Applying InnoDB Plugin 1.0.5 snapshot, part 7

From revisions r5792 to r5864

Detailed revision comments:

r5792 | vasil | 2009-09-09 08:35:58 -0500 (Wed, 09 Sep 2009) | 32 lines
branches/zip:

Fix a bug in manipulating the variable innodb_old_blocks_pct:

for any value assigned it got that value -1, except for 75. When
assigned 75, it got 75.

  mysql> set global innodb_old_blocks_pct=15;
  Query OK, 0 rows affected (0.00 sec)
  
  mysql> show variables like 'innodb_old_blocks_pct';
  +-----------------------+-------+
  | Variable_name         | Value |
  +-----------------------+-------+
  | innodb_old_blocks_pct | 14    | 
  +-----------------------+-------+
  1 row in set (0.00 sec)
  
  mysql> set global innodb_old_blocks_pct=75;
  Query OK, 0 rows affected (0.00 sec)
  
  mysql> show variables like 'innodb_old_blocks_pct';
  +-----------------------+-------+
  | Variable_name         | Value |
  +-----------------------+-------+
  | innodb_old_blocks_pct | 75    | 
  +-----------------------+-------+

After the fix it gets exactly what was assigned.

Approved by:	Marko (via IM)

r5798 | calvin | 2009-09-09 10:28:10 -0500 (Wed, 09 Sep 2009) | 5 lines
branches/zip:

HA_ERR_TOO_MANY_CONCURRENT_TRXS is added in 5.1.38.
But the plugin should still work with previous versions
of MySQL.
r5804 | marko | 2009-09-10 00:29:31 -0500 (Thu, 10 Sep 2009) | 1 line
branches/zip: trx_cleanup_at_db_startup(): Fix a typo in comment.
r5822 | marko | 2009-09-10 05:10:20 -0500 (Thu, 10 Sep 2009) | 1 line
branches/zip: buf_page_release(): De-stutter the function comment.
r5825 | marko | 2009-09-10 05:47:09 -0500 (Thu, 10 Sep 2009) | 20 lines
branches/zip: Reduce mutex contention that was introduced when
addressing Bug #45015 (Issue #316), in r5703.

buf_page_set_accessed_make_young(): New auxiliary function, called by
buf_page_get_zip(), buf_page_get_gen(),
buf_page_optimistic_get_func(). Call ut_time_ms() outside of
buf_pool_mutex. Use cached access_time.

buf_page_set_accessed(): Add the parameter time_ms, so that
ut_time_ms() need not be called while holding buf_pool_mutex.

buf_page_optimistic_get_func(), buf_page_get_known_nowait(): Read
buf_page_t::access_time without holding buf_pool_mutex. This should be
OK, because the field is only used for heuristic purposes.

buf_page_peek_if_too_old(): If buf_pool->freed_page_clock == 0, return
FALSE, so that we will not waste time moving blocks in the LRU list in
the warm-up phase or when the workload fits in the buffer pool.

rb://156 approved by Sunny Bains
r5826 | marko | 2009-09-10 06:29:46 -0500 (Thu, 10 Sep 2009) | 12 lines
branches/zip: Roll back recovered dictionary transactions before
dropping incomplete indexes (Issue #337).

trx_rollback_or_clean_recovered(ibool all): New function, split from
trx_rollback_or_clean_all_recovered().  all==FALSE will only roll back
dictionary transactions.

recv_recovery_from_checkpoint_finish(): Call
trx_rollback_or_clean_recovered(FALSE) before
row_merge_drop_temp_indexes().

rb://158 approved by Sunny Bains
r5858 | vasil | 2009-09-11 12:46:47 -0500 (Fri, 11 Sep 2009) | 4 lines
branches/zip:

Fix the indentation of the closing bracket.

r5863 | vasil | 2009-09-12 02:07:08 -0500 (Sat, 12 Sep 2009) | 10 lines
branches/zip:

Check that pthread_t can indeed be passed to Solaris atomic functions, instead
of assuming that it can be passed if 0 can be assigned to it. It could be that:
* 0 can be assigned, but pthread_t cannot be passed and
* 0 cannot be assigned but pthread_t can be passed

Better to check what we are interested in, not something else and make
assumptions.

r5864 | vasil | 2009-09-12 02:22:55 -0500 (Sat, 12 Sep 2009) | 4 lines
branches/zip:

Include string.h which is needed for memset().

storage/innodb_plugin/buf/buf0buf.c

@@ -1489,7 +1489,7 @@ buf_pool_resize(void)
 
