Commit 05dd9e49 authored by unknown's avatar unknown

NULL MERGE this ChangeSet to 5.1

Apply innodb-5.0-ss1696 snapshot

Fixes:
- Bug#20090: InnoDB: Error: trying to declare trx to enter InnoDB
- Bug#23710: crash_commit_before fails if innodb_file_per_table=1
  At InnoDB startup consider the case where log scan went beyond
  checkpoint_lsn as a crash and initiate crash recovery code path.
- Bug#28781: InnoDB increments auto-increment value incorrectly with ON DUPLICATE KEY UPDATE
  We need to do some special AUTOINC handling for the following case:
  INSERT INTO t (c1,c2) VALUES(x,y) ON DUPLICATE KEY UPDATE ...
  We need to use the AUTOINC counter that was actually used by
  MySQL in the UPDATE statement, which can be different from the
  value used in the INSERT statement.
- Bug#29097: fsp_get_available_space_in_free_extents() is capped at 4TB
  Fix by typecasting the variables before multiplying them, so that the
  result of the multiplication is of type "unsigned long long".
- Bug#29155: Innodb "Parallel recovery" is not prevented
  Fix by enabling file locking on FreeBSD.  It has been disabled because
  InnoDB has refused to start on FreeBSD & LinuxThreads, but now it
  starts just fine.


innobase/fsp/fsp0fsp.c:
  Apply innodb-5.0-ss1696 snapshot
  
  Revision r1614:
  branches/5.0:
  
  Merge r1605 from trunk:
  
  Fix Bug#29097 "fsp_get_available_space_in_free_extents() is capped at 4TB"
  by typecasting the variables before multiplying them, so that the result of
  the multiplication is of type "unsigned long long".
  
  I verified this fix by creating a sparse file of 6TB and forcing InnoDB to
  use it without overwriting it with zeroes (by commenting the code that
  overwrites :newraw files).
  
  New type ullint is introduced with the sole purpose of shortening
  "unsigned long long", please do not define it to something else than
  "unsigned long long".
  
  Approved by:	Heikki
innobase/include/fsp0fsp.h:
  Apply innodb-5.0-ss1696 snapshot
  
  Revision r1614:
  branches/5.0:
  
  Merge r1605 from trunk:
  
  Fix Bug#29097 "fsp_get_available_space_in_free_extents() is capped at 4TB"
  by typecasting the variables before multiplying them, so that the result of
  the multiplication is of type "unsigned long long".
  
  I verified this fix by creating a sparse file of 6TB and forcing InnoDB to
  use it without overwriting it with zeroes (by commenting the code that
  overwrites :newraw files).
  
  New type ullint is introduced with the sole purpose of shortening
  "unsigned long long", please do not define it to something else than
  "unsigned long long".
  
  Approved by:	Heikki
innobase/include/univ.i:
  Apply innodb-5.0-ss1696 snapshot
  
  Revision r1614:
  branches/5.0:
  
  Merge r1605 from trunk:
  
  Fix Bug#29097 "fsp_get_available_space_in_free_extents() is capped at 4TB"
  by typecasting the variables before multiplying them, so that the result of
  the multiplication is of type "unsigned long long".
  
  I verified this fix by creating a sparse file of 6TB and forcing InnoDB to
  use it without overwriting it with zeroes (by commenting the code that
  overwrites :newraw files).
  
  New type ullint is introduced with the sole purpose of shortening
  "unsigned long long", please do not define it to something else than
  "unsigned long long".
  
  Approved by:	Heikki
innobase/log/log0recv.c:
  Apply innodb-5.0-ss1696 snapshot
  
  Revision r1608:
  branches/5.0: Bug#23710
  Back port of r1607 from trunk
  
  At InnoDB startup consider the case where log scan went beyond checkpoint_lsn as a crash and initiate crash recovery code path.
  
  reviewed by: Heikki
innobase/os/os0file.c:
  Apply innodb-5.0-ss1696 snapshot
  
  Revision r1615:
  branches/5.0:
  
  Merge r1613 from trunk:
  
  Fix Bug#29155 by enabling file locking on FreeBSD.
  It has been disabled because InnoDB has refused to start on
  FreeBSD & LinuxThreads, but now it starts just fine.
  
