mariadb/que/que0que.c
sunny 37312683cc branches/innodb+: Merge revisions r5971:6130 from branches/zip.
------------------------------------------------------------------------
  r5971 | marko | 2009-09-23 23:03:51 +1000 (Wed, 23 Sep 2009) | 2 lines

  branches/zip: os_file_pwrite(): Make the code compile in InnoDB Hot Backup
  when the pwrite system call is not available.
  ------------------------------------------------------------------------
  r5972 | marko | 2009-09-24 05:44:52 +1000 (Thu, 24 Sep 2009) | 5 lines

  branches/zip: fil_node_open_file(): In InnoDB Hot Backup,
  determine the page size of single-file tablespaces before computing
    the file node size.  Otherwise, the space->size of compressed tablespaces
  would be computed with UNIV_PAGE_SIZE instead of key_block_size.
  This should fix Issue #313.
  ------------------------------------------------------------------------
  r5973 | marko | 2009-09-24 05:53:21 +1000 (Thu, 24 Sep 2009) | 2 lines

  branches/zip: recv_add_to_hash_table():
  Simplify obfuscated pointer arithmetics.
  ------------------------------------------------------------------------
  r5978 | marko | 2009-09-24 17:47:56 +1000 (Thu, 24 Sep 2009) | 1 line

  branches/zip: Fix warnings and errors when UNIV_HOTBACKUP is defined.
  ------------------------------------------------------------------------
  r5979 | marko | 2009-09-24 20:16:10 +1000 (Thu, 24 Sep 2009) | 4 lines

  branches/zip: ha_innodb.cc: Define MYSQL_PLUGIN_IMPORT when necessary.
  This preprocessor symbol has been recently introduced in MySQL 5.1.
  The InnoDB Plugin should remain source compatible with MySQL 5.1.24
  and later.
  ------------------------------------------------------------------------
  r5988 | calvin | 2009-09-26 05:14:43 +1000 (Sat, 26 Sep 2009) | 8 lines

  branches/zip: fix bug#47055 unconditional exit(1) on ERROR_WORKING_SET_QUOTA
  1453 (0x5AD) for InnoDB backend

  When error ERROR_WORKING_SET_QUOTA or ERROR_NO_SYSTEM_RESOURCES
  occurs, yields for 100ms and retries the operation.

  Approved by: Heikki (on IM)

  ------------------------------------------------------------------------
  r5992 | vasil | 2009-09-28 17:10:29 +1000 (Mon, 28 Sep 2009) | 4 lines

  branches/zip:

  Add ChangeLog entry for c5988.

  ------------------------------------------------------------------------
  r5994 | marko | 2009-09-28 18:33:59 +1000 (Mon, 28 Sep 2009) | 17 lines

  branches/zip: Try to prevent the reuse of tablespace identifiers after
  InnoDB has crashed during table creation.  Also, refuse to start if
  files with duplicate tablespace identifiers are encountered.

    fil_node_create(): Update fil_system->max_assigned_id.  This should
  prevent the reuse of a space->id when InnoDB does a full crash
  recovery and invokes fil_load_single_table_tablespaces().  Normally,
  fil_system->max_assigned_id is initialized from
  SELECT MAX(ID) FROM SYS_TABLES.

  fil_open_single_table_tablespace(): Return FALSE when
  fil_space_create() fails.

  fil_load_single_table_tablespace(): Exit if fil_space_create() fails
  and innodb_force_recovery=0.

  rb://173 approved by Heikki Tuuri.  This addresses Issue #335.
  ------------------------------------------------------------------------
  r5995 | marko | 2009-09-28 18:52:25 +1000 (Mon, 28 Sep 2009) | 17 lines

  branches/zip: Do not write to PAGE_INDEX_ID after page creation,
  not even when restoring an uncompressed page after a compression failure.

  btr_page_reorganize_low(): On compression failure, do not restore
  those page header fields that should not be affected by the
  reorganization.  Instead, compare the fields.

  page_zip_decompress(): Add the parameter ibool all, for copying all
  page header fields.  Pass the parameter all=TRUE on block read
  completion, redo log application, and page_zip_validate(); pass
  all=FALSE in all other cases.

  page_zip_reorganize(): Do not restore the uncompressed page on
  failure.  It will be restored (to pre-modification state) by the
  caller anyway.

  rb://167, Issue #346
  ------------------------------------------------------------------------
  r5996 | marko | 2009-09-28 22:46:02 +1000 (Mon, 28 Sep 2009) | 4 lines

  branches/zip: Address Issue #350 in comments.

  lock_rec_queue_validate(), lock_rec_queue_validate(): Note that
  this debug code may violate the latching order and cause deadlocks.
  ------------------------------------------------------------------------
  r5997 | marko | 2009-09-28 23:03:58 +1000 (Mon, 28 Sep 2009) | 12 lines

  branches/zip: Remove an assertion failure when the InnoDB data dictionary
  is inconsistent with the MySQL .frm file.

  ha_innobase::index_read(): When the index cannot be found,
  return an error.

  ha_innobase::change_active_index(): When prebuilt->index == NULL,
  set also prebuilt->index_usable = FALSE.  This is not needed for
  correctness, because prebuilt->index_usable is only checked by
  row_search_for_mysql(), which requires prebuilt->index != NULL.

  This addresses Issue #349.  Approved by Heikki Tuuri over IM.
  ------------------------------------------------------------------------
  r6005 | vasil | 2009-09-29 18:09:52 +1000 (Tue, 29 Sep 2009) | 4 lines

  branches/zip:

  ChangeLog: wrap around 78th column, not earlier.

  ------------------------------------------------------------------------
  r6006 | vasil | 2009-09-29 20:15:25 +1000 (Tue, 29 Sep 2009) | 4 lines

  branches/zip:

  Add ChangeLog entry for the release of 1.0.4.

  ------------------------------------------------------------------------
  r6007 | vasil | 2009-09-29 23:19:59 +1000 (Tue, 29 Sep 2009) | 6 lines

  branches/zip:

  Fix the year, should be 2009.

  Pointed by:	Calvin

  ------------------------------------------------------------------------
  r6026 | marko | 2009-09-30 17:18:24 +1000 (Wed, 30 Sep 2009) | 1 line

  branches/zip: Add some debug assertions for checking FSEG_MAGIC_N.
  ------------------------------------------------------------------------
  r6028 | marko | 2009-09-30 23:55:23 +1000 (Wed, 30 Sep 2009) | 3 lines

  branches/zip: recv_no_log_write: New debug flag for tracking down
  Mantis Issue #347.  No modifications should be made to the database
  while recv_apply_hashed_log_recs() is about to complete.
  ------------------------------------------------------------------------
  r6029 | calvin | 2009-10-01 06:32:02 +1000 (Thu, 01 Oct 2009) | 4 lines

    branches/zip: non-functional changes

  Fix typo.

  ------------------------------------------------------------------------
  r6031 | marko | 2009-10-01 21:24:33 +1000 (Thu, 01 Oct 2009) | 49 lines

  branches/zip: Clean up after a crash during DROP INDEX.
  When InnoDB crashes while dropping an index, ensure that
  the index will be completely dropped during crash recovery.

  row_merge_drop_index(): Before dropping an index, rename the index to
  start with TEMP_INDEX_PREFIX_STR and commit the change, so that
  row_merge_drop_temp_indexes() will drop the index after crash
  recovery if the server crashes while dropping the index.

  fseg_inode_try_get(): New function, forked from fseg_inode_get().
  Return NULL if the file segment index node is free.

  fseg_inode_get(): Assert that the file segment index node is not free.

  fseg_free_step(): If the file segment index node is already free,
  print a diagnostic message and return TRUE.

  fsp_free_seg_inode(): Write a nonzero number to FSEG_MAGIC_N, so that
  allocated-and-freed file segment index nodes can be better
  distinguished from uninitialized ones.

  This is rb://174, addressing Issue #348.

  Tested by restarting mysqld upon the completion of the added
  log_write_up_to() invocation below, during DROP INDEX.  The index was
  dropped after crash recovery, and re-issuing the DROP INDEX did not
  crash the server.
    