 /********************************************************************//**
 Moves a page to the start of the buffer pool LRU list. This high-level
-function can be used to prevent an important page from from slipping out of
+function can be used to prevent an important page from slipping out of
 the buffer pool. */
 UNIV_INTERN
 void
@@ -1506,6 +1506,36 @@ buf_page_make_young(
 	buf_pool_mutex_exit();
 }
 
+/********************************************************************//**
+Sets the time of the first access of a page and moves a page to the
+start of the buffer pool LRU list if it is too old.  This high-level
+function can be used to prevent an important page from slipping
+out of the buffer pool. */
+static
+void
+buf_page_set_accessed_make_young(
+/*=============================*/
+	buf_page_t*	bpage,		/*!< in/out: buffer block of a
+					file page */
+	unsigned	access_time)	/*!< in: bpage->access_time
+					read under mutex protection,
+					or 0 if unknown */
+{
+	ut_ad(!buf_pool_mutex_own());
+	ut_a(buf_page_in_file(bpage));
+
+	if (buf_page_peek_if_too_old(bpage)) {
+		buf_pool_mutex_enter();
+		buf_LRU_make_block_young(bpage);
+		buf_pool_mutex_exit();
+	} else if (!access_time) {
+		ulint	time_ms = ut_time_ms();
+		buf_pool_mutex_enter();
+		buf_page_set_accessed(bpage, time_ms);
+		buf_pool_mutex_exit();
+	}
+}
+
 /********************************************************************//**
 Resets the check_index_page_at_flush field of a page if found in the buffer
 pool. */
@@ -1637,6 +1667,7 @@ buf_page_get_zip(
 	buf_page_t*	bpage;
 	mutex_t*	block_mutex;
 	ibool		must_read;
+	unsigned	access_time;
 
 #ifndef UNIV_LOG_DEBUG
 	ut_ad(!ibuf_inside());
@@ -1704,17 +1735,14 @@ err_exit:
 
 got_block:
 	must_read = buf_page_get_io_fix(bpage) == BUF_IO_READ;
-
-	if (buf_page_peek_if_too_old(bpage)) {
-		buf_LRU_make_block_young(bpage);
-	}
-
-	buf_page_set_accessed(bpage);
+	access_time = buf_page_is_accessed(bpage);
 
 	buf_pool_mutex_exit();
 
 	mutex_exit(block_mutex);
 
+	buf_page_set_accessed_make_young(bpage, access_time);
+
 #ifdef UNIV_DEBUG_FILE_ACCESSES
 	ut_a(!bpage->file_page_was_freed);
 #endif
@@ -2244,14 +2272,10 @@ wait_until_unfixed:
 
 	access_time = buf_page_is_accessed(&block->page);
 
-	if (buf_page_peek_if_too_old(&block->page)) {
-		buf_LRU_make_block_young(&block->page);
-	}
-
-	buf_page_set_accessed(&block->page);
-
 	buf_pool_mutex_exit();
 
+	buf_page_set_accessed_make_young(&block->page, access_time);
+
 #ifdef UNIV_DEBUG_FILE_ACCESSES
 	ut_a(!block->page.file_page_was_freed);
 #endif
@@ -2353,18 +2377,13 @@ buf_page_optimistic_get_func(
 
 	mutex_exit(&block->mutex);
 
-	buf_pool_mutex_enter();
+	/* Check if this is the first access to the page.
+	We do a dirty read on purpose, to avoid mutex contention.
+	This field is only used for heuristic purposes; it does not
+	affect correctness. */
 
-	/* Check if this is the first access to the page */
 	access_time = buf_page_is_accessed(&block->page);
-
-	if (buf_page_peek_if_too_old(&block->page)) {
-		buf_LRU_make_block_young(&block->page);
-	}
-
-	buf_page_set_accessed(&block->page);
-
-	buf_pool_mutex_exit();
+	buf_page_set_accessed_make_young(&block->page, access_time);
 
 	ut_ad(!ibuf_inside()
 	      || ibuf_page(buf_block_get_space(block),
@@ -2477,15 +2496,21 @@ buf_page_get_known_nowait(
 
 	mutex_exit(&block->mutex);
 
-	buf_pool_mutex_enter();
-
 	if (mode == BUF_MAKE_YOUNG && buf_page_peek_if_too_old(&block->page)) {
+		buf_pool_mutex_enter();
 		buf_LRU_make_block_young(&block->page);
-	}
-
-	buf_page_set_accessed(&block->page);
+		buf_pool_mutex_exit();
+	} else if (!buf_page_is_accessed(&block->page)) {
+		/* Above, we do a dirty read on purpose, to avoid
+		mutex contention.  The field buf_page_t::access_time
+		is only used for heuristic purposes.  Writes to the
+		field must be protected by mutex, however. */
+		ulint	time_ms = ut_time_ms();
 
