/* Copyright (c) 2012 Red Hat, Inc. This file is part of GlusterFS. This file is licensed to you under your choice of the GNU Lesser General Public License, version 3 or any later version (LGPLv3 or later), or the GNU General Public License, version 2 (GPLv2), in all cases as published by the Free Software Foundation. */ #include #include #include #include #include #include #include #include "logging.h" #include "event.h" #include "mem-pool.h" #include "common-utils.h" #include "libglusterfs-messages.h" #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" #endif #ifdef HAVE_SYS_EPOLL_H #include struct event_slot_epoll { int fd; int events; int gen; int ref; int do_close; int in_handler; void *data; event_handler_t handler; gf_lock_t lock; }; struct event_thread_data { struct event_pool *event_pool; int event_index; }; static struct event_slot_epoll * __event_newtable (struct event_pool *event_pool, int table_idx) { struct event_slot_epoll *table = NULL; int i = -1; table = GF_CALLOC (sizeof (*table), EVENT_EPOLL_SLOTS, gf_common_mt_ereg); if (!table) return NULL; for (i = 0; i < EVENT_EPOLL_SLOTS; i++) { table[i].fd = -1; LOCK_INIT (&table[i].lock); } event_pool->ereg[table_idx] = table; event_pool->slots_used[table_idx] = 0; return table; } static int __event_slot_alloc (struct event_pool *event_pool, int fd) { int i = 0; int table_idx = -1; int gen = -1; struct event_slot_epoll *table = NULL; for (i = 0; i < EVENT_EPOLL_TABLES; i++) { switch (event_pool->slots_used[i]) { case EVENT_EPOLL_SLOTS: continue; case 0: if (!event_pool->ereg[i]) { table = __event_newtable (event_pool, i); if (!table) return -1; } else { table = event_pool->ereg[i]; } break; default: table = event_pool->ereg[i]; break; } if (table) /* break out of the loop */ break; } if (!table) return -1; table_idx = i; for (i = 0; i < EVENT_EPOLL_SLOTS; i++) { if (table[i].fd == -1) { /* wipe everything except bump the generation */ gen = table[i].gen; memset (&table[i], 0, sizeof (table[i])); table[i].gen = gen + 1; LOCK_INIT (&table[i].lock); table[i].fd = fd; event_pool->slots_used[table_idx]++; break; } } return table_idx * EVENT_EPOLL_SLOTS + i; } static int event_slot_alloc (struct event_pool *event_pool, int fd) { int idx = -1; pthread_mutex_lock (&event_pool->mutex); { idx = __event_slot_alloc (event_pool, fd); } pthread_mutex_unlock (&event_pool->mutex); return idx; } static void __event_slot_dealloc (struct event_pool *event_pool, int idx) { int table_idx = 0; int offset = 0; struct event_slot_epoll *table = NULL; struct event_slot_epoll *slot = NULL; table_idx = idx / EVENT_EPOLL_SLOTS; offset = idx % EVENT_EPOLL_SLOTS; table = event_pool->ereg[table_idx]; if (!table) return; slot = &table[offset]; slot->gen++; slot->fd = -1; event_pool->slots_used[table_idx]--; return; } static void event_slot_dealloc (struct event_pool *event_pool, int idx) { pthread_mutex_lock (&event_pool->mutex); { __event_slot_dealloc (event_pool, idx); } pthread_mutex_unlock (&event_pool->mutex); return; } static struct event_slot_epoll * event_slot_get (struct event_pool *event_pool, int idx) { struct event_slot_epoll *slot = NULL; struct event_slot_epoll *table = NULL; int table_idx = 0; int offset = 0; table_idx = idx / EVENT_EPOLL_SLOTS; offset = idx % EVENT_EPOLL_SLOTS; table = event_pool->ereg[table_idx]; if (!table) return NULL; slot = &table[offset]; LOCK (&slot->lock); { slot->ref++; } UNLOCK (&slot->lock); return slot; } static void event_slot_unref (struct event_pool *event_pool, struct event_slot_epoll *slot, int idx) { int ref = -1; int fd = -1; int do_close = 0; LOCK (&slot->lock); { ref = --slot->ref; fd = slot->fd; do_close = slot->do_close; } UNLOCK (&slot->lock); if (ref) /* slot still alive */ goto done; event_slot_dealloc (event_pool, idx); if (do_close) close (fd); done: return; } static struct event_pool * event_pool_new_epoll (int count, int eventthreadcount) { struct event_pool *event_pool = NULL; int epfd = -1; event_pool = GF_CALLOC (1, sizeof (*event_pool), gf_common_mt_event_pool); if (!event_pool) goto out; epfd = epoll_create (count); if (epfd == -1) { gf_msg ("epoll", GF_LOG_ERROR, errno, LG_MSG_EPOLL_FD_CREATE_FAILED, "epoll fd creation " "failed"); GF_FREE (event_pool->reg); GF_FREE (event_pool); event_pool = NULL; goto out; } event_pool->fd = epfd; event_pool->count = count; event_pool->eventthreadcount = eventthreadcount; pthread_mutex_init (&event_pool->mutex, NULL); out: return event_pool; } static void __slot_update_events (struct event_slot_epoll *slot, int poll_in, int poll_out) { switch (poll_in) { case 1: slot->events |= EPOLLIN; break; case 0: slot->events &= ~EPOLLIN; break; case -1: /* do nothing */ break; default: gf_msg ("epoll", GF_LOG_ERROR, 0, LG_MSG_INVALID_POLL_IN, "invalid poll_in value %d", poll_in); break; } switch (poll_out) { case 1: slot->events |= EPOLLOUT; break; case 0: slot->events &= ~EPOLLOUT; break; case -1: /* do nothing */ break; default: gf_msg ("epoll", GF_LOG_ERROR, 0, LG_MSG_INVALID_POLL_OUT, "invalid poll_out value %d", poll_out); break; } } int event_register_epoll (struct event_pool *event_pool, int fd, event_handler_t handler, void *data, int poll_in, int poll_out) { int idx = -1; int ret = -1; int destroy = 0; struct epoll_event epoll_event = {0, }; struct event_data *ev_data = (void *)&epoll_event.data; struct event_slot_epoll *slot = NULL; GF_VALIDATE_OR_GOTO ("event", event_pool, out); /* TODO: Even with the below check, there is a possiblity of race, * What if the destroy mode is set after the check is done. * Not sure of the best way to prevent this race, ref counting * is one possibility. * There is no harm in registering and unregistering the fd * even after destroy mode is set, just that such fds will remain * open until unregister is called, also the events on that fd will be * notified, until one of the poller thread is alive. */ pthread_mutex_lock (&event_pool->mutex); { destroy = event_pool->destroy; } pthread_mutex_unlock (&event_pool->mutex); if (destroy == 1) goto out; idx = event_slot_alloc (event_pool, fd); if (idx == -1) { gf_msg ("epoll", GF_LOG_ERROR, 0, LG_MSG_SLOT_NOT_FOUND, "could not find slot for fd=%d", fd); return -1; } slot = event_slot_get (event_pool, idx); assert (slot->fd == fd); LOCK (&slot->lock); { /* make epoll 'singleshot', which means we need to re-add the fd with epoll_ctl(EPOLL_CTL_MOD) after delivery of every single event. This assures us that while a poller thread has picked up and is processing an event, another poller will not try to pick this at the same time as well. */ slot->events = EPOLLPRI | EPOLLONESHOT; slot->handler = handler; slot->data = data; __slot_update_events (slot, poll_in, poll_out); epoll_event.events = slot->events; ev_data->idx = idx; ev_data->gen = slot->gen; ret = epoll_ctl (event_pool->fd, EPOLL_CTL_ADD, fd, &epoll_event); /* check ret after UNLOCK() to avoid deadlock in event_slot_unref() */ } UNLOCK (&slot->lock); if (ret == -1) { gf_msg ("epoll", GF_LOG_ERROR, errno, LG_MSG_EPOLL_FD_ADD_FAILED, "failed to add fd(=%d) to " "epoll fd(=%d)", fd, event_pool->fd); event_slot_unref (event_pool, slot, idx); idx = -1; } /* keep slot->ref (do not event_slot_unref) if successful */ out: return idx; } static int event_unregister_epoll_common (struct event_pool *event_pool, int fd, int idx, int do_close) { int ret = -1; struct event_slot_epoll *slot = NULL; GF_VALIDATE_OR_GOTO ("event", event_pool, out); slot = event_slot_get (event_pool, idx); assert (slot->fd == fd); LOCK (&slot->lock); { ret = epoll_ctl (event_pool->fd, EPOLL_CTL_DEL, fd, NULL); if (ret == -1) { gf_msg ("epoll", GF_LOG_ERROR, errno, LG_MSG_EPOLL_FD_DEL_FAILED, "fail to del " "fd(=%d) from epoll fd(=%d)", fd, event_pool->fd); goto unlock; } slot->do_close = do_close; slot->gen++; /* detect unregister in dispatch_handler() */ } unlock: UNLOCK (&slot->lock); event_slot_unref (event_pool, slot, idx); /* one for event_register() */ event_slot_unref (event_pool, slot, idx); /* one for event_slot_get() */ out: return ret; } static int event_unregister_epoll (struct event_pool *event_pool, int fd, int idx_hint) { int ret = -1; ret = event_unregister_epoll_common (event_pool, fd, idx_hint, 0); return ret; } static int event_unregister_close_epoll (struct event_pool *event_pool, int fd, int idx_hint) { int ret = -1; ret = event_unregister_epoll_common (event_pool, fd, idx_hint, 1); return ret; } static int event_select_on_epoll (struct event_pool *event_pool, int fd, int idx, int poll_in, int poll_out) { int ret = -1; struct event_slot_epoll *slot = NULL; struct epoll_event epoll_event = {0, }; struct event_data *ev_data = (void *)&epoll_event.data; GF_VALIDATE_OR_GOTO ("event", event_pool, out); slot = event_slot_get (event_pool, idx); assert (slot->fd == fd); LOCK (&slot->lock); { __slot_update_events (slot, poll_in, poll_out); epoll_event.events = slot->events; ev_data->idx = idx; ev_data->gen = slot->gen; if (slot->in_handler) /* in_handler indicates at least one thread executing event_dispatch_epoll_handler() which will perform epoll_ctl(EPOLL_CTL_MOD) anyways (because of EPOLLET) This not only saves a system call, but also avoids possibility of another epoll thread parallely picking up the next event while the ongoing handler is still in progress (and resulting in unnecessary contention on rpc_transport_t->mutex). */ goto unlock; ret = epoll_ctl (event_pool->fd, EPOLL_CTL_MOD, fd, &epoll_event); if (ret == -1) { gf_msg ("epoll", GF_LOG_ERROR, errno, LG_MSG_EPOLL_FD_MODIFY_FAILED, "failed to " "modify fd(=%d) events to %d", fd, epoll_event.events); } } unlock: UNLOCK (&slot->lock); event_slot_unref (event_pool, slot, idx); out: return idx; } static int event_dispatch_epoll_handler (struct event_pool *event_pool, struct epoll_event *event) { struct event_data *ev_data = NULL; struct event_slot_epoll *slot = NULL; event_handler_t handler = NULL; void *data = NULL; int idx = -1; int gen = -1; int ret = -1; int fd = -1; ev_data = (void *)&event->data; handler = NULL; data = NULL; idx = ev_data->idx; gen = ev_data->gen; slot = event_slot_get (event_pool, idx); LOCK (&slot->lock); { fd = slot->fd; if (fd == -1) { gf_msg ("epoll", GF_LOG_ERROR, 0, LG_MSG_STALE_FD_FOUND, "stale fd found on " "idx=%d, gen=%d, events=%d, slot->gen=%d", idx, gen, event->events, slot->gen); /* fd got unregistered in another thread */ goto pre_unlock; } if (gen != slot->gen) { gf_msg ("epoll", GF_LOG_ERROR, 0, LG_MSG_GENERATION_MISMATCH, "generation " "mismatch on idx=%d, gen=%d, slot->gen=%d, " "slot->fd=%d", idx, gen, slot->gen, slot->fd); /* slot