From e589d8de66d3325da8fbbbe44d1a5bd6335e08ab Mon Sep 17 00:00:00 2001 From: Mohit Agrawal Date: Wed, 4 Mar 2020 09:17:26 +0530 Subject: core[brick_mux]: brick crashed when creating and deleting volumes over time Problem: In brick_mux environment, while volumes are created/stopped in a loop after running a long time the main brick is crashed.The brick is crashed because the main brick process was not cleaned up memory for all objects at the time of detaching a volume. Below are the objects that are missed at the time of detaching a volume 1) xlator object for a brick graph 2) local_pool for posix_lock xlator 3) rpc object cleanup at quota xlator 4) inode leak at brick xlator Solution: To avoid the crash resolve all leak at the time of detaching a brick Change-Id: Ibb6e46c5fba22b9441a88cbaf6b3278823235913 updates: #977 Signed-off-by: Mohit Agrawal --- xlators/features/changelog/src/changelog.c | 1 + xlators/features/locks/src/posix.c | 4 ++ xlators/features/quota/src/quota-enforcer-client.c | 14 +++++- xlators/features/quota/src/quota.c | 54 ++++++++++++++++++++-- xlators/features/quota/src/quota.h | 3 ++ xlators/protocol/server/src/server.c | 12 +++-- 6 files changed, 79 insertions(+), 9 deletions(-) (limited to 'xlators') diff --git a/xlators/features/changelog/src/changelog.c b/xlators/features/changelog/src/changelog.c index 75d5ae267fd..37916f40882 100644 --- a/xlators/features/changelog/src/changelog.c +++ b/xlators/features/changelog/src/changelog.c @@ -2852,6 +2852,7 @@ fini(xlator_t *this) if (priv->active || priv->rpc_active) { /* terminate RPC server/threads */ changelog_cleanup_rpc(this, priv); + GF_FREE(priv->ev_dispatcher); } /* call barrier_disable to cancel timer */ if (priv->barrier_enabled) diff --git a/xlators/features/locks/src/posix.c b/xlators/features/locks/src/posix.c index af530aafd84..8b57627addf 100644 --- a/xlators/features/locks/src/posix.c +++ b/xlators/features/locks/src/posix.c @@ -4104,6 +4104,10 @@ fini(xlator_t *this) if (!priv) return; this->private = NULL; + if (this->local_pool) { + mem_pool_destroy(this->local_pool); + this->local_pool = NULL; + } GF_FREE(priv->brickname); GF_FREE(priv); diff --git a/xlators/features/quota/src/quota-enforcer-client.c b/xlators/features/quota/src/quota-enforcer-client.c index 1a4c2e30dd6..097439d86d6 100644 --- a/xlators/features/quota/src/quota-enforcer-client.c +++ b/xlators/features/quota/src/quota-enforcer-client.c @@ -362,16 +362,28 @@ quota_enforcer_notify(struct rpc_clnt *rpc, void *mydata, { xlator_t *this = NULL; int ret = 0; + quota_priv_t *priv = NULL; this = mydata; - + priv = this->private; switch (event) { case RPC_CLNT_CONNECT: { + pthread_mutex_lock(&priv->conn_mutex); + { + priv->conn_status = _gf_true; + } + pthread_mutex_unlock(&priv->conn_mutex); gf_msg_trace(this->name, 0, "got RPC_CLNT_CONNECT"); break; } case RPC_CLNT_DISCONNECT: { + pthread_mutex_lock(&priv->conn_mutex); + { + priv->conn_status = _gf_false; + pthread_cond_signal(&priv->conn_cond); + } + pthread_mutex_unlock(&priv->conn_mutex); gf_msg_trace(this->name, 0, "got RPC_CLNT_DISCONNECT"); break; } diff --git a/xlators/features/quota/src/quota.c b/xlators/features/quota/src/quota.c index 01b01b5f01b..73c008a2c00 100644 --- a/xlators/features/quota/src/quota.c +++ b/xlators/features/quota/src/quota.c @@ -5008,6 +5008,43 @@ quota_forget(xlator_t *this, inode_t *inode) return 0; } +int +notify(xlator_t *this, int event, void *data, ...) +{ + quota_priv_t *priv = NULL; + int ret = 0; + rpc_clnt_t *rpc = NULL; + gf_boolean_t conn_status = _gf_true; + xlator_t *victim = data; + + priv = this->private; + if (!priv || !priv->is_quota_on) + goto out; + + if (event == GF_EVENT_PARENT_DOWN) { + rpc = priv->rpc_clnt; + if (rpc) { + rpc_clnt_disable(rpc); + pthread_mutex_lock(&priv->conn_mutex); + { + conn_status = priv->conn_status; + while (conn_status) { + (void)pthread_cond_wait(&priv->conn_cond, + &priv->conn_mutex); + conn_status = priv->conn_status; + } + } + pthread_mutex_unlock(&priv->conn_mutex); + gf_log(this->name, GF_LOG_INFO, + "Notify GF_EVENT_PARENT_DOWN for brick %s", victim->name); + } + } + +out: + ret = default_notify(this, event, data); + return ret; +} + int32_t init(xlator_t *this) { @@ -5050,6 +5087,10 @@ init(xlator_t *this) goto err; } + pthread_mutex_init(&priv->conn_mutex, NULL); + pthread_cond_init(&priv->conn_cond, NULL); + priv->conn_status = _gf_false; + if (priv->is_quota_on) { rpc = quota_enforcer_init(this, this->options); if (rpc == NULL) { @@ -5163,20 +5204,22 @@ fini(xlator_t *this) { quota_priv_t *priv = NULL; rpc_clnt_t *rpc = NULL; - int i = 0, cnt = 0; priv = this->private; if (!priv) return; rpc = priv->rpc_clnt; priv->rpc_clnt = NULL; - this->private = NULL; if (rpc) { - cnt = GF_ATOMIC_GET(rpc->refcount); - for (i = 0; i < cnt; i++) - rpc_clnt_unref(rpc); + rpc_clnt_connection_cleanup(&rpc->conn); + rpc_clnt_unref(rpc); } + + this->private = NULL; LOCK_DESTROY(&priv->lock); + pthread_mutex_destroy(&priv->conn_mutex); + pthread_cond_destroy(&priv->conn_cond); + GF_FREE(priv); if (this->local_pool) { mem_pool_destroy(this->local_pool); @@ -5308,6 +5351,7 @@ struct volume_options options[] = { xlator_api_t xlator_api = { .init = init, .fini = fini, + .notify = notify, .reconfigure = reconfigure, .mem_acct_init = mem_acct_init, .op_version = {1}, /* Present from the initial version */ diff --git a/xlators/features/quota/src/quota.h b/xlators/features/quota/src/quota.h index 00012c22f46..8a3dc7a77f5 100644 --- a/xlators/features/quota/src/quota.h +++ b/xlators/features/quota/src/quota.h @@ -214,6 +214,9 @@ struct quota_priv { char *volume_uuid; uint64_t validation_count; int32_t quotad_conn_status; + pthread_mutex_t conn_mutex; + pthread_cond_t conn_cond; + gf_boolean_t conn_status; }; typedef struct quota_priv quota_priv_t; diff --git a/xlators/protocol/server/src/server.c b/xlators/protocol/server/src/server.c index d5775c6ff4b..5b76b49cc82 100644 --- a/xlators/protocol/server/src/server.c +++ b/xlators/protocol/server/src/server.c @@ -403,7 +403,13 @@ server_call_xlator_mem_cleanup(xlator_t *this, char *victim_name) arg = calloc(1, sizeof(*arg)); arg->this = this; - arg->victim_name = gf_strdup(victim_name); + arg->victim_name = strdup(victim_name); + if (!arg->victim_name) { + gf_smsg(this->name, GF_LOG_CRITICAL, ENOMEM, LG_MSG_NO_MEMORY, + "Memory allocation is failed"); + return; + } + th_ret = gf_thread_create_detached(&th_id, server_graph_janitor_threads, arg, "graphjanitor"); if (th_ret) { @@ -411,7 +417,7 @@ server_call_xlator_mem_cleanup(xlator_t *this, char *victim_name) "graph janitor Thread" " creation is failed for brick %s", victim_name); - GF_FREE(arg->victim_name); + free(arg->victim_name); free(arg); } } @@ -621,7 +627,7 @@ server_graph_janitor_threads(void *data) } out: - GF_FREE(arg->victim_name); + free(arg->victim_name); free(arg); return NULL; } -- cgit