remove experimental tags from all symbol definitions
[dpdk.git] / lib / librte_eal / common / eal_common_proc.c
index a329708..cbe8d10 100644 (file)
@@ -20,6 +20,7 @@
 #include <sys/un.h>
 #include <unistd.h>
 
+#include <rte_alarm.h>
 #include <rte_common.h>
 #include <rte_cycles.h>
 #include <rte_eal.h>
@@ -36,6 +37,7 @@ static int mp_fd = -1;
 static char mp_filter[PATH_MAX];   /* Filter for secondary process sockets */
 static char mp_dir_path[PATH_MAX]; /* The directory path for all mp sockets */
 static pthread_mutex_t mp_mutex_action = PTHREAD_MUTEX_INITIALIZER;
+static char peer_name[PATH_MAX];
 
 struct action_entry {
        TAILQ_ENTRY(action_entry) next;
@@ -94,11 +96,9 @@ TAILQ_HEAD(pending_request_list, pending_request);
 static struct {
        struct pending_request_list requests;
        pthread_mutex_t lock;
-       pthread_cond_t async_cond;
 } pending_requests = {
        .requests = TAILQ_HEAD_INITIALIZER(pending_requests.requests),
        .lock = PTHREAD_MUTEX_INITIALIZER,
-       .async_cond = PTHREAD_COND_INITIALIZER
        /**< used in async requests only */
 };
 
@@ -106,9 +106,19 @@ static struct {
 static int
 mp_send(struct rte_mp_msg *msg, const char *peer, int type);
 
+/* for use with alarm callback */
+static void
+async_reply_handle(void *arg);
+
+/* for use with process_msg */
+static struct pending_request *
+async_reply_handle_thread_unsafe(void *arg);
+
+static void
+trigger_async_action(struct pending_request *req);
 
 static struct pending_request *
-find_sync_request(const char *dst, const char *act_name)
+find_pending_request(const char *dst, const char *act_name)
 {
        struct pending_request *r;
 
@@ -187,14 +197,20 @@ validate_action_name(const char *name)
        return 0;
 }
 
-int __rte_experimental
+int
 rte_mp_action_register(const char *name, rte_mp_t action)
 {
        struct action_entry *entry;
 
-       if (validate_action_name(name))
+       if (validate_action_name(name) != 0)
                return -1;
 
+       if (internal_config.no_shconf) {
+               RTE_LOG(DEBUG, EAL, "No shared files mode enabled, IPC is disabled\n");
+               rte_errno = ENOTSUP;
+               return -1;
+       }
+
        entry = malloc(sizeof(struct action_entry));
        if (entry == NULL) {
                rte_errno = ENOMEM;
@@ -215,13 +231,18 @@ rte_mp_action_register(const char *name, rte_mp_t action)
        return 0;
 }
 
-void __rte_experimental
+void
 rte_mp_action_unregister(const char *name)
 {
        struct action_entry *entry;
 
-       if (validate_action_name(name))
+       if (validate_action_name(name) != 0)
+               return;
+
+       if (internal_config.no_shconf) {
+               RTE_LOG(DEBUG, EAL, "No shared files mode enabled, IPC is disabled\n");
                return;
+       }
 
        pthread_mutex_lock(&mp_mutex_action);
        entry = find_action_entry_by_name(name);
@@ -275,14 +296,22 @@ read_msg(struct mp_msg_internal *m, struct sockaddr_un *s)
                        break;
                }
        }
-
+       /* sanity-check the response */
+       if (m->msg.num_fds < 0 || m->msg.num_fds > RTE_MP_MAX_FD_NUM) {
+               RTE_LOG(ERR, EAL, "invalid number of fd's received\n");
+               return -1;
+       }
+       if (m->msg.len_param < 0 || m->msg.len_param > RTE_MP_MAX_PARAM_LEN) {
+               RTE_LOG(ERR, EAL, "invalid received data length\n");
+               return -1;
+       }
        return 0;
 }
 
 static void
 process_msg(struct mp_msg_internal *m, struct sockaddr_un *s)
 {
-       struct pending_request *sync_req;
+       struct pending_request *pending_req;
        struct action_entry *entry;
        struct rte_mp_msg *msg = &m->msg;
        rte_mp_t action = NULL;
@@ -290,21 +319,27 @@ process_msg(struct mp_msg_internal *m, struct sockaddr_un *s)
        RTE_LOG(DEBUG, EAL, "msg: %s\n", msg->name);
 
