When InnoDB isn't using memory its polite to other system
processes to allow that memory to be used.

Create buf_page_t::set_os_unused, to madvise(MADV_FREE)/
DiscardVirtualMemory innodb buffer pool pages to keep
virtual mapping but allow kernel to reclaim memory.

set_os_unused is called by buf_LRU_block_free_non_file_page
for flushing deallocations to free up memory.

set_os_used is called by buf_block_t::initialise (used by
buf_page_create_low and buf_LRU_get_free_only.

Move MSan (memory sanitizer) directive to be associated with
set_os_{un,}used so that the correctness can be determined by
instrumentation and test cases rather than relying on OS behaviour
at specific times.

Remove existing page information poisoning in
buf_LRU_block_free_non_file_page.

Tested by: Matthias Leich
Reviewer: Marko Mäkelä
Windows Review: Vladislav Vaintroub

trx_undo_rseg_free(): Revert an inadvertent change that was done
as part of the merge a42c80bd

- During shutdown, InnoDB fts fails to update synced doc id
when there is only one doc id about to sync. While starting
the server, InnoDB fetches the already synced doc id from
config table. In the subsequent sync operation, InnoDB fails
with DB_DUPLICATE_KEY error.

In commit 244fdc43 (MDEV-29438)
we made sure that if the preceding record is the page infimum record,
no more than 8 bytes will be read from it. But, if the data payload of
the being-inserted record is less than 8 bytes (this can happen in
secondary indexes), we must not compare all 8 bytes.

This was caught by a failure of the test gcol.innodb_virtual_basic
under MemorySanitizer and some builds with AddressSanitizer.

- This is caused by commit 1bd681c8(MDEV-25506)
InnoDB removes the index from the table object before freeing the
leaf and non-leaf segments.

The resources like uring in MariaDB aren't intended for spawned
processes so we restrict access using the io_uring_ring_dontfork
liburing library call.

This bug was found in MariaDB Server 10.6 thanks to the
OPT_PAGE_CHECKSUM record that was implemented
in commit 4179f93d for catching
this type of recovery failures.

page_cur_insert_rec_low(): If the previous record is the page infimum,
correctly limit the end of the record. We do not want to copy data from
the header of the page supremum. This omission caused the incorrect
recovery of DB_TRX_ID in an instant ALTER TABLE metadata record, because
part of the DB_TRX_ID was incorrectly copied from the n_owned of the
page supremum, which in recovery would be updated after the copying,
but in normal operation would already have been updated at the time the
common prefix was being determined.

log_phys_t::apply(): If a data page is found to be corrupted, do not
flag the log corrupted but instead return a new status APPLIED_CORRUPTED
so that the caller may discard all log for this page. We do not want
the recovery of unrelated pages to fail in recv_recover_page().

No test case is included, because the known test case would only work
in 10.6, and even after this fix, it would trigger another bug in
instant ALTER TABLE crash recovery.

The replication unsafe warning's pattern gets corrected in the
punctuation part.

MDEV-28530: Revoking privileges from a non-existing user on a master breaks replication on the slave in the presence of replication filters

Problem:
========
Replication can break while applying a query log event if its
respective command errors on the primary, but is ignored by the
replication filter within Grant_tables on the replica. The bug
reported by MDEV-28530 shows this with REVOKE ALL PRIVILEGES using a
non-existent user. The primary will binlog the REVOKE command with
an error code, and the replica will think the command executed with
success because the replication filter will ignore the command while
accessing the Grant_tables classes. When the replica performs an
error check, it sees the difference between the error codes, and
replication breaks.

Solution:
========
If the replication filter check done by Grant_tables logic ignores
the tables, reset thd->slave_expected_error to 0 so that
Query_log_event::do_apply_event() can be made aware that the
underlying query was ignored when it compares errors.

Note that this bug also effects DROP USER if not all users exist
in the provided list, and the patch fixes and tests this case.

Reviewed By:
============
andrei.elkin@mariadb.com

These files are currently not being used nor compiled in MariaDB. The
use of large lists of 'case' statements in these source files are also
not a great way to represent translated strings. This git history can
be referred to when a better translation interface can be implemented
in the future.

Therefore, these files can be removed to cleanup the MariaDB codebase.

