Second attempt to fix same bug:

Use the same queue for all READ operations.
Release queues for all used pages.

This fixes a hang in the s3.alter2 test case

When converting a table (test.s3_table) from S3 to another engine, the
following will be logged to the binary log:

DROP TABLE IF EXISTS test.t1;
CREATE OR REPLACE TABLE test.t1 (...) ENGINE=new_engine
INSERT rows to test.t1 in binary-row-log-format

The bug is that the above statements are logged one by one to the binary
log. This means that a fast slave, configured to use the same S3 storage
as the master, would be able to execute the DROP and CREATE from the
binary log before the master has finished the ALTER TABLE.
In this case the slave would ignore the DROP (as it's on a S3 table) but
it will stop on CREATE of the local tale, as the table is still exists in
S3. The REPLACE part will be ignored by the slave as it can't touch the
S3 table.

The fix is to ensure that all the above statements is written to binary
log AFTER the table has been deleted from S3.

- Added missing test for binlog_filter to ALTER TABLE

- Rewrote bool Query_compressed_log_event::write() to make it more readable
  (no logic changes).
- Changed DBUG_PRINT of 'is_error:' to 'is_error():' to make it easier to
  find error: in traces.
- Ensure that 'db' is never null in Query_log_event (Simplified code).

The rw_lock_s_lock() calls for the buf_pool.page_hash became a
clear bottleneck after MDEV-15053 reduced the contention on
buf_pool.mutex. We will replace that use of rw_lock_t with a
special implementation that is optimized for memory bus traffic.

The hash_table_locks instrumentation will be removed.

buf_pool_t::page_hash: Use a special implementation whose API is
compatible with hash_table_t, and store the custom rw-locks
directly in buf_pool.page_hash.array, intentionally sharing
cache lines with the hash table pointers.

rw_lock: A low-level rw-lock implementation based on std::atomic<uint32_t>
where read_trylock() becomes a simple fetch_add(1).

buf_pool_t::page_hash_latch: The special of rw_lock for the page_hash.

buf_pool_t::page_hash_latch::read_lock(): Assert that buf_pool.mutex
is not being held by the caller.

buf_pool_t::page_hash_latch::write_lock() may be called while not holding
buf_pool.mutex. buf_pool_t::watch_set() is such a caller.

buf_pool_t::page_hash_latch::read_lock_wait(),
page_hash_latch::write_lock_wait(): The spin loops.
These will obey the global parameters innodb_sync_spin_loops and
innodb_sync_spin_wait_delay.

buf_pool_t::freed_page_hash: A singly linked list of copies of
buf_pool.page_hash that ever existed. The fact that we never
free any buf_pool.page_hash.array guarantees that all
page_hash_latch that ever existed will remain valid until shutdown.

buf_pool_t::resize_hash(): Replaces buf_pool_resize_hash().
Prepend a shallow copy of the old page_hash to freed_page_hash.

buf_pool_t::page_hash_table::n_cells: Declare as Atomic_relaxed.

buf_pool_t::page_hash_table::lock(): Explain what prevents a
race condition with buf_pool_t::resize_hash().

hash_get_n_cells(): Remove. Access n_cells directly.

hash_get_nth_cell(): Remove. Access array directly.

hash_table_clear(): Replaced with hash_table_t::clear().

hash_table_create(), hash_table_free(): Remove.

hash0hash.cc: Remove.

btr_search_sys::parts[]: A single structure for the partitions of
the adaptive hash index. Replaces the 3 separate arrays:
btr_search_latches[], btr_search_sys->hash_tables,
btr_search_sys->hash_tables[i]->heap.

hash_table_t::heap, hash_table_t::adaptive: Remove.

ha0ha.cc: Remove. Move all code to btr0sea.cc.

HASH_TABLE_SYNC_MUTEX was kind-of used for the adaptive hash index,
even though that hash table is already protected by btr_search_latches[].

HASH_TABLE_SYNC_RWLOCK was only being used for buf_pool.page_hash.
It is cleaner to decouple that synchronization from hash_table_t,
and move it to the actual user.

buf_pool_t::page_hash_latches[]: Synchronization for buf_pool.page_hash.

LATCH_ID_HASH_TABLE_MUTEX: Remove.

hash_table_t::sync_obj, hash_table_t::n_sync_obj: Remove.

hash_table_t::type, hash_table_sync_t: Remove.