        if (m->type == MP_REP || m->type == MP_IGN) {
+               struct pending_request *req = NULL;
+
                pthread_mutex_lock(&pending_requests.lock);
-               sync_req = find_sync_request(s->sun_path, msg->name);
-               if (sync_req) {
-                       memcpy(sync_req->reply, msg, sizeof(*msg));
+               pending_req = find_pending_request(s->sun_path, msg->name);
+               if (pending_req) {
+                       memcpy(pending_req->reply, msg, sizeof(*msg));
                        /* -1 indicates that we've been asked to ignore */
-                       sync_req->reply_received = m->type == MP_REP ? 1 : -1;
-
-                       if (sync_req->type == REQUEST_TYPE_SYNC)
-                               pthread_cond_signal(&sync_req->sync.cond);
-                       else if (sync_req->type == REQUEST_TYPE_ASYNC)
-                               pthread_cond_signal(
-                                       &pending_requests.async_cond);
+                       pending_req->reply_received =
+                               m->type == MP_REP ? 1 : -1;
+
+                       if (pending_req->type == REQUEST_TYPE_SYNC)
+                               pthread_cond_signal(&pending_req->sync.cond);
+                       else if (pending_req->type == REQUEST_TYPE_ASYNC)
+                               req = async_reply_handle_thread_unsafe(
+                                               pending_req);
                } else
                        RTE_LOG(ERR, EAL, "Drop mp reply: %s\n", msg->name);
                pthread_mutex_unlock(&pending_requests.lock);
+
+               if (req != NULL)
+                       trigger_async_action(req);
                return;
        }
 
@@ -322,6 +357,7 @@ process_msg(struct mp_msg_internal *m, struct sockaddr_un *s)
                         * yet ready to process this request.
                         */
                        struct rte_mp_msg dummy;
+
                        memset(&dummy, 0, sizeof(dummy));
                        strlcpy(dummy.name, msg->name, sizeof(dummy.name));
                        mp_send(&dummy, s->sun_path, MP_IGN);
@@ -363,7 +399,6 @@ timespec_cmp(const struct timespec *a, const struct timespec *b)
 }
 
 enum async_action {
-       ACTION_NONE, /**< don't do anything */
        ACTION_FREE, /**< free the action entry, but don't trigger callback */
        ACTION_TRIGGER /**< trigger callback, then free action entry */
 };
@@ -373,7 +408,7 @@ process_async_request(struct pending_request *sr, const struct timespec *now)
 {
        struct async_request_param *param;
        struct rte_mp_reply *reply;
-       bool timeout, received, last_msg;
+       bool timeout, last_msg;
 
        param = sr->async.param;
        reply = &param->user_reply;
@@ -381,13 +416,6 @@ process_async_request(struct pending_request *sr, const struct timespec *now)
        /* did we timeout? */
        timeout = timespec_cmp(&param->end, now) <= 0;
 
-       /* did we receive a response? */
-       received = sr->reply_received != 0;
-
-       /* if we didn't time out, and we didn't receive a response, ignore */
-       if (!timeout && !received)
-               return ACTION_NONE;
-
        /* if we received a response, adjust relevant data and copy mesasge. */
        if (sr->reply_received == 1 && sr->reply) {
                struct rte_mp_msg *msg, *user_msgs, *tmp;
@@ -417,7 +445,13 @@ process_async_request(struct pending_request *sr, const struct timespec *now)
        } else if (sr->reply_received == -1) {
                /* we were asked to ignore this process */
                reply->nb_sent--;
+       } else if (timeout) {
+               /* count it as processed response, but don't increment
+                * nb_received.
+                */
+               param->n_responses_processed++;
        }
+
        free(sr->reply);
 
        last_msg = param->n_responses_processed == reply->nb_sent;
@@ -440,126 +474,66 @@ trigger_async_action(struct pending_request *sr)
        free(sr->async.param->user_reply.msgs);
        free(sr->async.param);
        free(sr->request);
+       free(sr);
 }
 
 static struct pending_request *
-check_trigger(struct timespec *ts)
+async_reply_handle_thread_unsafe(void *arg)
 {
-       struct pending_request *next, *cur, *trigger = NULL;
-
-       TAILQ_FOREACH_SAFE(cur, &pending_requests.requests, next, next) {
-               enum async_action action;
-               if (cur->type != REQUEST_TYPE_ASYNC)
-                       continue;
+       struct pending_request *req = (struct pending_request *)arg;
+       enum async_action action;
+       struct timespec ts_now;
+       struct timeval now;
 