All new code of the whole pull request, including one or several files
that are either new files or modified ones, are contributed under the
BSD-new license. I am contributing on behalf of my employer
Amazon Web Services, Inc.

Moving these members:

   CHARSET_INFO *cs;
   const MY_UCA_WEIGHT_LEVEL *level;

from my_uca_scanner to a new separate structure my_uca_scanner_param.

Rationale:

During a comparison of two strings these members were initialized two times
(one time for every string).

After the change these members initialized only one time inside
a shared instance of my_uca_scanner_param, and the instance is
shared between two scanners (its const address is passed as new a parameter
to the underlying scanner functions).

This change gives a slight performance improvement (~5%).

- During rollback of DDL, InnoDB should copy the collation
changed column into the index heap

Spider converts HA_READ_KEY_EXACT to the equality (=) in the
function spider_db_append_key_where_internal() but the conversion
is not necessarily correct for tables with prefix indices.

We fix the bug by converting HA_READ_KEY_EXACT to 'LIKE "foo%"' when
a target key is a prefix key. The fix is partly inspired by FEDERATED.
See ha_federated::create_where_from_key() for more details.

btr_validate_level(): Invoke mtr.commit() after a failure.
This omission was introduced in
commit 0b47c126 (MDEV-13542).

If multiple threads invoke buf_page_get_low() on a ROW_FORMAT=COMPRESSED
page that does not reside in the buffer pool, then one of the threads
will end up acquiring an exclusive page latch (the "if" statement
right before the new wait_for_unzip: label) and other threads will
end up waiting for a shared latch while holding a buffer-fix.
The exclusive latch holder would then wait for the buffer-fixes to
be released while the buffer-fix holders are waiting for the shared latch.

buf_page_get_low(): Prevent the hang that was introduced
in commit 9436c778 (MDEV-27058),
by releasing the buffer-fix, sleeping some time, and retrying the
page lookup.

recv_sys_t::free_corrupted_page(): Identify the corrupted page in
an error or warning message.

buf_page_free(): Just in case, register the page as modified.
This should already have been done in mtr_t::free() as part of
fseg_free_page_low().

mtr_t::memo_push(): Simplify a condition, so that when invoked
with MTR_MEMO_PAGE_X_MODIFY, we will do the right thing.

fseg_free_page_low(): Remove an accidentally added return statement
that prevented mtr_t::free() from being called. This fixes a regression
that was introduced in
commit 0b47c126 (MDEV-13542).

Reviewed by: Nayuta Yanagisawa

Use temporary constraint names for temporary tables. The constraints
are not added to cache (skipped in dict_table_rename_in_cache()).

The scheme for temporary constraint names is as follows:

    for old table: db_name/\xFFconstraint_name
    for new table: db_name/\xFF\xFFconstraint_name

normalize_table_name_c_low(): wrong comparison "less than FN_REFLEN -
1". Somewhere array of FN_REFLEN includes the trailing 0, somewhere
array of FN_REFLEN + 1 includes trailing 0, but nowhere array of
FN_REFLEN - 1 must include trailing 0.

All InnoDB internal SQL functions should be moved to sql_funcs.h

add_back_last_deleted_lock() was called when the lock was never
removed. Lock is removed in finalize_atomic_replace() in
close_all_tables_for_name(). finalize_atomic_replace() is done only
for successful operation.

In non-atomic codepath it drops the table first, if anything fails
later we don't need to return back the lock since there is no table
now. So the fix is required as well.

Related to MDEV-24176.

1. vcol_fix_expr() generates new tree changes:
Type_std_attributes::agg_item_set_converter() does change_item_tree().

The changes are allocated on expr_arena (via Vcol_expr_context as per
MDEV-24176).

2. vcol_cleanup_expr() doesn't remove these changes (can be a bug of
Type_std_attributes or per design).

3. Atomic CREATE OR REPLACE renames old table to backup
(finalize_atomic_replace()). It does that via
rename_table_and_triggers() and that closes table share and releases
expr_arena root. Hence now we have Item corpses in thd->change_list.

4. PS cleanup phase tries to rollback thd->change_list and accesses
already freed item corpses.

The fix saves and restores change_list on
vcol_fix_expr()/vcol_cleanup_expr().