ndb_global_schema_lock.cc

/*
   Copyright (c) 2011, Oracle and/or its affiliates. All rights reserved.

   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 <mysql/plugin.h>
#include <ndbapi/NdbApi.hpp>
#include <portlib/NdbTick.h>
#include <my_sys.h>               // my_sleep.h

/* perform random sleep in the range milli_sleep to 2*milli_sleep */
static inline
void do_retry_sleep(unsigned milli_sleep)
{
  my_sleep(1000*(milli_sleep + 5*(rand()%(milli_sleep/5))));
}


#include "ndb_table_guard.h"

/*
  The lock/unlock functions use the BACKUP_SEQUENCE row in SYSTAB_0

  retry_time == 0 means no retry
  retry_time <  0 means infinite retries
  retry_time >  0 means retries for max 'retry_time' seconds
*/
static NdbTransaction *
gsl_lock_ext(THD *thd, Ndb *ndb, NdbError &ndb_error,
             int retry_time= 10)
{
  ndb->setDatabaseName("sys");
  ndb->setDatabaseSchemaName("def");
  NdbDictionary::Dictionary *dict= ndb->getDictionary();
  Ndb_table_guard ndbtab_g(dict, "SYSTAB_0");
  const NdbDictionary::Table *ndbtab= NULL;
  NdbOperation *op;
  NdbTransaction *trans= NULL;
  int retry_sleep= 50; /* 50 milliseconds, transaction */
  NDB_TICKS time_end;

  if (retry_time > 0)
  {
    time_end= NdbTick_CurrentMillisecond();
    time_end+= retry_time * 1000;
  }
  while (1)
  {
    if (!ndbtab)
    {
      if (!(ndbtab= ndbtab_g.get_table()))
      {
        if (dict->getNdbError().status == NdbError::TemporaryError)
          goto retry;
        ndb_error= dict->getNdbError();
        goto error_handler;
      }
    }

    trans= ndb->startTransaction();
    if (trans == NULL)
    {
      ndb_error= ndb->getNdbError();
      goto error_handler;
    }

    op= trans->getNdbOperation(ndbtab);
    op->readTuple(NdbOperation::LM_Exclusive);
    op->equal("SYSKEY_0", NDB_BACKUP_SEQUENCE);

    if (trans->execute(NdbTransaction::NoCommit) == 0)
      break;

    if (trans->getNdbError().status != NdbError::TemporaryError)
      goto error_handler;
    else if (thd_killed(thd))
      goto error_handler;
  retry:
    if (retry_time == 0)
      goto error_handler;
    if (retry_time > 0 &&
        time_end < NdbTick_CurrentMillisecond())
      goto error_handler;
    if (trans)
    {
      ndb->closeTransaction(trans);
      trans= NULL;
    }
    do_retry_sleep(retry_sleep);
  }
  return trans;

 error_handler:
  if (trans)
  {
    ndb_error= trans->getNdbError();
    ndb->closeTransaction(trans);
  }
  return NULL;
}


static bool
gsl_unlock_ext(Ndb *ndb, NdbTransaction *trans,
               NdbError &ndb_error)
{
  if (trans->execute(NdbTransaction::Commit))
  {
    ndb_error= trans->getNdbError();
    ndb->closeTransaction(trans);
    return false;
  }
  ndb->closeTransaction(trans);
  return true;
}

/*
  lock/unlock calls are reference counted, so calls to lock
  must be matched to a call to unlock even if the lock call fails
*/
static int gsl_is_locked_or_queued= 0;
static int gsl_no_locking_allowed= 0;
static pthread_mutex_t gsl_mutex;

/*
  Indicates if ndb_global_schema_lock module is active/initialized, normally
  turned on/off in ndbcluster_init/deinit with LOCK_plugin held.
*/
static bool gsl_initialized= false;

// NOTE! 'thd_proc_info' is defined in myql/plugin.h but not implemented, only
// a #define available in sql_class.h -> include sql_class.h until
// bug#11844974 has been fixed. 
#include <sql_class.h> 

class Thd_proc_info_guard
{
public:
  Thd_proc_info_guard(THD *thd)
   : m_thd(thd), m_proc_info(NULL) {}
  void set(const char* message)
  {
    const char* old= thd_proc_info(m_thd, message);
    if (!m_proc_info)
    {
      // Save the original on first change
      m_proc_info = old;
    }
  }
  ~Thd_proc_info_guard()
  {
    if (m_proc_info)
      thd_proc_info(m_thd, m_proc_info);
  }
private:
  THD *m_thd;
  const char *m_proc_info;
};


#include "ndb_thd.h"
#include "ndb_thd_ndb.h"


extern ulong opt_ndb_extra_logging;

static
int
ndbcluster_global_schema_lock(THD *thd, bool no_lock_queue,
                              bool report_cluster_disconnected)
{
  if (!gsl_initialized)
    return 0;

