path: root/sql/wsrep_mysqld.cc

/* Copyright 2008-2015 Codership Oy <http://www.codership.com>

   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; version 2 of the License.x1

   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software
   Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1335 USA */

#include "sql_plugin.h"                         /* wsrep_plugins_pre_init() */
#include "my_global.h"
#include "wsrep_server_state.h"

#include "mariadb.h"
#include <mysqld.h>
#include <transaction.h>
#include <sql_class.h>
#include <sql_parse.h>
#include <sql_base.h> /* find_temporary_table() */
#include "slave.h"
#include "rpl_mi.h"
#include "sql_repl.h"
#include "rpl_filter.h"
#include "sql_callback.h"
#include "sp_head.h"
#include "sql_show.h"
#include "sp.h"
#include "wsrep_priv.h"
#include "wsrep_thd.h"
#include "wsrep_sst.h"
#include "wsrep_utils.h"
#include "wsrep_var.h"
#include "wsrep_binlog.h"
#include "wsrep_applier.h"
#include "wsrep_schema.h"
#include "wsrep_xid.h"
#include "wsrep_trans_observer.h"
#include "mysql/service_wsrep.h"
#include <cstdio>
#include <cstdlib>
#include <string>
#include "log_event.h"

#include <sstream>

/* wsrep-lib */
Wsrep_server_state* Wsrep_server_state::m_instance;

my_bool wsrep_emulate_bin_log  = FALSE; // activating parts of binlog interface
#ifdef GTID_SUPPORT
/* Sidno in global_sid_map corresponding to group uuid */
rpl_sidno wsrep_sidno= -1;
#endif /* GTID_SUPPORT */
my_bool wsrep_preordered_opt= FALSE;

/* Streaming Replication */
const char *wsrep_fragment_units[]= { "bytes", "rows", "statements", NullS };
const char *wsrep_SR_store_types[]= { "none", "table", NullS };

/*
 * Begin configuration options
 */

extern my_bool plugins_are_initialized;
extern uint kill_cached_threads;
extern mysql_cond_t COND_thread_cache;

/* System variables. */
const char *wsrep_provider;
const char *wsrep_provider_options;
const char *wsrep_cluster_address;
const char *wsrep_cluster_name;
const char *wsrep_node_name;
const char *wsrep_node_address;
const char *wsrep_node_incoming_address;
const char *wsrep_start_position;
const char *wsrep_data_home_dir;
const char *wsrep_dbug_option;
const char *wsrep_notify_cmd;

ulong   wsrep_debug;                            // Debug level logging
my_bool wsrep_convert_LOCK_to_trx;              // Convert locking sessions to trx
my_bool wsrep_auto_increment_control;           // Control auto increment variables
my_bool wsrep_drupal_282555_workaround;         // Retry autoinc insert after dupkey
my_bool wsrep_certify_nonPK;                    // Certify, even when no primary key
ulong   wsrep_certification_rules      = WSREP_CERTIFICATION_RULES_STRICT;
my_bool wsrep_recovery;                         // Recovery
my_bool wsrep_replicate_myisam;                 // Enable MyISAM replication
my_bool wsrep_log_conflicts;
my_bool wsrep_load_data_splitting= 0;           // Commit load data every 10K intervals
my_bool wsrep_slave_UK_checks;                  // Slave thread does UK checks
my_bool wsrep_slave_FK_checks;                  // Slave thread does FK checks
my_bool wsrep_restart_slave;                    // Should mysql slave thread be
                                                // restarted, when node joins back?
my_bool wsrep_desync;                           // De(re)synchronize the node from the
                                                // cluster
long wsrep_slave_threads;                       // No. of slave appliers threads
ulong wsrep_retry_autocommit;                   // Retry aborted autocommit trx
ulong wsrep_max_ws_size;                        // Max allowed ws (RBR buffer) size
ulong wsrep_max_ws_rows;                        // Max number of rows in ws
ulong wsrep_forced_binlog_format;
ulong wsrep_mysql_replication_bundle;
bool wsrep_gtid_mode;                           // Use wsrep_gtid_domain_id
                                                // for galera transactions?
uint32 wsrep_gtid_domain_id;                    // gtid_domain_id for galera
                                                // transactions

/* Other configuration variables and their default values. */
my_bool wsrep_incremental_data_collection= 0;   // Incremental data collection
my_bool wsrep_restart_slave_activated= 0;       // Node has dropped, and slave
                                                // restart will be needed
bool wsrep_new_cluster= false;                  // Bootstrap the cluster?
int wsrep_slave_count_change= 0;                // No. of appliers to stop/start
int wsrep_to_isolation= 0;                      // No. of active TO isolation threads
long wsrep_max_protocol_version= 4;             // Maximum protocol version to use
long int  wsrep_protocol_version= wsrep_max_protocol_version;
ulong wsrep_trx_fragment_unit= WSREP_FRAG_BYTES;
                                                // unit for fragment size
ulong wsrep_SR_store_type= WSREP_SR_STORE_TABLE;
uint  wsrep_ignore_apply_errors= 0;


/*
 * End configuration options
 */

/*
 * Other wsrep global variables.
 */

mysql_mutex_t LOCK_wsrep_ready;
mysql_cond_t  COND_wsrep_ready;
mysql_mutex_t LOCK_wsrep_sst;
mysql_cond_t  COND_wsrep_sst;
mysql_mutex_t LOCK_wsrep_sst_init;
mysql_cond_t  COND_wsrep_sst_init;
mysql_mutex_t LOCK_wsrep_replaying;
mysql_cond_t  COND_wsrep_replaying;
mysql_mutex_t LOCK_wsrep_slave_threads;
mysql_cond_t  COND_wsrep_slave_threads;
mysql_mutex_t LOCK_wsrep_cluster_config;
mysql_mutex_t LOCK_wsrep_desync;
mysql_mutex_t LOCK_wsrep_config_state;
mysql_mutex_t LOCK_wsrep_group_commit;
mysql_mutex_t LOCK_wsrep_SR_pool;
mysql_mutex_t LOCK_wsrep_SR_store;

int wsrep_replaying= 0;
ulong  wsrep_running_threads = 0; // # of currently running wsrep
				  // # threads
ulong  wsrep_running_applier_threads = 0; // # of running applier threads
ulong  wsrep_running_rollbacker_threads = 0; // # of running
					     // # rollbacker threads
ulong  my_bind_addr;

#ifdef HAVE_PSI_INTERFACE
PSI_mutex_key 
  key_LOCK_wsrep_replaying, key_LOCK_wsrep_ready, key_LOCK_wsrep_sst,
  key_LOCK_wsrep_sst_thread, key_LOCK_wsrep_sst_init,
  key_LOCK_wsrep_slave_threads, key_LOCK_wsrep_desync,
  key_LOCK_wsrep_config_state, key_LOCK_wsrep_cluster_config,
  key_LOCK_wsrep_group_commit,
  key_LOCK_wsrep_SR_pool,
  key_LOCK_wsrep_SR_store,
  key_LOCK_wsrep_thd_queue;

PSI_cond_key key_COND_wsrep_thd,
  key_COND_wsrep_replaying, key_COND_wsrep_ready, key_COND_wsrep_sst,
  key_COND_wsrep_sst_init, key_COND_wsrep_sst_thread,
  key_COND_wsrep_thd_queue, key_COND_wsrep_slave_threads;
  

PSI_file_key key_file_wsrep_gra_log;

static PSI_mutex_info wsrep_mutexes[]=
{
  { &key_LOCK_wsrep_ready, "LOCK_wsrep_ready", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_sst, "LOCK_wsrep_sst", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_sst_thread, "wsrep_sst_thread", 0},
  { &key_LOCK_wsrep_sst_init, "LOCK_wsrep_sst_init", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_sst, "LOCK_wsrep_sst", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_replaying, "LOCK_wsrep_replaying", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_slave_threads, "LOCK_wsrep_slave_threads", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_cluster_config, "LOCK_wsrep_cluster_config", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_desync, "LOCK_wsrep_desync", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_config_state, "LOCK_wsrep_config_state", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_group_commit, "LOCK_wsrep_group_commit", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_SR_pool, "LOCK_wsrep_SR_pool", PSI_FLAG_GLOBAL},
  { &key_LOCK_wsrep_SR_store, "LOCK_wsrep_SR_store", PSI_FLAG_GLOBAL}
};

static PSI_cond_info wsrep_conds[]=
{
  { &key_COND_wsrep_ready, "COND_wsrep_ready", PSI_FLAG_GLOBAL},
  { &key_COND_wsrep_sst, "COND_wsrep_sst", PSI_FLAG_GLOBAL},
  { &key_COND_wsrep_sst_init, "COND_wsrep_sst_init", PSI_FLAG_GLOBAL},
  { &key_COND_wsrep_sst_thread, "wsrep_sst_thread", 0},
  { &key_COND_wsrep_thd, "THD::COND_wsrep_thd", 0},
  { &key_COND_wsrep_replaying, "COND_wsrep_replaying", PSI_FLAG_GLOBAL},
  { &key_COND_wsrep_slave_threads, "COND_wsrep_wsrep_slave_threads", PSI_FLAG_GLOBAL}
};

static PSI_file_info wsrep_files[]=
{
  { &key_file_wsrep_gra_log, "wsrep_gra_log", 0}
};

PSI_thread_key key_wsrep_sst_joiner, key_wsrep_sst_donor,
  key_wsrep_rollbacker, key_wsrep_applier;

static PSI_thread_info wsrep_threads[]=
{
 {&key_wsrep_sst_joiner, "wsrep_sst_joiner_thread", PSI_FLAG_GLOBAL},
 {&key_wsrep_sst_donor, "wsrep_sst_donor_thread", PSI_FLAG_GLOBAL},
 {&key_wsrep_rollbacker, "wsrep_rollbacker_thread", PSI_FLAG_GLOBAL},
 {&key_wsrep_applier, "wsrep_applier_thread", PSI_FLAG_GLOBAL}
};

#endif /* HAVE_PSI_INTERFACE */

my_bool wsrep_inited= 0; // initialized ?

static wsrep_uuid_t node_uuid= WSREP_UUID_UNDEFINED;
static char         cluster_uuid_str[40]= { 0, };

static char provider_name[256]= { 0, };
static char provider_version[256]= { 0, };
static char provider_vendor[256]= { 0, };

/*
 * Wsrep status variables. LOCK_status must be locked When modifying
 * these variables,
 */
my_bool     wsrep_connected         = FALSE;
my_bool     wsrep_ready             = FALSE;
const char* wsrep_cluster_state_uuid= cluster_uuid_str;
long long   wsrep_cluster_conf_id   = WSREP_SEQNO_UNDEFINED;
const char* wsrep_cluster_status    = "Disconnected";
long        wsrep_cluster_size      = 0;
long        wsrep_local_index       = -1;
long long   wsrep_local_bf_aborts   = 0;
const char* wsrep_provider_name     = provider_name;
const char* wsrep_provider_version  = provider_version;
const char* wsrep_provider_vendor   = provider_vendor;
char* wsrep_provider_capabilities   = NULL;
char* wsrep_cluster_capabilities    = NULL;
/* End wsrep status variables */

wsp::Config_state *wsrep_config_state;


wsrep_uuid_t               local_uuid       = WSREP_UUID_UNDEFINED;
wsrep_seqno_t              local_seqno      = WSREP_SEQNO_UNDEFINED;
wsp::node_status           local_status;