  Approved by:	Heikki
innobase/srv/srv0srv.c:
  Apply innodb-5.0-ss1696 snapshot
  
  Revision r1552:
  branches/5.0:
  Fix Bug#20090 as suggested in the bug followup by Heikki.
  
  Approved by:	Heikki
innobase/trx/trx0trx.c:
  Apply innodb-5.0-ss1696 snapshot
  
  Revision r1596:
  branches/5.0: Merge r1595 from trunk:
  
  trx_commit_for_mysql(): Avoid acquiring and releasing kernel_mutex when
  trx->sess or trx_dummy_sess is non-NULL.
sql/ha_innodb.cc:
  Apply innodb-5.0-ss1696 snapshot
  
  Revision r1614:
  branches/5.0:
  
  Merge r1605 from trunk:
  
  Fix Bug#29097 "fsp_get_available_space_in_free_extents() is capped at 4TB"
  by typecasting the variables before multiplying them, so that the result of
  the multiplication is of type "unsigned long long".
  
  I verified this fix by creating a sparse file of 6TB and forcing InnoDB to
  use it without overwriting it with zeroes (by commenting the code that
  overwrites :newraw files).
  
  New type ullint is introduced with the sole purpose of shortening
  "unsigned long long", please do not define it to something else than
  "unsigned long long".
  
  Approved by:	Heikki
  
  
  Revision r1695:
  branches/5.0: Merge a change from MySQL AB:
  
  ChangeSet@1.2463.166.1  2007-06-20 19:22:27+03:00  monty@mysql.fi
  
  Allow multiple calls to mysql_server_end()
  (Part of fix for Bug 25621 Error in my_thread_global_end(): 1 threads
  didn't exit)
  Give correct error message if InnoDB table is not found
  (This allows us to drop a an innodb table that is not in the InnoDB registery)
  
  ha_innodb.cc:
    Give correct error message if InnoDB table is not found.
    (This allows us to drop a an innodb table that is not in the InnoDB
    registery)
  
  
  Revision r1606:
  branches/5.0: Formatting corrections.
  
  spotted by: Vasil
  
  
  Revision r1691:
  branches/5.0: Merge a change from MySQL AB:
  
  ChangeSet@1.2463.261.1  2007-07-20 14:17:15+03:00  gkodinov@magare.gmz
    Bug 29644: alter table hangs if records locked in share mode 
    by long running transaction
    
    On Windows opened files can't be deleted. There was a special
    upgraded lock mode (TL_WRITE instead of TL_WRITE_ALLOW_READ) 
    in ALTER TABLE to make sure nobody has the table opened
    when deleting the old table in ALTER TABLE. This special mode
    was causing ALTER TABLE to hang waiting on a lock inside InnoDB.
    This special lock is no longer necessary as the server is 
    closing the tables it needs to delete in ALTER TABLE.
    Fixed by removing the special lock.
    Note that this also reverses the fix for bug 17264 that deals with
    another consequence of this special lock mode being used.
  
  sql/ha_innodb.cc: Bug 29644: reverse the (now excessive) fix
  for bug 17264 (but leave the test case).
  
  
  Revision r1601:
  branches/5.0: Fix for bug#28781
  We need to do some special AUTOINC handling for the following case:
  
  INSERT INTO t (c1,c2) VALUES(x,y) ON DUPLICATE KEY UPDATE ...
  
  We need to use the AUTOINC counter that was actually used by
  MySQL in the UPDATE statement, which can be different from the
  value used in the INSERT statement.
  
  approved by: Sunny
  
  
  Revision r1692:
  branches/5.0: Merge a change from MySQL AB:
  
  ChangeSet@1.2463.267.1  2007-07-30 17:14:34+04:00  evgen@local
    Bug 24989: The DEADLOCK error is improperly handled by InnoDB.
    
    When innodb detects a deadlock it calls ha_rollback_trans() to rollback the 
    main transaction. But such action isn't allowed from inside of triggers and
    functions. When it happen the 'Explicit or implicit commit' error is thrown
    even if there is no commit/rollback statements in the trigger/function. This
    leads to the user confusion.
    