  Ndb *ndb= check_ndb_in_thd(thd);
  Thd_ndb *thd_ndb= get_thd_ndb(thd);
  NdbError ndb_error;
  if (thd_ndb->options & TNO_NO_LOCK_SCHEMA_OP)
    return 0;
  DBUG_ENTER("ndbcluster_global_schema_lock");
  DBUG_PRINT("enter", ("query: '%-.4096s', no_lock_queue: %d",
                       *thd_query(thd), no_lock_queue));
  if (thd_ndb->global_schema_lock_count)
  {
    if (thd_ndb->global_schema_lock_trans)
      thd_ndb->global_schema_lock_trans->refresh();
    else
      DBUG_ASSERT(thd_ndb->global_schema_lock_error != 0);
    thd_ndb->global_schema_lock_count++;
    DBUG_PRINT("exit", ("global_schema_lock_count: %d",
                        thd_ndb->global_schema_lock_count));
    DBUG_RETURN(0);
  }
  DBUG_ASSERT(thd_ndb->global_schema_lock_count == 0);
  thd_ndb->global_schema_lock_count= 1;
  thd_ndb->global_schema_lock_error= 0;
  DBUG_PRINT("exit", ("global_schema_lock_count: %d",
                      thd_ndb->global_schema_lock_count));

  /*
    Check that taking the lock is allowed
    - if not allowed to enter lock queue, return if lock exists
    - wait until allowed
    - increase global lock count
  */
  Thd_proc_info_guard proc_info(thd);
  pthread_mutex_lock(&gsl_mutex);
  /* increase global lock count */
  gsl_is_locked_or_queued++;
  if (no_lock_queue)
  {
    if (gsl_is_locked_or_queued != 1)
    {
      /* Other thread has lock and this thread may not enter lock queue */
      pthread_mutex_unlock(&gsl_mutex);
      thd_ndb->global_schema_lock_error= -1;
      DBUG_PRINT("exit", ("aborting as lock exists"));
      DBUG_RETURN(-1);
    }
    /* Mark that no other thread may be take lock */
    gsl_no_locking_allowed= 1;
  }
  else
  {
    while (gsl_no_locking_allowed)
    {
      proc_info.set("Waiting for allowed to take ndbcluster global schema lock");
      /* Wait until locking is allowed */
      pthread_mutex_unlock(&gsl_mutex);
      do_retry_sleep(50);
      if (thd_killed(thd))
      {
        thd_ndb->global_schema_lock_error= -1;
        DBUG_RETURN(-1);
      }
      pthread_mutex_lock(&gsl_mutex);
    }
  }
  pthread_mutex_unlock(&gsl_mutex);

  /*
    Take the lock
  */
  proc_info.set("Waiting for ndbcluster global schema lock");
  thd_ndb->global_schema_lock_trans= gsl_lock_ext(thd, ndb, ndb_error, -1);

  DBUG_EXECUTE_IF("sleep_after_global_schema_lock", my_sleep(6000000););

  if (no_lock_queue)
  {
    pthread_mutex_lock(&gsl_mutex);
    /* Mark that other thread may be take lock */
    gsl_no_locking_allowed= 0;
    pthread_mutex_unlock(&gsl_mutex);
  }

  if (thd_ndb->global_schema_lock_trans)
  {
    if (opt_ndb_extra_logging > 19)
    {
      sql_print_information("NDB: Global schema lock acquired");
    }

    // Count number of global schema locks taken by this thread
    thd_ndb->schema_locks_count++;
    DBUG_PRINT("info", ("schema_locks_count: %d",
                        thd_ndb->schema_locks_count));

    DBUG_RETURN(0);
  }

  if (ndb_error.code != 4009 || report_cluster_disconnected)
  {
    sql_print_warning("NDB: Could not acquire global schema lock (%d)%s",
                      ndb_error.code, ndb_error.message);
    push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN,
                        ER_GET_ERRMSG, ER_DEFAULT(ER_GET_ERRMSG),
                        ndb_error.code, ndb_error.message,
                        "NDB. Could not acquire global schema lock");
  }
  thd_ndb->global_schema_lock_error= ndb_error.code ? ndb_error.code : -1;
  DBUG_RETURN(-1);
}


static
int
ndbcluster_global_schema_unlock(THD *thd)
{
  if (!gsl_initialized)
    return 0;

