Fix for Bug#37408 - Compressed MyISAM files should not require/use mmap()

                  
When compressed myisam files are opened, they are always memory mapped
sometimes causing memory swapping problems.

When we mmap the myisam compressed tables of size greater than the memory 
available, the kswapd0 process utilization is very high consuming 30-40% of 
the cpu. This happens only with linux kernels older than 2.6.9

With newer linux kernels, we don't have this problem of high cpu consumption
and this option may not be required.
 
The option 'myisam_mmap_size' is added to limit the amount of memory used for
memory mapping of myisam files. This option is not dynamic.

The default value on 32 bit system is 4294967295 bytes and on 64 bit system it
is 18446744073709547520 bytes.

Note: Testcase only tests the option variable. The actual bug has be to 
tested manually.

Author: Satya B

include/my_global.h

@@ -793,6 +793,9 @@ typedef SOCKET_SIZE_TYPE size_socket;
 #define DBL_MAX		1.79769313486231470e+308
 #define FLT_MAX		((float)3.40282346638528860e+38)
 #endif
+#ifndef SIZE_T_MAX
+#define SIZE_T_MAX      (~((size_t) 0))
+#endif
 
 #ifndef HAVE_FINITE
 #define finite(x) (1.0 / fabs(x) > 0.0)

include/myisam.h

@@ -270,6 +270,8 @@ extern ulong myisam_bulk_insert_tree_size, myisam_data_pointer_size;
 /* usually used to check if a symlink points into the mysql data home */
 /* which is normally forbidden                                        */
 extern int (*myisam_test_invalid_symlink)(const char *filename);
+extern ulonglong myisam_mmap_size, myisam_mmap_used;
+extern pthread_mutex_t THR_LOCK_myisam_mmap;
 
 	/* Prototypes for myisam-functions */
 
@@ -315,6 +317,8 @@ extern int mi_delete_all_rows(struct st_myisam_info *info);
 extern ulong _mi_calc_blob_length(uint length , const byte *pos);
 extern uint mi_get_pointer_length(ulonglong file_length, uint def);
 
+#define MEMMAP_EXTRA_MARGIN     7       /* Write this as a suffix for mmap file */
+
 /* this is used to pass to mysql_myisamchk_table -- by Sasha Pachev */
 
 #define   MYISAMCHK_REPAIR 1  /* equivalent to myisamchk -r */

myisam/mi_packrec.c

@@ -1499,12 +1499,26 @@ my_bool _mi_memmap_file(MI_INFO *info)
 {
   byte *file_map;
   MYISAM_SHARE *share=info->s;
+  my_bool eom;
+
   DBUG_ENTER("mi_memmap_file");
 
   if (!share->file_map)
   {
     my_off_t data_file_length= share->state.state.data_file_length;
-    if (data_file_length > (my_off_t) (~((size_t) 0)) - MEMMAP_EXTRA_MARGIN)
+
+    if (myisam_mmap_size != SIZE_T_MAX)
+    {
+      pthread_mutex_lock(&THR_LOCK_myisam_mmap);
+      eom= data_file_length > myisam_mmap_size - myisam_mmap_used - MEMMAP_EXTRA_MARGIN;
+      if (!eom)
+        myisam_mmap_used+= data_file_length + MEMMAP_EXTRA_MARGIN;
+      pthread_mutex_unlock(&THR_LOCK_myisam_mmap);
+    }
+    else
+      eom= data_file_length > myisam_mmap_size - MEMMAP_EXTRA_MARGIN;
+
+    if (eom)
     {
       DBUG_PRINT("warning", ("File is too large for mmap"));
       DBUG_RETURN(0);
@@ -1513,6 +1527,12 @@ my_bool _mi_memmap_file(MI_INFO *info)
         data_file_length + MEMMAP_EXTRA_MARGIN)
     {
       DBUG_PRINT("warning",("File isn't extended for memmap"));
+      if (myisam_mmap_size != SIZE_T_MAX)
+      {
+        pthread_mutex_lock(&THR_LOCK_myisam_mmap);
+        myisam_mmap_used-= data_file_length + MEMMAP_EXTRA_MARGIN;
+        pthread_mutex_unlock(&THR_LOCK_myisam_mmap);
+      }
       DBUG_RETURN(0);
     }
     file_map=(byte*)
@@ -1522,6 +1542,12 @@ my_bool _mi_memmap_file(MI_INFO *info)
     {
       DBUG_PRINT("warning",("mmap failed: errno: %d",errno));
       my_errno=errno;
+      if (myisam_mmap_size != SIZE_T_MAX)
+      {
+        pthread_mutex_lock(&THR_LOCK_myisam_mmap);
+        myisam_mmap_used-= data_file_length + MEMMAP_EXTRA_MARGIN;
+        pthread_mutex_unlock(&THR_LOCK_myisam_mmap);
+      }
       DBUG_RETURN(0);
     }
     share->file_map= file_map;
@@ -1538,6 +1564,13 @@ void _mi_unmap_file(MI_INFO *info)
   VOID(my_munmap(info->s->file_map,
 	      (size_t) info->s->state.state.data_file_length+
 	      MEMMAP_EXTRA_MARGIN));
+
+  if (myisam_mmap_size != SIZE_T_MAX)
+  {
+    pthread_mutex_lock(&THR_LOCK_myisam_mmap);
+    myisam_mmap_used-= info->s->state.state.data_file_length + MEMMAP_EXTRA_MARGIN;
+    pthread_mutex_unlock(&THR_LOCK_myisam_mmap);
+  }
 }
 