    Index: btr/btr0btr.c
    ===================================================================
    --- btr/btr0btr.c	(revision 6026)
    +++ btr/btr0btr.c	(working copy)
    @@ -42,6 +42,7 @@ Created 6/2/1994 Heikki Tuuri
     #include "ibuf0ibuf.h"
     #include "trx0trx.h"
    +#include "log0log.h"
     
     /*
     Latching strategy of the InnoDB B-tree
     --------------------------------------
    @@ -873,6 +874,8 @@ leaf_loop:
     
   		  goto leaf_loop;
   	  }
    +
    +	log_write_up_to(mtr.end_lsn, LOG_WAIT_ALL_GROUPS, TRUE);
     top_loop:
   	  mtr_start(&mtr);

  ------------------------------------------------------------------------
  r6033 | calvin | 2009-10-02 06:19:46 +1000 (Fri, 02 Oct 2009) | 4 lines

  branches/zip: fix a typo in error message

  Reported as bug#47763.

  ------------------------------------------------------------------------
  r6043 | inaam | 2009-10-06 01:45:35 +1100 (Tue, 06 Oct 2009) | 12 lines

  branches/zip  rb://176

  Do not invalidate buffer pool while an LRU batch is active. Added
  code to buf_pool_invalidate() to wait for the running batches to finish.

  This patch also resets the state of buf_pool struct at invalidation. This
  addresses the concern where buf_pool->freed_page_clock becomes non-zero
  because we read in a system tablespace page for file format info at
  startup.

  Approved by: Marko

  ------------------------------------------------------------------------
  r6044 | pekka | 2009-10-07 01:44:54 +1100 (Wed, 07 Oct 2009) | 5 lines

  branches/zip:
  Add os_file_is_same() function for Hot Backup (inside ifdef UNIV_HOTBACKUP).
  This is part of the fix for Issue #186.
  Note! The Windows implementation is incomplete.

  ------------------------------------------------------------------------
  r6046 | pekka | 2009-10-08 20:24:56 +1100 (Thu, 08 Oct 2009) | 3 lines

  branches/zip: Revert r6044 which added os_file_is_same() function
  (issue#186). This functionality is moved to Hot Backup source tree.

  ------------------------------------------------------------------------
  r6048 | vasil | 2009-10-09 16:42:55 +1100 (Fri, 09 Oct 2009) | 16 lines

  branches/zip:

  When scanning a directory readdir() is called and stat() after it,
  if a file is deleted between the two calls stat will fail and the
  whole precedure will fail. Change this behavior to continue with the
  next entry if stat() fails because of nonexistent file. This is
  transparent change as it will make it look as if the file was deleted
  before the readdir() call.

  This change is needed in order to fix
  https://svn.innodb.com/mantis/view.php?id=174
  in which we need to abort if os_file_readdir_next_file()
  encounters "real" errors.

  Approved by:	Marko, Pekka (rb://177)

  ------------------------------------------------------------------------
  r6049 | vasil | 2009-10-10 03:05:26 +1100 (Sat, 10 Oct 2009) | 7 lines

  branches/zip:

  Fix compilation warning in Hot Backup:

  innodb/fil/fil0fil.c: In function 'fil_load_single_table_tablespace':
  innodb/fil/fil0fil.c:3253: warning: format '%lld' expects type 'long long int', but argument 6 has type 'ib_int64_t'

  ------------------------------------------------------------------------
  r6064 | calvin | 2009-10-14 02:23:35 +1100 (Wed, 14 Oct 2009) | 4 lines

  branches/zip: non-functional changes

  Changes from MySQL to fix build issue.

  ------------------------------------------------------------------------
  r6065 | inaam | 2009-10-14 04:43:13 +1100 (Wed, 14 Oct 2009) | 7 lines

  branches/zip rb://182

  Call fsync() on datafiles after a batch of pages is written to disk
  even when skip_innodb_doublewrite is set.

  Approved by: Heikki

  ------------------------------------------------------------------------
  r6080 | sunny | 2009-10-15 09:29:01 +1100 (Thu, 15 Oct 2009) | 3 lines

  branches/zip: Change page_mem_alloc_free() to inline.
  Fix Bug #47058 - Failure to compile innodb_plugin on solaris 10u7 + spro cc/CC 5.10

  ------------------------------------------------------------------------
  r6084 | vasil | 2009-10-15 16:21:17 +1100 (Thu, 15 Oct 2009) | 4 lines

  branches/zip:

  Add ChangeLog entry for r6080.

  ------------------------------------------------------------------------
  r6095 | vasil | 2009-10-20 00:04:59 +1100 (Tue, 20 Oct 2009) | 7 lines

  branches/zip:

  Fix Bug#47808 innodb_information_schema.test fails when run under valgrind 

  by using the wait_until_rows_count macro that loops until the number of
  rows becomes 14 instead of sleep 0.1, which is obviously very fragile.

  ------------------------------------------------------------------------
  r6096 | vasil | 2009-10-20 00:06:09 +1100 (Tue, 20 Oct 2009) | 4 lines

  branches/zip:

  Add ChangeLog entry for r6095.

  ------------------------------------------------------------------------
  r6099 | jyang | 2009-10-22 13:58:39 +1100 (Thu, 22 Oct 2009) | 7 lines

  branches/zip: Port bug #46000 related changes from 5.1 to zip
  branch. Due to different code path for creating index in zip
  branch comparing to 5.1), the index reserved name check function
  is extended to be used in ha_innobase::add_index(). 
  rb://190  Approved by: Marko


  ------------------------------------------------------------------------
  r6100 | jyang | 2009-10-22 14:51:07 +1100 (Thu, 22 Oct 2009) | 6 lines

  branches/zip: As a request from mysql, WARN_LEVEL_ERROR cannot
  be used for push_warning_* call any more. Switch to 
  WARN_LEVEL_WARN. Bug #47233.
  rb://172 approved by Sunny Bains and Marko.


  ------------------------------------------------------------------------
  r6101 | jyang | 2009-10-23 19:45:50 +1100 (Fri, 23 Oct 2009) | 7 lines
    
  branches/zip: Update test result with the WARN_LEVEL_ERROR
  to WARN_LEVEL_WARN change. This is the same result as 
  submitted in rb://172 review, which approved by Sunny Bains
  and Marko.

  ------------------------------------------------------------------------
  r6102 | marko | 2009-10-26 18:32:23 +1100 (Mon, 26 Oct 2009) | 1 line

  branches/zip: row_prebuilt_struct::prebuilts: Unused field, remove.
  ------------------------------------------------------------------------
  r6103 | marko | 2009-10-27 00:46:18 +1100 (Tue, 27 Oct 2009) | 4 lines

  branches/zip: row_ins_alloc_sys_fields(): Zero out the system columns
  DB_TRX_ID, DB_ROLL_PTR and DB_ROW_ID, in order to avoid harmless
  Valgrind warnings about uninitialized data.  (The warnings were
  harmless, because the fields would be initialized at a later stage.)
  ------------------------------------------------------------------------
  r6105 | calvin | 2009-10-28 09:05:52 +1100 (Wed, 28 Oct 2009) | 6 lines

  branches/zip: backport r3848 from 6.0 branch

  ----
      branches/6.0: innobase_start_or_create_for_mysql(): Make the 10 MB
      minimum tablespace limit independent of UNIV_PAGE_SIZE. (Bug #41490)

  ------------------------------------------------------------------------
  r6107 | marko | 2009-10-29 01:10:34 +1100 (Thu, 29 Oct 2009) | 5 lines

  branches/zip: buf_page_set_old(): Improve UNIV_LRU_DEBUG diagnostics
  in order to catch the buf_pool->LRU_old corruption reported in Issue #381.

  buf_LRU_old_init(): Set the property from the tail towards the front
  of the buf_pool->LRU list, in order not to trip the debug check.
  ------------------------------------------------------------------------
  r6108 | calvin | 2009-10-29 16:58:04 +1100 (Thu, 29 Oct 2009) | 5 lines

  branches/zip: close file handle when building with UNIV_HOTBACKUP

  The change does not affect regular InnoDB engine. Confirmed by
  Marko.