-	buf_pool_mutex_exit();
+		buf_pool_mutex_enter();
+		buf_page_set_accessed(&block->page, time_ms);
+		buf_pool_mutex_exit();
+	}
 
 	ut_ad(!ibuf_inside() || (mode == BUF_KEEP_OLD));
 
@@ -2917,6 +2942,7 @@ buf_page_create(
 	buf_frame_t*	frame;
 	buf_block_t*	block;
 	buf_block_t*	free_block	= NULL;
+	ulint		time_ms		= ut_time_ms();
 
 	ut_ad(mtr);
 	ut_ad(space || !zip_size);
@@ -3000,7 +3026,7 @@ buf_page_create(
 		rw_lock_x_unlock(&block->lock);
 	}
 
-	buf_page_set_accessed(&block->page);
+	buf_page_set_accessed(&block->page, time_ms);
 
 	buf_pool_mutex_exit();
 

storage/innodb_plugin/buf/buf0lru.c

@@ -1866,7 +1866,9 @@ buf_LRU_old_ratio_update(
 		buf_LRU_old_ratio = ratio;
 	}
 
-	return(ratio * 100 / BUF_LRU_OLD_RATIO_DIV);
+	/* the reverse of 
+	ratio = old_pct * BUF_LRU_OLD_RATIO_DIV / 100 */
+	return((uint) (ratio * 100 / (double) BUF_LRU_OLD_RATIO_DIV + 0.5));
 }
 
 /********************************************************************//**

storage/innodb_plugin/handler/ha_innodb.cc

@@ -868,17 +868,14 @@ convert_error_code_to_mysql(
 		return(ER_PRIMARY_CANT_HAVE_NULL);
 
 	case DB_TOO_MANY_CONCURRENT_TRXS:
-		/* Once MySQL add the appropriate code to errmsg.txt then
-		we can get rid of this #ifdef. NOTE: The code checked by
-		the #ifdef is the suggested name for the error condition
-		and the actual error code name could very well be different.
-		This will require some monitoring, ie. the status
-		of this request on our part.*/
-#ifdef ER_TOO_MANY_CONCURRENT_TRXS
-		return(ER_TOO_MANY_CONCURRENT_TRXS);
-#else
+		/* New error code HA_ERR_TOO_MANY_CONCURRENT_TRXS is only
+		available in 5.1.38 and later, but the plugin should still
+		work with previous versions of MySQL. */
+#ifdef HA_ERR_TOO_MANY_CONCURRENT_TRXS
+		return(HA_ERR_TOO_MANY_CONCURRENT_TRXS);
+#else /* HA_ERR_TOO_MANY_CONCURRENT_TRXS */
 		return(HA_ERR_RECORD_FILE_FULL);
-#endif
+#endif /* HA_ERR_TOO_MANY_CONCURRENT_TRXS */
 	case DB_UNSUPPORTED:
 		return(HA_ERR_UNSUPPORTED);
 	}

storage/innodb_plugin/include/buf0buf.h

@@ -346,7 +346,7 @@ buf_page_release(
 	mtr_t*		mtr);		/*!< in: mtr */
 /********************************************************************//**
 Moves a page to the start of the buffer pool LRU list. This high-level
-function can be used to prevent an important page from from slipping out of
+function can be used to prevent an important page from slipping out of
 the buffer pool. */
 UNIV_INTERN
 void
@@ -821,7 +821,8 @@ UNIV_INLINE
 void
 buf_page_set_accessed(
 /*==================*/
-	buf_page_t*	bpage)		/*!< in/out: control block */
+	buf_page_t*	bpage,		/*!< in/out: control block */
+	ulint		time_ms)	/*!< in: ut_time_ms() */
 	__attribute__((nonnull));
 /*********************************************************************//**
 Gets the buf_block_t handle of a buffered file block if an uncompressed

storage/innodb_plugin/include/buf0buf.ic

@@ -72,10 +72,16 @@ buf_page_peek_if_too_old(
 /*=====================*/
 	const buf_page_t*	bpage)	/*!< in: block to make younger */
 {
-	if (buf_LRU_old_threshold_ms && bpage->old) {
+	if (UNIV_UNLIKELY(buf_pool->freed_page_clock == 0)) {
+		/* If eviction has not started yet, do not update the
+		statistics or move blocks in the LRU list.  This is
+		either the warm-up phase or an in-memory workload. */
+		return(FALSE);
+	} else if (buf_LRU_old_threshold_ms && bpage->old) {
 		unsigned	access_time = buf_page_is_accessed(bpage);
 
