TOCTOU bug. The path is checked to be valid, symlinks are resolved.
Then the resolved path is opened. Between the check and the open,
there's a window when one can replace some path component with a
symlink, bypassing validity checks.
Fix: after we resolved all symlinks in the path, don't allow open()
to resolve symlinks, there should be none.
Compared to the old MyISAM/Aria code:
* fastpath. Opening of not-symlinked files is just one open(),
no fn_format() and lstat() anymore.
* opening of symlinked tables doesn't do fn_format() and lstat() either.
it also doesn't to realpath() (which was lstat-ing every path
component), instead if opens every path component with O_PATH.
* share->data_file_name stores realpath(path) not readlink(path). So,
SHOW CREATE TABLE needs to do lstat/readlink() now (see ::info()),
and certain error messages (cannot open file "XXX") show the real
file path with all symlinks resolved.
Fixed wrong calculation of buffer sizes. ulong datatype was used wrongly,
as were the casts to ulong. Buffer sizes should be of type size_t,
not ulong, or bad things happen on 64 bit Windows.
This patch changes pagecache struct to use size_t/ssize_t
where long/ulong were previously used. Also, removed several casts.
Both aria and myisam storage engines feature a logic path in
thr_find_all_keys that leads to undefined behaviour by bypassing the
initialization code of variables after my_thread_init().
By refactoring the nested logic into a separate function, this problem
is resolved.
Aria service threads are created "joinable", but they're not "joined" on
completion. This causes memory leaks around thread local storage.
Fixed by joining service thread. Simplified relevant code and cleaned up
relevant valgrind suppressions.
Problem was that repair() did lock and unlock tables, which leaved already locked tables in wrong state
include/my_check_opt.h:
Added option T_NO_LOCKS to disable locking during repair()
Fixed duplicated bit T_NO_CREATE_RENAME_LSN
mysql-test/suite/rpl/r/myisam_external_lock.result:
Test case for MDEV-6871
mysql-test/suite/rpl/t/myisam_external_lock-slave.opt:
Test case for MDEV-6871
mysql-test/suite/rpl/t/myisam_external_lock.test:
Test case for MDEV-6871
storage/maria/ha_maria.cc:
Don't lock tables during enable_indexes()
Removed some calls to current_thd
storage/myisam/ha_myisam.cc:
Don't lock tables during enable_indexes()
Removed some calls to current_thd
stat structure (from <sys/stat.h>) is conditionally defined
to have different layout and size depending on the defined macros.
The correct macro is defined in my_config.h, which means it MUST be
included first (or, at least before <features.h> - so, practically,
before including any system headers).
This bug only happens in case of paritioned tables used in LOCK TABLES and implicit_commit() was called
(as part of trying to execute a CREATE TABLE withing lock tables)
The problem was that Aria could not move the tables from one transaction to the new one, as thd->open_tables contained
a partitioned tables and not an Aria table.
Fix:
- Store a list of all open tables that are part of a share in share->open_tables
- In maria::implict_commit() use transaction->used_tables & share->open_tables to find out which tables
was part of the current transaction instead of using thd->open_tables, which may contain partitioned tables.
mysql-test/suite/maria/maria_partition.result:
Added test case
mysql-test/suite/maria/maria_partition.test:
Added test case
storage/maria/ha_maria.cc:
Use trn->used tables and share->open_tables to find out which tables was part of the current transaction instead of using thd->open_tables.
storage/maria/ma_close.c:
Remove closed table from share->open_list
storage/maria/ma_open.c:
Add table to share->open_list
storage/maria/ma_state.c:
Added comment
storage/maria/maria_def.h:
Added share->open_list, a list of all tables that is using this share.
MDEV-6281 Typo in mysql_install_db scripts
and collateral changes:
* remove mysql_tableinfo.1 and references to it (there's no mysql_tableinfo)
* for debian: create manpages for mysqlrepair, mysqlanalyze, mysqloptimize
(as symlinks for mysqlcheck.1, just as executables are symlinks to mysqlcheck)
* remove mysqlmanager.8 and references to it
* correct "very long line" error in mysqladmin.1
* simplify and fix table formatting in mysqlbinlog.1 and mysqldump.1
* fix a typo in the help text in mysql_install_db
* aria_chk: say "for Linux on x86_64", like other tools do
(not "for Linux at x86_64")
* add simple manpages for aria_* utilities
The problem was that a big record was allocated on the stack, which casued stack to run out.