-               action = process_async_request(cur, ts);
-               if (action == ACTION_FREE) {
-                       TAILQ_REMOVE(&pending_requests.requests, cur, next);
-                       free(cur);
-               } else if (action == ACTION_TRIGGER) {
-                       TAILQ_REMOVE(&pending_requests.requests, cur, next);
-                       trigger = cur;
-                       break;
-               }
+       if (gettimeofday(&now, NULL) < 0) {
+               RTE_LOG(ERR, EAL, "Cannot get current time\n");
+               goto no_trigger;
        }
-       return trigger;
-}
+       ts_now.tv_nsec = now.tv_usec * 1000;
+       ts_now.tv_sec = now.tv_sec;
 
-static void
-wait_for_async_messages(void)
-{
-       struct pending_request *sr;
-       struct timespec timeout;
-       bool timedwait = false;
-       bool nowait = false;
-       int ret;
+       action = process_async_request(req, &ts_now);
 
-       /* scan through the list and see if there are any timeouts that
-        * are earlier than our current timeout.
-        */
-       TAILQ_FOREACH(sr, &pending_requests.requests, next) {
-               if (sr->type != REQUEST_TYPE_ASYNC)
-                       continue;
-               if (!timedwait || timespec_cmp(&sr->async.param->end,
-                               &timeout) < 0) {
-                       memcpy(&timeout, &sr->async.param->end,
-                               sizeof(timeout));
-                       timedwait = true;
-               }
+       TAILQ_REMOVE(&pending_requests.requests, req, next);
 
-               /* sometimes, we don't even wait */
-               if (sr->reply_received) {
-                       nowait = true;
-                       break;
+       if (rte_eal_alarm_cancel(async_reply_handle, req) < 0) {
+               /* if we failed to cancel the alarm because it's already in
+                * progress, don't proceed because otherwise we will end up
+                * handling the same message twice.
+                */
+               if (rte_errno == EINPROGRESS) {
+                       RTE_LOG(DEBUG, EAL, "Request handling is already in progress\n");
+                       goto no_trigger;
                }
+               RTE_LOG(ERR, EAL, "Failed to cancel alarm\n");
        }
 
-       if (nowait)
-               return;
-
-       do {
-               ret = timedwait ?
-                       pthread_cond_timedwait(
-                               &pending_requests.async_cond,
-                               &pending_requests.lock,
-                               &timeout) :
-                       pthread_cond_wait(
-                               &pending_requests.async_cond,
-                               &pending_requests.lock);
-       } while (ret != 0 && ret != ETIMEDOUT);
-
-       /* we've been woken up or timed out */
+       if (action == ACTION_TRIGGER)
+               return req;
+no_trigger:
+       free(req);
+       return NULL;
 }
 
-static void *
-async_reply_handle(void *arg __rte_unused)
+static void
+async_reply_handle(void *arg)
 {
-       struct timeval now;
-       struct timespec ts_now;
-       while (1) {
-               struct pending_request *trigger = NULL;
+       struct pending_request *req;
 
-               pthread_mutex_lock(&pending_requests.lock);
-
-               /* we exit this function holding the lock */
-               wait_for_async_messages();
-
-               if (gettimeofday(&now, NULL) < 0) {
-                       pthread_mutex_unlock(&pending_requests.lock);
-                       RTE_LOG(ERR, EAL, "Cannot get current time\n");
-                       break;
-               }
-               ts_now.tv_nsec = now.tv_usec * 1000;
-               ts_now.tv_sec = now.tv_sec;
-
-               do {
-                       trigger = check_trigger(&ts_now);
-                       /* unlock request list */
-                       pthread_mutex_unlock(&pending_requests.lock);
-
-                       if (trigger) {
-                               trigger_async_action(trigger);
-                               free(trigger);
-
-                               /* we've triggered a callback, but there may be
-                                * more, so lock the list and check again.
-                                */
-                               pthread_mutex_lock(&pending_requests.lock);
-                       }
-               } while (trigger);
-       }
-
-       RTE_LOG(ERR, EAL, "ERROR: asynchronous requests disabled\n");
+       pthread_mutex_lock(&pending_requests.lock);
+       req = async_reply_handle_thread_unsafe(arg);
+       pthread_mutex_unlock(&pending_requests.lock);
 
