mariadb/sql/wsrep_client_service.cc

/* Copyright 2018-2023 Codership Oy <info@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.

   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 St, Fifth Floor, Boston, MA 02110-1301  USA */

#include "wsrep_client_service.h"
#include "wsrep_high_priority_service.h"
#include "wsrep_binlog.h"  /* wsrep_dump_rbr_buf() */
#include "wsrep_schema.h"  /* remove_fragments() */
#include "wsrep_thd.h"
#include "wsrep_xid.h"
#include "wsrep_trans_observer.h"
#include "wsrep_server_state.h"
#include "wsrep_mysqld.h"

#include "sql_base.h"    /* close_temporary_table() */
#include "sql_class.h"   /* THD */
#include "sql_parse.h"   /* stmt_causes_implicit_commit() */
#include "rpl_filter.h"  /* binlog_filter */
#include "rpl_rli.h"     /* Relay_log_info */
#include "slave.h"   /* opt_log_slave_updates */
#include "transaction.h" /* trans_commit()... */
#include "log.h"      /* stmt_has_updated_trans_table() */
#include "mysql/service_debug_sync.h"
#include "mysql/psi/mysql_thread.h" /* mysql_mutex_assert_owner() */

namespace
{

void debug_sync_caller(THD* thd, const char* sync_point)
{
#ifdef ENABLED_DEBUG_SYNC_OUT
  debug_sync_set_action(thd, sync_point, strlen(sync_point));
#endif
#ifdef ENABLED_DEBUG_SYNC
  if (debug_sync_service) debug_sync_service(thd,sync_point,strlen(sync_point));
#endif

}
}

Wsrep_client_service::Wsrep_client_service(THD* thd,
                                           Wsrep_client_state& client_state)
  : wsrep::client_service()
  , m_thd(thd)
  , m_client_state(client_state)
{ }

void Wsrep_client_service::store_globals()
{
  wsrep_store_threadvars(m_thd);
}

void Wsrep_client_service::reset_globals()
{
  wsrep_reset_threadvars(m_thd);
}

bool Wsrep_client_service::interrupted(
  wsrep::unique_lock<wsrep::mutex>& lock WSREP_UNUSED) const
{
  DBUG_ASSERT(m_thd == current_thd);
  mysql_mutex_assert_owner(static_cast<mysql_mutex_t*>(lock.mutex()->native()));
  bool ret= (m_thd->killed != NOT_KILLED);
  if (ret)
  {
    WSREP_DEBUG("wsrep state is interrupted, THD::killed %d trx state %d",
                m_thd->killed,  m_thd->wsrep_trx().state());
  }
  return ret;
}

int Wsrep_client_service::prepare_data_for_replication()
{
  DBUG_ASSERT(m_thd == current_thd);
  DBUG_ENTER("Wsrep_client_service::prepare_data_for_replication");
  size_t data_len= 0;
  IO_CACHE* transactional_cache= wsrep_get_cache(m_thd, true);
  IO_CACHE* stmt_cache= wsrep_get_cache(m_thd, false);

  if (transactional_cache || stmt_cache)
  {
    m_thd->binlog_flush_pending_rows_event(true);

    size_t transactional_data_len= 0;
    size_t stmt_data_len= 0;

    // Write transactional cache
    if (transactional_cache &&
        wsrep_write_cache(m_thd, transactional_cache, &transactional_data_len))
    {
      WSREP_ERROR("rbr write fail, data_len: %zu",
                  data_len);
      // wsrep_override_error(m_thd, ER_ERROR_DURING_COMMIT);
      DBUG_RETURN(1);
    }

    // Write stmt cache
    if (stmt_cache && wsrep_write_cache(m_thd, stmt_cache, &stmt_data_len))
    {
      WSREP_ERROR("rbr write fail, data_len: %zu",
                  data_len);
      // wsrep_override_error(m_thd, ER_ERROR_DURING_COMMIT);
      DBUG_RETURN(1);
    }

