path: root/storage/ndb/src/ndbapi/TransporterFacade.hpp

/* Copyright (C) 2003 MySQL AB

   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; either version 2 of the License, or
   (at your option) any later version.

   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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */

#ifndef TransporterFacade_H
#define TransporterFacade_H

#include <kernel_types.h>
#include <ndb_limits.h>
#include <NdbThread.h>
#include <TransporterRegistry.hpp>
#include <NdbMutex.h>
#include "DictCache.hpp"
#include <BlockNumbers.h>
#include <mgmapi.h>

class ClusterMgr;
class ArbitMgr;
class IPCConfig;
struct ndb_mgm_configuration;
class ConfigRetriever;

class Ndb;
class NdbApiSignal;
class NdbWaiter;

typedef void (* ExecuteFunction)(void *, NdbApiSignal *, LinearSectionPtr ptr[3]);
typedef void (* NodeStatusFunction)(void *, Uint32, bool nodeAlive, bool nfComplete);

extern "C" {
  void* runSendRequest_C(void*);
  void* runReceiveResponse_C(void*);
  void atexit_stop_instance();
}

class TransporterFacade
{
public:
  /**
   * Max number of Ndb objects.  
   * (Ndb objects should not be shared by different threads.)
   */
  STATIC_CONST( MAX_NO_THREADS = 4711 );
  TransporterFacade();
  virtual ~TransporterFacade();
  bool init(Uint32, const ndb_mgm_configuration *);

  static TransporterFacade* instance();
  int start_instance(int, const ndb_mgm_configuration*);
  static void stop_instance();
  
  /**
   * Register this block for sending/receiving signals
   * @return BlockNumber or -1 for failure
   */
  int open(void* objRef, ExecuteFunction, NodeStatusFunction);
  
  // Close this block number
  int close(BlockNumber blockNumber, Uint64 trans_id);

  // Only sends to nodes which are alive
  int sendSignal(NdbApiSignal * signal, NodeId nodeId);
  int sendSignal(NdbApiSignal*, NodeId, 
		 LinearSectionPtr ptr[3], Uint32 secs);
  int sendFragmentedSignal(NdbApiSignal*, NodeId, 
			   LinearSectionPtr ptr[3], Uint32 secs);

  // Is node available for running transactions
  bool   get_node_alive(NodeId nodeId) const;
  bool   get_node_stopping(NodeId nodeId) const;
  bool   getIsDbNode(NodeId nodeId) const;
  bool   getIsNodeSendable(NodeId nodeId) const;
  Uint32 getNodeGrp(NodeId nodeId) const;
  Uint32 getNodeSequence(NodeId nodeId) const;

  // Is there space in sendBuffer to send messages
  bool   check_send_size(Uint32 node_id, Uint32 send_size);

  // My own processor id
  NodeId ownId() const;

  void connected();

  void doConnect(int NodeId);
  void reportConnected(int NodeId);
  void doDisconnect(int NodeId);
  void reportDisconnected(int NodeId);

  NodeId get_an_alive_node();
  void ReportNodeAlive(NodeId nodeId);
  void ReportNodeDead(NodeId nodeId);
  void ReportNodeFailureComplete(NodeId nodeId);
  
  void lock_mutex();
  void unlock_mutex();

  // Improving the API performance
  void forceSend(Uint32 block_number);
  void checkForceSend(Uint32 block_number);

  // Close this block number
  int close_local(BlockNumber blockNumber);

  // Scan batch configuration parameters
  Uint32 get_scan_batch_size();
  Uint32 get_batch_byte_size();
  Uint32 get_batch_size();

  TransporterRegistry* get_registry() { return theTransporterRegistry;};