HASH_ASSERT_OWN(), hash_get_mutex(), hash_get_nth_mutex(): Remove.

ib_recreate(): Merge to the only caller, buf_pool_resize_hash().

ib_create(): Merge to the callers.

ha_clear(): Merge to the only caller buf_pool_t::close().

buf_pool_t::create(): Merge the ib_create() and
hash_create_sync_obj() invocations.

ha_insert_for_fold_func(): Clarify an assertion.

buf_pool_t::page_hash_lock(): Simplify the logic.

hash_assert_can_search(), hash_assert_can_modify(): Remove.
These predicates were only being invoked for the adaptive hash index,
while they only are effective for buf_pool.page_hash.

HASH_DELETE_AND_COMPACT(): Merge to ha_delete_hash_node().

hash_get_sync_obj_index(): Remove.

hash_table_t::heaps[], hash_get_nth_heap(): Remove. It was actually unused!

hash_get_heap(): Remove. It was only used in ha_delete_hash_node(),
where we always use hash_table_t::heap.

hash_table_t::calc_hash(): Replaces hash_calc_hash().

MRI scripts cannot handle + in paths, and ubuntu CI makes use of
these.

So we remove the top level build dir from the script and
transform it into a relative path script.

Because of common dependencies between the
static libraries list can contain duplicates.

We reduce these down to the single last one in
the list.

This reduces the relative time of a rebuild from:

$ (cd builddir/; time make -j)
...
real	0m30.789s
user	1m33.477s
sys	0m19.678s

and the LIB entries
$ grep ADDLIB builddir/libmysqld/mysqlserver-\$\<CONFIG\>.mri.tpl  | wc -l
179

$ du -h builddir/libmysqld/libmariadbd.a
4.1G	builddir/libmysqld/libmariadbd.a

To:

$ (cd builddir/; time make -j)
...
real	0m20.139s
user	1m32.423s
sys	0m12.208s

$ grep ADDLIB builddir/libmysqld/mysqlserver-\$\<CONFIG\>.mri.tpl  | wc -l
25

$ du -h builddir/libmysqld/libmariadbd.a
688M	builddir/libmysqld/libmariadbd.a

from configuration files

Post-push fix: add mysqd options in backup string in mariabackup sst
script for the case of logging not in syslog.

The test mariabackup.mdev-14447 started failing due to the option
--apply-log-only that became invalid since
commit 9bdf35e9
and had been removed in
commit 8c71c6aa.

and put them all together in mysqld.cc

(if this option is present). perf, and other tools that do stack walking
works much better with it.

- There is a possiblity that metadata record is the only record in
the leftmost leaf page. So change the assertion to check only
previous page.

- This is a regression of MDEV-21174(56f6dab1).
InnoDB resets the BTR_EXTERN_LEN value at wrong offset.

galera_ipv6_mariabackup MTR tests

The galera.galera_parallel_autoinc_manytrx mtr test opens and runs test
scenario through 3 connections to node 1 and one connection to node 2.
In the test initialization phase, the test creates two tables 't1' and 'ten'
and then creates a stored procedure 'p1' to operate on these tables.
These 3 create DDL statements are issued through same connection to node 1.

In the next test phase, the mtr script uses send command to launch the call
for the p1 stored procedure through all 3 connections to node 1 and through
one connection to node 2. As the mtr send command is asynchronous,
this test phase is non blocking and fast operation.
Now, if the replication between nodes is slow, it may happen that the
initialization phase DDL statements have not been received or have not been
fully applied in node 2. Therefore there is no guarantee that the test tables
and the stored procedure have been created in node 2. Yet, the test is trying
to call p1 in node 2.

In the failure case error logs, there is error message
"MTR failed: query 'reap' failed: 1305: PROCEDURE test.p1 does not exist"

The reap command through connection to node 2, is the first place where test
execution may observe that test tables and/or stored procedure are not yet
created in node 2.

The fix in this commit adds a wait condition in connection to node 2, to wait
until the stored procedure is created before calling the stored procedure.
The wait is implemented by looking in information_schema.routines for the p1
stored procedure.

MDEV-22125 : galera.galera_drop_multi MTR failed: InnoDB: MySQL is trying to drop database `fts`.`` though there are still open handles

MDEV-22140 galera.galera_drop_database MTR failed: InnoDB: MySQL is trying to drop database `fts`.`` though there are still open handles

Add wait conditions to wait that all operations are done in both
nodes.

Make sure to initialize SSL early enough, when encryption plugins is loaded

            increased scalability through even distribution

Rollback segments are allocated to transactions in round-robin fashion.
This is controlled by incrementing a static-scope counter named rseg_slot.

