This fixes up 4489a89c
and the test binlog.binlog_truncate_multi_engine
that indirectly invokes the DBUG_SUICIDE.

In commit 1193a793 we
set innodb_use_native_aio=OFF when using io_uring
on a kernel where write requests could potentially be lost.

The last reproducible issue was fixed in Linux 5.16-rc1
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v5.16-rc1&id=d3e3c102d107bb84251455a298cf475f24bab995
and the fix was backported to 5.15.3.

Hence, using a 5.16 or later kernel should be fine.
The Debian kernel 5.15.0-1-amd64 (5.15.3-1) was tested.
On Debian, utsname::release or uname -r does not reflect the
exact minor version while utsname::version and uname -v does.

On Fedora however the utsname::version is rather different:

  $ uname -r
  5.14.20-200.fc34.x86_64
  $ uname -v
  #1 SMP Thu Nov 18 22:03:20 UTC 2021

As such we use the version, but fall back to the release if
there isn't the beginnings of a kernel version in the version.

Thanks to Daniel Black for reporting the Linux kernel bug and
Jens Axboe for actually fixing it.
Co-Authored-By: Daniel Black <daniel@mariadb.org>

Closes: #1953

Sys_lc_messages, Sys_lc_time_names: Use offsetof(MY_LOCALE, ...)
because the type has standard layout.

The macro my_offsetof() performs pointer arithmetics that may be
undefined behavior. As reported in MDEV-26272, it may cause
clang -fsanitize=undefined to generate invalid memory references.

struct PFS_events_statements: Convert to std::is_standard_layout
by encapsulating the standard-layout struct PFS_events instead of
deriving from it, so that the standard macro offsetof() can be used.

PFS_events_statements::copy(): Renamed from copy_events_statements().
A cast to void* is now needed in memcpy() to avoid GCC -Wclass-memaccess
"writing to an object ... leaves 64 bytes unchanged".

mdl_hash_function(): Use plain offsetof; MDL_key has standard layout.

There's no need for Debian to set config items to their
default. Left commented user, datadir and tmpdir as
these may want to be changed. lc-messages and skip-external-locks
are so infrequently set even listing them looks overly verbose.

socket left uncommented in [client-server] as various client
implementations may have different defaults compiled in.

Currently, SST scripts assume that the filename specified in
the --log-bin-index argument either does not contain an extension
or uses the standard ".index" extension. Similar assumptions are
used for the log_bin_index parameter read from the configuration
file. This commit adds support for arbitrary extensions for the
index file paths.

If the server is started with the --innodb-force-recovery argument
on the command line, then during SST this argument can be passed to
mariabackup only at the --prepare stage, and accordingly it must be
removed from the --mysqld-args list (and it is not should be passed
to mariabackup otherwise).

This commit fixes a flaw in the SST scripts and add a test that
checks the ability to run the joiner node in a configuration that
uses --innodb-force-recovery=1.

buf_LRU_scan_and_free_block(): It turns out that even with
-fno-expensive-optimizations, GCC 4.8.5 may fail to split an instruction.
For the non-embedded server, -O1 would fail and -Og would seem to work,
while the embedded server build seems to require -O0.

buf_block_init(): Correct the MemorySanitizer instrumentation.

buf_page_get_low(): Do not read dirty data from read-fixed blocks.
These data races were identified by MemorySanitizer. If a read-fixed
block is being accessed, we must acquire and release a page latch,
so that the read-fix (and the exclusive page latch) will be released
and it will be safe to read the page frame contents if needed,
even before acquiring the final page latch. We do that in
buf_read_ahead_linear() and for the allow_ibuf_merge check.

mtr_t::page_lock(): Assert that the block is not read-fixed.

buf_page_get_low(): When we are creating an uncompressed page frame
for a ROW_FORMAT=COMPRESSED page, we must release the buf_pool.page_hash
latch and buf_pool.mutex while waiting for other threads to release their
fixes on the block.

This was caught by an occasional hang of the test innodb_zip.bug56680.