  ------------------------------------------------------------------------
  r6109 | jyang | 2009-10-29 19:37:32 +1100 (Thu, 29 Oct 2009) | 7 lines

  branches/zip: In os_mem_alloc_large(), if we fail to attach
  the shared memory, reset memory pointer ptr to NULL, and
   allocate memory from conventional pool.
  Bug #48237 Error handling in os_mem_alloc_large appears to be incorrect
  rb://198  Approved by: Marko


  ------------------------------------------------------------------------
  r6110 | marko | 2009-10-29 21:44:57 +1100 (Thu, 29 Oct 2009) | 2 lines

  branches/zip: Makefile.am (INCLUDES): Merge a change from MySQL:
  Use $(srcdir)/include instead of $(top_srcdir)/storage/innobase/include.
  ------------------------------------------------------------------------
  r6111 | marko | 2009-10-29 22:04:11 +1100 (Thu, 29 Oct 2009) | 33 lines

  branches/zip: Fix corruption of buf_pool->LRU_old and improve debug assertions.
  This was reported as Issue #381.

  buf_page_set_old(): Assert that blocks may only be set old if
  buf_pool->LRU_old is initialized and buf_pool->LRU_old_len is nonzero.
  Assert that buf_pool->LRU_old points to the block at the old/new boundary.

  buf_LRU_old_adjust_len(): Invoke buf_page_set_old() after adjusting
  buf_pool->LRU_old and buf_pool->LRU_old_len, in order not to violate
  the added assertions.

  buf_LRU_old_init(): Replace buf_page_set_old() with a direct
  assignment to bpage->old, because these loops that initialize all the
  blocks would temporarily violate the assertions about
  buf_pool->LRU_old.

  buf_LRU_remove_block(): When setting buf_pool->LRU_old = NULL, also
  clear all bpage->old flags and set buf_pool->LRU_old_len = 0.

  buf_LRU_add_block_to_end_low(), buf_LRU_add_block_low(): Move the
  buf_page_set_old() call later in order not to violate the debug
  assertions.  If buf_pool->LRU_old is NULL, set old=FALSE.

  buf_LRU_free_block(): Replace the UNIV_LRU_DEBUG assertion with a
  dummy buf_page_set_old() call that performs more thorough checks.

  buf_LRU_validate(): Do not tolerate garbage in buf_pool->LRU_old_len
  even if buf_pool->LRU_old is NULL.  Check that bpage->old is monotonic.

  buf_relocate(): Make the UNIV_LRU_DEBUG checks stricter.

  buf0buf.h: Revise the documentation of buf_page_t::old and
  buf_pool_t::LRU_old_len.
  ------------------------------------------------------------------------
  r6112 | calvin | 2009-10-30 01:21:15 +1100 (Fri, 30 Oct 2009) | 4 lines

  branches/zip: consideration for icc compilers

  Proposed by MySQL, and approved by Marko.

  ------------------------------------------------------------------------
  r6113 | vasil | 2009-10-30 03:15:50 +1100 (Fri, 30 Oct 2009) | 93 lines

  branches/zip: Merge r5912:6112 from branches/5.1:

  (after this merge the innodb-autoinc test starts to fail, but
  I commit anyway because it would be easier to investigate the
  failure this way)

    ------------------------------------------------------------------------
    r5952 | calvin | 2009-09-22 19:45:07 +0300 (Tue, 22 Sep 2009) | 7 lines
    Changed paths:
       M /branches/5.1/handler/ha_innodb.cc
    
    branches/5.1: fix bug#42383: Can't create table 'test.bug39438'
    
    For embedded server, MySQL may pass in full path, which is
    currently disallowed. It is needed to relax the condition by
    accepting full paths in the embedded case.
    
    Approved by: Heikki (on IM)
    ------------------------------------------------------------------------
    r6032 | vasil | 2009-10-01 15:55:49 +0300 (Thu, 01 Oct 2009) | 8 lines
    Changed paths:
       M /branches/5.1/handler/ha_innodb.cc
    
    branches/5.1:
    
    Fix Bug#38996 Race condition in ANALYZE TABLE
    
    by serializing ANALYZE TABLE inside InnoDB.
    
    Approved by:	Heikki (rb://175)
    
    ------------------------------------------------------------------------
    r6045 | jyang | 2009-10-08 02:27:08 +0300 (Thu, 08 Oct 2009) | 7 lines
    Changed paths:
       M /branches/5.1/handler/ha_innodb.cc
       A /branches/5.1/mysql-test/innodb_bug47777.result
       A /branches/5.1/mysql-test/innodb_bug47777.test
    
    branches/5.1: Fix bug #47777. Treat the Geometry data same as
    Binary BLOB in ha_innobase::store_key_val_for_row(), since the
    Geometry data is stored as Binary BLOB in Innodb.
    
    Review: rb://180 approved by Marko Makela.
    
    
      ------------------------------------------------------------------------
    r6051 | sunny | 2009-10-12 07:05:00 +0300 (Mon, 12 Oct 2009) | 6 lines
    Changed paths:
       M /branches/5.1/handler/ha_innodb.cc
       M /branches/5.1/mysql-test/innodb-autoinc.result
       M /branches/5.1/mysql-test/innodb-autoinc.test
    
    branches/5.1: Ignore negative values supplied by the user when calculating the
    next value to store in dict_table_t. Setting autoincrement columns top negative
    values is undefined behavior and this change should bring the behavior of
    InnoDB closer to what users expect. Added several tests to check.
    rb://162
    
    ------------------------------------------------------------------------
    r6052 | sunny | 2009-10-12 07:09:56 +0300 (Mon, 12 Oct 2009) | 4 lines
    Changed paths:
       M /branches/5.1/handler/ha_innodb.cc
       M /branches/5.1/mysql-test/innodb-autoinc.result
       M /branches/5.1/mysql-test/innodb-autoinc.test
    
    branches/5.1: Reset the statement level autoinc counter on ROLLBACK. Fix
    the test results too.
    rb://164
    
    ------------------------------------------------------------------------
    r6053 | sunny | 2009-10-12 07:37:49 +0300 (Mon, 12 Oct 2009) | 6 lines
    Changed paths:
       M /branches/5.1/handler/ha_innodb.cc
       M /branches/5.1/mysql-test/innodb-autoinc.result
       M /branches/5.1/mysql-test/innodb-autoinc.test
    
    branches/5.1: Copy the maximum AUTOINC value from the old table to the new
    table when MySQL does a CREATE INDEX ON T. This is required because MySQL
    does a table copy, rename and drops the old table.
    Fix Bug#47125: auto_increment start value is ignored if an index is created and engine=innodb
    rb://168
    
    ------------------------------------------------------------------------
    r6076 | vasil | 2009-10-14 19:30:12 +0300 (Wed, 14 Oct 2009) | 4 lines
    Changed paths:
       M /branches/5.1/row/row0mysql.c
    
    branches/5.1:
      
  Fix typo.
      
    ------------------------------------------------------------------------

  ------------------------------------------------------------------------
  r6114 | vasil | 2009-10-30 03:43:51 +1100 (Fri, 30 Oct 2009) | 6 lines

  branches/zip:

  * Add ChangeLog entries for latest changes
  * Obey alphabetical order in the list of the files
  * White-space fixup

  ------------------------------------------------------------------------
  r6121 | sunny | 2009-10-30 10:42:11 +1100 (Fri, 30 Oct 2009) | 7 lines

  branches/zip: This test has been problematic for sometime now. The underlying
  bug is that the data dictionaries get out of sync. In the AUTOINC code we
  try and apply salve to the symptoms. In the past MySQL made some unrelated
  change and the dictionaries stopped getting out of sync and this test started
  to fail. Now, it seems they have reverted that changed and the test is
  passing again. I suspect this is not he last time that this test will change.

  ------------------------------------------------------------------------
  r6124 | jyang | 2009-10-30 19:02:31 +1100 (Fri, 30 Oct 2009) | 5 lines

  branches/zip: Correct the bug number for -r6109 change
  from # 48273 to #48237



  ------------------------------------------------------------------------
  r6126 | vasil | 2009-10-30 19:36:07 +1100 (Fri, 30 Oct 2009) | 45 lines

  branches/zip: Merge r6112:6125 from branches/5.1:

  (skipping r6122 and r6123, Jimmy says these are already present and need
  not be merged):

    ------------------------------------------------------------------------
    r6122 | jyang | 2009-10-30 05:18:38 +0200 (Fri, 30 Oct 2009) | 7 lines
    Changed paths:
       M /branches/5.1/handler/ha_innodb.cc
       M /branches/5.1/mysql-test/innodb_bug44369.result
       M /branches/5.1/mysql-test/innodb_bug44369.test
       M /branches/5.1/mysql-test/innodb_bug46000.result
       M /branches/5.1/mysql-test/innodb_bug46000.test
    
    branches/5.1: Chnage WARN_LEVEL_ERROR to WARN_LEVEL_WARN
    for push_warning_printf() call in innodb.
    Fix Bug#47233: Innodb calls push_warning(MYSQL_ERROR::WARN_LEVEL_ERROR)
    
    rb://170 approved by Marko.
    
    
    ------------------------------------------------------------------------
    r6123 | jyang | 2009-10-30 05:43:06 +0200 (Fri, 30 Oct 2009) | 8 lines
    Changed paths:
       M /branches/5.1/os/os0proc.c
    
    branches/5.1: In os_mem_alloc_large(), if we fail to attach
    the shared memory, reset memory pointer ptr to NULL, and
    allocate memory from conventional pool. This is a port
    from branches/zip.
    Bug #48237 Error handling in os_mem_alloc_large appears to be incorrect
    rb://198  Approved by: Marko
    
    
    ------------------------------------------------------------------------
    r6125 | vasil | 2009-10-30 10:31:23 +0200 (Fri, 30 Oct 2009) | 4 lines
    Changed paths:
       M /branches/5.1/handler/ha_innodb.cc
    
    branches/5.1:
    
    White-space fixup.
    
    ------------------------------------------------------------------------

  ------------------------------------------------------------------------
  r6130 | marko | 2009-11-02 20:42:56 +1100 (Mon, 02 Nov 2009) | 9 lines

  branches/zip: Free all resources at shutdown. Set pointers to NULL, so
  that Valgrind will not complain about freed data structures that are
  reachable via pointers.  This addresses Bug #45992 and Bug #46656.

  This patch is mostly based on changes copied from branches/embedded-1.0,
  mainly c5432, c3439, c3134, c2994, c2978, but also some other code was
  copied.  Some added cleanup code is specific to MySQL/InnoDB.

  rb://199 approved by Sunny Bains
    ------------------------------------------------------------------------
2009-11-04 06:02:00 +00:00

