MDEV-26110: Do not rely on alignment on static allocation

It is implementation-defined whether alignment requirements
that are larger than std::max_align_t (typically 8 or 16 bytes)
will be honored by the compiler and linker.

It turns out that on IBM AIX, both alignas() and MY_ALIGNED()
only guarantees alignment up to 16 bytes.

For some data structures, specifying alignment to the CPU
cache line size (typically 64 or 128 bytes) is a mere performance
optimization, and we do not really care whether the requested
alignment is guaranteed.

But, for the correct operation of direct I/O, we do require that
the buffers be aligned at a block size boundary.

field_ref_zero: Define as a pointer, not an array.
For innochecksum, we can make this point to unaligned memory;
for anything else, we will allocate an aligned buffer from the heap.
This buffer will be used for overwriting freed data pages when
innodb_immediate_scrub_data_uncompressed=ON. And exactly that code
hit an assertion failure on AIX, in the test innodb.innodb_scrub.

log_sys.checkpoint_buf: Define as a pointer to aligned memory
that is allocated from heap.

log_t::file::write_header_durable(): Reuse log_sys.checkpoint_buf
instead of trying to allocate an aligned buffer from the stack.

1 files changed, 28 insertions, 11 deletions

diff --git a/extra/mariabackup/xtrabackup.cc b/extra/mariabackup/xtrabackup.cc
index 7ec6fbb8f59..bf4a80832dc 100644
--- a/extra/mariabackup/xtrabackup.cc
+++ b/extra/mariabackup/xtrabackup.cc
@@ -4501,6 +4501,14 @@ fail:
 		goto fail;
 	}
 
+
+	if (auto b = aligned_malloc(UNIV_PAGE_SIZE_MAX, 4096)) {
+		field_ref_zero = static_cast<byte*>(
+			memset_aligned<4096>(b, 0, UNIV_PAGE_SIZE_MAX));
+	} else {
+		goto fail;
+	}
+
         {
 	/* definition from recv_recovery_from_checkpoint_start() */
 	ulint		max_cp_field;
@@ -4515,14 +4523,17 @@ reread_log_header:
 
 	if (err != DB_SUCCESS) {
 		msg("Error: cannot read redo log header");
+unlock_and_fail:
 		mysql_mutex_unlock(&log_sys.mutex);
+free_and_fail:
+		aligned_free(const_cast<byte*>(field_ref_zero));
+		field_ref_zero = nullptr;
 		goto fail;
 	}
 
 	if (log_sys.log.format == 0) {
 		msg("Error: cannot process redo log before MariaDB 10.2.2");
-		mysql_mutex_unlock(&log_sys.mutex);
-		goto fail;
+		goto unlock_and_fail;
 	}
 
 	byte* buf = log_sys.checkpoint_buf;
@@ -4543,7 +4554,7 @@ reread_log_header:
 	xtrabackup_init_datasinks();
 
 	if (!select_history()) {
-		goto fail;
+		goto free_and_fail;
 	}
 
 	/* open the log file */
@@ -4552,12 +4563,13 @@ reread_log_header:
 	if (dst_log_file == NULL) {
 		msg("Error: failed to open the target stream for '%s'.",
 		    LOG_FILE_NAME);
-		goto fail;
+		goto free_and_fail;
 	}
 
 	/* label it */
-	byte MY_ALIGNED(OS_FILE_LOG_BLOCK_SIZE) log_hdr_buf[LOG_FILE_HDR_SIZE];
-	memset(log_hdr_buf, 0, sizeof log_hdr_buf);
+	byte* log_hdr_buf = static_cast<byte*>(
+		aligned_malloc(LOG_FILE_HDR_SIZE, OS_FILE_LOG_BLOCK_SIZE));
+	memset(log_hdr_buf, 0, LOG_FILE_HDR_SIZE);
 
 	byte *log_hdr_field = log_hdr_buf;
 	mach_write_to_4(LOG_HEADER_FORMAT + log_hdr_field, log_sys.log.format);
@@ -4586,11 +4598,13 @@ reread_log_header:
 			log_block_calc_checksum_crc32(log_hdr_field));
 
 	/* Write log header*/
-	if (ds_write(dst_log_file, log_hdr_buf, sizeof(log_hdr_buf))) {
+	if (ds_write(dst_log_file, log_hdr_buf, LOG_FILE_HDR_SIZE)) {
 		msg("error: write to logfile failed");
-		goto fail;
+		aligned_free(log_hdr_buf);
+		goto free_and_fail;
 	}
 
+	aligned_free(log_hdr_buf);
 	log_copying_running = true;
 	/* start io throttle */
 	if(xtrabackup_throttle) {
@@ -4608,7 +4622,7 @@ reread_log_header:
 		    " error %s.", ut_strerr(err));
 fail_before_log_copying_thread_start:
 		log_copying_running = false;
-		goto fail;
+		goto free_and_fail;
 	}
 
 	/* copy log file by current position */
@@ -4625,7 +4639,7 @@ fail_before_log_copying_thread_start:
 
 	/* FLUSH CHANGED_PAGE_BITMAPS call */
 	if (!flush_changed_page_bitmaps()) {
-		goto fail;
+		goto free_and_fail;
 	}
 
 	ut_a(xtrabackup_parallel > 0);
@@ -4647,7 +4661,7 @@ fail_before_log_copying_thread_start:
 	datafiles_iter_t *it = datafiles_iter_new();
 	if (it == NULL) {
 		msg("mariabackup: Error: datafiles_iter_new() failed.");
-		goto fail;
+		goto free_and_fail;
 	}
 
 	/* Create data copying threads */
@@ -4702,6 +4716,9 @@ fail_before_log_copying_thread_start:
 	if (opt_log_innodb_page_corruption)
 		ok = corrupted_pages.print_to_file(MB_CORRUPTED_PAGES_FILE);
 
+	aligned_free(const_cast<byte*>(field_ref_zero));
+	field_ref_zero = nullptr;
+
 	if (!ok) {
 		goto fail;
 	}