The test main.mysqltest could crash or hang with
cmake -DWITH_ASAN=ON builds. The reason appears to be
a memory leak, which was found out by manually invoking
echo --replace_regex a > file
ASAN_OPTIONS=log_path=/dev/shm/asan mysqltest ... < file
and then examining the /dev/shm/asan.* file.
Improve the test that was imported and adapted for MariaDB in
commit fb217449dc.
row_undo_step(): Move the DEBUG_SYNC point from trx_rollback_for_mysql().
This DEBUG_SYNC point is executed after rolling back one row.
trx_rollback_for_mysql(): Clarify the comments that describe the scenario,
and remove the DEBUG_SYNC point.
If the statement "if (trx->has_logged_persistent())" and its body are
removed from trx_rollback_for_mysql(), then the test
innodb.xa_recovery_debug will fail because the transaction would still
exist in the XA PREPARE state. If we allow the XA COMMIT statement
to succeed in the test, we would observe an incorrect state of the
XA transaction where the table would contain row (1,NULL). Depending
on whether the XA transaction was committed, the table should either
be empty or contain the record (1,1). The intermediate state of
(1,NULL) should never be observed after completed recovery.
Adapt the test that was added in
mysql/mysql-server@6b65d9032c
but omitted in commit 2e814d4702.
Instead of triggering a log checkpoint, we will only trigger
a redo log flush before killing the server.
Note: the mtr.commit() call in trx_rollback_for_mysql()
will not actually make the undo log header page state change durable.
A call to log_write_up_to(mtr.commit_lsn(), true) would do that.
It is unclear what the originally reported bug scenario was.
As long as innobase_rollback_by_xid() will not return without
ensuring that the redo log has been durably written, we should be safe.
When binlog is disabled, WSREP will not behave correctly when
SAVEPOINT ROLLBACK is executed since we don't register handlers for such case.
Fixed by registering WSREP handlerton for SAVEPOINT related commands.
FOREIGN_KEY_CHECKS is disabled
- dict_foreign_find_index() can return NULL if InnoDB already dropped
the foreign index when FOREIGN_KEY_CHECKS is disabled.
If async slave thread (slave SQL handler), becomes a BF victim, it may occasionally happen that rollbacker thread is used to carry out the rollback instead of the async slave thread.
This can happen, if async slave thread has flagged "idle" state when BF thread tries to figure out how to kill the victim.
The issue was possible to test by using a galera cluster as slave for external master, and issuing high load of conflicting writes through async replication and directly against galera cluster nodes.
However, a deterministic mtr test for the "conflict window" has not yet been worked on.
The fix, in this patch makes sure that async slave thread state is never set to IDLE. This prevents the rollbacker thread to intervene.
The wsrep_query_state change was refactored to happen by dedicated function to make controlling the idle state change in one place.
redo log during recovery
- InnoDB unnecessarily reads the page even though it has fully initialized
buffered redo log records. Allow the page initialization redo log to
apply for the page in buf_page_get_gen() during recovery.
- Renamed buf_page_get_gen() to buf_page_get_low()
- Newly added buf_page_get_gen() will check for buffered redo log for
the particular page id during recovery
- Added new function buf_page_mtr_lock() which basically latches the page
for the given latch type.
- recv_recovery_create_page() is inline function which creates a page
if it has page initialization redo log records.
Update submodule, change the source file list
rename CACHE_LINE_SIZE in ut0counter, so it does not conflics with
rocksdb sources, which also defines this constant now.
Making a linked list of dtuple_t is needed only for inserting
records. It's better to store tuples in a non-intrusive
container to not affect all other use cases of dtuple_t
dtuple_t::tuple_list: removed, it was 2 * sizeof(void*) bytes
ins_node_t::entry_list: now it's std::vector<dtuple_t*>
ins_node_t::entry: now it's std::vector<dtuple_t*>::iterator
DBUG_EXECUTE_IF("row_ins_skip_sec": this dead code removed
move span.h to a proper place to make it available for the whole server
Reformat it.
Constuctors from a contigous container are fixed
to use cont.data() instead of cont.begin()
span<>::index_type is replaced with span<>::size_type
In my micro-benchmarks memcmp(4196) 3 times faster than old
implementation. Also, it's generally better to use as less
reinterpret_casts<> as possible.
buf_is_zeroes(): renamed from buf_page_is_zeroes() and
argument changed to span<> for convenience.
st_::span<T>::const_iterator: fixed
page_zip-verify_checksum(): make argument byte* instead of void*
prepared statement
The method Item_func_in::build_clone() that builds a clone item for an
Item_func_in item first calls a generic method Item_func::build_item()
that builds the the clones for the arguments of the Item_func_in item
to be cloned, creates a copy of the Item_func_in object and attaches the
clones for the arguments to this copy. Then the method Item_func_in::build_clone()
makes the copy fully independent on the copied object in order to
guarantee a proper destruction of the clone. The fact is the copy of the
Item_func_in object is registered as any other item object and should be
destructed as any other item object.