was re-used and therefore is another fd! */ goto pre_unlock; } handler = slot->handler; data = slot->data; slot->in_handler++; } pre_unlock: UNLOCK (&slot->lock); if (!handler) goto out; ret = handler (fd, idx, data, (event->events & (EPOLLIN|EPOLLPRI)), (event->events & (EPOLLOUT)), (event->events & (EPOLLERR|EPOLLHUP))); LOCK (&slot->lock); { slot->in_handler--; if (gen != slot->gen) { /* event_unregister() happened while we were in handler() */ gf_msg_debug ("epoll", 0, "generation bumped on idx=%d" " from gen=%d to slot->gen=%d, fd=%d, " "slot->fd=%d", idx, gen, slot->gen, fd, slot->fd); goto post_unlock; } /* This call also picks up the changes made by another thread calling event_select_on_epoll() while this thread was busy in handler() */ if (slot->in_handler == 0) { event->events = slot->events; ret = epoll_ctl (event_pool->fd, EPOLL_CTL_MOD, fd, event); } } post_unlock: UNLOCK (&slot->lock); out: event_slot_unref (event_pool, slot, idx); return ret; } static void * event_dispatch_epoll_worker (void *data) { struct epoll_event event; int ret = -1; struct event_thread_data *ev_data = data; struct event_pool *event_pool; int myindex = -1; int timetodie = 0; GF_VALIDATE_OR_GOTO ("event", ev_data, out); event_pool = ev_data->event_pool; myindex = ev_data->event_index; GF_VALIDATE_OR_GOTO ("event", event_pool, out); gf_msg ("epoll", GF_LOG_INFO, 0, LG_MSG_STARTED_EPOLL_THREAD, "Started" " thread with index %d", myindex); pthread_mutex_lock (&event_pool->mutex); { event_pool->activethreadcount++; } pthread_mutex_unlock (&event_pool->mutex); for (;;) { if (event_pool->eventthreadcount < myindex) { /* ...time to die, thread count was decreased below * this threads index */ /* Start with extra safety at this point, reducing * lock conention in normal case when threads are not * reconfigured always */ pthread_mutex_lock (&event_pool->mutex); { if (event_pool->eventthreadcount < myindex) { /* if found true in critical section, * die */ event_pool->pollers[myindex - 1] = 0; event_pool->activethreadcount--; timetodie = 1; pthread_cond_broadcast (&event_pool->cond); } } pthread_mutex_unlock (&event_pool->mutex); if (timetodie) { gf_msg ("epoll", GF_LOG_INFO, 0, LG_MSG_EXITED_EPOLL_THREAD, "Exited " "thread with index %d", myindex); goto out; } } ret = epoll_wait (event_pool->fd, &event, 1, -1); if (ret == 0) /* timeout */ continue; if (ret == -1 && errno == EINTR) /* sys call */ continue; ret = event_dispatch_epoll_handler (event_pool, &event); } out: if (ev_data) GF_FREE (ev_data); return NULL; } /* Attempts to start the # of configured pollers, ensuring at least the first * is started in a joinable state */ static int event_dispatch_epoll (struct event_pool *event_pool) { int i = 0; pthread_t t_id; int pollercount = 0; int ret = -1; struct event_thread_data *ev_data = NULL; /* Start the configured number of pollers */ pthread_mutex_lock (&event_pool->mutex); { pollercount = event_pool->eventthreadcount; /* Set to MAX if greater */ if (pollercount > EVENT_MAX_THREADS) pollercount = EVENT_MAX_THREADS; /* Default pollers to 1 in case this is incorrectly set */ if (pollercount <= 0) pollercount = 1; event_pool->activethreadcount++; for (i = 0; i < pollercount; i++) { ev_data = GF_CALLOC (1, sizeof (*ev_data), gf_common_mt_event_pool); if (!ev_data) { if (i == 0) { /* Need to suceed creating 0'th * thread, to joinable and wait */ break; } else { /* Inability to create other threads * are a lesser evil, and ignored */ continue; } } ev_data->event_pool = event_pool; ev_data->event_index = i + 1; ret = pthread_create (&t_id, NULL, event_dispatch_epoll_worker, ev_data); if (!ret) { event_pool->pollers[i] = t_id; /* mark all threads other than one in index 0 * as detachable. Errors can be ignored, they * spend their time as zombies if not detched * and the thread counts are decreased */ if (i != 0) pthread_detach (event_pool->pollers[i]); } else { gf_msg ("epoll", GF_LOG_WARNING, 0, LG_MSG_START_EPOLL_THREAD_FAILED, "Failed to start thread for index %d", i); if (i == 0) { GF_FREE (ev_data); break; } else { GF_FREE (ev_data); continue; } } } } pthread_mutex_unlock (&event_pool->mutex); /* Just wait for the first thread, that is created in a joinable state * and will never die, ensuring this function never returns */ if (event_pool->pollers[0] != 0) pthread_join (event_pool->pollers[0], NULL); pthread_mutex_lock (&event_pool->mutex); { event_pool->activethreadcount--; } pthread_mutex_unlock (&event_pool->mutex); return ret; } int event_reconfigure_threads_epoll (struct event_pool *event_pool, int value) { int i; int ret = 0; pthread_t t_id; int oldthreadcount; struct event_thread_data *ev_data = NULL; pthread_mutex_lock (&event_pool->mutex); { /* Reconfigure to 0 threads is allowed only in destroy mode */ if (event_pool->destroy == 1) { value = 0; } else { /* Set to MAX if greater */ if (value > EVENT_MAX_THREADS) value = EVENT_MAX_THREADS; /* Default pollers to 1 in case this is set incorrectly */ if (value <= 0) value = 1; } oldthreadcount = event_pool->eventthreadcount; if (oldthreadcount < value) { /* create more poll threads */ for (i = oldthreadcount; i < value; i++) { /* Start a thread if the index at this location * is a 0, so that the older thread is confirmed * as dead */ if (event_pool->pollers[i] == 0) { ev_data = GF_CALLOC (1, sizeof (*ev_data), gf_common_mt_event_pool); if (!ev_data) { continue; } ev_data->event_pool = event_pool; ev_data->event_index = i + 1; ret = pthread_create (&t_id, NULL, event_dispatch_epoll_worker, ev_data); if (ret) { gf_msg ("epoll", GF_LOG_WARNING, 0, LG_MSG_START_EPOLL_THREAD_FAILED, "Failed to start thread" " for index %d", i); GF_FREE (ev_data); } else { pthread_detach (t_id); event_pool->pollers[i] = t_id; } } } } /* if value decreases, threads will terminate, themselves */ event_pool->eventthreadcount = value; } pthread_mutex_unlock (&event_pool->mutex); return 0; } /* This function is the destructor for the event_pool data structure * Should be called only after poller_threads_destroy() is called, * else will lead to crashes. */ static int event_pool_destroy_epoll (struct event_pool *event_pool) { int ret = 0, i = 0, j = 0; struct event_slot_epoll *table = NULL; ret = close (event_pool->fd); for (i = 0; i < EVENT_EPOLL_TABLES; i++) { if (event_pool->ereg[i]) { table = event_pool->ereg[i]; event_pool->ereg[i] = NULL; for (j = 0; j < EVENT_EPOLL_SLOTS; j++) { LOCK_DESTROY (&table[j].lock); } GF_FREE (table); } } pthread_mutex_destroy (&event_pool->mutex); pthread_cond_destroy (&event_pool->cond); GF_FREE (event_pool->evcache); GF_FREE (event_pool->reg); GF_FREE (event_pool); return ret; } struct event_ops event_ops_epoll = { .new = event_pool_new_epoll, .event_register = event_register_epoll, .event_select_on = event_select_on_epoll, .event_unregister = event_unregister_epoll, .event_unregister_close = event_unregister_close_epoll, .event_dispatch = event_dispatch_epoll, .event_reconfigure_threads = event_reconfigure_threads_epoll, .event_pool_destroy = event_pool_destroy_epoll }; #endif