jobs: Add starting mode for jobs
authorJose Bollo <jose.bollo@iot.bzh>
Wed, 20 Feb 2019 09:07:21 +0000 (10:07 +0100)
committerJosé Bollo <jose.bollo@iot.bzh>
Fri, 22 Mar 2019 11:21:54 +0000 (12:21 +0100)
Also improve naming of variables.

Change-Id: I8efa74e27256425df18f57b3de257c234ff60e69
Signed-off-by: Jose Bollo <jose.bollo@iot.bzh>
src/CMakeLists.txt
src/jobs.c
src/jobs.h

index 64a8d97..fbefcda 100644 (file)
@@ -26,12 +26,12 @@ ADD_DEFINITIONS(-DINFER_EXTENSION)
 
 ############################################################################
 # TODO: improve below setting by using config file
-option(WITH_SIG_MONITOR_DUMPSTACK "activate dump stack on error"           ON)
-option(WITH_SIG_MONITOR_SIGNALS   "activate handling of signals"           ON)
-option(WITH_SIG_MONITOR_FOR_CALL  "activate monitoring of calls"           ON)
-option(WITH_SIG_MONITOR_TIMERS    "activate monitoring of call expiration" ON)
-option(WITH_AFB_HOOK              "include hooking"                        ON)
-option(WITH_AFB_TRACE             "include monitoring trace"               ON)
+option(WITH_SIG_MONITOR_DUMPSTACK "Activate dump stack on error"           ON)
+option(WITH_SIG_MONITOR_SIGNALS   "Activate handling of signals"           ON)
+option(WITH_SIG_MONITOR_FOR_CALL  "Activate monitoring of calls"           ON)
+option(WITH_SIG_MONITOR_TIMERS    "Activate monitoring of call expiration" ON)
+option(WITH_AFB_HOOK              "Include hooking"                        ON)
+option(WITH_AFB_TRACE             "Include monitoring trace"               ON)
 option(WITH_SUPERVISOR            "Activates installation of supervisor"   OFF)
 option(WITH_DBUS_TRANSPARENCY     "Allows API transparency over DBUS"      OFF)
 option(WITH_LEGACY_BINDING_V1     "Includes the legacy Binding API version 1" OFF)
index 4a9ce18..029ecae 100644 (file)
 #define EVENT_TIMEOUT_TOP      ((uint64_t)-1)
 #define EVENT_TIMEOUT_CHILD    ((uint64_t)10000)
 
-struct thread;
-
 /** Internal shortcut for callback */
 typedef void (*job_cb_t)(int, void*);
 
+/** starting mode for jobs */
+enum start_mode
+{
+       Start_Default,  /**< Start a thread if more than one jobs is pending */
+       Start_Urgent,   /**< Always start a thread */
+       Start_Lazy      /**< Never start a thread */
+};
+
 /** Description of a pending job */
 struct job
 {
@@ -83,19 +89,22 @@ struct sync
 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; /** allowed count of waiting jobs */
+/* 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 *free_jobs;
+static struct job *first_pending_job;
+static struct job *first_free_job;
 
 /* event loop */
 static struct evmgr *evmgr;
@@ -119,9 +128,9 @@ static struct job *job_create(
        struct job *job;
 
        /* try recyle existing job */
-       job = free_jobs;
+       job = first_free_job;
        if (job)
-               free_jobs = job->next;
+               first_free_job = job->next;
        else {
                /* allocation without blocking */
                pthread_mutex_unlock(&mutex);
@@ -159,8 +168,8 @@ static void job_add(struct job *job)
        job->next = NULL;
 
        /* search end and blockers */
-       pjob = &first_job;
-       ijob = first_job;
+       pjob = &first_pending_job;
+       ijob = first_pending_job;
        while (ijob) {
                if (group && ijob->group == group)
                        job->blocked = 1;
@@ -170,7 +179,7 @@ static void job_add(struct job *job)
 
        /* queue the jobs */
        *pjob = job;
-       remains--;
+       remaining_job_count--;
 }
 
 /**
@@ -179,11 +188,11 @@ static void job_add(struct job *job)
  */
 static inline struct job *job_get()
 {
-       struct job *job = first_job;
+       struct job *job = first_pending_job;
        while (job && job->blocked)
                job = job->next;
        if (job)
-               remains++;
+               remaining_job_count++;
        return job;
 }
 
@@ -199,8 +208,8 @@ static inline void job_release(struct job *job)
        const void *group;
 