/*
 */
Wsrep_schema *wsrep_schema= 0;

static void wsrep_log_cb(wsrep::log::level level, const char *msg)
{
  /*
    Silence all wsrep related logging from lib and provider if
    wsrep is not enabled.
  */
  if (WSREP_ON)
  {
    switch (level) {
    case wsrep::log::info:
      sql_print_information("WSREP: %s", msg);
      break;
    case wsrep::log::warning:
      sql_print_warning("WSREP: %s", msg);
      break;
    case wsrep::log::error:
      sql_print_error("WSREP: %s", msg);
    break;
    case wsrep::log::debug:
      if (wsrep_debug) sql_print_information ("[Debug] WSREP: %s", msg);
    default:
      break;
    }
  }
}

void wsrep_init_sidno(const wsrep::id& uuid)
{
  /*
    Protocol versions starting from 4 use group gtid as it is.
    For lesser protocol versions generate new Sid map entry from inverted
    uuid.
  */
  rpl_gtid sid;
  if (wsrep_protocol_version >= 4)
  {
    memcpy((void*)&sid, (const uchar*)uuid.data(),16);
  }
  else
  {
    wsrep_uuid_t ltid_uuid;
    for (size_t i= 0; i < sizeof(ltid_uuid.data); ++i)
    {
      ltid_uuid.data[i]= ~((const uchar*)uuid.data())[i];
    }
    memcpy((void*)&sid, (const uchar*)ltid_uuid.data,16);
  }
#ifdef GTID_SUPPORT
  global_sid_lock->wrlock();
  wsrep_sidno= global_sid_map->add_sid(sid);
  WSREP_INFO("Initialized wsrep sidno %d", wsrep_sidno);
  global_sid_lock->unlock();
#endif
}

void wsrep_init_schema()
{
  DBUG_ASSERT(!wsrep_schema);

  WSREP_INFO("wsrep_init_schema_and_SR %p", wsrep_schema);
  if (!wsrep_schema)
  {
    wsrep_schema= new Wsrep_schema();
    if (wsrep_schema->init())
    {
      WSREP_ERROR("Failed to init wsrep schema");
      unireg_abort(1);
    }
  }
}

void wsrep_deinit_schema()
{
  delete wsrep_schema;
  wsrep_schema= 0;
}

void wsrep_recover_sr_from_storage(THD *orig_thd)
{
  switch (wsrep_SR_store_type)
  {
  case WSREP_SR_STORE_TABLE:
    if (!wsrep_schema)
    {
      WSREP_ERROR("Wsrep schema not initialized when trying to recover "
                  "streaming transactions");
      unireg_abort(1);
    }
    if (wsrep_schema->recover_sr_transactions(orig_thd))
    {
      WSREP_ERROR("Failed to recover SR transactions from schema");
      unireg_abort(1);
    }
    break;
  default:
    /* */
    WSREP_ERROR("Unsupported wsrep SR store type: %lu", wsrep_SR_store_type);
    unireg_abort(1);
    break;
  }
}

/** Export the WSREP provider's capabilities as a human readable string.
 * The result is saved in a dynamically allocated string of the form:
 * :cap1:cap2:cap3:
 */
static void wsrep_capabilities_export(wsrep_cap_t const cap, char** str)
{
  static const char* names[] =
  {
    /* Keep in sync with wsrep/wsrep_api.h WSREP_CAP_* macros. */
    "MULTI_MASTER",
    "CERTIFICATION",
    "PARALLEL_APPLYING",
    "TRX_REPLAY",
    "ISOLATION",
    "PAUSE",
    "CAUSAL_READS",
    "CAUSAL_TRX",
    "INCREMENTAL_WRITESET",
    "SESSION_LOCKS",
    "DISTRIBUTED_LOCKS",
    "CONSISTENCY_CHECK",
    "UNORDERED",
    "ANNOTATION",
    "PREORDERED",
    "STREAMING",
    "SNAPSHOT",
    "NBO",
  };

  std::string s;
  for (size_t i= 0; i < sizeof(names) / sizeof(names[0]); ++i)
  {
    if (cap & (1ULL << i))
    {
      if (s.empty())
      {
        s= ":";
      }
      s += names[i];
      s += ":";
    }
  }

  /* A read from the string pointed to by *str may be started at any time,
   * so it must never point to free(3)d memory or non '\0' terminated string. */

  char* const previous= *str;

  *str= strdup(s.c_str());

  if (previous != NULL)
  {
    free(previous);
  }
}

/* Verifies that SE position is consistent with the group position
 * and initializes other variables */
void wsrep_verify_SE_checkpoint(const wsrep_uuid_t& uuid,
                                wsrep_seqno_t const seqno)
{
}

/*
  Wsrep is considered ready if
  1) Provider is not loaded (native mode)
  2) Server has reached synced state
  3) Server is in joiner mode and mysqldump SST method has been
     specified
  See Wsrep_server_service::log_state_change() for further details.
 */
my_bool wsrep_ready_get (void)
{
  if (mysql_mutex_lock (&LOCK_wsrep_ready)) abort();
  my_bool ret= wsrep_ready;
  mysql_mutex_unlock (&LOCK_wsrep_ready);
  return ret;
}

int wsrep_show_ready(THD *thd, SHOW_VAR *var, char *buff)
{
  var->type= SHOW_MY_BOOL;
  var->value= buff;
  *((my_bool *)buff)= wsrep_ready_get();
  return 0;
}

void wsrep_update_cluster_state_uuid(const char* uuid)
{
  strncpy(cluster_uuid_str, uuid, sizeof(cluster_uuid_str) - 1);
}

static void wsrep_init_position()
{
}

/****************************************************************************
                         Helpers for wsrep_init()
 ****************************************************************************/
static std::string wsrep_server_name()
{
  std::string ret(wsrep_node_name ? wsrep_node_name : "");
  return ret;
}

static std::string wsrep_server_id()
{
  /* using empty server_id, which enables view change handler to
     set final server_id later on
  */
  std::string ret("");
  return ret;
}

static std::string wsrep_server_node_address()
{

  std::string ret;
  if (!wsrep_data_home_dir || strlen(wsrep_data_home_dir) == 0)
    wsrep_data_home_dir= mysql_real_data_home;

  /* Initialize node address */
  if (!wsrep_node_address || !strcmp(wsrep_node_address, ""))
  {
    char node_addr[512]= {0, };
    const size_t node_addr_max= sizeof(node_addr) - 1;
    size_t guess_ip_ret= wsrep_guess_ip(node_addr, node_addr_max);
    if (!(guess_ip_ret > 0 && guess_ip_ret < node_addr_max))
    {
      WSREP_WARN("Failed to guess base node address. Set it explicitly via "
                 "wsrep_node_address.");
    }
    else
    {
      ret= node_addr;
    }
  }
  else
  {
    ret= wsrep_node_address;
  }
  return ret;
}

static std::string wsrep_server_incoming_address()
{
  std::string ret;
  const std::string node_addr(wsrep_server_node_address());
  char inc_addr[512]= { 0, };
  size_t const inc_addr_max= sizeof (inc_addr);

  /*
    In case wsrep_node_incoming_address is either not set or set to AUTO,
    we need to use mysqld's my_bind_addr_str:mysqld_port, lastly fallback
    to wsrep_node_address' value if mysqld's bind-address is not set either.
  */
  if ((!wsrep_node_incoming_address ||
       !strcmp (wsrep_node_incoming_address, WSREP_NODE_INCOMING_AUTO)))
  {
    bool is_ipv6= false;
    unsigned int my_bind_ip= INADDR_ANY; // default if not set

    if (my_bind_addr_str && strlen(my_bind_addr_str) && 
        strcmp(my_bind_addr_str, "*") != 0)
    {
      my_bind_ip= wsrep_check_ip(my_bind_addr_str, &is_ipv6);
    }

    if (INADDR_ANY != my_bind_ip)
    {
      /*
        If its a not a valid address, leave inc_addr as empty string. mysqld
        is not listening for client connections on network interfaces.
      */
      if (INADDR_NONE != my_bind_ip && INADDR_LOOPBACK != my_bind_ip)
      {
        const char *fmt= (is_ipv6) ? "[%s]:%u" : "%s:%u";
        snprintf(inc_addr, inc_addr_max, fmt, my_bind_addr_str, mysqld_port);
      }
    }
    else /* mysqld binds to 0.0.0.0, try taking IP from wsrep_node_address. */
    {
      if (node_addr.size())
      {
        size_t const ip_len_mdb= wsrep_host_len(node_addr.c_str(), node_addr.size());
        if (ip_len_mdb + 7 /* :55555\0 */ < inc_addr_max)
        {
          memcpy (inc_addr, node_addr.c_str(), ip_len_mdb);
          snprintf(inc_addr + ip_len_mdb, inc_addr_max - ip_len_mdb, ":%u",
                   (int)mysqld_port);
        }
        else
        {
          WSREP_WARN("Guessing address for incoming client connections: "
                     "address too long.");
          inc_addr[0]= '\0';
        }
      }

      if (!strlen(inc_addr))
      {
        WSREP_WARN("Guessing address for incoming client connections failed. "
                   "Try setting wsrep_node_incoming_address explicitly.");
        WSREP_INFO("Node addr: %s", node_addr.c_str());
      }
    }
  }
  else
  {
    wsp::Address addr(wsrep_node_incoming_address);

    if (!addr.is_valid())
    {
      WSREP_WARN("Could not parse wsrep_node_incoming_address : %s",
                 wsrep_node_incoming_address);
      goto done;
    }

    /*
      In case port is not specified in wsrep_node_incoming_address, we use
      mysqld_port.
    */
    int port= (addr.get_port() > 0) ? addr.get_port() : (int) mysqld_port;
    const char *fmt= (addr.is_ipv6()) ? "[%s]:%u" : "%s:%u";

    snprintf(inc_addr, inc_addr_max, fmt, addr.get_address(), port);
  }
  
 done:
  ret= wsrep_node_incoming_address;
  return ret;
}

static std::string wsrep_server_working_dir()
{
  std::string ret;
  if (!wsrep_data_home_dir || strlen(wsrep_data_home_dir) == 0)
  {
    ret= mysql_real_data_home;
  }
  else
  {
    ret= wsrep_data_home_dir;
  }
  return ret;
}

static wsrep::gtid wsrep_server_initial_position()
{
  wsrep::gtid ret;
  WSREP_DEBUG("Server initial position: %s", wsrep_start_position);
  std::istringstream is(wsrep_start_position);
  is >> ret;
  return ret;
}

/*
  Intitialize provider specific status variables
 */
static void wsrep_init_provider_status_variables()
{
  wsrep_inited= 1;
  const wsrep::provider& provider=
    Wsrep_server_state::instance().provider();
  strncpy(provider_name,
          provider.name().c_str(),    sizeof(provider_name) - 1);
  strncpy(provider_version,
          provider.version().c_str(), sizeof(provider_version) - 1);
  strncpy(provider_vendor,
          provider.vendor().c_str(),  sizeof(provider_vendor) - 1);
}

int wsrep_init_server()
{
  wsrep::log::logger_fn(wsrep_log_cb);
  try
  {
    std::string server_name;
    std::string server_id;
    std::string node_address;
    std::string incoming_address;
    std::string working_dir;
    wsrep::gtid initial_position;

    server_name= wsrep_server_name();
    server_id= wsrep_server_id();
    node_address= wsrep_server_node_address();
    incoming_address= wsrep_server_incoming_address();
    working_dir= wsrep_server_working_dir();
    initial_position= wsrep_server_initial_position();

    Wsrep_server_state::init_once(server_name,
                                  incoming_address,
                                  node_address,
                                  working_dir,
                                  initial_position,
                                  wsrep_max_protocol_version);
    Wsrep_server_state::instance().debug_log_level(wsrep_debug);
  }
  catch (const wsrep::runtime_error& e)
  {
    WSREP_ERROR("Failed to init wsrep server %s", e.what());
    return 1;
  }
  catch (const std::exception& e)
  {
    WSREP_ERROR("Failed to init wsrep server %s", e.what());
  }
  return 0;
}

void wsrep_init_globals()
{
  wsrep_init_sidno(Wsrep_server_state::instance().connected_gtid().id());
  wsrep_init_schema();
  if (WSREP_ON)
  {
    Wsrep_server_state::instance().initialized();
  }
}

void wsrep_deinit_server()
{
  wsrep_deinit_schema();
  Wsrep_server_state::destroy();
}

int wsrep_init()
{
  assert(wsrep_provider);

  wsrep_init_position();
  wsrep_sst_auth_init();

  if (strlen(wsrep_provider)== 0 ||
      !strcmp(wsrep_provider, WSREP_NONE))
  {
    // enable normal operation in case no provider is specified
    global_system_variables.wsrep_on= 0;
    int err= Wsrep_server_state::instance().load_provider(wsrep_provider, wsrep_provider_options ? wsrep_provider_options : "");
    if (err)
    {
      DBUG_PRINT("wsrep",("wsrep::init() failed: %d", err));
      WSREP_ERROR("wsrep::init() failed: %d, must shutdown", err);
    }
    else
      wsrep_init_provider_status_variables();
    return err;
  }

  global_system_variables.wsrep_on= 1;

  if (wsrep_gtid_mode && opt_bin_log && !opt_log_slave_updates)
  {
    WSREP_ERROR("Option --log-slave-updates is required if "
                "binlog is enabled, GTID mode is on and wsrep provider "
                "is specified");
    return 1;
  }

  if (!wsrep_data_home_dir || strlen(wsrep_data_home_dir) == 0)
    wsrep_data_home_dir= mysql_real_data_home;

  if (Wsrep_server_state::instance().load_provider(wsrep_provider,
                                                   wsrep_provider_options))
  {
    WSREP_ERROR("Failed to load provider");
    return 1;
  }

  if (!wsrep_provider_is_SR_capable() &&
      global_system_variables.wsrep_trx_fragment_size > 0)
  {
    WSREP_ERROR("The WSREP provider (%s) does not support streaming "
                "replication but wsrep_trx_fragment_size is set to a "
                "value other than 0 (%llu). Cannot continue. Either set "
                "wsrep_trx_fragment_size to 0 or use wsrep_provider that "
                "supports streaming replication.",
                wsrep_provider, global_system_variables.wsrep_trx_fragment_size);
    Wsrep_server_state::instance().unload_provider();
    return 1;
  }

  wsrep_init_provider_status_variables();
  wsrep_capabilities_export(Wsrep_server_state::instance().provider().capabilities(),
                            &wsrep_provider_capabilities);

