WL#3244 "CSV engine: convert mmap to read/write calls"

mysql-test/r/csv.result:
  Correct result file
storage/csv/ha_tina.cc:
   Get rid of mmap in tina. Use usual reads/writes
   to access the file.
storage/csv/ha_tina.h:
    Add Transparent_file class, which hides details of access
    to the file. Perform a cleanup & add new functions.

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;

storage/csv/ha_tina.cc

@@ -61,7 +61,7 @@
 
 /* The file extension */
 #define CSV_EXT ".CSV"               // The data file
-#define CSN_EXT ".CSN"               // Files used during repair
+#define CSN_EXT ".CSN"               // Files used during repair and update
 #define CSM_EXT ".CSM"               // Meta file
 
 
@@ -114,9 +114,59 @@ handlerton tina_hton= {
   NULL,    /* Fill FILES Table */
   HTON_CAN_RECREATE,
   NULL, /* binlog_func */
-  NULL /* binlog_log_query */
+  NULL, /* binlog_log_query */
+  NULL  /* release_temporary_latches */
 };
 
+
+off_t Transparent_file::read_next()
+{
+  off_t bytes_read;
+
+  /*
+     No need to seek here, as the file managed by Transparent_file class
+     always points to upper_bound byte
+  */
+  if ((bytes_read= my_read(filedes, buff, buff_size, MYF(0))) == MY_FILE_ERROR)
+    return -1;
+
+  /* end of file */
+  if (!bytes_read)
+    return -1;
+
+  lower_bound= upper_bound;
+  upper_bound+= bytes_read;
+
+  return lower_bound;
+}
+
+
+char Transparent_file::get_value(off_t offset)
+{
+  off_t bytes_read;
+
+  /* check boundaries */
+  if ((lower_bound <= offset) && (offset < upper_bound))
+    return buff[offset - lower_bound];
+  else
+  {
+    VOID(my_seek(filedes, offset, MY_SEEK_SET, MYF(0)));
+    /* read appropriate portion of the file */
+    if ((bytes_read= my_read(filedes, buff, buff_size,
+                             MYF(0))) == MY_FILE_ERROR)
+      return 0;
+
+    lower_bound= offset;
+    upper_bound= lower_bound + bytes_read;
+
+    /* end of file */
+    if (upper_bound == offset)
+      return 0;
+
+    return buff[0];
+  }
+}
+
 /*****************************************************************************
  ** TINA tables
  *****************************************************************************/
@@ -130,7 +180,7 @@ int sort_set (tina_set *a, tina_set *b)
     We assume that intervals do not intersect. So, it is enought to compare
     any two points. Here we take start of intervals for comparison.
   */
-  return ( a->begin > b->begin ? -1 : ( a->begin < b->begin ? 1 : 0 ) );
+  return ( a->begin > b->begin ? 1 : ( a->begin < b->begin ? -1 : 0 ) );
 }
 
 static byte* tina_get_key(TINA_SHARE *share,uint *length,
@@ -140,48 +190,6 @@ static byte* tina_get_key(TINA_SHARE *share,uint *length,
   return (byte*) share->table_name;
 }
 
-/*
-  Reloads the mmap file.
-*/
-int get_mmap(TINA_SHARE *share, int write)
-{
-  DBUG_ENTER("ha_tina::get_mmap");
-  if (share->mapped_file && my_munmap(share->mapped_file,
-                                      share->file_stat.st_size))
-    DBUG_RETURN(1);
-
-  if (my_fstat(share->data_file, &share->file_stat, MYF(MY_WME)) == -1)
-    DBUG_RETURN(1);
-
-  if (share->file_stat.st_size)
-  {
-    if (write)
-      share->mapped_file= (byte *)my_mmap(NULL, share->file_stat.st_size,
-                                          PROT_READ|PROT_WRITE, MAP_SHARED,
-                                          share->data_file, 0);
-    else
-      share->mapped_file= (byte *)my_mmap(NULL, share->file_stat.st_size,
-                                          PROT_READ, MAP_PRIVATE,
-                                          share->data_file, 0);
-    if ((share->mapped_file == MAP_FAILED))
-    {
-      /*
-        Bad idea you think? See the problem is that nothing actually checks
-        the return value of ::rnd_init(), so tossing an error is about
-        it for us.
-        Never going to happen right? :)
-      */
-      my_message(errno, "Woops, blew up opening a mapped file", 0);
-      DBUG_ASSERT(0);
-      DBUG_RETURN(1);
-    }
-  }
-  else
-    share->mapped_file= NULL;
-
-  DBUG_RETURN(0);
-}
-
 