The following options were introduced in
commit 2e814d47 (mariadb-10.2.2)
and have little use:

innodb_disable_resize_buffer_pool_debug had no effect even in
MariaDB 10.2.2 or MySQL 5.7.9. It was introduced in
mysql/mysql-server@5c4094cf4971eebab89da4ee4ae92c71f69cd524
to work around a problem that was fixed in
mysql/mysql-server@2957ae4f990bf3aed25822b0ce15d3ccad0b54b6
(but the parameter was not removed).

innodb_page_cleaner_disabled_debug and innodb_master_thread_disabled_debug
are only used by the test innodb.redo_log_during_checkpoint
that will be removed as part of this commit.

innodb_dict_stats_disabled_debug is only used by that test,
and it is redundant because one could simply use
innodb_stats_persistent=OFF or the STATS_PERSISTENT=0 attribute
of the table in the test to achieve the same effect.

32-bit variable must be expanded to 64-bit before shift left.

buf_page_t::frame: Moved from buf_block_t::frame.
All 'thin' buf_page_t describing compressed-only ROW_FORMAT=COMPRESSED
pages will have frame=nullptr, while all 'fat' buf_block_t
will have a non-null frame pointing to aligned innodb_page_size bytes.
This eliminates the need for separate states for
BUF_BLOCK_FILE_PAGE and BUF_BLOCK_ZIP_PAGE.

buf_page_t::lock: Moved from buf_block_t::lock. That is, all block
descriptors will have a page latch. The IO_PIN state that was used
for discarding or creating the uncompressed page frame of a
ROW_FORMAT=COMPRESSED block is replaced by a combination of read-fix
and page X-latch.

page_zip_des_t::fix: Replaces state_, buf_fix_count_, io_fix_, status
of buf_page_t with a single std::atomic<uint32_t>. All modifications
will use store(), fetch_add(), fetch_sub(). This space was previously
wasted to alignment on 64-bit systems. We will use the following encoding
that combines a state (partly read-fix or write-fix) and a buffer-fix
count:

buf_page_t::NOT_USED=0 (previously BUF_BLOCK_NOT_USED)
buf_page_t::MEMORY=1 (previously BUF_BLOCK_MEMORY)
buf_page_t::REMOVE_HASH=2 (previously BUF_BLOCK_REMOVE_HASH)
buf_page_t::FREED=3 + fix: pages marked as freed in the file
buf_page_t::UNFIXED=1U<<29 + fix: normal pages
buf_page_t::IBUF_EXIST=2U<<29 + fix: normal pages; may need ibuf merge
buf_page_t::REINIT=3U<<29 + fix: reinitialized pages (skip doublewrite)
buf_page_t::READ_FIX=4U<<29 + fix: read-fixed pages (also X-latched)
buf_page_t::WRITE_FIX=5U<<29 + fix: write-fixed pages (also U-latched)
buf_page_t::WRITE_FIX_IBUF=6U<<29 + fix: write-fixed; may have ibuf
buf_page_t::WRITE_FIX_REINIT=7U<<29 + fix: write-fixed (no doublewrite)

buf_page_t::write_complete(): Change WRITE_FIX or WRITE_FIX_REINIT to
UNFIXED, and WRITE_FIX_IBUF to IBUF_EXIST, before releasing the U-latch.

buf_page_t::read_complete(): Renamed from buf_page_read_complete().
Change READ_FIX to UNFIXED or IBUF_EXIST, before releasing the X-latch.

buf_page_t::can_relocate(): If the page latch is being held or waited for,
or the block is buffer-fixed or io-fixed, return false. (The condition
on the page latch is new.)

Outside buf_page_get_gen(), buf_page_get_low() and buf_page_free(), we
will acquire the page latch before fix(), and unfix() before unlocking.

buf_page_t::flush(): Replaces buf_flush_page(). Optimize the
handling of FREED pages.

buf_pool_t::release_freed_page(): Assume that buf_pool.mutex is held
by the caller.

buf_page_t::is_read_fixed(), buf_page_t::is_write_fixed(): New predicates.

buf_page_get_low(): Ignore guesses that are read-fixed because they
may not yet be registered in buf_pool.page_hash and buf_pool.LRU.

buf_page_optimistic_get(): Acquire latch before buffer-fixing.

buf_page_make_young(): Leave read-fixed blocks alone, because they
might not be registered in buf_pool.LRU yet.

recv_sys_t::recover_deferred(), recv_sys_t::recover_low():
Possibly fix MDEV-26326, by holding a page X-latch instead of
only buffer-fixing the page.