  WSREP_DEBUG("SR storage init for: %s",
              (wsrep_SR_store_type == WSREP_SR_STORE_TABLE) ? "table" : "void");

  return 0;
}

/* Initialize wsrep thread LOCKs and CONDs */
void wsrep_thr_init()
{
  DBUG_ENTER("wsrep_thr_init");
  wsrep_config_state= new wsp::Config_state;
#ifdef HAVE_PSI_INTERFACE
  mysql_mutex_register("sql", wsrep_mutexes, array_elements(wsrep_mutexes));
  mysql_cond_register("sql", wsrep_conds, array_elements(wsrep_conds));
  mysql_file_register("sql", wsrep_files, array_elements(wsrep_files));
  mysql_thread_register("sql", wsrep_threads, array_elements(wsrep_threads));
#endif

  mysql_mutex_init(key_LOCK_wsrep_ready, &LOCK_wsrep_ready, MY_MUTEX_INIT_FAST);
  mysql_cond_init(key_COND_wsrep_ready, &COND_wsrep_ready, NULL);
  mysql_mutex_init(key_LOCK_wsrep_sst, &LOCK_wsrep_sst, MY_MUTEX_INIT_FAST);
  mysql_cond_init(key_COND_wsrep_sst, &COND_wsrep_sst, NULL);
  mysql_mutex_init(key_LOCK_wsrep_sst_init, &LOCK_wsrep_sst_init, MY_MUTEX_INIT_FAST);
  mysql_cond_init(key_COND_wsrep_sst_init, &COND_wsrep_sst_init, NULL);
  mysql_mutex_init(key_LOCK_wsrep_replaying, &LOCK_wsrep_replaying, MY_MUTEX_INIT_FAST);
  mysql_cond_init(key_COND_wsrep_replaying, &COND_wsrep_replaying, NULL);
  mysql_mutex_init(key_LOCK_wsrep_slave_threads, &LOCK_wsrep_slave_threads, MY_MUTEX_INIT_FAST);
  mysql_cond_init(key_COND_wsrep_slave_threads, &COND_wsrep_slave_threads, NULL);
  mysql_mutex_init(key_LOCK_wsrep_cluster_config, &LOCK_wsrep_cluster_config, MY_MUTEX_INIT_FAST);
  mysql_mutex_init(key_LOCK_wsrep_desync, &LOCK_wsrep_desync, MY_MUTEX_INIT_FAST);
  mysql_mutex_init(key_LOCK_wsrep_config_state, &LOCK_wsrep_config_state, MY_MUTEX_INIT_FAST);
  mysql_mutex_init(key_LOCK_wsrep_group_commit, &LOCK_wsrep_group_commit, MY_MUTEX_INIT_FAST);
  mysql_mutex_init(key_LOCK_wsrep_SR_pool,
                   &LOCK_wsrep_SR_pool, MY_MUTEX_INIT_FAST);
  mysql_mutex_init(key_LOCK_wsrep_SR_store,
                   &LOCK_wsrep_SR_store, MY_MUTEX_INIT_FAST);
  DBUG_VOID_RETURN;
}

void wsrep_init_startup (bool sst_first)
{
  if (wsrep_init()) unireg_abort(1);

  wsrep_thr_lock_init(wsrep_thd_is_BF, wsrep_thd_bf_abort,
                      wsrep_debug, wsrep_convert_LOCK_to_trx, wsrep_on);

  /*
    Pre-initialize global_system_variables.table_plugin with a dummy engine
    (placeholder) required during the initialization of wsrep threads (THDs).
    (see: plugin_thdvar_init())
    Note: This only needs to be done for rsync & mariabackup based SST methods.
    In case of mysqldump SST method, the wsrep threads are created after the
    server plugins & global system variables are initialized.
  */
  if (wsrep_before_SE())
    wsrep_plugins_pre_init();

  /* Skip replication start if dummy wsrep provider is loaded */
  if (!strcmp(wsrep_provider, WSREP_NONE)) return;

  /* Skip replication start if no cluster address */
  if (!wsrep_cluster_address || wsrep_cluster_address[0] == 0) return;

  /*
    Read value of wsrep_new_cluster before wsrep_start_replication(),
    the value is reset to FALSE inside wsrep_start_replication.
  */
  if (!wsrep_start_replication()) unireg_abort(1);

  wsrep_create_rollbacker();
  wsrep_create_appliers(1);

  Wsrep_server_state& server_state= Wsrep_server_state::instance();
  /*
    If the SST happens before server initialization, wait until the server
    state reaches initializing. This indicates that
    either SST was not necessary or SST has been delivered.

    With mysqldump SST (!sst_first) wait until the server reaches
    joiner state and procedd to accepting connections.
  */
  if (sst_first)
  {
    server_state.wait_until_state(Wsrep_server_state::s_initializing);
  }
  else
  {
    server_state.wait_until_state(Wsrep_server_state::s_joiner);
  }
}


void wsrep_deinit(bool free_options)
{
  DBUG_ASSERT(wsrep_inited == 1);
  WSREP_DEBUG("wsrep_deinit");

  Wsrep_server_state::instance().unload_provider();
  provider_name[0]=    '\0';
  provider_version[0]= '\0';
  provider_vendor[0]=  '\0';

  wsrep_inited= 0;

  if (wsrep_provider_capabilities != NULL)
  {
    char* p= wsrep_provider_capabilities;
    wsrep_provider_capabilities= NULL;
    free(p);
  }

  if (free_options)
  {
    wsrep_sst_auth_free();
  }
}

/* Destroy wsrep thread LOCKs and CONDs */
void wsrep_thr_deinit()
{
  if (!wsrep_config_state)
    return;                                     // Never initialized
  WSREP_DEBUG("wsrep_thr_deinit");
  mysql_mutex_destroy(&LOCK_wsrep_ready);
  mysql_cond_destroy(&COND_wsrep_ready);
  mysql_mutex_destroy(&LOCK_wsrep_sst);
  mysql_cond_destroy(&COND_wsrep_sst);
  mysql_mutex_destroy(&LOCK_wsrep_sst_init);
  mysql_cond_destroy(&COND_wsrep_sst_init);
  mysql_mutex_destroy(&LOCK_wsrep_replaying);
  mysql_cond_destroy(&COND_wsrep_replaying);
  mysql_mutex_destroy(&LOCK_wsrep_slave_threads);
  mysql_cond_destroy(&COND_wsrep_slave_threads);
  mysql_mutex_destroy(&LOCK_wsrep_cluster_config);
  mysql_mutex_destroy(&LOCK_wsrep_desync);
  mysql_mutex_destroy(&LOCK_wsrep_config_state);
  mysql_mutex_destroy(&LOCK_wsrep_group_commit);
  mysql_mutex_destroy(&LOCK_wsrep_SR_pool);
  mysql_mutex_destroy(&LOCK_wsrep_SR_store);

  delete wsrep_config_state;
  wsrep_config_state= 0;                        // Safety

  if (wsrep_cluster_capabilities != NULL)
  {
    char* p= wsrep_cluster_capabilities;
    wsrep_cluster_capabilities= NULL;
    free(p);
  }
}

void wsrep_recover()
{
  char uuid_str[40];

  if (wsrep_uuid_compare(&local_uuid, &WSREP_UUID_UNDEFINED) == 0 &&
      local_seqno == -2)
  {
    wsrep_uuid_print(&local_uuid, uuid_str, sizeof(uuid_str));
    WSREP_INFO("Position %s:%lld given at startup, skipping position recovery",
               uuid_str, (long long)local_seqno);
    return;
  }
  wsrep::gtid gtid= wsrep_get_SE_checkpoint();
  std::ostringstream oss;
  oss << gtid;
  WSREP_INFO("Recovered position: %s", oss.str().c_str());
}


void wsrep_stop_replication(THD *thd)
{
  WSREP_INFO("Stop replication by %llu", (thd) ? thd->thread_id : 0);
  if (Wsrep_server_state::instance().state() !=
      Wsrep_server_state::s_disconnected)
  {
    WSREP_DEBUG("Disconnect provider");
    Wsrep_server_state::instance().disconnect();
    Wsrep_server_state::instance().wait_until_state(Wsrep_server_state::s_disconnected);
  }

  /* my connection, should not terminate with wsrep_close_client_connection(),
     make transaction to rollback
  */
  if (thd && !thd->wsrep_applier) trans_rollback(thd);
  wsrep_close_client_connections(TRUE, thd);
 
  /* wait until appliers have stopped */
  wsrep_wait_appliers_close(thd);

  node_uuid= WSREP_UUID_UNDEFINED;
}

void wsrep_shutdown_replication()
{
  WSREP_INFO("Shutdown replication");
  if (Wsrep_server_state::instance().state() != wsrep::server_state::s_disconnected)
  {
    WSREP_DEBUG("Disconnect provider");
    Wsrep_server_state::instance().disconnect();
    Wsrep_server_state::instance().wait_until_state(Wsrep_server_state::s_disconnected);
  }

  wsrep_close_client_connections(TRUE);

  /* wait until appliers have stopped */
  wsrep_wait_appliers_close(NULL);
  node_uuid= WSREP_UUID_UNDEFINED;

  /* Undocking the thread specific data. */
  my_pthread_setspecific_ptr(THR_THD, NULL);
}

bool wsrep_start_replication()
{
  int rcode;
  WSREP_DEBUG("wsrep_start_replication");

  /*
    if provider is trivial, don't even try to connect,
    but resume local node operation
  */
  if (!WSREP_PROVIDER_EXISTS)
  {
    // enable normal operation in case no provider is specified
    return true;
  }

  if (!wsrep_cluster_address || wsrep_cluster_address[0]== 0)
  {
    // if provider is non-trivial, but no address is specified, wait for address
    WSREP_DEBUG("wsrep_start_replication exit due to empty address");
    return true;
  }

  bool const bootstrap(TRUE == wsrep_new_cluster);
  wsrep_new_cluster= FALSE;