 static int tina_init_func()
 {
@@ -218,6 +226,7 @@ static TINA_SHARE *get_share(const char *table_name, TABLE *table)
 {
   TINA_SHARE *share;
   char meta_file_name[FN_REFLEN];
+  MY_STAT file_stat;                /* Stat information for the data file */
   char *tmp_name;
   uint length;
 
@@ -250,6 +259,8 @@ static TINA_SHARE *get_share(const char *table_name, TABLE *table)
     share->table_name= tmp_name;
     share->crashed= FALSE;
     share->rows_recorded= 0;
+    share->update_file_opened= FALSE;
+    share->tina_write_opened= FALSE;
     strmov(share->table_name, table_name);
     fn_format(share->data_file_name, table_name, "", CSV_EXT,
               MY_REPLACE_EXT|MY_UNPACK_FILENAME);
@@ -277,27 +288,16 @@ static TINA_SHARE *get_share(const char *table_name, TABLE *table)
     */
     if (read_meta_file(share->meta_file, &share->rows_recorded))
       share->crashed= TRUE;
-    else
-      (void)write_meta_file(share->meta_file, share->rows_recorded, TRUE);
 
-    if ((share->data_file= my_open(share->data_file_name, O_RDWR|O_APPEND,
-                                   MYF(0))) == -1)
+    if (my_stat(share->data_file_name, &file_stat, MYF(MY_WME)) == NULL)
       goto error2;
-
-    share->mapped_file= NULL; // We don't know the state as we just allocated it
-    if (get_mmap(share, 0) > 0)
-      goto error3;
-
-    /* init file length value used by readers */
-    share->saved_data_file_length= share->file_stat.st_size;
+    share->saved_data_file_length= file_stat.st_size;
   }
   share->use_count++;
   pthread_mutex_unlock(&tina_mutex);
 
   return share;
 
-error3:
-  my_close(share->data_file,MYF(0));
 error2:
   thr_lock_delete(&share->lock);
   pthread_mutex_destroy(&share->mutex);
@@ -425,6 +425,30 @@ bool ha_tina::check_and_repair(THD *thd)
 }
 