/*
  When a thread has sent its signals and is ready to wait for reception
  of these it does normally always wait on a conditional mutex and
  the actual reception is handled by the receiver thread in the NDB API.
  With the below new methods and variables each thread has the possibility
  of becoming owner of the "right" to poll for signals. Effectually this
  means that the thread acts temporarily as a receiver thread.
  For the thread that succeeds in grabbing this "ownership" it will avoid
  a number of expensive calls to conditional mutex and even more expensive
  context switches to wake up.
  When an owner of the poll "right" has completed its own task it is likely
  that there are others still waiting. In this case we pick one of the
  threads as new owner of the poll "right". Since we want to switch owner
  as seldom as possible we always pick the last thread which is likely to
  be the last to complete its reception.
*/
  void external_poll(Uint32 wait_time);
  NdbWaiter* get_poll_owner(void) const { return poll_owner; }
  void set_poll_owner(NdbWaiter* new_owner) { poll_owner= new_owner; }
  Uint32 put_in_cond_wait_queue(NdbWaiter *aWaiter);
  void remove_from_cond_wait_queue(NdbWaiter *aWaiter);
  NdbWaiter* rem_last_from_cond_wait_queue();
  // heart beat received from a node (e.g. a signal came)
  void hb_received(NodeId n);

private:
  void init_cond_wait_queue();
  struct CondWaitQueueElement {
    NdbWaiter *cond_wait_object;
    Uint32 next_cond_wait;
    Uint32 prev_cond_wait;
  };
  NdbWaiter *poll_owner;
  CondWaitQueueElement cond_wait_array[MAX_NO_THREADS];
  Uint32 first_in_cond_wait;
  Uint32 first_free_cond_wait;
  Uint32 last_in_cond_wait;
  /* End poll owner stuff */
  /**
   * Send a signal unconditional of node status (used by ClusterMgr)
   */
  friend class ClusterMgr;
  friend class ArbitMgr;
  friend class MgmtSrvr;
  friend class SignalSender;
  friend class GrepPS;
  friend class ExtSender; ///< @todo Hack to be able to sendSignalUnCond
  friend class GrepSS;
  friend class Ndb;
  friend class Ndb_cluster_connection_impl;

  int sendSignalUnCond(NdbApiSignal *, NodeId nodeId);

  bool isConnected(NodeId aNodeId);
  void doStop();

  TransporterRegistry* theTransporterRegistry;
  SocketServer m_socket_server;
  int sendPerformedLastInterval;
  int theOwnId;

  NodeId theStartNodeId;

  ClusterMgr* theClusterMgr;
  ArbitMgr* theArbitMgr;
  
  // Improving the API response time
  int checkCounter;
  Uint32 currentSendLimit;
  
  void calculateSendLimit();

  // Scan batch configuration parameters
  Uint32 m_scan_batch_size;
  Uint32 m_batch_byte_size;
  Uint32 m_batch_size;

  // Declarations for the receive and send thread
  int  theStopReceive;

  void threadMainSend(void);
  NdbThread* theSendThread;
  void threadMainReceive(void);
  NdbThread* theReceiveThread;

  friend void* runSendRequest_C(void*);
  friend void* runReceiveResponse_C(void*);
  friend void atexit_stop_instance();

  /**
   * Block number handling
   */
private:

  struct ThreadData {
    STATIC_CONST( ACTIVE = (1 << 16) | 1 );
    STATIC_CONST( INACTIVE = (1 << 16) );
    STATIC_CONST( END_OF_LIST = MAX_NO_THREADS + 1 );
    
    ThreadData(Uint32 initialSize = 32);
    
    /**
     * Split "object" into 3 list
     *   This to improve locality
     *   when iterating over lists
     */
    struct Object_Execute {
      void * m_object;
      ExecuteFunction m_executeFunction;
    };
    struct NodeStatus_NextFree {
      NodeStatusFunction m_statusFunction;
    };

    Uint32 m_firstFree;
    Vector<Uint32> m_statusNext;
    Vector<Object_Execute> m_objectExecute;
    Vector<NodeStatusFunction> m_statusFunction;
    
    int open(void* objRef, ExecuteFunction, NodeStatusFunction);
    int close(int number);
    void expand(Uint32 size);

    inline Object_Execute get(Uint16 blockNo) const {
      blockNo -= MIN_API_BLOCK_NO;
      if(blockNo < m_objectExecute.size()){
	return m_objectExecute[blockNo];
      }
      Object_Execute oe = { 0, 0 };
      return oe;
    }

    /**
     * Is the block number used currently
     */
    inline bool getInUse(Uint16 index) const {
      return (m_statusNext[index] & (1 << 16)) != 0;
    }
  } m_threads;
  
  Uint32 m_max_trans_id;
  Uint32 m_fragmented_signal_id;