1436 lines
34 KiB
C

/*****************************************************************************
Copyright (c) 1996, 2009, Innobase Oy. All Rights Reserved.
This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with
this program; if not, write to the Free Software Foundation, Inc., 59 Temple
Place, Suite 330, Boston, MA 02111-1307 USA
*****************************************************************************/
/**************************************************//**
@file que/que0que.c
Query graph
Created 5/27/1996 Heikki Tuuri
*******************************************************/
#include "que0que.h"
#ifdef UNIV_NONINL
#include "que0que.ic"
#endif
#include "srv0que.h"
#include "usr0sess.h"
#include "trx0trx.h"
#include "trx0roll.h"
#include "row0undo.h"
#include "row0ins.h"
#include "row0upd.h"
#include "row0sel.h"
#include "row0purge.h"
#include "dict0crea.h"
#include "log0log.h"
#include "eval0proc.h"
#include "eval0eval.h"
#include "pars0types.h"
#define QUE_PARALLELIZE_LIMIT (64 * 256 * 256 * 256)
#define QUE_ROUND_ROBIN_LIMIT (64 * 256 * 256 * 256)
#define QUE_MAX_LOOPS_WITHOUT_CHECK 16
#ifdef UNIV_DEBUG
/* If the following flag is set TRUE, the module will print trace info
of SQL execution in the UNIV_SQL_DEBUG version */
UNIV_INTERN ibool que_trace_on = FALSE;
#endif /* UNIV_DEBUG */
/* Short introduction to query graphs
==================================
A query graph consists of nodes linked to each other in various ways. The
execution starts at que_run_threads() which takes a que_thr_t parameter.
que_thr_t contains two fields that control query graph execution: run_node
and prev_node. run_node is the next node to execute and prev_node is the
last node executed.
Each node has a pointer to a 'next' statement, i.e., its brother, and a
pointer to its parent node. The next pointer is NULL in the last statement
of a block.
Loop nodes contain a link to the first statement of the enclosed statement
list. While the loop runs, que_thr_step() checks if execution to the loop
node came from its parent or from one of the statement nodes in the loop. If
it came from the parent of the loop node it starts executing the first
statement node in the loop. If it came from one of the statement nodes in
the loop, then it checks if the statement node has another statement node
following it, and runs it if so.
To signify loop ending, the loop statements (see e.g. while_step()) set
que_thr_t->run_node to the loop node's parent node. This is noticed on the
next call of que_thr_step() and execution proceeds to the node pointed to by
the loop node's 'next' pointer.
For example, the code:
X := 1;
WHILE X < 5 LOOP
X := X + 1;
X := X + 1;
X := 5
will result in the following node hierarchy, with the X-axis indicating
'next' links and the Y-axis indicating parent/child links:
A - W - A
|
|
A - A
A = assign_node_t, W = while_node_t. */
/* How a stored procedure containing COMMIT or ROLLBACK commands
is executed?
The commit or rollback can be seen as a subprocedure call.
The problem is that if there are several query threads
currently running within the transaction, their action could
mess the commit or rollback operation. Or, at the least, the
operation would be difficult to visualize and keep in control.
Therefore the query thread requesting a commit or a rollback
sends to the transaction a signal, which moves the transaction
to TRX_QUE_SIGNALED state. All running query threads of the
transaction will eventually notice that the transaction is now in
this state and voluntarily suspend themselves. Only the last
query thread which suspends itself will trigger handling of
the signal.
When the transaction starts to handle a rollback or commit
signal, it builds a query graph which, when executed, will
roll back or commit the incomplete transaction. The transaction
is moved to the TRX_QUE_ROLLING_BACK or TRX_QUE_COMMITTING state.
If specified, the SQL cursors opened by the transaction are closed.
When the execution of the graph completes, it is like returning
from a subprocedure: the query thread which requested the operation
starts running again. */
/**********************************************************************//**
Moves a thread from another state to the QUE_THR_RUNNING state. Increments
the n_active_thrs counters of the query graph and transaction.
***NOTE***: This is the only function in which such a transition is allowed
to happen! */
static
void
que_thr_move_to_run_state(
/*======================*/
que_thr_t* thr); /*!< in: an query thread */
/***********************************************************************//**
Adds a query graph to the session's list of graphs. */
UNIV_INTERN
void
que_graph_publish(
/*==============*/
que_t* graph, /*!< in: graph */
sess_t* sess) /*!< in: session */
{
ut_ad(mutex_own(&kernel_mutex));
UT_LIST_ADD_LAST(graphs, sess->graphs, graph);
}
/***********************************************************************//**
Creates a query graph fork node.
@return own: fork node */
UNIV_INTERN
que_fork_t*
que_fork_create(
/*============*/
que_t* graph, /*!< in: graph, if NULL then this
fork node is assumed to be the
graph root */
que_node_t* parent, /*!< in: parent node */
ulint fork_type, /*!< in: fork type */
mem_heap_t* heap) /*!< in: memory heap where created */
{
que_fork_t* fork;
ut_ad(heap);
fork = mem_heap_alloc(heap, sizeof(que_fork_t));
fork->common.type = QUE_NODE_FORK;
fork->n_active_thrs = 0;
fork->state = QUE_FORK_COMMAND_WAIT;
if (graph != NULL) {
fork->graph = graph;
} else {
fork->graph = fork;
}
fork->common.parent = parent;
fork->fork_type = fork_type;
fork->caller = NULL;
UT_LIST_INIT(fork->thrs);
fork->sym_tab = NULL;
fork->info = NULL;
fork->heap = heap;
return(fork);
}
/***********************************************************************//**
Creates a query graph thread node.
@return own: query thread node */
UNIV_INTERN
que_thr_t*
que_thr_create(
/*===========*/
que_fork_t* parent, /*!< in: parent node, i.e., a fork node */
mem_heap_t* heap) /*!< in: memory heap where created */
{
que_thr_t* thr;
ut_ad(parent && heap);
thr = mem_heap_alloc(heap, sizeof(que_thr_t));
thr->common.type = QUE_NODE_THR;
thr->common.parent = parent;
thr->magic_n = QUE_THR_MAGIC_N;
thr->graph = parent->graph;
thr->state = QUE_THR_COMMAND_WAIT;
thr->is_active = FALSE;
thr->run_node = NULL;
thr->resource = 0;
thr->lock_state = QUE_THR_LOCK_NOLOCK;
UT_LIST_ADD_LAST(thrs, parent->thrs, thr);
return(thr);
}
/**********************************************************************//**
Moves a suspended query thread to the QUE_THR_RUNNING state and may release
a single worker thread to execute it. This function should be used to end
the wait state of a query thread waiting for a lock or a stored procedure
completion. */
UNIV_INTERN
void
que_thr_end_wait(
/*=============*/
que_thr_t* thr, /*!< in: query thread in the
QUE_THR_LOCK_WAIT,
or QUE_THR_PROCEDURE_WAIT, or
QUE_THR_SIG_REPLY_WAIT state */
que_thr_t** next_thr) /*!< in/out: next query thread to run;
if the value which is passed in is
a pointer to a NULL pointer, then the
calling function can start running
a new query thread; if NULL is passed
as the parameter, it is ignored */
{
ibool was_active;
ut_ad(mutex_own(&kernel_mutex));
ut_ad(thr);
ut_ad((thr->state == QUE_THR_LOCK_WAIT)
|| (thr->state == QUE_THR_PROCEDURE_WAIT)
|| (thr->state == QUE_THR_SIG_REPLY_WAIT));
ut_ad(thr->run_node);
thr->prev_node = thr->run_node;
was_active = thr->is_active;
que_thr_move_to_run_state(thr);
if (was_active) {
return;
}
if (next_thr && *next_thr == NULL) {
*next_thr = thr;
} else {
ut_a(0);
srv_que_task_enqueue_low(thr);
}
}
/**********************************************************************//**
Same as que_thr_end_wait, but no parameter next_thr available. */
UNIV_INTERN
void
que_thr_end_wait_no_next_thr(
/*=========================*/
que_thr_t* thr) /*!< in: query thread in the QUE_THR_LOCK_WAIT,
or QUE_THR_PROCEDURE_WAIT, or
QUE_THR_SIG_REPLY_WAIT state */
{
ibool was_active;
ut_a(thr->state == QUE_THR_LOCK_WAIT); /* In MySQL this is the
only possible state here */
ut_ad(mutex_own(&kernel_mutex));
ut_ad(thr);
ut_ad((thr->state == QUE_THR_LOCK_WAIT)
