Applying InnoDB Plugin 1.0.5 snapshot ,part 12

From r5995 to r6043

Detailed revision comments:

r5995 | marko | 2009-09-28 03:52:25 -0500 (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 07:46:02 -0500 (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 08:03:58 -0500 (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 03:09:52 -0500 (Tue, 29 Sep 2009) | 4 lines
branches/zip:

ChangeLog: wrap around 78th column, not earlier.

r6006 | vasil | 2009-09-29 05:15:25 -0500 (Tue, 29 Sep 2009) | 4 lines
branches/zip:

Add ChangeLog entry for the release of 1.0.4.

r6007 | vasil | 2009-09-29 08:19:59 -0500 (Tue, 29 Sep 2009) | 6 lines
branches/zip:

Fix the year, should be 2009.

Pointed by:	Calvin

r6026 | marko | 2009-09-30 02:18:24 -0500 (Wed, 30 Sep 2009) | 1 line
branches/zip: Add some debug assertions for checking FSEG_MAGIC_N.
r6028 | marko | 2009-09-30 08:55:23 -0500 (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 | 2009-09-30 15:32:02 -0500 (Wed, 30 Sep 2009) | 4 lines
branches/zip: non-functional changes

Fix typo.

r6031 | marko | 2009-10-01 06:24:33 -0500 (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 | 2009-10-01 15:19:46 -0500 (Thu, 01 Oct 2009) | 4 lines
branches/zip: fix a typo in error message

Reported as bug#47763.

r6043 | inaam | 2009-10-05 09:45:35 -0500 (Mon, 05 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

storage/innodb_plugin/ChangeLog

+2009-10-01	The InnoDB Team
+
+	* fsp/fsp0fsp.c, row/row0merge.c:
+	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.  The MySQL .frm file may still
+	contain the dropped index, but there is little that we can do
+	about it.
+
+2009-09-28	The InnoDB Team
+
+	* handler/ha_innodb.cc:
+	When a secondary index exists in the MySQL .frm file but not in
+	the InnoDB data dictionary, return an error instead of letting an
+	assertion fail in index_read.
+
+2009-09-28	The InnoDB Team
+
+	* btr/btr0btr.c, buf/buf0buf.c, include/page0page.h,
+	include/page0zip.h, page/page0cur.c, page/page0page.c,
+	page/page0zip.c:
+	Do not write to PAGE_INDEX_ID when restoring an uncompressed page
+	after a compression failure.  The field should only be written
+	when creating a B-tree page.  This fix addresses a race condition
+	in a debug assertion.
+
 2009-09-28	The InnoDB Team
 
 	* fil/fil0fil.c:
@@ -114,6 +140,10 @@
 	Fix Bug#46657 InnoDB plugin: invalid read in index_merge_innodb test
 	(Valgrind)
 
+2009-08-11	The InnoDB Team
+
+	InnoDB Plugin 1.0.4 released
+
 2009-07-20	The InnoDB Team
 
 	* handler/ha_innodb.cc, buf/buf0rea.c, include/srv0srv.h, srv/srv0srv.c

storage/innodb_plugin/btr/btr0btr.c

@@ -41,6 +41,7 @@ Created 6/2/1994 Heikki Tuuri
 #include "lock0lock.h"
 #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);
 
@@ -1011,7 +1014,26 @@ btr_page_reorganize_low(
 	    (!page_zip_compress(page_zip, page, index, NULL))) {
 
 		/* Restore the old page and exit. */
-		buf_frame_copy(page, temp_page);
+
+#if defined UNIV_DEBUG || defined UNIV_ZIP_DEBUG
+		/* Check that the bytes that we skip are identical. */
+		ut_a(!memcmp(page, temp_page, PAGE_HEADER));
+		ut_a(!memcmp(PAGE_HEADER + PAGE_N_RECS + page,
+			     PAGE_HEADER + PAGE_N_RECS + temp_page,
+			     PAGE_DATA - (PAGE_HEADER + PAGE_N_RECS)));
+		ut_a(!memcmp(UNIV_PAGE_SIZE - FIL_PAGE_DATA_END + page,
+			     UNIV_PAGE_SIZE - FIL_PAGE_DATA_END + temp_page,
+			     FIL_PAGE_DATA_END));
+#endif /* UNIV_DEBUG || UNIV_ZIP_DEBUG */
+
+		memcpy(PAGE_HEADER + page, PAGE_HEADER + temp_page,
+		       PAGE_N_RECS - PAGE_N_DIR_SLOTS);
+		memcpy(PAGE_DATA + page, PAGE_DATA + temp_page,
+		       UNIV_PAGE_SIZE - PAGE_DATA - FIL_PAGE_DATA_END);
+
+#if defined UNIV_DEBUG || defined UNIV_ZIP_DEBUG
+		ut_a(!memcmp(page, temp_page, UNIV_PAGE_SIZE));
+#endif /* UNIV_DEBUG || UNIV_ZIP_DEBUG */
 