  WSREP_INFO("Start replication");

  if ((rcode= Wsrep_server_state::instance().connect(
        wsrep_cluster_name,
        wsrep_cluster_address,
        wsrep_sst_donor,
        bootstrap)))
  {
    DBUG_PRINT("wsrep",("wsrep_ptr->connect(%s) failed: %d",
                        wsrep_cluster_address, rcode));
    WSREP_ERROR("wsrep::connect(%s) failed: %d",
                wsrep_cluster_address, rcode);
    return false;
  }
  else
  {
    try
    {
      std::string opts= Wsrep_server_state::instance().provider().options();
      wsrep_provider_options_init(opts.c_str());
    }
    catch (const wsrep::runtime_error&)
    {
      WSREP_WARN("Failed to get wsrep options");
    }
  }

  return true;
}

bool wsrep_must_sync_wait (THD* thd, uint mask)
{
  bool ret;
  mysql_mutex_lock(&thd->LOCK_thd_data);
  ret= (thd->variables.wsrep_sync_wait & mask) &&
    thd->wsrep_client_thread &&
    thd->variables.wsrep_on &&
    !(thd->variables.wsrep_dirty_reads &&
      !is_update_query(thd->lex->sql_command)) &&
    !thd->in_active_multi_stmt_transaction() &&
    thd->wsrep_trx().state() !=
    wsrep::transaction::s_replaying &&
    thd->wsrep_cs().sync_wait_gtid().is_undefined();
  mysql_mutex_unlock(&thd->LOCK_thd_data);
  return ret;
}

bool wsrep_sync_wait (THD* thd, uint mask)
{
  if (wsrep_must_sync_wait(thd, mask))
  {
    WSREP_DEBUG("wsrep_sync_wait: thd->variables.wsrep_sync_wait= %u, "
                "mask= %u, thd->variables.wsrep_on= %d",
                thd->variables.wsrep_sync_wait, mask,
                thd->variables.wsrep_on);
    /*
      This allows autocommit SELECTs and a first SELECT after SET AUTOCOMMIT=0
      TODO: modify to check if thd has locked any rows.
    */
    if (thd->wsrep_cs().sync_wait(-1))
    {
      const char* msg;
      int err;

      /*
        Possibly relevant error codes:
        ER_CHECKREAD, ER_ERROR_ON_READ, ER_INVALID_DEFAULT, ER_EMPTY_QUERY,
        ER_FUNCTION_NOT_DEFINED, ER_NOT_ALLOWED_COMMAND, ER_NOT_SUPPORTED_YET,
        ER_FEATURE_DISABLED, ER_QUERY_INTERRUPTED
      */

      switch (thd->wsrep_cs().current_error())
      {
      case wsrep::e_not_supported_error:
        msg= "synchronous reads by wsrep backend. "
          "Please unset wsrep_causal_reads variable.";
        err= ER_NOT_SUPPORTED_YET;
        break;
      default:
        msg= "Synchronous wait failed.";
        err= ER_LOCK_WAIT_TIMEOUT; // NOTE: the above msg won't be displayed
                                   //       with ER_LOCK_WAIT_TIMEOUT
      }

      my_error(err, MYF(0), msg);

      return true;
    }
  }

  return false;
}

enum wsrep::provider::status
wsrep_sync_wait_upto (THD*          thd,
                      wsrep_gtid_t* upto,
                      int           timeout)
{
  DBUG_ASSERT(upto);
  enum wsrep::provider::status ret;
  if (upto)
  {
    wsrep::gtid upto_gtid(wsrep::id(upto->uuid.data, sizeof(upto->uuid.data)),
                          wsrep::seqno(upto->seqno));
    ret= Wsrep_server_state::instance().wait_for_gtid(upto_gtid, timeout);
  }
  else
  {
    ret= Wsrep_server_state::instance().causal_read(timeout).second;
  }
  WSREP_DEBUG("wsrep_sync_wait_upto: %d", ret);
  return ret;
}

void wsrep_keys_free(wsrep_key_arr_t* key_arr)
{
    for (size_t i= 0; i < key_arr->keys_len; ++i)
    {
        my_free((void*)key_arr->keys[i].key_parts);
    }
    my_free(key_arr->keys);
    key_arr->keys= 0;
    key_arr->keys_len= 0;
}

/*!
 * @param db      Database string
 * @param table   Table string
 * @param key     Array of wsrep_key_t
 * @param key_len In: number of elements in key array, Out: number of
 *                elements populated
 *
 * @return true if preparation was successful, otherwise false.
 */

static bool wsrep_prepare_key_for_isolation(const char* db,
                                           const char* table,
                                           wsrep_buf_t* key,
                                           size_t* key_len)
{
  if (*key_len < 2) return false;

  switch (wsrep_protocol_version)
  {
  case 0:
    *key_len= 0;
    break;
  case 1:
  case 2:
  case 3:
  case 4:
  {
    *key_len= 0;
    if (db)
    {
      key[*key_len].ptr= db;
      key[*key_len].len= strlen(db);
      ++(*key_len);
      if (table)
      {
        key[*key_len].ptr= table;
        key[*key_len].len= strlen(table);
        ++(*key_len);
      }
    }
    break;
  }
  default:
    assert(0);
    WSREP_ERROR("Unsupported protocol version: %ld", wsrep_protocol_version);
    unireg_abort(1);
    return false;
  }

    return true;
}

static bool wsrep_prepare_key_for_isolation(const char* db,
                                            const char* table,
                                            wsrep_key_arr_t* ka)
{
  wsrep_key_t* tmp;
  tmp= (wsrep_key_t*)my_realloc(ka->keys,
                                (ka->keys_len + 1) * sizeof(wsrep_key_t),
                                MYF(MY_ALLOW_ZERO_PTR));
  if (!tmp)
  {
    WSREP_ERROR("Can't allocate memory for key_array");
    return false;
  }
  ka->keys= tmp;
  if (!(ka->keys[ka->keys_len].key_parts= (wsrep_buf_t*)
        my_malloc(sizeof(wsrep_buf_t)*2, MYF(0))))
  {
    WSREP_ERROR("Can't allocate memory for key_parts");
    return false;
  }
  ka->keys[ka->keys_len].key_parts_num= 2;
  ++ka->keys_len;
  if (!wsrep_prepare_key_for_isolation(db, table,
                                       (wsrep_buf_t*)ka->keys[ka->keys_len - 1].key_parts,
                                       &ka->keys[ka->keys_len - 1].key_parts_num))
  {
    WSREP_ERROR("Preparing keys for isolation failed");
    return false;
  }

  return true;
}

static bool wsrep_prepare_keys_for_alter_add_fk(const char* child_table_db,
                                                Alter_info* alter_info,
                                                wsrep_key_arr_t* ka)
{
  Key *key;
  List_iterator<Key> key_iterator(alter_info->key_list);
  while ((key= key_iterator++))
  {
    if (key->type == Key::FOREIGN_KEY)
    {
      Foreign_key *fk_key= (Foreign_key *)key;
      const char *db_name= fk_key->ref_db.str;
      const char *table_name= fk_key->ref_table.str;
      if (!db_name)
      {
        db_name= child_table_db;
      }
      if (!wsrep_prepare_key_for_isolation(db_name, table_name, ka))
      {
        return false;
      }
    }
  }
  return true;
}

static bool wsrep_prepare_keys_for_isolation(THD*              thd,
                                             const char*       db,
                                             const char*       table,
                                             const TABLE_LIST* table_list,
                                             Alter_info*       alter_info,
                                             wsrep_key_arr_t*  ka)
{
  ka->keys= 0;
  ka->keys_len= 0;

  if (db || table)
  {
    if (!wsrep_prepare_key_for_isolation(db, table, ka))
      goto err;
  }

  for (const TABLE_LIST* table= table_list; table; table= table->next_global)
  {
    if (!wsrep_prepare_key_for_isolation(table->db.str, table->table_name.str, ka))
      goto err;
  }

  if (alter_info && (alter_info->flags & (ALTER_ADD_FOREIGN_KEY)))
  {
    if (!wsrep_prepare_keys_for_alter_add_fk(table_list->db.str, alter_info, ka))
      goto err;
  }
  return false;

err:
    wsrep_keys_free(ka);
    return true;
}

/*
 * Prepare key list from db/table and table_list
 *
 * Return zero in case of success, 1 in case of failure.
 */

bool wsrep_prepare_keys_for_isolation(THD*              thd,
                                      const char*       db,
                                      const char*       table,
                                      const TABLE_LIST* table_list,
                                      wsrep_key_arr_t*  ka)
{
  return wsrep_prepare_keys_for_isolation(thd, db, table, table_list, NULL, ka);
}

bool wsrep_prepare_key(const uchar* cache_key, size_t cache_key_len,
                       const uchar* row_id, size_t row_id_len,
                       wsrep_buf_t* key, size_t* key_len)
{
    if (*key_len < 3) return false;

    *key_len= 0;
    switch (wsrep_protocol_version)
    {
    case 0:
    {
        key[0].ptr= cache_key;
        key[0].len= cache_key_len;

        *key_len= 1;
        break;
    }
    case 1:
    case 2:
    case 3:
    case 4:
    {
        key[0].ptr= cache_key;
        key[0].len= strlen( (char*)cache_key );

        key[1].ptr= cache_key + strlen( (char*)cache_key ) + 1;
        key[1].len= strlen( (char*)(key[1].ptr) );

        *key_len= 2;
        break;
    }
    default:
        return false;
    }

    key[*key_len].ptr= row_id;
    key[*key_len].len= row_id_len;
    ++(*key_len);

    return true;
}

bool wsrep_prepare_key_for_innodb(THD* thd,
                                  const uchar* cache_key,
                                  size_t cache_key_len,
                                  const uchar* row_id,
                                  size_t row_id_len,
                                  wsrep_buf_t* key,
                                  size_t* key_len)
{

  return wsrep_prepare_key(cache_key, cache_key_len, row_id, row_id_len, key, key_len);
}

wsrep::key wsrep_prepare_key_for_toi(const char* db, const char* table,
                                     enum wsrep::key::type type)
{
  wsrep::key ret(type);
  DBUG_ASSERT(db);
  ret.append_key_part(db, strlen(db));
  if (table) ret.append_key_part(table, strlen(table));
  return ret;
}

wsrep::key_array
wsrep_prepare_keys_for_alter_add_fk(const char* child_table_db,
                                    Alter_info* alter_info)

{
  wsrep::key_array ret;
  Key *key;
  List_iterator<Key> key_iterator(alter_info->key_list);
  while ((key= key_iterator++))
  {
    if (key->type == Key::FOREIGN_KEY)
    {
      Foreign_key *fk_key= (Foreign_key *)key;
      const char *db_name= fk_key->ref_db.str;
      const char *table_name= fk_key->ref_table.str;
      if (!db_name)
      {
        db_name= child_table_db;
      }
      ret.push_back(wsrep_prepare_key_for_toi(db_name, table_name,
                                              wsrep::key::exclusive));
    }
  }
  return ret;
}

wsrep::key_array wsrep_prepare_keys_for_toi(const char* db,
                                            const char* table,
                                            const TABLE_LIST* table_list,
                                            Alter_info* alter_info)
{
  wsrep::key_array ret;
  if (db || table)
  {
    ret.push_back(wsrep_prepare_key_for_toi(db, table, wsrep::key::exclusive));
  }
  for (const TABLE_LIST* table= table_list; table; table= table->next_global)
  {
    ret.push_back(wsrep_prepare_key_for_toi(table->db.str, table->table_name.str,
                                            wsrep::key::exclusive));
  }
  if (alter_info && (alter_info->flags & ALTER_ADD_FOREIGN_KEY))
  {
    wsrep::key_array fk(wsrep_prepare_keys_for_alter_add_fk(table_list->db.str, alter_info));
    if (!fk.empty())
    {
      ret.insert(ret.end(), fk.begin(), fk.end());
    }
  }
  return ret;
}
/*
 * Construct Query_log_Event from thd query and serialize it
 * into buffer.
 *
 * Return 0 in case of success, 1 in case of error.
 */
int wsrep_to_buf_helper(
    THD* thd, const char *query, uint query_len, uchar** buf, size_t* buf_len)
{
  IO_CACHE tmp_io_cache;
  Log_event_writer writer(&tmp_io_cache, 0);
  if (open_cached_file(&tmp_io_cache, mysql_tmpdir, TEMP_PREFIX,
                       65536, MYF(MY_WME)))
    return 1;
  int ret(0);
  enum enum_binlog_checksum_alg current_binlog_check_alg=
    (enum_binlog_checksum_alg) binlog_checksum_options;

  Format_description_log_event *tmp_fd= new Format_description_log_event(4);
  tmp_fd->checksum_alg= current_binlog_check_alg;
  writer.write(tmp_fd);
  delete tmp_fd;