MDEV-23855 and MDEV-23399 already moved some transient data fields
from buffer pool page descriptors to IORequest, but the write buffer
of PAGE_COMPRESSED or ENCRYPTED tables was missed. Since is only needed
during asynchronous page write requests, it belongs to IORequest.

btr_cur_optimistic_latch_leaves(): Use transactional_shared_lock_guard.

btr_cur_latch_leaves(): Avoid acquiring some page latches, because
the changes are already blocked by index->lock.

btr_cur_search_to_nth_level_func(): Remove a redundant variable
retrying_for_search_prev=!!prev_tree_blocks, and avoid acquiring
some page latches.

dict_stats_recalc_pool_del(): Always reposition the iterators after
releasing and reacquiring the mutex. Another thread could have modified
recalc_pool, causing reallocation of the underlying memory while
we were waiting.

This fixes a regression that was caused by
commit 45a05fda (MDEV-25919).

MDEV-23805 simplified the treatment of empty tables during ALTER TABLE,
which could prevent the scenarios that were previously reported and
fixed as MDEV-16131 and MDEV-24730.

With the MDEV-23805 fix, the statement
SET DEBUG_SYNC = 'now WAIT_FOR copied';
could occasionally time out, depending on timing.
Apparently, there was a race condition where purge could resume
(and empty the table) before ALTER TABLE got the chance to execute.
We must prevent the purge of history from running before
ALTER TABLE has started executing.

fil_space_decrypt(): change signature to return status via dberr_t only.
Also replace impossible condition with an assertion and prove it via
test cases.

  This bug affected queries with two or more references to a CTE referring
another CTE if the definition of the latter contained an invocation of
a stored function that used a base table. The bug could lead to a bogus
error message or to an assertion failure.
  For any non-first reference to CTE cte1 With_element::clone_parsed_spec()
is called that parses the specification of cte1 to construct the unit
structure for this usage of cte1. If cte1 refers to another CTE cte2
outside of the specification of cte1 then With_element::clone_parsed_spec()
has to be called for cte2 as well. This call is made by the function
LEX::resolve_references_to_cte() within the invocation of the function
With_element::clone_parsed_spec() for cte1.
  When the specification of a CTE is parsed all table references encountered
in it must be added to the global list of table references for the query.
As the specification for the non-first usage of a CTE is parsed at a
recursive call of the parser the function With_element::clone_parsed_spec()
invoked at this recursive call should takes care of appending the list of
table references encountered in the specification of this CTE cte1 to the
list of table references created for the query. And it should do it after
the call of LEX::resolve_references_to_cte() that resolves references to
CTEs defined outside of the specification of cte1 because this call may
invoke the parser again for specifications of other CTEs and  the table
references from their specifications must ultimately appear in the global
list of table references of the query.
  The code of With_element::clone_parsed_spec() misplaced the call of
LEX::resolve_references_to_cte(). As a result LEX::query_tables_last used
for the query that was supposed to point to the field 'next_global' of the
last element in the global list of table references actually pointed to
'next_global' of the previous element.
  The above inconsistency certainly caused serious problems when table
references used in the stored functions invoked in cloned specifications
of CTEs were added to the global list of table references.

In commit 7ae21b18 (MDEV-12353)
the recovery of ROW_FORMAT=COMPRESSED tables was changed.
Changes would be logged in a physical format for the compressed
page image, so that the page need not be decompressed or compressed
during recovery.

page_zip_write_rec(): Log any update of the delete-mark flag in the
ROW_FORMAT=COMPRESSED page.

page_zip_dir_insert(): Copy the delete-mark flag. A delete-marked
record may be inserted by btr_cur_pessimistic_update() via
btr_cur_insert_if_possible(), page_cur_tuple_insert(),
page_cur_insert_rec_zip(). In the observed scenario, it was
an ROLLBACK. Presumably, the test case involved repeated DELETE
and INSERT of the same key, or updating a key back and forth.
This change alone might make the adjustment in page_zip_write_rec()
redundant, but we play it safe because we failed to create a
minimal test case for this scenario.

If the server is killed during any DDL operation that is about to
delete an .ibd file, recovery could crash when attempting to load
the table definition of the being-dropped table. By design of
commit 1bd681c8 (MDEV-25506 part 3),
a table whose name starts with #sql-ib in the data dictionary may
belong to an uncommitted transaction. So, we must ignore any missing
SYS_COLUMNS, SYS_FIELDS, and SYS_VIRTUAL records for such tables.

The "ID mismatch" error messages were misleading; they really mean
"record not found".