 		goto func_exit;
 	}

storage/innodb_plugin/buf/buf0buf.c

@@ -1834,7 +1834,7 @@ buf_zip_decompress(
 	switch (fil_page_get_type(frame)) {
 	case FIL_PAGE_INDEX:
 		if (page_zip_decompress(&block->page.zip,
-					block->frame)) {
+					block->frame, TRUE)) {
 			return(TRUE);
 		}
 
@@ -3287,7 +3287,32 @@ void
 buf_pool_invalidate(void)
 /*=====================*/
 {
-	ibool	freed;
+	ibool		freed;
+	enum buf_flush	i;
+
+	buf_pool_mutex_enter();
+
+	for (i = BUF_FLUSH_LRU; i < BUF_FLUSH_N_TYPES; i++) {
+
+		/* As this function is called during startup and
+		during redo application phase during recovery, InnoDB
+		is single threaded (apart from IO helper threads) at
+		this stage. No new write batch can be in intialization
+		stage at this point. */
+		ut_ad(buf_pool->init_flush[i] == FALSE);
+
+		/* However, it is possible that a write batch that has
+		been posted earlier is still not complete. For buffer
+		pool invalidation to proceed we must ensure there is NO
+		write activity happening. */
+		if (buf_pool->n_flush[i] > 0) {
+			buf_pool_mutex_exit();
+			buf_flush_wait_batch_end(i);
+			buf_pool_mutex_enter();
+		}
+	}
+
+	buf_pool_mutex_exit();
 
 	ut_ad(buf_all_freed());
 
@@ -3302,6 +3327,14 @@ buf_pool_invalidate(void)
 	ut_ad(UT_LIST_GET_LEN(buf_pool->LRU) == 0);
 	ut_ad(UT_LIST_GET_LEN(buf_pool->unzip_LRU) == 0);
 
+	buf_pool->freed_page_clock = 0;
+	buf_pool->LRU_old = NULL;
+	buf_pool->LRU_old_len = 0;
+	buf_pool->LRU_flush_ended = 0;
+
+	memset(&buf_pool->stat, 0x00, sizeof(buf_pool->stat));
+	buf_refresh_io_stats();
+
 	buf_pool_mutex_exit();
 }
 

storage/innodb_plugin/fsp/fsp0fsp.c

@@ -1848,6 +1848,8 @@ fsp_seg_inode_page_find_used(
 		if (!ut_dulint_is_zero(mach_read_from_8(inode + FSEG_ID))) {
 			/* This is used */
 
+			ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N)
+			      == FSEG_MAGIC_N_VALUE);
 			return(i);
 		}
 	}
@@ -1879,6 +1881,9 @@ fsp_seg_inode_page_find_free(
 
 			return(i);
 		}
+
+		ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N)
+		      == FSEG_MAGIC_N_VALUE);
 	}
 
 	return(ULINT_UNDEFINED);
@@ -1997,6 +2002,8 @@ fsp_alloc_seg_inode(
 			      page + FSEG_INODE_PAGE_NODE, mtr);
 	}
 
+	ut_ad(ut_dulint_is_zero(mach_read_from_8(inode + FSEG_ID))
+	      || mach_read_from_4(inode + FSEG_MAGIC_N) == FSEG_MAGIC_N_VALUE);
 	return(inode);
 }
 
@@ -2034,7 +2041,7 @@ fsp_free_seg_inode(
 	}
 
 	mlog_write_dulint(inode + FSEG_ID, ut_dulint_zero, mtr);
-	mlog_write_ulint(inode + FSEG_MAGIC_N, 0, MLOG_4BYTES, mtr);
+	mlog_write_ulint(inode + FSEG_MAGIC_N, 0xfa051ce3, MLOG_4BYTES, mtr);
 
 	if (ULINT_UNDEFINED
 	    == fsp_seg_inode_page_find_used(page, zip_size, mtr)) {
@@ -2050,11 +2057,11 @@ fsp_free_seg_inode(
 