#ifdef GTID_SUPPORT
  if (thd->variables.gtid_next.type == GTID_GROUP)
  {
      Gtid_log_event gtid_ev(thd, FALSE, &thd->variables.gtid_next);
      if (!gtid_ev.is_valid()) ret= 0;
      if (!ret && writer.write(&gtid_ev)) ret= 1;
  }
#endif /* GTID_SUPPORT */
  if (wsrep_gtid_mode && thd->variables.gtid_seq_no)
  {
    Gtid_log_event gtid_event(thd, thd->variables.gtid_seq_no,
                          thd->variables.gtid_domain_id,
                          true, LOG_EVENT_SUPPRESS_USE_F,
                          true, 0);
    gtid_event.server_id= thd->variables.server_id;
    if (!gtid_event.is_valid()) ret= 0;
    ret= writer.write(&gtid_event);
  }

  /* if there is prepare query, add event for it */
  if (!ret && thd->wsrep_TOI_pre_query)
  {
    Query_log_event ev(thd, thd->wsrep_TOI_pre_query,
		       thd->wsrep_TOI_pre_query_len,
		       FALSE, FALSE, FALSE, 0);
    ev.checksum_alg= current_binlog_check_alg;
    if (writer.write(&ev)) ret= 1;
  }

  /* continue to append the actual query */
  Query_log_event ev(thd, query, query_len, FALSE, FALSE, FALSE, 0);
  ev.checksum_alg= current_binlog_check_alg;
  if (!ret && writer.write(&ev)) ret= 1;
  if (!ret && wsrep_write_cache_buf(&tmp_io_cache, buf, buf_len)) ret= 1;
  close_cached_file(&tmp_io_cache);
  return ret;
}

static int
wsrep_alter_query_string(THD *thd, String *buf)
{
  /* Append the "ALTER" part of the query */
  if (buf->append(STRING_WITH_LEN("ALTER ")))
    return 1;
  /* Append definer */
  append_definer(thd, buf, &(thd->lex->definer->user), &(thd->lex->definer->host));
  /* Append the left part of thd->query after event name part */
  if (buf->append(thd->lex->stmt_definition_begin,
                  thd->lex->stmt_definition_end -
                  thd->lex->stmt_definition_begin))
    return 1;

  return 0;
}

static int wsrep_alter_event_query(THD *thd, uchar** buf, size_t* buf_len)
{
  String log_query;

  if (wsrep_alter_query_string(thd, &log_query))
  {
    WSREP_WARN("events alter string failed: schema: %s, query: %s",
               thd->get_db(), thd->query());
    return 1;
  }
  return wsrep_to_buf_helper(thd, log_query.ptr(), log_query.length(), buf, buf_len);
}

#include "sql_show.h"
static int
create_view_query(THD *thd, uchar** buf, size_t* buf_len)
{
    LEX *lex= thd->lex;
    SELECT_LEX *select_lex= lex->first_select_lex();
    TABLE_LIST *first_table= select_lex->table_list.first;
    TABLE_LIST *views= first_table;
    LEX_USER *definer;
    String buff;
    const LEX_CSTRING command[3]=
      {{ STRING_WITH_LEN("CREATE ") },
       { STRING_WITH_LEN("ALTER ") },
       { STRING_WITH_LEN("CREATE OR REPLACE ") }};

    buff.append(&command[thd->lex->create_view->mode]);

    if (lex->definer)
      definer= get_current_user(thd, lex->definer);
    else
    {
      /*
        DEFINER-clause is missing; we have to create default definer in
        persistent arena to be PS/SP friendly.
        If this is an ALTER VIEW then the current user should be set as
        the definer.
      */
      definer= create_default_definer(thd, false);
    }

    if (definer)
    {
      views->definer.user= definer->user;
      views->definer.host= definer->host;
    } else {
      WSREP_ERROR("Failed to get DEFINER for VIEW.");
      return 1;
    }

    views->algorithm   = lex->create_view->algorithm;
    views->view_suid   = lex->create_view->suid;
    views->with_check  = lex->create_view->check;

    view_store_options(thd, views, &buff);
    buff.append(STRING_WITH_LEN("VIEW "));
    /* Test if user supplied a db (ie: we did not use thd->db) */
    if (views->db.str && views->db.str[0] &&
        (thd->db.str == NULL || cmp(&views->db, &thd->db)))
    {
      append_identifier(thd, &buff, &views->db);
      buff.append('.');
    }
    append_identifier(thd, &buff, &views->table_name);
    if (lex->view_list.elements)
    {
      List_iterator_fast<LEX_CSTRING> names(lex->view_list);
      LEX_CSTRING *name;
      int i;

      for (i= 0; (name= names++); i++)
      {
        buff.append(i ? ", " : "(");
        append_identifier(thd, &buff, name);
      }
      buff.append(')');
    }
    buff.append(STRING_WITH_LEN(" AS "));
    buff.append(thd->lex->create_view->select.str,
                thd->lex->create_view->select.length);
    return wsrep_to_buf_helper(thd, buff.ptr(), buff.length(), buf, buf_len);
}

/*
  Rewrite DROP TABLE for TOI. Temporary tables are eliminated from
  the query as they are visible only to client connection.

  TODO: See comments for sql_base.cc:drop_temporary_table() and refine
  the function to deal with transactional locked tables.
 */
static int wsrep_drop_table_query(THD* thd, uchar** buf, size_t* buf_len)
{

  LEX* lex= thd->lex;
  SELECT_LEX* select_lex= lex->first_select_lex();
  TABLE_LIST* first_table= select_lex->table_list.first;
  String buff;

  DBUG_ASSERT(!lex->create_info.tmp_table());

  bool found_temp_table= false;
  for (TABLE_LIST* table= first_table; table; table= table->next_global)
  {
    if (thd->find_temporary_table(table->db.str, table->table_name.str))
    {
      found_temp_table= true;
      break;
    }
  }

  if (found_temp_table)
  {
    buff.append("DROP TABLE ");
    if (lex->check_exists)
      buff.append("IF EXISTS ");

    for (TABLE_LIST* table= first_table; table; table= table->next_global)
    {
      if (!thd->find_temporary_table(table->db.str, table->table_name.str))
      {
        append_identifier(thd, &buff, table->db.str, table->db.length);
        buff.append(".");
        append_identifier(thd, &buff,
                          table->table_name.str, table->table_name.length);
        buff.append(",");
      }
    }

    /* Chop the last comma */
    buff.chop();
    buff.append(" /* generated by wsrep */");

    WSREP_DEBUG("Rewrote '%s' as '%s'", thd->query(), buff.ptr());

    return wsrep_to_buf_helper(thd, buff.ptr(), buff.length(), buf, buf_len);
  }
  else
  {
    return wsrep_to_buf_helper(thd, thd->query(), thd->query_length(),
                               buf, buf_len);
  }
}


/* Forward declarations. */
int wsrep_create_trigger_query(THD *thd, uchar** buf, size_t* buf_len);

/*
  Decide if statement should run in TOI.

  Look if table or table_list contain temporary tables. If the
  statement affects only temporary tables,   statement should not run
  in TOI. If the table list contains mix of regular and temporary tables
  (DROP TABLE, OPTIMIZE, ANALYZE), statement should be run in TOI but
  should be rewritten at later time for replication to contain only
  non-temporary tables.
 */
static bool wsrep_can_run_in_toi(THD *thd, const char *db, const char *table,
                                 const TABLE_LIST *table_list)
{
  DBUG_ASSERT(!table || db);
  DBUG_ASSERT(table_list || db);

  LEX* lex= thd->lex;
  SELECT_LEX* select_lex= lex->first_select_lex();
  TABLE_LIST* first_table= select_lex->table_list.first;

  switch (lex->sql_command)
  {
  case SQLCOM_CREATE_TABLE:
    DBUG_ASSERT(!table_list);
    if (thd->lex->create_info.options & HA_LEX_CREATE_TMP_TABLE)
    {
      return false;
    }
    /*
      If mariadb master has replicated a CTAS, we should not replicate the create table
      part separately as TOI, but to replicate both create table and following inserts
      as one write set.
      Howver, if CTAS creates empty table, we should replicate the create table alone
      as TOI. We have to do relay log event lookup to see if row events follow the
      create table event.
    */
    if (thd->slave_thread && !(thd->rgi_slave->gtid_ev_flags2 & Gtid_log_event::FL_STANDALONE))
    {
      /* this is CTAS, either empty or populated table */
      ulonglong event_size = 0;
      enum Log_event_type ev_type= wsrep_peak_event(thd->rgi_slave, &event_size);
      switch (ev_type)
      {
      case QUERY_EVENT:
        /* CTAS with empty table, we replicate create table as TOI */
        break;

      case TABLE_MAP_EVENT:
        WSREP_DEBUG("replicating CTAS of empty table as TOI");
        // fall through
      case WRITE_ROWS_EVENT:
        /* CTAS with populated table, we replicate later at commit time */
        WSREP_DEBUG("skipping create table of CTAS replication");
        return false;

      default:
        WSREP_WARN("unexpected async replication event: %d", ev_type);
      }
      return true;
    }
    /* no next async replication event */
    return true;

  case SQLCOM_CREATE_VIEW:

    DBUG_ASSERT(!table_list);
    DBUG_ASSERT(first_table); /* First table is view name */
    /*
      If any of the remaining tables refer to temporary table error
      is returned to client, so TOI can be skipped
    */
    for (TABLE_LIST* it= first_table->next_global; it; it= it->next_global)
    {
      if (thd->find_temporary_table(it))
      {
        return false;
      }
    }
    return true;

  case SQLCOM_CREATE_TRIGGER:

    DBUG_ASSERT(first_table);

    if (thd->find_temporary_table(first_table))
    {
      return false;
    }
    return true;

  case SQLCOM_DROP_TRIGGER:
    DBUG_ASSERT(table_list);
    if (thd->find_temporary_table(table_list))
    {
      return false;
    }
    return true;

  default:
    if (table && !thd->find_temporary_table(db, table))
    {
      return true;
    }

    if (table_list)
    {
      for (TABLE_LIST* table= first_table; table; table= table->next_global)
      {
        if (!thd->find_temporary_table(table->db.str, table->table_name.str))
        {
          return true;
        }
      }
    }
    return !(table || table_list);
  }
}


static int wsrep_create_sp(THD *thd, uchar** buf, size_t* buf_len)
{
  String log_query;
  sp_head *sp= thd->lex->sphead;
  sql_mode_t saved_mode= thd->variables.sql_mode;
  String retstr(64);
  LEX_CSTRING returns= empty_clex_str;
  retstr.set_charset(system_charset_info);

  log_query.set_charset(system_charset_info);

  if (sp->m_handler->type() == TYPE_ENUM_FUNCTION)
  {
    sp_returns_type(thd, retstr, sp);
    returns= retstr.lex_cstring();
  }
  if (sp->m_handler->
      show_create_sp(thd, &log_query,
                     sp->m_explicit_name ? sp->m_db : null_clex_str,
                     sp->m_name, sp->m_params, returns,
                     sp->m_body, sp->chistics(),
                     thd->lex->definer[0],
                     thd->lex->create_info,
                     saved_mode))
  {
    WSREP_WARN("SP create string failed: schema: %s, query: %s",
               thd->get_db(), thd->query());
    return 1;
  }

  return wsrep_to_buf_helper(thd, log_query.ptr(), log_query.length(), buf, buf_len);
}

static int wsrep_TOI_event_buf(THD* thd, uchar** buf, size_t* buf_len)
{
  int err;
  switch (thd->lex->sql_command)
  {
  case SQLCOM_CREATE_VIEW:
    err= create_view_query(thd, buf, buf_len);
    break;
  case SQLCOM_CREATE_PROCEDURE:
  case SQLCOM_CREATE_SPFUNCTION:
    err= wsrep_create_sp(thd, buf, buf_len);
    break;
  case SQLCOM_CREATE_TRIGGER:
    err= wsrep_create_trigger_query(thd, buf, buf_len);
    break;
  case SQLCOM_CREATE_EVENT:
    err= wsrep_create_event_query(thd, buf, buf_len);
    break;
  case SQLCOM_ALTER_EVENT:
    err= wsrep_alter_event_query(thd, buf, buf_len);
    break;
  case SQLCOM_DROP_TABLE:
    err= wsrep_drop_table_query(thd, buf, buf_len);
    break;
  case SQLCOM_CREATE_ROLE:
    if (sp_process_definer(thd))
    {
      WSREP_WARN("Failed to set CREATE ROLE definer for TOI.");
    }
    /* fallthrough */
  default:
    err= wsrep_to_buf_helper(thd, thd->query(), thd->query_length(), buf,
                             buf_len);
    break;
  }

  return err;
}

static void wsrep_TOI_begin_failed(THD* thd, const wsrep_buf_t* /* const err */)
{
  if (wsrep_thd_trx_seqno(thd) > 0)
  {
    /* GTID was granted and TO acquired - need to log event and release TO */
    if (wsrep_emulate_bin_log) wsrep_thd_binlog_trx_reset(thd);
    if (wsrep_write_dummy_event(thd, "TOI begin failed")) { goto fail; }
    wsrep::client_state& cs(thd->wsrep_cs());
    std::string const err(wsrep::to_c_string(cs.current_error()));
    wsrep::mutable_buffer err_buf;
    err_buf.push_back(err);
    int const ret= cs.leave_toi_local(err_buf);
    if (ret)
    {
      WSREP_ERROR("Leaving critical section for failed TOI failed: thd: %lld, "
                  "schema: %s, SQL: %s, rcode: %d wsrep_error: %s",
                  (long long)thd->real_id, thd->db.str,
                  thd->query(), ret, err.c_str());
      goto fail;
    }
  }
  return;
fail:
  WSREP_ERROR("Failed to release TOI resources. Need to abort.");
  unireg_abort(1);
}