-		if (access_time && ut_time_ms() - access_time
+		if (access_time > 0
+		    && (ut_time_ms() - access_time)
 		    >= buf_LRU_old_threshold_ms) {
 			return(TRUE);
 		}
@@ -85,10 +91,10 @@ buf_page_peek_if_too_old(
 	} else {
 		/* FIXME: bpage->freed_page_clock is 31 bits */
 		return((buf_pool->freed_page_clock & ((1UL << 31) - 1))
-		       > bpage->freed_page_clock
-		       + (buf_pool->curr_size
-			  * (BUF_LRU_OLD_RATIO_DIV - buf_LRU_old_ratio)
-			  / (BUF_LRU_OLD_RATIO_DIV * 4)));
+		       > ((ulint) bpage->freed_page_clock
+			  + (buf_pool->curr_size
+			     * (BUF_LRU_OLD_RATIO_DIV - buf_LRU_old_ratio)
+			     / (BUF_LRU_OLD_RATIO_DIV * 4))));
 	}
 }
 
@@ -490,14 +496,15 @@ UNIV_INLINE
 void
 buf_page_set_accessed(
 /*==================*/
-	buf_page_t*	bpage)		/*!< in/out: control block */
+	buf_page_t*	bpage,		/*!< in/out: control block */
+	ulint		time_ms)	/*!< in: ut_time_ms() */
 {
 	ut_a(buf_page_in_file(bpage));
 	ut_ad(buf_pool_mutex_own());
 
 	if (!bpage->access_time) {
 		/* Make this the time of the first access. */
-		bpage->access_time = ut_time_ms();
+		bpage->access_time = time_ms;
 	}
 }
 

storage/innodb_plugin/include/trx0roll.h

@@ -133,6 +133,17 @@ trx_rollback(
 Rollback or clean up any incomplete transactions which were
 encountered in crash recovery.  If the transaction already was
 committed, then we clean up a possible insert undo log. If the
+transaction was not yet committed, then we roll it back. */
+UNIV_INTERN
+void
+trx_rollback_or_clean_recovered(
+/*============================*/
+	ibool	all);	/*!< in: FALSE=roll back dictionary transactions;
+			TRUE=roll back all non-PREPARED transactions */
+/*******************************************************************//**
+Rollback or clean up any incomplete transactions which were
+encountered in crash recovery.  If the transaction already was
+committed, then we clean up a possible insert undo log. If the
 transaction was not yet committed, then we roll it back.
 Note: this is done in a background thread.
 @return	a dummy parameter */

storage/innodb_plugin/include/trx0trx.h

@@ -179,7 +179,7 @@ trx_commit_off_kernel(
 /****************************************************************//**
 Cleans up a transaction at database startup. The cleanup is needed if
 the transaction already got to the middle of a commit when the database
-crashed, andf we cannot roll it back. */
+crashed, and we cannot roll it back. */
 UNIV_INTERN
 void
 trx_cleanup_at_db_startup(

storage/innodb_plugin/log/log0recv.c

@@ -3118,6 +3118,11 @@ recv_recovery_from_checkpoint_finish(void)
 #ifndef UNIV_LOG_DEBUG
 	recv_sys_free();
 #endif
+	/* Roll back any recovered data dictionary transactions, so
+	that the data dictionary tables will be free of any locks.
+	The data dictionary latch should guarantee that there is at
+	most one data dictionary transaction active at a time. */
+	trx_rollback_or_clean_recovered(FALSE);
 
 	/* Drop partially created indexes. */
 	row_merge_drop_temp_indexes();

storage/innodb_plugin/plug.in

@@ -112,7 +112,7 @@ MYSQL_PLUGIN_ACTIONS(innodb_plugin,  [
     [
       AC_MSG_RESULT(no)
     ]
-    )
+  )
 
   AC_MSG_CHECKING(whether pthread_t can be used by GCC atomic builtins)
   AC_TRY_RUN(
@@ -142,7 +142,7 @@ MYSQL_PLUGIN_ACTIONS(innodb_plugin,  [
     [
       AC_MSG_RESULT(no)
     ]
-    )
+  )
 