    Now the convert_error_code_to_mysql() function doesn't call the 
    ha_rollback_trans() function directly but rather calls the
    mark_transaction_to_rollback function and returns an error.
    The sp_rcontext::find_handler() now doesn't allow errors to be caught by the
    trigger/function error handlers when the thd->is_fatal_sub_stmt_error flag
    is set. Procedures are still allowed to catch such errors.
    The sp_rcontext::find_handler function now accepts a THD handle as a parameter.
    The transaction_rollback_request and the is_fatal_sub_stmt_error flags are 
    added to the THD class. The are initialized by the THD class constructor.
    Now the ha_autocommit_or_rollback function rolls back main transaction
    when not in a sub statement and the thd->transaction_rollback_request
    is set.
    The THD::restore_sub_statement_state function now resets the 
    thd->is_fatal_sub_stmt_error flag on exit from a sub-statement.
  
  innodb-big.test, innodb-big.result:
    Added a test case for the bug 24989: The DEADLOCK error is improperly
    handled by InnoDB.
  
  sql/ha_innodb.cc:
    Bug 24989: The DEADLOCK error is improperly handled by InnoDB.
    Now the convert_error_code_to_mysql() function doesn't call the
    ha_rollback_trans() function directly but rather calls the
    mark_transaction_to_rollback function and returns an error.
  
  
  Revision r1693:
  branches/5.0: Merge a change from MySQL AB:
  