Fixed by using my_safe_alloca() instead of my_alloca() when allocating records.
Now only records <= 16384 are allocated on the stack.
mysql-test/r/stack-crash.result:
Added test case
mysql-test/t/stack-crash.test:
Added test case
storage/maria/ma_blockrec.c:
Use my_safe_alloca() instead of my_alloca()
storage/maria/ma_dynrec.c:
Use my_safe_alloca() instead of my_alloca()
storage/maria/maria_def.h:
Added MARIA_MAX_RECORD_ON_STACK
storage/maria/maria_pack.c:
Use my_safe_alloca() instead of my_alloca()
Problem was that ALTER TABLE DISABLE KEYS incremented create_trid for the table,
which made the new index entries invisible until the global trid catched up.
Fixed by only updating create_trid if we are rewriting all rows and indexes.
mysql-test/suite/maria/alter.result:
Added test case
mysql-test/suite/maria/alter.test:
Added test case
storage/maria/ha_maria.cc:
Only updating create_trid if we are doing a full repair (and thus rewriting all rows and indexes).
storage/maria/trnman.c:
More DBUG_PRINT
There was 2 problems:
1) coping/moving of the same type (usually casting) as sizeof() (solved in different ways depends on the cause);
2) using 'const' in SSL_CTX::getVerifyCallback() which return object (not reference) and so copy of the object will be created and 'const' has no sens.
- Let _ma_record_pos() set SEARCH_PART_KEY when doing a search on
a prefix of a [unique] key. Otherwise, _ma_search_pos() would
find the first key equal to search key, and assume it is also
the last one, which will make a wrong estimate of key's position.
A wrong key position may cause min_pos > max_pos and records_in_range()
will return 0, which will make the optimizer think it's an impossible
range while in fact it is not.
Backport the fix olav.sandstaa@sun.com-20101102184747-qfuntqwj021imy9r:
"Fix for Bug#52660 Perf. regr. using ICP for MyISAM on range queries on an index containing TEXT"
(together with further fixes in that code) into MyISAM and Aria.
Partitioning didn't store the name of default storage engine for partitions
in the frm file - it only store the typecode. Typecodes aren't stable, and
might vary depending on the order in which storage engines are loaded (can
be changed even from my.cnf without recompilation).
As a temporary workaround for 5.5, we hard-code Aria's typecode, to make sure it
never changes.
Fixed some cases that didn't work with > 4G buffers.
Fixed compiler warnings
include/mysql_com.h:
Avoid compiler warning with strncmp()
sql-common/client.c:
Fixed long comment; Added ()
sql/filesort.cc:
Fix code to get filesort to work with big buffers
sql/sys_vars.cc:
Fixed some cache variables that could be set to higher value than the size_t
Limit query cache to ULONG_MAX as the query cache buffer variables are ulong
storage/federatedx/ha_federatedx.cc:
Remove not used variable
storage/maria/ha_maria.cc:
Fix that bulk_insert() works with big buffers
storage/maria/ma_write.c:
Fix that bulk_insert() works with big buffers
storage/myisam/ha_myisam.cc:
Fix that bulk_insert() works with big buffers
storage/myisam/mi_write.c:
Fix that bulk_insert() works with big buffers
storage/sphinx/snippets_udf.cc:
Fixed compiler warnings
mysql-test/suite/maria/maria-autozerofill.result:
Updated result
mysql-test/suite/maria/maria-autozerofill.test:
Added test that zerofilled table should not give any warnings when table is used
mysql-test/suite/maria/maria-recovery2.result:
More tests to make it easier to find bugs
mysql-test/suite/maria/maria-recovery2.test:
More tests to make it easier to find bugs
storage/maria/ha_maria.cc:
Set create_trid after repair (needed if table was moved from another system)
Set uuid after repair (needed if table was moved from another system)
storage/maria/maria_chk.c:
Reset share->state.create_trid if we reset share->state.create_rename_lsn.
Make the table moveable
outside datafile) on INSERT into an Aria table.
The isssue was that the check if a table was moved between systems didn't take into account that create_trid could be bigger than the current max trid on the new system.
This could only happen if one tried to move a table that one had just done a 'REPAIR TABLE' on.
