MDEV-33585 The maximum innodb_log_buffer_size is too large

On Microsoft Windows, ReadFile() as well as WriteFile() limit the size
of the request to DWORD, which is 32 bits (at most 4 GiB - 1) also on
64-bit systems.

On FreeBSD, sysctl debug.iosize_max_clamp could limit the size of a
write request to INT_MAX. The size of a read request is always limited
to INT_MAX. This would allow the request size to be 4095 bytes more than
the Linux limit (0x7ffff000 according to "man 2 read" and "man 2 write").

On OpenBSD, Solaris and possibly NetBSD, the read request size is limited
to SSIZE_T_MAX, which would be half the current maximum
innodb_log_buffer_size. This should be not much of an issue anyway,
because on contemporary 64-bit platforms, the virtual addresses are
limited to 48 bits.

IBM AIX documentation mentions OFF_MAX which would apply when
a 64-bit application is running on a 32-bit kernel.

Let us declare innodb_log_buffer_size as 32-bit unsigned and make the
maximum 0x7ffff000, to be compatible with the least common
denominator (Linux).

The maximum innodb_sort_buffer_size already was 64 MiB,
which is not a problem.

SyncFileIO::execute(): Assert that the size of a synchronous read or
write request is limited to the maximum.

Reviewed by: Vladislav Vaintroub

extra/mariabackup/xtrabackup.cc

@@ -1888,8 +1888,8 @@ struct my_option xb_server_options[] =
   {"innodb_log_buffer_size", OPT_INNODB_LOG_BUFFER_SIZE,
    "Redo log buffer size in bytes.",
    (G_PTR*) &log_sys.buf_size, (G_PTR*) &log_sys.buf_size, 0,
-   IF_WIN(GET_ULL,GET_ULONG), REQUIRED_ARG, 2U << 20,
-   2U << 20, SIZE_T_MAX, 0, 4096, 0},
+   GET_UINT, REQUIRED_ARG, 2U << 20,
+   2U << 20, log_sys.buf_size_max, 0, 4096, 0},
 #if defined __linux__ || defined _WIN32
   {"innodb_log_file_buffering", OPT_INNODB_LOG_FILE_BUFFERING,
    "Whether the file system cache for ib_logfile0 is enabled during --backup",