-       return NULL;
+       if (req != NULL)
+               trigger_async_action(req);
 }
 
 static int
 open_socket_fd(void)
 {
-       char peer_name[PATH_MAX] = {0};
        struct sockaddr_un un;
 
+       peer_name[0] = '\0';
        if (rte_eal_process_type() == RTE_PROC_SECONDARY)
                snprintf(peer_name, sizeof(peer_name),
                                "%d_%"PRIx64, getpid(), rte_rdtsc());
@@ -588,36 +562,34 @@ open_socket_fd(void)
        return mp_fd;
 }
 
-static int
-unlink_sockets(const char *filter)
+static void
+close_socket_fd(void)
 {
-       int dir_fd;
-       DIR *mp_dir;
-       struct dirent *ent;
-
-       mp_dir = opendir(mp_dir_path);
-       if (!mp_dir) {
-               RTE_LOG(ERR, EAL, "Unable to open directory %s\n", mp_dir_path);
-               return -1;
-       }
-       dir_fd = dirfd(mp_dir);
+       char path[PATH_MAX];
 
-       while ((ent = readdir(mp_dir))) {
-               if (fnmatch(filter, ent->d_name, 0) == 0)
-                       unlinkat(dir_fd, ent->d_name, 0);
-       }
+       if (mp_fd < 0)
+               return;
 
-       closedir(mp_dir);
-       return 0;
+       close(mp_fd);
+       create_socket_path(peer_name, path, sizeof(path));
+       unlink(path);
 }
 
 int
 rte_mp_channel_init(void)
 {
-       char thread_name[RTE_MAX_THREAD_NAME_LEN];
        char path[PATH_MAX];
        int dir_fd;
-       pthread_t mp_handle_tid, async_reply_handle_tid;
+       pthread_t mp_handle_tid;
+
+       /* in no shared files mode, we do not have secondary processes support,
+        * so no need to initialize IPC.
+        */
+       if (internal_config.no_shconf) {
+               RTE_LOG(DEBUG, EAL, "No shared files mode enabled, IPC will be disabled\n");
+               rte_errno = ENOTSUP;
+               return -1;
+       }
 
        /* create filter path */
        create_socket_path("*", path, sizeof(path));
@@ -642,19 +614,13 @@ rte_mp_channel_init(void)
                return -1;
        }
 
-       if (rte_eal_process_type() == RTE_PROC_PRIMARY &&
-                       unlink_sockets(mp_filter)) {
-               RTE_LOG(ERR, EAL, "failed to unlink mp sockets\n");
-               close(dir_fd);
-               return -1;
-       }
-
        if (open_socket_fd() < 0) {
                close(dir_fd);
                return -1;
        }
 
-       if (pthread_create(&mp_handle_tid, NULL, mp_handle, NULL) < 0) {
+       if (rte_ctrl_thread_create(&mp_handle_tid, "rte_mp_handle",
+                       NULL, mp_handle, NULL) < 0) {
                RTE_LOG(ERR, EAL, "failed to create mp thead: %s\n",
                        strerror(errno));
                close(mp_fd);
@@ -663,24 +629,6 @@ rte_mp_channel_init(void)
                return -1;
        }
 
-       if (pthread_create(&async_reply_handle_tid, NULL,
-                       async_reply_handle, NULL) < 0) {
-               RTE_LOG(ERR, EAL, "failed to create mp thead: %s\n",
-                       strerror(errno));
-               close(mp_fd);
-               close(dir_fd);
-               mp_fd = -1;
-               return -1;
-       }
-
-       /* try best to set thread name */
-       strlcpy(thread_name, "rte_mp_handle", RTE_MAX_THREAD_NAME_LEN);
-       rte_thread_setname(mp_handle_tid, thread_name);
-
-       /* try best to set thread name */
-       strlcpy(thread_name, "rte_mp_async_handle", RTE_MAX_THREAD_NAME_LEN);
-       rte_thread_setname(async_reply_handle_tid, thread_name);
-
        /* unlock the directory */
        flock(dir_fd, LOCK_UN);
        close(dir_fd);
@@ -688,6 +636,12 @@ rte_mp_channel_init(void)
        return 0;
 }
 