If the method Item_func::build_item fails to build a clone of an argument
then it returns 0. In this case no copy of the Item_func_in object should
be created. Otherwise the finalizing actions for this copy would not be
performed and the copy would remain in a state that would prevent its
proper destruction.
The code of Item_func_in::build_clone() before this patch created the copy
of the Item_func_in object before cloning the argument items. If this
cloning failed the server crashed when trying to destruct the copy item.
The code of Item_row::build_clone() was changed similarly to the code of
Item_func::build_clone though this code could not cause any problems.
Remove CREATE/DROP database.
Remove some unnecessary suppressions, replacements, and
SQL statements.
Populate tables via have_sequence.inc to avoid the creation of
explicit InnoDB record locks in INSERT...SELECT. This will remove
some gaps in AUTO_INCREMENT values.
After MDEV-12353, the consistency check that I originally added for
commit 1b9fe0bbac
(InnoDB Plugin for MySQL 5.1) started randomly failing.
It turns out that the IMPORT TABLESPACE code was always incorrect:
it did not update the (redundantly stored) tablespace ID
in index tree root pages. It only does that for page headers
and BLOB pointers.
PageConverter::update_index_page(): Update the tablespace ID
in the BTR_SEG_TOP and BTR_SEG_LEAF of index root pages.
The problem happened in these line:
uval0= (ulonglong) (val0_negative ? -val0 : val0);
uval1= (ulonglong) (val1_negative ? -val1 : val1);
return check_integer_overflow(val0_negative ? -(longlong) res : res,
!val0_negative);
when unary minus was performed on -9223372036854775808.
This behavior is undefined in C/C++.
Several macros such as sint2korr() and uint4korr() are using the
arithmetic + operator while a bitwise or operator would suffice.
GCC 5 and clang 5 and later can detect patterns consisting of
bitwise or and shifts by multiples of 8 bits, such as those used
in the InnoDB function mach_read_from_4(). They actually translate
that verbose low-level code into high-level machine language
(i486 bswap instruction or fused into the Haswell movbe instruction).
We should do the same for MariaDB Server code that is outside InnoDB.
Note: The Microsoft C compiler is lacking this optimization.
There, we might consider using _byteswap_ushort(), _byteswap_ulong(),
_byteswap_uint64(). But, those would lead to unaligned reads, which are
bad for reasons stated in MDEV-20277. Besides, outside InnoDB,
most data is already being stored in the native little-endian format
of that compiler.
fil_delete_tablespace(): Remove the unused parameter drop_ahi,
and add the parameter if_exists=false. We want to suppress
error messages if we know that the tablespace has been discarded.
dict_table_rename_in_cache(): Pass the new parameter to
fil_delete_tablespace(), that is, do not complain about
missing tablespace if the tablespace has been discarded.
row_make_new_pathname(): Declare as static.
row_drop_table_for_mysql(): Tolerate !table->data_dir_path
when the tablespace has been discarded.
row_rename_table_for_mysql(): Skip part of the RENAME TABLE
when fil_space_get_first_path() returns NULL.
buf_pool_resize(): Simplify the fault injection
for innodb.buf_pool_resize_oom.
innodb.buf_pool_resize_oom: Use a small buffer pool.
innodb.innodb_buffer_pool_load_now: Make use of the sequence engine,
to avoid creating explicit InnoDB record locks. Clean up the
accesses to information_schema.innodb_buffer_page_lru.
- This issue is caused by MDEV-19176
(bba59abb03).
- Problem is that there is miscalculation of available memory during
recovery if innodb_buffer_pool_instances > 1.
- Ignore the buffer pool instance while calculating available_memory
- Removed recv_n_pool_free_frames variable and use buf_pool_get_n_pages()
instead.
Temporary tables are typically short-lived, and temporary tables
are assumed to be accessed only by the thread that is handling
the owning connection. Hence, they must not be subject to
defragmenting.
ha_innobase::optimize(): Do not add temporary tables to
the defragment_table() queue.
AUTO_INCREMENT values are nondeterministic after crash recovery.
While MDEV-6076 guarantees that the AUTO_INCREMENT values of committed
transactions will not roll back, it is possible that the AUTO_INCREMENT
values will be durably incremented for incomplete transactions. So
changing the test case to avoid showing the result of AUTO_INCREMENT value.
Rows_log_event::change_to_flashback_event(): Reduce the scope
of the variable swap_buff2, and do not duplicate conditions.
GCC 9.3.0 flagged the -Wmaybe-uninitialized when compiling the
10.5 branch using cmake -DWITH_ASAN=ON -DCMAKE_CXX_FLAGS=-O2
myrg_open(): Reduce the scope of the variable 'end' and
simplify the code.
For some reason, I got no warning for this code in the 10.2
branch, only 10.3 or later.
The ENGINE=MERGE is covered by the tests main.merge, main.merge_debug,
and main.merge-big.
