Pass the line numbers (only) to PERFORMANCE_SCHEMA

This reverts commit e34e53b5
and defines os_thread_sleep() is a macro on Windows.

The latching order checks for rw-locks have not caught many bugs
in the past few years and they are greatly complicating the code.

Last time the debug checks were useful was in
commit 59caf2c3 (MDEV-13485).

The B-tree hang MDEV-14637 was not caught by LatchDebug,
because the granularity of the checks is not sufficient
to distinguish the levels of non-leaf B-tree pages.

The interface was already made dead code by the parent commit.

This will also disable the LatchDebug and sync_array interface
to InnoDB rw-locks.

FIXME: Pass line number information to PERFORMANCE_SCHEMA
for dict_index_t::lock

FIXME: Test&document the changes to innotop

Both dict_index_t::lock and buf_block_t::lock depend on
re-entrant 'update' and 'write' locks. The two main users
of recursion are the change buffer merge and operations on
BLOB columns. In both cases, a 'sub-mini-transaction' is being used.
It is much easier to implement re-entrant locks than to implement
the migration of blocks between mini-transactions.

The 'update' lock mode allows any number of concurrent reads,
but no concurrent 'update' or 'write'. While an 'update' lock is
present, 'read' lock requests may be served immediately, even if
'write' lock requests are pending. This allows the 'update' lock
holder thread to acquire one redundant 'read' lock.

An 'update' lock may be upgraded to 'write', but it involves a special
operation, separate from x_lock(). All existing 'update' lock references
in the mini-transaction must be changed to 'write'. This is because after
an x_unlock() operation on mtr_t::commit(), u_unlock() would not be
permitted. In the old rw_lock_t this was permitted, because the lock word
itself included an X-lock re-entrancy count, in addition to including
the S-lock count.

mtr_t::page_lock(): Renamed from buf_page_mtr_lock().

buf_page_try_get_func(): Try U latch only, to avoid recursive S latch.

There only exist two types of interesting InnoDB rw_lock_t anymore:
buf_block_t::lock and dict_index_t::lock. We can remove more of
the debug fields.

FIXME: Iterate dict_index_t::lock via dict_sys everywhere

FIXME: Move count_os_waits to dict_index_t, and introduce something
similar for block mutexes

FIXME: Pass the line number information to performance_schema

FIXME: Remove PSI_RWLOCK_SHAREDLOCK and friends, just use
PSI_RWLOCK_READLOCK and friends.

For some reason, InnoDB debug tests on Windows fail due to rw_lock_t
if the function call overhead for some os_thread_ code is removed.

This change worked fine on Windows in combination with MDEV-24142.

Let us always base srw_lock on our own std::atomic<uint32_t>
based rw_lock. In this way, we can extend the locks in a portable
way across all platforms.

We will use futex system calls where available:
Linux, OpenBSD, and Microsoft Windows.

Elsewhere, we will emulate futex with a mutex and a condition variable.

Thanks to Daniel Black for testing this on OpenBSD.

os_thread_pf(): Remove.

os_thread_eq(), os_thread_yield(), os_thread_get_curr_id():
Define as macros.

ut_print_timestamp(), ut_sprintf_timestamp(): Simplify.

When executing set operations in a pipeline using only one temporary table
additional scans of intermediate results may be needed. The scans are
performed with usage of the rnd_next() handler function that might
leave record buffers used for the temporary table not in a state that
is good for following writes into the table. For example it happens for
aria engine when the last call of rnd_next() encounters only deleted
records. Thus a cleanup of record buffers is needed after each such scan
of the temporary table.

Approved by Oleksandr Byelkin <sanja@mariadb.com>

This change is needed in 10.5 to avoid extra malloc calls in val_str().
In 10.6 it's not needed anymore but the extra +1 byte doesn't harm
that much.

- Fold long comment rows and updated comments
- Moved one private function in class Item_func_rand among other private
  functions

This is for Oracle compatiblity. ENABLED is in Oracle the default case
and just ensures that the NOT NULL constraints will be tested, which is
also default in MariaDB

A side effect of MDEV-16264 is that a large number of threads will
be created at server startup, to be destroyed after a minute or two.

