Bug#31621: Windows server hanging during shutdown using named pipes
and idle connection
Problem: when idle pipe connection is forcefully closed with KILL
statement or when the server goes down, thread that is closing connection
would hang infinitely in CloseHandle(). The reason for the hang is that
named pipe operations are performed synchronously. In this mode all IOs
on pipe are serialized, that is CloseHandle() will not abort ReadFile()
in another thread, but wait for ReadFile() to complete.
The fix implements asynchrnous mode for named pipes, where operation of file
are not synchronized. Read/Write operation would fire an async IO and wait for
either IO completion or timeout.
Note, that with this patch timeouts are properly handled for named pipes.
Post-review: Win32 timeout code has been fixed for named pipes and shared
memory. We do not store pointer to NET in vio structure, only the read and
write timeouts.
Conflicts
=========
Text conflict in .bzr-mysql/default.conf
Text conflict in libmysqld/CMakeLists.txt
Text conflict in libmysqld/Makefile.am
Text conflict in mysql-test/collections/default.experimental
Text conflict in mysql-test/extra/rpl_tests/rpl_row_sp006.test
Text conflict in mysql-test/suite/binlog/r/binlog_tmp_table.result
Text conflict in mysql-test/suite/rpl/r/rpl_loaddata.result
Text conflict in mysql-test/suite/rpl/r/rpl_loaddata_fatal.result
Text conflict in mysql-test/suite/rpl/r/rpl_row_create_table.result
Text conflict in mysql-test/suite/rpl/r/rpl_row_sp006_InnoDB.result
Text conflict in mysql-test/suite/rpl/r/rpl_stm_log.result
Text conflict in mysql-test/suite/rpl_ndb/r/rpl_ndb_circular_simplex.result
Text conflict in mysql-test/suite/rpl_ndb/r/rpl_ndb_sp006.result
Text conflict in mysql-test/t/mysqlbinlog.test
Text conflict in sql/CMakeLists.txt
Text conflict in sql/Makefile.am
Text conflict in sql/log_event_old.cc
Text conflict in sql/rpl_rli.cc
Text conflict in sql/slave.cc
Text conflict in sql/sql_binlog.cc
Text conflict in sql/sql_lex.h
21 conflicts encountered.
NOTE
====
mysql-5.1-rpl-merge has been made a mirror of mysql-next-mr:
- "mysql-5.1-rpl-merge$ bzr pull ../mysql-next-mr"
This is the first cset (merge/...) committed after pulling
from mysql-next-mr.
2630.39.1, 2630.28.29, 2630.34.3, 2630.34.2, 2630.34.1, 2630.29.29,
2630.29.28, 2630.31.1, 2630.28.13, 2630.28.10, 2617.23.14 and
some other minor revisions.
This patch implements:
WL#4264 "Backup: Stabilize Service Interface" -- all the
server prerequisites except si_objects.{h,cc} themselves (they can
be just copied over, when needed).
WL#4435: Support OUT-parameters in prepared statements.
(and all issues in the initial patches for these two
tasks, that were discovered in pushbuild and during testing).
Bug#39519: mysql_stmt_close() should flush all data
associated with the statement.
After execution of a prepared statement, send OUT parameters of the invoked
stored procedure, if any, to the client.
When using the binary protocol, send the parameters in an additional result
set over the wire. When using the text protocol, assign out parameters to
the user variables from the CALL(@var1, @var2, ...) specification.
The following refactoring has been made:
- Protocol::send_fields() was renamed to Protocol::send_result_set_metadata();
- A new Protocol::send_result_set_row() was introduced to incapsulate
common functionality for sending row data.
- Signature of Protocol::prepare_for_send() was changed: this operation
does not need a list of items, the number of items is fully sufficient.
The following backward incompatible changes have been made:
- CLIENT_MULTI_RESULTS is now enabled by default in the client;
- CLIENT_PS_MULTI_RESUTLS is now enabled by default in the client.
------------------------------------------------------------
revno: 2476.657.208
committer: anozdrin/alik@station.
timestamp: Tue 2007-12-04 17:22:53 +0300
message:
Remove rpl_probe, enable_rpl_parse and disable_rpl_parse commands
from mysqltest, the corresponding funtctions have been removed from
MySQL C API
------------------------------------------------------------
revno: 2476.981.1
committer: msvensson@pilot.mysql.com
timestamp: Mon 2007-11-26 19:03:23 +0100
message:
Bug#31952 Remove undocumented mysql_rpl_* functions.
- Functions removed + variables in st_mysql and st_mysql_options
renamed to "unused".
- Code updated to work without those functions and variables
given an already connected MYSQL handle
mysql_real_connect() did not check whether the MYSQL connection
handler was already connected and connected again even if so.
