Bug 12635227 - 61188: DROP TABLE EXTREMELY SLOW

approved by: Marko
rb://681

Coalescing of free buf_page_t descriptors can prove to be one severe
bottleneck in performance of compression. One such workload where it
hurts badly is DROP TABLE. This patch removes buf_page_t allocations
from buf_buddy and uses ut_malloc instead.
In order to further reduce overhead of colaescing we no longer attempt
to coalesce a block if the corresponding free_list is less than 16 in
size.

storage/innodb_plugin/include/buf0lru.h

@@ -30,18 +30,6 @@ Created 11/5/1995 Heikki Tuuri
 #include "ut0byte.h"
 #include "buf0types.h"
 
-/** The return type of buf_LRU_free_block() */
-enum buf_lru_free_block_status {
-	/** freed */
-	BUF_LRU_FREED = 0,
-	/** not freed because the caller asked to remove the
-	uncompressed frame but the control block cannot be
-	relocated */
-	BUF_LRU_CANNOT_RELOCATE,
-	/** not freed because of some other reason */
-	BUF_LRU_NOT_FREED
-};
-
 /******************************************************************//**
 Tries to remove LRU flushed blocks from the end of the LRU list and put them
 to the free list. This is beneficial for the efficiency of the insert buffer
@@ -98,17 +86,16 @@ buf_LRU_insert_zip_clean(
 Try to free a block.  If bpage is a descriptor of a compressed-only
 page, the descriptor object will be freed as well.
 
-NOTE: If this function returns BUF_LRU_FREED, it will temporarily
+NOTE: If this function returns TRUE, it will temporarily
 release buf_pool_mutex.  Furthermore, the page frame will no longer be
 accessible via bpage.
 
 The caller must hold buf_pool_mutex and buf_page_get_mutex(bpage) and
 release these two mutexes after the call.  No other
 buf_page_get_mutex() may be held when calling this function.
-@return BUF_LRU_FREED if freed, BUF_LRU_CANNOT_RELOCATE or
-BUF_LRU_NOT_FREED otherwise. */
+@return TRUE if freed, FALSE otherwise. */
 UNIV_INTERN
-enum buf_lru_free_block_status
+ibool
 buf_LRU_free_block(
 /*===============*/
 	buf_page_t*	bpage,	/*!< in: block to be freed */