/* Copyright (c) 2008-2012 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. */ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" #endif #include "dict.h" #include "glusterfs.h" #include "logging.h" #include "rdma.h" #include "name.h" #include "byte-order.h" #include "xlator.h" #include "xdr-rpc.h" #include #define GF_RDMA_LOG_NAME "rpc-transport/rdma" static int32_t __gf_rdma_ioq_churn (gf_rdma_peer_t *peer); gf_rdma_post_t * gf_rdma_post_ref (gf_rdma_post_t *post); int gf_rdma_post_unref (gf_rdma_post_t *post); static void * gf_rdma_send_completion_proc (void *data); static void * gf_rdma_recv_completion_proc (void *data); void * gf_rdma_async_event_thread (void *context); static int32_t gf_rdma_create_qp (rpc_transport_t *this); static int32_t __gf_rdma_teardown (rpc_transport_t *this); static int32_t gf_rdma_teardown (rpc_transport_t *this); static int32_t gf_rdma_disconnect (rpc_transport_t *this); static void gf_rdma_cm_handle_disconnect (rpc_transport_t *this); static int gf_rdma_cm_handle_connect_init (struct rdma_cm_event *event); static void gf_rdma_put_post (gf_rdma_queue_t *queue, gf_rdma_post_t *post) { post->ctx.is_request = 0; pthread_mutex_lock (&queue->lock); { if (post->prev) { queue->active_count--; post->prev->next = post->next; } if (post->next) { post->next->prev = post->prev; } post->prev = &queue->passive_posts; post->next = post->prev->next; post->prev->next = post; post->next->prev = post; queue->passive_count++; } pthread_mutex_unlock (&queue->lock); } static gf_rdma_post_t * gf_rdma_new_post (rpc_transport_t *this, gf_rdma_device_t *device, int32_t len, gf_rdma_post_type_t type) { gf_rdma_post_t *post = NULL; int ret = -1; post = (gf_rdma_post_t *) GF_CALLOC (1, sizeof (*post), gf_common_mt_rdma_post_t); if (post == NULL) { goto out; } pthread_mutex_init (&post->lock, NULL); post->buf_size = len; post->buf = valloc (len); if (!post->buf) { gf_msg_nomem (GF_RDMA_LOG_NAME, GF_LOG_ERROR, len); goto out; } post->mr = ibv_reg_mr (device->pd, post->buf, post->buf_size, IBV_ACCESS_LOCAL_WRITE); if (!post->mr) { gf_log (this->name, GF_LOG_WARNING, "memory registration failed (%s)", strerror (errno)); goto out; } post->device = device; post->type = type; ret = 0; out: if (ret != 0) { free (post->buf); GF_FREE (post); post = NULL; } return post; } static gf_rdma_post_t * gf_rdma_get_post (gf_rdma_queue_t *queue) { gf_rdma_post_t *post = NULL; pthread_mutex_lock (&queue->lock); { post = queue->passive_posts.next; if (post == &queue->passive_posts) post = NULL; if (post) { if (post->prev) post->prev->next = post->next; if (post->next) post->next->prev = post->prev; post->prev = &queue->active_posts; post->next = post->prev->next; post->prev->next = post; post->next->prev = post; post->reused++; queue->active_count++; } } pthread_mutex_unlock (&queue->lock); return post; } void gf_rdma_destroy_post (gf_rdma_post_t *post) { ibv_dereg_mr (post->mr); free (post->buf); GF_FREE (post); } static int32_t __gf_rdma_quota_get (gf_rdma_peer_t *peer) { int32_t ret = -1; gf_rdma_private_t *priv = NULL; priv = peer->trans->private; if (priv->connected && peer->quota > 0) { ret = peer->quota--; } return ret; } static void __gf_rdma_ioq_entry_free (gf_rdma_ioq_t *entry) { list_del_init (&entry->list); if (entry->iobref) { iobref_unref (entry->iobref); entry->iobref = NULL; } if (entry->msg.request.rsp_iobref) { iobref_unref (entry->msg.request.rsp_iobref); entry->msg.request.rsp_iobref = NULL; } mem_put (entry); } static void __gf_rdma_ioq_flush (gf_rdma_peer_t *peer) { gf_rdma_ioq_t *entry = NULL, *dummy = NULL; list_for_each_entry_safe (entry, dummy, &peer->ioq, list) { __gf_rdma_ioq_entry_free (entry); } } static int32_t __gf_rdma_disconnect (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; priv = this->private; if (priv->connected) { rdma_disconnect (priv->peer.cm_id); } return 0; } static void gf_rdma_queue_init (gf_rdma_queue_t *queue) { pthread_mutex_init (&queue->lock, NULL); queue->active_posts.next = &queue->active_posts; queue->active_posts.prev = &queue->active_posts; queue->passive_posts.next = &queue->passive_posts; queue->passive_posts.prev = &queue->passive_posts; } static void __gf_rdma_destroy_queue (gf_rdma_post_t *post) { gf_rdma_post_t *tmp = NULL; while (post->next != post) { tmp = post->next; post->next = post->next->next; post->next->prev = post; gf_rdma_destroy_post (tmp); } } static void gf_rdma_destroy_queue (gf_rdma_queue_t *queue) { if (queue == NULL) { goto out; } pthread_mutex_lock (&queue->lock); { if (queue->passive_count > 0) { __gf_rdma_destroy_queue (&queue->passive_posts); queue->passive_count = 0; } if (queue->active_count > 0) { __gf_rdma_destroy_queue (&queue->active_posts); queue->active_count = 0; } } pthread_mutex_unlock (&queue->lock); out: return; } static void gf_rdma_destroy_posts (rpc_transport_t *this) { gf_rdma_device_t *device = NULL; gf_rdma_private_t *priv = NULL; if (this == NULL) { goto out; } priv = this->private; device = priv->device; gf_rdma_destroy_queue (&device->sendq); gf_rdma_destroy_queue (&device->recvq); out: return; } static int32_t __gf_rdma_create_posts (rpc_transport_t *this, int32_t count, int32_t size, gf_rdma_queue_t *q, gf_rdma_post_type_t type) { int32_t i = 0; int32_t ret = 0; gf_rdma_private_t *priv = NULL; gf_rdma_device_t *device = NULL; priv = this->private; device = priv->device; for (i=0 ; iname, GF_LOG_ERROR, "post creation failed"); ret = -1; break; } gf_rdma_put_post (q, post); } return ret; } static int32_t gf_rdma_post_recv (struct ibv_srq *srq, gf_rdma_post_t *post) { struct ibv_sge list = { .addr = (unsigned long) post->buf, .length = post->buf_size, .lkey = post->mr->lkey }; struct ibv_recv_wr wr = { .wr_id = (unsigned long) post, .sg_list = &list, .num_sge = 1, }, *bad_wr; gf_rdma_post_ref (post); return ibv_post_srq_recv (srq, &wr, &bad_wr); } static int32_t gf_rdma_create_posts (rpc_transport_t *this) { int32_t i = 0, ret = 0; gf_rdma_post_t *post = NULL; gf_rdma_private_t *priv = NULL; gf_rdma_options_t *options = NULL; gf_rdma_device_t *device = NULL; priv = this->private; options = &priv->options; device = priv->device; ret = __gf_rdma_create_posts (this, options->send_count, options->send_size, &device->sendq, GF_RDMA_SEND_POST); if (!ret) ret = __gf_rdma_create_posts (this, options->recv_count, options->recv_size, &device->recvq, GF_RDMA_RECV_POST); if (!ret) { for (i=0 ; irecv_count ; i++) { post = gf_rdma_get_post (&device->recvq); if (gf_rdma_post_recv (device->srq, post) != 0) { ret = -1; break; } } } if (ret) gf_rdma_destroy_posts (this); return ret; } static void gf_rdma_destroy_cq (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; gf_rdma_device_t *device = NULL; priv = this->private; device = priv->device; if (device->recv_cq) ibv_destroy_cq (device->recv_cq); device->recv_cq = NULL; if (device->send_cq) ibv_destroy_cq (device->send_cq); device->send_cq = NULL; return; } static int32_t gf_rdma_create_cq (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; gf_rdma_options_t *options = NULL; gf_rdma_device_t *device = NULL; uint64_t send_cqe = 0; int32_t ret = 0; struct ibv_device_attr device_attr = {{0}, }; priv = this->private; options = &priv->options; device = priv->device; device->recv_cq = ibv_create_cq (priv->device->context, options->recv_count * 2, device, device->recv_chan, 0); if (!device->recv_cq) { gf_log (this->name, GF_LOG_ERROR, "creation of CQ for device %s failed", device->device_name); ret = -1; goto out; } else if (ibv_req_notify_cq (device->recv_cq, 0)) { gf_log (this->name, GF_LOG_ERROR, "ibv_req_notify_cq on recv CQ of device %s failed", device->device_name); ret = -1; goto out; } do { ret = ibv_query_device (priv->device->context, &device_attr); if (ret != 0) { gf_log (this->name, GF_LOG_ERROR, "ibv_query_device on %s returned %d (%s)", priv->device->device_name, ret, (ret > 0) ? strerror (ret) : ""); ret = -1; goto out; } send_cqe = options->send_count * 128; send_cqe = (send_cqe > device_attr.max_cqe) ? device_attr.max_cqe : send_cqe; /* TODO: make send_cq size dynamically adaptive */ device->send_cq = ibv_create_cq (priv->device->context, send_cqe, device, device->send_chan, 0); if (!device->send_cq) { gf_log (this->name, GF_LOG_ERROR, "creation of send_cq for device %s failed", device->device_name); ret = -1; goto out; } if (ibv_req_notify_cq (device->send_cq, 0)) { gf_log (this->name, GF_LOG_ERROR, "ibv_req_notify_cq on send_cq for device %s" " failed", device->device_name); ret = -1; goto out; } } while (0); out: if (ret != 0) gf_rdma_destroy_cq (this); return ret; } static gf_rdma_device_t * gf_rdma_get_device (rpc_transport_t *this, struct ibv_context *ibctx, char *device_name) { glusterfs_ctx_t *ctx = NULL; gf_rdma_private_t *priv = NULL; gf_rdma_options_t *options = NULL; int32_t ret = 0; int32_t i = 0; gf_rdma_device_t *trav = NULL, *device = NULL; gf_rdma_ctx_t *rdma_ctx = NULL; priv = this->private; options = &priv->options; ctx = this->ctx; rdma_ctx = ctx->ib; trav = rdma_ctx->device; while (trav) { if (!strcmp (trav->device_name, device_name)) break; trav = trav->next; } if (!trav) { trav = GF_CALLOC (1, sizeof (*trav), gf_common_mt_rdma_device_t); if (trav == NULL) { goto out; } priv->device = trav; trav->context = ibctx; trav->request_ctx_pool = mem_pool_new (gf_rdma_request_context_t, GF_RDMA_POOL_SIZE); if (trav->request_ctx_pool == NULL) { goto out; } trav->ioq_pool = mem_pool_new (gf_rdma_ioq_t, GF_RDMA_POOL_SIZE); if (trav->ioq_pool == NULL) { goto out; } trav->reply_info_pool = mem_pool_new (gf_rdma_reply_info_t, GF_RDMA_POOL_SIZE); if (trav->reply_info_pool == NULL) { goto out; } trav->device_name = gf_strdup (device_name); trav->next = rdma_ctx->device; rdma_ctx->device = trav; trav->send_chan = ibv_create_comp_channel (trav->context); if (!trav->send_chan) { gf_log (this->name, GF_LOG_ERROR, "could not create send completion channel for " "device (%s)", device_name); goto out; } trav->recv_chan = ibv_create_comp_channel (trav->context); if (!trav->recv_chan) { gf_log (this->name, GF_LOG_ERROR, "could not create recv completion channel for " "device (%s)", device_name); /* TODO: cleanup current mess */ goto out; } if (gf_rdma_create_cq (this) < 0) { gf_log (this->name, GF_LOG_ERROR, "could not create CQ for device (%s)", device_name); goto out; } /* protection domain */ trav->pd = ibv_alloc_pd (trav->context); if (!trav->pd) { gf_log (this->name, GF_LOG_ERROR, "could not allocate protection domain for " "device (%s)", device_name); goto out; } struct ibv_srq_init_attr attr = { .attr = { .max_wr = options->recv_count, .max_sge = 1, .srq_limit = 10 } }; trav->srq = ibv_create_srq (trav->pd, &attr); if (!trav->srq) { gf_log (this->name, GF_LOG_ERROR, "could not create SRQ for device (%s)", device_name); goto out; } /* queue init */ gf_rdma_queue_init (&trav->sendq); gf_rdma_queue_init (&trav->recvq); if (gf_rdma_create_posts (this) < 0) { gf_log (this->name, GF_LOG_ERROR, "could not allocate posts for device (%s)", device_name); goto out; } /* completion threads */ ret = gf_thread_create (&trav->send_thread, NULL, gf_rdma_send_completion_proc, trav->send_chan); if (ret) { gf_log (this->name, GF_LOG_ERROR, "could not create send completion thread for " "device (%s)", device_name); goto out; } ret = gf_thread_create (&trav->recv_thread, NULL, gf_rdma_recv_completion_proc, trav->recv_chan); if (ret) { gf_log (this->name, GF_LOG_ERROR, "could not create recv completion thread " "for device (%s)", device_name); return NULL; } ret = gf_thread_create (&trav->async_event_thread, NULL, gf_rdma_async_event_thread, ibctx); if (ret) { gf_log (this->name, GF_LOG_ERROR, "could not create async_event_thread"); return NULL; } /* qpreg */ pthread_mutex_init (&trav->qpreg.lock, NULL); for (i=0; i<42; i++) { trav->qpreg.ents[i].next = &trav->qpreg.ents[i]; trav->qpreg.ents[i].prev = &trav->qpreg.ents[i]; } } device = trav; trav = NULL; out: if (trav != NULL) { gf_rdma_destroy_posts (this); mem_pool_destroy (trav->ioq_pool); mem_pool_destroy (trav->request_ctx_pool); mem_pool_destroy (trav->reply_info_pool); ibv_dealloc_pd (trav->pd); gf_rdma_destroy_cq (this); ibv_destroy_comp_channel (trav->recv_chan); ibv_destroy_comp_channel (trav->send_chan); GF_FREE ((char *)trav->device_name); GF_FREE (trav); } return device; } static rpc_transport_t * gf_rdma_transport_new (rpc_transport_t *listener, struct rdma_cm_id *cm_id) { gf_rdma_private_t *listener_priv = NULL, *priv = NULL; rpc_transport_t *this = NULL, *new = NULL; gf_rdma_options_t *options = NULL; char *device_name = NULL; listener_priv = listener->private; this = GF_CALLOC (1, sizeof (rpc_transport_t), gf_common_mt_rpc_transport_t); if (this == NULL) { goto out; } this->listener = listener; priv = GF_CALLOC (1, sizeof (gf_rdma_private_t), gf_common_mt_rdma_private_t); if (priv == NULL) { goto out; } this->private = priv; priv->options = listener_priv->options; priv->listener = listener; priv->entity = GF_RDMA_SERVER; options = &priv->options; this->ops = listener->ops; this->init = listener->init; this->fini = listener->fini; this->ctx = listener->ctx; this->name = gf_strdup (listener->name); this->notify = listener->notify; this->mydata = listener->mydata; this->xl = listener->xl; this->myinfo.sockaddr_len = sizeof (cm_id->route.addr.src_addr); memcpy (&this->myinfo.sockaddr, &cm_id->route.addr.src_addr, this->myinfo.sockaddr_len); this->peerinfo.sockaddr_len = sizeof (cm_id->route.addr.dst_addr); memcpy (&this->peerinfo.sockaddr, &cm_id->route.addr.dst_addr, this->peerinfo.sockaddr_len); priv->peer.trans = this; gf_rdma_get_transport_identifiers (this); device_name = (char *)ibv_get_device_name (cm_id->verbs->device); if (device_name == NULL) { gf_log (listener->name, GF_LOG_WARNING, "cannot get device name (peer:%s me:%s)", this->peerinfo.identifier, this->myinfo.identifier); goto out; } priv->device = gf_rdma_get_device (this, cm_id->verbs, device_name); if (priv->device == NULL) { gf_log (listener->name, GF_LOG_WARNING, "cannot get infiniband device %s (peer:%s me:%s)", device_name, this->peerinfo.identifier, this->myinfo.identifier); goto out; } priv->peer.send_count = options->send_count; priv->peer.recv_count = options->recv_count; priv->peer.send_size = options->send_size; priv->peer.recv_size = options->recv_size; priv->peer.cm_id = cm_id; INIT_LIST_HEAD (&priv->peer.ioq); pthread_mutex_init (&priv->write_mutex, NULL); pthread_mutex_init (&priv->recv_mutex, NULL); cm_id->context = this; new = rpc_transport_ref (this); this = NULL; out: if (this != NULL) { if (this->private != NULL) { GF_FREE (this->private); } if (this->name != NULL) { GF_FREE (this->name); } GF_FREE (this); } return new; } static int gf_rdma_cm_handle_connect_request (struct rdma_cm_event *event) { int ret = -1; rpc_transport_t *this = NULL, *listener = NULL; struct rdma_cm_id *child_cm_id = NULL, *listener_cm_id = NULL; struct rdma_conn_param conn_param = {0, }; gf_rdma_private_t *priv = NULL; gf_rdma_options_t *options = NULL; child_cm_id = event->id; listener_cm_id = event->listen_id; listener = listener_cm_id->context; priv = listener->private; options = &priv->options; this = gf_rdma_transport_new (listener, child_cm_id); if (this == NULL) { gf_log (listener->name, GF_LOG_WARNING, "could not create a transport for incoming connection" " (me.name:%s me.identifier:%s)", listener->name, listener->myinfo.identifier); rdma_destroy_id (child_cm_id); goto out; } gf_log (listener->name, GF_LOG_TRACE, "got a connect request (me:%s peer:%s)", listener->myinfo.identifier, this->peerinfo.identifier); ret = gf_rdma_create_qp (this); if (ret < 0) { gf_log (listener->name, GF_LOG_WARNING, "could not create QP (peer:%s me:%s)", this->peerinfo.identifier, this->myinfo.identifier); gf_rdma_cm_handle_disconnect (this); goto out; } conn_param.responder_resources = 1; conn_param.initiator_depth = 1; conn_param.retry_count = options->attr_retry_cnt; conn_param.rnr_retry_count = options->attr_rnr_retry; ret = rdma_accept(child_cm_id, &conn_param); if (ret < 0) { gf_log (listener->name, GF_LOG_WARNING, "rdma_accept failed " "peer:%s me:%s (%s)", this->peerinfo.identifier, this->myinfo.identifier, strerror (errno)); gf_rdma_cm_handle_disconnect (this); goto out; } gf_rdma_cm_handle_connect_init (event); ret = 0; out: return ret; } static int gf_rdma_cm_handle_route_resolved (struct rdma_cm_event *event) { struct rdma_conn_param conn_param = {0, }; int ret = 0; rpc_transport_t *this = NULL; gf_rdma_private_t *priv = NULL; gf_rdma_peer_t *peer = NULL; gf_rdma_options_t *options = NULL; if (event == NULL) { goto out; } this = event->id->context; priv = this->private; peer = &priv->peer; options = &priv->options; ret = gf_rdma_create_qp (this); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "could not create QP (peer:%s me:%s)", this->peerinfo.identifier, this->myinfo.identifier); gf_rdma_cm_handle_disconnect (this); goto out; } memset(&conn_param, 0, sizeof conn_param); conn_param.responder_resources = 1; conn_param.initiator_depth = 1; conn_param.retry_count = options->attr_retry_cnt; conn_param.rnr_retry_count = options->attr_rnr_retry; ret = rdma_connect(peer->cm_id, &conn_param); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "rdma_connect failed (%s)", strerror (errno)); gf_rdma_cm_handle_disconnect (this); goto out; } gf_log (this->name, GF_LOG_TRACE, "route resolved (me:%s peer:%s)", this->myinfo.identifier, this->peerinfo.identifier); ret = 0; out: return ret; } static int gf_rdma_cm_handle_addr_resolved (struct rdma_cm_event *event) { rpc_transport_t *this = NULL; gf_rdma_peer_t *peer = NULL; gf_rdma_private_t *priv = NULL; int ret = 0; this = event->id->context; priv = this->private; peer = &priv->peer; GF_ASSERT (peer->cm_id == event->id); this->myinfo.sockaddr_len = sizeof (peer->cm_id->route.addr.src_addr); memcpy (&this->myinfo.sockaddr, &peer->cm_id->route.addr.src_addr, this->myinfo.sockaddr_len); this->peerinfo.sockaddr_len = sizeof (peer->cm_id->route.addr.dst_addr); memcpy (&this->peerinfo.sockaddr, &peer->cm_id->route.addr.dst_addr, this->peerinfo.sockaddr_len); gf_rdma_get_transport_identifiers (this); ret = rdma_resolve_route(peer->cm_id, 2000); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "rdma_resolve_route failed (me:%s peer:%s) (%s)", this->myinfo.identifier, this->peerinfo.identifier, strerror (errno)); gf_rdma_cm_handle_disconnect (this); } gf_log (this->name, GF_LOG_TRACE, "Address resolved (me:%s peer:%s)", this->myinfo.identifier, this->peerinfo.identifier); return ret; } static void gf_rdma_cm_handle_disconnect (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; char need_unref = 0, connected = 0; priv = this->private; gf_log (this->name, GF_LOG_DEBUG, "peer disconnected, cleaning up"); pthread_mutex_lock (&priv->write_mutex); { if (priv->peer.cm_id != NULL) { need_unref = 1; connected = priv->connected; priv->connected = 0; } __gf_rdma_teardown (this); } pthread_mutex_unlock (&priv->write_mutex); if (connected) { rpc_transport_notify (this, RPC_TRANSPORT_DISCONNECT, this); } if (need_unref) rpc_transport_unref (this); } static int gf_rdma_cm_handle_connect_init (struct rdma_cm_event *event) { rpc_transport_t *this = NULL; gf_rdma_private_t *priv = NULL; struct rdma_cm_id *cm_id = NULL; int ret = 0; cm_id = event->id; this = cm_id->context; priv = this->private; if (priv->connected == 1) { gf_log (this->name, GF_LOG_TRACE, "received event RDMA_CM_EVENT_ESTABLISHED (me:%s peer:%s)", this->myinfo.identifier, this->peerinfo.identifier); return ret; } priv->connected = 1; pthread_mutex_lock (&priv->write_mutex); { priv->peer.quota = 1; priv->peer.quota_set = 0; } pthread_mutex_unlock (&priv->write_mutex); if (priv->entity == GF_RDMA_CLIENT) { gf_log (this->name, GF_LOG_TRACE, "received event RDMA_CM_EVENT_ESTABLISHED (me:%s peer:%s)", this->myinfo.identifier, this->peerinfo.identifier); ret = rpc_transport_notify (this, RPC_TRANSPORT_CONNECT, this); } else if (priv->entity == GF_RDMA_SERVER) { ret = rpc_transport_notify (priv->listener, RPC_TRANSPORT_ACCEPT, this); } if (ret < 0) { gf_rdma_disconnect (this); } return ret; } static int gf_rdma_cm_handle_event_error (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; priv = this->private; if (priv->entity != GF_RDMA_SERVER_LISTENER) { gf_rdma_cm_handle_disconnect (this); } return 0; } static int gf_rdma_cm_handle_device_removal (struct rdma_cm_event *event) { return 0; } static void * gf_rdma_cm_event_handler (void *data) { struct rdma_cm_event *event = NULL; int ret = 0; rpc_transport_t *this = NULL; struct rdma_event_channel *event_channel = NULL; event_channel = data; while (1) { ret = rdma_get_cm_event (event_channel, &event); if (ret != 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma_cm_get_event failed (%s)", strerror (errno)); break; } switch (event->event) { case RDMA_CM_EVENT_ADDR_RESOLVED: gf_rdma_cm_handle_addr_resolved (event); break; case RDMA_CM_EVENT_ROUTE_RESOLVED: gf_rdma_cm_handle_route_resolved (event); break; case RDMA_CM_EVENT_CONNECT_REQUEST: gf_rdma_cm_handle_connect_request (event); break; case RDMA_CM_EVENT_ESTABLISHED: gf_rdma_cm_handle_connect_init (event); break; case RDMA_CM_EVENT_ADDR_ERROR: case RDMA_CM_EVENT_ROUTE_ERROR: case RDMA_CM_EVENT_CONNECT_ERROR: case RDMA_CM_EVENT_UNREACHABLE: case RDMA_CM_EVENT_REJECTED: this = event->id->context; gf_log (this->name, GF_LOG_WARNING, "cma event %s, error %d (me:%s peer:%s)\n", rdma_event_str(event->event), event->status, this->myinfo.identifier, this->peerinfo.identifier); rdma_ack_cm_event (event); event = NULL; gf_rdma_cm_handle_event_error (this); continue; case RDMA_CM_EVENT_DISCONNECTED: this = event->id->context; gf_log (this->name, GF_LOG_DEBUG, "received disconnect (me:%s peer:%s)\n", this->myinfo.identifier, this->peerinfo.identifier); rdma_ack_cm_event (event); event = NULL; gf_rdma_cm_handle_disconnect (this); continue; case RDMA_CM_EVENT_DEVICE_REMOVAL: gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "device removed"); gf_rdma_cm_handle_device_removal (event); break; default: gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "unhandled event: %s, ignoring", rdma_event_str(event->event)); break; } rdma_ack_cm_event (event); } return NULL; } static int32_t gf_rdma_post_send (struct ibv_qp *qp, gf_rdma_post_t *post, int32_t len) { struct ibv_sge list = { .addr = (unsigned long) post->buf, .length = len, .lkey = post->mr->lkey }; struct ibv_send_wr wr = { .wr_id = (unsigned long) post, .sg_list = &list, .num_sge = 1, .opcode = IBV_WR_SEND, .send_flags = IBV_SEND_SIGNALED, }, *bad_wr; if (!qp) return EINVAL; return ibv_post_send (qp, &wr, &bad_wr); } int __gf_rdma_encode_error(gf_rdma_peer_t *peer, gf_rdma_reply_info_t *reply_info, struct iovec *rpchdr, gf_rdma_header_t *hdr, gf_rdma_errcode_t err) { struct rpc_msg *rpc_msg = NULL; if (reply_info != NULL) { hdr->rm_xid = hton32(reply_info->rm_xid); } else { rpc_msg = rpchdr[0].iov_base; /* assume rpchdr contains * only one vector. * (which is true) */ hdr->rm_xid = rpc_msg->rm_xid; } hdr->rm_vers = hton32(GF_RDMA_VERSION); hdr->rm_credit = hton32(peer->send_count); hdr->rm_type = hton32(GF_RDMA_ERROR); hdr->rm_body.rm_error.rm_type = hton32(err); if (err == ERR_VERS) { hdr->rm_body.rm_error.rm_version.gf_rdma_vers_low = hton32(GF_RDMA_VERSION); hdr->rm_body.rm_error.rm_version.gf_rdma_vers_high = hton32(GF_RDMA_VERSION); } return sizeof (*hdr); } int32_t __gf_rdma_send_error (gf_rdma_peer_t *peer, gf_rdma_ioq_t *entry, gf_rdma_post_t *post, gf_rdma_reply_info_t *reply_info, gf_rdma_errcode_t err) { int32_t ret = -1, len = 0; len = __gf_rdma_encode_error (peer, reply_info, entry->rpchdr, (gf_rdma_header_t *)post->buf, err); if (len == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_ERROR, "encode error returned -1"); goto out; } gf_rdma_post_ref (post); ret = gf_rdma_post_send (peer->qp, post, len); if (!ret) { ret = len; } else { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "gf_rdma_post_send (to %s) failed with ret = %d (%s)", peer->trans->peerinfo.identifier, ret, (ret > 0) ? strerror (ret) : ""); gf_rdma_post_unref (post); __gf_rdma_disconnect (peer->trans); ret = -1; } out: return ret; } int32_t __gf_rdma_create_read_chunks_from_vector (gf_rdma_peer_t *peer, gf_rdma_read_chunk_t **readch_ptr, int32_t *pos, struct iovec *vector, int count, gf_rdma_request_context_t *request_ctx) { int i = 0; gf_rdma_private_t *priv = NULL; gf_rdma_device_t *device = NULL; struct ibv_mr *mr = NULL; gf_rdma_read_chunk_t *readch = NULL; int32_t ret = -1; GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, peer, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, readch_ptr, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, *readch_ptr, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, request_ctx, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, vector, out); priv = peer->trans->private; device = priv->device; readch = *readch_ptr; for (i = 0; i < count; i++) { readch->rc_discrim = hton32 (1); readch->rc_position = hton32 (*pos); mr = ibv_reg_mr (device->pd, vector[i].iov_base, vector[i].iov_len, IBV_ACCESS_REMOTE_READ); if (!mr) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "memory registration failed (%s) (peer:%s)", strerror (errno), peer->trans->peerinfo.identifier); goto out; } request_ctx->mr[request_ctx->mr_count++] = mr; readch->rc_target.rs_handle = hton32 (mr->rkey); readch->rc_target.rs_length = hton32 (vector[i].iov_len); readch->rc_target.rs_offset = hton64 ((uint64_t)(unsigned long)vector[i].iov_base); *pos = *pos + vector[i].iov_len; readch++; } *readch_ptr = readch; ret = 0; out: return ret; } int32_t __gf_rdma_create_read_chunks (gf_rdma_peer_t *peer, gf_rdma_ioq_t *entry, gf_rdma_chunktype_t type, uint32_t **ptr, gf_rdma_request_context_t *request_ctx) { int32_t ret = -1; int pos = 0; GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, peer, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, entry, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, ptr, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, *ptr, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, request_ctx, out); request_ctx->iobref = iobref_ref (entry->iobref); if (type == gf_rdma_areadch) { pos = 0; ret = __gf_rdma_create_read_chunks_from_vector (peer, (gf_rdma_read_chunk_t **)ptr, &pos, entry->rpchdr, entry->rpchdr_count, request_ctx); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot create read chunks from vector " "entry->rpchdr"); goto out; } ret = __gf_rdma_create_read_chunks_from_vector (peer, (gf_rdma_read_chunk_t **)ptr, &pos, entry->proghdr, entry->proghdr_count, request_ctx); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot create read chunks from vector " "entry->proghdr"); } if (entry->prog_payload_count != 0) { ret = __gf_rdma_create_read_chunks_from_vector (peer, (gf_rdma_read_chunk_t **)ptr, &pos, entry->prog_payload, entry->prog_payload_count, request_ctx); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot create read chunks from vector" " entry->prog_payload"); } } } else { pos = iov_length (entry->rpchdr, entry->rpchdr_count); ret = __gf_rdma_create_read_chunks_from_vector (peer, (gf_rdma_read_chunk_t **)ptr, &pos, entry->prog_payload, entry->prog_payload_count, request_ctx); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot create read chunks from vector " "entry->prog_payload"); } } /* terminate read-chunk list*/ **ptr = 0; *ptr = *ptr + 1; out: return ret; } int32_t __gf_rdma_create_write_chunks_from_vector (gf_rdma_peer_t *peer, gf_rdma_write_chunk_t **writech_ptr, struct iovec *vector, int count, gf_rdma_request_context_t *request_ctx) { int i = 0; gf_rdma_private_t *priv = NULL; gf_rdma_device_t *device = NULL; struct ibv_mr *mr = NULL; gf_rdma_write_chunk_t *writech = NULL; int32_t ret = -1; GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, peer, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, writech_ptr, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, *writech_ptr, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, request_ctx, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, vector, out); writech = *writech_ptr; priv = peer->trans->private; device = priv->device; for (i = 0; i < count; i++) { mr = ibv_reg_mr (device->pd, vector[i].iov_base, vector[i].iov_len, IBV_ACCESS_REMOTE_WRITE | IBV_ACCESS_LOCAL_WRITE); if (!mr) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "memory registration failed (%s) (peer:%s)", strerror (errno), peer->trans->peerinfo.identifier); goto out; } request_ctx->mr[request_ctx->mr_count++] = mr; writech->wc_target.rs_handle = hton32 (mr->rkey); writech->wc_target.rs_length = hton32 (vector[i].iov_len); writech->wc_target.rs_offset = hton64 (((uint64_t)(unsigned long)vector[i].iov_base)); writech++; } *writech_ptr = writech; ret = 0; out: return ret; } int32_t __gf_rdma_create_write_chunks (gf_rdma_peer_t *peer, gf_rdma_ioq_t *entry, gf_rdma_chunktype_t chunk_type, uint32_t **ptr, gf_rdma_request_context_t *request_ctx) { int32_t ret = -1; gf_rdma_write_array_t *warray = NULL; GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, peer, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, ptr, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, *ptr, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, request_ctx, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, entry, out); if ((chunk_type == gf_rdma_replych) && ((entry->msg.request.rsphdr_count != 1) || (entry->msg.request.rsphdr_vec[0].iov_base == NULL))) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, (entry->msg.request.rsphdr_count == 1) ? "chunktype specified as reply chunk but the vector " "specifying the buffer to be used for holding reply" " header is not correct" : "chunktype specified as reply chunk, but more than one " "buffer provided for holding reply"); goto out; } /* if ((chunk_type == gf_rdma_writech) && ((entry->msg.request.rsphdr_count == 0) || (entry->msg.request.rsphdr_vec[0].iov_base == NULL))) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_DEBUG, "vector specifying buffer to hold the program's reply " "header should also be provided when buffers are " "provided for holding the program's payload in reply"); goto out; } */ if (chunk_type == gf_rdma_writech) { warray = (gf_rdma_write_array_t *)*ptr; warray->wc_discrim = hton32 (1); warray->wc_nchunks = hton32 (entry->msg.request.rsp_payload_count); *ptr = (uint32_t *)&warray->wc_array[0]; ret = __gf_rdma_create_write_chunks_from_vector (peer, (gf_rdma_write_chunk_t **)ptr, entry->msg.request.rsp_payload, entry->msg.request.rsp_payload_count, request_ctx); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot create write chunks from vector " "entry->rpc_payload"); goto out; } /* terminate write chunklist */ **ptr = 0; *ptr = *ptr + 1; /* no reply chunklist */ **ptr = 0; *ptr = *ptr + 1; } else { /* no write chunklist */ **ptr = 0; *ptr = *ptr + 1; warray = (gf_rdma_write_array_t *)*ptr; warray->wc_discrim = hton32 (1); warray->wc_nchunks = hton32 (entry->msg.request.rsphdr_count); *ptr = (uint32_t *)&warray->wc_array[0]; ret = __gf_rdma_create_write_chunks_from_vector (peer, (gf_rdma_write_chunk_t **)ptr, entry->msg.request.rsphdr_vec, entry->msg.request.rsphdr_count, request_ctx); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot create write chunks from vector " "entry->rpchdr"); goto out; } /* terminate reply chunklist */ **ptr = 0; *ptr = *ptr + 1; } out: return ret; } static inline void __gf_rdma_deregister_mr (struct ibv_mr **mr, int count) { int i = 0; if (mr == NULL) { goto out; } for (i = 0; i < count; i++) { ibv_dereg_mr (mr[i]); } out: return; } static int32_t __gf_rdma_quota_put (gf_rdma_peer_t *peer) { int32_t ret = 0; peer->quota++; ret = peer->quota; if (!list_empty (&peer->ioq)) { ret = __gf_rdma_ioq_churn (peer); } return ret; } static int32_t gf_rdma_quota_put (gf_rdma_peer_t *peer) { int32_t ret = 0; gf_rdma_private_t *priv = NULL; priv = peer->trans->private; pthread_mutex_lock (&priv->write_mutex); { ret = __gf_rdma_quota_put (peer); } pthread_mutex_unlock (&priv->write_mutex); return ret; } /* to be called with priv->mutex held */ void __gf_rdma_request_context_destroy (gf_rdma_request_context_t *context) { gf_rdma_peer_t *peer = NULL; gf_rdma_private_t *priv = NULL; int32_t ret = 0; if (context == NULL) { goto out; } peer = context->peer; __gf_rdma_deregister_mr (context->mr, context->mr_count); priv = peer->trans->private; if (priv->connected) { ret = __gf_rdma_quota_put (peer); if (ret < 0) { gf_log ("rdma", GF_LOG_DEBUG, "failed to send " "message"); mem_put (context); __gf_rdma_disconnect (peer->trans); goto out; } } if (context->iobref != NULL) { iobref_unref (context->iobref); context->iobref = NULL; } if (context->rsp_iobref != NULL) { iobref_unref (context->rsp_iobref); context->rsp_iobref = NULL; } mem_put (context); out: return; } void gf_rdma_post_context_destroy (gf_rdma_post_context_t *ctx) { if (ctx == NULL) { goto out; } __gf_rdma_deregister_mr (ctx->mr, ctx->mr_count); if (ctx->iobref != NULL) { iobref_unref (ctx->iobref); } if (ctx->hdr_iobuf != NULL) { iobuf_unref (ctx->hdr_iobuf); } memset (ctx, 0, sizeof (*ctx)); out: return; } int gf_rdma_post_unref (gf_rdma_post_t *post) { int refcount = -1; if (post == NULL) { goto out; } pthread_mutex_lock (&post->lock); { refcount = --post->refcount; } pthread_mutex_unlock (&post->lock); if (refcount == 0) { gf_rdma_post_context_destroy (&post->ctx); if (post->type == GF_RDMA_SEND_POST) { gf_rdma_put_post (&post->device->sendq, post); } else { gf_rdma_post_recv (post->device->srq, post); } } out: return refcount; } int gf_rdma_post_get_refcount (gf_rdma_post_t *post) { int refcount = -1; if (post == NULL) { goto out; } pthread_mutex_lock (&post->lock); { refcount = post->refcount; } pthread_mutex_unlock (&post->lock); out: return refcount; } gf_rdma_post_t * gf_rdma_post_ref (gf_rdma_post_t *post) { if (post == NULL) { goto out; } pthread_mutex_lock (&post->lock); { post->refcount++; } pthread_mutex_unlock (&post->lock); out: return post; } int32_t __gf_rdma_ioq_churn_request (gf_rdma_peer_t *peer, gf_rdma_ioq_t *entry, gf_rdma_post_t *post) { gf_rdma_chunktype_t rtype = gf_rdma_noch; gf_rdma_chunktype_t wtype = gf_rdma_noch; uint64_t send_size = 0; gf_rdma_header_t *hdr = NULL; struct rpc_msg *rpc_msg = NULL; uint32_t *chunkptr = NULL; char *buf = NULL; int32_t ret = 0; gf_rdma_private_t *priv = NULL; gf_rdma_device_t *device = NULL; int chunk_count = 0; gf_rdma_request_context_t *request_ctx = NULL; uint32_t prog_payload_length = 0, len = 0; struct rpc_req *rpc_req = NULL; GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, peer, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, entry, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, post, out); if ((entry->msg.request.rsphdr_count != 0) && (entry->msg.request.rsp_payload_count != 0)) { ret = -1; gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "both write-chunklist and reply-chunk cannot be " "present"); goto out; } post->ctx.is_request = 1; priv = peer->trans->private; device = priv->device; hdr = (gf_rdma_header_t *)post->buf; send_size = iov_length (entry->rpchdr, entry->rpchdr_count) + iov_length (entry->proghdr, entry->proghdr_count) + GLUSTERFS_RDMA_MAX_HEADER_SIZE; if (entry->prog_payload_count != 0) { prog_payload_length = iov_length (entry->prog_payload, entry->prog_payload_count); } if (send_size > GLUSTERFS_RDMA_INLINE_THRESHOLD) { rtype = gf_rdma_areadch; } else if ((send_size + prog_payload_length) < GLUSTERFS_RDMA_INLINE_THRESHOLD) { rtype = gf_rdma_noch; } else if (entry->prog_payload_count != 0) { rtype = gf_rdma_readch; } if (entry->msg.request.rsphdr_count != 0) { wtype = gf_rdma_replych; } else if (entry->msg.request.rsp_payload_count != 0) { wtype = gf_rdma_writech; } if (rtype == gf_rdma_readch) { chunk_count += entry->prog_payload_count; } else if (rtype == gf_rdma_areadch) { chunk_count += entry->rpchdr_count; chunk_count += entry->proghdr_count; } if (wtype == gf_rdma_writech) { chunk_count += entry->msg.request.rsp_payload_count; } else if (wtype == gf_rdma_replych) { chunk_count += entry->msg.request.rsphdr_count; } if (chunk_count > GF_RDMA_MAX_SEGMENTS) { ret = -1; gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "chunk