Now a CR_ALREADY_CONNECTED error is returned.
Options loaded from config files were added before command line
arguments, and they were parsed together, which could interprete
the following:
option-a
option-b
as --option-a=--option-b if 'option-a' requires a value, and
caused confusing.
Because all options that requires a value are always given in
the form '--option=value', so it's an error if there is no
'=value' part for such an option read from config file.
This patch added a separator to separate the arguments from
config files and that from command line, so that they can be
handled differently. And report an error for options loaded
from config files that requires a value and is not given in the
form '--option=value'.
We cann connect() in a non-blocking mode to be able to specify a
non-standard timeout.
The problem was that we did not fetch the status from the
non-blocking connect(). We assumed that poll() would not return
a POLLIN flag if the connect failed. But on some platforms this
is not true.
After a successful poll() we do now retrieve the status value
from connect() with getsockopt(...SO_ERROR...). Now we do know
if (and how) the connect failed.
The test case for my investigation was rpl.rlp_ssl1 on an
Ubuntu 9.04 x86_64 machine. Both, IPV4 and IPV6 were active.
'localhost' resolved first for IPV6 and then for IPV4. The
connection over IPV6 was blocked. rpl.rlp_ssl1 timed out
as it did not notice the failed connect(). The first read()
failed, which was interpreted as a master crash and the
connection was tried to reestablish with the same result
until the retry limit was reached.
With the fix, the connect() problem is immediately recognized,
and the connect() is retried on the second resolution for
'localhost', which is successful.
DECIMAL and TIMESTAMP used to have NUM_FLAG, but NEWDECIMAL was forgotten.
It's correct that TIMESTAMP does not have the flag nowadays (manual will be updated, connectors
developers will be notified).
Failing to connect would release parts of the MYSQL struct.
We would then proceed to try again to connect without re-
initializing the struct.
We prevent the unwanted freeing of data we'll still need now.
with gcc 4.3.2
This patch fixes a number of GCC warnings about variables used
before initialized. A new macro UNINIT_VAR() is introduced for
use in the variable declaration, and LINT_INIT() usage will be
gradually deprecated. (A workaround is used for g++, pending a
patch for a g++ bug.)
GCC warnings for unused results (attribute warn_unused_result)
for a number of system calls (present at least in later
Ubuntus, where the usual void cast trick doesn't work) are
also fixed.
Bug#45243: crash on win in sql thread clear_tables_to_lock() -> free()
Bug#45242: crash on win in mysql_close() -> free()
Bug#45238: rpl_slave_skip, rpl_change_master failed (lost connection) for STOP SLAVE
Bug#46030: rpl_truncate_3innodb causes server crash on windows
Bug#46014: rpl_stm_reset_slave crashes the server sporadically in pb2
When killing a user session on the server, it's necessary to
interrupt (notify) the thread associated with the session that
the connection is being killed so that the thread is woken up
if waiting for I/O. On a few platforms (Mac, Windows and HP-UX)
where the SIGNAL_WITH_VIO_CLOSE flag is defined, this interruption
procedure is to asynchronously close the underlying socket of
the connection.
In order to enable this schema, each connection serving thread
registers its VIO (I/O interface) so that other threads can
access it and close the connection. But only the owner thread of
the VIO might delete it as to guarantee that other threads won't
see freed memory (the thread unregisters the VIO before deleting
it). A side note: closing the socket introduces a harmless race
that might cause a thread attempt to read from a closed socket,
but this is deemed acceptable.
The problem is that this infrastructure was meant to only be used
by server threads, but the slave I/O thread was registering the
VIO of a mysql handle (a client API structure that represents a
connection to another server instance) as a active connection of
the thread. But under some circumstances such as network failures,
the client API might destroy the VIO associated with a handle at
will, yet the VIO wouldn't be properly unregistered. This could
lead to accesses to freed data if a thread attempted to kill a
slave I/O thread whose connection was already broken.
There was a attempt to work around this by checking whether
the socket was being interrupted, but this hack didn't work as
intended due to the aforementioned race -- attempting to read
from the socket would yield a "bad file descriptor" error.
The solution is to add a hook to the client API that is called
from the client code before the VIO of a handle is deleted.
This hook allows the slave I/O thread to detach the active vio
so it does not point to freed memory.
The problem is that the C API function mysql_real_connect
only attempts to connect to the first IP address returned
for a hostname. This can be a problem if a hostname maps
to multiple IP address and the server is not bound to the
first one that is returned.
The solution is to augment mysql_real_connect so that it
attempts to connect to all IPv4 addresses that a domain
name maps to. The function goes over the list of address
until a successful connection is established.
No test case is provided as its not possible to test this
automatically with the current testing infrastructure.