  ChangeSet@1.2489.5.1  2007-07-31 17:42:48+04:00  evgen@local
    ha_innodb.cc:
      Warning fixed.
parent 0739c70e
......@@ -2806,7 +2806,7 @@ will be able to insert new data to the database without running out the
tablespace. Only free extents are taken into account and we also subtract
the safety margin required by the above function fsp_reserve_free_extents. */
ulint
ullint
fsp_get_available_space_in_free_extents(
/*====================================*/
/* out: available space in kB */
......@@ -2872,7 +2872,8 @@ fsp_get_available_space_in_free_extents(
return(0);
}
return(((n_free - reserve) * FSP_EXTENT_SIZE)
return((ullint)(n_free - reserve)
* FSP_EXTENT_SIZE
* (UNIV_PAGE_SIZE / 1024));
}
......
......@@ -245,7 +245,7 @@ will be able to insert new data to the database without running out the
tablespace. Only free extents are taken into account and we also subtract
the safety margin required by the above function fsp_reserve_free_extents. */
ulint
ullint
fsp_get_available_space_in_free_extents(
/*====================================*/
/* out: available space in kB */
......
......@@ -195,6 +195,8 @@ typedef __int64 ib_longlong;
typedef longlong ib_longlong;
#endif
typedef unsigned long long int ullint;
#ifndef __WIN__
#if SIZEOF_LONG != SIZEOF_VOIDP
#error "Error: InnoDB's ulint must be of the same size as void*"
......
......@@ -57,6 +57,16 @@ ibool recv_needed_recovery = FALSE;
ibool recv_lsn_checks_on = FALSE;
/* There are two conditions under which we scan the logs, the first
is normal startup and the second is when we do a recovery from an
archive.
This flag is set if we are doing a scan from the last checkpoint during
startup. If we find log entries that were written after the last checkpoint
we know that the server was not cleanly shutdown. We must then initialize
the crash recovery environment before attempting to store these entries in
the log hash table. */
ibool recv_log_scan_is_startup_type = FALSE;
/* If the following is TRUE, the buffer pool file pages must be invalidated
after recovery and no ibuf operations are allowed; this becomes TRUE if
the log record hash table becomes too full, and log records must be merged
......@@ -99,6 +109,16 @@ the recovery failed and the database may be corrupt. */
dulint recv_max_page_lsn;
/* prototypes */
/***********************************************************
Initialize crash recovery environment. Can be called iff
recv_needed_recovery == FALSE. */
static
void
recv_init_crash_recovery(void);
/*===========================*/
/************************************************************
Creates the recovery system. */
......@@ -2438,6 +2458,23 @@ recv_scan_log_recs(
if (ut_dulint_cmp(scanned_lsn, recv_sys->scanned_lsn) > 0) {
/* We have found more entries. If this scan is
of startup type, we must initiate crash recovery
environment before parsing these log records. */
if (recv_log_scan_is_startup_type
&& !recv_needed_recovery) {
fprintf(stderr,
"InnoDB: Log scan progressed"
" past the checkpoint lsn %lu %lu\n",
(ulong) ut_dulint_get_high(
recv_sys->scanned_lsn),
(ulong) ut_dulint_get_low(
recv_sys->scanned_lsn));
recv_init_crash_recovery();
}
/* We were able to find more log data: add it to the
parsing buffer if parse_start_lsn is already
non-zero */
......@@ -2557,6 +2594,48 @@ recv_group_scan_log_recs(
#endif /* UNIV_DEBUG */
}
/***********************************************************
Initialize crash recovery environment. Can be called iff
recv_needed_recovery == FALSE. */
static
void
recv_init_crash_recovery(void)
/*==========================*/
{
ut_a(!recv_needed_recovery);
recv_needed_recovery = TRUE;
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: Database was not"
" shut down normally!\n"
"InnoDB: Starting crash recovery.\n");
fprintf(stderr,
"InnoDB: Reading tablespace information"
" from the .ibd files...\n");
fil_load_single_table_tablespaces();
/* If we are using the doublewrite method, we will
check if there are half-written pages in data files,
and restore them from the doublewrite buffer if
possible */
if (srv_force_recovery < SRV_FORCE_NO_LOG_REDO) {
fprintf(stderr,
"InnoDB: Restoring possible"
" half-written data pages from"