    // Complete data written from both caches
    data_len = transactional_data_len + stmt_data_len;
  }

  if (data_len == 0)
  {
    if (m_thd->get_stmt_da()->is_ok()              &&
        m_thd->get_stmt_da()->affected_rows() > 0  &&
        !binlog_filter->is_on() &&
        !m_thd->wsrep_trx().is_streaming())
    {
      WSREP_DEBUG("empty rbr buffer, query: %s, "
                  "affected rows: %llu, "
                  "changed tables: %d, "
                  "sql_log_bin: %d",
                  wsrep_thd_query(m_thd),
                  m_thd->get_stmt_da()->affected_rows(),
                  stmt_has_updated_trans_table(m_thd),
                  m_thd->variables.sql_log_bin);
    }
    else
    {
      WSREP_DEBUG("empty rbr buffer, query: %s", wsrep_thd_query(m_thd));
    }
  }
  DBUG_RETURN(0);
}


void Wsrep_client_service::cleanup_transaction()
{
  DBUG_ASSERT(m_thd == current_thd);
  if (WSREP_EMULATE_BINLOG(m_thd)) wsrep_thd_binlog_trx_reset(m_thd);
  m_thd->wsrep_affected_rows= 0;
}


int Wsrep_client_service::prepare_fragment_for_replication(
  wsrep::mutable_buffer& buffer, size_t& log_position)
{
  DBUG_ASSERT(m_thd == current_thd);
  THD* thd= m_thd;
  DBUG_ENTER("Wsrep_client_service::prepare_fragment_for_replication");
  IO_CACHE* cache= wsrep_get_cache(thd, true);
  thd->binlog_flush_pending_rows_event(true);

  if (!cache)
  {
    DBUG_RETURN(0);
  }

  const my_off_t saved_pos(my_b_tell(cache));
  if (reinit_io_cache(cache, READ_CACHE, thd->wsrep_sr().log_position(), 0, 0))
  {
    DBUG_RETURN(1);
  }

  int ret= 0;
  size_t total_length= 0;
  size_t length= my_b_bytes_in_cache(cache);

  if (!length)
  {
    length= my_b_fill(cache);
  }

  if (length > 0)
  {
    do
    {
      total_length+= length;
      if (total_length > wsrep_max_ws_size)
      {
        WSREP_WARN("transaction size limit (%lu) exceeded: %zu",
                   wsrep_max_ws_size, total_length);
        ret= 1;
        goto cleanup;
      }

      buffer.push_back(reinterpret_cast<const char*>(cache->read_pos),
                       reinterpret_cast<const char*>(cache->read_pos + length));
      cache->read_pos= cache->read_end;
    }
    while (cache->file >= 0 && (length= my_b_fill(cache)));
  }
  DBUG_ASSERT(total_length == buffer.size());
  log_position= saved_pos;
cleanup:
  if (reinit_io_cache(cache, WRITE_CACHE, saved_pos, 0, 0))
  {
    WSREP_WARN("Failed to reinitialize IO cache");
    ret= 1;
  }
  DBUG_RETURN(ret);
}

int Wsrep_client_service::remove_fragments()
{
  DBUG_ENTER("Wsrep_client_service::remove_fragments");
  DEBUG_SYNC(m_thd, "wsrep_before_fragment_removal");
  if (wsrep_schema->remove_fragments(m_thd,
                                     Wsrep_server_state::instance().id(),
                                     m_thd->wsrep_trx().id(),
                                     m_thd->wsrep_sr().fragments()))
  {
    WSREP_DEBUG("Failed to remove fragments from SR storage for transaction "
                "%llu, %llu",
                m_thd->thread_id, m_thd->wsrep_trx().id().get());
    DBUG_RETURN(1);
  }
  DBUG_RETURN(0);
}

bool Wsrep_client_service::statement_allowed_for_streaming() const
{
  /*
    Todo: Decide if implicit commit is allowed with streaming
    replication.
    !stmt_causes_implicit_commit(m_thd, CF_IMPLICIT_COMMIT_BEGIN);
  */
  return true;
}

size_t Wsrep_client_service::bytes_generated() const
{
  IO_CACHE* cache= wsrep_get_cache(m_thd, true);
  if (cache)
  {
    size_t pending_rows_event_length= 0;
    if (Rows_log_event* ev= m_thd->binlog_get_pending_rows_event(true))
    {
      pending_rows_event_length= ev->get_data_size();
    }
    return my_b_tell(cache) + pending_rows_event_length;
  }
  return 0;
}

void Wsrep_client_service::will_replay()
{
  DBUG_ASSERT(m_thd == current_thd);
  mysql_mutex_lock(&LOCK_wsrep_replaying);
  ++wsrep_replaying;
  mysql_mutex_unlock(&LOCK_wsrep_replaying);
}

void Wsrep_client_service::signal_replayed()
{
  DBUG_ASSERT(m_thd == current_thd);
  mysql_mutex_lock(&LOCK_wsrep_replaying);
  --wsrep_replaying;
  DBUG_ASSERT(wsrep_replaying >= 0);
  mysql_cond_broadcast(&COND_wsrep_replaying);
  mysql_mutex_unlock(&LOCK_wsrep_replaying);
}

enum wsrep::provider::status Wsrep_client_service::replay()
{

  DBUG_ASSERT(m_thd == current_thd);
  DBUG_ENTER("Wsrep_client_service::replay");

  /*
    Allocate separate THD for replaying to avoid tampering
    original THD state during replication event applying.
   */
  THD *replayer_thd= new THD(true, true);
  // Replace the security context of the replayer with the security context
  // of the original THD. Since security context class doesn't have proper
  // copy constructors, we need to store the original one and set it back
  // before destruction so that THD desctruction doesn't cause double-free
  // on the replaced security context.
  Security_context old_ctx = replayer_thd->main_security_ctx;
  replayer_thd->main_security_ctx = m_thd->main_security_ctx;
  replayer_thd->thread_stack= m_thd->thread_stack;
  replayer_thd->real_id= pthread_self();
  replayer_thd->prior_thr_create_utime=
      replayer_thd->start_utime= microsecond_interval_timer();
  replayer_thd->mark_connection_idle();
  replayer_thd->reset_for_next_command(true);

  enum wsrep::provider::status ret;
  {
    Wsrep_replayer_service replayer_service(replayer_thd, m_thd);
    wsrep::provider& provider(replayer_thd->wsrep_cs().provider());
    ret= provider.replay(replayer_thd->wsrep_trx().ws_handle(),
                         &replayer_service);
    replayer_service.replay_status(ret);
  }

  // In Galera we allow only InnoDB sequences, thus
  // sequence table updates are in writeset.
  // Binlog cache needs reset so that binlog_close
  // does not write cache to binlog file yet.
  binlog_reset_cache(m_thd);

  replayer_thd->main_security_ctx = old_ctx;
  delete replayer_thd;
  DBUG_RETURN(ret);
}

enum wsrep::provider::status Wsrep_client_service::replay_unordered()
{
  DBUG_ASSERT(0);
  return wsrep::provider::error_not_implemented;
}

void Wsrep_client_service::wait_for_replayers(wsrep::unique_lock<wsrep::mutex>& lock)
{
  DBUG_ASSERT(m_thd == current_thd);
  lock.unlock();
  mysql_mutex_lock(&LOCK_wsrep_replaying);
  /* We need to check if the THD is BF aborted during condition wait.
     Because the aborter does not know which condition this thread is waiting,
     use timed wait and check if the THD is BF aborted in the loop. */
  while (wsrep_replaying > 0 && !wsrep_is_bf_aborted(m_thd))
  {
    struct timespec wait_time;
    set_timespec_nsec(wait_time, 10000000L);
    mysql_cond_timedwait(&COND_wsrep_replaying, &LOCK_wsrep_replaying,
                         &wait_time);
  }
  mysql_mutex_unlock(&LOCK_wsrep_replaying);
  lock.lock();
}

enum wsrep::provider::status Wsrep_client_service::commit_by_xid()
{
  DBUG_ASSERT(0);
  return wsrep::provider::error_not_implemented;
}

void Wsrep_client_service::debug_sync(const char* sync_point)
{
  DBUG_ASSERT(m_thd == current_thd);
  debug_sync_caller(m_thd, sync_point);
}

void Wsrep_client_service::debug_crash(const char* crash_point)
{
  // DBUG_ASSERT(m_thd == current_thd);
  DBUG_EXECUTE_IF(crash_point, DBUG_SUICIDE(); );
}

int Wsrep_client_service::bf_rollback()
{
  DBUG_ASSERT(m_thd == current_thd);
  DBUG_ENTER("Wsrep_client_service::bf_rollback");

  int ret= (trans_rollback_stmt(m_thd) || trans_rollback(m_thd));

  WSREP_DEBUG("::bf_rollback() thread: %lu, client_state %s "
              "client_mode %s trans_state %s killed %d",
              thd_get_thread_id(m_thd),
              wsrep_thd_client_state_str(m_thd),
              wsrep_thd_client_mode_str(m_thd),
              wsrep_thd_transaction_state_str(m_thd),
              m_thd->killed);

  /* If client is quiting all below will be done in THD::cleanup()
     TODO: why we need this any other case?  */
  if (m_thd->wsrep_cs().state() != wsrep::client_state::s_quitting)
  {
    if (m_thd->locked_tables_mode && m_thd->lock)
    {
      if (m_thd->locked_tables_list.unlock_locked_tables(m_thd))
        ret= 1;
      m_thd->variables.option_bits&= ~OPTION_TABLE_LOCK;
    }
    if (m_thd->global_read_lock.is_acquired())
    {
      m_thd->global_read_lock.unlock_global_read_lock(m_thd);
    }
    m_thd->release_transactional_locks();
  }

  DBUG_RETURN(ret);
}