 
+int ha_tina::init_tina_writer()
+{
+  DBUG_ENTER("ha_tina::init_tina_writer");
+
+  /*
+    Mark the file as crashed. We will set the flag back when we close
+    the file. In the case of the crash it will remain marked crashed,
+    which enforce recovery.
+  */
+  (void)write_meta_file(share->meta_file, share->rows_recorded, TRUE);
+
+  if ((share->tina_write_filedes=
+        my_open(share->data_file_name, O_RDWR|O_APPEND, MYF(0))) == -1)
+  {
+    DBUG_PRINT("info", ("Could not open tina file writes"));
+    share->crashed= TRUE;
+    DBUG_RETURN(1);
+  }
+  share->tina_write_opened= TRUE;
+
+  DBUG_RETURN(0);
+}
+
+
 bool ha_tina::is_crashed() const
 {
   DBUG_ENTER("ha_tina::is_crashed");
@@ -445,10 +469,13 @@ static int free_share(TINA_SHARE *share)
                           share->crashed ? TRUE :FALSE);
     if (my_close(share->meta_file, MYF(0)))
       result_code= 1;
-    if (share->mapped_file)
-      my_munmap(share->mapped_file, share->file_stat.st_size);
-    share->mapped_file= NULL;
-    result_code= my_close(share->data_file,MYF(0));
+    if (share->tina_write_opened)
+    {
+      if (my_close(share->tina_write_filedes, MYF(0)))
+        result_code= 1;
+      share->tina_write_opened= FALSE;
+    }
+
     hash_delete(&tina_open_tables, (byte*) share);
     thr_lock_delete(&share->lock);
     pthread_mutex_destroy(&share->mutex);
@@ -468,16 +495,16 @@ int tina_end(ha_panic_function type)
   Finds the end of a line.
   Currently only supports files written on a UNIX OS.
 */
-byte * find_eoln(byte *data, off_t begin, off_t end)
+
+off_t find_eoln_buff(Transparent_file *data_buff, off_t begin, off_t end)
 {
   for (off_t x= begin; x < end; x++)
-    if (data[x] == '\n')
-      return data + x;
+    if (data_buff->get_value(x) == '\n')
+      return x;
 
   return 0;
 }
 
-
 static handler *tina_create_handler(TABLE_SHARE *table)
 {
   return new ha_tina(table);
@@ -491,7 +518,7 @@ ha_tina::ha_tina(TABLE_SHARE *table_arg)
     They are not probably completely right.
   */
   current_position(0), next_position(0), local_saved_data_file_length(0),
-  chain_alloced(0), chain_size(DEFAULT_CHAIN_LENGTH),
+  file_buff(0), chain_alloced(0), chain_size(DEFAULT_CHAIN_LENGTH),
   records_is_known(0)
 {
   /* Set our original buffers from pre-allocated memory */
@@ -619,50 +646,50 @@ int ha_tina::chain_append()
 */
 int ha_tina::find_current_row(byte *buf)
 {
-  byte *mapped_ptr;
-  byte *end_ptr;
+  off_t end_offset, curr_offset= current_position;
   DBUG_ENTER("ha_tina::find_current_row");
 
-  mapped_ptr= (byte *)share->mapped_file + current_position;
-
   /*
     We do not read further then local_saved_data_file_length in order
     not to conflict with undergoing concurrent insert.
   */
-  if ((end_ptr=  find_eoln(share->mapped_file, current_position,
+  if ((end_offset= find_eoln_buff(file_buff, current_position,
                                   local_saved_data_file_length)) == 0)
     DBUG_RETURN(HA_ERR_END_OF_FILE);
 