 /**********************************************************************//**
 Returns the file segment inode, page x-latched.
-@return	segment inode, page x-latched */
+@return	segment inode, page x-latched; NULL if the inode is free */
 static
 fseg_inode_t*
-fseg_inode_get(
-/*===========*/
+fseg_inode_try_get(
+/*===============*/
 	fseg_header_t*	header,	/*!< in: segment header */
 	ulint		space,	/*!< in: space id */
 	ulint		zip_size,/*!< in: compressed page size in bytes
@@ -2070,8 +2077,34 @@ fseg_inode_get(
 
 	inode = fut_get_ptr(space, zip_size, inode_addr, RW_X_LATCH, mtr);
 
-	ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N) == FSEG_MAGIC_N_VALUE);
+	if (UNIV_UNLIKELY
+	    (ut_dulint_is_zero(mach_read_from_8(inode + FSEG_ID)))) {
 
+		inode = NULL;
+	} else {
+		ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N)
+		      == FSEG_MAGIC_N_VALUE);
+	}
+
+	return(inode);
+}
+
+/**********************************************************************//**
+Returns the file segment inode, page x-latched.
+@return	segment inode, page x-latched */
+static
+fseg_inode_t*
+fseg_inode_get(
+/*===========*/
+	fseg_header_t*	header,	/*!< in: segment header */
+	ulint		space,	/*!< in: space id */
+	ulint		zip_size,/*!< in: compressed page size in bytes
+				or 0 for uncompressed pages */
+	mtr_t*		mtr)	/*!< in: mtr handle */
+{
+	fseg_inode_t*	inode
+		= fseg_inode_try_get(header, space, zip_size, mtr);
+	ut_a(inode);
 	return(inode);
 }
 
@@ -2089,6 +2122,7 @@ fseg_get_nth_frag_page_no(
 	ut_ad(inode && mtr);
 	ut_ad(n < FSEG_FRAG_ARR_N_SLOTS);
 	ut_ad(mtr_memo_contains_page(mtr, inode, MTR_MEMO_PAGE_X_FIX));
+	ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N) == FSEG_MAGIC_N_VALUE);
 	return(mach_read_from_4(inode + FSEG_FRAG_ARR
 				+ n * FSEG_FRAG_SLOT_SIZE));
 }
@@ -2107,6 +2141,7 @@ fseg_set_nth_frag_page_no(
 	ut_ad(inode && mtr);
 	ut_ad(n < FSEG_FRAG_ARR_N_SLOTS);
 	ut_ad(mtr_memo_contains_page(mtr, inode, MTR_MEMO_PAGE_X_FIX));
+	ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N) == FSEG_MAGIC_N_VALUE);
 
 	mlog_write_ulint(inode + FSEG_FRAG_ARR + n * FSEG_FRAG_SLOT_SIZE,
 			 page_no, MLOG_4BYTES, mtr);
@@ -2467,6 +2502,8 @@ fseg_fill_free_list(
 		xdes_set_state(descr, XDES_FSEG, mtr);
 
 		seg_id = mtr_read_dulint(inode + FSEG_ID, mtr);
+		ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N)
+		      == FSEG_MAGIC_N_VALUE);
 		mlog_write_dulint(descr + XDES_ID, seg_id, mtr);
 
 		flst_add_last(inode + FSEG_FREE, descr + XDES_FLST_NODE, mtr);
@@ -2495,6 +2532,7 @@ fseg_alloc_free_extent(
 	fil_addr_t	first;
 
 	ut_ad(!((page_offset(inode) - FSEG_ARR_OFFSET) % FSEG_INODE_SIZE));
+	ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N) == FSEG_MAGIC_N_VALUE);
 
 	if (flst_get_len(inode + FSEG_FREE, mtr) > 0) {
 		/* Segment free list is not empty, allocate from it */
@@ -3152,6 +3190,8 @@ fseg_mark_page_used(
 
 	ut_ad(seg_inode && mtr);
 	ut_ad(!((page_offset(seg_inode) - FSEG_ARR_OFFSET) % FSEG_INODE_SIZE));
+	ut_ad(mach_read_from_4(seg_inode + FSEG_MAGIC_N)
+	      == FSEG_MAGIC_N_VALUE);
 
 	descr = xdes_get_descriptor(space, zip_size, page, mtr);
 