count(%d) exceeding maximum allowed RDMA " "segment count(%d)", chunk_count, GF_RDMA_MAX_SEGMENTS); goto out; } if ((wtype != gf_rdma_noch) || (rtype != gf_rdma_noch)) { request_ctx = mem_get (device->request_ctx_pool); if (request_ctx == NULL) { ret = -1; goto out; } memset (request_ctx, 0, sizeof (*request_ctx)); request_ctx->pool = device->request_ctx_pool; request_ctx->peer = peer; entry->msg.request.rpc_req->conn_private = request_ctx; if (entry->msg.request.rsp_iobref != NULL) { request_ctx->rsp_iobref = iobref_ref (entry->msg.request.rsp_iobref); } } rpc_msg = (struct rpc_msg *) entry->rpchdr[0].iov_base; hdr->rm_xid = rpc_msg->rm_xid; /* no need of hton32(rpc_msg->rm_xid), * since rpc_msg->rm_xid is already * hton32ed value of actual xid */ hdr->rm_vers = hton32 (GF_RDMA_VERSION); hdr->rm_credit = hton32 (peer->send_count); if (rtype == gf_rdma_areadch) { hdr->rm_type = hton32 (GF_RDMA_NOMSG); } else { hdr->rm_type = hton32 (GF_RDMA_MSG); } chunkptr = &hdr->rm_body.rm_chunks[0]; if (rtype != gf_rdma_noch) { ret = __gf_rdma_create_read_chunks (peer, entry, rtype, &chunkptr, request_ctx); if (ret != 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "creation of read chunks failed"); goto out; } } else { *chunkptr++ = 0; /* no read chunks */ } if (wtype != gf_rdma_noch) { ret = __gf_rdma_create_write_chunks (peer, entry, wtype, &chunkptr, request_ctx); if (ret != 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "creation of write/reply chunk failed"); goto out; } } else { *chunkptr++ = 0; /* no write chunks */ *chunkptr++ = 0; /* no reply chunk */ } buf = (char *)chunkptr; if (rtype != gf_rdma_areadch) { iov_unload (buf, entry->rpchdr, entry->rpchdr_count); buf += iov_length (entry->rpchdr, entry->rpchdr_count); iov_unload (buf, entry->proghdr, entry->proghdr_count); buf += iov_length (entry->proghdr, entry->proghdr_count); if (rtype != gf_rdma_readch) { iov_unload (buf, entry->prog_payload, entry->prog_payload_count); buf += iov_length (entry->prog_payload, entry->prog_payload_count); } } len = buf - post->buf; gf_rdma_post_ref (post); ret = gf_rdma_post_send (peer->qp, post, len); if (!ret) { ret = len; } else { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "gf_rdma_post_send (to %s) failed with ret = %d (%s)", peer->trans->peerinfo.identifier, ret, (ret > 0) ? strerror (ret) : ""); gf_rdma_post_unref (post); __gf_rdma_disconnect (peer->trans); ret = -1; } out: if (ret == -1) { rpc_req = entry->msg.request.rpc_req; if (request_ctx != NULL) { __gf_rdma_request_context_destroy (rpc_req->conn_private); } rpc_req->conn_private = NULL; } return ret; } static inline void __gf_rdma_fill_reply_header (gf_rdma_header_t *header, struct iovec *rpchdr, gf_rdma_reply_info_t *reply_info, int credits) { struct rpc_msg *rpc_msg = NULL; if (reply_info != NULL) { header->rm_xid = hton32 (reply_info->rm_xid); } else { rpc_msg = rpchdr[0].iov_base; /* assume rpchdr contains * only one vector. * (which is true) */ header->rm_xid = rpc_msg->rm_xid; } header->rm_type = hton32 (GF_RDMA_MSG); header->rm_vers = hton32 (GF_RDMA_VERSION); header->rm_credit = hton32 (credits); header->rm_body.rm_chunks[0] = 0; /* no read chunks */ header->rm_body.rm_chunks[1] = 0; /* no write chunks */ header->rm_body.rm_chunks[2] = 0; /* no reply chunks */ return; } int32_t __gf_rdma_send_reply_inline (gf_rdma_peer_t *peer, gf_rdma_ioq_t *entry, gf_rdma_post_t *post, gf_rdma_reply_info_t *reply_info) { gf_rdma_header_t *header = NULL; int32_t send_size = 0, ret = 0; char *buf = NULL; send_size = iov_length (entry->rpchdr, entry->rpchdr_count) + iov_length (entry->proghdr, entry->proghdr_count) + iov_length (entry->prog_payload, entry->prog_payload_count) + sizeof (gf_rdma_header_t); /* * remember, no chunklists in the * reply */ if (send_size > GLUSTERFS_RDMA_INLINE_THRESHOLD) { ret = __gf_rdma_send_error (peer, entry, post, reply_info, ERR_CHUNK); gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "msg size (%d) is greater than maximum size " "of msg that can be sent inlined (%d)", send_size, GLUSTERFS_RDMA_INLINE_THRESHOLD); goto out; } header = (gf_rdma_header_t *)post->buf; __gf_rdma_fill_reply_header (header, entry->rpchdr, reply_info, peer->send_count); buf = (char *)&header->rm_body.rm_chunks[3]; if (entry->rpchdr_count != 0) { iov_unload (buf, entry->rpchdr, entry->rpchdr_count); buf += iov_length (entry->rpchdr, entry->rpchdr_count); } if (entry->proghdr_count != 0) { iov_unload (buf, entry->proghdr, entry->proghdr_count); buf += iov_length (entry->proghdr, entry->proghdr_count); } if (entry->prog_payload_count != 0) { iov_unload (buf, entry->prog_payload, entry->prog_payload_count); buf += iov_length (entry->prog_payload, entry->prog_payload_count); } gf_rdma_post_ref (post); ret = gf_rdma_post_send (peer->qp, post, (buf - post->buf)); if (!ret) { ret = send_size; } else { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "posting send (to %s) failed with ret = %d (%s)", peer->trans->peerinfo.identifier, ret, (ret > 0) ? strerror (ret) : ""); gf_rdma_post_unref (post); __gf_rdma_disconnect (peer->trans); ret = -1; } out: return ret; } int32_t __gf_rdma_reply_encode_write_chunks (gf_rdma_peer_t *peer, uint32_t payload_size, gf_rdma_post_t *post, gf_rdma_reply_info_t *reply_info, uint32_t **ptr) { uint32_t chunk_size = 0; int32_t ret = -1; gf_rdma_write_array_t *target_array = NULL; int i = 0; target_array = (gf_rdma_write_array_t *)*ptr; for (i = 0; i < reply_info->wc_array->wc_nchunks; i++) { chunk_size += reply_info->wc_array->wc_array[i].wc_target.rs_length; } if (chunk_size < payload_size) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_DEBUG, "length of payload (%d) is exceeding the total " "write chunk length (%d)", payload_size, chunk_size); goto out; } target_array->wc_discrim = hton32 (1); for (i = 0; (i < reply_info->wc_array->wc_nchunks) && (payload_size != 0); i++) { target_array->wc_array[i].wc_target.rs_offset = hton64 (reply_info->wc_array->wc_array[i].wc_target.rs_offset); target_array->wc_array[i].wc_target.rs_length = hton32 (min (payload_size, reply_info->wc_array->wc_array[i].wc_target.rs_length)); } target_array->wc_nchunks = hton32 (i); target_array->wc_array[i].wc_target.rs_handle = 0; /* terminate chunklist */ ret = 0; *ptr = &target_array->wc_array[i].wc_target.rs_length; out: return ret; } static inline int32_t __gf_rdma_register_local_mr_for_rdma (gf_rdma_peer_t *peer, struct iovec *vector, int count, gf_rdma_post_context_t *ctx) { int i = 0; int32_t ret = -1; gf_rdma_private_t *priv = NULL; gf_rdma_device_t *device = NULL; GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, ctx, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, vector, out); priv = peer->trans->private; device = priv->device; for (i = 0; i < count; i++) { /* what if the memory is registered more than once? * Assume that a single write buffer is passed to afr, which * then passes it to its children. If more than one children * happen to use rdma, then the buffer is registered more than * once. * Ib-verbs specification says that multiple registrations of * same memory location is allowed. Refer to 10.6.3.8 of * Infiniband Architecture Specification Volume 1 * (Release 1.2.1) */ ctx->mr[ctx->mr_count] = ibv_reg_mr (device->pd, vector[i].iov_base, vector[i].iov_len, IBV_ACCESS_LOCAL_WRITE); if (ctx->mr[ctx->mr_count] == NULL) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "registering memory for IBV_ACCESS_LOCAL_WRITE " "failed (%s)", strerror (errno)); goto out; } ctx->mr_count++; } ret = 0; out: return ret; } /* 1. assumes xfer_len of data is pointed by vector(s) starting from vec[*idx] * 2. modifies vec */ int32_t __gf_rdma_write (gf_rdma_peer_t *peer, gf_rdma_post_t *post, struct iovec *vec, uint32_t xfer_len, int *idx, gf_rdma_write_chunk_t *writech) { int size = 0, num_sge = 0, i = 0; int32_t ret = -1; struct ibv_sge *sg_list = NULL; struct ibv_send_wr wr = { .opcode = IBV_WR_RDMA_WRITE, .send_flags = IBV_SEND_SIGNALED, }, *bad_wr; if ((peer == NULL) || (writech == NULL) || (idx == NULL) || (post == NULL) || (vec == NULL) || (xfer_len == 0)) { goto out; } for (i = *idx; size < xfer_len; i++) { size += vec[i].iov_len; } num_sge = i - *idx; sg_list = GF_CALLOC (num_sge, sizeof (struct ibv_sge), gf_common_mt_sge); if (sg_list == NULL) { ret = -1; goto out; } for ((i = *idx), (num_sge = 0); (xfer_len != 0); i++, num_sge++) { size = min (xfer_len, vec[i].iov_len); sg_list [num_sge].addr = (unsigned long)vec[i].iov_base; sg_list [num_sge].length = size; sg_list [num_sge].lkey = post->ctx.mr[i]->lkey; xfer_len -= size; } *idx = i; if (size < vec[i - 1].iov_len) { vec[i - 1].iov_base += size; vec[i - 1].iov_len -= size; *idx = i - 1; } wr.sg_list = sg_list; wr.num_sge = num_sge; wr.wr_id = (unsigned long) gf_rdma_post_ref (post); wr.wr.rdma.rkey = writech->wc_target.rs_handle; wr.wr.rdma.remote_addr = writech->wc_target.rs_offset; ret = ibv_post_send(peer->qp, &wr, &bad_wr); if (ret) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma write to " "client (%s) failed with ret = %d (%s)", peer->trans->peerinfo.identifier, ret, (ret > 0) ? strerror (ret) : ""); ret = -1; } GF_FREE (sg_list); out: return ret; } int32_t __gf_rdma_do_gf_rdma_write (gf_rdma_peer_t *peer, gf_rdma_post_t *post, struct iovec *vector, int count, struct iobref *iobref, gf_rdma_reply_info_t *reply_info) { int i = 0, payload_idx = 0; uint32_t payload_size = 0, xfer_len = 0; int32_t ret = -1; if (count != 0) { payload_size = iov_length (vector, count); } if (payload_size == 0) { ret = 0; goto out; } ret = __gf_rdma_register_local_mr_for_rdma (peer, vector, count, &post->ctx); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "registering memory region for rdma failed"); goto out; } post->ctx.iobref = iobref_ref (iobref); for (i = 0; (i < reply_info->wc_array->wc_nchunks) && (payload_size != 0); i++) { xfer_len = min (payload_size, reply_info->wc_array->wc_array[i].wc_target.rs_length); ret = __gf_rdma_write (peer, post, vector, xfer_len, &payload_idx, &reply_info->wc_array->wc_array[i]); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma write to client (%s) failed", peer->trans->peerinfo.identifier); goto out; } payload_size -= xfer_len; } ret = 0; out: return ret; } int32_t __gf_rdma_send_reply_type_nomsg (gf_rdma_peer_t *peer, gf_rdma_ioq_t *entry, gf_rdma_post_t *post, gf_rdma_reply_info_t *reply_info) { gf_rdma_header_t *header = NULL; char *buf = NULL; uint32_t payload_size = 0; int count = 0, i = 0; int32_t ret = 0; struct iovec vector[MAX_IOVEC]; header = (gf_rdma_header_t *)post->buf; __gf_rdma_fill_reply_header (header, entry->rpchdr, reply_info, peer->send_count); header->rm_type = hton32 (GF_RDMA_NOMSG); payload_size = iov_length (entry->rpchdr, entry->rpchdr_count) + iov_length (entry->proghdr, entry->proghdr_count); /* encode reply chunklist */ buf = (char *)&header->rm_body.rm_chunks[2]; ret = __gf_rdma_reply_encode_write_chunks (peer, payload_size, post, reply_info, (uint32_t **)&buf); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "encoding write chunks failed"); ret = __gf_rdma_send_error (peer, entry, post, reply_info, ERR_CHUNK); goto out; } gf_rdma_post_ref (post); for (i = 0; i < entry->rpchdr_count; i++) { vector[count++] = entry->rpchdr[i]; } for (i = 0; i < entry->proghdr_count; i++) { vector[count++] = entry->proghdr[i]; } ret = __gf_rdma_do_gf_rdma_write (peer, post, vector, count, entry->iobref, reply_info); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma write to peer (%s) failed", peer->trans->peerinfo.identifier); gf_rdma_post_unref (post); goto out; } ret = gf_rdma_post_send (peer->qp, post, (buf - post->buf)); if (ret) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "posting a send request to client (%s) failed with " "ret = %d (%s)", peer->trans->peerinfo.identifier, ret, (ret > 0) ? strerror (ret) : ""); ret = -1; gf_rdma_post_unref (post); } else { ret = payload_size; } out: return ret; } int32_t __gf_rdma_send_reply_type_msg (gf_rdma_peer_t *peer, gf_rdma_ioq_t *entry, gf_rdma_post_t *post, gf_rdma_reply_info_t *reply_info) { gf_rdma_header_t *header = NULL; int32_t send_size = 0, ret = 0; char *ptr = NULL; uint32_t payload_size = 0; send_size = iov_length (entry->rpchdr, entry->rpchdr_count) + iov_length (entry->proghdr, entry->proghdr_count) + GLUSTERFS_RDMA_MAX_HEADER_SIZE; if (send_size > GLUSTERFS_RDMA_INLINE_THRESHOLD) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "client has provided only write chunks, but the " "combined size of rpc and program header (%d) is " "exceeding the size of msg that can be sent using " "RDMA send (%d)", send_size, GLUSTERFS_RDMA_INLINE_THRESHOLD); ret = __gf_rdma_send_error (peer, entry, post, reply_info, ERR_CHUNK); goto out; } header = (gf_rdma_header_t *)post->buf; __gf_rdma_fill_reply_header (header, entry->rpchdr, reply_info, peer->send_count); payload_size = iov_length (entry->prog_payload, entry->prog_payload_count); ptr = (char *)&header->rm_body.rm_chunks[1]; ret = __gf_rdma_reply_encode_write_chunks (peer, payload_size, post, reply_info, (uint32_t **)&ptr); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "encoding write chunks failed"); ret = __gf_rdma_send_error (peer, entry, post, reply_info, ERR_CHUNK); goto out; } *(uint32_t *)ptr = 0; /* terminate reply chunklist */ ptr += sizeof (uint32_t); gf_rdma_post_ref (post); ret = __gf_rdma_do_gf_rdma_write (peer, post, entry->prog_payload, entry->prog_payload_count, entry->iobref, reply_info); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma write to peer " "(%s) failed", peer->trans->peerinfo.identifier); gf_rdma_post_unref (post); goto out; } iov_unload (ptr, entry->rpchdr, entry->rpchdr_count); ptr += iov_length (entry->rpchdr, entry->rpchdr_count); iov_unload (ptr, entry->proghdr, entry->proghdr_count); ptr += iov_length (entry->proghdr, entry->proghdr_count); ret = gf_rdma_post_send (peer->qp, post, (ptr - post->buf)); if (ret) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma send to client (%s) failed with ret = %d (%s)", peer->trans->peerinfo.identifier, ret, (ret > 0) ? strerror (ret) : ""); gf_rdma_post_unref (post); ret = -1; } else { ret = send_size + payload_size; } out: return ret; } void gf_rdma_reply_info_destroy (gf_rdma_reply_info_t *reply_info) { if (reply_info == NULL) { goto out; } if (reply_info->wc_array != NULL) { GF_FREE (reply_info->wc_array); reply_info->wc_array = NULL; } mem_put (reply_info); out: return; } gf_rdma_reply_info_t * gf_rdma_reply_info_alloc (gf_rdma_peer_t *peer) { gf_rdma_reply_info_t *reply_info = NULL; gf_rdma_private_t *priv = NULL; priv = peer->trans->private; reply_info = mem_get (priv->device->reply_info_pool); if (reply_info == NULL) { goto out; } memset (reply_info, 0, sizeof (*reply_info)); reply_info->pool = priv->device->reply_info_pool; out: return reply_info; } int32_t __gf_rdma_ioq_churn_reply (gf_rdma_peer_t *peer, gf_rdma_ioq_t *entry, gf_rdma_post_t *post) { gf_rdma_reply_info_t *reply_info = NULL; int32_t ret = -1; gf_rdma_chunktype_t type = gf_rdma_noch; GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, peer, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, entry, out); GF_VALIDATE_OR_GOTO (GF_RDMA_LOG_NAME, post, out); reply_info = entry->msg.reply_info; if (reply_info != NULL) { type = reply_info->type; } switch (type) { case gf_rdma_noch: ret = __gf_rdma_send_reply_inline (peer, entry, post, reply_info); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "failed to send reply to peer (%s) as an " "inlined rdma msg", peer->trans->peerinfo.identifier); } break; case gf_rdma_replych: ret = __gf_rdma_send_reply_type_nomsg (peer, entry, post, reply_info); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "failed to send reply to peer (%s) as " "RDMA_NOMSG", peer->trans->peerinfo.identifier); } break; case gf_rdma_writech: ret = __gf_rdma_send_reply_type_msg (peer, entry, post, reply_info); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "failed to send reply with write chunks " "to peer (%s)", peer->trans->peerinfo.identifier); } break; default: gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "invalid chunktype (%d) specified for sending reply " " (peer:%s)", type, peer->trans->peerinfo.identifier); break; } if (reply_info != NULL) { gf_rdma_reply_info_destroy (reply_info); } out: return ret; } int32_t __gf_rdma_ioq_churn_entry (gf_rdma_peer_t *peer, gf_rdma_ioq_t *entry) { int32_t ret = 0, quota = 0; gf_rdma_private_t *priv = NULL; gf_rdma_device_t *device = NULL; gf_rdma_options_t *options = NULL; gf_rdma_post_t *post = NULL; priv = peer->trans->private; options = &priv->options; device = priv->device; quota = __gf_rdma_quota_get (peer); if (quota > 0) { post = gf_rdma_get_post (&device->sendq); if (post == NULL) { post = gf_rdma_new_post (peer->trans, device, (options->send_size + 2048), GF_RDMA_SEND_POST); } if (post == NULL) { ret = -1; gf_log_callingfn (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "not able to get a post to send msg"); goto out; } if (entry->is_request) { ret = __gf_rdma_ioq_churn_request (peer, entry, post); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "failed to process request ioq entry " "to peer(%s)", peer->trans->peerinfo.identifier); } } else { ret = __gf_rdma_ioq_churn_reply (peer, entry, post); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "failed to process reply ioq entry " "to peer (%s)", peer->trans->peerinfo.identifier); } } if (ret != 0) { __gf_rdma_ioq_entry_free (entry); } } else { ret = 0; } out: return ret; } static int32_t __gf_rdma_ioq_churn (gf_rdma_peer_t *peer) { gf_rdma_ioq_t *entry = NULL; int32_t ret = 0; while (!list_empty (&peer->ioq)) { /* pick next entry */ entry = peer->ioq_next; ret = __gf_rdma_ioq_churn_entry (peer, entry); if (ret <= 0) break; } /* list_for_each_entry_safe (entry, dummy, &peer->ioq, list) { ret = __gf_rdma_ioq_churn_entry (peer, entry); if (ret <= 0) { break; } } */ return ret; } static int32_t gf_rdma_writev (rpc_transport_t *this, gf_rdma_ioq_t *entry) { int32_t ret = 0, need_append = 1; gf_rdma_private_t *priv = NULL; gf_rdma_peer_t *peer = NULL; priv = this->private; pthread_mutex_lock (&priv->write_mutex); { if (!priv->connected) { gf_log (this->name, GF_LOG_WARNING, "rdma is not connected to peer (%s)", this->peerinfo.identifier); ret = -1; goto unlock; } peer = &priv->peer; if (list_empty (&peer->ioq)) { ret = __gf_rdma_ioq_churn_entry (peer, entry); if (ret != 0) { need_append = 0; if (ret < 0) { gf_log (this->name, GF_LOG_WARNING, "processing ioq entry destined " "to (%s) failed", this->peerinfo.identifier); } } } if (need_append) { list_add_tail (&entry->list, &peer->ioq); } } unlock: pthread_mutex_unlock (&priv->write_mutex); return ret; } gf_rdma_ioq_t * gf_rdma_ioq_new (rpc_transport_t *this, rpc_transport_data_t *data) { gf_rdma_ioq_t *entry = NULL; int count = 0, i = 0; rpc_transport_msg_t *msg = NULL; gf_rdma_private_t *priv = NULL; if ((data == NULL) || (this == NULL)) { goto out; } priv = this->private; entry = mem_get (priv->device->ioq_pool); if (entry == NULL) { goto out; } memset (entry, 0, sizeof (*entry)); entry->pool = priv->device->ioq_pool; if (data->is_request) { msg = &data->data.req.msg; if (data->data.req.rsp.rsphdr_count != 0) { for (i = 0; i < data->data.req.rsp.rsphdr_count; i++) { entry->msg.request.rsphdr_vec[i] = data->data.req.rsp.rsphdr[i]; } entry->msg.request.rsphdr_count = data->data.req.rsp.rsphdr_count; } if (data->data.req.rsp.rsp_payload_count != 0) { for (i = 0; i < data->data.req.rsp.rsp_payload_count; i++) { entry->msg.request.rsp_payload[i] = data->data.req.rsp.rsp_payload[i]; } entry->msg.request.rsp_payload_count = data->data.req.rsp.rsp_payload_count; } entry->msg.request.rpc_req = data->data.req.rpc_req; if (data->data.req.rsp.rsp_iobref != NULL) { entry->msg.request.rsp_iobref = iobref_ref (data->data.req.rsp.rsp_iobref); } } else { msg = &data->data.reply.msg; entry->msg.reply_info = data->data.reply.private; } entry->is_request = data->is_request; count = msg->rpchdrcount + msg->proghdrcount + msg->progpayloadcount; GF_ASSERT (count <= MAX_IOVEC); if (msg->rpchdr != NULL) { memcpy (&entry->rpchdr[0], msg->rpchdr, sizeof (struct iovec) * msg->rpchdrcount); entry->rpchdr_count = msg->rpchdrcount; } if (msg->proghdr != NULL) { memcpy (&entry->proghdr[0], msg->proghdr, sizeof (struct iovec) * msg->proghdrcount); entry->proghdr_count = msg->proghdrcount; } if (msg->progpayload != NULL) { memcpy (&entry->prog_payload[0], msg->progpayload, sizeof (struct iovec) * msg->progpayloadcount); entry->prog_payload_count = msg->progpayloadcount; } if (msg->iobref != NULL) { entry->iobref = iobref_ref (msg->iobref); } INIT_LIST_HEAD (&entry->list); out: return entry; } int32_t gf_rdma_submit_request (rpc_transport_t *this, rpc_transport_req_t *req) { int32_t ret = 0; gf_rdma_ioq_t *entry = NULL; rpc_transport_data_t data = {0, }; if (req == NULL) { goto out; } data.is_request = 1; data.data.req = *req; entry = gf_rdma_ioq_new (this, &data); if (entry == NULL) { gf_log (this->name, GF_LOG_WARNING, "getting a new ioq entry failed (peer:%s)", this->peerinfo.identifier); goto out; } ret = gf_rdma_writev (this, entry); if (ret > 0) { ret = 0; } else if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "sending request to peer (%s) failed", this->peerinfo.identifier); rpc_transport_disconnect (this); } out: return ret; } int32_t gf_rdma_submit_reply (rpc_transport_t *this, rpc_transport_reply_t *reply) { int32_t ret = 0; gf_rdma_ioq_t *entry = NULL; rpc_transport_data_t data = {0, }; if (reply == NULL) { goto out; } data.data.reply = *reply; entry = gf_rdma_ioq_new (this, &data); if (entry == NULL) { gf_log (this->name, GF_LOG_WARNING, "getting a new ioq entry failed (peer:%s)", this->peerinfo.identifier); goto out; } ret = gf_rdma_writev (this, entry); if (ret > 0) { ret = 0; } else if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "sending request to peer (%s) failed", this->peerinfo.identifier); rpc_transport_disconnect (this); } out: return ret; } static int gf_rdma_register_peer (gf_rdma_device_t *device, int32_t qp_num, gf_rdma_peer_t *peer) { struct _qpent *ent = NULL; gf_rdma_qpreg_t *qpreg = NULL; int32_t hash = 0; int ret = -1; qpreg = &device->qpreg; hash = qp_num % 42; pthread_mutex_lock (&qpreg->lock); { ent = qpreg->ents[hash].next; while ((ent != &qpreg->ents[hash]) && (ent->qp_num != qp_num)) { ent = ent->next; } if (ent->qp_num == qp_num) { ret = 0; goto unlock; } ent = (struct _qpent *) GF_CALLOC (1, sizeof (*ent), gf_common_mt_qpent); if (ent == NULL) { goto unlock; } /* TODO: ref reg->peer */ ent->peer = peer; ent->next = &qpreg->ents[hash]; ent->prev = ent->next->prev; ent->next->prev = ent; ent->prev->next = ent; ent->qp_num = qp_num; qpreg->count++; ret = 0; } unlock: pthread_mutex_unlock (&qpreg->lock); return ret; } static void gf_rdma_unregister_peer (gf_rdma_device_t *device, int32_t qp_num) { struct _qpent *ent = NULL; gf_rdma_qpreg_t *qpreg = NULL; int32_t hash = 0; qpreg = &device->qpreg; hash = qp_num % 42; pthread_mutex_lock (&qpreg->lock); { ent = qpreg->ents[hash].next; while ((ent != &qpreg->ents[hash]) && (ent->qp_num != qp_num)) ent = ent->next; if (ent->qp_num != qp_num) { pthread_mutex_unlock (&qpreg->lock); return; } ent->prev->next = ent->next; ent->next->prev = ent->prev; /* TODO: unref reg->peer */ GF_FREE (ent); qpreg->count--; } pthread_mutex_unlock (&qpreg->lock); } static gf_rdma_peer_t * __gf_rdma_lookup_peer (gf_rdma_device_t *device, int32_t qp_num) { struct _qpent *ent = NULL; gf_rdma_peer_t *peer = NULL; gf_rdma_qpreg_t *qpreg = NULL; int32_t hash = 0; qpreg = &device->qpreg; hash = qp_num % 42; ent = qpreg->ents[hash].next; while ((ent != &qpreg->ents[hash]) && (ent->qp_num != qp_num)) ent = ent->next; if (ent != &qpreg->ents[hash]) { peer = ent->peer; } return peer; } static void __gf_rdma_destroy_qp (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; priv = this->private; if (priv->peer.qp) { gf_rdma_unregister_peer (priv->device, priv->peer.qp->qp_num); rdma_destroy_qp (priv->peer.cm_id); } priv->peer.qp = NULL; return; } static int32_t gf_rdma_create_qp (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; gf_rdma_device_t *device = NULL; int32_t ret = 0; gf_rdma_peer_t *peer = NULL; char *device_name = NULL; priv = this->private; peer = &priv->peer; device_name = (char *)ibv_get_device_name (peer->cm_id->verbs->device); if (device_name == NULL) { ret = -1; gf_log (this->name, GF_LOG_WARNING, "cannot get device_name"); goto out; } device = gf_rdma_get_device (this, peer->cm_id->verbs, device_name); if (device == NULL) { ret = -1; gf_log (this->name, GF_LOG_WARNING, "cannot get device for " "device %s", device_name); goto out; } if (priv->device == NULL) { priv->device = device; } struct ibv_qp_init_attr init_attr = { .send_cq = device->send_cq, .recv_cq = device->recv_cq, .srq = device->srq, .cap = { .max_send_wr = peer->send_count, .max_recv_wr = peer->recv_count, .max_send_sge = 2, .max_recv_sge = 1 }, .qp_type = IBV_QPT_RC }; ret = rdma_create_qp(peer->cm_id, device->pd, &init_attr); if (ret != 0) { gf_log (peer->trans->name, GF_LOG_CRITICAL, "%s: could not create QP (%s)", this->name, strerror (errno)); ret = -1; goto out; } peer->qp = peer->cm_id->qp; ret = gf_rdma_register_peer (device, peer->qp->qp_num, peer); out: if (ret == -1) __gf_rdma_destroy_qp (this); return ret; } static int32_t __gf_rdma_teardown (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; gf_rdma_peer_t *peer = NULL; priv = this->private; peer = &priv->peer; if (peer->cm_id->qp != NULL) { __gf_rdma_destroy_qp (this); } if (!list_empty (&priv->peer.ioq)) { __gf_rdma_ioq_flush (peer); } if (peer->cm_id != NULL) { rdma_destroy_id (peer->cm_id); peer->cm_id = NULL; } /* TODO: decrement cq size */ return 0; } static int32_t gf_rdma_teardown (rpc_transport_t *this) { int32_t ret = 0; gf_rdma_private_t *priv = NULL; if (this == NULL) { goto out; } priv = this->private; pthread_mutex_lock (&priv->write_mutex); { ret = __gf_rdma_teardown (this); } pthread_mutex_unlock (&priv->write_mutex); out: return ret; } /* * allocates new memory to hold write-chunklist. New memory is needed since * write-chunklist will be used while sending reply and the post holding initial * write-chunklist sent from client will be put back to srq before a pollin * event is sent to upper layers. */ int32_t gf_rdma_get_write_chunklist (char **ptr, gf_rdma_write_array_t **write_ary) { gf_rdma_write_array_t *from = NULL, *to = NULL; int32_t ret = -1, size = 0, i = 0; from = (gf_rdma_write_array_t *) *ptr; if (from->wc_discrim == 0) { ret = 0; goto out; } from->wc_nchunks = ntoh32 (from->wc_nchunks); size = sizeof (*from) + (sizeof (gf_rdma_write_chunk_t) * from->wc_nchunks); to = GF_CALLOC (1, size, gf_common_mt_char); if (to == NULL) { ret = -1; goto out; } to->wc_discrim = ntoh32 (from->wc_discrim); to->wc_nchunks = from->wc_nchunks; for (i = 0; i < to->wc_nchunks; i++) { to->wc_array[i].wc_target.rs_handle = ntoh32 (from->wc_array[i].wc_target.rs_handle); to->wc_array[i].wc_target.rs_length = ntoh32 (from->wc_array[i].wc_target.rs_length); to->wc_array[i].wc_target.rs_offset = ntoh64 (from->wc_array[i].wc_target.rs_offset); } *write_ary = to; ret = 0; *ptr = (char *)&from->wc_array[i].wc_target.rs_handle; out: return ret; } /* * does not allocate new memory to hold read-chunklist. New memory is not * needed, since post is not put back to srq till we've completed all the * rdma-reads and hence readchunk-list can point to memory held by post. */ int32_t gf_rdma_get_read_chunklist (char **ptr, gf_rdma_read_chunk_t **readch) { int32_t ret = -1; gf_rdma_read_chunk_t *chunk = NULL; int i = 0; chunk = (gf_rdma_read_chunk_t *)*ptr; if (chunk[0].rc_discrim == 0) { ret = 0; goto out; } for (i = 0; chunk[i].rc_discrim != 0; i++) { chunk[i].rc_discrim = ntoh32 (chunk[i].rc_discrim); chunk[i].rc_position = ntoh32 (chunk[i].rc_position); chunk[i].rc_target.rs_handle = ntoh32 (chunk[i].rc_target.rs_handle); chunk[i].rc_target.rs_length = ntoh32 (chunk[i].rc_target.rs_length); chunk[i].rc_target.rs_offset = ntoh64 (chunk[i].rc_target.rs_offset); } *readch = &chunk[0]; ret = 0; *ptr = (char *)&chunk[i].rc_discrim; out: return ret; } static inline int32_t gf_rdma_decode_error_msg (gf_rdma_peer_t *peer, gf_rdma_post_t *post, size_t bytes_in_post) { gf_rdma_header_t *header = NULL; struct iobuf *iobuf = NULL; struct iobref *iobref = NULL; int32_t ret = -1; struct rpc_msg rpc_msg = {0, }; header = (gf_rdma_header_t *)post->buf; header->rm_body.rm_error.rm_type = ntoh32 (header->rm_body.rm_error.rm_type); if (header->rm_body.rm_error.rm_type == ERR_VERS) { header->rm_body.rm_error.rm_version.gf_rdma_vers_low = ntoh32 (header->rm_body.rm_error.rm_version.gf_rdma_vers_low); header->rm_body.rm_error.rm_version.gf_rdma_vers_high = ntoh32 (header->rm_body.rm_error.rm_version.gf_rdma_vers_high); } rpc_msg.rm_xid = header->rm_xid; rpc_msg.rm_direction = REPLY; rpc_msg.rm_reply.rp_stat = MSG_DENIED; iobuf = iobuf_get2 (peer->trans->ctx->iobuf_pool, bytes_in_post); if (iobuf == NULL) { ret = -1; goto out; } post->ctx.iobref = iobref = iobref_new (); if (iobref == NULL) { ret = -1; goto out; } iobref_add (iobref, iobuf); iobuf_unref (iobuf); ret = rpc_reply_to_xdr (&rpc_msg, iobuf_ptr (iobuf), iobuf_pagesize (iobuf), &post->ctx.vector[0]); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "Failed to create RPC reply"); goto out; } post->ctx.count = 1; iobuf = NULL; iobref = NULL; out: if (ret == -1) { if (iobuf != NULL) { iobuf_unref (iobuf); } if (iobref != NULL) { iobref_unref (iobref); } } return 0; } int32_t gf_rdma_decode_msg (gf_rdma_peer_t *peer, gf_rdma_post_t *post, gf_rdma_read_chunk_t **readch, size_t bytes_in_post) { int32_t ret = -1; gf_rdma_header_t *header = NULL; gf_rdma_reply_info_t *reply_info = NULL; char *ptr = NULL; gf_rdma_write_array_t *write_ary = NULL; size_t header_len = 0; header = (gf_rdma_header_t *)post->buf; ptr = (char *)&header->rm_body.rm_chunks[0]; ret = gf_rdma_get_read_chunklist (&ptr, readch); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot get read chunklist from msg"); goto out; } /* skip terminator of read-chunklist */ ptr = ptr + sizeof (uint32_t); ret = gf_rdma_get_write_chunklist (&ptr, &write_ary); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot get write chunklist from msg"); goto out; } /* skip terminator of write-chunklist */ ptr = ptr + sizeof (uint32_t); if (write_ary != NULL) { reply_info = gf_rdma_reply_info_alloc (peer); if (reply_info == NULL) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "reply_info_alloc failed"); ret = -1; goto out; } reply_info->type = gf_rdma_writech; reply_info->wc_array = write_ary; reply_info->rm_xid = header->rm_xid; } else { ret = gf_rdma_get_write_chunklist (&ptr, &write_ary); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot get reply chunklist from msg"); goto out; } if (write_ary != NULL) { reply_info = gf_rdma_reply_info_alloc (peer); if (reply_info == NULL) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "reply_info_alloc_failed"); ret = -1; goto out; } reply_info->type = gf_rdma_replych; reply_info->wc_array = write_ary; reply_info->rm_xid = header->rm_xid; } } /* skip terminator of reply chunk */ ptr = ptr + sizeof (uint32_t); if (header->rm_type != GF_RDMA_NOMSG) { header_len = (long)ptr - (long)post->buf; post->ctx.vector[0].iov_len = (bytes_in_post - header_len); post->ctx.hdr_iobuf = iobuf_get2 (peer->trans->ctx->iobuf_pool, (bytes_in_post - header_len)); if (post->ctx.hdr_iobuf == NULL) { ret = -1; goto out; } post->ctx.vector[0].iov_base = iobuf_ptr (post->ctx.hdr_iobuf); memcpy (post->ctx.vector[0].iov_base, ptr, post->ctx.vector[0].iov_len); post->ctx.count = 1; } post->ctx.reply_info = reply_info; out: if (ret == -1) { if (*readch != NULL) { GF_FREE (*readch); *readch = NULL; } GF_FREE (write_ary); } return ret; } /* Assumes only one of either write-chunklist or a reply chunk is present */ int32_t gf_rdma_decode_header (gf_rdma_peer_t *peer, gf_rdma_post_t *post, gf_rdma_read_chunk_t **readch, size_t bytes_in_post) { int32_t ret = -1; gf_rdma_header_t *header = NULL; header = (gf_rdma_header_t *)post->buf; header->rm_xid = ntoh32 (header->rm_xid); header->rm_vers = ntoh32 (header->rm_vers); header->rm_credit = ntoh32 (header->rm_credit); header->rm_type = ntoh32 (header->rm_type); switch (header->rm_type) { case GF_RDMA_MSG: case GF_RDMA_NOMSG: ret = gf_rdma_decode_msg (peer, post, readch, bytes_in_post); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot decode msg of type (%d)", header->rm_type); } break; case GF_RDMA_MSGP: gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma msg of msg-type GF_RDMA_MSGP should not have " "been received"); ret = -1; break; case GF_RDMA_DONE: gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma msg of msg-type GF_RDMA_DONE should not have " "been received"); ret = -1; break; case GF_RDMA_ERROR: gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "received a msg of type RDMA_ERROR"); ret = gf_rdma_decode_error_msg (peer, post, bytes_in_post); break; default: gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "unknown rdma msg-type (%d)", header->rm_type); } return ret; } int32_t __gf_rdma_read (gf_rdma_peer_t *peer, gf_rdma_post_t *post, struct iovec *to, gf_rdma_read_chunk_t *readch) { int32_t ret = -1; struct ibv_sge list = {0, }; struct ibv_send_wr wr = {0, }, *bad_wr = NULL; ret = __gf_rdma_register_local_mr_for_rdma (peer, to, 1, &post->ctx); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "registering local memory for rdma read failed"); goto out; } list.addr = (unsigned long) to->iov_base; list.length = to->iov_len; list.lkey = post->ctx.mr[post->ctx.mr_count - 1]->lkey; wr.wr_id = (unsigned long) gf_rdma_post_ref (post); wr.sg_list = &list; wr.num_sge = 1; wr.opcode = IBV_WR_RDMA_READ; wr.send_flags = IBV_SEND_SIGNALED; wr.wr.rdma.remote_addr = readch->rc_target.rs_offset; wr.wr.rdma.rkey = readch->rc_target.rs_handle; ret = ibv_post_send (peer->qp, &wr, &bad_wr); if (ret) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma read from client " "(%s) failed with ret = %d (%s)", peer->trans->peerinfo.identifier, ret, (ret > 0) ? strerror (ret) : ""); ret = -1; gf_rdma_post_unref (post); } out: return ret; } int32_t gf_rdma_do_reads (gf_rdma_peer_t *peer, gf_rdma_post_t *post, gf_rdma_read_chunk_t *readch) { int32_t ret = -1, i = 0, count = 0; size_t size = 0; char *ptr = NULL; struct iobuf *iobuf = NULL; gf_rdma_private_t *priv = NULL; priv = peer->trans->private; for (i = 0; readch[i].rc_discrim != 0; i++) { size += readch[i].rc_target.rs_length; } if (i == 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "message type specified as rdma-read but there are no " "rdma read-chunks present"); goto out; } post->ctx.gf_rdma_reads = i; iobuf = iobuf_get2 (peer->trans->ctx->iobuf_pool, size); if (iobuf == NULL) { goto out; } if (post->ctx.iobref == NULL) { post->ctx.iobref = iobref_new (); if (post->ctx.iobref == NULL) { iobuf_unref (iobuf); goto out; } } iobref_add (post->ctx.iobref, iobuf); iobuf_unref (iobuf); ptr = iobuf_ptr (iobuf); iobuf = NULL; pthread_mutex_lock (&priv->write_mutex); { if (!priv->connected) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "transport not connected to peer (%s), " "not doing rdma reads", peer->trans->peerinfo.identifier); goto unlock; } for (i = 0; readch[i].rc_discrim != 0; i++) { count = post->ctx.count++; post->ctx.vector[count].iov_base = ptr; post->ctx.vector[count].iov_len = readch[i].rc_target.rs_length; ret = __gf_rdma_read (peer, post, &post->ctx.vector[count], &readch[i]); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma read from peer (%s) failed", peer->trans->peerinfo.identifier); goto unlock; } ptr += readch[i].rc_target.rs_length; } ret = 0; } unlock: pthread_mutex_unlock (&priv->write_mutex); out: if (ret == -1) { if (iobuf != NULL) { iobuf_unref (iobuf); } } return ret; } int32_t gf_rdma_pollin_notify (gf_rdma_peer_t *peer, gf_rdma_post_t *post) { int32_t ret = -1; enum msg_type msg_type = 0; struct rpc_req *rpc_req = NULL; gf_rdma_request_context_t *request_context = NULL; rpc_request_info_t request_info = {0, }; gf_rdma_private_t *priv = NULL; uint32_t *ptr = NULL; rpc_transport_pollin_t *pollin = NULL; if ((peer == NULL) || (post == NULL)) { goto out; } if (post->ctx.iobref == NULL) { post->ctx.iobref = iobref_new (); if (post->ctx.iobref == NULL) { goto out; } /* handling the case where both hdr and payload of * GF_FOP_READ_CBK were received in a single iobuf * because of server sending entire msg as inline without * doing rdma writes. */ if (post->ctx.hdr_iobuf) iobref_add (post->ctx.iobref, post->ctx.hdr_iobuf); } pollin = rpc_transport_pollin_alloc (peer->trans, post->ctx.vector, post->ctx.count, post->ctx.hdr_iobuf, post->ctx.iobref, post->ctx.reply_info); if (pollin == NULL) { goto out; } ptr = (uint32_t *)pollin->vector[0].iov_base; request_info.xid = ntoh32 (*ptr); msg_type = ntoh32 (*(ptr + 1)); if (msg_type == REPLY) { ret = rpc_transport_notify (peer->trans, RPC_TRANSPORT_MAP_XID_REQUEST, &request_info); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_DEBUG, "cannot get request information from rpc " "layer"); goto out; } rpc_req = request_info.rpc_req; if (rpc_req == NULL) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_DEBUG, "rpc request structure not found"); ret = -1; goto out; } request_context = rpc_req->conn_private; rpc_req->conn_private = NULL; priv = peer->trans->private; if (request_context != NULL) { pthread_mutex_lock (&priv->write_mutex); { __gf_rdma_request_context_destroy (request_context); } pthread_mutex_unlock (&priv->write_mutex); } else { gf_rdma_quota_put (peer); } pollin->is_reply = 1; } ret = rpc_transport_notify (peer->trans, RPC_TRANSPORT_MSG_RECEIVED, pollin); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "transport_notify failed"); } out: if (pollin != NULL) { pollin->private = NULL; rpc_transport_pollin_destroy (pollin); } return ret; } int32_t gf_rdma_recv_reply (gf_rdma_peer_t *peer, gf_rdma_post_t *post) { int32_t ret = -1; gf_rdma_header_t *header = NULL; gf_rdma_reply_info_t *reply_info = NULL; gf_rdma_write_array_t *wc_array = NULL; int i = 0; uint32_t *ptr = NULL; gf_rdma_request_context_t *ctx = NULL; rpc_request_info_t request_info = {0, }; struct rpc_req *rpc_req = NULL; header = (gf_rdma_header_t *)post->buf; reply_info = post->ctx.reply_info; /* no write chunklist, just notify upper layers */ if (reply_info == NULL) { ret = 0; goto out; } wc_array = reply_info->wc_array; if (header->rm_type == GF_RDMA_NOMSG) { post->ctx.vector[0].iov_base = (void *)(long)wc_array->wc_array[0].wc_target.rs_offset; post->ctx.vector[0].iov_len = wc_array->wc_array[0].wc_target.rs_length; post->ctx.count = 1; } else { for (i = 0; i < wc_array->wc_nchunks; i++) { post->ctx.vector[i + 1].iov_base = (void *)(long)wc_array->wc_array[i].wc_target.rs_offset; post->ctx.vector[i + 1].iov_len = wc_array->wc_array[i].wc_target.rs_length; } post->ctx.count += wc_array->wc_nchunks; } ptr = (uint32_t *)post->ctx.vector[0].iov_base; request_info.xid = ntoh32 (*ptr); ret = rpc_transport_notify (peer->trans, RPC_TRANSPORT_MAP_XID_REQUEST, &request_info); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "cannot get request information (peer:%s) from rpc " "layer", peer->trans->peerinfo.identifier); goto out; } rpc_req = request_info.rpc_req; if (rpc_req == NULL) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rpc request structure not found"); ret = -1; goto out; } ctx = rpc_req->conn_private; if ((post->ctx.iobref == NULL) && ctx->rsp_iobref) { post->ctx.iobref = iobref_ref (ctx->rsp_iobref); } ret = 0; gf_rdma_reply_info_destroy (reply_info); out: if (ret == 0) { ret = gf_rdma_pollin_notify (peer, post); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "pollin notify failed"); } } return ret; } static inline int32_t gf_rdma_recv_request (gf_rdma_peer_t *peer, gf_rdma_post_t *post, gf_rdma_read_chunk_t *readch) { int32_t ret = -1; if (readch != NULL) { ret = gf_rdma_do_reads (peer, post, readch); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma read from peer (%s) failed", peer->trans->peerinfo.identifier); } } else { ret = gf_rdma_pollin_notify (peer, post); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "pollin notification failed"); } } return ret; } void gf_rdma_process_recv (gf_rdma_peer_t *peer, struct ibv_wc *wc) { gf_rdma_post_t *post = NULL; gf_rdma_read_chunk_t *readch = NULL; int ret = -1; uint32_t *ptr = NULL; enum msg_type msg_type = 0; gf_rdma_header_t *header = NULL; gf_rdma_private_t *priv = NULL; post = (gf_rdma_post_t *) (long) wc->wr_id; if (post == NULL) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "no post found in successful work completion element"); goto out; } ret = gf_rdma_decode_header (peer, post, &readch, wc->byte_len); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "decoding of header failed"); goto out; } header = (gf_rdma_header_t *)post->buf; priv = peer->trans->private; pthread_mutex_lock (&priv->write_mutex); { if (!priv->peer.quota_set) { priv->peer.quota_set = 1; /* Initially peer.quota is set to 1 as per RFC 5666. We * have to account for the quota used while sending * first msg (which may or may not be returned to pool * at this point) while deriving peer.quota from * header->rm_credit. Hence the arithmatic below, * instead of directly setting it to header->rm_credit. */ priv->peer.quota = header->rm_credit - ( 1 - priv->peer.quota); } } pthread_mutex_unlock (&priv->write_mutex); switch (header->rm_type) { case GF_RDMA_MSG: ptr = (uint32_t *)post->ctx.vector[0].iov_base; msg_type = ntoh32 (*(ptr + 1)); break; case GF_RDMA_NOMSG: if (readch != NULL) { msg_type = CALL; } else { msg_type = REPLY; } break; case GF_RDMA_ERROR: if (header->rm_body.rm_error.rm_type == ERR_CHUNK) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "peer (%s), couldn't encode or decode the msg " "properly or write chunks were not provided " "for replies that were bigger than " "RDMA_INLINE_THRESHOLD (%d)", peer->trans->peerinfo.identifier, GLUSTERFS_RDMA_INLINE_THRESHOLD); ret = gf_rdma_pollin_notify (peer, post); if (ret == -1) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_DEBUG, "pollin notification failed"); } goto out; } else { gf_log (GF_RDMA_LOG_NAME, GF_LOG_ERROR, "an error has happened while transmission of " "msg, disconnecting the transport"); ret = -1; goto out; } default: gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "invalid rdma msg-type (%d)", header->rm_type); goto out; } if (msg_type == CALL) { ret = gf_rdma_recv_request (peer, post, readch); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "receiving a request from peer (%s) failed", peer->trans->peerinfo.identifier); } } else { ret = gf_rdma_recv_reply (peer, post); if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "receiving a reply from peer (%s) failed", peer->trans->peerinfo.identifier); } } out: if (ret == -1) { rpc_transport_disconnect (peer->trans); } return; } void * gf_rdma_async_event_thread (void *context) { struct ibv_async_event event; int ret; while (1) { do { ret = ibv_get_async_event((struct ibv_context *)context, &event); if (ret && errno != EINTR) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "Error getting event (%s)", strerror (errno)); } } while(ret && errno == EINTR); switch (event.event_type) { case IBV_EVENT_SRQ_LIMIT_REACHED: gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "received srq_limit reached"); break; default: gf_log (GF_RDMA_LOG_NAME, GF_LOG_DEBUG, "event (%d) received", event.event_type); break; } ibv_ack_async_event(&event); } return 0; } static void * gf_rdma_recv_completion_proc (void *data) { struct ibv_comp_channel *chan = NULL; gf_rdma_device_t *device = NULL;; gf_rdma_post_t *post = NULL; gf_rdma_peer_t *peer = NULL; struct ibv_cq *event_cq = NULL; struct ibv_wc wc = {0, }; void *event_ctx = NULL; int32_t ret = 0; chan = data; while (1) { ret = ibv_get_cq_event (chan, &event_cq, &event_ctx); if (ret) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_ERROR, "ibv_get_cq_event failed, terminating recv " "thread %d (%d)", ret, errno); continue; } device = event_ctx; ret = ibv_req_notify_cq (event_cq, 0); if (ret) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_ERROR, "ibv_req_notify_cq on %s failed, terminating " "recv thread: %d (%d)", device->device_name, ret, errno); continue; } device = (gf_rdma_device_t *) event_ctx; while ((ret = ibv_poll_cq (event_cq, 1, &wc)) > 0) { post = (gf_rdma_post_t *) (long) wc.wr_id; pthread_mutex_lock (&device->qpreg.lock); { peer = __gf_rdma_lookup_peer (device, wc.qp_num); /* * keep a refcount on transport so that it * does not get freed because of some error * indicated by wc.status till we are done * with usage of peer and thereby that of trans. */ if (peer != NULL) { rpc_transport_ref (peer->trans); } } pthread_mutex_unlock (&device->qpreg.lock); if (wc.status != IBV_WC_SUCCESS) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_ERROR, "recv work request on `%s' returned " "error (%d)", device->device_name, wc.status); if (peer) { rpc_transport_unref (peer->trans); rpc_transport_disconnect (peer->trans); } if (post) { gf_rdma_post_unref (post); } continue; } if (peer) { gf_rdma_process_recv (peer, &wc); rpc_transport_unref (peer->trans); } else { gf_log (GF_RDMA_LOG_NAME, GF_LOG_DEBUG, "could not lookup peer for qp_num: %d", wc.qp_num); } gf_rdma_post_unref (post); } if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_ERROR, "ibv_poll_cq on `%s' returned error " "(ret = %d, errno = %d)", device->device_name, ret, errno); continue; } ibv_ack_cq_events (event_cq, 1); } return NULL; } void gf_rdma_handle_failed_send_completion (gf_rdma_peer_t *peer, struct ibv_wc *wc) { gf_rdma_post_t *post = NULL; gf_rdma_device_t *device = NULL; gf_rdma_private_t *priv = NULL; if (peer != NULL) { priv = peer->trans->private; if (priv != NULL) { device = priv->device; } } post = (gf_rdma_post_t *) (long) wc->wr_id; gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "send work request on `%s' returned error " "wc.status = %d, wc.vendor_err = %d, post->buf = %p, " "wc.byte_len = %d, post->reused = %d", (device != NULL) ? device->device_name : NULL, wc->status, wc->vendor_err, post->buf, wc->byte_len, post->reused); if (wc->status == IBV_WC_RETRY_EXC_ERR) { gf_log ("rdma", GF_LOG_ERROR, "connection between client and" " server not working. check by running " "'ibv_srq_pingpong'. also make sure subnet manager" " is running (eg: 'opensm'), or check if rdma port is " "valid (or active) by running 'ibv_devinfo'. contact " "Gluster Support Team if the problem persists."); } if (peer) { rpc_transport_disconnect (peer->trans); } return; } void gf_rdma_handle_successful_send_completion (gf_rdma_peer_t *peer, struct ibv_wc *wc) { gf_rdma_post_t *post = NULL; int reads = 0, ret = 0; gf_rdma_header_t *header = NULL; if (wc->opcode != IBV_WC_RDMA_READ) { goto out; } post = (gf_rdma_post_t *)(long) wc->wr_id; pthread_mutex_lock (&post->lock); { reads = --post->ctx.gf_rdma_reads; } pthread_mutex_unlock (&post->lock); if (reads != 0) { /* if it is not the last rdma read, we've got nothing to do */ goto out; } header = (gf_rdma_header_t *)post->buf; if (header->rm_type == GF_RDMA_NOMSG) { post->ctx.count = 1; post->ctx.vector[0].iov_len += post->ctx.vector[1].iov_len; } ret = gf_rdma_pollin_notify (peer, post); if ((ret == -1) && (peer != NULL)) { rpc_transport_disconnect (peer->trans); } out: return; } static void * gf_rdma_send_completion_proc (void *data) { struct ibv_comp_channel *chan = NULL; gf_rdma_post_t *post = NULL; gf_rdma_peer_t *peer = NULL; struct ibv_cq *event_cq = NULL; void *event_ctx = NULL; gf_rdma_device_t *device = NULL; struct ibv_wc wc = {0, }; char is_request = 0; int32_t ret = 0, quota_ret = 0; chan = data; while (1) { ret = ibv_get_cq_event (chan, &event_cq, &event_ctx); if (ret) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_ERROR, "ibv_get_cq_event on failed, terminating " "send thread: %d (%d)", ret, errno); continue; } device = event_ctx; ret = ibv_req_notify_cq (event_cq, 0); if (ret) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_ERROR, "ibv_req_notify_cq on %s failed, terminating " "send thread: %d (%d)", device->device_name, ret, errno); continue; } while ((ret = ibv_poll_cq (event_cq, 1, &wc)) > 0) { post = (gf_rdma_post_t *) (long) wc.wr_id; pthread_mutex_lock (&device->qpreg.lock); { peer = __gf_rdma_lookup_peer (device, wc.qp_num); /* * keep a refcount on transport so that it * does not get freed because of some error * indicated by wc.status, till we are done * with usage of peer and thereby that of trans. */ if (peer != NULL) { rpc_transport_ref (peer->trans); } } pthread_mutex_unlock (&device->qpreg.lock); if (wc.status != IBV_WC_SUCCESS) { gf_rdma_handle_failed_send_completion (peer, &wc); } else { gf_rdma_handle_successful_send_completion (peer, &wc); } if (post) { is_request = post->ctx.is_request; ret = gf_rdma_post_unref (post); if ((ret == 0) && (wc.status == IBV_WC_SUCCESS) && !is_request && (post->type == GF_RDMA_SEND_POST) && (peer != NULL)) { /* An GF_RDMA_RECV_POST can end up in * gf_rdma_send_completion_proc for * rdma-reads, and we do not take * quota for getting an GF_RDMA_RECV_POST. */ /* * if it is request, quota is returned * after reply has come. */ quota_ret = gf_rdma_quota_put (peer); if (quota_ret < 0) { gf_log ("rdma", GF_LOG_DEBUG, "failed to send " "message"); } } } if (peer) { rpc_transport_unref (peer->trans); } else { gf_log (GF_RDMA_LOG_NAME, GF_LOG_DEBUG, "could not lookup peer for qp_num: %d", wc.qp_num); } } if (ret < 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_ERROR, "ibv_poll_cq on `%s' returned error (ret = %d," " errno = %d)", device->device_name, ret, errno); continue; } ibv_ack_cq_events (event_cq, 1); } return NULL; } static void gf_rdma_options_init (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; gf_rdma_options_t *options = NULL; int32_t mtu = 0; data_t *temp = NULL; /* TODO: validate arguments from options below */ priv = this->private; options = &priv->options; options->send_size = GLUSTERFS_RDMA_INLINE_THRESHOLD;/*this->ctx->page_size * 4; 512 KB*/ options->recv_size = GLUSTERFS_RDMA_INLINE_THRESHOLD;/*this->ctx->page_size * 4; 512 KB*/ options->send_count = 4096; options->recv_count = 4096; options->attr_timeout = GF_RDMA_TIMEOUT; options->attr_retry_cnt = GF_RDMA_RETRY_CNT; options->attr_rnr_retry = GF_RDMA_RNR_RETRY; temp = dict_get (this->options, "transport.rdma.work-request-send-count"); if (temp) options->send_count = data_to_int32 (temp); temp = dict_get (this->options, "transport.rdma.work-request-recv-count"); if (temp) options->recv_count = data_to_int32 (temp); temp = dict_get (this->options, "transport.rdma.attr-timeout"); if (temp) options->attr_timeout = data_to_uint8 (temp); temp = dict_get (this->options, "transport.rdma.attr-retry-cnt"); if (temp) options->attr_retry_cnt = data_to_uint8 (temp); temp = dict_get (this->options, "transport.rdma.attr-rnr-retry"); if (temp) options->attr_rnr_retry = data_to_uint8 (temp); options->port = 1; temp = dict_get (this->options, "transport.rdma.port"); if (temp) options->port = data_to_uint64 (temp); options->mtu = mtu = IBV_MTU_2048; temp = dict_get (this->options, "transport.rdma.mtu"); if (temp) mtu = data_to_int32 (temp); switch (mtu) { case 256: options->mtu = IBV_MTU_256; break; case 512: options->mtu = IBV_MTU_512; break; case 1024: options->mtu = IBV_MTU_1024; break; case 2048: options->mtu = IBV_MTU_2048; break; case 4096: options->mtu = IBV_MTU_4096; break; default: if (temp) gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "%s: unrecognized MTU value '%s', defaulting " "to '2048'", this->name, data_to_str (temp)); else gf_log (GF_RDMA_LOG_NAME, GF_LOG_TRACE, "%s: defaulting MTU to '2048'", this->name); options->mtu = IBV_MTU_2048; break; } temp = dict_get (this->options, "transport.rdma.device-name"); if (temp) options->device_name = gf_strdup (temp->data); return; } gf_rdma_ctx_t * __gf_rdma_ctx_create (void) { gf_rdma_ctx_t *rdma_ctx = NULL; int ret = -1; rdma_ctx = GF_CALLOC (1, sizeof (*rdma_ctx), gf_common_mt_char); if (rdma_ctx == NULL) { goto out; } rdma_ctx->rdma_cm_event_channel = rdma_create_event_channel (); if (rdma_ctx->rdma_cm_event_channel == NULL) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "rdma_cm event channel creation failed (%s)", strerror (errno)); goto out; } ret = gf_thread_create (&rdma_ctx->rdma_cm_thread, NULL, gf_rdma_cm_event_handler, rdma_ctx->rdma_cm_event_channel); if (ret != 0) { gf_log (GF_RDMA_LOG_NAME, GF_LOG_WARNING, "creation of thread to handle rdma-cm events " "failed (%s)", strerror (ret)); goto out; } out: if (ret < 0) { if (rdma_ctx->rdma_cm_event_channel != NULL) { rdma_destroy_event_channel (rdma_ctx->rdma_cm_event_channel); } GF_FREE (rdma_ctx); rdma_ctx = NULL; } return rdma_ctx; } static int32_t gf_rdma_init (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; int32_t ret = 0; glusterfs_ctx_t *ctx = NULL; gf_rdma_options_t *options = NULL; ctx= this->ctx; priv = this->private; ibv_fork_init (); gf_rdma_options_init (this); options = &priv->options; priv->peer.send_count = options->send_count; priv->peer.recv_count = options->recv_count; priv->peer.send_size = options->send_size; priv->peer.recv_size = options->recv_size; priv->peer.trans = this; INIT_LIST_HEAD (&priv->peer.ioq); pthread_mutex_init (&priv->write_mutex, NULL); pthread_mutex_init (&priv->recv_mutex, NULL); pthread_cond_init (&priv->recv_cond, NULL); pthread_mutex_lock (&ctx->lock); { if (ctx->ib == NULL) { ctx->ib = __gf_rdma_ctx_create (); if (ctx->ib == NULL) { ret = -1; } } } pthread_mutex_unlock (&ctx->lock); return ret; } static int32_t gf_rdma_disconnect (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; int32_t ret = 0; priv = this->private; gf_log_callingfn (this->name, GF_LOG_WARNING, "disconnect called (peer:%s)", this->peerinfo.identifier); pthread_mutex_lock (&priv->write_mutex); { ret = __gf_rdma_disconnect (this); } pthread_mutex_unlock (&priv->write_mutex); return ret; } static int32_t gf_rdma_connect (struct rpc_transport *this, int port) { gf_rdma_private_t *priv = NULL; int32_t ret = 0; union gf_sock_union sock_union = {{0, }, }; socklen_t sockaddr_len = 0; gf_rdma_peer_t *peer = NULL; gf_rdma_ctx_t *rdma_ctx = NULL; gf_boolean_t connected = _gf_false; priv = this->private; peer = &priv->peer; rpc_transport_ref (this); ret = gf_rdma_client_get_remote_sockaddr (this, &sock_union.sa, &sockaddr_len, port); if (ret != 0) { gf_log (this->name, GF_LOG_DEBUG, "cannot get remote address to connect"); goto out; } rdma_ctx = this->ctx->ib; pthread_mutex_lock (&priv->write_mutex); { if (peer->cm_id != NULL) { ret = -1; errno = EINPROGRESS; connected = _gf_true; goto unlock; } priv->entity = GF_RDMA_CLIENT; ret = rdma_create_id (rdma_ctx->rdma_cm_event_channel, &peer->cm_id, this, RDMA_PS_TCP); if (ret != 0) { gf_log (this->name, GF_LOG_ERROR, "creation of rdma_cm_id failed (%s)", strerror (errno)); ret = -errno; goto unlock; } memcpy (&this->peerinfo.sockaddr, &sock_union.storage, sockaddr_len); this->peerinfo.sockaddr_len = sockaddr_len; if (port > 0) sock_union.sin.sin_port = htons (port); ((struct sockaddr *) &this->myinfo.sockaddr)->sa_family = ((struct sockaddr *)&this->peerinfo.sockaddr)->sa_family; ret = gf_rdma_client_bind (this, (struct sockaddr *)&this->myinfo.sockaddr, &this->myinfo.sockaddr_len, peer->cm_id); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "client bind failed: %s", strerror (errno)); goto unlock; } ret = rdma_resolve_addr (peer->cm_id, NULL, &sock_union.sa, 2000); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "rdma_resolve_addr failed (%s)", strerror (errno)); goto unlock; } priv->connected = 0; } unlock: pthread_mutex_unlock (&priv->write_mutex); out: if (ret != 0) { if (!connected) { gf_rdma_teardown (this); } rpc_transport_unref (this); } return ret; } static int32_t gf_rdma_listen (rpc_transport_t *this) { union gf_sock_union sock_union = {{0, }, }; socklen_t sockaddr_len = 0; gf_rdma_private_t *priv = NULL; gf_rdma_peer_t *peer = NULL; int ret = 0; gf_rdma_ctx_t *rdma_ctx = NULL; char service[NI_MAXSERV], host[NI_MAXHOST]; int optval = 2; priv = this->private; peer = &priv->peer; priv->entity = GF_RDMA_SERVER_LISTENER; rdma_ctx = this->ctx->ib; ret = gf_rdma_server_get_local_sockaddr (this, &sock_union.sa, &sockaddr_len); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "cannot find network address of server to bind to"); goto err; } ret = rdma_create_id (rdma_ctx->rdma_cm_event_channel, &peer->cm_id, this, RDMA_PS_TCP); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "creation of rdma_cm_id failed (%s)", strerror (errno)); goto err; } memcpy (&this->myinfo.sockaddr, &sock_union.storage, sockaddr_len); this->myinfo.sockaddr_len = sockaddr_len; ret = getnameinfo ((struct sockaddr *)&this->myinfo.sockaddr, this->myinfo.sockaddr_len, host, sizeof (host), service, sizeof (service), NI_NUMERICHOST); if (ret != 0) { gf_log (this->name, GF_LOG_ERROR, "getnameinfo failed (%s)", gai_strerror (ret)); goto err; } sprintf (this->myinfo.identifier, "%s:%s", host, service); ret = rdma_set_option(peer->cm_id, RDMA_OPTION_ID, RDMA_OPTION_ID_REUSEADDR, (void *)&optval, sizeof(optval)); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "rdma option set failed (%s)", strerror (errno)); goto err; } ret = rdma_bind_addr (peer->cm_id, &sock_union.sa); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "rdma_bind_addr failed (%s)", strerror (errno)); goto err; } ret = rdma_listen (peer->cm_id, 10); if (ret != 0) { gf_log (this->name, GF_LOG_WARNING, "rdma_listen failed (%s)", strerror (errno)); goto err; } rpc_transport_ref (this); ret = 0; err: if (ret < 0) { if (peer->cm_id != NULL) { rdma_destroy_id (peer->cm_id); peer->cm_id = NULL; } } return ret; } struct rpc_transport_ops tops = { .submit_request = gf_rdma_submit_request, .submit_reply = gf_rdma_submit_reply, .connect = gf_rdma_connect, .disconnect = gf_rdma_disconnect, .listen = gf_rdma_listen, }; int32_t init (rpc_transport_t *this) { gf_rdma_private_t *priv = NULL; priv = GF_CALLOC (1, sizeof (*priv), gf_common_mt_rdma_private_t); if (!priv) return -1; this->private = priv; if (gf_rdma_init (this)) { gf_log (this->name, GF_LOG_ERROR, "Failed to initialize IB Device"); return -1; } return 0; } void fini (struct rpc_transport *this) { /* TODO: verify this function does graceful finish */ gf_rdma_private_t *priv = NULL; priv = this->private; this->private = NULL; if (priv) { pthread_mutex_destroy (&priv->recv_mutex); pthread_mutex_destroy (&priv->write_mutex); gf_log (this->name, GF_LOG_TRACE, "called fini on transport: %p", this); GF_FREE (priv); } return; } /* TODO: expand each option */ struct volume_options options[] = { { .key = {"transport.rdma.port", "rdma-port"}, .type = GF_OPTION_TYPE_INT, .min = 1, .max = 4, .description = "check the option by 'ibv_devinfo'" }, { .key = {"transport.rdma.mtu", "rdma-mtu"}, .type = GF_OPTION_TYPE_INT, }, { .key = {"transport.rdma.device-name", "rdma-device-name"}, .type = GF_OPTION_TYPE_ANY, .description = "check by 'ibv_devinfo'" }, { .key = {"transport.rdma.work-request-send-count", "rdma-work-request-send-count"}, .type = GF_OPTION_TYPE_INT, }, { .key = {"transport.rdma.work-request-recv-count", "rdma-work-request-recv-count"}, .type = GF_OPTION_TYPE_INT, }, { .key = {"remote-port", "transport.remote-port", "transport.rdma.remote-port"}, .type = GF_OPTION_TYPE_INT }, { .key = {"transport.rdma.attr-timeout", "rdma-attr-timeout"}, .type = GF_OPTION_TYPE_INT }, { .key = {"transport.rdma.attr-retry-cnt", "rdma-attr-retry-cnt"}, .type = GF_OPTION_TYPE_INT }, { .key = {"transport.rdma.attr-rnr-retry", "rdma-attr-rnr-retry"}, .type = GF_OPTION_TYPE_INT }, { .key = {"transport.rdma.listen-port", "listen-port"}, .type = GF_OPTION_TYPE_INT }, { .key = {"transport.rdma.connect-path", "connect-path"}, .type = GF_OPTION_TYPE_ANY }, { .key = {"transport.rdma.bind-path", "bind-path"}, .type = GF_OPTION_TYPE_ANY }, { .key = {"transport.rdma.listen-path", "listen-path"}, .type = GF_OPTION_TYPE_ANY }, { .key = {"transport.address-family", "address-family"}, .value = {"inet", "inet6", "inet/inet6", "inet6/inet", "unix", "inet-sdp" }, .type = GF_OPTION_TYPE_STR }, { .key = {"transport.socket.lowlat"}, .type = GF_OPTION_TYPE_BOOL }, { .key = {NULL} } };