MDEV-19745 BACKUP STAGE BLOCK_DDL hangs on flush sequence table

Problem was that FLUSH TABLES where trying to read latest sequence state
which conflicted with a running ALTER SEQUENCE. Removed the reading
of the state, when opening a table for FLUSH, as it's not needed in this
case.

Other thing:
- Fixed a potential issue with concurrently running ALTER SEQUENCE where
  the later ALTER could potentially read old data

1 files changed, 6 insertions, 3 deletions

diff --git a/sql/sql_base.cc b/sql/sql_base.cc
index 090ced59f60..d7be718b1ca 100644
--- a/sql/sql_base.cc
+++ b/sql/sql_base.cc
@@ -598,12 +598,15 @@ bool flush_tables(THD *thd, flush_tables_type flag)
     else
     {
       /*
-        HA_OPEN_FOR_ALTER is used to allow us to open the table even if
-        TABLE_SHARE::incompatible_version is set.
+        HA_OPEN_FOR_FLUSH is used to allow us to open the table even if
+        TABLE_SHARE::incompatible_version is set. It also will tell
+        SEQUENCE engine that we don't have to read the sequence information
+        (which may cause deadlocks with concurrently running ALTER TABLE or
+        ALTER SEQUENCE) as we will close the table at once.
       */
       if (!open_table_from_share(thd, share, &empty_clex_str,
                                  HA_OPEN_KEYFILE, 0,
-                                 HA_OPEN_FOR_ALTER,
+                                 HA_OPEN_FOR_ALTER | HA_OPEN_FOR_FLUSH,
                                  tmp_table, FALSE,
                                  NULL))
       {