At each mini-transaction commit, the log sequence number of the
mini-transaction must be written to each modified page, so that
it will be available in the FIL_PAGE_LSN field when the page is
being read in crash recovery.
InnoDB was unnecessarily allocating redundant storage for the
field, in buf_page_t::newest_modification. Let us access
FIL_PAGE_LSN directly.
Furthermore, on ALTER TABLE...IMPORT TABLESPACE, let us write
0 to FIL_PAGE_LSN instead of using log_sys.lsn.
buf_flush_init_for_writing(), buf_flush_update_zip_checksum(),
fil_encrypt_buf_for_full_crc32(), fil_encrypt_buf(),
fil_space_encrypt(): Remove the parameter lsn.
buf_page_get_newest_modification(): Merge with the only caller.
buf_tmp_reserve_compression_buf(), buf_tmp_page_encrypt(),
buf_page_encrypt(): Define static in the same compilation unit
with the only caller.
PageConverter::m_current_lsn: Remove. Write 0 to FIL_PAGE_LSN
on ALTER TABLE...IMPORT TABLESPACE.
Almost all threads have gone
- the "ticking" threads, that sleep a while then do some work)
(srv_monitor_thread, srv_error_monitor_thread, srv_master_thread)
were replaced with timers. Some timers are periodic,
e.g the "master" timer.
- The btr_defragment_thread is also replaced by a timer , which
reschedules it self when current defragment "item" needs throttling
- the buf_resize_thread and buf_dump_threads are substitutes with tasks
Ditto with page cleaner workers.
- purge workers threads are not tasks as well, and purge cleaner
coordinator is a combination of a task and timer.
- All AIO is outsourced to tpool, Innodb just calls thread_pool::submit_io()
and provides the callback.
- The srv_slot_t was removed, and innodb_debug_sync used in purge
is currently not working, and needs reimplementation.
The page allocation bitmaps in the extent descriptor pages
contain two bits per page: XDES_FREE_BIT and XDES_CLEAN_BIT,
which is unused. Simplify read access.
xdes_is_free(descr,mtr): Remove. Use !xdes_get_n_used(descr) instead.
xdes_is_free(): Replaces xdes_get_bit(), xdes_mtr_get_bit().
xdes_find_free(): Replaces xdes_find_bit().
fsp_seg_inode_page_get_nth_inode(): Remove the redundant parameters
physical_size, mtr.
fsp_seg_inode_page_find_used(), fsp_seg_inode_page_find_free():
Remove the redundant parameter mtr.
The function mach_read_ulint() is a wrapper for the lower-level
functions mach_read_from_1(), mach_read_from_2(), mach_read_from_8().
Invoke those functions directly, for better readability of the code.
mtr_t::read_ulint(), mtr_read_ulint(): Remove. Yes, we will lose the
ability to assert that the read is covered by the mini-transaction.
We would still check that on writes, and any writes that
wrongly bypass mini-transaction logging would likely be caught by
stress testing with Mariabackup.
A conflict between MDEV-19514 (b42294bc64)
and MDEV-20934 (d7a2401750)
was resolved. We will not invoke the function ibuf_delete_recs()
from ibuf_merge_or_delete_for_page(). Instead, we will add that
logic to the function ibuf_read_merge_pages().
dict_table_rename_in_cache(): Use strcpy() instead of strncpy(),
because they are known to be equivalent in this case (the length
of old_name was already validated).
mariabackup: Invoke strncpy() with one less than the buffer size,
and explicitly add NUL as the last byte of the buffer.
Based on the performance testing that was conducted in MDEV-17492,
the InnoDB adaptive hash index could only help performance in specific,
almost-read-only workloads. It could slow down all kinds of workloads
(especially DROP TABLE, TRUNCATE TABLE, ALTER TABLE, or DROP INDEX
operations), and it can become corrupted, causing crashes (such as
MDEV-18815, MDEV-20203) and possibly data corruption. Furthermore,
the adaptive hash index consumes space from the InnoDB buffer pool,
which could hurt performance when the working set would almost fit
in the buffer pool.
Given all this, it is best to disable the adaptive hash index by default.
if ${CRC32_LIBRARY} target has no COMPILE_FLAGS yet,
GET_TARGET_PROPERTY returns COMPILE_FLAGS-NOTFOUND, which
doesn't work very well when it's later fed back into COMPILE_FLAGS.
GET_PROPERTY() returns an empty string in this case.
Cherry-picked from 10.4 - 13e8f728ec
Signed-off-by: Vicențiu Ciorbaru <vicentiu@mariadb.org>
only my_getopt should use it, because it changes my_getopt's behavior.