+void
+rte_mp_channel_cleanup(void)
+{
+       close_socket_fd();
+}
+
 /**
  * Return -1, as fail to send message and it's caused by the local side.
  * Return 0, as fail to send message and it's caused by the remote side.
@@ -744,11 +698,6 @@ send_msg(const char *dst_path, struct rte_mp_msg *msg, int type)
                        unlink(dst_path);
                        return 0;
                }
-               if (errno == ENOBUFS) {
-                       RTE_LOG(ERR, EAL, "Peer cannot receive message %s\n",
-                               dst_path);
-                       return 0;
-               }
                RTE_LOG(ERR, EAL, "failed to send to (%s) due to %s\n",
                        dst_path, strerror(errno));
                return -1;
@@ -785,7 +734,7 @@ mp_send(struct rte_mp_msg *msg, const char *peer, int type)
 
        dir_fd = dirfd(mp_dir);
        /* lock the directory to prevent processes spinning up while we send */
-       if (flock(dir_fd, LOCK_EX)) {
+       if (flock(dir_fd, LOCK_SH)) {
                RTE_LOG(ERR, EAL, "Unable to lock directory %s\n",
                        mp_dir_path);
                rte_errno = errno;
@@ -812,73 +761,88 @@ mp_send(struct rte_mp_msg *msg, const char *peer, int type)
        return ret;
 }
 
-static bool
+static int
 check_input(const struct rte_mp_msg *msg)
 {
        if (msg == NULL) {
                RTE_LOG(ERR, EAL, "Msg cannot be NULL\n");
                rte_errno = EINVAL;
-               return false;
+               return -1;
        }
 
-       if (validate_action_name(msg->name))
-               return false;
+       if (validate_action_name(msg->name) != 0)
+               return -1;
+
+       if (msg->len_param < 0) {
+               RTE_LOG(ERR, EAL, "Message data length is negative\n");
+               rte_errno = EINVAL;
+               return -1;
+       }
+
+       if (msg->num_fds < 0) {
+               RTE_LOG(ERR, EAL, "Number of fd's is negative\n");
+               rte_errno = EINVAL;
+               return -1;
+       }
 
        if (msg->len_param > RTE_MP_MAX_PARAM_LEN) {
                RTE_LOG(ERR, EAL, "Message data is too long\n");
                rte_errno = E2BIG;
-               return false;
+               return -1;
        }
 
        if (msg->num_fds > RTE_MP_MAX_FD_NUM) {
                RTE_LOG(ERR, EAL, "Cannot send more than %d FDs\n",
                        RTE_MP_MAX_FD_NUM);
                rte_errno = E2BIG;
-               return false;
+               return -1;
        }
 
-       return true;
+       return 0;
 }
 
-int __rte_experimental
+int
 rte_mp_sendmsg(struct rte_mp_msg *msg)
 {
-       if (!check_input(msg))
+       if (check_input(msg) != 0)
                return -1;
 
+       if (internal_config.no_shconf) {
+               RTE_LOG(DEBUG, EAL, "No shared files mode enabled, IPC is disabled\n");
+               rte_errno = ENOTSUP;
+               return -1;
+       }
+
        RTE_LOG(DEBUG, EAL, "sendmsg: %s\n", msg->name);
        return mp_send(msg, NULL, MP_MSG);
 }
 
 static int
 mp_request_async(const char *dst, struct rte_mp_msg *req,
-               struct async_request_param *param)
+               struct async_request_param *param, const struct timespec *ts)
 {
        struct rte_mp_msg *reply_msg;
-       struct pending_request *sync_req, *exist;
-       int ret;
+       struct pending_request *pending_req, *exist;
+       int ret = -1;
 