/*
  returns:
   0: statement was replicated as TOI
   1: TOI replication was skipped
  -1: TOI replication failed
 */
static int wsrep_TOI_begin(THD *thd, const char *db, const char *table,
                           const TABLE_LIST* table_list,
                           Alter_info* alter_info)
{
  DBUG_ASSERT(thd->variables.wsrep_OSU_method == WSREP_OSU_TOI);

  WSREP_DEBUG("TOI Begin");
  if (wsrep_can_run_in_toi(thd, db, table, table_list) == false)
  {
    WSREP_DEBUG("No TOI for %s", WSREP_QUERY(thd));
    return 1;
  }

  uchar* buf= 0;
  size_t buf_len(0);
  int buf_err;
  int rc;

  buf_err= wsrep_TOI_event_buf(thd, &buf, &buf_len);
  if (buf_err) {
    WSREP_ERROR("Failed to create TOI event buf: %d", buf_err);
    my_message(ER_UNKNOWN_ERROR,
               "WSREP replication failed to prepare TOI event buffer. "
               "Check your query.",
               MYF(0));
    return -1;
  }
  struct wsrep_buf buff= { buf, buf_len };

  wsrep::key_array key_array=
    wsrep_prepare_keys_for_toi(db, table, table_list, alter_info);

  if (thd->has_read_only_protection())
  {
    /* non replicated DDL, affecting temporary tables only */
    WSREP_DEBUG("TO isolation skipped, sql: %s."
                "Only temporary tables affected.",
                WSREP_QUERY(thd));
    if (buf) my_free(buf);
    return -1;
  }

  thd_proc_info(thd, "acquiring total order isolation");

  wsrep::client_state& cs(thd->wsrep_cs());
  int ret= cs.enter_toi_local(key_array,
                              wsrep::const_buffer(buff.ptr, buff.len),
                              wsrep::provider::flag::start_transaction |
                              wsrep::provider::flag::commit);

  if (ret)
  {
    DBUG_ASSERT(cs.current_error());
    WSREP_DEBUG("to_execute_start() failed for %llu: %s, seqno: %lld",
                thd->thread_id, WSREP_QUERY(thd),
                (long long)wsrep_thd_trx_seqno(thd));

    /* jump to error handler in mysql_execute_command() */
    switch (cs.current_error())
    {
    case wsrep::e_size_exceeded_error:
      WSREP_WARN("TO isolation failed for: %d, schema: %s, sql: %s. "
                 "Maximum size exceeded.",
                 ret,
                 (thd->db.str ? thd->db.str : "(null)"),
                 WSREP_QUERY(thd));
      my_error(ER_ERROR_DURING_COMMIT, MYF(0), WSREP_SIZE_EXCEEDED);
      break;
    default:
      WSREP_WARN("TO isolation failed for: %d, schema: %s, sql: %s. "
                 "Check wsrep connection state and retry the query.",
                 ret,
                 (thd->db.str ? thd->db.str : "(null)"),
                 WSREP_QUERY(thd));
      if (!thd->is_error())
      {
        my_error(ER_LOCK_DEADLOCK, MYF(0), "WSREP replication failed. Check "
                 "your wsrep connection state and retry the query.");
      }
    }
    rc= -1;
  }
  else {
    ++wsrep_to_isolation;
    rc= 0;
  }

  if (buf) my_free(buf);

  if (rc) wsrep_TOI_begin_failed(thd, NULL);

  return rc;
}

static void wsrep_TOI_end(THD *thd) {
  wsrep_to_isolation--;
  wsrep::client_state& client_state(thd->wsrep_cs());
  DBUG_ASSERT(wsrep_thd_is_local_toi(thd));
  WSREP_DEBUG("TO END: %lld: %s", client_state.toi_meta().seqno().get(),
              WSREP_QUERY(thd));

  if (wsrep_thd_is_local_toi(thd))
  {
    wsrep_set_SE_checkpoint(client_state.toi_meta().gtid());
    wsrep::mutable_buffer err;
    if (thd->is_error() && !wsrep_must_ignore_error(thd))
    {
        wsrep_store_error(thd, err);
    }
    int const ret= client_state.leave_toi_local(err);
    if (!ret)
    {
      WSREP_DEBUG("TO END: %lld", client_state.toi_meta().seqno().get());
    }
    else
    {
      WSREP_WARN("TO isolation end failed for: %d, schema: %s, sql: %s",
                 ret, (thd->db.str ? thd->db.str : "(null)"), WSREP_QUERY(thd));
    }
  }
}

static int wsrep_RSU_begin(THD *thd, const char *db_, const char *table_)
{
  WSREP_DEBUG("RSU BEGIN: %lld, : %s", wsrep_thd_trx_seqno(thd),
              WSREP_QUERY(thd));
  if (thd->wsrep_cs().begin_rsu(5000))
  {
    WSREP_WARN("RSU begin failed");
  }
  else
  {
    thd->variables.wsrep_on= 0;
  }
  return 0;
}

static void wsrep_RSU_end(THD *thd)
{
  WSREP_DEBUG("RSU END: %lld : %s", wsrep_thd_trx_seqno(thd),
              WSREP_QUERY(thd));
  if (thd->wsrep_cs().end_rsu())
  {
    WSREP_WARN("Failed to end RSU, server may need to be restarted");
  }
  thd->variables.wsrep_on= 1;
}

int wsrep_to_isolation_begin(THD *thd, const char *db_, const char *table_,
                             const TABLE_LIST* table_list,
                             Alter_info* alter_info)
{
  /*
    No isolation for applier or replaying threads.
   */
  if (!wsrep_thd_is_local(thd)) return 0;

  int ret= 0;
  mysql_mutex_lock(&thd->LOCK_thd_data);

  if (thd->wsrep_trx().state() == wsrep::transaction::s_must_abort)
  {
    WSREP_INFO("thread: %lld  schema: %s  query: %s has been aborted due to multi-master conflict",
               (longlong) thd->thread_id, thd->get_db(), thd->query());
    mysql_mutex_unlock(&thd->LOCK_thd_data);
    return WSREP_TRX_FAIL;
  }
  mysql_mutex_unlock(&thd->LOCK_thd_data);

  DBUG_ASSERT(wsrep_thd_is_local(thd));
  DBUG_ASSERT(thd->wsrep_trx().ws_meta().seqno().is_undefined());

  if (thd->global_read_lock.is_acquired())
  {
    WSREP_DEBUG("Aborting TOI: Global Read-Lock (FTWRL) in place: %s %llu",
                WSREP_QUERY(thd), thd->thread_id);
    return -1;
  }

  if (wsrep_debug && thd->mdl_context.has_locks())
  {
    WSREP_DEBUG("thread holds MDL locks at TI begin: %s %llu",
                WSREP_QUERY(thd), thd->thread_id);
  }

  /*
    It makes sense to set auto_increment_* to defaults in TOI operations.
    Must be done before wsrep_TOI_begin() since Query_log_event encapsulating
    TOI statement and auto inc variables for wsrep replication is constructed
    there. Variables are reset back in THD::reset_for_next_command() before
    processing of next command.
   */
  if (wsrep_auto_increment_control)
  {
    thd->variables.auto_increment_offset= 1;
    thd->variables.auto_increment_increment= 1;
  }

  if (thd->variables.wsrep_on && wsrep_thd_is_local(thd))
  {
    switch (thd->variables.wsrep_OSU_method) {
    case WSREP_OSU_TOI:
      ret= wsrep_TOI_begin(thd, db_, table_, table_list, alter_info);
      break;
    case WSREP_OSU_RSU:
      ret= wsrep_RSU_begin(thd, db_, table_);
      break;
    default:
      WSREP_ERROR("Unsupported OSU method: %lu",
                  thd->variables.wsrep_OSU_method);
      ret= -1;
      break;
    }
    switch (ret) {
    case 0: /* wsrep_TOI_begin sould set toi mode */ break;
    case 1:
      /* TOI replication skipped, treat as success */
      ret= 0;
      break;
    case -1:
      /* TOI replication failed, treat as error */
      break;
    }
  }

  return ret;
}

void wsrep_to_isolation_end(THD *thd)
{
  DBUG_ASSERT(wsrep_thd_is_local_toi(thd) ||
              wsrep_thd_is_in_rsu(thd));
  if (wsrep_thd_is_local_toi(thd))
  {
    DBUG_ASSERT(thd->variables.wsrep_OSU_method == WSREP_OSU_TOI);
    wsrep_TOI_end(thd);
  }
  else if (wsrep_thd_is_in_rsu(thd))
  {
    DBUG_ASSERT(thd->variables.wsrep_OSU_method == WSREP_OSU_RSU);
    wsrep_RSU_end(thd);
  }
  else
  {
    DBUG_ASSERT(0);
  }
  if (wsrep_emulate_bin_log) wsrep_thd_binlog_trx_reset(thd);
}

#define WSREP_MDL_LOG(severity, msg, schema, schema_len, req, gra)             \
    WSREP_##severity(                                                          \
      "%s\n"                                                                   \
      "schema:  %.*s\n"                                                        \
      "request: (%llu \tseqno %lld \twsrep (%s, %s, %s) cmd %d %d \t%s)\n"     \
      "granted: (%llu \tseqno %lld \twsrep (%s, %s, %s) cmd %d %d \t%s)",      \
      msg, schema_len, schema,                                                 \
      req->thread_id, (long long)wsrep_thd_trx_seqno(req),                     \
      wsrep_thd_client_mode_str(req), wsrep_thd_client_state_str(req), wsrep_thd_transaction_state_str(req), \
      req->get_command(), req->lex->sql_command, req->query(),                 \
      gra->thread_id, (long long)wsrep_thd_trx_seqno(gra),                     \
      wsrep_thd_client_mode_str(gra), wsrep_thd_client_state_str(gra), wsrep_thd_transaction_state_str(gra), \
      gra->get_command(), gra->lex->sql_command, gra->query());

/**
  Check if request for the metadata lock should be granted to the requester.

