MDEV-19738: Doublewrite buffer is unnecessarily used for newly (re)initialized pages

Thanks to MDEV-12699, the doublewrite buffer will only be needed in
those cases when a page is being updated in the data file. If the page
had never been written to the data file since it was initialized,
then recovery will be able to reconstruct the page based solely on
the contents of the redo log files.

The doublewrite buffer is only really needed when recovery needs to read
the page in order to apply redo log.

Note: As noted in MDEV-19739, we cannot safely disable the doublewrite
buffer if any MLOG_INDEX_LOAD records were written in the past or will
be written in the future. These records denote that redo logging was
disabled for some pages in a tablespace. Ideally, we would have
the setting innodb_log_optimize_ddl=OFF by default, and would not allow
it to be set while the server is running. If we wanted to make this
safe, assignments with SET GLOBAL innodb_log_optimize_ddl=...
should not only issue a redo log checkpoint (including a write of all
dirty pages from the entire buffer pool), but it should also wait for
all pending ALTER TABLE activity to complete. We elect not to do this.

Avoiding unnecessary use of the doublewrite buffer should improve the
write performance of InnoDB.

buf_page_t::init_on_flush: A new flag to indicate whether it is safe to
skip doublewrite buffering when writing the page.

fsp_init_file_page(): When writing a MLOG_INIT_FILE_PAGE2 record,
set the init_on_flush flag if innodb_log_optimize_ddl=OFF.
This is the only function that writes that log record.

buf_flush_write_block_low(): Skip doublewrite if init_on_flush is set.

fil_aio_wait(): Clear init_on_flush.

7 files changed, 32 insertions, 6 deletions

diff --git a/storage/innobase/buf/buf0buf.cc b/storage/innobase/buf/buf0buf.cc
index bfd3ffce69b..eeabc6aa8ab 100644
--- a/storage/innobase/buf/buf0buf.cc
+++ b/storage/innobase/buf/buf0buf.cc
@@ -1523,6 +1523,7 @@ buf_block_init(
 	block->page.io_fix = BUF_IO_NONE;
 	block->page.flush_observer = NULL;
 	block->page.encrypted = false;
+	block->page.init_on_flush = false;
 	block->page.real_size = 0;
 	block->page.write_size = 0;
 	block->modify_clock = 0;
@@ -5471,6 +5472,7 @@ buf_page_init_for_read(
 		bpage->state = BUF_BLOCK_ZIP_PAGE;
 		bpage->id = page_id;
 		bpage->flush_observer = NULL;
+		bpage->init_on_flush = false;
 
 		ut_d(bpage->in_page_hash = FALSE);
 		ut_d(bpage->in_zip_hash = FALSE);
diff --git a/storage/innobase/buf/buf0flu.cc b/storage/innobase/buf/buf0flu.cc
index 373f6eb4539..aa44a1c5e4c 100644
--- a/storage/innobase/buf/buf0flu.cc
+++ b/storage/innobase/buf/buf0flu.cc
@@ -1040,7 +1040,10 @@ buf_flush_write_block_low(
 
 	ut_ad(space->purpose == FIL_TYPE_TABLESPACE
 	      || space->atomic_write_supported);
-	if (!space->use_doublewrite()) {
+	const bool use_doublewrite = !bpage->init_on_flush
+		&& space->use_doublewrite();
+
+	if (!use_doublewrite) {
 		ulint	type = IORequest::WRITE | IORequest::DO_NOT_WAKE;
 
 		IORequest	request(type, bpage);
@@ -1081,7 +1084,7 @@ buf_flush_write_block_low(
 #endif
 		/* true means we want to evict this page from the
 		LRU list as well. */
-		buf_page_io_complete(bpage, space->use_doublewrite(), true);
+		buf_page_io_complete(bpage, use_doublewrite, true);
 
 		ut_ad(err == DB_SUCCESS);
 	}
diff --git a/storage/innobase/fil/fil0fil.cc b/storage/innobase/fil/fil0fil.cc
index be422c7c467..d395e21f79f 100644
--- a/storage/innobase/fil/fil0fil.cc
+++ b/storage/innobase/fil/fil0fil.cc
@@ -4357,7 +4357,7 @@ fil_aio_wait(
 	fil_node_complete_io(node, type);
 	const fil_type_t	purpose	= node->space->purpose;
 	const ulint		space_id= node->space->id;
-	const bool		dblwr	= node->space->use_doublewrite();
+	bool			dblwr	= node->space->use_doublewrite();
 
 	mutex_exit(&fil_system.mutex);
 