" the doublewrite\n"
"InnoDB: buffer...\n");
trx_sys_doublewrite_init_or_restore_pages(TRUE);
}
}
/************************************************************
Recovers from a checkpoint. When this function returns, the database is able
to start processing of new user transactions, but the function
......@@ -2681,72 +2760,6 @@ recv_recovery_from_checkpoint_start(
recv_sys->recovered_lsn = checkpoint_lsn;
srv_start_lsn = checkpoint_lsn;
/* NOTE: we always do a 'recovery' at startup, but only if
there is something wrong we will print a message to the
user about recovery: */
if (ut_dulint_cmp(checkpoint_lsn, max_flushed_lsn) != 0
|| ut_dulint_cmp(checkpoint_lsn, min_flushed_lsn) != 0) {
if (ut_dulint_cmp(checkpoint_lsn, max_flushed_lsn)
< 0) {
fprintf(stderr,
"InnoDB: ##########################################################\n"
"InnoDB: WARNING!\n"
"InnoDB: The log sequence number in ibdata files is higher\n"
"InnoDB: than the log sequence number in the ib_logfiles! Are you sure\n"
"InnoDB: you are using the right ib_logfiles to start up the database?\n"
"InnoDB: Log sequence number in ib_logfiles is %lu %lu, log\n"
"InnoDB: sequence numbers stamped to ibdata file headers are between\n"
"InnoDB: %lu %lu and %lu %lu.\n"
"InnoDB: ##########################################################\n",
(ulong) ut_dulint_get_high(checkpoint_lsn),
(ulong) ut_dulint_get_low(checkpoint_lsn),
(ulong) ut_dulint_get_high(min_flushed_lsn),
(ulong) ut_dulint_get_low(min_flushed_lsn),
(ulong) ut_dulint_get_high(max_flushed_lsn),
(ulong) ut_dulint_get_low(max_flushed_lsn));
}
recv_needed_recovery = TRUE;
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: Database was not shut down normally!\n"
"InnoDB: Starting crash recovery.\n");
fprintf(stderr,
"InnoDB: Reading tablespace information from the .ibd files...\n");
fil_load_single_table_tablespaces();
/* If we are using the doublewrite method, we will
check if there are half-written pages in data files,
and restore them from the doublewrite buffer if
possible */
if (srv_force_recovery < SRV_FORCE_NO_LOG_REDO) {
fprintf(stderr,
"InnoDB: Restoring possible half-written data pages from the doublewrite\n"
"InnoDB: buffer...\n");
trx_sys_doublewrite_init_or_restore_pages(
TRUE);
}
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: Starting log scan based on checkpoint at\n"
"InnoDB: log sequence number %lu %lu.\n",
(ulong) ut_dulint_get_high(checkpoint_lsn),
(ulong) ut_dulint_get_low(checkpoint_lsn));
} else {
/* Init the doublewrite buffer memory structure */
trx_sys_doublewrite_init_or_restore_pages(FALSE);
}
}
contiguous_lsn = ut_dulint_align_down(recv_sys->scanned_lsn,
......@@ -2798,6 +2811,8 @@ recv_recovery_from_checkpoint_start(
group = UT_LIST_GET_NEXT(log_groups, group);
}
/* Set the flag to publish that we are doing startup scan. */
recv_log_scan_is_startup_type = (type == LOG_CHECKPOINT);
while (group) {
old_scanned_lsn = recv_sys->scanned_lsn;
......@@ -2819,6 +2834,69 @@ recv_recovery_from_checkpoint_start(
group = UT_LIST_GET_NEXT(log_groups, group);
}
/* Done with startup scan. Clear the flag. */
recv_log_scan_is_startup_type = FALSE;
if (type == LOG_CHECKPOINT) {
/* NOTE: we always do a 'recovery' at startup, but only if
there is something wrong we will print a message to the
user about recovery: */
if (ut_dulint_cmp(checkpoint_lsn, max_flushed_lsn) != 0
|| ut_dulint_cmp(checkpoint_lsn, min_flushed_lsn) != 0) {
if (ut_dulint_cmp(checkpoint_lsn, max_flushed_lsn)
< 0) {
fprintf(stderr,
"InnoDB: #########################"
"#################################\n"
"InnoDB: "
"WARNING!\n"
"InnoDB: The log sequence number"
" in ibdata files is higher\n"
"InnoDB: than the log sequence number"
" in the ib_logfiles! Are you sure\n"
"InnoDB: you are using the right"
" ib_logfiles to start up"
" the database?\n"
"InnoDB: Log sequence number in"
" ib_logfiles is %lu %lu, log\n"
"InnoDB: sequence numbers stamped"
" to ibdata file headers are between\n"
"InnoDB: %lu %lu and %lu %lu.\n"
"InnoDB: #########################"
"#################################\n",
(ulong) ut_dulint_get_high(
checkpoint_lsn),
(ulong) ut_dulint_get_low(
checkpoint_lsn),
(ulong) ut_dulint_get_high(
min_flushed_lsn),