If one simply wants to skip the separator - don't ask it to be added
in the first place
process all --defaults* options uniformly,
get rid of special case for --no-defaults and --print-defaults
use realpath instead of blindly concatenating pwd and relative path.
When "--export" mariabackup option is used, mariabackup starts the server in
bootstrap mode to generate *.cfg files for the certain innodb tables.
The started instance of the server reads options from the file, pointed
out in "--defaults-file" mariabackup option.
If the server uses the same config file as mariabackup, and binlog is
switched on in that config file, then "mariabackup --prepare --export"
will create binary log files in the server's binary log directory, what
can cause issues.
The fix is to add "--skip-log-bin" in mysld options when the server is
started to generate *.cfg files.
The general reason why innodb redo log file is limited by 512G is that
log_block_convert_lsn_to_no() returns value limited by 1G. But there is no
need to have unique log block numbers in log group. The fix removes 512G
limit and limits log group size by
(uint32_t maximum value) * (minimum page size), which, in turns, can be
removed if fil_io() is no longer used for innodb redo log io.
Added the condition in innochecksum tool to check page id mismatch.
This could catch the write corruption caused by InnoDB.
Added the debug insert inside fil_io() to check whether it writes
the page to wrong offset.
cmake -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_BUILD_TYPE=Debug
Maintainer mode makes all warnings errors. This patch fix warnings. Mostly about
deprecated `register` keyword.
Too much warnings came from Mroonga and I gave up on it.
if ${CRC32_LIBRARY} target has no COMPILE_FLAGS yet,
GET_TARGET_PROPERTY returns COMPILE_FLAGS-NOTFOUND, which
doesn't work very well when it's later fed back into COMPILE_FLAGS.
GET_PROPERTY() returns an empty string in this case.
Server and command line tools now support option --tls_version to specify the
TLS version between client and server. Valid values are TLSv1.0, TLSv1.1, TLSv1.2, TLSv1.3
or a combination of them. E.g.
--tls_version=TLSv1.3
--tls_version=TLSv1.2,TLSv1.3
In case there is a gap between versions, the lowest version will be used:
--tls_version=TLSv1.1,TLSv1.3 -> Only TLSv1.1 will be available.
If the used TLS library doesn't support the specified TLS version, it will use
the default configuration.
Limitations:
SSLv3 is not supported. The default configuration doesn't support TLSv1.0 anymore.
TLSv1.3 protocol currently is only supported by OpenSSL 1.1.0 (client and server) and
GnuTLS 3.6.5 (client only).
Overview of TLS implementations and protocols
Server:
+-----------+-----------------------------------------+
| Library | Supported TLS versions |
+-----------+-----------------------------------------+
| WolfSSL | TLSv1.1, TLSv1,2 |
+-----------+-----------------------------------------+
| OpenSSL | (TLSv1.0), TLSv1.1, TLSv1,2, TLSv1.3 |
+-----------+-----------------------------------------+
| LibreSSL | (TLSv1.0), TLSv1.1, TLSv1,2, TLSv1.3 |
+-----------+-----------------------------------------+
Client (MariaDB Connector/C)
+-----------+-----------------------------------------+
| Library | Supported TLS versions |
+-----------+-----------------------------------------+
| GnuTLS | (TLSv1.0), TLSv1.1, TLSv1.2, TLSv1.3 |
+-----------+-----------------------------------------+
| Schannel | (TLSv1.0), TLSv1.1, TLSv1.2 |
+-----------+-----------------------------------------+
| OpenSSL | (TLSv1.0), TLSv1.1, TLSv1,2, TLSv1.3 |
+-----------+-----------------------------------------+
| LibreSSL | (TLSv1.0), TLSv1.1, TLSv1,2, TLSv1.3 |
+-----------+-----------------------------------------+
documentation
Apparently, WolfSSL wants to have *exactly* the same defines for
the user of the library as the was when building library itself.
A lot of #defines have an impact on ABI (structure sizes, alignment etc)
Using different recommended speedup options for WolfSSL.
- Enable x64 assembly code on Intel.
- in my_crypt.cc, align EVP_CIPHER_CTX buffer, since some members need
alignment of 16 (for AESNI instructions), when assembler is enabled.
- Adjust MY_AES_CTX_SIZE
- Enable fastmath in wolfssl (large integer math).
ARMv8 (AArch64) CPUs implement the CRC32 extension which is implemented by inline assembly ,
so they can also benefit from hardware acceleration in IO-intensive workloads.