One source of such thread creation is srv_start_periodic_timer().
InnoDB is creating 3 periodic tasks: srv_master_callback (1Hz)
srv_error_monitor_task (1Hz), and srv_monitor_task (0.2Hz).

It appears that we can merge srv_error_monitor_task and srv_monitor_task
and have them invoked 4 times per minute (every 15 seconds). This will
affect our ability to enforce innodb_fatal_semaphore_wait_threshold and
some computations around BUF_LRU_STAT_N_INTERVAL.

We could remove srv_master_callback along with the DROP TABLE queue
at some point of time in the future. We must keep it independent
of the innodb_fatal_semaphore_wait_threshold detection, because
the background DROP TABLE queue could get stuck due to dict_sys
being locked by another thread. For now, srv_master_callback
must be invoked once per second, so that
innodb_flush_log_at_timeout=1 can work.

BUF_LRU_STAT_N_INTERVAL: Reduce the precision and extend the time
from 50*1 second to 4*15 seconds.

srv_error_monitor_timer: Remove.

MAX_MUTEX_NOWAIT: Increase from 20*1 second to 2*15 seconds.

srv_refresh_innodb_monitor_stats(): Avoid a repeated call to time(NULL).
Change the interval to less than 60 seconds.

srv_monitor(): Renamed from srv_monitor_task.

srv_monitor_task(): Renamed from srv_error_monitor_task().
Invoked only once in 15 seconds. Invoke also srv_monitor().
Increase the fatal_cnt threshold from 10*1 second to 1*15 seconds.

sync_array_print_long_waits_low(): Invoke time(NULL) only once.
Remove a bogus message about printouts for 30 seconds. Those
printouts were effectively already disabled in MDEV-16264
(commit 5e62b6a5).

The purpose of the InnoDB page cleaner subsystem is to write out
modified pages from the buffer pool to data files. When the
innodb_max_dirty_pages_pct_lwm is not exceeded or
innodb_adaptive_flushing=ON decides not to write out anything,
the page cleaner should keep sleeping indefinitely until the state
of the system changes: a dirty page is added to the buffer pool such
that the page cleaner would no longer be idle.

buf_flush_page_cleaner(): Explicitly note when the page cleaner is idle.
When that happens, use mysql_cond_wait() instead of mysql_cond_timedwait().

buf_flush_insert_into_flush_list(): Wake up the page cleaner if needed.

innodb_max_dirty_pages_pct_update(),
innodb_max_dirty_pages_pct_lwm_update():
Wake up the page cleaner just in case.

Note: buf_flush_ahead(), buf_flush_wait_flushed() and shutdown are
already waking up the page cleaner thread.

Kudos to Marko for finding.

This partially reverts commit 6479006e.

Remove the constant tpool::aio::N_PENDING, which has no
intrinsic meaning for the tpool.

tpool::aio::N_PENDING: Replaces OS_AIO_N_PENDING_IOS_PER_THREAD.
This limits two similar things: the number of outstanding requests
that a thread may io_submit(), and the number of completed requests
collected at a time by io_getevents().

In the asynchronous I/O interface, InnoDB is invoking io_getevents()
with a timeout value of half a second, and requesting exactly 1 event
at a time.

The reason to have such a short timeout is to facilitate shutdown.

We can do better: Use an infinite timeout, wait for a larger maximum
number of events. On shutdown, we will invoke io_destroy(), which
should lead to the io_getevents system call reporting EINVAL.

my_getevents(): Reimplement the libaio io_getevents() by only invoking
the system call. The library implementation would try to elide the
system call and return 0 immediately if aio_ring_is_empty() holds.
Here, we do want a blocking system call, not 100% CPU usage. Neither
do we want the aio_ring_is_empty() trigger SIGSEGV because it is
dereferencing some memory that was freed by io_destroy().

For reading trx_t::state we can avoid acquiring trx_t::mutex.
Atomic load and store should be similar to normal load and store
on most instruction set architectures. The atomicity of the operation
would merely prohibit the compiler from reordering some operations.

We must avoid acquiring a latch while we are already holding one.
The tablespace latch was being acquired recursively in some
operations that allocate or free pages.