@@ -3389,6 +3429,8 @@ fseg_free_extent(
 	ut_a(xdes_get_state(descr, mtr) == XDES_FSEG);
 	ut_a(0 == ut_dulint_cmp(mtr_read_dulint(descr + XDES_ID, mtr),
 				mtr_read_dulint(seg_inode + FSEG_ID, mtr)));
+	ut_ad(mach_read_from_4(seg_inode + FSEG_MAGIC_N)
+	      == FSEG_MAGIC_N_VALUE);
 
 	first_page_in_extent = page - (page % FSP_EXTENT_SIZE);
 
@@ -3479,7 +3521,13 @@ fseg_free_step(
 	ut_a(descr);
 	ut_a(xdes_get_bit(descr, XDES_FREE_BIT,
 			  header_page % FSP_EXTENT_SIZE, mtr) == FALSE);
-	inode = fseg_inode_get(header, space, zip_size, mtr);
+	inode = fseg_inode_try_get(header, space, zip_size, mtr);
+
+	if (UNIV_UNLIKELY(inode == NULL)) {
+		fprintf(stderr, "double free of inode from %u:%u\n",
+			(unsigned) space, (unsigned) header_page);
+		return(TRUE);
+	}
 
 	descr = fseg_get_first_extent(inode, space, zip_size, mtr);
 
@@ -3603,6 +3651,7 @@ fseg_get_first_extent(
 	ut_ad(inode && mtr);
 
 	ut_ad(space == page_get_space_id(page_align(inode)));
+	ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N) == FSEG_MAGIC_N_VALUE);
 
 	first = fil_addr_null;
 
@@ -3817,6 +3866,7 @@ fseg_print_low(
 		(ulong) reserved, (ulong) used, (ulong) n_full,
 		(ulong) n_frag, (ulong) n_free, (ulong) n_not_full,
 		(ulong) n_used);
+	ut_ad(mach_read_from_4(inode + FSEG_MAGIC_N) == FSEG_MAGIC_N_VALUE);
 }
 
 #ifdef UNIV_BTR_PRINT

storage/innodb_plugin/handler/ha_innodb.cc

@@ -3159,7 +3159,7 @@ retry:
 
 		if (is_part) {
 			sql_print_error("Failed to open table %s after "
-					"%lu attemtps.\n", norm_name,
+					"%lu attempts.\n", norm_name,
 					retries);
 		}
 
@@ -5060,6 +5060,11 @@ ha_innobase::index_read(
 
 	index = prebuilt->index;
 
+	if (UNIV_UNLIKELY(index == NULL)) {
+		prebuilt->index_usable = FALSE;
+		DBUG_RETURN(HA_ERR_CRASHED);
+	}
+
 	/* Note that if the index for which the search template is built is not
 	necessarily prebuilt->index, but can also be the clustered index */
 
@@ -5219,6 +5224,7 @@ ha_innobase::change_active_index(
 	if (UNIV_UNLIKELY(!prebuilt->index)) {
 		sql_print_warning("InnoDB: change_active_index(%u) failed",
 				  keynr);
+		prebuilt->index_usable = FALSE;
 		DBUG_RETURN(1);
 	}
 

storage/innodb_plugin/include/log0recv.h

@@ -433,6 +433,11 @@ are allowed yet: the variable name is misleading. */
 extern ibool		recv_no_ibuf_operations;
 /** TRUE when recv_init_crash_recovery() has been called. */
 extern ibool		recv_needed_recovery;
+#ifdef UNIV_DEBUG
+/** TRUE if writing to the redo log (mtr_commit) is forbidden.
+Protected by log_sys->mutex. */
+extern ibool		recv_no_log_write;
+#endif /* UNIV_DEBUG */
 
