) {
+               memset(&lt, 0, sizeof lt);
+               current_thread = &lt;
        }
 
-       /* cancel pending jobs */
-       while (first_job) {
-               job = first_job;
-               first_job = job->next;
-               sig_monitor(0, job_cancel, job);
-               free(job);
+       /* lock */
+       pthread_mutex_lock(&mutex);
+
+       /* creates the evloop on need */
+       if (!evmgr)
+               evmgr_create(&evmgr);
+
+       /* acquire the event loop under lock */
+       if (evmgr)
+               evloop_acquire();
+
+       /* unlock */
+       pthread_mutex_unlock(&mutex);
+
+       /* release the faked thread environment if needed */
+       if (current_thread == &lt) {
+               /*
+                * Releasing it is needed because there is no way to guess
+                * when it has to be released really. But here is where it is
+                * hazardous: if the caller modifies the eventloop when it
+                * is waiting, there is no way to make the change effective.
+                * A workaround to achieve that goal is for the caller to
+                * require the event loop a second time after having modified it.
+                */
+               NOTICE("Requiring event manager/loop from outside of binder's callback is hazardous!");
+               if (verbose_wants(Log_Level_Info))
+                       sig_monitor_dumpstack();
+               evloop_release();
+               current_thread = NULL;
        }
 }
 
-int jobs_add_event_loop(void *key, int timeout, void (*evloop)(int signum, void*), void *closure)
+/**
+ * Enter the jobs processing loop.
+ * @param allowed_count Maximum count of thread for jobs including this one
+ * @param start_count   Count of thread to start now, must be lower.
+ * @param waiter_count  Maximum count of jobs that can be waiting.
+ * @param start         The start routine to activate (can't be NULL)
+ * @return 0 in case of success or -1 in case of error.
+ */
+int jobs_start(
+       int allowed_count,
+       int start_count,
+       int waiter_count,
+       void (*start)(int signum, void* arg),
+       void *arg)
 {
+       int rc, launched;
        struct job *job;
 
+       assert(allowed_count >= 1);
+       assert(start_count >= 0);
+       assert(waiter_count > 0);
+       assert(start_count <= allowed_count);
+
+       rc = -1;
        pthread_mutex_lock(&mutex);
-       job = job_create(key, timeout, (void (*)(int,  void *, void *, void *))evloop, closure, NULL, NULL);
-       if (job) {
-               /* adds the loop */
-               job->next = first_evloop;
-               first_evloop = job;
-
-               /* signal the loop */
-               pthread_cond_signal(&cond);
+
+       /* check whether already running */
+       if (current_thread || allowed_thread_count) {
+               ERROR("thread already started");
+               errno = EINVAL;
+               goto error;
        }
+
+       /* records the allowed count */
+       allowed_thread_count = allowed_count;
+       started_thread_count = 0;
+       busy_thread_count = 0;
+       remaining_job_count = waiter_count;
+       allowed_job_count = waiter_count;
+
+       /* start at least one thread: the current one */
+       launched = 1;
+       while (launched < start_count) {
+               if (start_one_thread() != 0) {
+                       ERROR("Not all threads can be started");
+                       goto error;
+               }
+               launched++;
+       }
+
+       /* queue the start job */
+       job = job_create(NULL, 0, start, arg);
+       if (!job)
+               goto error;
+       job_add(job);
+
+       /* run until end */
+       thread_main();
+       rc = 0;
+error:
        pthread_mutex_unlock(&mutex);
-       return -!job;
+       if (exit_handler)
+               exit_handler();
+       return rc;
 }
 
-int jobs_add_me()
+/**
+ * Exit jobs threads and call handler if not NULL.
+ */
+void jobs_exit(void (*handler)())
 {
-       pthread_t me;
        struct thread *t;
 
-       /* how am i? */
-       me = pthread_self();
-
        /* request all threads to stop */
        pthread_mutex_lock(&mutex);
+
+       /* set the handler */
+       exit_handler = handler;
+
+       /* stops the threads */
        t = threads;
        while (t) {
-               if (pthread_equal(t->tid, me)) {
-                       pthread_mutex_unlock(&mutex);
-                       ERROR("thread already running");
-                       errno = EINVAL;
-                       return -1;
-               }
+               t->stop = 1;
                t = t->next;
        }
 
-       /* allowed... */
-       allowed++;
-       pthread_mutex_unlock(&mutex);
-
-       /* run */
-       thread_main_loop(NULL);
+       /* wake up the threads */
+       evloop_wakeup();
+       pthread_cond_broadcast(&cond);
 
-       /* returns */
-       pthread_mutex_lock(&mutex);
-       allowed--;
+       /* leave */
        pthread_mutex_unlock(&mutex);
-       return 0;
 }
-
-