        /* first unqueue the job */
-       pjob = &first_job;
-       ijob = first_job;
+       pjob = &first_pending_job;
+       ijob = first_pending_job;
        while (ijob != job) {
                pjob = &ijob->next;
                ijob = ijob->next;
@@ -218,8 +227,8 @@ static inline void job_release(struct job *job)
        }
 
        /* recycle the job */
-       job->next = free_jobs;
-       free_jobs = job;
+       job->next = first_free_job;
+       first_free_job = job;
 }
 
 /**
@@ -388,13 +397,13 @@ static void thread_run_internal(volatile struct thread *me)
                        pthread_mutex_lock(&mutex);
                } else {
                        /* no job and no event loop */
-                       running--;
-                       if (!running)
+                       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;
-                       running++;
+                       busy_thread_count++;
                }
        }
        /* cleanup */
@@ -428,13 +437,13 @@ static void thread_main()
 {
        struct thread me;
 
-       running++;
-       started++;
+       busy_thread_count++;
+       started_thread_count++;
        sig_monitor_init_timeouts();
        thread_run_internal(&me);
        sig_monitor_clean_timeouts();
-       started--;
-       running--;
+       started_thread_count--;
+       busy_thread_count--;
 }
 
 /**
@@ -483,7 +492,7 @@ static int start_one_thread()
  *                 The remaining parameter is the parameter 'arg1'
  *                 given here.
  * @param arg      The second argument for 'callback'
- * @param start    Allow to start a thread if not zero
+ * @param start    The start mode for threads
  * @return 0 in case of success or -1 in case of error
  */
 static int queue_job(
@@ -491,30 +500,33 @@ static int queue_job(
                int timeout,
                void (*callback)(int, void*),
                void *arg,
-               int start)
+               enum start_mode start_mode)
 {
        struct job *job;
        int rc;
 
        pthread_mutex_lock(&mutex);
 
-       /* allocates the job */
-       job = job_create(group, timeout, callback, arg);
-       if (!job)
-               goto error;
-
        /* check availability */
-       if (remains <= 0) {
+       if (remaining_job_count <= 0) {
                ERROR("can't process job with threads: too many jobs");
                errno = EBUSY;
-               goto error2;
+               goto error;
        }
 
+       /* allocates the job */
+       job = job_create(group, timeout, callback, arg);
+       if (!job)
+               goto error;
+
        /* start a thread if needed */
-       if (start && running == started && started < allowed) {
+       if (start_mode != Start_Lazy
+        && busy_thread_count == started_thread_count
+        && (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) {
+               if (rc < 0 && started_thread_count == 0) {
                        ERROR("can't start initial thread: %m");
                        goto error2;
                }
@@ -529,8 +541,8 @@ static int queue_job(
        return 0;
 
 error2:
-       job->next = free_jobs;
-       free_jobs = job;
+       job->next = first_free_job;
+       first_free_job = job;
 error:
        pthread_mutex_unlock(&mutex);
        return -1;
@@ -559,7 +571,59 @@ int jobs_queue(
                void (*callback)(int, void*),
                void *arg)
 {
-       return queue_job(group, timeout, callback, arg, 1);
+       return queue_job(group, timeout, callback, arg, Start_Default);
+}
+
+/**
+ * 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);
 }
 
 /**
@@ -760,7 +824,12 @@ struct evmgr *jobs_acquire_event_manager()
  * @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 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;
@@ -774,17 +843,18 @@ int jobs_start(int allowed_count, int start_count, int waiter_count, void (*star
        pthread_mutex_lock(&mutex);
 
        /* check whether already running */
-       if (current_thread || allowed) {
+       if (current_thread || allowed_thread_count) {
                ERROR("thread already started");
                errno = EINVAL;
                goto error;
        }
 
        /* records the allowed count */
-       allowed = allowed_count;
-       started = 0;
-       running = 0;
-       remains = waiter_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;
@@ -832,7 +902,7 @@ void jobs_exit(void (*handler)())
                t = t->next;
        }
 
-       /* wait the threads */
+       /* wake up the threads */
        pthread_cond_broadcast(&cond);
 
        /* leave */
index 60ee383..5ce422c 100644 (file)
@@ -26,6 +26,18 @@ extern int jobs_queue(
                void (*callback)(int signum, void* arg),
                void *arg);
 
+extern int jobs_queue_lazy(
+               const void *group,
+               int timeout,
+               void (*callback)(int signum, void* arg),
+               void *arg);
+
+extern int jobs_queue_urgent(
+               const void *group,
+               int timeout,
+               void (*callback)(int signum, void* arg),
+               void *arg);
+
 extern int jobs_enter(
                const void *group,
                int timeout,