 /** TRUE if buf_page_is_corrupted() should check if the log sequence
 number (FIL_PAGE_LSN) is in the future.  Initially FALSE, and set by

storage/innodb_plugin/include/page0page.h

@@ -76,8 +76,11 @@ typedef	byte		page_header_t;
 				header which are set in a page create */
 /*----*/
 #define	PAGE_LEVEL	 26	/* level of the node in an index tree; the
-				leaf level is the level 0 */
-#define	PAGE_INDEX_ID	 28	/* index id where the page belongs */
+				leaf level is the level 0.  This field should
+				not be written to after page creation. */
+#define	PAGE_INDEX_ID	 28	/* index id where the page belongs.
+				This field should not be written to after
+				page creation. */
 #define PAGE_BTR_SEG_LEAF 36	/* file segment header for the leaf pages in
 				a B-tree: defined only on the root page of a
 				B-tree, but not in the root of an ibuf tree */

storage/innodb_plugin/include/page0zip.h

@@ -127,8 +127,12 @@ page_zip_decompress(
 /*================*/
 	page_zip_des_t*	page_zip,/*!< in: data, ssize;
 				out: m_start, m_end, m_nonempty, n_blobs */
-	page_t*		page)	/*!< out: uncompressed page, may be trashed */
-	__attribute__((nonnull));
+	page_t*		page,	/*!< out: uncompressed page, may be trashed */
+	ibool		all)	/*!< in: TRUE=decompress the whole page;
+				FALSE=verify but do not copy some
+				page header fields that should not change
+				after page creation */
+	__attribute__((nonnull(1,2)));
 
 #ifdef UNIV_DEBUG
 /**********************************************************************//**
@@ -385,8 +389,8 @@ IMPORTANT: if page_zip_reorganize() is invoked on a leaf page of a
 non-clustered index, the caller must update the insert buffer free
 bits in the same mini-transaction in such a way that the modification
 will be redo-logged.
-@return TRUE on success, FALSE on failure; page and page_zip will be
-left intact on failure. */
+@return TRUE on success, FALSE on failure; page_zip will be left
+intact on failure, but page will be overwritten. */
 UNIV_INTERN
 ibool
 page_zip_reorganize(

storage/innodb_plugin/include/row0ins.h

@@ -45,7 +45,7 @@ row_ins_check_foreign_constraint(
 /*=============================*/
 	ibool		check_ref,/*!< in: TRUE If we want to check that
 				the referenced table is ok, FALSE if we
-				want to to check the foreign key table */
+				want to check the foreign key table */
 	dict_foreign_t*	foreign,/*!< in: foreign constraint; NOTE that the
 				tables mentioned in it must be in the
 				dictionary cache if they exist at all */

storage/innodb_plugin/lock/lock0lock.c

@@ -4633,6 +4633,10 @@ lock_rec_queue_validate(
 		next function call: we have to release lock table mutex
 		to obey the latching order */
 
+		/* If this thread is holding the file space latch
+		(fil_space_t::latch), the following check WILL break
+		latching order and may cause a deadlock of threads. */
+
 		impl_trx = lock_sec_rec_some_has_impl_off_kernel(
 			rec, index, offsets);
 
@@ -4756,6 +4760,11 @@ loop:
 
 			lock_mutex_exit_kernel();
 
+			/* If this thread is holding the file space
+			latch (fil_space_t::latch), the following
+			check WILL break the latching order and may
+			cause a deadlock of threads. */
+
 			lock_rec_queue_validate(block, rec, index, offsets);
 
 			lock_mutex_enter_kernel();

storage/innodb_plugin/log/log0log.c

@@ -241,6 +241,7 @@ log_reserve_and_open(
 	ut_a(len < log->buf_size / 2);
 loop:
 	mutex_enter(&(log->mutex));
+	ut_ad(!recv_no_log_write);
 
 	/* Calculate an upper limit for the space the string may take in the
 	log buffer */
@@ -309,6 +310,7 @@ log_write_low(
 
 	ut_ad(mutex_own(&(log->mutex)));
 part_loop:
+	ut_ad(!recv_no_log_write);
 	/* Calculate a part length */
 
 	data_len = (log->buf_free % OS_FILE_LOG_BLOCK_SIZE) + str_len;
@@ -377,6 +379,7 @@ log_close(void)
 	ib_uint64_t	checkpoint_age;
 
 	ut_ad(mutex_own(&(log->mutex)));
+	ut_ad(!recv_no_log_write);
 
 	lsn = log->lsn;
 
@@ -668,8 +671,6 @@ log_calc_max_ages(void)
 	ulint		archive_margin;
 	ulint		smallest_archive_margin;
 
