Merge pchardin@bk-internal.mysql.com:/home/bk/mysql-5.1

into  mysql.com:/home/cps/mysql/devel/5.1-csv-remove-mmap

mysql-test/r/csv.result

@@ -4944,10 +4944,10 @@ val
 UPDATE  bug13894 SET val=6 WHERE val=10;
 SELECT * FROM bug13894;
 val
-5
-11
 6
 6
+5
+11
 DROP TABLE bug13894;
 DROP TABLE IF EXISTS bug14672;
 CREATE TABLE bug14672 (c1 integer) engine = CSV;

sql/ha_ndbcluster_binlog.cc

@@ -453,7 +453,7 @@ static void ndbcluster_binlog_wait(THD *thd)
 }
 
 /*
- Called from MYSQL_LOG::reset_logs in log.cc when binlog is emptied
+ Called from MYSQL_BIN_LOG::reset_logs in log.cc when binlog is emptied
 */
 static int ndbcluster_reset_logs(THD *thd)
 {
@@ -477,7 +477,7 @@ static int ndbcluster_reset_logs(THD *thd)
 }
 
 /*
-  Called from MYSQL_LOG::purge_logs in log.cc when the binlog "file"
+  Called from MYSQL_BIN_LOG::purge_logs in log.cc when the binlog "file"
   is removed
 */
 

sql/log.h

@@ -147,33 +147,85 @@ typedef struct st_log_info
 class Log_event;
 class Rows_log_event;
 
-enum enum_log_type { LOG_CLOSED, LOG_TO_BE_OPENED, LOG_NORMAL, LOG_NEW, LOG_BIN};
+enum enum_log_type { LOG_UNKNOWN, LOG_NORMAL, LOG_BIN };
+enum enum_log_state { LOG_OPENED, LOG_CLOSED, LOG_TO_BE_OPENED };
 
 /*
-  TODO split MYSQL_LOG into base MYSQL_LOG and
-  MYSQL_QUERY_LOG, MYSQL_SLOW_LOG, MYSQL_BIN_LOG
-  most of the code from MYSQL_LOG should be in the MYSQL_BIN_LOG
-  only (TC_LOG included)
-
   TODO use mmap instead of IO_CACHE for binlog
   (mmap+fsync is two times faster than write+fsync)
 */
 
