TABLE ... WRITE".
Memory and CPU hogging occured when connection which had to wait for table
lock was serviced by thread which previously serviced connection that was
killed (note that connections can reuse threads if thread cache is enabled).
One possible scenario which exposed this problem was when thread which
provided binlog dump to replication slave was implicitly/automatically
killed when the same slave reconnected and started pulling data through
different thread/connection.
The problem also occured when one killed particular query in connection
(using KILL QUERY) and later this connection had to wait for some table
lock.
This problem was caused by the fact that thread-specific mysys_var::abort
variable, which indicates that waiting operations on mysys layer should
be aborted (this includes waiting for table locks), was set by kill
operation but was never reset back. So this value was "inherited" by the
following statements or even other connections (which reused the same
physical thread). Such discrepancy between this variable and THD::killed
flag broke logic on SQL-layer and caused CPU and memory hogging.
This patch tries to fix this problem by properly resetting this member.
There is no test-case associated with this patch since it is hard to test
for memory/CPU hogging conditions in our test-suite.
differences in tables
Certain merge tables were wrongly reported as having incorrect definition:
- Some fields that are 1 byte long (e.g. TINYINT, CHAR(1)), might
be internally casted (in certain cases) to a different type on a
storage engine layer. (affects 4.1 and up)
- If tables in a merge (and a MERGE table itself) had short VARCHAR column (less
than 4 bytes) and at least one (but not all) tables were ALTER'ed (even to an
identical table: ALTER TABLE xxx ENGINE=yyy), table definitions went ouf of
sync. (affects 4.1 only)
This is fixed by relaxing a check for underlying conformance and setting
field type to FIELD_TYPE_STRING in case varchar is shorter than 4
when a table is created.
fixes).
The legend: on a replication slave, in case a trigger creation
was filtered out because of application of replicate-do-table/
replicate-ignore-table rule, the parsed definition of a trigger was not
cleaned up properly. LEX::sphead member was left around and leaked
memory. Until the actual implementation of support of
replicate-ignore-table rules for triggers by the patch for Bug 24478 it
was never the case that "case SQLCOM_CREATE_TRIGGER"
was not executed once a trigger was parsed,
so the deletion of lex->sphead there worked and the memory did not leak.
The fix:
The real cause of the bug is that there is no 1 or 2 places where
we can clean up the main LEX after parse. And the reason we
can not have just one or two places where we clean up the LEX is
asymmetric behaviour of MYSQLparse in case of success or error.
One of the root causes of this behaviour is the code in Item::Item()
constructor. There, a newly created item adds itself to THD::free_list
- a single-linked list of Items used in a statement. Yuck. This code
is unaware that we may have more than one statement active at a time,
and always assumes that the free_list of the current statement is
located in THD::free_list. One day we need to be able to explicitly
allocate an item in a given Query_arena.
Thus, when parsing a definition of a stored procedure, like
CREATE PROCEDURE p1() BEGIN SELECT a FROM t1; SELECT b FROM t1; END;
we actually need to reset THD::mem_root, THD::free_list and THD::lex
to parse the nested procedure statement (SELECT *).
The actual reset and restore is implemented in semantic actions
attached to sp_proc_stmt grammar rule.
The problem is that in case of a parsing error inside a nested statement
Bison generated parser would abort immediately, without executing the
restore part of the semantic action. This would leave THD in an
in-the-middle-of-parsing state.
This is why we couldn't have had a single place where we clean up the LEX
after MYSQLparse - in case of an error we needed to do a clean up
immediately, in case of success a clean up could have been delayed.
This left the door open for a memory leak.
One of the following possibilities were considered when working on a fix:
- patch the replication logic to do the clean up. Rejected
as breaks module borders, replication code should not need to know the
gory details of clean up procedure after CREATE TRIGGER.
- wrap MYSQLparse with a function that would do a clean up.
Rejected as ideally we should fix the problem when it happens, not
adjust for it outside of the problematic code.
- make sure MYSQLparse cleans up after itself by invoking the clean up
functionality in the appropriate places before return. Implemented in
this patch.
- use %destructor rule for sp_proc_stmt to restore THD - cleaner
than the prevoius approach, but rejected
because needs a careful analysis of the side effects, and this patch is
for 5.0, and long term we need to use the next alternative anyway
- make sure that sp_proc_stmt doesn't juggle with THD - this is a
large work that will affect many modules.