@@ -4406,6 +4406,10 @@ fil_aio_wait(
 		}
 
 		ulint offset = bpage->id.page_no();
+		if (dblwr && bpage->init_on_flush) {
+			bpage->init_on_flush = false;
+			dblwr = false;
+		}
 		dberr_t err = buf_page_io_complete(bpage, dblwr);
 		if (err == DB_SUCCESS) {
 			return;
diff --git a/storage/innobase/include/btr0bulk.h b/storage/innobase/include/btr0bulk.h
index 0fcf4a36d6b..d1a8af22d68 100644
--- a/storage/innobase/include/btr0bulk.h
+++ b/storage/innobase/include/btr0bulk.h
@@ -34,8 +34,6 @@ Created 03/11/2014 Shaohua Wang
 
 /** Innodb B-tree index fill factor for bulk load. */
 extern	uint	innobase_fill_factor;
-/** whether to reduce redo logging during ALTER TABLE */
-extern	my_bool	innodb_log_optimize_ddl;
 
 /*
 The proper function call sequence of PageBulk is as below:
diff --git a/storage/innobase/include/buf0buf.h b/storage/innobase/include/buf0buf.h
index ea2348d61fd..fa22da5192c 100644
--- a/storage/innobase/include/buf0buf.h
+++ b/storage/innobase/include/buf0buf.h
@@ -1493,6 +1493,13 @@ public:
 
 	bool            encrypted;	/*!< page is still encrypted */
 
+	/** whether the page will be (re)initialized at the time it will
+	be written to the file, that is, whether the doublewrite buffer
+	can be safely skipped. Protected under similar conditions as
+	buf_block_t::frame. Can be set while holding buf_block_t::lock
+	X-latch and reset during page flush, while io_fix is in effect. */
+	bool		init_on_flush;
+
 	ulint           real_size;	/*!< Real size of the page
 					Normal pages == srv_page_size
 					page compressed pages, payload
diff --git a/storage/innobase/include/fil0fil.h b/storage/innobase/include/fil0fil.h
index 3e04d90e04e..ea6d8e235c7 100644
--- a/storage/innobase/include/fil0fil.h
+++ b/storage/innobase/include/fil0fil.h
@@ -38,6 +38,8 @@ Created 10/25/1995 Heikki Tuuri
 # include <set>
 #endif
 
+/** whether to reduce redo logging during ALTER TABLE */
+extern	my_bool	innodb_log_optimize_ddl;
 // Forward declaration
 extern my_bool srv_use_doublewrite_buf;
 extern struct buf_dblwr_t* buf_dblwr;
diff --git a/storage/innobase/include/fsp0fsp.h b/storage/innobase/include/fsp0fsp.h
index 75c448c956d..26da55bb2ad 100644
--- a/storage/innobase/include/fsp0fsp.h
+++ b/storage/innobase/include/fsp0fsp.h
@@ -625,7 +625,17 @@ inline void fsp_init_file_page(
 	ut_d(space->modify_check(*mtr));
 	ut_ad(space->id == block->page.id.space());
 	fsp_apply_init_file_page(block);
-	mlog_write_initial_log_record(block->frame, MLOG_INIT_FILE_PAGE2, mtr);
+
+	if (byte* log_ptr = mlog_open(mtr, 11)) {
+		log_ptr = mlog_write_initial_log_record_low(
+			MLOG_INIT_FILE_PAGE2,
+			block->page.id.space(), block->page.id.page_no(),
+			log_ptr, mtr);
+		mlog_close(mtr, log_ptr);
+		if (!innodb_log_optimize_ddl) {
+			block->page.init_on_flush = true;
+		}
+	}
 }
 
 #ifndef UNIV_DEBUG