  /**
   * execute function
   */
  friend void execute(void * callbackObj, SignalHeader * const header, 
                      Uint8 prio, 
                      Uint32 * const theData, LinearSectionPtr ptr[3]);
  
public:
  NdbMutex* theMutexPtr;
private:
  static TransporterFacade* theFacadeInstance;

public:
  GlobalDictCache m_globalDictCache;
};

class PollGuard
{
  public:
  PollGuard(TransporterFacade *tp, NdbWaiter *aWaiter, Uint32 block_no);
  ~PollGuard() { unlock_and_signal(); }
  int wait_n_unlock(int wait_time, NodeId nodeId, Uint32 state,
                    bool forceSend= false);
  int wait_for_input_in_loop(int wait_time, bool forceSend);
  void wait_for_input(int wait_time);
  int wait_scan(int wait_time, NodeId nodeId, bool forceSend);
  void unlock_and_signal();
  private:
  TransporterFacade *m_tp;
  NdbWaiter *m_waiter;
  Uint32 m_block_no;
  bool m_locked;
};

inline
TransporterFacade*
TransporterFacade::instance()
{
  return theFacadeInstance;
}

inline
void 
TransporterFacade::lock_mutex()
{
  NdbMutex_Lock(theMutexPtr);
}

inline
void 
TransporterFacade::unlock_mutex()
{
  NdbMutex_Unlock(theMutexPtr);
}

#include "ClusterMgr.hpp"

inline
bool
TransporterFacade::check_send_size(Uint32 node_id, Uint32 send_size)
{
  return true;
}

inline
bool
TransporterFacade::getIsDbNode(NodeId n) const {
  return 
    theClusterMgr->getNodeInfo(n).defined && 
    theClusterMgr->getNodeInfo(n).m_info.m_type == NodeInfo::DB;
}

inline
Uint32
TransporterFacade::getNodeGrp(NodeId n) const {
  return theClusterMgr->getNodeInfo(n).m_state.nodeGroup;
}


inline
bool
TransporterFacade::get_node_alive(NodeId n) const {

  const ClusterMgr::Node & node = theClusterMgr->getNodeInfo(n);
  return node.m_alive;
}

inline
void
TransporterFacade::hb_received(NodeId n) {
  theClusterMgr->hb_received(n);
}

inline
bool
TransporterFacade::get_node_stopping(NodeId n) const {
  const ClusterMgr::Node & node = theClusterMgr->getNodeInfo(n);
  return ((node.m_state.startLevel == NodeState::SL_STOPPING_1) ||
          (node.m_state.startLevel == NodeState::SL_STOPPING_2));
}

inline
bool
TransporterFacade::getIsNodeSendable(NodeId n) const {
  const ClusterMgr::Node & node = theClusterMgr->getNodeInfo(n);
  const Uint32 startLevel = node.m_state.startLevel;

  if (node.m_info.m_type == NodeInfo::DB) {
    if(node.m_state.singleUserMode && 
       ownId() == node.m_state.singleUserApi) {
      return (node.compatible && 
              (node.m_state.startLevel == NodeState::SL_STOPPING_1 ||
               node.m_state.startLevel == NodeState::SL_STARTED ||
               node.m_state.startLevel == NodeState::SL_SINGLEUSER));
      }
      else
        return node.compatible && (startLevel == NodeState::SL_STARTED ||
                                 startLevel == NodeState::SL_STOPPING_1);
  } else if (node.m_info.m_type == NodeInfo::REP) {
    /**
     * @todo Check that REP node actually has received API_REG_REQ
     */
    return node.compatible;
  } else {
    ndbout_c("TransporterFacade::getIsNodeSendable: Illegal node type: "
             "%d of node: %d", 
             node.m_info.m_type, n);
    abort();
    return false; // to remove compiler warning
  }
}

inline
Uint32
TransporterFacade::getNodeSequence(NodeId n) const {
  return theClusterMgr->getNodeInfo(n).m_info.m_connectCount;
}

inline
Uint32
TransporterFacade::get_scan_batch_size() {
  return m_scan_batch_size;
}

inline
Uint32
TransporterFacade::get_batch_byte_size() {
  return m_batch_byte_size;
}

inline
Uint32
TransporterFacade::get_batch_size() {
  return m_batch_size;
}



#endif // TransporterFacade_H