 

myisam/mi_static.c

@@ -40,7 +40,8 @@ ulong myisam_concurrent_insert= 0;
 my_off_t myisam_max_temp_length= MAX_FILE_SIZE;
 ulong    myisam_bulk_insert_tree_size=8192*1024;
 ulong    myisam_data_pointer_size=4;
-
+ulonglong    myisam_mmap_size= SIZE_T_MAX, myisam_mmap_used= 0;
+pthread_mutex_t  THR_LOCK_myisam_mmap;
 
 static int always_valid(const char *filename __attribute__((unused)))
 {

myisam/myisamdef.h

@@ -428,7 +428,6 @@ typedef struct st_mi_sort_param
 #define MI_MAX_BLOCK_LENGTH	((((ulong) 1 << 24)-1) & (~ (ulong) (MI_DYN_ALIGN_SIZE-1)))
 #define MI_REC_BUFF_OFFSET      ALIGN_SIZE(MI_DYN_DELETE_BLOCK_HEADER+sizeof(uint32))
 
-#define MEMMAP_EXTRA_MARGIN	7	/* Write this as a suffix for file */
 
 #define PACK_TYPE_SELECTED	1	/* Bits in field->pack_type */
 #define PACK_TYPE_SPACE_FIELDS	2

mysql-test/r/variables.result

@@ -904,3 +904,9 @@ set global server_id                 =@my_server_id;
 set global slow_launch_time          =@my_slow_launch_time;
 set global storage_engine            =@my_storage_engine;
 set global thread_cache_size         =@my_thread_cache_size;
+#
+# BUG#37408 - Compressed MyISAM files should not require/use mmap()
+#
+# Test 'myisam_mmap_size' option is not dynamic
+SET @@myisam_mmap_size= 500M;
+ERROR HY000: Variable 'myisam_mmap_size' is a read only variable

mysql-test/t/variables.test

@@ -765,3 +765,9 @@ set global slow_launch_time          =@my_slow_launch_time;
 set global storage_engine            =@my_storage_engine;
 set global thread_cache_size         =@my_thread_cache_size;
 
+--echo #
+--echo # BUG#37408 - Compressed MyISAM files should not require/use mmap()
+--echo #
+--echo # Test 'myisam_mmap_size' option is not dynamic
+--error ER_INCORRECT_GLOBAL_LOCAL_VAR
+SET @@myisam_mmap_size= 500M;

mysys/my_thr_init.c

@@ -30,7 +30,9 @@ pthread_key(struct st_my_thread_var, THR_KEY_mysys);
 #endif /* USE_TLS */
 pthread_mutex_t THR_LOCK_malloc,THR_LOCK_open,
 	        THR_LOCK_lock,THR_LOCK_isam,THR_LOCK_myisam,THR_LOCK_heap,
-                THR_LOCK_net, THR_LOCK_charset, THR_LOCK_threads;
+                THR_LOCK_net, THR_LOCK_charset, THR_LOCK_threads,
+                THR_LOCK_myisam_mmap;
+
 pthread_cond_t  THR_COND_threads;
 uint            THR_thread_count= 0;
 uint 		my_thread_end_wait_time= 5;
@@ -143,6 +145,7 @@ my_bool my_thread_global_init(void)
   pthread_mutex_init(&THR_LOCK_lock,MY_MUTEX_INIT_FAST);
   pthread_mutex_init(&THR_LOCK_isam,MY_MUTEX_INIT_SLOW);
   pthread_mutex_init(&THR_LOCK_myisam,MY_MUTEX_INIT_SLOW);
+  pthread_mutex_init(&THR_LOCK_myisam_mmap,MY_MUTEX_INIT_FAST);
   pthread_mutex_init(&THR_LOCK_heap,MY_MUTEX_INIT_FAST);
   pthread_mutex_init(&THR_LOCK_net,MY_MUTEX_INIT_FAST);
   pthread_mutex_init(&THR_LOCK_charset,MY_MUTEX_INIT_FAST);
@@ -208,6 +211,7 @@ void my_thread_global_end(void)
   pthread_mutex_destroy(&THR_LOCK_lock);
   pthread_mutex_destroy(&THR_LOCK_isam);
   pthread_mutex_destroy(&THR_LOCK_myisam);
+  pthread_mutex_destroy(&THR_LOCK_myisam_mmap);
   pthread_mutex_destroy(&THR_LOCK_heap);
   pthread_mutex_destroy(&THR_LOCK_net);
   pthread_mutex_destroy(&THR_LOCK_charset);