  @param  requestor_ctx        The MDL context of the requestor
  @param  ticket               MDL ticket for the requested lock

  @retval TRUE   Lock request can be granted
  @retval FALSE  Lock request cannot be granted
*/

void wsrep_handle_mdl_conflict(MDL_context *requestor_ctx,
                               MDL_ticket *ticket,
                               const MDL_key *key)
{
  /* Fallback to the non-wsrep behaviour */
  if (!WSREP_ON) return;

  THD *request_thd= requestor_ctx->get_thd();
  THD *granted_thd= ticket->get_ctx()->get_thd();

  const char* schema= key->db_name();
  int schema_len= key->db_name_length();

  mysql_mutex_lock(&request_thd->LOCK_thd_data);
  if (wsrep_thd_is_toi(request_thd) ||
      wsrep_thd_is_applying(request_thd)) {

    mysql_mutex_unlock(&request_thd->LOCK_thd_data);
    WSREP_MDL_LOG(DEBUG, "MDL conflict ", schema, schema_len,
                  request_thd, granted_thd);
    ticket->wsrep_report(wsrep_debug);

    mysql_mutex_lock(&granted_thd->LOCK_thd_data);
    if (wsrep_thd_is_toi(granted_thd) ||
        wsrep_thd_is_applying(granted_thd))
    {
      if (wsrep_thd_is_aborting(granted_thd))
      {
        WSREP_DEBUG("BF thread waiting for SR in aborting state");
        ticket->wsrep_report(wsrep_debug);
        mysql_mutex_unlock(&granted_thd->LOCK_thd_data);
      }
      else if (wsrep_thd_is_SR(granted_thd) && !wsrep_thd_is_SR(request_thd))
      {
        WSREP_MDL_LOG(INFO, "MDL conflict, DDL vs SR", 
                      schema, schema_len, request_thd, granted_thd);
        mysql_mutex_unlock(&granted_thd->LOCK_thd_data);
        wsrep_abort_thd(request_thd, granted_thd, 1);
      }
      else
      {
        WSREP_MDL_LOG(INFO, "MDL BF-BF conflict", schema, schema_len,
                      request_thd, granted_thd);
        ticket->wsrep_report(true);
        mysql_mutex_unlock(&granted_thd->LOCK_thd_data);
        unireg_abort(1);
      }
    }
    else if (granted_thd->lex->sql_command == SQLCOM_FLUSH ||
             granted_thd->mdl_context.has_explicit_locks())
    {
      WSREP_DEBUG("BF thread waiting for FLUSH");
      ticket->wsrep_report(wsrep_debug);
      mysql_mutex_unlock(&granted_thd->LOCK_thd_data);
    }
    else if (request_thd->lex->sql_command == SQLCOM_DROP_TABLE)
    {
      WSREP_DEBUG("DROP caused BF abort, conf %s",
                  wsrep_thd_transaction_state_str(granted_thd));
      ticket->wsrep_report(wsrep_debug);
      mysql_mutex_unlock(&granted_thd->LOCK_thd_data);
      wsrep_abort_thd(request_thd, granted_thd, 1);
    }
    else
    {
      WSREP_MDL_LOG(DEBUG, "MDL conflict-> BF abort", schema, schema_len,
                    request_thd, granted_thd);
      ticket->wsrep_report(wsrep_debug);
      if (granted_thd->wsrep_trx().active())
      {
        mysql_mutex_unlock(&granted_thd->LOCK_thd_data);
        wsrep_abort_thd(request_thd, granted_thd, 1);
      }
      else
      {
        /*
          Granted_thd is likely executing with wsrep_on=0. If the requesting
          thd is BF, BF abort and wait.
        */
        mysql_mutex_unlock(&granted_thd->LOCK_thd_data);
        if (wsrep_thd_is_BF(request_thd, FALSE))
        {
          ha_abort_transaction(request_thd, granted_thd, TRUE);
        }
        else
        {
	  WSREP_MDL_LOG(INFO, "MDL unknown BF-BF conflict", schema, schema_len,
                      request_thd, granted_thd);
	  ticket->wsrep_report(true);
	  unireg_abort(1);
        }
      }
    }
  }
  else
  {
    mysql_mutex_unlock(&request_thd->LOCK_thd_data);
  }
}

/**/
static bool abort_replicated(THD *thd)
{
  bool ret_code= false;
  if (thd->wsrep_trx().state() == wsrep::transaction::s_committing)
  {
    WSREP_DEBUG("aborting replicated trx: %llu", (ulonglong)(thd->real_id));

    (void)wsrep_abort_thd(thd, thd, TRUE);
    ret_code= true;
  }
  return ret_code;
}

/**/
static inline bool is_client_connection(THD *thd)
{
  return (thd->wsrep_client_thread && thd->variables.wsrep_on);
}

static inline bool is_replaying_connection(THD *thd)
{
  bool ret;

  mysql_mutex_lock(&thd->LOCK_thd_data);
  ret=  (thd->wsrep_trx().state() == wsrep::transaction::s_replaying) ? true : false;
  mysql_mutex_unlock(&thd->LOCK_thd_data);

  return ret;
}

static inline bool is_committing_connection(THD *thd)
{
  bool ret;

  mysql_mutex_lock(&thd->LOCK_thd_data);
  ret=  (thd->wsrep_trx().state() == wsrep::transaction::s_committing) ? true : false;
  mysql_mutex_unlock(&thd->LOCK_thd_data);

  return ret;
}

static my_bool have_client_connections(THD *thd, void*)
{
  DBUG_PRINT("quit",("Informing thread %lld that it's time to die",
                     (longlong) thd->thread_id));
  if (is_client_connection(thd) && thd->killed == KILL_CONNECTION)
  {
    (void)abort_replicated(thd);
    return 1;
  }
  return 0;
}

static void wsrep_close_thread(THD *thd)
{
  thd->set_killed(KILL_CONNECTION);
  MYSQL_CALLBACK(thread_scheduler, post_kill_notification, (thd));
  mysql_mutex_lock(&thd->LOCK_thd_kill);
  if (thd->mysys_var)
  {
    thd->mysys_var->abort=1;
    mysql_mutex_lock(&thd->mysys_var->mutex);
    if (thd->mysys_var->current_cond)
    {
      mysql_mutex_lock(thd->mysys_var->current_mutex);
      mysql_cond_broadcast(thd->mysys_var->current_cond);
      mysql_mutex_unlock(thd->mysys_var->current_mutex);
    }
    mysql_mutex_unlock(&thd->mysys_var->mutex);
  }
  mysql_mutex_unlock(&thd->LOCK_thd_kill);
}

static my_bool have_committing_connections(THD *thd, void *)
{
  return is_client_connection(thd) && is_committing_connection(thd) ? 1 : 0;
}

int wsrep_wait_committing_connections_close(int wait_time)
{
  int sleep_time= 100;

  while (server_threads.iterate(have_committing_connections) && wait_time > 0)
  {
    WSREP_DEBUG("wait for committing transaction to close: %d", wait_time);
    my_sleep(sleep_time);
    wait_time -= sleep_time;
  }
  return server_threads.iterate(have_committing_connections);
}

static my_bool kill_all_threads(THD *thd, THD *caller_thd)
{
  DBUG_PRINT("quit", ("Informing thread %lld that it's time to die",
                      (longlong) thd->thread_id));
  /* We skip slave threads & scheduler on this first loop through. */
  if (is_client_connection(thd) && thd != caller_thd)
  {
    if (is_replaying_connection(thd))
      thd->set_killed(KILL_CONNECTION);
    else if (!abort_replicated(thd))
    {
      /* replicated transactions must be skipped */
      WSREP_DEBUG("closing connection %lld", (longlong) thd->thread_id);
      /* instead of wsrep_close_thread() we do now  soft kill by THD::awake */
      thd->awake(KILL_CONNECTION);
    }
  }
  return 0;
}

static my_bool kill_remaining_threads(THD *thd, THD *caller_thd)
{
#ifndef __bsdi__				// Bug in BSDI kernel
  if (is_client_connection(thd) &&
      !abort_replicated(thd)    &&
      !is_replaying_connection(thd) &&
      thd != caller_thd)
  {
    WSREP_INFO("killing local connection: %lld", (longlong) thd->thread_id);
    close_connection(thd, 0);
  }
#endif
  return 0;
}

void wsrep_close_client_connections(my_bool wait_to_end, THD* except_caller_thd)
{
  /* Clear thread cache */
  kill_cached_threads++;
  flush_thread_cache();
  
  /*
    First signal all threads that it's time to die
  */
  server_threads.iterate(kill_all_threads, except_caller_thd);

  /*
    Force remaining threads to die by closing the connection to the client
  */
  server_threads.iterate(kill_remaining_threads, except_caller_thd);

  DBUG_PRINT("quit", ("Waiting for threads to die (count=%u)",
             uint32_t(thread_count)));
  WSREP_DEBUG("waiting for client connections to close: %u",
              uint32_t(thread_count));

  while (wait_to_end && server_threads.iterate(have_client_connections))
  {
    sleep(1);
    DBUG_PRINT("quit",("One thread died (count=%u)", uint32_t(thread_count)));
  }

  /* All client connection threads have now been aborted */
}


void wsrep_close_applier(THD *thd)
{
  WSREP_DEBUG("closing applier %lld", (longlong) thd->thread_id);
  wsrep_close_thread(thd);
}

static my_bool wsrep_close_threads_callback(THD *thd, THD *caller_thd)
{
  DBUG_PRINT("quit",("Informing thread %lld that it's time to die",
                     (longlong) thd->thread_id));
  /* We skip slave threads & scheduler on this first loop through. */
  if (thd->wsrep_applier && thd != caller_thd)
  {
    WSREP_DEBUG("closing wsrep thread %lld", (longlong) thd->thread_id);
    wsrep_close_thread(thd);
  }
  return 0;
}

void wsrep_close_threads(THD *thd)
{
  server_threads.iterate(wsrep_close_threads_callback, thd);
}

void wsrep_wait_appliers_close(THD *thd)
{
  /* Wait for wsrep appliers to gracefully exit */
  mysql_mutex_lock(&LOCK_wsrep_slave_threads);
  while (wsrep_running_threads > 2)
    /*
      2 is for rollbacker thread which needs to be killed explicitly.
      This gotta be fixed in a more elegant manner if we gonna have arbitrary
      number of non-applier wsrep threads.
    */
  {
    mysql_cond_wait(&COND_wsrep_slave_threads, &LOCK_wsrep_slave_threads);
  }
  mysql_mutex_unlock(&LOCK_wsrep_slave_threads);
  DBUG_PRINT("quit",("applier threads have died (count=%u)",
                     uint32_t(wsrep_running_threads)));

  /* Now kill remaining wsrep threads: rollbacker */
  wsrep_close_threads (thd);
  /* and wait for them to die */
  mysql_mutex_lock(&LOCK_wsrep_slave_threads);
  while (wsrep_running_threads > 0)
  {
    mysql_cond_wait(&COND_wsrep_slave_threads, &LOCK_wsrep_slave_threads);
  }
  mysql_mutex_unlock(&LOCK_wsrep_slave_threads);
  DBUG_PRINT("quit",("all wsrep system threads have died"));

  /* All wsrep applier threads have now been aborted. However, if this thread
     is also applier, we are still running...
  */
}

void
wsrep_last_committed_id(wsrep_gtid_t* gtid)
{
  wsrep::gtid ret= Wsrep_server_state::instance().last_committed_gtid();
  memcpy(gtid->uuid.data, ret.id().data(), sizeof(gtid->uuid.data));
  gtid->seqno= ret.seqno().get();
}

void
wsrep_node_uuid(wsrep_uuid_t& uuid)
{
  uuid= node_uuid;
}

int wsrep_must_ignore_error(THD* thd)
{
  const int error= thd->get_stmt_da()->sql_errno();
  const uint flags= sql_command_flags[thd->lex->sql_command];

  DBUG_ASSERT(error);
  DBUG_ASSERT(wsrep_thd_is_toi(thd));

  if ((wsrep_ignore_apply_errors & WSREP_IGNORE_ERRORS_ON_DDL))
    goto ignore_error;

  if ((flags & CF_WSREP_MAY_IGNORE_ERRORS) &&
      (wsrep_ignore_apply_errors & WSREP_IGNORE_ERRORS_ON_RECONCILING_DDL))
  {
    switch (error)
    {
    case ER_DB_DROP_EXISTS:
    case ER_BAD_TABLE_ERROR:
    case ER_CANT_DROP_FIELD_OR_KEY:
      goto ignore_error;
    }
  }

  return 0;

ignore_error:
  WSREP_WARN("Ignoring error '%s' on query. "
             "Default database: '%s'. Query: '%s', Error_code: %d",
             thd->get_stmt_da()->message(),
             print_slave_db_safe(thd->db.str),
             thd->query(),
             error);
  return 1;
}

int wsrep_ignored_error_code(Log_event* ev, int error)
{
  const THD* thd= ev->thd;