Cleanup: move main_lex and main_mem_root from Statement to its
only two descendants Prepared_statement and THD. This ensures that
when a Statement instance was created for purposes of statement backup,
we do not involve LEX constructor/destructor, which is fairly expensive.
In order to track that the transformation produces equivalent
functionality please check the respective constructors and destructors
of Statement, Prepared_statement and THD - these members were
used only there.
This cleanup is unrelated to the patch.
When INSERT is done over a view the table being inserted into is
checked to be unique among all views tables. But if the view contains
self-joined table an error will be thrown even if all tables are used under
different aliases.
The unique_table() function now also checks tables' aliases when needed.
Problem: DROP TRIGGER was not properly handled in combination
with slave filters, which made replication stop
Fix: loading table name before checking slave filters when
dropping a trigger.
- Starting time of a query sent by bootstrapping wasn't initialized
and starting time defaulted to 0. This later used value by NOW-
item and was translated to 1970-01-01 11:00:00.
- Marketing the time with thd->set_time() before the call to
mysql_parse resolves this issue.
- set_time was refactored to be part of the thd->init_for_queries-
process.
- Starting time of a query sent by file bootstrapping wasn't initialized
and starting time defaulted to 0. This later used value by the Now-
item and is translated to 1970-01-01 11:00:00.
- marking the time with thd->set_time() before the call to
mysql_parse resolves this issue.
Two problems here:
Problem 1:
While constructing the join columns list the optimizer does as follows:
1. Sets the join_using_fields/natural_join members of the right JOIN
operand.
2. Makes a "table reference" (TABLE_LIST) to parent the two tables.
3. Assigns the join_using_fields/is_natural_join of the wrapper table
using join_using_fields/natural_join of the rightmost table
4. Sets join_using_fields to NULL for the right JOIN operand.
5. Passes the parent table up to the same procedure on the upper
level.
Step 1 overrides the the join_using_fields that are set for a nested
join wrapping table in step 4.
Fixed by making a designated variable SELECT_LEX::prev_join_using to
pass the data from step 1 to step 4 without destroying the wrapping
table data.
Problem 2:
The optimizer checks for ambiguous columns while transforming
NATURAL JOIN/JOIN USING to JOIN ON. While doing that there was no
distinction between columns that are used in the generated join
condition (where ambiguity can be checked) and the other columns
(where ambiguity can be checked only when resolving references
coming from outside the JOIN construct itself).
Fixed by allowing the non-USING columns to be present in multiple
copies in both sides of the join and moving the ambiguity check
to the place where unqualified references to the join columns are
resolved (find_field_in_natural_join()).
- Removed not used variables and functions
- Added #ifdef around code that is not used
- Renamed variables and functions to avoid conflicts
- Removed some not used arguments
Fixed some class/struct warnings in ndb
Added define IS_LONGDATA() to simplify code in libmysql.c
I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
Bug#4968 "Stored procedure crash if cursor opened on altered table"
Bug#19733 "Repeated alter, or repeated create/drop, fails"
Bug#19182 "CREATE TABLE bar (m INT) SELECT n FROM foo; doesn't work from
stored procedure."
Bug#6895 "Prepared Statements: ALTER TABLE DROP COLUMN does nothing"
Bug#22060 "ALTER TABLE x AUTO_INCREMENT=y in SP crashes server"
Test cases for bugs 4968, 19733, 6895 will be added in 5.0.
Re-execution of CREATE DATABASE, CREATE TABLE and ALTER TABLE
statements in stored routines or as prepared statements caused
incorrect results (and crashes in versions prior to 5.0.25).
In 5.1 the problem occured only for CREATE DATABASE, CREATE TABLE
SELECT and CREATE TABLE with INDEX/DATA DIRECTOY options).
The problem of bugs 4968, 19733, 19282 and 6895 was that functions
mysql_prepare_table, mysql_create_table and mysql_alter_table were not
re-execution friendly: during their operation they used to modify contents
of LEX (members create_info, alter_info, key_list, create_list),
thus making the LEX unusable for the next execution.
In particular, these functions removed processed columns and keys from
create_list, key_list and drop_list. Search the code in sql_table.cc
for drop_it.remove() and similar patterns to find evidence.
The fix is to supply to these functions a usable copy of each of the
above structures at every re-execution of an SQL statement.