(ulong) ut_dulint_get_low(
min_flushed_lsn),
(ulong) ut_dulint_get_high(
max_flushed_lsn),
(ulong) ut_dulint_get_low(
max_flushed_lsn));
}
if (!recv_needed_recovery) {
fprintf(stderr,
"InnoDB: The log sequence number"
" in ibdata files does not match\n"
"InnoDB: the log sequence number"
" in the ib_logfiles!\n");
recv_init_crash_recovery();
}
}
if (!recv_needed_recovery) {
/* Init the doublewrite buffer memory structure */
trx_sys_doublewrite_init_or_restore_pages(FALSE);
}
}
/* We currently have only one log group */
if (ut_dulint_cmp(group_scanned_lsn, checkpoint_lsn) < 0) {
ut_print_timestamp(stderr);
......@@ -2871,16 +2949,9 @@ recv_recovery_from_checkpoint_start(
recv_synchronize_groups(up_to_date_group);
if (!recv_needed_recovery) {
if (ut_dulint_cmp(checkpoint_lsn, recv_sys->recovered_lsn)
!= 0) {
fprintf(stderr,
"InnoDB: Warning: we did not need to do crash recovery, but log scan\n"
"InnoDB: progressed past the checkpoint lsn %lu %lu up to lsn %lu %lu\n",
(ulong) ut_dulint_get_high(checkpoint_lsn),
(ulong) ut_dulint_get_low(checkpoint_lsn),
(ulong) ut_dulint_get_high(recv_sys->recovered_lsn),
(ulong) ut_dulint_get_low(recv_sys->recovered_lsn));
}
ut_a(ut_dulint_cmp(checkpoint_lsn,
recv_sys->recovered_lsn) == 0);
} else {
srv_start_lsn = recv_sys->recovered_lsn;
}
......
......@@ -436,10 +436,9 @@ os_file_handle_error_no_exit(
#undef USE_FILE_LOCK
#define USE_FILE_LOCK
#if defined(UNIV_HOTBACKUP) || defined(__WIN__) || defined(__FreeBSD__) || defined(__NETWARE__)
#if defined(UNIV_HOTBACKUP) || defined(__WIN__) || defined(__NETWARE__)
/* InnoDB Hot Backup does not lock the data files.
* On Windows, mandatory locking is used.
* On FreeBSD with LinuxThreads, advisory locking does not work properly.
*/
# undef USE_FILE_LOCK
#endif
......
......@@ -1151,7 +1151,7 @@ srv_conc_force_enter_innodb(
srv_conc_n_threads++;
trx->declared_to_be_inside_innodb = TRUE;
trx->n_tickets_to_enter_innodb = 0;
trx->n_tickets_to_enter_innodb = 1;
os_fast_mutex_unlock(&srv_conc_mutex);
}
......
......@@ -1606,19 +1606,21 @@ trx_commit_for_mysql(
the transaction object does not have an InnoDB session object, and we
set the dummy session that we use for all MySQL transactions. */
mutex_enter(&kernel_mutex);
if (trx->sess == NULL) {
/* Open a dummy session */
if (!trx_dummy_sess) {
mutex_enter(&kernel_mutex);
if (!trx_dummy_sess) {
trx_dummy_sess = sess_open();
}
trx->sess = trx_dummy_sess;
mutex_exit(&kernel_mutex);
}
mutex_exit(&kernel_mutex);
trx->sess = trx_dummy_sess;
}
trx_start_if_not_started(trx);
......
......@@ -3301,8 +3301,6 @@ no_commit:
if (error == DB_DUPLICATE_KEY && auto_inc_used
&& (user_thd->lex->sql_command == SQLCOM_REPLACE
|| user_thd->lex->sql_command == SQLCOM_REPLACE_SELECT
|| (user_thd->lex->sql_command == SQLCOM_INSERT
&& user_thd->lex->duplicates == DUP_UPDATE)
|| (user_thd->lex->sql_command == SQLCOM_LOAD
&& user_thd->lex->duplicates == DUP_REPLACE))) {
......@@ -3533,6 +3531,27 @@ ha_innobase::update_row(
error = row_update_for_mysql((byte*) old_row, prebuilt);
/* We need to do some special AUTOINC handling for the following case:
INSERT INTO t (c1,c2) VALUES(x,y) ON DUPLICATE KEY UPDATE ...
We need to use the AUTOINC counter that was actually used by
MySQL in the UPDATE statement, which can be different from the
value used in the INSERT statement.*/
if (error == DB_SUCCESS
&& table->next_number_field && new_row == table->record[0]
&& user_thd->lex->sql_command == SQLCOM_INSERT
&& user_thd->lex->duplicates == DUP_UPDATE) {
longlong auto_inc;
auto_inc = table->next_number_field->val_int();
if (auto_inc != 0) {
dict_table_autoinc_update(prebuilt->table, auto_inc);
}
}
innodb_srv_conc_exit_innodb(prebuilt->trx);
error = convert_error_code_to_mysql(error, user_thd);
......@@ -5609,8 +5628,8 @@ ha_innobase::update_table_comment(
mutex_enter_noninline(&srv_dict_tmpfile_mutex);
rewind(srv_dict_tmpfile);
fprintf(srv_dict_tmpfile, "InnoDB free: %lu kB",
(ulong) fsp_get_available_space_in_free_extents(
fprintf(srv_dict_tmpfile, "InnoDB free: %llu kB",
fsp_get_available_space_in_free_extents(
prebuilt->table->space));
dict_print_info_on_foreign_keys(FALSE, srv_dict_tmpfile,
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment