ibuf.size, ibuf.max_size: Changed the type to Atomic_relaxed<ulint>
in order to fix some (not all) race conditions.
ibuf_contract(): Renamed from ibuf_merge_pages(ulint*).
ibuf_merge(), ibuf_merge_all(): Removed.
srv_shutdown(): Invoke log_free_check() and ibuf_contract(). Even though
ibuf_contract() is not writing anything, it will trigger calls of
ibuf_merge_or_delete_for_page(), which will write something. Because
we cannot invoke log_free_check() at that low level, we must invoke
it at the high level.
srv_shutdown_print(): Replaces srv_shutdown_print_master_pending().
Report progress and remaining work every 15 seconds. For the
change buffer merge, the remaining work is indicated by ibuf.size.
The InnoDB write-ahead log ib_logfile0 is of fixed size,
specified by innodb_log_file_size. If the tail of the log
manages to overwrite the head (latest checkpoint) of the log,
crash recovery will be broken.
Let us clarify the messages about this, including adding
a message on the completion of a log checkpoint that notes
that the dangerous situation is over.
To reproduce the dangerous scenario, we will introduce the
debug injection label ib_log_checkpoint_avoid_hard, which will
avoid log checkpoints even harder than the previous
ib_log_checkpoint_avoid.
log_t::overwrite_warned: The first known dangerous log sequence number.
Set in log_close() and cleared in log_write_checkpoint_info(),
which will output a "Crash recovery was broken" message.
Every operation that is going to write redo log is supposed to
invoke log_free_check() before acquiring any latches. If there
is a risk of log buffer overrun, a log checkpoint would be
triggered by that call.
ibuf_merge_space(), ibuf_merge_in_background(),
ibuf_delete_for_discarded_space(): Invoke log_free_check()
when the current thread is not holding any page latches.
Unfortunately, in lower-level code called from ibuf_insert()
or ibuf_merge_or_delete_for_page(), some page latches may be
held and a call to log_free_check() could hang.
ibuf_set_bitmap_for_bulk_load(): Use the caller's mini-transaction.
The caller should have invoked log_free_check() while not holding
any page latches.
The lock is created during page splitting after moving records and
locks(lock_move_rec_list_(start|end)()) to the new page, and inheriting
the locks to the supremum of left page from the successor of the infimum
on right page.
There is no need in such inheritance for READ COMMITTED isolation level
and not-gap locks, so the fix is to add the corresponding condition in
gap lock inheritance function.
One more fix is to forbid gap lock inheritance if XA was prepared. Use the
most significant bit of trx_t::n_ref to indicate that gap lock inheritance
is forbidden. This fix is based on
mysql/mysql-server@b063e52a83
See also commits aa8a31da and 64678c for a Bug #22990029 fix.
In this scenario INSERT chose to check if delete unmarking is available for
a just deleted record. To build an update vector, it needed to calculate
the vcols as well. Since this INSERT was not IGNORE-flagged, recalculation
failed.
Solutiuon: temporarily set abort_on_warning=true, while calculating the
column for delete-unmarked insert.
As of now innodb does not store trx_id for each record in secondary index.
The idea behind is following: let us store only per-page max_trx_id, and
delete-mark the records when they are deleted/updated.
If the read starts, it rememders the lowest id of currently active
transaction. Innodb refers to it as trx->read_view->m_up_limit_id.
See also ReadView::open.
When the page is fetched, its max_trx_id is compared to m_up_limit_id.
If the value is lower, and the secondary index record is not delete-marked,
then this page is just safe to read as is. Else, a clustered index could be
needed ato access. See page_get_max_trx_id call in row_search_mvcc, and the
corresponding switch (row_search_idx_cond_check(...)) below.
Virtual columns are required to be updated in case if the record was
delete-marked. The motivation behind it is documented in
Row_sel_get_clust_rec_for_mysql::operator() near
row_sel_sec_rec_is_for_clust_rec call.
This was basically a description why virtual column computation can
normally happen during SELECT, and, generally, a vcol index access.
Sometimes stats tables are updated by innodb. This starts a new
transaction, and it can happen that it didn't finish to the moment of
SELECT execution, forcing virtual columns recomputation. If the result was
a something that normally outputs a warning, like division by zero, then
it could be outputted in a racy manner.
The solution is to suppress the warnings when a column is computed
for the described purpose.
ignore_wrnings argument is added innobase_get_computed_value.
Currently, it is only true for a call from
row_sel_sec_rec_is_for_clust_rec.
btr_search_guess_on_hash() would only acquire an index page latch if it
is invoked with ahi_latch=NULL. If it's invoked from
row_sel_try_search_shortcut_for_mysql() with ahi_latch!=NULL, a page
will not be latched, and row_search_mvcc() will get a pointer to the
record, which can be changed by some other transaction before the record
was stored in result buffer with row_sel_store_mysql_rec() call.
ahi_latch argument of btr_cur_search_to_nth_level_func() and
btr_pcur_open_with_no_init_func() is used only for
row_sel_try_search_shortcut_for_mysql().
btr_cur_search_to_nth_level_func(..., ahi_latch !=0, ...) is invoked
only from btr_pcur_open_with_no_init_func(..., ahi_latch !=0, ...),
which, in turns, is invoked only from
row_sel_try_search_shortcut_for_mysql().
I suppose that separate case with ahi_latch!=0 was intentionally
implemented to protect row_sel_store_mysql_rec() call in
row_search_mvcc() just after row_sel_try_search_shortcut_for_mysql()
call. After the ahi_latch was moved from row_seach_mvcc() to
row_sel_try_search_shortcut_for_mysql(), there is no need in it at all
if btr_search_guess_on_hash() latches a page unconditionally. And if
btr_search_guess_on_hash() latched the page, any access to the record in
row_sel_try_search_shortcut_for_mysql() after btr_pcur_open_with_no_init()
call will be protected with the page latch.
The fix is to remove ahi_latch argument from
btr_pcur_open_with_no_init_func(), btr_cur_search_to_nth_level_func()
and btr_search_guess_on_hash().
There will not be test, as to test it we need to freeze some SELECT
execution in the point between row_sel_try_search_shortcut_for_mysql()
and row_sel_store_mysql_rec() calls in row_search_mvcc(), and to change
the record in some other transaction to let row_sel_store_mysql_rec() to
store changed record in result buffer. Buf we can't do this with the
fix, as the page will be latched in btr_search_guess_on_hash() call.
row_purge_get_partial(): Replaces trx_undo_rec_get_partial_row().
Also copy the purge_node_t::ref to the purge_node_t::row.
In this way, the clustered index key fields will always be
available, even if thanks to
commit d384ead0f0 (MDEV-14799)
they would no longer be repeated in the remaining part of the
undo log record.
The futex system calls were introduced in Linux 2.6.0,
which was released in December 2003. It should be safe to assume
that the system calls are always available on the Linux kernels
that MariaDB Server 10.3 would run on.
Let us use the normal platform-specific preprocessor symbols
__linux__, __sun__, _AIX instead of some homebrew ones.
The preprocessor symbol UNIV_HPUX must have lost its meaning
by f6deb00a56 (note: the symbol
UNIV_HPUX10 is being checked for, but only UNIV_HPUX is defined).
Reason:
=======
Race condition between btr_search_drop_hash_index() and
btr_search_lazy_free(). One thread does resizing of buffer pool
and clears the ahi on all pages in the buffer pool, frees the
index and table while removing the last reference. At the same time,
other thread access index->heap in btr_search_drop_hash_index().
Solution:
=========
Acquire the respective ahi latch before checking index->freed()
btr_search_drop_page_hash_index(): Added new parameter to indicate
that drop ahi entries only if the index is marked as freed
btr_search_check_marked_free_index(): Acquire all ahi latches and
return true if the index was freed
Starting with commit da094188f6 (MDEV-24393),
MariaDB will no longer acquire advisory file locks on InnoDB data
files by default, because it would create a large number of
entries in Linux /proc/locks.