-class MYSQL_LOG: public TC_LOG
+class MYSQL_LOG
+{
+public:
+  MYSQL_LOG();
+  void init_pthread_objects();
+  void cleanup();
+  bool open(const char *log_name,
+            enum_log_type log_type,
+            const char *new_name,
+            enum cache_type io_cache_type_arg);
+  void init(enum_log_type log_type_arg,
+            enum cache_type io_cache_type_arg);
+  void close(uint exiting);
+  inline bool is_open() { return log_state != LOG_CLOSED; }
+  const char *generate_name(const char *log_name, const char *suffix,
+                            bool strip_ext, char *buff);
+  int generate_new_name(char *new_name, const char *log_name);
+ protected:
+  /* LOCK_log is inited by init_pthread_objects() */
+  pthread_mutex_t LOCK_log;
+  char *name;
+  char log_file_name[FN_REFLEN];
+  char time_buff[20], db[NAME_LEN + 1];
+  bool write_error, inited;
+  IO_CACHE log_file;
+  enum_log_type log_type;
+  volatile enum_log_state log_state;
+  enum cache_type io_cache_type;
+  friend class Log_event;
+};
+
+class MYSQL_QUERY_LOG: public MYSQL_LOG
+{
+public:
+  MYSQL_QUERY_LOG() : last_time(0) {}
+  void reopen_file();
+  bool write(time_t event_time, const char *user_host,
+             uint user_host_len, int thread_id,
+             const char *command_type, uint command_type_len,
+             const char *sql_text, uint sql_text_len);
+  bool write(THD *thd, time_t current_time, time_t query_start_arg,
+             const char *user_host, uint user_host_len,
+             longlong query_time, longlong lock_time, bool is_command,
+             const char *sql_text, uint sql_text_len);
+  bool open_slow_log(const char *log_name)
+  {
+    char buf[FN_REFLEN];
+    return open(generate_name(log_name, "-slow.log", 0, buf), LOG_NORMAL, 0,
+                WRITE_CACHE);
+  }
+  bool open_query_log(const char *log_name)
+  {
+    char buf[FN_REFLEN];
+    return open(generate_name(log_name, ".log", 0, buf), LOG_NORMAL, 0,
+                WRITE_CACHE);
+  }
+private:
+  time_t last_time;
+};
+
+class MYSQL_BIN_LOG: public TC_LOG, private MYSQL_LOG
 {
  private:
   /* LOCK_log and LOCK_index are inited by init_pthread_objects() */
-  pthread_mutex_t LOCK_log, LOCK_index;
+  pthread_mutex_t LOCK_index;
   pthread_mutex_t LOCK_prep_xids;
   pthread_cond_t  COND_prep_xids;
   pthread_cond_t update_cond;
   ulonglong bytes_written;
-  time_t last_time,query_start;
-  IO_CACHE log_file;
   IO_CACHE index_file;
-  char *name;
-  char time_buff[20],db[NAME_LEN+1];
-  char log_file_name[FN_REFLEN],index_file_name[FN_REFLEN];
+  char index_file_name[FN_REFLEN];
   /*
      The max size before rotation (usable only if log_type == LOG_BIN: binary
      logs and relay logs).
@@ -186,13 +238,10 @@ class MYSQL_LOG: public TC_LOG
   */
   ulong max_size;
   ulong prepared_xids; /* for tc log - number of xids to remember */
-  volatile enum_log_type log_type;
-  enum cache_type io_cache_type;
   // current file sequence number for load data infile binary logging
   uint file_id;
   uint open_count;				// For replication
   int readers_count;
-  bool write_error, inited;
   bool need_start_event;
   /*
     no_auto_events means we don't want any of these automatic events :
@@ -202,13 +251,21 @@ class MYSQL_LOG: public TC_LOG
     In 5.0 it's 0 for relay logs too!
   */
   bool no_auto_events;
-  friend class Log_event;
 
   ulonglong m_table_map_version;
 
   int write_to_file(IO_CACHE *cache);
+  /*
+    This is used to start writing to a new log file. The difference from
+    new_file() is locking. new_file_without_locking() does not acquire
+    LOCK_log.
+  */
+  void new_file_without_locking();
+  void new_file_impl(bool need_lock);
 
 public:
+  MYSQL_LOG::generate_name;
+  MYSQL_LOG::is_open;
   /*
     These describe the log's format. This is used only for relay logs.
     _for_exec is used by the SQL thread, _for_queue by the I/O thread. It's
@@ -220,9 +277,9 @@ class MYSQL_LOG: public TC_LOG
   Format_description_log_event *description_event_for_exec,
     *description_event_for_queue;
 