-	ut_ad(!mutex_own(&(log_sys->mutex)));
-
 	mutex_enter(&(log_sys->mutex));
 
 	group = UT_LIST_GET_FIRST(log_sys->log_groups);
@@ -1117,6 +1118,7 @@ log_io_complete(
 	}
 
 	mutex_enter(&(log_sys->mutex));
+	ut_ad(!recv_no_log_write);
 
 	ut_a(group->n_pending_writes > 0);
 	ut_a(log_sys->n_pending_writes > 0);
@@ -1148,6 +1150,7 @@ log_group_file_header_flush(
 	ulint	dest_offset;
 
 	ut_ad(mutex_own(&(log_sys->mutex)));
+	ut_ad(!recv_no_log_write);
 	ut_a(nth_file < group->n_files);
 
 	buf = *(group->file_header_bufs + nth_file);
@@ -1219,6 +1222,7 @@ log_group_write_buf(
 	ulint	i;
 
 	ut_ad(mutex_own(&(log_sys->mutex)));
+	ut_ad(!recv_no_log_write);
 	ut_a(len % OS_FILE_LOG_BLOCK_SIZE == 0);
 	ut_a(((ulint) start_lsn) % OS_FILE_LOG_BLOCK_SIZE == 0);
 
@@ -1361,6 +1365,7 @@ loop:
 #endif
 
 	mutex_enter(&(log_sys->mutex));
+	ut_ad(!recv_no_log_write);
 
 	if (flush_to_disk
 	    && log_sys->flushed_to_disk_lsn >= lsn) {
@@ -1974,6 +1979,7 @@ log_checkpoint(
 
 	mutex_enter(&(log_sys->mutex));
 
+	ut_ad(!recv_no_log_write);
 	oldest_lsn = log_buf_pool_get_oldest_modification();
 
 	mutex_exit(&(log_sys->mutex));
@@ -2086,6 +2092,7 @@ loop:
 	do_checkpoint = FALSE;
 
 	mutex_enter(&(log->mutex));
+	ut_ad(!recv_no_log_write);
 
 	if (log->check_flush_or_checkpoint == FALSE) {
 		mutex_exit(&(log->mutex));
@@ -3035,6 +3042,7 @@ loop:
 #endif /* UNIV_LOG_ARCHIVE */
 
 	mutex_enter(&(log_sys->mutex));
+	ut_ad(!recv_no_log_write);
 
 	if (log_sys->check_flush_or_checkpoint) {
 

storage/innodb_plugin/log/log0recv.c

@@ -78,6 +78,11 @@ UNIV_INTERN ibool	recv_recovery_from_backup_on = FALSE;
 #ifndef UNIV_HOTBACKUP
 /** TRUE when recv_init_crash_recovery() has been called. */
 UNIV_INTERN ibool	recv_needed_recovery = FALSE;
+# ifdef UNIV_DEBUG
+/** TRUE if writing to the redo log (mtr_commit) is forbidden.
+Protected by log_sys->mutex. */
+UNIV_INTERN ibool	recv_no_log_write = FALSE;
+# endif /* UNIV_DEBUG */
 
 /** TRUE if buf_page_is_corrupted() should check if the log sequence
 number (FIL_PAGE_LSN) is in the future.  Initially FALSE, and set by
@@ -1705,6 +1710,7 @@ loop:
 		/* Flush all the file pages to disk and invalidate them in
 		the buffer pool */
 
+		ut_d(recv_no_log_write = TRUE);
 		mutex_exit(&(recv_sys->mutex));
 		mutex_exit(&(log_sys->mutex));
 
@@ -1718,6 +1724,7 @@ loop:
 
 		mutex_enter(&(log_sys->mutex));
 		mutex_enter(&(recv_sys->mutex));
+		ut_d(recv_no_log_write = FALSE);
 
 		recv_no_ibuf_operations = FALSE;
 	}

storage/innodb_plugin/mtr/mtr0mtr.c

@@ -35,6 +35,7 @@ Created 11/26/1995 Heikki Tuuri
 #include "log0log.h"
 
 #ifndef UNIV_HOTBACKUP
+# include "log0recv.h"
 /*****************************************************************//**
 Releases the item in the slot given. */
 UNIV_INLINE
@@ -181,6 +182,8 @@ mtr_commit(
 	ut_d(mtr->state = MTR_COMMITTING);
 