-       sync_req = malloc(sizeof(*sync_req));
-       reply_msg = malloc(sizeof(*reply_msg));
-       if (sync_req == NULL || reply_msg == NULL) {
+       pending_req = calloc(1, sizeof(*pending_req));
+       reply_msg = calloc(1, sizeof(*reply_msg));
+       if (pending_req == NULL || reply_msg == NULL) {
                RTE_LOG(ERR, EAL, "Could not allocate space for sync request\n");
                rte_errno = ENOMEM;
                ret = -1;
                goto fail;
        }
 
-       memset(sync_req, 0, sizeof(*sync_req));
-       memset(reply_msg, 0, sizeof(*reply_msg));
-
-       sync_req->type = REQUEST_TYPE_ASYNC;
-       strlcpy(sync_req->dst, dst, sizeof(sync_req->dst));
-       sync_req->request = req;
-       sync_req->reply = reply_msg;
-       sync_req->async.param = param;
+       pending_req->type = REQUEST_TYPE_ASYNC;
+       strlcpy(pending_req->dst, dst, sizeof(pending_req->dst));
+       pending_req->request = req;
+       pending_req->reply = reply_msg;
+       pending_req->async.param = param;
 
        /* queue already locked by caller */
 
-       exist = find_sync_request(dst, req->name);
+       exist = find_pending_request(dst, req->name);
        if (exist) {
                RTE_LOG(ERR, EAL, "A pending request %s:%s\n", dst, req->name);
                rte_errno = EEXIST;
@@ -896,13 +860,21 @@ mp_request_async(const char *dst, struct rte_mp_msg *req,
                ret = 0;
                goto fail;
        }
-       TAILQ_INSERT_TAIL(&pending_requests.requests, sync_req, next);
-
        param->user_reply.nb_sent++;
 
+       /* if alarm set fails, we simply ignore the reply */
+       if (rte_eal_alarm_set(ts->tv_sec * 1000000 + ts->tv_nsec / 1000,
+                             async_reply_handle, pending_req) < 0) {
+               RTE_LOG(ERR, EAL, "Fail to set alarm for request %s:%s\n",
+                       dst, req->name);
+               ret = -1;
+               goto fail;
+       }
+       TAILQ_INSERT_TAIL(&pending_requests.requests, pending_req, next);
+
        return 0;
 fail:
-       free(sync_req);
+       free(pending_req);
        free(reply_msg);
        return ret;
 }
@@ -913,16 +885,16 @@ mp_request_sync(const char *dst, struct rte_mp_msg *req,
 {
        int ret;
        struct rte_mp_msg msg, *tmp;
-       struct pending_request sync_req, *exist;
+       struct pending_request pending_req, *exist;
 
-       sync_req.type = REQUEST_TYPE_SYNC;
-       sync_req.reply_received = 0;
-       strlcpy(sync_req.dst, dst, sizeof(sync_req.dst));
-       sync_req.request = req;
-       sync_req.reply = &msg;
-       pthread_cond_init(&sync_req.sync.cond, NULL);
+       pending_req.type = REQUEST_TYPE_SYNC;
+       pending_req.reply_received = 0;
+       strlcpy(pending_req.dst, dst, sizeof(pending_req.dst));
+       pending_req.request = req;
+       pending_req.reply = &msg;
+       pthread_cond_init(&pending_req.sync.cond, NULL);
 
-       exist = find_sync_request(dst, req->name);
+       exist = find_pending_request(dst, req->name);
        if (exist) {
                RTE_LOG(ERR, EAL, "A pending request %s:%s\n", dst, req->name);
                rte_errno = EEXIST;
@@ -937,24 +909,24 @@ mp_request_sync(const char *dst, struct rte_mp_msg *req,
        } else if (ret == 0)
                return 0;
 
-       TAILQ_INSERT_TAIL(&pending_requests.requests, &sync_req, next);
+       TAILQ_INSERT_TAIL(&pending_requests.requests, &pending_req, next);
 
        reply->nb_sent++;
 
        do {
-               ret = pthread_cond_timedwait(&sync_req.sync.cond,
+               ret = pthread_cond_timedwait(&pending_req.sync.cond,
                                &pending_requests.lock, ts);
        } while (ret != 0 && ret != ETIMEDOUT);
 
-       TAILQ_REMOVE(&pending_requests.requests, &sync_req, next);
+       TAILQ_REMOVE(&pending_requests.requests, &pending_req, next);
 