mysql-test/suite/sys_vars/r/sysvars_innodb,32bit.rdiff

@@ -1,4 +1,6 @@
-@@ -49,7 +49,7 @@
+--- sysvars_innodb.result
++++ sysvars_innodb.result,32bit
+@@ -47,7 +47,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	8
  VARIABLE_SCOPE	GLOBAL
@@ -7,7 +9,7 @@
  VARIABLE_COMMENT	Number of InnoDB Adaptive Hash Index Partitions (default 8)
  NUMERIC_MIN_VALUE	1
  NUMERIC_MAX_VALUE	512
-@@ -73,7 +73,7 @@
+@@ -71,7 +71,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	1
  VARIABLE_SCOPE	GLOBAL
@@ -16,7 +18,7 @@
  VARIABLE_COMMENT	The AUTOINC lock modes supported by InnoDB: 0 => Old style AUTOINC locking (for backward compatibility); 1 => New style AUTOINC locking; 2 => No AUTOINC locking (unsafe for SBR)
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	2
-@@ -85,10 +85,10 @@
+@@ -83,10 +83,10 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	0
  VARIABLE_SCOPE	GLOBAL
@@ -29,7 +31,7 @@
  NUMERIC_BLOCK_SIZE	1048576
  ENUM_VALUE_LIST	NULL
  READ_ONLY	YES
-@@ -121,7 +121,7 @@
+@@ -119,7 +119,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	25
  VARIABLE_SCOPE	GLOBAL
@@ -38,7 +40,7 @@
  VARIABLE_COMMENT	Dump only the hottest N% of each buffer pool, defaults to 25
  NUMERIC_MIN_VALUE	1
  NUMERIC_MAX_VALUE	100
-@@ -205,7 +205,7 @@
+@@ -203,7 +203,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	0
  VARIABLE_SCOPE	GLOBAL
@@ -47,7 +49,7 @@
  VARIABLE_COMMENT	A number between [0, 100] that tells how oftern buffer pool dump status in percentages should be printed. E.g. 10 means that buffer pool dump status is printed when every 10% of number of buffer pool pages are dumped. Default is 0 (only start and end status is printed).
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	100
-@@ -325,7 +325,7 @@
+@@ -323,7 +323,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	5
  VARIABLE_SCOPE	GLOBAL
@@ -56,7 +58,7 @@
  VARIABLE_COMMENT	If the compression failure rate of a table is greater than this number more padding is added to the pages to reduce the failures. A value of zero implies no padding
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	100
-@@ -349,7 +349,7 @@
+@@ -347,7 +347,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	50
  VARIABLE_SCOPE	GLOBAL
@@ -65,7 +67,7 @@
  VARIABLE_COMMENT	Percentage of empty space on a data page that can be reserved to make the page compressible.
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	75
-@@ -625,7 +625,7 @@
+@@ -623,7 +623,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	600
  VARIABLE_SCOPE	GLOBAL
@@ -74,7 +76,7 @@
  VARIABLE_COMMENT	Maximum number of seconds that semaphore times out in InnoDB.
  NUMERIC_MIN_VALUE	1
  NUMERIC_MAX_VALUE	4294967295
-@@ -673,7 +673,7 @@
+@@ -671,7 +671,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	30
  VARIABLE_SCOPE	GLOBAL
@@ -83,7 +85,7 @@
  VARIABLE_COMMENT	Number of iterations over which the background flushing is averaged.
  NUMERIC_MIN_VALUE	1
  NUMERIC_MAX_VALUE	1000
-@@ -697,7 +697,7 @@
+@@ -695,7 +695,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	1
  VARIABLE_SCOPE	GLOBAL
@@ -92,7 +94,7 @@
  VARIABLE_COMMENT	Controls the durability/speed trade-off for commits. Set to 0 (write and flush redo log to disk only once per second), 1 (flush to disk at each commit), 2 (write to log at commit but flush to disk only once per second) or 3 (flush to disk at prepare and at commit, slower and usually redundant). 1 and 3 guarantees that after a crash, committed transactions will not be lost and will be consistent with the binlog and other transactional engines. 2 can get inconsistent and lose transactions if there is a power failure or kernel crash but not if mysqld crashes. 0 has no guarantees in case of crash. 0 and 2 can be faster than 1 or 3.
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	3
-@@ -721,7 +721,7 @@
+@@ -719,7 +719,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	1
  VARIABLE_SCOPE	GLOBAL
@@ -101,7 +103,7 @@
  VARIABLE_COMMENT	Set to 0 (don't flush neighbors from buffer pool), 1 (flush contiguous neighbors from buffer pool) or 2 (flush neighbors from buffer pool), when flushing a block
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	2
-@@ -757,7 +757,7 @@
+@@ -755,7 +755,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	0
  VARIABLE_SCOPE	GLOBAL
@@ -110,7 +112,7 @@
  VARIABLE_COMMENT	Helps to save your data in case the disk image of the database becomes corrupt. Value 5 can return bogus data, and 6 can permanently corrupt data.
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	6
-@@ -781,10 +781,10 @@
+@@ -779,10 +779,10 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	8000000
  VARIABLE_SCOPE	GLOBAL
@@ -123,7 +125,7 @@
  NUMERIC_BLOCK_SIZE	0