The patch optimizes crc32c calculate with the armv8 crypto instruction(Intrinsics) when available
rather than original linear crc instructions.
Change-Id: I05d36a64c726d910c47befad93390108f4e6567f
Signed-off-by: Yuqi Gu <yuqi.gu@arm.com>
There is only one InnoDB crash recovery subsystem.
Allocating recv_sys statically removes one level of pointer indirection
and makes code more readable, and removes the awkward initialization of
recv_sys->dblwr.
recv_sys_t::create(): Replaces recv_sys_init().
recv_sys_t::debug_free(): Replaces recv_sys_debug_free().
recv_sys_t::close(): Replaces recv_sys_close().
recv_sys_t::add(): Replaces recv_add_to_hash_table().
recv_sys_t::empty(): Replaces recv_sys_empty_hash().
In some cases it's possible that InnoDB redo log file header is re-written so,
that checkpoint lsn and checkpoint lsn offset are updated, but checkpoint
number stays the same. The fix is to re-read redo log header if at least one
of those three parametes is changed at backup start.
Repeat the logic of log_group_checkpoint() on choosing InnoDB checkpoint info
field on backup start. This does not influence backup correctness, but
simplifies bugs analysis.
The option innodb_rollback_segments was deprecated already in
MariaDB Server 10.0. Its misleadingly named replacement innodb_undo_logs
is of very limited use. It makes sense to always create and use the
maximum number of rollback segments.
Let us remove the deprecated parameter innodb_rollback_segments and
deprecate&ignore the parameter innodb_undo_logs (to be removed in a
later major release).
This work involves some cleanup of InnoDB startup. Similar to other
write operations, DROP TABLE will no longer be allowed if
innodb_force_recovery is set to a value larger than 3.
The transaction isolation levels READ COMMITTED and READ UNCOMMITTED
should behave similarly to the old deprecated setting
innodb_locks_unsafe_for_binlog=1, that is, avoid acquiring gap locks.
row_search_mvcc(): Reduce the scope of some variables, and clean up
the initialization and use of the variable set_also_gap_locks.
The parameter innodb_log_checksums that was introduced in MariaDB 10.2.2
via mysql/mysql-server@af0acedd88
does not make much sense. The original motivation of introducing this
parameter (initially called innodb_log_checksum_algorithm in
mysql/mysql-server@22ba38218e)
was that the InnoDB redo log used the slow and insecure innodb algorithm.
With hardware or SIMD accelerated CRC-32C, there should be no reason to
allow checksums to be disabled on the redo log.
The parameter innodb_encrypt_log already implies innodb_log_checksums=ON.
Let us deprecate the parameter innodb_log_checksums and always compute
redo log checksums, even if innodb_log_checksums=OFF is specified.
An upgrade from MariaDB 10.2.2 or later will only be possible after
using the default value innodb_log_checksums=ON. If the non-default
value innodb_log_checksums=OFF was in effect when the server was shut down,
a log block checksum mismatch will be reported and the upgraded server
will fail to start up.
- Add new submodule for WolfSSL
- Build and use wolfssl and wolfcrypt instead of yassl/taocrypt
- Use HAVE_WOLFSSL instead of HAVE_YASSL
- Increase MY_AES_CTX_SIZE, to avoid compile time asserts in my_crypt.cc
(sizeof(EVP_CIPHER_CTX) is larger on WolfSSL)
If required privilege is missing, dump the output from "SHOW GRANTS"
into mariabackup log.
This will help troubleshooting, and make the bug reproducible.
It's a micro optimization. On most platforms CPUs has instructions to
compare with 0 fast. DB_SUCCESS is the most popular outcome of functions
and this patch optimized code like (err == DB_SUCCESS)
BtrBulk::finish(): bogus assertion fixed
fil_node_t::read_page0(): corrected usage of os_file_read()
que_eval_sql(): bugus assertion removed. Apparently it checked that
the field was assigned after having been zero-initialized at
object creation.
It turns out that the return type of os_file_read_func() was changed
in mysql/mysql-server@98909cefbc (MySQL 5.7)
from ibool to dberr_t. The reviewer (if there was any) failed to
point out that because of future merges, it could be a bad idea to
change the return type of a function without changing the function name.
This change was applied to MariaDB 10.2.2 in
commit 2e814d4702 but the
MariaDB-specific code was not fully adjusted accordingly,
e.g. in fil_node_open_file(). Essentially, code like
!os_file_read(...) became dead code in MariaDB and later
in Mariabackup 10.2, and we could be dealing with an uninitialized
buffer after a failed page read.