sql/mysqld.cc

@@ -4975,7 +4975,8 @@ enum options_mysqld
   OPT_MAX_WRITE_LOCK_COUNT, OPT_BULK_INSERT_BUFFER_SIZE,
   OPT_MAX_ERROR_COUNT, OPT_MULTI_RANGE_COUNT, OPT_MYISAM_DATA_POINTER_SIZE,
   OPT_MYISAM_BLOCK_SIZE, OPT_MYISAM_MAX_EXTRA_SORT_FILE_SIZE,
-  OPT_MYISAM_MAX_SORT_FILE_SIZE, OPT_MYISAM_SORT_BUFFER_SIZE,
+  OPT_MYISAM_MAX_SORT_FILE_SIZE, OPT_MYISAM_MMAP_SIZE,
+  OPT_MYISAM_SORT_BUFFER_SIZE,
   OPT_MYISAM_STATS_METHOD,
   OPT_NET_BUFFER_LENGTH, OPT_NET_RETRY_COUNT,
   OPT_NET_READ_TIMEOUT, OPT_NET_WRITE_TIMEOUT,
@@ -6255,6 +6256,10 @@ The minimum value for this variable is 4096.",
    (gptr*) &max_system_variables.myisam_max_sort_file_size, 0,
    GET_ULL, REQUIRED_ARG, (longlong) LONG_MAX, 0, (ulonglong) MAX_FILE_SIZE,
    0, 1024*1024, 0},
+  {"myisam_mmap_size", OPT_MYISAM_MMAP_SIZE,
+   "Can be used to restrict the total memory used for memory mmaping of myisam files",
+   (gptr*) &myisam_mmap_size, (gptr*) &myisam_mmap_size, 0,
+   GET_ULL, REQUIRED_ARG, SIZE_T_MAX, MEMMAP_EXTRA_MARGIN, SIZE_T_MAX, 0, 1, 0},
   {"myisam_repair_threads", OPT_MYISAM_REPAIR_THREADS,
    "Number of threads to use when repairing MyISAM tables. The value of 1 disables parallel repair.",
    (gptr*) &global_system_variables.myisam_repair_threads,

sql/set_var.cc

@@ -122,6 +122,7 @@ static byte *get_error_count(THD *thd);
 static byte *get_warning_count(THD *thd);
 static byte *get_have_innodb(THD *thd);
 static byte *get_tmpdir(THD *thd);
+static byte *get_myisam_mmap_size(THD *thd);
 
 /*
   Variable definition list
@@ -623,6 +624,10 @@ sys_var_thd_bool  sys_keep_files_on_create("keep_files_on_create",
                                            &SV::keep_files_on_create);
 
 
+static sys_var_readonly         sys_myisam_mmap_size("myisam_mmap_size",
+                                                     OPT_GLOBAL,
+                                                     SHOW_LONGLONG,
+                                                     get_myisam_mmap_size);
 
 
 /*
@@ -723,6 +728,7 @@ sys_var *sys_variables[]=
   &sys_multi_range_count,
   &sys_myisam_data_pointer_size,
   &sys_myisam_max_sort_file_size,
+  &sys_myisam_mmap_size,
   &sys_myisam_repair_threads,
   &sys_myisam_sort_buffer_size,
   &sys_myisam_stats_method,
@@ -1026,6 +1032,7 @@ struct show_var_st init_vars[]= {
   {sys_myisam_data_pointer_size.name, (char*) &sys_myisam_data_pointer_size, SHOW_SYS},
   {sys_myisam_max_sort_file_size.name, (char*) &sys_myisam_max_sort_file_size,
    SHOW_SYS},
+  {sys_myisam_mmap_size.name, (char*) &sys_myisam_mmap_size, SHOW_SYS},
   {"myisam_recover_options",  (char*) &myisam_recover_options_str,  SHOW_CHAR_PTR},
   {sys_myisam_repair_threads.name, (char*) &sys_myisam_repair_threads,
    SHOW_SYS},
@@ -3181,6 +3188,11 @@ static byte *get_tmpdir(THD *thd)
   return (byte*)mysql_tmpdir;
 }
 
+static byte *get_myisam_mmap_size(THD *thd)
+{
+  return (byte *)&myisam_mmap_size;
+}
+
 /****************************************************************************
   Main handling of variables:
   - Initialisation