-  MYSQL_LOG();
+  MYSQL_BIN_LOG();
   /*
-    note that there's no destructor ~MYSQL_LOG() !
+    note that there's no destructor ~MYSQL_BIN_LOG() !
     The reason is that we don't want it to be automatically called
     on exit() - but only during the correct shutdown process
   */
@@ -264,9 +321,7 @@ class MYSQL_LOG: public TC_LOG
   void signal_update();
   void wait_for_update(THD* thd, bool master_or_slave);
   void set_need_start_event() { need_start_event = 1; }
-  void init(enum_log_type log_type_arg,
-	    enum cache_type io_cache_type_arg,
-	    bool no_auto_events_arg, ulong max_size);
+  void init(bool no_auto_events_arg, ulong max_size);
   void init_pthread_objects();
   void cleanup();
   bool open(const char *log_name,
@@ -275,35 +330,10 @@ class MYSQL_LOG: public TC_LOG
 	    enum cache_type io_cache_type_arg,
 	    bool no_auto_events_arg, ulong max_size,
             bool null_created);
-  const char *generate_name(const char *log_name, const char *suffix,
-                            bool strip_ext, char *buff);
-  /* simplified open_xxx wrappers for the gigantic open above */
-  bool open_query_log(const char *log_name)
-  {
-    char buf[FN_REFLEN];
-    return open(generate_name(log_name, ".log", 0, buf),
-                LOG_NORMAL, 0, WRITE_CACHE, 0, 0, 0);
-  }
-  bool open_slow_log(const char *log_name)
-  {
-    char buf[FN_REFLEN];
-    return open(generate_name(log_name, "-slow.log", 0, buf),
-                LOG_NORMAL, 0, WRITE_CACHE, 0, 0, 0);
-  }
   bool open_index_file(const char *index_file_name_arg,
                        const char *log_name);
-  void new_file(bool need_lock);
-  /* log a command to the old-fashioned general log */
-  bool write(time_t event_time, const char *user_host,
-             uint user_host_len, int thread_id,
-             const char *command_type, uint command_type_len,
-             const char *sql_text, uint sql_text_len);
-
-  /* log a query to the old-fashioned slow query log */
-  bool write(THD *thd, time_t current_time, time_t query_start_arg,
-             const char *user_host, uint user_host_len,
-             longlong query_time, longlong lock_time, bool is_command,
-             const char *sql_text, uint sql_text_len);
+  /* Use this to start writing a new log file */
+  void new_file();
 
   bool write(Log_event* event_info); // binary log write
   bool write(THD *thd, IO_CACHE *cache, Log_event *commit_event);
@@ -319,7 +349,6 @@ class MYSQL_LOG: public TC_LOG
   bool appendv(const char* buf,uint len,...);
   bool append(Log_event* ev);
 
-  int generate_new_name(char *new_name,const char *old_name);
   void make_log_name(char* buf, const char* log_ident);
   bool is_active(const char* log_file_name);
   int update_log_index(LOG_INFO* linfo, bool need_update_threads);
@@ -339,7 +368,6 @@ class MYSQL_LOG: public TC_LOG
   int find_next_log(LOG_INFO* linfo, bool need_mutex);
   int get_current_log(LOG_INFO* linfo);
   uint next_file_id();
-  inline bool is_open() { return log_type != LOG_CLOSED; }
   inline char* get_index_fname() { return index_file_name;}
   inline char* get_log_fname() { return log_file_name; }
   inline char* get_name() { return name; }
@@ -416,7 +444,8 @@ class Log_to_csv_event_handler: public Log_event_handler
 
 class Log_to_file_event_handler: public Log_event_handler
 {
-  MYSQL_LOG mysql_log, mysql_slow_log;
+  MYSQL_QUERY_LOG mysql_log;
+  MYSQL_QUERY_LOG mysql_slow_log;
   bool is_initialized;
 public:
   Log_to_file_event_handler(): is_initialized(FALSE)

sql/log_event.h

@@ -470,7 +470,7 @@ enum Int_event_type
 
 #ifndef MYSQL_CLIENT
 class String;
-class MYSQL_LOG;
+class MYSQL_BIN_LOG;
 class THD;
 #endif
 

sql/mysql_priv.h

@@ -1537,7 +1537,7 @@ extern char *default_tz_name;
 extern my_bool opt_large_pages;
 extern uint opt_large_page_size;
 
-extern MYSQL_LOG mysql_bin_log;
+extern MYSQL_BIN_LOG mysql_bin_log;
 extern LOGGER logger;
 extern TABLE_LIST general_log, slow_log;
 extern FILE *bootstrap_file;

sql/mysqld.cc

@@ -2635,10 +2635,10 @@ static int init_common_variables(const char *conf_file_name, int argc,
   global_system_variables.time_zone= my_tz_SYSTEM;
   