-       if (sync_req.reply_received == 0) {
+       if (pending_req.reply_received == 0) {
                RTE_LOG(ERR, EAL, "Fail to recv reply for request %s:%s\n",
                        dst, req->name);
                rte_errno = ETIMEDOUT;
                return -1;
        }
-       if (sync_req.reply_received == -1) {
+       if (pending_req.reply_received == -1) {
                RTE_LOG(DEBUG, EAL, "Asked to ignore response\n");
                /* not receiving this message is not an error, so decrement
                 * number of sent messages
@@ -976,11 +948,11 @@ mp_request_sync(const char *dst, struct rte_mp_msg *req,
        return 0;
 }
 
-int __rte_experimental
+int
 rte_mp_request_sync(struct rte_mp_msg *req, struct rte_mp_reply *reply,
                const struct timespec *ts)
 {
-       int dir_fd, ret = 0;
+       int dir_fd, ret = -1;
        DIR *mp_dir;
        struct dirent *ent;
        struct timeval now;
@@ -988,28 +960,35 @@ rte_mp_request_sync(struct rte_mp_msg *req, struct rte_mp_reply *reply,
 
        RTE_LOG(DEBUG, EAL, "request: %s\n", req->name);
 
-       if (check_input(req) == false)
+       reply->nb_sent = 0;
+       reply->nb_received = 0;
+       reply->msgs = NULL;
+
+       if (check_input(req) != 0)
+               goto end;
+
+       if (internal_config.no_shconf) {
+               RTE_LOG(DEBUG, EAL, "No shared files mode enabled, IPC is disabled\n");
+               rte_errno = ENOTSUP;
                return -1;
+       }
+
        if (gettimeofday(&now, NULL) < 0) {
-               RTE_LOG(ERR, EAL, "Faile to get current time\n");
+               RTE_LOG(ERR, EAL, "Failed to get current time\n");
                rte_errno = errno;
-               return -1;
+               goto end;
        }
 
        end.tv_nsec = (now.tv_usec * 1000 + ts->tv_nsec) % 1000000000;
        end.tv_sec = now.tv_sec + ts->tv_sec +
                        (now.tv_usec * 1000 + ts->tv_nsec) / 1000000000;
 
-       reply->nb_sent = 0;
-       reply->nb_received = 0;
-       reply->msgs = NULL;
-
        /* for secondary process, send request to the primary process only */
        if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
                pthread_mutex_lock(&pending_requests.lock);
                ret = mp_request_sync(eal_mp_socket_path(), req, reply, &end);
                pthread_mutex_unlock(&pending_requests.lock);
-               return ret;
+               goto end;
        }
 
        /* for primary process, broadcast request, and collect reply 1 by 1 */
@@ -1017,17 +996,16 @@ rte_mp_request_sync(struct rte_mp_msg *req, struct rte_mp_reply *reply,
        if (!mp_dir) {
                RTE_LOG(ERR, EAL, "Unable to open directory %s\n", mp_dir_path);
                rte_errno = errno;
-               return -1;
+               goto end;
        }
 
        dir_fd = dirfd(mp_dir);
        /* lock the directory to prevent processes spinning up while we send */
-       if (flock(dir_fd, LOCK_EX)) {
+       if (flock(dir_fd, LOCK_SH)) {
                RTE_LOG(ERR, EAL, "Unable to lock directory %s\n",
                        mp_dir_path);
-               closedir(mp_dir);
                rte_errno = errno;
-               return -1;
+               goto close_end;
        }
 
        pthread_mutex_lock(&pending_requests.lock);
@@ -1044,18 +1022,29 @@ rte_mp_request_sync(struct rte_mp_msg *req, struct rte_mp_reply *reply,
                 * locks on receive
                 */
                if (mp_request_sync(path, req, reply, &end))
-                       ret = -1;
+                       goto unlock_end;
        }
+       ret = 0;
+
+unlock_end:
        pthread_mutex_unlock(&pending_requests.lock);
        /* unlock the directory */
        flock(dir_fd, LOCK_UN);
 
+close_end:
        /* dir_fd automatically closed on closedir */
        closedir(mp_dir);
+
+end:
+       if (ret) {
+               free(reply->msgs);
+               reply->nb_received = 0;
+               reply->msgs = NULL;
+       }
        return ret;
 }
 