 #ifndef UNIV_HOTBACKUP
+	/* This is a dirty read, for debugging. */
+	ut_ad(!recv_no_log_write);
 	write_log = mtr->modifications && mtr->n_log_recs;
 
 	if (write_log) {

storage/innodb_plugin/page/page0cur.c

@@ -1195,7 +1195,7 @@ page_cur_insert_rec_zip_reorg(
 	}
 
 	/* Out of space: restore the page */
-	if (!page_zip_decompress(page_zip, page)) {
+	if (!page_zip_decompress(page_zip, page, FALSE)) {
 		ut_error; /* Memory corrupted? */
 	}
 	ut_ad(page_validate(page, index));

storage/innodb_plugin/page/page0page.c

@@ -679,7 +679,7 @@ page_copy_rec_list_end(
 
 				if (UNIV_UNLIKELY
 				    (!page_zip_decompress(new_page_zip,
-							  new_page))) {
+							  new_page, FALSE))) {
 					ut_error;
 				}
 				ut_ad(page_validate(new_page, index));
@@ -792,7 +792,7 @@ page_copy_rec_list_start(
 
 				if (UNIV_UNLIKELY
 				    (!page_zip_decompress(new_page_zip,
-							  new_page))) {
+							  new_page, FALSE))) {
 					ut_error;
 				}
 				ut_ad(page_validate(new_page, index));

storage/innodb_plugin/page/page0zip.c

@@ -2821,7 +2821,11 @@ page_zip_decompress(
 /*================*/
 	page_zip_des_t*	page_zip,/*!< in: data, ssize;
 				out: m_start, m_end, m_nonempty, n_blobs */
-	page_t*		page)	/*!< out: uncompressed page, may be trashed */
+	page_t*		page,	/*!< out: uncompressed page, may be trashed */
+	ibool		all)	/*!< in: TRUE=decompress the whole page;
+				FALSE=verify but do not copy some
+				page header fields that should not change
+				after page creation */
 {
 	z_stream	d_stream;
 	dict_index_t*	index	= NULL;
@@ -2851,13 +2855,36 @@ page_zip_decompress(
 	heap = mem_heap_create(n_dense * (3 * sizeof *recs) + UNIV_PAGE_SIZE);
 	recs = mem_heap_alloc(heap, n_dense * (2 * sizeof *recs));
 
+	if (all) {
+		/* Copy the page header. */
+		memcpy(page, page_zip->data, PAGE_DATA);
+	} else {
+		/* Check that the bytes that we skip are identical. */
+#if defined UNIV_DEBUG || defined UNIV_ZIP_DEBUG
+		ut_a(!memcmp(FIL_PAGE_TYPE + page,
+			     FIL_PAGE_TYPE + page_zip->data,
+			     PAGE_HEADER - FIL_PAGE_TYPE));
+		ut_a(!memcmp(PAGE_HEADER + PAGE_LEVEL + page,
+			     PAGE_HEADER + PAGE_LEVEL + page_zip->data,
+			     PAGE_DATA - (PAGE_HEADER + PAGE_LEVEL)));
+#endif /* UNIV_DEBUG || UNIV_ZIP_DEBUG */
+
+		/* Copy the mutable parts of the page header. */
+		memcpy(page, page_zip->data, FIL_PAGE_TYPE);
+		memcpy(PAGE_HEADER + page, PAGE_HEADER + page_zip->data,
+		       PAGE_LEVEL - PAGE_N_DIR_SLOTS);
+
+#if defined UNIV_DEBUG || defined UNIV_ZIP_DEBUG
+		/* Check that the page headers match after copying. */
+		ut_a(!memcmp(page, page_zip->data, PAGE_DATA));
+#endif /* UNIV_DEBUG || UNIV_ZIP_DEBUG */
+	}
+
 #ifdef UNIV_ZIP_DEBUG
-	/* Clear the page. */
-	memset(page, 0x55, UNIV_PAGE_SIZE);
+	/* Clear the uncompressed page, except the header. */
+	memset(PAGE_DATA + page, 0x55, UNIV_PAGE_SIZE - PAGE_DATA);
 #endif /* UNIV_ZIP_DEBUG */
-	UNIV_MEM_INVALID(page, UNIV_PAGE_SIZE);
-	/* Copy the page header. */
-	memcpy(page, page_zip->data, PAGE_DATA);
+	UNIV_MEM_INVALID(PAGE_DATA + page, UNIV_PAGE_SIZE - PAGE_DATA);
 