   for (Field **field=table->field ; *field ; field++)
   {
     buffer.length(0);
-    if (*mapped_ptr == '"')
-      mapped_ptr++; // Increment past the first quote
+    if (file_buff->get_value(curr_offset) == '"')
+      curr_offset++; // Incrementpast the first quote
     else
       DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);
-    for(;mapped_ptr != end_ptr; mapped_ptr++)
+    for(;curr_offset != end_offset; curr_offset++)
     {
       // Need to convert line feeds!
-      if (*mapped_ptr == '"' &&
-          (((mapped_ptr[1] == ',') && (mapped_ptr[2] == '"')) ||
-           (mapped_ptr == end_ptr -1 )))
+      if (file_buff->get_value(curr_offset) == '"' &&
+          (((file_buff->get_value(curr_offset + 1) == ',') &&
+            (file_buff->get_value(curr_offset + 2) == '"')) ||
+           (curr_offset == end_offset -1 )))
       {
-        mapped_ptr += 2; // Move past the , and the "
+        curr_offset+= 2; // Move past the , and the "
         break;
       }
-      if (*mapped_ptr == '\\' && mapped_ptr != (end_ptr - 1))
+      if (file_buff->get_value(curr_offset) == '\\' &&
+          curr_offset != (end_offset - 1))
       {
-        mapped_ptr++;
-        if (*mapped_ptr == 'r')
+        curr_offset++;
+        if (file_buff->get_value(curr_offset) == 'r')
           buffer.append('\r');
-        else if (*mapped_ptr == 'n' )
+        else if (file_buff->get_value(curr_offset) == 'n' )
           buffer.append('\n');
-        else if ((*mapped_ptr == '\\') || (*mapped_ptr == '"'))
-          buffer.append(*mapped_ptr);
+        else if ((file_buff->get_value(curr_offset) == '\\') ||
+                 (file_buff->get_value(curr_offset) == '"'))
+          buffer.append(file_buff->get_value(curr_offset));
         else  /* This could only happed with an externally created file */
         {
           buffer.append('\\');
-          buffer.append(*mapped_ptr);
+          buffer.append(file_buff->get_value(curr_offset));
         }
       }
       else // ordinary symbol
@@ -671,14 +698,14 @@ int ha_tina::find_current_row(byte *buf)
           We are at final symbol and no last quote was found =>
           we are working with a damaged file.
         */
-        if (mapped_ptr == end_ptr -1)
+        if (curr_offset == end_offset - 1)
           DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);
-        buffer.append(*mapped_ptr);
+        buffer.append(file_buff->get_value(curr_offset));
       }
     }
     (*field)->store(buffer.ptr(), buffer.length(), system_charset_info);
   }
-  next_position= (end_ptr - share->mapped_file)+1;
+  next_position= end_offset + 1;
   /* Maybe use \N for null? */
   memset(buf, 0, table->s->null_bytes); /* We do not implement nulls! */
 
@@ -776,7 +803,7 @@ void ha_tina::get_status()
 void ha_tina::update_status()
 {
   /* correct local_saved_data_file_length for writers */
-  share->saved_data_file_length= share->file_stat.st_size;
+  share->saved_data_file_length= local_saved_data_file_length;
 }
 
 
@@ -827,6 +854,11 @@ int ha_tina::open(const char *name, int mode, uint open_options)
     DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);
   }
 
+  if ((data_file= my_open(share->data_file_name, O_RDONLY, MYF(0))) == -1)
+    DBUG_RETURN(0);
+
+  file_buff= new Transparent_file(data_file);
+
   /*
     Init locking. Pass handler object to the locking routines,
     so that they could save/update local_saved_data_file_length value
@@ -845,12 +877,14 @@ int ha_tina::open(const char *name, int mode, uint open_options)
 
 /*
   Close a database file. We remove ourselves from the shared strucutre.
-  If it is empty we destroy it and free the mapped file.
+  If it is empty we destroy it.
 */
 int ha_tina::close(void)
 {
+  int rc= 0;
   DBUG_ENTER("ha_tina::close");
-  DBUG_RETURN(free_share(share));
+  rc= my_close(data_file, MYF(0));
+  DBUG_RETURN(free_share(share) || rc);
 }
 
 /*
@@ -873,22 +907,17 @@ int ha_tina::write_row(byte * buf)
 
   size= encode_quote(buf);
 
-  if (my_write(share->data_file, (byte*)buffer.ptr(), size,
-               MYF(MY_WME | MY_NABP)))
+  if (!share->tina_write_opened)
+    if (init_tina_writer())
       DBUG_RETURN(-1);
 
-  /*
-    Ok, this is means that we will be doing potentially bad things
-    during a bulk insert on some OS'es. What we need is a cleanup
-    call for ::write_row that would let us fix up everything after the bulk
-    insert. The archive handler does this with an extra mutx call, which
-    might be a solution for this.
-  */
-  if (get_mmap(share, 0) > 0)
+   /* use pwrite, as concurrent reader could have changed the position */
+  if (my_write(share->tina_write_filedes, buffer.ptr(), size,
+               MYF(MY_WME | MY_NABP)))
     DBUG_RETURN(-1);
 