   /*
-    Init mutexes for the global MYSQL_LOG objects.
+    Init mutexes for the global MYSQL_BIN_LOG objects.
     As safe_mutex depends on what MY_INIT() does, we can't init the mutexes of
-    global MYSQL_LOGs in their constructors, because then they would be inited
-    before MY_INIT(). So we do it here.
+    global MYSQL_BIN_LOGs in their constructors, because then they would be
+    inited before MY_INIT(). So we do it here.
   */
   mysql_bin_log.init_pthread_objects();
 

sql/rpl_injector.cc

@@ -25,7 +25,7 @@
 
 /* inline since it's called below */
 inline
-injector::transaction::transaction(MYSQL_LOG *log, THD *thd)
+injector::transaction::transaction(MYSQL_BIN_LOG *log, THD *thd)
   : m_state(START_STATE), m_thd(thd)
 {
   /* 

sql/rpl_injector.h

@@ -26,7 +26,7 @@
 
 /* Forward declarations */
 class handler;
-class MYSQL_LOG;
+class MYSQL_BIN_LOG;
 class st_table;
 
 typedef st_table TABLE;
@@ -219,7 +219,7 @@ class injector
 
     private:
       /* Only the injector may construct these object */
-      transaction(MYSQL_LOG *, THD *);
+      transaction(MYSQL_BIN_LOG *, THD *);
 
       void swap(transaction& o) {
         /* std::swap(m_start_pos, o.m_start_pos); */

sql/rpl_rli.h

@@ -69,7 +69,7 @@ typedef struct st_relay_log_info
     Protected with internal locks.
     Must get data_lock when resetting the logs.
   */
-  MYSQL_LOG relay_log;
+  MYSQL_BIN_LOG relay_log;
   LOG_INFO linfo;
   IO_CACHE cache_buf,*cur_log;
 

sql/slave.cc

@@ -4658,7 +4658,7 @@ static Log_event* next_event(RELAY_LOG_INFO* rli)
       When the relay log is created when the I/O thread starts, easy: the
       master will send the description event and we will queue it.
       But if the relay log is created by new_file(): then the solution is:
-      MYSQL_LOG::open() will write the buffered description event.
+      MYSQL_BIN_LOG::open() will write the buffered description event.
     */
     if ((ev=Log_event::read_log_event(cur_log,0,
                                       rli->relay_log.description_event_for_exec)))
@@ -4920,7 +4920,8 @@ event(errno: %d  cur_log->error: %d)",
   Rotate a relay log (this is used only by FLUSH LOGS; the automatic rotation
   because of size is simpler because when we do it we already have all relevant
   locks; here we don't, so this function is mainly taking locks). 
-  Returns nothing as we cannot catch any error (MYSQL_LOG::new_file() is void).
+  Returns nothing as we cannot catch any error (MYSQL_BIN_LOG::new_file()
+  is void).
 */
 
 void rotate_relay_log(MASTER_INFO* mi)
@@ -4942,7 +4943,7 @@ void rotate_relay_log(MASTER_INFO* mi)
   }
 
   /* If the relay log is closed, new_file() will do nothing. */
-  rli->relay_log.new_file(1);
+  rli->relay_log.new_file();
 
   /*
     We harvest now, because otherwise BIN_LOG_HEADER_SIZE will not immediately

sql/slave.h

@@ -73,7 +73,7 @@
   run_lock protects all information about the run state: slave_running, and the
   existence of the I/O thread (to stop/start it, you need this mutex).
   data_lock protects some moving members of the struct: counters (log name,
-  position) and relay log (MYSQL_LOG object).
+  position) and relay log (MYSQL_BIN_LOG object).
 
   In RELAY_LOG_INFO: run_lock, data_lock
   see MASTER_INFO
@@ -81,7 +81,7 @@
   Order of acquisition: if you want to have LOCK_active_mi and a run_lock, you
   must acquire LOCK_active_mi first.
 