   # Try using solaris atomics on SunOS if GCC atomics are not available
   AC_CHECK_DECLS(

storage/innodb_plugin/trx/trx0roll.c

@@ -532,28 +532,26 @@ trx_rollback_active(
 Rollback or clean up any incomplete transactions which were
 encountered in crash recovery.  If the transaction already was
 committed, then we clean up a possible insert undo log. If the
-transaction was not yet committed, then we roll it back.
-Note: this is done in a background thread.
-@return	a dummy parameter */
+transaction was not yet committed, then we roll it back. */
 UNIV_INTERN
-os_thread_ret_t
-trx_rollback_or_clean_all_recovered(
-/*================================*/
-	void*	arg __attribute__((unused)))
-			/*!< in: a dummy parameter required by
-			os_thread_create */
+void
+trx_rollback_or_clean_recovered(
+/*============================*/
+	ibool	all)	/*!< in: FALSE=roll back dictionary transactions;
+			TRUE=roll back all non-PREPARED transactions */
 {
 	trx_t*	trx;
 
 	mutex_enter(&kernel_mutex);
 
-	if (UT_LIST_GET_FIRST(trx_sys->trx_list)) {
+	if (!UT_LIST_GET_FIRST(trx_sys->trx_list)) {
+		goto leave_function;
+	}
 
+	if (all) {
 		fprintf(stderr,
 			"InnoDB: Starting in background the rollback"
 			" of uncommitted transactions\n");
-	} else {
-		goto leave_function;
 	}
 
 	mutex_exit(&kernel_mutex);
@@ -582,18 +580,42 @@ loop:
 			goto loop;
 
 		case TRX_ACTIVE:
-			mutex_exit(&kernel_mutex);
-			trx_rollback_active(trx);
-			goto loop;
+			if (all || trx_get_dict_operation(trx)
+			    != TRX_DICT_OP_NONE) {
+				mutex_exit(&kernel_mutex);
+				trx_rollback_active(trx);
+				goto loop;
+			}
 		}
 	}
 
-	ut_print_timestamp(stderr);
-	fprintf(stderr,
-		"  InnoDB: Rollback of non-prepared transactions completed\n");
+	if (all) {
+		ut_print_timestamp(stderr);
+		fprintf(stderr,
+			"  InnoDB: Rollback of non-prepared"
+			" transactions completed\n");
+	}
 
 leave_function:
 	mutex_exit(&kernel_mutex);
+}
+
+/*******************************************************************//**
+Rollback or clean up any incomplete transactions which were
+encountered in crash recovery.  If the transaction already was
+committed, then we clean up a possible insert undo log. If the
+transaction was not yet committed, then we roll it back.
+Note: this is done in a background thread.
+@return	a dummy parameter */
+UNIV_INTERN
+os_thread_ret_t
+trx_rollback_or_clean_all_recovered(
+/*================================*/
+	void*	arg __attribute__((unused)))
+			/*!< in: a dummy parameter required by
+			os_thread_create */
+{
+	trx_rollback_or_clean_recovered(TRUE);
 
 	/* We count the number of threads in os_thread_exit(). A created
 	thread should always use that to exit and not use return() to exit. */

storage/innodb_plugin/trx/trx0trx.c

@@ -950,7 +950,7 @@ trx_commit_off_kernel(
 /****************************************************************//**
 Cleans up a transaction at database startup. The cleanup is needed if
 the transaction already got to the middle of a commit when the database
-crashed, andf we cannot roll it back. */
+crashed, and we cannot roll it back. */
 UNIV_INTERN
 void
 trx_cleanup_at_db_startup(

storage/innodb_plugin/ut/ut0auxconf_atomic_pthread_t_solaris.c

@@ -17,18 +17,38 @@ Place, Suite 330, Boston, MA 02111-1307 USA
 *****************************************************************************/
 
 /*****************************************************************************
-If this program compiles, then pthread_t objects can be used as arguments
-to Solaris libc atomic functions.
+If this program compiles and returns 0, then pthread_t objects can be used as
+arguments to Solaris libc atomic functions.
 
 Created April 18, 2009 Vasil Dimov
 *****************************************************************************/
 
 #include <pthread.h>
+#include <string.h>
 
 int
 main(int argc, char** argv)
 {
-	pthread_t	x = 0;
+	pthread_t	x1;
+	pthread_t	x2;
+	pthread_t	x3;
+
+	memset(&x1, 0x0, sizeof(x1));
+	memset(&x2, 0x0, sizeof(x2));
+	memset(&x3, 0x0, sizeof(x3));
+
+	if (sizeof(pthread_t) == 4) {
+
+		atomic_cas_32(&x1, x2, x3);
+
+	} else if (sizeof(pthread_t) == 8) {
+
+		atomic_cas_64(&x1, x2, x3);
+
+	} else {
+
+		return(1);
+	}
 
 	return(0);
 }