To simplify memory management, LEX::key_list and LEX::create_list
were added to LEX::alter_info, a fresh copy of which is created for
every execution.
The problem of crashing bug 22060 stemmed from the fact that the above
metnioned functions were not only modifying HA_CREATE_INFO structure in
LEX, but also were changing it to point to areas in volatile memory of
the execution memory root.
The patch solves this problem by creating and using an on-stack
copy of HA_CREATE_INFO (note that code in 5.1 already creates and
uses a copy of this structure in mysql_create_table()/alter_table(),
but this approach didn't work well for CREATE TABLE SELECT statement).
- Using DATA/INDEX DIRECTORY option on Windows put data/index file into
default directory because the OS doesn't support readlink().
- The procedure for changing data/index file directory is
different under Windows.
- With this fix we report a warning if DATA/INDEX option is used,
but OS doesn't support readlink().
table
ROW_FORMAT option is lost during CREATE/DROP INDEX.
This fix forces CREATE/DROP INDEX to retain ROW_FORMAT by instructing
mysql_alter_table() that ROW_FORMAT is not used during creating/dropping
indexes.
Problem: replication of LC_TIME_NAMES didn't work.
Thus, INSERTS or UPDATES using date_format() always
worked with en_US on the slave side.
Fix: adding ONE_SHOT implementation for LC_TIME_NAMES.
- CREATE PROCEDURE stores database name based on query context instead
of 'current database' as set by 'USE' according to manual.
The bug reporter interpret the filtering statements as bug for
DROP PROCEDURE based on this behavior.
- Removed the code which changes db context.
- Added code to check that a valid db was supplied.
(Mostly in DBUG_PRINT() and unused arguments)
Fixed bug in query cache when used with traceing (--with-debug)
Fixed memory leak in mysqldump
Removed warnings from mysqltest scripts (replaced -- with #)
Do not issue a 'read-only' error in case of DROP TEMPORARY TABLE on a non-existing temporary table.
Instead produce the correct "Unknown table" error or warning (in cases when the IF EXISTS clause was specified).
To a documentor: the part of the manual describing the 'read_only' system variable should be clarified to state the following:
"When the read_only variable is set to ON, all operations which create/update/drop tables are rejected with the exceptions for:
1. Any operation performed by the replication thread on a slave server
2. Any operation performed by a user that have the SUPER privilege
3. Any operation that creates/updates/drops only temporary tables"
The problem was that THD::row_count_func was zeroed too. It was zeroed
as a fix for bug 4905 "Stored procedure doesn't clear for "Rows affected"
However, the proper solution is not to zero, because THD::row_count_func has
been set to -1 already in mysql_execute_command(), a later fix, which obsoletes
the incorrect fix of #4095
The regression is caused by the fix for bug 14767. When INSERT ... SELECT
used a view in the SELECT list that was not inlined, and there was an
active transaction, the server could crash in Query_cache::invalidate.
On INSERT ... SELECT only the table being inserted into is invalidated.
Thus views that can't be inlined are skipped from invalidation.
The bug manifests itself in two ways so there is 2 test cases.
One checks that the only the table being inserted into is invalidated.
And the second one checks that there is no crash on INSERT ... SELECT.
sync using replicate-wild-ignore-table
Problem: changes in character set variables
before an action on an replication-ignored table
makes slave to forget new variable values.
Fix: initialize one_shot variables only when
4.1 -> 5.x replication is running.
Problem: SHOW CREATE TABLE printed garbage in table
name for tables having TURKISH I
(i.e. LATIN CAPITABLE LETTER I WITH DOT ABOVE)
when lower-case-table-name=1.
Reason: In some cases during lower/upper conversion in utf8,
the result string can be shorter the original string
(including the above letter). Old implementation of caseup_str()
and casedn_str() didn't handle the result length properly,
assuming that length cannot change.
This fix changes the result type of cs->cset->casedn_str()
and cs->cset->caseup_str() from VOID to UINT, to return
the result length, as well as put '\0' terminator on a
proper place.
Also, my_caseup_str_utf8() and my_casedn_str_utf8() were
rewritten not to use strlen() for performance purposes.
It was done with help of adding of new functions - my_utf8_uni_no_range()
and my_uni_utf8_no_range() - for null terminated strings.
a updatable view.
When there's a VIEW on a base table that have AUTO_INCREMENT column, and
this VIEW doesn't provide an access such column, after INSERT to such
VIEW LAST_INSERT_ID() did not return the value just generated.