 	/* Copy the page directory. */
 	if (UNIV_UNLIKELY(!page_zip_dir_decode(page_zip, page, recs,
@@ -3098,7 +3125,7 @@ page_zip_validate_low(
 #endif /* UNIV_DEBUG_VALGRIND */
 
 	temp_page_zip = *page_zip;
-	valid = page_zip_decompress(&temp_page_zip, temp_page);
+	valid = page_zip_decompress(&temp_page_zip, temp_page, TRUE);
 	if (!valid) {
 		fputs("page_zip_validate(): failed to decompress\n", stderr);
 		goto func_exit;
@@ -4376,8 +4403,8 @@ IMPORTANT: if page_zip_reorganize() is invoked on a leaf page of a
 non-clustered index, the caller must update the insert buffer free
 bits in the same mini-transaction in such a way that the modification
 will be redo-logged.
-@return TRUE on success, FALSE on failure; page and page_zip will be
-left intact on failure. */
+@return TRUE on success, FALSE on failure; page_zip will be left
+intact on failure, but page will be overwritten. */
 UNIV_INTERN
 ibool
 page_zip_reorganize(
@@ -4442,9 +4469,6 @@ page_zip_reorganize(
 
 	if (UNIV_UNLIKELY(!page_zip_compress(page_zip, page, index, mtr))) {
 
-		/* Restore the old page and exit. */
-		buf_frame_copy(page, temp_page);
-
 #ifndef UNIV_HOTBACKUP
 		buf_block_free(temp_block);
 #endif /* !UNIV_HOTBACKUP */
@@ -4605,7 +4629,8 @@ corrupt:
 		memcpy(page_zip->data + page_zip_get_size(page_zip)
 		       - trailer_size, ptr + 8 + size, trailer_size);
 
-		if (UNIV_UNLIKELY(!page_zip_decompress(page_zip, page))) {
+		if (UNIV_UNLIKELY(!page_zip_decompress(page_zip, page,
+						       TRUE))) {
 
 			goto corrupt;
 		}

storage/innodb_plugin/row/row0ins.c

@@ -1191,7 +1191,7 @@ row_ins_check_foreign_constraint(
 /*=============================*/
 	ibool		check_ref,/*!< in: TRUE if we want to check that
 				the referenced table is ok, FALSE if we
-				want to to check the foreign key table */
+				want to check the foreign key table */
 	dict_foreign_t*	foreign,/*!< in: foreign constraint; NOTE that the
 				tables mentioned in it must be in the
 				dictionary cache if they exist at all */

storage/innodb_plugin/row/row0merge.c

@@ -1947,7 +1947,15 @@ row_merge_drop_index(
 	static const char str1[] =
 		"PROCEDURE DROP_INDEX_PROC () IS\n"
 		"BEGIN\n"
+		/* Rename the index, so that it will be dropped by
+		row_merge_drop_temp_indexes() at crash recovery
+		if the server crashes before this trx is committed. */
+		"UPDATE SYS_INDEXES SET NAME=CONCAT('"
+		TEMP_INDEX_PREFIX_STR "', NAME) WHERE ID = :indexid;\n"
+		"COMMIT WORK;\n"
+		/* Drop the field definitions of the index. */
 		"DELETE FROM SYS_FIELDS WHERE INDEX_ID = :indexid;\n"
+		/* Drop the index definition and the B-tree. */
 		"DELETE FROM SYS_INDEXES WHERE ID = :indexid\n"
 		"		AND TABLE_ID = :tableid;\n"
 		"END;\n";

storage/innodb_plugin/trx/trx0trx.c

@@ -803,7 +803,7 @@ trx_commit_off_kernel(
 		in exactly the same order as commit lsn's, if the transactions
 		have different rollback segments. To get exactly the same
 		order we should hold the kernel mutex up to this point,
-		adding to to the contention of the kernel mutex. However, if
+		adding to the contention of the kernel mutex. However, if
 		a transaction T2 is able to see modifications made by
 		a transaction T1, T2 will always get a bigger transaction
 		number and a bigger commit lsn than T1. */