  ENUM_VALUE_LIST	NULL
  READ_ONLY	NO
-@@ -817,7 +817,7 @@
+@@ -815,7 +815,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	84
  VARIABLE_SCOPE	GLOBAL
@@ -132,7 +134,7 @@
  VARIABLE_COMMENT	InnoDB Fulltext search maximum token size in characters
  NUMERIC_MIN_VALUE	10
  NUMERIC_MAX_VALUE	84
-@@ -829,7 +829,7 @@
+@@ -827,7 +827,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	3
  VARIABLE_SCOPE	GLOBAL
@@ -141,7 +143,7 @@
  VARIABLE_COMMENT	InnoDB Fulltext search minimum token size in characters
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	16
-@@ -841,7 +841,7 @@
+@@ -839,7 +839,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	2000
  VARIABLE_SCOPE	GLOBAL
@@ -150,7 +152,7 @@
  VARIABLE_COMMENT	InnoDB Fulltext search number of words to optimize for each optimize table call 
  NUMERIC_MIN_VALUE	1000
  NUMERIC_MAX_VALUE	10000
-@@ -853,10 +853,10 @@
+@@ -851,10 +851,10 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	2000000000
  VARIABLE_SCOPE	GLOBAL
@@ -163,7 +165,7 @@
  NUMERIC_BLOCK_SIZE	0
  ENUM_VALUE_LIST	NULL
  READ_ONLY	NO
-@@ -877,7 +877,7 @@
+@@ -875,7 +875,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	2
  VARIABLE_SCOPE	GLOBAL
@@ -172,7 +174,7 @@
  VARIABLE_COMMENT	InnoDB Fulltext search parallel sort degree, will round up to nearest power of 2 number
  NUMERIC_MIN_VALUE	1
  NUMERIC_MAX_VALUE	16
-@@ -889,10 +889,10 @@
+@@ -887,10 +887,10 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	640000000
  VARIABLE_SCOPE	GLOBAL
@@ -185,7 +187,7 @@
  NUMERIC_BLOCK_SIZE	0
  ENUM_VALUE_LIST	NULL
  READ_ONLY	NO
-@@ -937,22 +937,22 @@
+@@ -935,22 +935,22 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	200
  VARIABLE_SCOPE	GLOBAL
@@ -213,20 +215,7 @@
  NUMERIC_BLOCK_SIZE	0
  ENUM_VALUE_LIST	NULL
  READ_ONLY	NO
-@@ -985,10 +985,10 @@
- SESSION_VALUE	NULL
- DEFAULT_VALUE	16777216
- VARIABLE_SCOPE	GLOBAL
--VARIABLE_TYPE	BIGINT UNSIGNED
-+VARIABLE_TYPE	INT UNSIGNED
- VARIABLE_COMMENT	Redo log buffer size in bytes.
- NUMERIC_MIN_VALUE	2097152
--NUMERIC_MAX_VALUE	18446744073709551615
-+NUMERIC_MAX_VALUE	4294967295
- NUMERIC_BLOCK_SIZE	4096
- ENUM_VALUE_LIST	NULL
- READ_ONLY	YES
-@@ -1033,10 +1033,10 @@
+@@ -1043,10 +1043,10 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	32
  VARIABLE_SCOPE	GLOBAL
@@ -239,7 +228,7 @@
  NUMERIC_BLOCK_SIZE	0
  ENUM_VALUE_LIST	NULL
  READ_ONLY	NO
-@@ -1045,10 +1045,10 @@
+@@ -1055,10 +1055,10 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	1536
  VARIABLE_SCOPE	GLOBAL
@@ -252,7 +241,7 @@
  NUMERIC_BLOCK_SIZE	0
  ENUM_VALUE_LIST	NULL
  READ_ONLY	NO
-@@ -1081,10 +1081,10 @@
+@@ -1091,10 +1091,10 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	0
  VARIABLE_SCOPE	GLOBAL
@@ -265,7 +254,7 @@
  NUMERIC_BLOCK_SIZE	0
  ENUM_VALUE_LIST	NULL
  READ_ONLY	NO
-@@ -1093,7 +1093,7 @@
+@@ -1103,7 +1103,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	0
  VARIABLE_SCOPE	GLOBAL
@@ -274,7 +263,7 @@
  VARIABLE_COMMENT	Maximum delay of user threads in micro-seconds
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	10000000
-@@ -1225,10 +1225,10 @@
+@@ -1235,10 +1235,10 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	0
  VARIABLE_SCOPE	GLOBAL
@@ -287,7 +276,7 @@
  NUMERIC_BLOCK_SIZE	0
  ENUM_VALUE_LIST	NULL
  READ_ONLY	YES
-@@ -1249,7 +1249,7 @@
+@@ -1259,7 +1259,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	16384
  VARIABLE_SCOPE	GLOBAL
@@ -296,7 +285,7 @@
  VARIABLE_COMMENT	Page size to use for all InnoDB tablespaces.
  NUMERIC_MIN_VALUE	4096
  NUMERIC_MAX_VALUE	65536
-@@ -1285,7 +1285,7 @@
+@@ -1295,7 +1295,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	1000
  VARIABLE_SCOPE	GLOBAL
@@ -305,7 +294,7 @@
  VARIABLE_COMMENT	Number of UNDO log pages to purge in one batch from the history list.
  NUMERIC_MIN_VALUE	1
  NUMERIC_MAX_VALUE	5000
-@@ -1297,7 +1297,7 @@
+@@ -1307,7 +1307,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	128
  VARIABLE_SCOPE	GLOBAL
@@ -314,7 +303,7 @@
  VARIABLE_COMMENT	Deprecated parameter with no effect
  NUMERIC_MIN_VALUE	1
  NUMERIC_MAX_VALUE	128
-@@ -1333,7 +1333,7 @@
+@@ -1343,7 +1343,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	56
  VARIABLE_SCOPE	GLOBAL
@@ -323,7 +312,7 @@
  VARIABLE_COMMENT	Number of pages that must be accessed sequentially for InnoDB to trigger a readahead.
  NUMERIC_MIN_VALUE	0
  NUMERIC_MAX_VALUE	64
-@@ -1405,7 +1405,7 @@
+@@ -1427,7 +1427,7 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	1048576
  VARIABLE_SCOPE	GLOBAL
@@ -332,7 +321,7 @@
  VARIABLE_COMMENT	Memory buffer size for index creation
  NUMERIC_MIN_VALUE	65536
  NUMERIC_MAX_VALUE	67108864
-@@ -1573,10 +1573,10 @@
+@@ -1595,10 +1595,10 @@
  SESSION_VALUE	NULL
  DEFAULT_VALUE	30
  VARIABLE_SCOPE	GLOBAL