  Thd_ndb *thd_ndb= get_thd_ndb(thd);
  DBUG_ASSERT(thd_ndb != 0);
  if (thd_ndb == 0 || (thd_ndb->options & TNO_NO_LOCK_SCHEMA_OP))
    return 0;
  Ndb *ndb= thd_ndb->ndb;
  DBUG_ENTER("ndbcluster_global_schema_unlock");
  NdbTransaction *trans= thd_ndb->global_schema_lock_trans;
  thd_ndb->global_schema_lock_count--;
  DBUG_PRINT("exit", ("global_schema_lock_count: %d",
                      thd_ndb->global_schema_lock_count));
  DBUG_ASSERT(ndb != NULL);
  if (ndb == NULL)
    return 0;
  DBUG_ASSERT(trans != NULL || thd_ndb->global_schema_lock_error != 0);
  if (thd_ndb->global_schema_lock_count != 0)
  {
    DBUG_RETURN(0);
  }
  thd_ndb->global_schema_lock_error= 0;

  /*
    Decrease global lock count
  */
  pthread_mutex_lock(&gsl_mutex);
  gsl_is_locked_or_queued--;
  pthread_mutex_unlock(&gsl_mutex);

  if (trans)
  {
    thd_ndb->global_schema_lock_trans= NULL;
    NdbError ndb_error;
    if (!gsl_unlock_ext(ndb, trans, ndb_error))
    {
      sql_print_warning("NDB: Releasing global schema lock (%d)%s",
                        ndb_error.code, ndb_error.message);
      push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN,
                          ER_GET_ERRMSG, ER_DEFAULT(ER_GET_ERRMSG),
                          ndb_error.code,
                          ndb_error.message,
                          "ndb. Releasing global schema lock");
      DBUG_RETURN(-1);
    }
    if (opt_ndb_extra_logging > 19)
    {
      sql_print_information("NDB: Global schema lock release");
    }
  }
  DBUG_RETURN(0);
}


#ifndef NDB_WITHOUT_GLOBAL_SCHEMA_LOCK
static
int
ndbcluster_global_schema_func(THD *thd, bool lock, void* args)
{
  if (lock)
  {
    bool no_lock_queue = (bool)args;
    return ndbcluster_global_schema_lock(thd, no_lock_queue, true);
  }

  return ndbcluster_global_schema_unlock(thd);
}
#endif


#include "ndb_global_schema_lock.h"

void ndbcluster_global_schema_lock_init(handlerton *hton)
{
  assert(gsl_initialized == false);
  assert(gsl_is_locked_or_queued == 0);
  assert(gsl_no_locking_allowed == 0);
  gsl_initialized= true;
  pthread_mutex_init(&gsl_mutex, MY_MUTEX_INIT_FAST);

#ifndef NDB_WITHOUT_GLOBAL_SCHEMA_LOCK
  hton->global_schema_func= ndbcluster_global_schema_func;
#endif
}


void ndbcluster_global_schema_lock_deinit(void)
{
  assert(gsl_initialized == true);
  assert(gsl_is_locked_or_queued == 0);
  assert(gsl_no_locking_allowed == 0);
  gsl_initialized= false;
  pthread_mutex_destroy(&gsl_mutex);
}


bool
Thd_ndb::has_required_global_schema_lock(const char* func)
{
#ifdef NDB_WITHOUT_GLOBAL_SCHEMA_LOCK
  // The global schema lock hook is not installed ->
  //  no thd has gsl
  return true;
#else
  if (global_schema_lock_error)
  {
    // An error occured while locking, either because
    // no connection to cluster or another user has locked
    // the lock -> ok, but caller should not allow to continue
    return false;
  }

  if (global_schema_lock_trans)
  {
    global_schema_lock_trans->refresh();
    return true; // All OK
  }

  // No attempt at taking global schema lock has been done, neither
  // error or trans set -> programming error
  LEX_STRING* query= thd_query_string(m_thd);
  sql_print_error("NDB: programming error, no lock taken while running "
                  "query '%*s' in function '%s'",
                  (int)query->length, query->str, func);
  abort();
  return false;
#endif
}


#include "ndb_global_schema_lock_guard.h"

Ndb_global_schema_lock_guard::Ndb_global_schema_lock_guard(THD *thd)
  : m_thd(thd), m_locked(false)
{
}


Ndb_global_schema_lock_guard::~Ndb_global_schema_lock_guard()
{
  if (m_locked)
    ndbcluster_global_schema_unlock(m_thd);
}


int Ndb_global_schema_lock_guard::lock(bool no_lock_queue,
                                       bool report_cluster_disconnected)
{
  /* only one lock call allowed */
  assert(!m_locked);

  /*
    Always set m_locked, even if lock fails. Since the
    lock/unlock calls are reference counted, the number
    of calls to lock and unlock need to match up.
  */
  m_locked= true;

  return ndbcluster_global_schema_lock(m_thd, no_lock_queue,
                                       report_cluster_disconnected);  
}