|| (thr->state == QUE_THR_PROCEDURE_WAIT)
|| (thr->state == QUE_THR_SIG_REPLY_WAIT));
was_active = thr->is_active;
que_thr_move_to_run_state(thr);
if (was_active) {
return;
}
/* In MySQL we let the OS thread (not just the query thread) to wait
for the lock to be released: */
srv_release_mysql_thread_if_suspended(thr);
/* srv_que_task_enqueue_low(thr); */
}
/**********************************************************************//**
Inits a query thread for a command. */
UNIV_INLINE
void
que_thr_init_command(
/*=================*/
que_thr_t* thr) /*!< in: query thread */
{
thr->run_node = thr;
thr->prev_node = thr->common.parent;
que_thr_move_to_run_state(thr);
}
/**********************************************************************//**
Starts execution of a command in a query fork. Picks a query thread which
is not in the QUE_THR_RUNNING state and moves it to that state. If none
can be chosen, a situation which may arise in parallelized fetches, NULL
is returned.
@return a query thread of the graph moved to QUE_THR_RUNNING state, or
NULL; the query thread should be executed by que_run_threads by the
caller */
UNIV_INTERN
que_thr_t*
que_fork_start_command(
/*===================*/
que_fork_t* fork) /*!< in: a query fork */
{
que_thr_t* thr;
que_thr_t* suspended_thr = NULL;
que_thr_t* completed_thr = NULL;
fork->state = QUE_FORK_ACTIVE;
fork->last_sel_node = NULL;
suspended_thr = NULL;
completed_thr = NULL;
/* Choose the query thread to run: usually there is just one thread,
but in a parallelized select, which necessarily is non-scrollable,
there may be several to choose from */
/* First we try to find a query thread in the QUE_THR_COMMAND_WAIT
state. Then we try to find a query thread in the QUE_THR_SUSPENDED
state, finally we try to find a query thread in the QUE_THR_COMPLETED
state */
thr = UT_LIST_GET_FIRST(fork->thrs);
/* We make a single pass over the thr list within which we note which
threads are ready to run. */
while (thr) {
switch (thr->state) {
case QUE_THR_COMMAND_WAIT:
/* We have to send the initial message to query thread
to start it */
que_thr_init_command(thr);
return(thr);
case QUE_THR_SUSPENDED:
/* In this case the execution of the thread was
suspended: no initial message is needed because
execution can continue from where it was left */
if (!suspended_thr) {
suspended_thr = thr;
}
break;
case QUE_THR_COMPLETED:
if (!completed_thr) {
completed_thr = thr;
}
break;
case QUE_THR_LOCK_WAIT:
ut_error;
}
thr = UT_LIST_GET_NEXT(thrs, thr);
}
if (suspended_thr) {
thr = suspended_thr;
que_thr_move_to_run_state(thr);
} else if (completed_thr) {
thr = completed_thr;
que_thr_init_command(thr);
}
return(thr);
}
/**********************************************************************//**
After signal handling is finished, returns control to a query graph error
handling routine. (Currently, just returns the control to the root of the
graph so that the graph can communicate an error message to the client.) */
UNIV_INTERN
void
que_fork_error_handle(
/*==================*/
trx_t* trx __attribute__((unused)), /*!< in: trx */
que_t* fork) /*!< in: query graph which was run before signal
handling started, NULL not allowed */
{
que_thr_t* thr;
ut_ad(mutex_own(&kernel_mutex));
ut_ad(trx->sess->state == SESS_ERROR);
ut_ad(UT_LIST_GET_LEN(trx->reply_signals) == 0);
ut_ad(UT_LIST_GET_LEN(trx->wait_thrs) == 0);
thr = UT_LIST_GET_FIRST(fork->thrs);
while (thr != NULL) {
ut_ad(!thr->is_active);
ut_ad(thr->state != QUE_THR_SIG_REPLY_WAIT);
ut_ad(thr->state != QUE_THR_LOCK_WAIT);
thr->run_node = thr;
thr->prev_node = thr->child;
thr->state = QUE_THR_COMPLETED;
thr = UT_LIST_GET_NEXT(thrs, thr);
}
thr = UT_LIST_GET_FIRST(fork->thrs);
que_thr_move_to_run_state(thr);
ut_a(0);
srv_que_task_enqueue_low(thr);
}
/****************************************************************//**
Tests if all the query threads in the same fork have a given state.
@return TRUE if all the query threads in the same fork were in the
given state */
UNIV_INLINE
ibool
que_fork_all_thrs_in_state(
/*=======================*/
que_fork_t* fork, /*!< in: query fork */
ulint state) /*!< in: state */
{
que_thr_t* thr_node;
thr_node = UT_LIST_GET_FIRST(fork->thrs);
while (thr_node != NULL) {
if (thr_node->state != state) {
return(FALSE);
}
thr_node = UT_LIST_GET_NEXT(thrs, thr_node);
}
return(TRUE);
}
/**********************************************************************//**
Calls que_graph_free_recursive for statements in a statement list. */
static
void
que_graph_free_stat_list(
/*=====================*/
que_node_t* node) /*!< in: first query graph node in the list */
{
while (node) {
que_graph_free_recursive(node);
node = que_node_get_next(node);
}
}
/**********************************************************************//**
Frees a query graph, but not the heap where it was created. Does not free
explicit cursor declarations, they are freed in que_graph_free. */
UNIV_INTERN
void
que_graph_free_recursive(
/*=====================*/
que_node_t* node) /*!< in: query graph node */
{
que_fork_t* fork;
que_thr_t* thr;
undo_node_t* undo;
sel_node_t* sel;
ins_node_t* ins;
upd_node_t* upd;
tab_node_t* cre_tab;
ind_node_t* cre_ind;
purge_node_t* purge;
if (node == NULL) {
return;
}
switch (que_node_get_type(node)) {
case QUE_NODE_FORK:
fork = node;
thr = UT_LIST_GET_FIRST(fork->thrs);
while (thr) {
que_graph_free_recursive(thr);
thr = UT_LIST_GET_NEXT(thrs, thr);
}
break;
case QUE_NODE_THR:
thr = node;
if (thr->magic_n != QUE_THR_MAGIC_N) {
fprintf(stderr,
"que_thr struct appears corrupt;"
" magic n %lu\n",
(unsigned long) thr->magic_n);
mem_analyze_corruption(thr);
ut_error;
}
thr->magic_n = QUE_THR_MAGIC_FREED;
que_graph_free_recursive(thr->child);
break;
case QUE_NODE_UNDO:
undo = node;
mem_heap_free(undo->heap);
break;
case QUE_NODE_SELECT:
sel = node;
sel_node_free_private(sel);
break;
case QUE_NODE_INSERT:
ins = node;
que_graph_free_recursive(ins->select);
mem_heap_free(ins->entry_sys_heap);
break;
case QUE_NODE_PURGE:
purge = node;
mem_heap_free(purge->heap);
break;
case QUE_NODE_UPDATE:
upd = node;
if (upd->in_mysql_interface) {
btr_pcur_free_for_mysql(upd->pcur);
}
que_graph_free_recursive(upd->cascade_node);
if (upd->cascade_heap) {
mem_heap_free(upd->cascade_heap);
}
que_graph_free_recursive(upd->select);
mem_heap_free(upd->heap);
break;
case QUE_NODE_CREATE_TABLE:
cre_tab = node;
que_graph_free_recursive(cre_tab->tab_def);
que_graph_free_recursive(cre_tab->col_def);
que_graph_free_recursive(cre_tab->commit_node);
mem_heap_free(cre_tab->heap);
break;
case QUE_NODE_CREATE_INDEX:
cre_ind = node;
que_graph_free_recursive(cre_ind->ind_def);
que_graph_free_recursive(cre_ind->field_def);
que_graph_free_recursive(cre_ind->commit_node);
mem_heap_free(cre_ind->heap);
break;
case QUE_NODE_PROC:
que_graph_free_stat_list(((proc_node_t*)node)->stat_list);
break;
case QUE_NODE_IF:
que_graph_free_stat_list(((if_node_t*)node)->stat_list);
que_graph_free_stat_list(((if_node_t*)node)->else_part);
que_graph_free_stat_list(((if_node_t*)node)->elsif_list);
break;
case QUE_NODE_ELSIF:
que_graph_free_stat_list(((elsif_node_t*)node)->stat_list);
break;
case QUE_NODE_WHILE:
que_graph_free_stat_list(((while_node_t*)node)->stat_list);
break;
case QUE_NODE_FOR:
que_graph_free_stat_list(((for_node_t*)node)->stat_list);
break;
case QUE_NODE_ASSIGNMENT:
case QUE_NODE_EXIT:
case QUE_NODE_RETURN:
case QUE_NODE_COMMIT:
case QUE_NODE_ROLLBACK:
case QUE_NODE_LOCK:
case QUE_NODE_FUNC:
case QUE_NODE_ORDER:
case QUE_NODE_ROW_PRINTF:
case QUE_NODE_OPEN:
case QUE_NODE_FETCH:
/* No need to do anything */
break;
default:
fprintf(stderr,
"que_node struct appears corrupt; type %lu\n",
(unsigned long) que_node_get_type(node));
mem_analyze_corruption(node);
ut_error;
}
}
/**********************************************************************//**
Frees a query graph. */
UNIV_INTERN
void
que_graph_free(
/*===========*/
que_t* graph) /*!< in: query graph; we assume that the memory
heap where this graph was created is private
to this graph: if not, then use
que_graph_free_recursive and free the heap
afterwards! */
{
ut_ad(graph);
if (graph->sym_tab) {
/* The following call frees dynamic memory allocated