   /* update local copy of the max position to see our own changes */
-  local_saved_data_file_length= share->file_stat.st_size;
+  local_saved_data_file_length+= size;
 
   /* update shared info */
   pthread_mutex_lock(&share->mutex);
@@ -903,6 +932,23 @@ int ha_tina::write_row(byte * buf)
 }
 
 
+int ha_tina::open_update_temp_file_if_needed()
+{
+  char updated_fname[FN_REFLEN];
+
+  if (!share->update_file_opened)
+  {
+    if ((update_temp_file=
+           my_create(fn_format(updated_fname, share->table_name,
+                               "", CSN_EXT,
+                               MY_REPLACE_EXT | MY_UNPACK_FILENAME),
+                     0, O_RDWR | O_TRUNC, MYF(MY_WME))) < 0)
+      return 1;
+    share->update_file_opened= TRUE;
+  }
+  return 0;
+}
+
 /*
   This is called for an update.
   Make sure you put in code to increment the auto increment, also
@@ -926,16 +972,16 @@ int ha_tina::update_row(const byte * old_data, byte * new_data)
   if (chain_append())
     DBUG_RETURN(-1);
 
-  if (my_write(share->data_file, (byte*)buffer.ptr(), size,
+  if (open_update_temp_file_if_needed())
+    DBUG_RETURN(-1);
+
+  if (my_write(update_temp_file, (byte*)buffer.ptr(), size,
                MYF(MY_WME | MY_NABP)))
     DBUG_RETURN(-1);
 
   /* UPDATE should never happen on the log tables */
   DBUG_ASSERT(!share->is_log_table);
 
-  /* update local copy of the max position to see our own changes */
-  local_saved_data_file_length= share->file_stat.st_size;
-
   DBUG_RETURN(0);
 }
 
@@ -1001,6 +1047,8 @@ int ha_tina::rnd_init(bool scan)
 {
   DBUG_ENTER("ha_tina::rnd_init");
 
+  /* set buffer to the beginning of the file */
+  file_buff->init_buff(data_file);
   if (share->crashed)
       DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);
 
@@ -1008,11 +1056,6 @@ int ha_tina::rnd_init(bool scan)
   records= 0;
   records_is_known= 0;
   chain_ptr= chain;
-#ifdef HAVE_MADVISE
-  if (scan)
-    (void) madvise(share->mapped_file, share->file_stat.st_size,
-                   MADV_SEQUENTIAL);
-#endif
 
   DBUG_RETURN(0);
 }
@@ -1042,8 +1085,11 @@ int ha_tina::rnd_next(byte *buf)
   ha_statistic_increment(&SSV::ha_read_rnd_next_count);
 
   current_position= next_position;
-  if (!share->mapped_file)
+
+  /* don't scan an empty file */
+  if (!local_saved_data_file_length)
     DBUG_RETURN(HA_ERR_END_OF_FILE);
+
   if ((rc= find_current_row(buf)))
     DBUG_RETURN(rc);
 
@@ -1112,6 +1158,22 @@ int ha_tina::extra(enum ha_extra_function operation)
   DBUG_RETURN(0);
 }
 
+/*
+  Set end_pos to the last valid byte of continuous area, closest
+  to the given "hole", stored in the buffer. "Valid" here means,
+  not listed in the chain of deleted records ("holes").
+*/
+bool ha_tina::get_write_pos(off_t *end_pos, tina_set *closest_hole)
+{
+  if (closest_hole == chain_ptr) /* no more chains */
+    *end_pos= file_buff->end();
+  else
+    *end_pos= min(file_buff->end(),
+                  closest_hole->begin);
+  return (closest_hole != chain_ptr) && (*end_pos == closest_hole->begin);
+}
+
+
 /*
   Called after each table scan. In particular after deletes,
   and updates. In the last case we employ chain of deleted
@@ -1120,53 +1182,99 @@ int ha_tina::extra(enum ha_extra_function operation)
 */
 int ha_tina::rnd_end()
 {
+  char updated_fname[FN_REFLEN];
+  off_t file_buffer_start= 0;
   DBUG_ENTER("ha_tina::rnd_end");
 
   records_is_known= 1;
 
-  /* First position will be truncate position, second will be increment */
   if ((chain_ptr - chain)  > 0)
   {
-    tina_set *ptr;
-    size_t length;
+    tina_set *ptr= chain;
 
     /*
-      Setting up writable map, this will contain all of the data after the
-      get_mmap call that we have added to the file.
+      Re-read the beginning of a file (as the buffer should point to the
+      end of file after the scan).
     */
-    if (get_mmap(share, 1) > 0)
-      DBUG_RETURN(-1);
-    length= share->file_stat.st_size;
+    file_buff->init_buff(data_file);
 
     /*
-      The sort handles updates/deletes with random orders.
-      It also sorts so that we move the final blocks to the
-      beginning so that we move the smallest amount of data possible.
+      The sort is needed when there were updates/deletes with random orders.
+      It sorts so that we move the firts blocks to the beginning.
     */
     qsort(chain, (size_t)(chain_ptr - chain), sizeof(tina_set),
           (qsort_cmp)sort_set);
-    for (ptr= chain; ptr < chain_ptr; ptr++)
+
+    off_t write_begin= 0, write_end;
+
+    /* create the file to write updated table if it wasn't yet created */
+    if (open_update_temp_file_if_needed())
+      DBUG_RETURN(-1);
+
+    /* write the file with updated info */
+    while ((file_buffer_start != -1))     // while not end of file
     {
-      memmove(share->mapped_file + ptr->begin, share->mapped_file + ptr->end,
-              length - (size_t)ptr->end);
-      length= length - (size_t)(ptr->end - ptr->begin);
+      bool in_hole= get_write_pos(&write_end, ptr);
+
+      /* if there is something to write, write it */
+      if ((write_end - write_begin) &&
+          (my_write(update_temp_file,
+                    file_buff->ptr() + (write_begin - file_buff->start()),
+                    write_end - write_begin, MYF_RW)))
+        goto error;
+
+      if (in_hole)
+      {
+        /* skip hole */
+        while (file_buff->end() <= ptr->end && file_buffer_start != -1)
+          file_buffer_start= file_buff->read_next();
+        write_begin= ptr->end;
+        ptr++;
       }
+      else
+        write_begin= write_end;
+
+      if (write_end == file_buff->end())
+        file_buffer_start= file_buff->read_next(); /* shift the buffer */
+
+    }
+
+    if (my_close(update_temp_file, MYF(0)))
+      DBUG_RETURN(-1);
+    share->update_file_opened= FALSE;
 
-    /* Unmap the file before the new size is set */
-    if (my_munmap(share->mapped_file, share->file_stat.st_size))
+    if (share->tina_write_opened)
+    {
+      if (my_close(share->tina_write_filedes, MYF(0)))
         DBUG_RETURN(-1);
-    /* We set it to null so that get_mmap() won't try to unmap it */
-    share->mapped_file= NULL;
+      /*
+        Mark that the writer fd is closed, so that init_tina_writer()
+        will reopen it later.
+      */
+      share->tina_write_opened= FALSE;
+    }
 
-    /* Set the file to the new size */
-    if (my_chsize(share->data_file, length, 0, MYF(MY_WME)))
+    /*
+      Close opened fildes's. Then move updated file in place
+      of the old datafile.
+    */
+    if (my_close(data_file, MYF(0)) ||
+        my_rename(fn_format(updated_fname, share->table_name, "", CSN_EXT,
+                            MY_REPLACE_EXT | MY_UNPACK_FILENAME),
+                  share->data_file_name, MYF(0)))
       DBUG_RETURN(-1);
 
-    if (get_mmap(share, 0) > 0)
+    /* Open the file again and sync it */
+    if (((data_file= my_open(share->data_file_name, O_RDONLY, MYF(0))) == -1)
+        || my_sync(data_file, MYF(MY_WME)))
       DBUG_RETURN(-1);
   }
 
   DBUG_RETURN(0);
+error:
+  my_close(update_temp_file, MYF(0));
+  share->update_file_opened= FALSE;
+  DBUG_RETURN(-1);
 }
 
 
@@ -1195,10 +1303,11 @@ int ha_tina::repair(THD* thd, HA_CHECK_OPT* check_opt)
   File repair_file;
   int rc;
   ha_rows rows_repaired= 0;
+  off_t write_begin= 0, write_end;
   DBUG_ENTER("ha_tina::repair");
 
   /* empty file */
-  if (!share->mapped_file)
+  if (!share->saved_data_file_length)
   {
     share->rows_recorded= 0;
     goto end;
@@ -1207,12 +1316,15 @@ int ha_tina::repair(THD* thd, HA_CHECK_OPT* check_opt)
   if (!(buf= (byte*) my_malloc(table->s->reclength, MYF(MY_WME))))
     DBUG_RETURN(HA_ERR_OUT_OF_MEM);
 
+  /* position buffer to the start of the file */
+  file_buff->init_buff(data_file);
+
   /*
     Local_saved_data_file_length is initialized during the lock phase.
     Sometimes this is not getting executed before ::repair (e.g. for
     the log tables). We set it manually here.
   */
-  local_saved_data_file_length= share->file_stat.st_size;
+  local_saved_data_file_length= share->saved_data_file_length;
   /* set current position to the beginning of the file */
   current_position= next_position= 0;
 
@@ -1227,11 +1339,10 @@ int ha_tina::repair(THD* thd, HA_CHECK_OPT* check_opt)
 
   if (rc == HA_ERR_END_OF_FILE)
   {
-    /* All rows were read ok until end of file, the file does not need repair. */
-
     /*
-      If rows_recorded != rows_repaired, we should update
-      rows_recorded value to the current amount of rows.
+      All rows were read ok until end of file, the file does not need repair.
+      If rows_recorded != rows_repaired, we should update rows_recorded value
+      to the current amount of rows.
     */
     share->rows_recorded= rows_repaired;
     goto end;
@@ -1247,36 +1358,45 @@ int ha_tina::repair(THD* thd, HA_CHECK_OPT* check_opt)
                            0, O_RDWR | O_TRUNC,MYF(MY_WME))) < 0)
     DBUG_RETURN(HA_ERR_CRASHED_ON_REPAIR);
 
-  if (my_write(repair_file, (byte*)share->mapped_file, current_position,
-               MYF(MY_NABP)))
-    DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);
-  my_close(repair_file, MYF(0));
+  file_buff->init_buff(data_file);
+
+
   /* we just truncated the file up to the first bad row. update rows count. */
   share->rows_recorded= rows_repaired;
 
-  if (my_munmap(share->mapped_file, share->file_stat.st_size))
+  /* write repaired file */
+  while (1)
+  {
+    write_end= min(file_buff->end(), current_position);
+    if ((write_end - write_begin) &&
+        (my_write(repair_file, file_buff->ptr(),
+                  write_end - write_begin, MYF_RW)))
       DBUG_RETURN(-1);
-  /* We set it to null so that get_mmap() won't try to unmap it */
-  share->mapped_file= NULL;
+
+    write_begin= write_end;
+    if (write_end== current_position)
+      break;
+    else
+      file_buff->read_next(); /* shift the buffer */
+  }
 
   /*
-    Close the "to"-file before renaming
-    On Windows one cannot rename a file, which descriptor
-    is still open. EACCES will be returned when trying to delete 
-    the "to"-file in my_rename()
+    Close the files and rename repaired file to the datafile.
+    We have to close the files, as on Windows one cannot rename
+    a file, which descriptor is still open. EACCES will be returned
+    when trying to delete the "to"-file in my_rename().
   */
-  my_close(share->data_file,MYF(0));
-
-  if (my_rename(repaired_fname, share->data_file_name, MYF(0)))
+  if (my_close(data_file,MYF(0)) || my_close(repair_file, MYF(0)) ||
+      my_rename(repaired_fname, share->data_file_name, MYF(0)))
     DBUG_RETURN(-1);
 
   /* Open the file again, it should now be repaired */
-  if ((share->data_file= my_open(share->data_file_name, O_RDWR|O_APPEND,
+  if ((data_file= my_open(share->data_file_name, O_RDWR|O_APPEND,
                           MYF(0))) == -1)
      DBUG_RETURN(-1);
 
-  if (get_mmap(share, 0) > 0)
-    DBUG_RETURN(-1);
+  /* Set new file size. The file size will be updated by ::update_status() */
+  local_saved_data_file_length= (size_t) current_position;
 
 end:
   share->crashed= FALSE;
@@ -1295,17 +1415,12 @@ int ha_tina::delete_all_rows()
   if (!records_is_known)
     DBUG_RETURN(my_errno=HA_ERR_WRONG_COMMAND);
 
-  /* Unmap the file before the new size is set */
-  if (share->mapped_file && my_munmap(share->mapped_file,
-                                      share->file_stat.st_size))
+  if (!share->tina_write_opened)
+    if (init_tina_writer())
       DBUG_RETURN(-1);
-  share->mapped_file= NULL;
 
   /* Truncate the file to zero size */
-  rc= my_chsize(share->data_file, 0, 0, MYF(MY_WME));
-
-  if (get_mmap(share, 0) > 0)
-    DBUG_RETURN(-1);
+  rc= my_chsize(share->tina_write_filedes, 0, 0, MYF(MY_WME));
 
   records=0;
   DBUG_RETURN(rc);
@@ -1367,12 +1482,15 @@ int ha_tina::check(THD* thd, HA_CHECK_OPT* check_opt)
   if (!(buf= (byte*) my_malloc(table->s->reclength, MYF(MY_WME))))
     DBUG_RETURN(HA_ERR_OUT_OF_MEM);
 
+  /* position buffer to the start of the file */
+  file_buff->init_buff(data_file);
+
   /*
     Local_saved_data_file_length is initialized during the lock phase.
     Check does not use store_lock in certain cases. So, we set it
     manually here.
   */
-  local_saved_data_file_length= share->file_stat.st_size;
+  local_saved_data_file_length= share->saved_data_file_length;
   /* set current position to the beginning of the file */
   current_position= next_position= 0;
   /* Read the file row-by-row. If everything is ok, repair is not needed. */
@@ -1412,6 +1530,7 @@ mysql_declare_plugin(csv)
   tina_init_func, /* Plugin Init */
   tina_done_func, /* Plugin Deinit */
   0x0100 /* 1.0 */,
+  0
 }
 mysql_declare_plugin_end;
 

storage/csv/ha_tina.h

@@ -19,6 +19,7 @@
 #include <my_dir.h>
 
 #define DEFAULT_CHAIN_LENGTH 512
+#define DEFAULT_FILE_WINDOW_SIZE 4096
 /*
   Version for file format.
   1 - Initial Version. That is, the version when the metafile was introduced.
@@ -29,15 +30,12 @@
 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 +44,10 @@ 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 +57,54 @@ 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(File filedes_arg) : lower_bound(0),
+                                        buff_size(DEFAULT_FILE_WINDOW_SIZE)
+   {
+     buff= (byte *) my_malloc(buff_size*sizeof(byte),  MYF(MY_WME));
+     /* read the beginning of the file */
+     init_buff(filedes_arg);
+   }
+
+   ~Transparent_file()
+   { my_free(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 +113,9 @@ 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 +129,19 @@ 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"; }