Bug#21483 "Server abort or deadlock on INSERT DELAYED with another
implicit insert"
Also fixes and adds test cases for bugs:
20497 "Trigger with INSERT DELAYED causes Error 1165"
21714 "Wrong NEW.value and server abort on INSERT DELAYED to a
table with a trigger".
Post-review fixes.
Problem:
In MySQL INSERT DELAYED is a way to pipe all inserts into a
given table through a dedicated thread. This is necessary for
simplistic storage engines like MyISAM, which do not have internal
concurrency control or threading and thus can not
achieve efficient INSERT throughput without support from SQL layer.
DELAYED INSERT works as follows:
For every distinct table, which can accept DELAYED inserts and has
pending data to insert, a dedicated thread is created to write data
to disk. All user connection threads that attempt to
delayed-insert into this table interact with the dedicated thread in
producer/consumer fashion: all records to-be inserted are pushed
into a queue of the dedicated thread, which fetches the records and
writes them.
In this design, client connection threads never open or lock
the delayed insert table.
This functionality was introduced in version 3.23 and does not take
into account existence of triggers, views, or pre-locking.
E.g. if INSERT DELAYED is called from a stored function, which,
in turn, is called from another stored function that uses the delayed
table, a deadlock can occur, because delayed locking by-passes
pre-locking. Besides:
* the delayed thread works directly with the subject table through
the storage engine API and does not invoke triggers
* even if it was patched to invoke triggers, if triggers,
in turn, used other tables, the delayed thread would
have to open and lock involved tables (use pre-locking).
* even if it was patched to use pre-locking, without deadlock
detection the delayed thread could easily lock out user
connection threads in case when the same table is used both
in a trigger and on the right side of the insert query:
the delayed thread would not release locks until all inserts
are complete, and user connection can not complete inserts
without having locks on the tables used on the right side of the
query.
Solution:
These considerations suggest two general alternatives for the
future of INSERT DELAYED:
* it is considered a full-fledged alternative to normal INSERT
* it is regarded as an optimisation that is only relevant
for simplistic engines.
Since we missed our chance to provide complete support of new
features when 5.0 was in development, the first alternative
currently renders infeasible.
However, even the second alternative, which is to detect
new features and convert DELAYED insert into a normal insert,
is not easy to implement.
The catch-22 is that we don't know if the subject table has triggers
or is a view before we open it, and we only open it in the
delayed thread. We don't know if the query involves pre-locking
until we have opened all tables, and we always first create
the delayed thread, and only then open the remaining tables.
This patch detects the problematic scenarios and converts
DELAYED INSERT to a normal INSERT using the following approach:
* if the statement is executed under pre-locking (e.g. from
within a stored function or trigger) or the right
side may require pre-locking, we detect the situation
before creating a delayed insert thread and convert the statement
to a conventional INSERT.
* if the subject table is a view or has triggers, we shutdown
the delayed thread and convert the statement to a conventional
INSERT.
When inserting into a join-based view the update fields from the ON DUPLICATE
KEY UPDATE wasn't checked to be from the table being inserted into and were
silently ignored.
The new check_view_single_update() function is added to check that
insert/update fields are being from the same single table of the view.
columns
Fixed confusing warning.
Quoting INSERT section of the manual:
----
Inserting NULL into a column that has been declared NOT NULL. For
multiple-row INSERT statements or INSERT INTO ... SELECT statements, the
column is set to the implicit default value for the column data type. This
is 0 for numeric types, the empty string ('') for string types, and the
"zero" value for date and time types. INSERT INTO ... SELECT statements are
handled the same way as multiple-row inserts because the server does not
examine the result set from the SELECT to see whether it returns a single
row. (For a single-row INSERT, no warning occurs when NULL is inserted into
a NOT NULL column. Instead, the statement fails with an error.)
----
This is also true for LOAD DATA INFILE. For INSERT user can specify
DEFAULT keyword as a value to set column default. There is no similiar
feature available for LOAD DATA INFILE.
INSERT triggers".
In cases when REPLACE was internally executed via update and table had
on update (on delete) triggers defined we exposed the fact that such
optimization used by callng on update (not calling on delete) triggers.