for variables etc. during execution. Frees also explicit
cursor definitions. */
sym_tab_free_private(graph->sym_tab);
}
if (graph->info && graph->info->graph_owns_us) {
pars_info_free(graph->info);
}
que_graph_free_recursive(graph);
mem_heap_free(graph->heap);
}
/****************************************************************//**
Performs an execution step on a thr node.
@return query thread to run next, or NULL if none */
static
que_thr_t*
que_thr_node_step(
/*==============*/
que_thr_t* thr) /*!< in: query thread where run_node must
be the thread node itself */
{
ut_ad(thr->run_node == thr);
if (thr->prev_node == thr->common.parent) {
/* If control to the node came from above, it is just passed
on */
thr->run_node = thr->child;
return(thr);
}
mutex_enter(&kernel_mutex);
if (que_thr_peek_stop(thr)) {
mutex_exit(&kernel_mutex);
return(thr);
}
/* Thread execution completed */
thr->state = QUE_THR_COMPLETED;
mutex_exit(&kernel_mutex);
return(NULL);
}
/**********************************************************************//**
Moves a thread from another state to the QUE_THR_RUNNING state. Increments
the n_active_thrs counters of the query graph and transaction if thr was
not active.
***NOTE***: This and ..._mysql are the only functions in which such a
transition is allowed to happen! */
static
void
que_thr_move_to_run_state(
/*======================*/
que_thr_t* thr) /*!< in: an query thread */
{
trx_t* trx;
ut_ad(thr->state != QUE_THR_RUNNING);
trx = thr_get_trx(thr);
if (!thr->is_active) {
(thr->graph)->n_active_thrs++;
trx->n_active_thrs++;
thr->is_active = TRUE;
ut_ad((thr->graph)->n_active_thrs == 1);
ut_ad(trx->n_active_thrs == 1);
}
thr->state = QUE_THR_RUNNING;
}
/**********************************************************************//**
Decrements the query thread reference counts in the query graph and the
transaction. May start signal handling, e.g., a rollback.
*** NOTE ***:
This and que_thr_stop_for_mysql are the only functions where the reference
count can be decremented and this function may only be called from inside
que_run_threads or que_thr_check_if_switch! These restrictions exist to make
the rollback code easier to maintain. */
static
void
que_thr_dec_refer_count(
/*====================*/
que_thr_t* thr, /*!< in: query thread */
que_thr_t** next_thr) /*!< in/out: next query thread to run;
if the value which is passed in is
a pointer to a NULL pointer, then the
calling function can start running
a new query thread */
{
que_fork_t* fork;
trx_t* trx;
ulint fork_type;
ibool stopped;
fork = thr->common.parent;
trx = thr_get_trx(thr);
mutex_enter(&kernel_mutex);
ut_a(thr->is_active);
if (thr->state == QUE_THR_RUNNING) {
stopped = que_thr_stop(thr);
if (!stopped) {
/* The reason for the thr suspension or wait was
already canceled before we came here: continue
running the thread */
/* fputs("!!!!!!!! Wait already ended: continue thr\n",
stderr); */
if (next_thr && *next_thr == NULL) {
/* Normally srv_suspend_mysql_thread resets
the state to DB_SUCCESS before waiting, but
in this case we have to do it here,
otherwise nobody does it. */
trx->error_state = DB_SUCCESS;
*next_thr = thr;
} else {
ut_error;
srv_que_task_enqueue_low(thr);
}
mutex_exit(&kernel_mutex);
return;
}
}
ut_ad(fork->n_active_thrs == 1);
ut_ad(trx->n_active_thrs == 1);
fork->n_active_thrs--;
trx->n_active_thrs--;
thr->is_active = FALSE;
if (trx->n_active_thrs > 0) {
mutex_exit(&kernel_mutex);
return;
}
fork_type = fork->fork_type;
/* Check if all query threads in the same fork are completed */
if (que_fork_all_thrs_in_state(fork, QUE_THR_COMPLETED)) {
switch (fork_type) {
case QUE_FORK_ROLLBACK:
/* This is really the undo graph used in rollback,
no roll_node in this graph */
ut_ad(UT_LIST_GET_LEN(trx->signals) > 0);
ut_ad(trx->handling_signals == TRUE);
trx_finish_rollback_off_kernel(fork, trx, next_thr);
break;
case QUE_FORK_PURGE:
case QUE_FORK_RECOVERY:
case QUE_FORK_MYSQL_INTERFACE:
/* Do nothing */
break;
default:
ut_error; /*!< not used in MySQL */
}
}
if (UT_LIST_GET_LEN(trx->signals) > 0 && trx->n_active_thrs == 0) {
/* If the trx is signaled and its query thread count drops to
zero, then we start processing a signal; from it we may get
a new query thread to run */
trx_sig_start_handle(trx, next_thr);
}
if (trx->handling_signals && UT_LIST_GET_LEN(trx->signals) == 0) {
trx_end_signal_handling(trx);
}
mutex_exit(&kernel_mutex);
}
/**********************************************************************//**
Stops a query thread if graph or trx is in a state requiring it. The
conditions are tested in the order (1) graph, (2) trx. The kernel mutex has
to be reserved.
@return TRUE if stopped */
UNIV_INTERN
ibool
que_thr_stop(
/*=========*/
que_thr_t* thr) /*!< in: query thread */
{
trx_t* trx;
que_t* graph;
ibool ret = TRUE;
ut_ad(mutex_own(&kernel_mutex));
graph = thr->graph;
trx = graph->trx;
if (graph->state == QUE_FORK_COMMAND_WAIT) {
thr->state = QUE_THR_SUSPENDED;
} else if (trx->que_state == TRX_QUE_LOCK_WAIT) {
UT_LIST_ADD_FIRST(trx_thrs, trx->wait_thrs, thr);
thr->state = QUE_THR_LOCK_WAIT;
} else if (trx->error_state != DB_SUCCESS
&& trx->error_state != DB_LOCK_WAIT) {
/* Error handling built for the MySQL interface */
thr->state = QUE_THR_COMPLETED;
} else if (UT_LIST_GET_LEN(trx->signals) > 0
&& graph->fork_type != QUE_FORK_ROLLBACK) {
thr->state = QUE_THR_SUSPENDED;
} else {
ut_ad(graph->state == QUE_FORK_ACTIVE);
ret = FALSE;
}
return(ret);
}
/**********************************************************************//**
A patch for MySQL used to 'stop' a dummy query thread used in MySQL. The
query thread is stopped and made inactive, except in the case where
it was put to the lock wait state in lock0lock.c, but the lock has already
been granted or the transaction chosen as a victim in deadlock resolution. */
UNIV_INTERN
void
que_thr_stop_for_mysql(
/*===================*/
que_thr_t* thr) /*!< in: query thread */
{
trx_t* trx;
trx = thr_get_trx(thr);
mutex_enter(&kernel_mutex);
if (thr->state == QUE_THR_RUNNING) {
if (trx->error_state != DB_SUCCESS
&& trx->error_state != DB_LOCK_WAIT) {
/* Error handling built for the MySQL interface */
thr->state = QUE_THR_COMPLETED;
} else {
/* It must have been a lock wait but the lock was
already released, or this transaction was chosen
as a victim in selective deadlock resolution */
mutex_exit(&kernel_mutex);
return;
}
}
ut_ad(thr->is_active == TRUE);
ut_ad(trx->n_active_thrs == 1);
ut_ad(thr->graph->n_active_thrs == 1);
thr->is_active = FALSE;
(thr->graph)->n_active_thrs--;
trx->n_active_thrs--;
mutex_exit(&kernel_mutex);
}
/**********************************************************************//**
Moves a thread from another state to the QUE_THR_RUNNING state. Increments
the n_active_thrs counters of the query graph and transaction if thr was
not active. */
UNIV_INTERN
void
que_thr_move_to_run_state_for_mysql(
/*================================*/
que_thr_t* thr, /*!< in: an query thread */
trx_t* trx) /*!< in: transaction */
{
if (thr->magic_n != QUE_THR_MAGIC_N) {
fprintf(stderr,
"que_thr struct appears corrupt; magic n %lu\n",
(unsigned long) thr->magic_n);
mem_analyze_corruption(thr);
ut_error;
}
if (!thr->is_active) {
thr->graph->n_active_thrs++;
trx->n_active_thrs++;
thr->is_active = TRUE;
}
thr->state = QUE_THR_RUNNING;
}
/**********************************************************************//**
A patch for MySQL used to 'stop' a dummy query thread used in MySQL
select, when there is no error or lock wait. */
UNIV_INTERN
void
que_thr_stop_for_mysql_no_error(
/*============================*/
que_thr_t* thr, /*!< in: query thread */
trx_t* trx) /*!< in: transaction */
{
ut_ad(thr->state == QUE_THR_RUNNING);
ut_ad(thr->is_active == TRUE);
ut_ad(trx->n_active_thrs == 1);
ut_ad(thr->graph->n_active_thrs == 1);
if (thr->magic_n != QUE_THR_MAGIC_N) {
fprintf(stderr,
"que_thr struct appears corrupt; magic n %lu\n",
(unsigned long) thr->magic_n);
mem_analyze_corruption(thr);
ut_error;
}
thr->state = QUE_THR_COMPLETED;
thr->is_active = FALSE;
(thr->graph)->n_active_thrs--;
trx->n_active_thrs--;
}
/****************************************************************//**