-  In MYSQL_LOG: LOCK_log, LOCK_index of the binlog and the relay log
+  In MYSQL_BIN_LOG: LOCK_log, LOCK_index of the binlog and the relay log
   LOCK_log: when you write to it. LOCK_index: when you create/delete a binlog
   (so that you have to update the .index file).
 */

storage/csv/ha_tina.h

@@ -29,15 +29,12 @@
 typedef struct st_tina_share {
   char *table_name;
   char data_file_name[FN_REFLEN];
-  byte *mapped_file;                /* mapped region of file */
   uint table_name_length, use_count;
   /*
     Below flag is needed to make log tables work with concurrent insert.
     For more details see comment to ha_tina::update_status.
   */
   my_bool is_log_table;
-  MY_STAT file_stat;                /* Stat information for the data file */
-  File data_file;                   /* Current open data file */
   /*
     Here we save the length of the file for readers. This is updated by
     inserts, updates and deletes. The var is initialized along with the
@@ -46,7 +43,10 @@ typedef struct st_tina_share {
   off_t saved_data_file_length;
   pthread_mutex_t mutex;
   THR_LOCK lock;
+  bool update_file_opened;
+  bool tina_write_opened;
   File meta_file;           /* Meta file we use */
+  File tina_write_filedes;  /* File handler for readers */
   bool crashed;             /* Meta file is crashed */
   ha_rows rows_recorded;    /* Number of rows in tables */
 } TINA_SHARE;
@@ -56,6 +56,49 @@ struct tina_set {
   off_t end;
 };
 
+class Transparent_file
+{
+   File filedes;
+   byte *buff;  /* in-memory window to the file or mmaped area */
+   /* current window sizes */
+   off_t lower_bound;
+   off_t upper_bound;
+   uint buff_size;
+
+ public:
+
+   Transparent_file() : lower_bound(0), buff_size(IO_SIZE)
+   { buff= (byte *) my_malloc(buff_size*sizeof(byte),  MYF(MY_WME)); }
+
+   ~Transparent_file()
+   { my_free((gptr)buff, MYF(MY_ALLOW_ZERO_PTR)); }
+
+   void init_buff(File filedes_arg)
+   {
+     filedes= filedes_arg;
+     /* read the beginning of the file */
+     lower_bound= 0;
+     VOID(my_seek(filedes, 0, MY_SEEK_SET, MYF(0)));
+     if (filedes && buff)
+       upper_bound= my_read(filedes, buff, buff_size, MYF(0));
+   }
+
+   byte *ptr()
+   { return buff; }
+
+   off_t start()
+   { return lower_bound; }
+
+   off_t end()
+   { return upper_bound; }
+
+   /* get a char from the given position in the file */
+   char get_value (off_t offset);
+   /* shift a buffer windows to see the next part of the file */
+   off_t read_next();
+
+};
+
 class ha_tina: public handler
 {
   THR_LOCK_DATA lock;      /* MySQL lock */
@@ -64,6 +107,9 @@ class ha_tina: public handler
   off_t next_position;     /* Next position in the file scan */
   off_t local_saved_data_file_length; /* save position for reads */
   byte byte_buffer[IO_SIZE];
+  Transparent_file *file_buff;
+  File data_file;                   /* File handler for readers */
+  File update_temp_file;
   String buffer;
   /*
     The chain contains "holes" in the file, occured because of
@@ -77,12 +123,19 @@ class ha_tina: public handler
   uint32 chain_size;
   bool records_is_known;
 
+private:
+  bool get_write_pos(off_t *end_pos, tina_set *closest_hole);
+  int open_update_temp_file_if_needed();
+  int init_tina_writer();
+
 public:
   ha_tina(TABLE_SHARE *table_arg);
   ~ha_tina()
   {
     if (chain_alloced)
-      my_free((gptr)chain,0);
+      my_free((gptr)chain, 0);
+    if (file_buff)
+      delete file_buff;
   }
   const char *table_type() const { return "CSV"; }
   const char *index_type(uint inx) { return "NONE"; }