This behaviour is intended and correct, because if the VIEW doesn't list
some columns then these columns are effectively hidden from the user,
and so any side effects of inserting default values to them.
However, there was a bug that such statement inserting into a view would
reset LAST_INSERT_ID() instead of leaving it unchanged.
This patch restores the original value of LAST_INSERT_ID() instead of
resetting it to zero.
When executing the init_connect statement, thd->net.vio is set to 0, to
forbid sending any results to the client. As a side effect we don't log
possible errors, either.
Now we write warnings to the error log if an init_connect query
fails.
When statement to be prepared contained CREATE PROCEDURE, CREATE FUNCTION
or CREATE TRIGGER statements with a syntax error in it, the preparation
would fail with syntax error message, but the memory could be corrupted.
The problem occurred because we switch memroot when parse stored
routine or trigger definitions, and on parse error we restored the
original memroot only after performing some memory operations. In more
detail:
- prepared statement would activate its own memory root to parse
the definition of the stored procedure.
- SP would reset this memory root with its own memory root to
parse SP statements
- a syntax error would happen
- prepared statement would restore the original memory root
- stored procedure would restore what it thinks was the original
memory root, but actually was the statement memory root.
That led to double free - in destruction of the statement and in
a next call to mysql_parse().
The solution is to restore memroot right after the failed parsing.
Do not consider SHOW commands slow queries, just because they don't use proper indexes.
This bug fix is not needed in 5.1, and the code changes will be null merged. However, the test cases will be propogated up to 5.1.
Note: bug#21726 does not directly apply to 4.1, as it doesn't have stored
procedures. However, 4.1 had some bugs that were fixed in 5.0 by the
patch for bug#21726, and this patch is a backport of those fixes.
Namely, in 4.1 it fixes:
- LAST_INSERT_ID(expr) didn't return value of expr (4.1 specific).
- LAST_INSERT_ID() could return the value generated by current
statement if the call happens after the generation, like in
CREATE TABLE t1 (i INT AUTO_INCREMENT PRIMARY KEY, j INT);
INSERT INTO t1 VALUES (NULL, 0), (NULL, LAST_INSERT_ID());
- Redundant binary log LAST_INSERT_ID_EVENTs could be generated.
Set a flag when a SHOW command is parsed, and check it in log_slow_statement(). SHOW commands are not counted as slow queries, even if they use table scans.
(race cond)
It was possible for one thread to interrupt a Data Definition Language
statement and thereby get messages to the binlog out of order. Consider:
Connection 1: Drop Foo x
Connection 2: Create or replace Foo x
Connection 2: Log "Create or replace Foo x"
Connection 1: Log "Drop Foo x"
Local end would have Foo x, but the replicated slaves would not.
The fix for this is to wrap all DDL and logging of a kind in the same mutex.
Since we already use mutexes for the various parts of altering the server,
this only entails moving the logging events down close to the action, inside
the mutex protection.
invocations of LAST_INSERT_ID.
Reding of LAST_INSERT_ID inside stored function wasn't noted by caller,
and no LAST_INSERT_ID_EVENT was issued for binary log.
The solution is to add THD::last_insert_id_used_bin_log, which is much
like THD::last_insert_id_used, but is reset only for upper-level
statements. This new variable is used to issue LAST_INSERT_ID_EVENT.
Non-upper-level INSERTs (the ones in the body of stored procedure,
stored function, or trigger) into a table that have AUTO_INCREMENT
column didn't affected the result of LAST_INSERT_ID() on this level.
The problem was introduced with the fix of bug 6880, which in turn was
introduced with the fix of bug 3117, where current insert_id value was
remembered on the first call to LAST_INSERT_ID() (bug 3117) and was
returned from that function until it was reset before the next
_upper-level_ statement (bug 6880).
The fix for bug#21726 brings back the behaviour of version 4.0, and
implements the following: remember insert_id value at the beginning
of the statement or expression (which at that point equals to
the first insert_id value generated by the previous statement), and
return that remembered value from LAST_INSERT_ID() or @@LAST_INSERT_ID.
Thus, the value returned by LAST_INSERT_ID() is not affected by values
generated by current statement, nor by LAST_INSERT_ID(expr) calls in
this statement.
Version 5.1 does not have this bug (it was fixed by WL 3146).