Backport to MySQL 5.0/1 fix by Vladislav Vaintroub:
In Vista and later and also in when using terminal services, when
server is started from command line, client cannot connect to it
via shared memory protocol.
This is a regression introduced when Bug#24731 was fixed. The
reason is that client is trying to attach to shared memory using
global kernel object namespace (all kernel objects are prefixed
with Global\). However, server started from the command line in
Vista and later will create shared memory and events using current
session namespace. Thus, client is unable to find the server and
connection fails.
The fix for the client is to first try to find server using "local"
names (omitting Global\ prefix) and only if server is not found,
trying global namespace.
in load_defaults()
load_defaults(), my_search_option_files() and
my_print_default_files() utilized a global variable
containing a pointer to thread local memory. This could lead
to race conditions when those functions were called with high
concurrency.
Fixed by changing the interface of the said functions to avoid
the necessity for using a global variable.
Since we cannot change load_defaults() prototype for API
compatibility reasons, it was renamed my_load_defaults().
Now load_defaults() is a thread-unsafe wrapper around
a thread-safe version, my_load_defaults().
- Remove bothersome warning messages. This change focuses on the warnings
that are covered by the ignore file: support-files/compiler_warnings.supp.
- Strings are guaranteed to be max uint in length
- Remove bothersome warning messages. This change focuses on the warnings
that are covered by the ignore file: support-files/compiler_warnings.supp.
- Strings are guaranteed to be max uint in length
The problem was that the COM_STMT_SEND_LONG_DATA was sending a response
packet if the prepared statement wasn't found in the server (due to
reconnection). The commands COM_STMT_SEND_LONG_DATA and COM_STMT_CLOSE
should not send any packets, even error packets should not be sent since
they are not expected by the client API.
The solution is to clear generated during the execution of the aforementioned
commands and to skip resend of prepared statement commands. Another fix is
that if the connection breaks during the send of prepared statement command,
the command is not sent again since the prepared statement is no longer in the
server.
in some case.
ER_CON_COUNT_ERROR is defined with SQL state 08004. However, this SQL state is not always
returned.
This error can be thrown in two cases:
1. when an ordinary user (a user w/o SUPER privilege) is connecting,
and the number of active user connections is equal or greater than
max_connections.
2. when a user is connecting and the number of active user connections is
already (max_connections + 1) -- that means that no more connections will
be accepted regardless of the user credentials.
In the 1-st case, SQL state is correct.
The bug happens in the 2-nd case -- on UNIX the client gets 00000 SQL state, which is
absolutely wrong (00000 means "not error SQL state); on Windows
the client accidentally gets HY000 (which means "unknown SQL state).
The cause of the problem is that the server rejects extra connection
prior to read a packet with client capabilities. Thus, the server
does not know if the client supports SQL states or not (if the client
supports 4.1 protocol or not). So, the server supposes the worst and
does not send SQL state at all.
The difference in behavior on UNIX and Windows occurs because on Windows
CLI_MYSQL_REAL_CONNECT() invokes create_shared_memory(), which returns
an error (in default configuration, where shared memory is not configured).
Then, the client does not reset this error, so when the connection is
rejected, SQL state is HY000 (from the error from create_shared_memory()).
The bug appeared after test case for Bug#33507 -- before that, this behavior
just had not been tested.
The fix is to 1) reset the error after create_shared_memory();
2) set SQL state to 'unknown error' if it was not received from
the server.
A separate test case is not required, since the behavior is already
tested in connect.test.
Note for doc-team: the manual should be updated to say that under
some circumstances, 'Too many connections' has HY000 SQL state.
Rename client_last_error to last_error and client_last_errno to last_errno
to not break connectors which use the internal net structure for error handling.
There was no way to return an error from the client library
if no MYSQL connections was established.
So here i added variables to store that king of errors and
made functions like mysql_error(NULL) to return these.
--local-infile=0 checks can be bypassed by sending a FETCH LOCAL FILE response
Add a check for CLIENT_LOCAL_FILES before sending a local file.
Beware, that all binary distributions enable sending of local files and it's up
to the programs which use libmysql to disable it, if they don't use this functionality.
Otherwise they are not safe.
cause ROLLBACK of statement", part 1. Review fixes.
Do not send OK/EOF packets to the client until we reached the end of
the current statement.
This is a consolidation, to keep the functionality that is shared by all
SQL statements in one place in the server.
Currently this functionality includes:
- close_thread_tables()
- log_slow_statement().
After this patch and the subsequent patch for Bug#12713, it shall also include:
- ha_autocommit_or_rollback()
- net_end_statement()
- query_cache_end_of_result().
In future it may also include:
- mysql_reset_thd_for_next_command().