Said logic is not protected by any mutex or use of atomic for the counter.
This potentially can cause the same rollback segment to get allocated to
N different transactions (requesting allocation at the same time).
While this is not an issue as a rollback segment can host multiple
transactions from contention (performance) perspective it is better to
allocate these rollback segments in round-robin fashion.

Fix for the said issue ports use of atomic for the said counter that would
ensure the original design semantic (even distribution through round-robin)
is retained.

MDEV-22370 safe_mutex: Trying to lock uninitialized mutex at /data/src/10.4-bug/sql/rpl_parallel.cc, line 470 upon shutdown during FTWRL

Problem:- When we issue FTWRL with shutdown in parallel, there is race between
FTWRL and shutdown. Shutdown might destroy the mutex (pool->LOCK_rpl_thread_pool)
before FTWRL can lock it. So we can get crash on FTWRL thread

Solution:- mysql_mutex_destroy(pool->LOCK_rpl_thread_pool) should wait for
FTWRL thread to complete its work , and then destroy.
So slave_prepare_for_shutdown will just deactivate the pool, and mutex is destroyed
later in end_slave()

The galera.galera_parallel_autoinc_manytrx mtr test opens and runs test
scenario through 3 connections to node 1 and one connection to node 2.
In the test initialization phase, the test creates two tables 't1' and 'ten'
and then creates a stored procedure 'p1' to operate on these tables.
These 3 create DDL statements are issued through same connection to node 1.

In the next test phase, the mtr script uses send command to launch the call
for the p1 stored procedure through all 3 connections to node 1 and through
one connection to node 2. As the mtr send command is asynchronous,
this test phase is non blocking and fast operation.
Now, if the replication between nodes is slow, it may happen that the
initialization phase DDL statements have not been received or have not been
fully applied in node 2. Therefore there is no guarantee that the test tables
and the stored procedure have been created in node 2. Yet, the test is trying
to call p1 in node 2.

In the failure case error logs, there is error message
"MTR failed: query 'reap' failed: 1305: PROCEDURE test.p1 does not exist"

The reap command through connection to node 2, is the first place where test
execution may observe that test tables and/or stored procedure are not yet
created in node 2.

The fix in this commit adds a wait condition in connection to node 2, to wait
until the stored procedure is created before calling the stored procedure.
The wait is implemented by looking in information_schema.routines for the p1
stored procedure.

galera_ipv6_mariabackup MTR tests

MDEV-21298: mariabackup doesn't read from the [mariadbd] and [mariadbd-X.Y]
server option groups from configuration files
MDEV-21301: mariabackup doesn't read [mariadb-backup] option group in
configuration file

All three issues require to change the same code, that is why their
fixes are joined in one commit.

The fix is in invoking load_defaults_or_exit() and handle_options() for
backup-specific groups separately from client-server groups to let the last
handle_options() call fail on unknown backup-specific options.

The order of options procesing is the following:
1) Load server groups and process server options, ignore unknown
options
2) Load client groups and process client options, ignore unknown
options
3) Load backup groups and process client-server options, exit on
unknown option
4) Process --mysqld-args command line options, ignore unknown options

New global flag my_handle_options_init_variables was added to have
ability to invoke handle_options() for the same allowed options set
several times without re-initialising previously set option values.

--password value destroying is moved from option processing callback to
mariabackup's handle_options() function to have ability to invoke server's
handle_options() several times for the same possible allowed options
set.

Galera invokes wsrep_sst_mariabackup.sh with mysqld command line
options to configure mariabackup as close to the server as possible.
It is not known what server options are supported by mariabackup when the
script is invoked. That is why new mariabackup option "--mysqld-args" is added,
all unknown options that follow this option will be silently ignored.

wsrep_sst_mariabackup.sh was also changed to:
- use "--mysqld-args" mariabackup option to pass mysqld options,
- remove deprecated innobackupex mode,
- remove unsupported mariabackup options:
    --encrypt
    --encrypt-key
    --rebuild-indexes
    --rebuild-threads

MEM_GET_VBITS(): Save information about uninitialized data.

MEM_SET_VBITS(): Restore information about uninitialized data.

Make ut_new_get_key_by_file event less expensive
remove binary search, compute auto_event_keys offset at compile time.

Replace all references to /usr/sbin/mysqld (and bin and libexec) with
mariadbd, so that the binary server will always be 'mariadbd'.

Also update all places that reference the server binary in other ways,
such as AppArmor profiles and scripts that previously expected to find
a 'mysqld' in process lists.