Tables that one had run 'aria_chk --zerofill' on worked.
Fixed this by assuming that if create_trid is too big then the table has been moved from one system to another and we have to do an automatic zerofill.
Other fixed:
- Added a check to detect a wrong create_trid in 'check table'.
- aria_chk -dvv will now write out also the create_trid (to make future error finding easier)
- aria_chk --zerofill doesn't anymore require a aria_control_file
- Removed some warnings from safemalloc when using aria_chk, ma_test1 and ma_test2.
include/myisamchk.h:
Removed wrong 'QQ' flags (the flags are used by myisamchk and aria_chk)
storage/maria/ha_maria.cc:
maria_chk_status() can now also return an error.
storage/maria/ma_check.c:
In maria_chk_status() check if create_trid value is too big.
storage/maria/ma_open.c:
Changed check if table is moved so that we can detect wrong create_trid values.
Don't set STATE_NOT_MOVABLE flag if we are doing repair/check. This was done so that aria_chk can print out the movable flag.
storage/maria/ma_test1.c:
Added code to suppress memory leaks from safemalloc
storage/maria/ma_test2.c:
Added code to suppress memory leaks from safemalloc
storage/maria/maria_chk.c:
Added code to suppress memory leaks from safemalloc.
Make help text a bit better for --HELP and --zerofill.
Incresed version number.
Don't require a control file if we are only doing --zerofill
Print out 'create_trid' when doing --describe --verbose
storage/maria/unittest/ma_test_recovery.expected:
Updated result file
MDEV-3989: Server crashes on import from MariaDB mysqldump export with partitioned Aria table.
Problem was that bulk insert in aria was not properly protected against concurrent selects.
storage/maria/ha_maria.cc:
Move settings of file->state to _ma_block_start_trans() to ensure that lock_key_trees is not changed by a concurrent bulk_insert.
storage/maria/ma_check.c:
Added DBUG_ASSERT()
storage/maria/ma_open.c:
Set start_trans to ma_start_trans for default behaviour.
storage/maria/ma_pagecrc.c:
Removed test for 'non_transactional' as a now_transactinal could be reset while a flush was happening.
storage/maria/ma_state.c:
Moved setting of info->state from external_lock to start_trans to protect against concurrent running bulk inserts.
This works as the other threads will wait in thr_lock() until bulk_insert is done and keys are re-generated.
storage/maria/ma_state.h:
Added _ma_start_trans()
This could happen when using Aria for internal temporary files (default case) and using DISTINCT.
_ma_scan_restore_block_record() didn't work correctly if there was rows inserted, updated or deleted on the handler
between calls to _ma_scan_remember_block_record() and _ma_scan_restore_block_record().
The effect was that some DISTINCT queries that used remove_dup_with_compare() could fail.
.bzrignore:
Ignore sql_yacc.hh
mysql-test/suite/maria/r/distinct.result:
Test case for MDEV-4280
mysql-test/suite/maria/t/distinct.test:
Test case for MDEV-4280
mysql-test/t/mysql.test:
Fixed test suite (we could get error -1 in some cases)
sql/sql_select.cc:
Break loop if restart_rnd_next() gives an error
storage/maria/ha_maria.cc:
scan_restore_pos() can return disk fault error.
storage/maria/ma_blockrec.c:
_ma_scan_remember_block_record() did incorrectly update scan.dir instead of scan_save.dir .
_ma_scan_restore_block_record() didn't work correctly if there was rows inserted,updated or deleted on the handler
between calls to _ma_scan_remember_block_record() and _ma_scan_restore_block_record().
Fixed by adding counters for row changes and reading the current scan page if changes had been made.
storage/maria/ma_blockrec.h:
scan_restore_pos() can return disk fault error.
storage/maria/ma_delete.c:
Increment row_changes
storage/maria/ma_scan.c:
scan_restore_pos() can return disk fault error.
storage/maria/ma_update.c:
Increment row_changes
storage/maria/ma_write.c:
Increment row_changes
storage/maria/maria_def.h:
scan_restore_pos() can return disk fault error.
Added comment to clearify the code.
storage/maria/ma_blockrec.c:
Added comment to clearify the code
In case of out of memory or disk error, mark pages with LSN_IMPOSSIBLE to make it easier to know which pages have wrong information.