Get the first containing loop node (e.g. while_node_t or for_node_t) for the
given node, or NULL if the node is not within a loop.
@return containing loop node, or NULL. */
UNIV_INTERN
que_node_t*
que_node_get_containing_loop_node(
/*==============================*/
que_node_t* node) /*!< in: node */
{
ut_ad(node);
for (;;) {
ulint type;
node = que_node_get_parent(node);
if (!node) {
break;
}
type = que_node_get_type(node);
if ((type == QUE_NODE_FOR) || (type == QUE_NODE_WHILE)) {
break;
}
}
return(node);
}
/**********************************************************************//**
Prints info of an SQL query graph node. */
UNIV_INTERN
void
que_node_print_info(
/*================*/
que_node_t* node) /*!< in: query graph node */
{
ulint type;
const char* str;
type = que_node_get_type(node);
if (type == QUE_NODE_SELECT) {
str = "SELECT";
} else if (type == QUE_NODE_INSERT) {
str = "INSERT";
} else if (type == QUE_NODE_UPDATE) {
str = "UPDATE";
} else if (type == QUE_NODE_WHILE) {
str = "WHILE";
} else if (type == QUE_NODE_ASSIGNMENT) {
str = "ASSIGNMENT";
} else if (type == QUE_NODE_IF) {
str = "IF";
} else if (type == QUE_NODE_FETCH) {
str = "FETCH";
} else if (type == QUE_NODE_OPEN) {
str = "OPEN";
} else if (type == QUE_NODE_PROC) {
str = "STORED PROCEDURE";
} else if (type == QUE_NODE_FUNC) {
str = "FUNCTION";
} else if (type == QUE_NODE_LOCK) {
str = "LOCK";
} else if (type == QUE_NODE_THR) {
str = "QUERY THREAD";
} else if (type == QUE_NODE_COMMIT) {
str = "COMMIT";
} else if (type == QUE_NODE_UNDO) {
str = "UNDO ROW";
} else if (type == QUE_NODE_PURGE) {
str = "PURGE ROW";
} else if (type == QUE_NODE_ROLLBACK) {
str = "ROLLBACK";
} else if (type == QUE_NODE_CREATE_TABLE) {
str = "CREATE TABLE";
} else if (type == QUE_NODE_CREATE_INDEX) {
str = "CREATE INDEX";
} else if (type == QUE_NODE_FOR) {
str = "FOR LOOP";
} else if (type == QUE_NODE_RETURN) {
str = "RETURN";
} else if (type == QUE_NODE_EXIT) {
str = "EXIT";
} else {
str = "UNKNOWN NODE TYPE";
}
fprintf(stderr, "Node type %lu: %s, address %p\n",
(ulong) type, str, (void*) node);
}
/**********************************************************************//**
Performs an execution step on a query thread.
@return query thread to run next: it may differ from the input
parameter if, e.g., a subprocedure call is made */
UNIV_INLINE
que_thr_t*
que_thr_step(
/*=========*/
que_thr_t* thr) /*!< in: query thread */
{
que_node_t* node;
que_thr_t* old_thr;
trx_t* trx;
ulint type;
trx = thr_get_trx(thr);
ut_ad(thr->state == QUE_THR_RUNNING);
ut_a(trx->error_state == DB_SUCCESS);
thr->resource++;
node = thr->run_node;
type = que_node_get_type(node);
old_thr = thr;
#ifdef UNIV_DEBUG
if (que_trace_on) {
fputs("To execute: ", stderr);
que_node_print_info(node);
}
#endif
if (type & QUE_NODE_CONTROL_STAT) {
if ((thr->prev_node != que_node_get_parent(node))
&& que_node_get_next(thr->prev_node)) {
/* The control statements, like WHILE, always pass the
control to the next child statement if there is any
child left */
thr->run_node = que_node_get_next(thr->prev_node);
} else if (type == QUE_NODE_IF) {
if_step(thr);
} else if (type == QUE_NODE_FOR) {
for_step(thr);
} else if (type == QUE_NODE_PROC) {
/* We can access trx->undo_no without reserving
trx->undo_mutex, because there cannot be active query
threads doing updating or inserting at the moment! */
if (thr->prev_node == que_node_get_parent(node)) {
trx->last_sql_stat_start.least_undo_no
= trx->undo_no;
}
proc_step(thr);
} else if (type == QUE_NODE_WHILE) {
while_step(thr);
} else {
ut_error;
}
} else if (type == QUE_NODE_ASSIGNMENT) {
assign_step(thr);
} else if (type == QUE_NODE_SELECT) {
thr = row_sel_step(thr);
} else if (type == QUE_NODE_INSERT) {
thr = row_ins_step(thr);
} else if (type == QUE_NODE_UPDATE) {
thr = row_upd_step(thr);
} else if (type == QUE_NODE_FETCH) {
thr = fetch_step(thr);
} else if (type == QUE_NODE_OPEN) {
thr = open_step(thr);
} else if (type == QUE_NODE_FUNC) {
proc_eval_step(thr);
} else if (type == QUE_NODE_LOCK) {
ut_error;
/*
thr = que_lock_step(thr);
*/
} else if (type == QUE_NODE_THR) {
thr = que_thr_node_step(thr);
} else if (type == QUE_NODE_COMMIT) {
thr = trx_commit_step(thr);
} else if (type == QUE_NODE_UNDO) {
thr = row_undo_step(thr);
} else if (type == QUE_NODE_PURGE) {
thr = row_purge_step(thr);
} else if (type == QUE_NODE_RETURN) {
thr = return_step(thr);
} else if (type == QUE_NODE_EXIT) {
thr = exit_step(thr);
} else if (type == QUE_NODE_ROLLBACK) {
thr = trx_rollback_step(thr);
} else if (type == QUE_NODE_CREATE_TABLE) {
thr = dict_create_table_step(thr);
} else if (type == QUE_NODE_CREATE_INDEX) {
thr = dict_create_index_step(thr);
} else if (type == QUE_NODE_ROW_PRINTF) {
thr = row_printf_step(thr);
} else {
ut_error;
}
if (type == QUE_NODE_EXIT) {
old_thr->prev_node = que_node_get_containing_loop_node(node);
} else {
old_thr->prev_node = node;
}
if (thr) {
ut_a(thr_get_trx(thr)->error_state == DB_SUCCESS);
}
return(thr);
}
/**********************************************************************//**
Run a query thread until it finishes or encounters e.g. a lock wait. */
static
void
que_run_threads_low(
/*================*/
que_thr_t* thr) /*!< in: query thread */
{
que_thr_t* next_thr;
ulint cumul_resource;
ulint loop_count;
ut_ad(thr->state == QUE_THR_RUNNING);
ut_a(thr_get_trx(thr)->error_state == DB_SUCCESS);
ut_ad(!mutex_own(&kernel_mutex));
/* cumul_resource counts how much resources the OS thread (NOT the
query thread) has spent in this function */
loop_count = QUE_MAX_LOOPS_WITHOUT_CHECK;
cumul_resource = 0;
loop:
/* Check that there is enough space in the log to accommodate
possible log entries by this query step; if the operation can touch
more than about 4 pages, checks must be made also within the query
step! */
log_free_check();
/* Perform the actual query step: note that the query thread
may change if, e.g., a subprocedure call is made */
/*-------------------------*/
next_thr = que_thr_step(thr);
/*-------------------------*/
ut_a(!next_thr || (thr_get_trx(next_thr)->error_state == DB_SUCCESS));
loop_count++;
if (next_thr != thr) {
ut_a(next_thr == NULL);
/* This can change next_thr to a non-NULL value if there was
a lock wait that already completed. */
que_thr_dec_refer_count(thr, &next_thr);
if (next_thr == NULL) {
return;
}
loop_count = QUE_MAX_LOOPS_WITHOUT_CHECK;
thr = next_thr;
}
goto loop;
}
/**********************************************************************//**
Run a query thread. Handles lock waits. */
UNIV_INTERN
void
que_run_threads(
/*============*/
que_thr_t* thr) /*!< in: query thread */
{
loop:
ut_a(thr_get_trx(thr)->error_state == DB_SUCCESS);
que_run_threads_low(thr);
mutex_enter(&kernel_mutex);
switch (thr->state) {
case QUE_THR_RUNNING:
/* There probably was a lock wait, but it already ended
before we came here: continue running thr */
mutex_exit(&kernel_mutex);
goto loop;
case QUE_THR_LOCK_WAIT:
mutex_exit(&kernel_mutex);
/* The ..._mysql_... function works also for InnoDB's
internal threads. Let us wait that the lock wait ends. */
srv_suspend_mysql_thread(thr);
if (thr_get_trx(thr)->error_state != DB_SUCCESS) {
/* thr was chosen as a deadlock victim or there was
a lock wait timeout */
que_thr_dec_refer_count(thr, NULL);
return;
}
goto loop;
case QUE_THR_COMPLETED:
case QUE_THR_COMMAND_WAIT:
/* Do nothing */
break;
default:
ut_error;
}
mutex_exit(&kernel_mutex);
}
/*********************************************************************//**
Evaluate the given SQL.
@return error code or DB_SUCCESS */
UNIV_INTERN
ulint
que_eval_sql(
/*=========*/
pars_info_t* info, /*!< in: info struct, or NULL */
const char* sql, /*!< in: SQL string */
ibool reserve_dict_mutex,
/*!< in: if TRUE, acquire/release
dict_sys->mutex around call to pars_sql. */
trx_t* trx) /*!< in: trx */
{
que_thr_t* thr;
que_t* graph;
ut_a(trx->error_state == DB_SUCCESS);
if (reserve_dict_mutex) {
mutex_enter(&dict_sys->mutex);
}
graph = pars_sql(info, sql);
if (reserve_dict_mutex) {
mutex_exit(&dict_sys->mutex);
}
ut_a(graph);
graph->trx = trx;
trx->graph = NULL;
graph->fork_type = QUE_FORK_MYSQL_INTERFACE;
ut_a(thr = que_fork_start_command(graph));
que_run_threads(thr);
que_graph_free(graph);
return(trx->error_state);
}