mysql-test/suite/sys_vars/r/sysvars_innodb.result

@@ -983,10 +983,10 @@ VARIABLE_NAME	INNODB_LOG_BUFFER_SIZE
 SESSION_VALUE	NULL
 DEFAULT_VALUE	16777216
 VARIABLE_SCOPE	GLOBAL
-VARIABLE_TYPE	BIGINT UNSIGNED
+VARIABLE_TYPE	INT UNSIGNED
 VARIABLE_COMMENT	Redo log buffer size in bytes.
 NUMERIC_MIN_VALUE	2097152
-NUMERIC_MAX_VALUE	18446744073709551615
+NUMERIC_MAX_VALUE	2147479552
 NUMERIC_BLOCK_SIZE	4096
 ENUM_VALUE_LIST	NULL
 READ_ONLY	YES

storage/innobase/handler/ha_innodb.cc

@@ -19316,10 +19316,10 @@ static MYSQL_SYSVAR_ULONG(page_size, srv_page_size,
   NULL, NULL, UNIV_PAGE_SIZE_DEF,
   UNIV_PAGE_SIZE_MIN, UNIV_PAGE_SIZE_MAX, 0);
 
-static MYSQL_SYSVAR_SIZE_T(log_buffer_size, log_sys.buf_size,
+static MYSQL_SYSVAR_UINT(log_buffer_size, log_sys.buf_size,
   PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY,
   "Redo log buffer size in bytes.",
-  NULL, NULL, 16U << 20, 2U << 20, SIZE_T_MAX, 4096);
+  NULL, NULL, 16U << 20, 2U << 20, log_sys.buf_size_max, 4096);
 
 #if defined __linux__ || defined _WIN32
 static MYSQL_SYSVAR_BOOL(log_file_buffering, log_sys.log_buffered,

storage/innobase/include/log0log.h

@@ -132,6 +132,9 @@ class log_file_t
 /** Redo log buffer */
 struct log_t
 {
+  /** The maximum buf_size */
+  static constexpr unsigned buf_size_max= os_file_request_size_max;
+
   /** The original (not version-tagged) InnoDB redo log format */
   static constexpr uint32_t FORMAT_3_23= 0;
   /** The MySQL 5.7.9/MariaDB 10.2.2 log format */
@@ -187,7 +190,7 @@ struct log_t
   /** number of append_prepare_wait(); protected by lock_lsn() or lsn_lock */
   size_t waits;
   /** innodb_log_buffer_size (size of buf,flush_buf if !is_pmem(), in bytes) */
-  size_t buf_size;
+  unsigned buf_size;
   /** log file size in bytes, including the header */
   lsn_t file_size;
 

storage/innobase/include/os0file.h

@@ -46,6 +46,18 @@ Created 10/21/1995 Heikki Tuuri
 #include <time.h>
 #endif /* !_WIN32 */
 
+/** The maximum size of a read or write request.
+
+According to Linux "man 2 read" and "man 2 write" this applies to
+both 32-bit and 64-bit systems.
+
+On FreeBSD, the limit is close to the Linux one, INT_MAX.
+
+On Microsoft Windows, the limit is UINT_MAX (4 GiB - 1).
+
+On other systems, the limit typically is up to SSIZE_T_MAX. */
+static constexpr unsigned os_file_request_size_max= 0x7ffff000;
+
 extern bool	os_has_said_disk_full;
 
 /** File offset in bytes */

storage/innobase/os/os0file.cc

@@ -678,6 +678,8 @@ ssize_t
 SyncFileIO::execute(const IORequest& request)
 {
 	ssize_t	n_bytes;
+	ut_ad(m_n > 0);
+	ut_ad(m_n <= os_file_request_size_max);
 
 	if (request.is_read()) {
 #ifdef _WIN32