The motivation for acquiring the file locks is to prevent accidental
concurrent startup of multiple server processes on the same data files.
Such mistake still turns out to be relatively common, based on
corruption bug reports from the community.
To prevent corruption due to concurrent startup attempts, the
Aria storage engine would unconditionally acquire an advisory lock
on one of its log files.
Solution: InnoDB will always lock its system tablespace files.
(Ever since commit 685d958e38
the InnoDB log file will not necessarily be open while the
server is running, because it can be accessed via memory-mapped I/O.)
If more protection is desired, then the option --external-locking
can be used.
The mandatory advisory lock also fixes intermittent failures of
some crash recovery tests. It turns out that when the mtr test harness
kills and restarts the server, it will not actually ensure that the
old process has terminated before starting the new one.
fil_space_t::acquire_low(): Introduce a parameter that specifies
which flags should be avoided. At all times, referenced() must not
be incremented if the STOPPING flag is set. When fil_system.mutex
is not being held by the current thread, the reference must not be
incremented if the CLOSING flag is set (unless NEEDS_FSYNC is set,
in fil_space_t::flush()).
fil_space_t::acquire(): Invoke acquire_low(STOPPING | CLOSING).
In this way, the reference count cannot be incremented after
fil_space_t::try_to_close() invoked fil_space_t::set_closing().
If the CLOSING flag was set, we must retry acquire_low() after
acquiring fil_system.mutex.
fil_space_t::prepare_acquired(): Replaces prepare(true).
fil_space_t::acquire_and_prepare(): Replaces prepare().
This basically retries fil_space_t::acquire() after
acquiring fil_system.mutex.
buf_page_print(): Dump the buffer page 32 bytes (64 hexadecimal digits)
per line. In this way, the limitation in mtr
("Data too long for column 'line'") will not be triggered.
Also, do not bother decoding the page contents, because everything
is present in the hexadecimal output.
dict_index_find_on_id_low(): Merge to dict_index_get_if_in_cache_low().
The direct call in buf_page_print() was prone to crashing, in case the
table definition was concurrently evicted or dropped from the
data dictionary cache.
MariaDB never supported this form of preemption via high-priority
transactions. This error code shold not have been added in the
first place, in commit 2e814d4702.
srv_do_purge(): In commit edde1f6e0d
when the de-facto 32-bit trx_sys_t::history_size() was replaced with
32-bit trx_sys.rseg_history_len, some more variables were changed
from ulint (size_t) to uint32_t.
The history list length is the number of committed transactions whose
undo logs are waiting to be purged. Each TRX_RSEG_HISTORY list is
storing the number of entries in a 32-bit field and each transaction
will occupy at least one undo log page. It is thinkable that the
length of each TRX_RSEG_HISTORY list may approach the maximum
representable number. The number cannot be exceeded, because the
rollback segment header is allocated from the same tablespace as
the undo log header pages it is pointing to, and because the page
numbers of a tablespace are stored in 32 bits. In any case, it is
possible that the total number of unpurged committed transactions
cannot be represented in 32 but 39 bits (corresponding to
128 rollback segments and undo tablespaces).
page_cur_insert_rec_low(): When checking for common bytes with
the preceding record, exclude the header bytes of next_rec
that could have been updated by this function.
The scenario where this caused corruption was an insert of
a node pointer record. The child page number was written as
0x203 but recovered as 0x103 because the n_owned field of next_rec
was changed from 1 to 2 before the comparison was invoked.
The counter srv_stats.key_rotation_list_length is never updated, and
therefore Innodb_encryption_key_rotation_list_length will always be 0.
The view INFORMATION_SCHEMA.INNODB_TABLESPACES_ENCRYPTION comes close
to reporting this information.
The counters were added in commit 5e55d1ced5
and any code to update them was
inadvertently removed in commit 2e814d4702
when applying InnoDB changes from MySQL 5.7.
Let us remove these counters that never reported anything useful. If such
statistics are really needed in a special case, they can be obtained by
instrumenting the code by some means, such as eBPF or a source code patch.