Such behavior contradicts our documentation which describes REPLACE as
INSERT with optional DELETE.
This fix just disables this optimization for tables with on delete triggers.
The optimization is still applied for tables which have on update but have
no on delete triggers, we just don't invoke on update triggers in this case
and thus don't expose information about optimization to user.
Also added test coverage for values returned by ROW_COUNT() function (and
thus for values returned by mysql_affected_rows()) for various forms of
INSERT.
The bug was caused by wrong behaviour of mysql_insert() which in case
of INSERT DELAYED into a view exited with thd->net.report_error == 0.
This blocked error reporting to the client which started waiting
infinitely for response to the query.
FOUND is not a reserved keyword anymore
Added Item_field::set_no_const_sub() to be able to mark fields that can't be substituted
Added 'simple_select' method to be able to quickly determinate if a select_result is a normal SELECT
Note that the 5.0 tree is not yet up to date: Sanja will have to fix multi-update-locks for this merge to be complete
More tests.
Better error messages.
Fixed bug when checking if we updated all needed columns for INSERT.
Give an error if we encounter a wrong float value during parsing.
Don't print DEFAULT for columns without a default value in SHOW CREATE/SHOW FIELDS.
Fixed UPDATE IGNORE when using STRICT mode.
Under strict mode MySQL will generate an error message if there was any conversion when assigning data to a field.
Added checking of date/datetime fields.
If strict mode, give error if we have not given value to field without a default value (for INSERT)
sigsegv protection (exp overflow)
don't return inf!
use errno=EOVERFLOW to signal an overflow (as my_strntod uses errno anyway)
if errno will be too slow, my_strtod can be changed to return overflow status in a parameter (like my_strntod does)
- get my_strtod to return inf
- get Field_float::store(double) and Field_double::store(float) to set null for
nan value
(as extra serg's recomendations to fix for patch on
Bug #2082 'mysqldump converts "inf" to null')
Added more DBUG statements
Ensure that we are comparing end space with BINARY strings
Use 'any_db' instead of '' to mean any database. (For HANDLER command)
Only strip ' ' when comparing CHAR, not other space-like characters (like \t)
Cleaned up embedded library access and query cache handling
Changed min stack size to 128K (to allow longer MyISAM keys)
Fixed wrong priority for XOR (should be less than NEG to get -1^1 to work)
Don't dump data for MRG_ISAM or MRG_MYISAM tables. (Bug #1846)
Ensure that 'lower_case_table_names' is always set on case insensitive file systems. (Bug #1812)
One can now configure MySQL as windows service as a normal user. (Bug #1802)
Database names is now compared with lower case in ON clause when lower_case_table_names is set. (Bug #1736)
IGNORE ... LINES option didn't work when used with fixed length rows. (Bug #1704)
Change INSERT DELAYED ... SELECT... to INSERT .... SELECT (Bug #1983)
Safety fix for service 'mysql start' (Bug #1815)
Add quoting for use `database` for mysqlbinlog
Removed test ins0000001
Add support for --replace for exec in mysqltest
Don't refer to install dir in mysqlbinlog.result
Fixed bugs in my last changeset that made MySQL hard to compile.
Added mutex around some data that could cause table cache corruptions when using OPTIMIZE TABLE / REPAIR TABLE or automatic repair of MyISAM tables.
Added mutex around some data in the slave start/stop code that could cause THD linked list corruptions
Extended my_chsize() to allow one to specify a filler character.
Extend vio_blocking to return the old state (This made some usage of this function much simpler)
Added testing for some functions that they caller have got the required mutexes before calling the function.
Use setrlimit() to ensure that we can write core file if one specifies --core-file.
Added --slave-compressed-protocol
Made 2 the minimum length for ft_min_word_len
Added variables foreign_key_checks & unique_checks.
Less logging from replication code (if not started with --log-warnings)
Changed that SHOW INNODB STATUS requre the SUPER privilege
More DBUG statements and a lot of new code comments
added support for quiet
increased line buffer size
client/mysqltest.c
fixed memory leak
added query logging to result file
added error message logging to result file
added enable_query_log/disable_query_log
mysql-test/mysql-test-run.sh
converted tests to use mysqlmanager
Updated test results