-int __rte_experimental
+int
 rte_mp_request_async(struct rte_mp_msg *req, const struct timespec *ts,
                rte_mp_async_reply_t clb)
 {
@@ -1072,26 +1061,29 @@ rte_mp_request_async(struct rte_mp_msg *req, const struct timespec *ts,
 
        RTE_LOG(DEBUG, EAL, "request: %s\n", req->name);
 
-       if (check_input(req) == false)
+       if (check_input(req) != 0)
                return -1;
+
+       if (internal_config.no_shconf) {
+               RTE_LOG(DEBUG, EAL, "No shared files mode enabled, IPC is disabled\n");
+               rte_errno = ENOTSUP;
+               return -1;
+       }
+
        if (gettimeofday(&now, NULL) < 0) {
                RTE_LOG(ERR, EAL, "Faile to get current time\n");
                rte_errno = errno;
                return -1;
        }
-       copy = malloc(sizeof(*copy));
-       dummy = malloc(sizeof(*dummy));
-       param = malloc(sizeof(*param));
+       copy = calloc(1, sizeof(*copy));
+       dummy = calloc(1, sizeof(*dummy));
+       param = calloc(1, sizeof(*param));
        if (copy == NULL || dummy == NULL || param == NULL) {
                RTE_LOG(ERR, EAL, "Failed to allocate memory for async reply\n");
                rte_errno = ENOMEM;
                goto fail;
        }
 
-       memset(copy, 0, sizeof(*copy));
-       memset(dummy, 0, sizeof(*dummy));
-       memset(param, 0, sizeof(*param));
-
        /* copy message */
        memcpy(copy, req, sizeof(*copy));
 
@@ -1125,7 +1117,7 @@ rte_mp_request_async(struct rte_mp_msg *req, const struct timespec *ts,
 
        /* for secondary process, send request to the primary process only */
        if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
-               ret = mp_request_async(eal_mp_socket_path(), copy, param);
+               ret = mp_request_async(eal_mp_socket_path(), copy, param, ts);
 
                /* if we didn't send anything, put dummy request on the queue */
                if (ret == 0 && reply->nb_sent == 0) {
@@ -1152,7 +1144,7 @@ rte_mp_request_async(struct rte_mp_msg *req, const struct timespec *ts,
        dir_fd = dirfd(mp_dir);
 
        /* lock the directory to prevent processes spinning up while we send */
-       if (flock(dir_fd, LOCK_EX)) {
+       if (flock(dir_fd, LOCK_SH)) {
                RTE_LOG(ERR, EAL, "Unable to lock directory %s\n",
                        mp_dir_path);
                rte_errno = errno;
@@ -1168,7 +1160,7 @@ rte_mp_request_async(struct rte_mp_msg *req, const struct timespec *ts,
                snprintf(path, sizeof(path), "%s/%s", mp_dir_path,
                         ent->d_name);
 
-               if (mp_request_async(path, copy, param))
+               if (mp_request_async(path, copy, param, ts))
                        ret = -1;
        }
        /* if we didn't send anything, put dummy request on the queue */
@@ -1177,9 +1169,6 @@ rte_mp_request_async(struct rte_mp_msg *req, const struct timespec *ts,
                dummy_used = true;
        }
 
-       /* trigger async request thread wake up */
-       pthread_cond_signal(&pending_requests.async_cond);
-
        /* finally, unlock the queue */
        pthread_mutex_unlock(&pending_requests.lock);
 
@@ -1205,12 +1194,12 @@ fail:
        return -1;
 }
 
-int __rte_experimental
+int
 rte_mp_reply(struct rte_mp_msg *msg, const char *peer)
 {
        RTE_LOG(DEBUG, EAL, "reply: %s\n", msg->name);
 
-       if (check_input(msg) == false)
+       if (check_input(msg) != 0)
                return -1;
 
        if (peer == NULL) {
@@ -1219,5 +1208,10 @@ rte_mp_reply(struct rte_mp_msg *msg, const char *peer)
                return -1;
        }
 
+       if (internal_config.no_shconf) {
+               RTE_LOG(DEBUG, EAL, "No shared files mode enabled, IPC is disabled\n");
+               return 0;
+       }
+
        return mp_send(msg, peer, MP_REP);
 }