  DBUG_ASSERT(error);
  DBUG_ASSERT(wsrep_thd_is_applying(thd) &&
              !wsrep_thd_is_local_toi(thd));

  if ((wsrep_ignore_apply_errors & WSREP_IGNORE_ERRORS_ON_RECONCILING_DML))
  {
    const int ev_type= ev->get_type_code();
    if ((ev_type == DELETE_ROWS_EVENT || ev_type == DELETE_ROWS_EVENT_V1)
        && error == ER_KEY_NOT_FOUND)
      goto ignore_error;
  }

  return 0;

ignore_error:
  WSREP_WARN("Ignoring error '%s' on %s event. Error_code: %d",
             thd->get_stmt_da()->message(),
             ev->get_type_str(),
             error);
  return 1;
}

bool wsrep_provider_is_SR_capable()
{
  return Wsrep_server_state::has_capability(wsrep::provider::capability::streaming);
}

int wsrep_thd_retry_counter(const THD *thd)
{
  return thd->wsrep_retry_counter;
}

extern  bool wsrep_thd_ignore_table(THD *thd)
{
  return thd->wsrep_ignore_table;
}

bool wsrep_is_show_query(enum enum_sql_command command)
{
  DBUG_ASSERT(command >= 0 && command <= SQLCOM_END);
  return (sql_command_flags[command] & CF_STATUS_COMMAND) != 0;
}
bool wsrep_create_like_table(THD* thd, TABLE_LIST* table,
                             TABLE_LIST* src_table,
                             HA_CREATE_INFO *create_info)
{
  if (create_info->tmp_table())
  {
    /* CREATE TEMPORARY TABLE LIKE must be skipped from replication */
    WSREP_DEBUG("CREATE TEMPORARY TABLE LIKE... skipped replication\n %s", 
                thd->query());
  }
  else if (!(thd->find_temporary_table(src_table)))
  {
    /* this is straight CREATE TABLE LIKE... with no tmp tables */
    WSREP_TO_ISOLATION_BEGIN(table->db.str, table->table_name.str, NULL);
  }
  else
  {
    /* here we have CREATE TABLE LIKE <temporary table> 
       the temporary table definition will be needed in slaves to
       enable the create to succeed
     */
    TABLE_LIST tbl;
    bzero((void*) &tbl, sizeof(tbl));
    tbl.db= src_table->db;
    tbl.table_name= tbl.alias= src_table->table_name;
    tbl.table= src_table->table;
    char buf[2048];
    String query(buf, sizeof(buf), system_charset_info);
    query.length(0);  // Have to zero it since constructor doesn't

    (void)  show_create_table(thd, &tbl, &query, NULL, WITH_DB_NAME);
    WSREP_DEBUG("TMP TABLE: %s", query.ptr());

    thd->wsrep_TOI_pre_query=     query.ptr();
    thd->wsrep_TOI_pre_query_len= query.length();

    WSREP_TO_ISOLATION_BEGIN(table->db.str, table->table_name.str, NULL);

    thd->wsrep_TOI_pre_query=      NULL;
    thd->wsrep_TOI_pre_query_len= 0;
  }

  return(false);
#ifdef WITH_WSREP
wsrep_error_label:
  thd->wsrep_TOI_pre_query= NULL;
  return (true);
#endif
}

int wsrep_create_trigger_query(THD *thd, uchar** buf, size_t* buf_len)
{
  LEX *lex= thd->lex;
  String stmt_query;

  LEX_CSTRING definer_user;
  LEX_CSTRING definer_host;

  if (!lex->definer)
  {
    if (!thd->slave_thread)
    {
      if (!(lex->definer= create_default_definer(thd, false)))
        return 1;
    }
  }

  if (lex->definer)
  {
    /* SUID trigger. */
    LEX_USER *d= get_current_user(thd, lex->definer);

    if (!d)
      return 1;

    definer_user= d->user;
    definer_host= d->host;
  }
  else
  {
    /* non-SUID trigger. */

    definer_user.str= 0;
    definer_user.length= 0;

    definer_host.str= 0;
    definer_host.length= 0;
  }

  stmt_query.append(STRING_WITH_LEN("CREATE "));

  append_definer(thd, &stmt_query, &definer_user, &definer_host);

  LEX_CSTRING stmt_definition;
  stmt_definition.str= (char*) thd->lex->stmt_definition_begin;
  stmt_definition.length= thd->lex->stmt_definition_end
    - thd->lex->stmt_definition_begin;
  trim_whitespace(thd->charset(), &stmt_definition);

  stmt_query.append(stmt_definition.str, stmt_definition.length);

  return wsrep_to_buf_helper(thd, stmt_query.c_ptr(), stmt_query.length(),
                             buf, buf_len);
}

void* start_wsrep_THD(void *arg)
{
  THD *thd;

  Wsrep_thd_args* thd_args= (Wsrep_thd_args*) arg;

  if (my_thread_init() || (!(thd= new THD(next_thread_id(), true))))
  {
    goto error;
  }

  statistic_increment(thread_created, &LOCK_status);

  if (wsrep_gtid_mode)
  {
    /* Adjust domain_id. */
    thd->variables.gtid_domain_id= wsrep_gtid_domain_id;
  }

  thd->real_id=pthread_self(); // Keep purify happy

  my_net_init(&thd->net,(st_vio*) 0, thd, MYF(0));

  DBUG_PRINT("wsrep",(("creating thread %lld"), (long long)thd->thread_id));
  thd->prior_thr_create_utime= thd->start_utime= microsecond_interval_timer();

  server_threads.insert(thd);

  /* from bootstrap()... */
  thd->bootstrap=1;
  thd->max_client_packet_length= thd->net.max_packet;
  thd->security_ctx->master_access= ~(ulong)0;

  /* from handle_one_connection... */
  pthread_detach_this_thread();

  mysql_thread_set_psi_id(thd->thread_id);
  thd->thr_create_utime=  microsecond_interval_timer();

// </5.1.17>
  /*
    handle_one_connection() is normally the only way a thread would
    start and would always be on the very high end of the stack ,
    therefore, the thread stack always starts at the address of the
    first local variable of handle_one_connection, which is thd. We
    need to know the start of the stack so that we could check for
    stack overruns.
  */
  DBUG_PRINT("wsrep", ("handle_one_connection called by thread %lld",
                       (long long)thd->thread_id));
  /* now that we've called my_thread_init(), it is safe to call DBUG_* */

  thd->thread_stack= (char*) &thd;
  wsrep_assign_from_threadvars(thd);
  if (wsrep_store_threadvars(thd))
  {
    close_connection(thd, ER_OUT_OF_RESOURCES);
    statistic_increment(aborted_connects,&LOCK_status);
    unlink_thd(thd);
    delete thd;
    delete thd_args;
    goto error;
  }

  thd->system_thread= SYSTEM_THREAD_SLAVE_SQL;
  thd->security_ctx->skip_grants();

  /* handle_one_connection() again... */
  thd->proc_info= 0;
  thd->set_command(COM_SLEEP);
  thd->init_for_queries();
  mysql_mutex_lock(&LOCK_wsrep_slave_threads);

  wsrep_running_threads++;

  switch (thd_args->thread_type()) {
    case WSREP_APPLIER_THREAD:
      wsrep_running_applier_threads++;
      break;
    case WSREP_ROLLBACKER_THREAD:
      wsrep_running_rollbacker_threads++;
      break;
    default:
      WSREP_ERROR("Incorrect wsrep thread type: %d", thd_args->thread_type());
      break;
  }

  mysql_cond_broadcast(&COND_wsrep_slave_threads);
  mysql_mutex_unlock(&LOCK_wsrep_slave_threads);

  WSREP_DEBUG("wsrep system thread %llu, %p starting",
              thd->thread_id, thd);
  thd_args->fun()(thd, static_cast<void *>(thd_args));

  WSREP_DEBUG("wsrep system thread: %llu, %p closing",
              thd->thread_id, thd);

  /* Wsrep may reset globals during thread context switches, store globals
     before cleanup. */
  wsrep_store_threadvars(thd);

  close_connection(thd, 0);

  mysql_mutex_lock(&LOCK_wsrep_slave_threads);
  DBUG_ASSERT(wsrep_running_threads > 0);
  wsrep_running_threads--;

  switch (thd_args->thread_type()) {
    case WSREP_APPLIER_THREAD:
      DBUG_ASSERT(wsrep_running_applier_threads > 0);
      wsrep_running_applier_threads--;
      break;
    case WSREP_ROLLBACKER_THREAD:
      DBUG_ASSERT(wsrep_running_rollbacker_threads > 0);
      wsrep_running_rollbacker_threads--;
      break;
    default:
      WSREP_ERROR("Incorrect wsrep thread type: %d", thd_args->thread_type());
      break;
  }

  delete thd_args;
  WSREP_DEBUG("wsrep running threads now: %lu", wsrep_running_threads);
  mysql_cond_broadcast(&COND_wsrep_slave_threads);
  mysql_mutex_unlock(&LOCK_wsrep_slave_threads);
  /*
    Note: We can't call THD destructor without crashing
    if plugins have not been initialized. However, in most of the
    cases this means that pre SE initialization SST failed and
    we are going to exit anyway.
  */
  if (plugins_are_initialized)
  {
    net_end(&thd->net);
    unlink_thd(thd);
  }
  else
  {
    /*
      TODO: lightweight cleanup to get rid of:
      'Error in my_thread_global_end(): 2 threads didn't exit'
      at server shutdown
    */
    server_threads.erase(thd);
  }

  delete thd;
  my_thread_end();
  return(NULL);

error:
  WSREP_ERROR("Failed to create/initialize system thread");

  /* Abort if its the first applier/rollbacker thread. */
  if (!mysqld_server_initialized)
    unireg_abort(1);
  else
    return NULL;
}

enum wsrep::streaming_context::fragment_unit wsrep_fragment_unit(ulong unit)
{
  switch (unit)
  {
  case WSREP_FRAG_BYTES: return wsrep::streaming_context::bytes;
  case WSREP_FRAG_ROWS: return wsrep::streaming_context::row;
  case WSREP_FRAG_STATEMENTS: return wsrep::streaming_context::statement;
  default:
    DBUG_ASSERT(0);
    return wsrep::streaming_context::bytes;
  }
}

/***** callbacks for wsrep service ************/

my_bool get_wsrep_recovery()
{
  return wsrep_recovery;
}

bool wsrep_consistency_check(THD *thd)
{
  return thd->wsrep_consistency_check == CONSISTENCY_CHECK_RUNNING;
}


/*
  Commit an empty transaction.

  If the transaction is real and the wsrep transaction is still active,
  the transaction did not generate any rows or keys and is committed
  as empty. Here the wsrep transaction is rolled back and after statement
  step is performed to leave the wsrep transaction in the state as it
  never existed.

  This should not be an inline functions as it requires a lot of stack space
  because of WSREP_DBUG() usage.  It's also not a function that is
  frequently called.
*/

void wsrep_commit_empty(THD* thd, bool all)
{
  DBUG_ENTER("wsrep_commit_empty");
  WSREP_DEBUG("wsrep_commit_empty(%llu)", thd->thread_id);
  if (wsrep_is_real(thd, all) &&
      wsrep_thd_is_local(thd) &&
      thd->wsrep_trx().active() &&
      thd->wsrep_trx().state() != wsrep::transaction::s_committed)
  {
    /* @todo CTAS with STATEMENT binlog format and empty result set
       seems to be committing empty. Figure out why and try to fix
       elsewhere. */
    DBUG_ASSERT(!wsrep_has_changes(thd) ||
                (thd->lex->sql_command == SQLCOM_CREATE_TABLE &&
                 !thd->is_current_stmt_binlog_format_row()));
    bool have_error= wsrep_current_error(thd);
    int ret= wsrep_before_rollback(thd, all) ||
      wsrep_after_rollback(thd, all) ||
      wsrep_after_statement(thd);
    /* The committing transaction was empty but it held some locks and
       got BF aborted. As there were no certified changes in the
       data, we ignore the deadlock error and rely on error reporting
       by storage engine/server. */
    if (!ret && !have_error && wsrep_current_error(thd))
    {
      DBUG_ASSERT(wsrep_current_error(thd) == wsrep::e_deadlock_error);
      thd->wsrep_cs().reset_error();
    }
    if (ret)
    {
      WSREP_DEBUG("wsrep_commit_empty failed: %d", wsrep_current_error(thd));
    }
  }
  DBUG_VOID_RETURN;
}