- DROP USER command didn't reload the acl_check_hosts cache causing subsequent
connect's via TCP to fail randomly.
- 4.1 version
mysql-test/r/grant2.result:
Update test results
mysql-test/t/grant2.test:
Add test for to check that connect via TCP work after "drop user"
sql/sql_acl.cc:
Reload acl_check_hosts as its memory is mapped to acl_user
Use acl_user->host.hostname when searching the acl_check_hosts list
into mysql.com:/home/dlenev/src/mysql-4.1-bg12423
sql/mysqld.cc:
Auto merged
sql/sql_acl.cc:
Auto merged
sql/sql_parse.cc:
Auto merged
mysql-test/r/grant2.result:
Manual merge.
mysql-test/t/grant2.test:
Manual merge.
multi-threaded environment".
To avoid deadlocks between several simultaneously run account management
commands (particularly between FLUSH PRIVILEGES/SET PASSWORD and GRANT
commands) we should always take table and internal locks during their
execution in the same order. In other words we should first open and lock
privilege tables and only then obtain acl_cache::lock/LOCK_grant locks.
mysql-test/r/grant2.result:
Added test for bug #12423 "Deadlock when doing FLUSH PRIVILEGES and GRANT in
multi-threaded environment".
mysql-test/t/grant2.test:
Added test for bug #12423 "Deadlock when doing FLUSH PRIVILEGES and GRANT in
multi-threaded environment".
sql/mysqld.cc:
acl_init/grant_init() are now used only at server start up so they always
allocate temporary THD object and don't need argument for passing pointer
to it.
sql/sql_acl.cc:
To avoid deadlocks between several simultaneously run account management
commands (particularly between FLUSH PRIVILEGES/SET PASSWORD and GRANT
commands) we should always take table and internal locks during their
execution in the same order. In other words we should first open and lock
privilege tables and only then obtain acl_cache::lock/LOCK_grant locks.
Changed acl_reload()/grant_reload() and change_password()/update_user_table()
in such way that they obey this principle. Now in acl_reload()/grant_reload()/
change_password() we open and lock privilege tables, then obtain internal
locks and then call acl_load()/grant_load()/update_user_table() functions to
do actual loading or updating.
sql/sql_acl.h:
acl_init/grant_init() are now used only at server start up so they always
allocate temporary THD object and don't need argument for passing pointer
to it. acl_reload()/grant_reload() now are able to report about their
success or failure through return value.
sql/sql_parse.cc:
If reload_acl_and_cache() is called from SIGHUP handler we have to
allocate temporary THD for execution of acl_reload()/grant_reload().
user to update with 'SET PASSWORD = ...'. (Bug #12302)
mysql-test/r/grant2.result:
Add new results
mysql-test/t/grant2.test:
Add new tests
sql/set_var.cc:
Pass priv_host into check_change_password().
sql/sql_acl.cc:
Add exact flag for find_acl_user, so we can specify that we want
an exact match on the hostname.
hostnames to not be matched correctly. (Bug #3309)
sql/sql_acl.cc:
Use an acl_host_or_ip in the GRANT_TABLE struct instead of storing the host
itself, and use compare_hostname() in testing it. This adds support for IP
addresses with netmasks to table-specific grants, and results in simplified
code by making use of compare_hostname() in more places.
mysql-test/r/grant2.result:
Add new results
mysql-test/t/grant2.test:
Add new regression test
mysql-test/r/grant2.result:
new test case
mysql-test/r/variables.result:
don't fail w/o innodb
mysql-test/t/grant2.test:
new test case
mysql-test/t/multi_update.test:
don't fail w/o innodb
mysql-test/t/variables.test:
don't fail w/o innodb
sql/sql_acl.cc:
cleanup
he has SELECT and INSERT privileges for table with primary key"
Now we set lex->duplicates= DUP_UPDATE right in parser if INSERT has
ON DUPLICATE KEY UPDATE clause, this simplifies insert_precheck()
function (this also fixes a bug) and some other code.
mysql-test/r/grant2.result:
Added test for bug #6173 "One can circumvent missing UPDATE privilege if
he has SELECT and INSERT privileges for table with primary key"
mysql-test/t/grant2.test:
Added test for bug #6173 "One can circumvent missing UPDATE privilege if
he has SELECT and INSERT privileges for table with primary key"
sql/mysql_priv.h:
insert_precheck() don't need "update" parameter any longer since
now we set lex->duplicates to DUP_UPDATE if INSERT has ON DUPLICATE
KEY UPDATE clause.
sql/sql_parse.cc:
insert_precheck() don't need "update" parameter any longer since
now we set lex->duplicates to DUP_UPDATE if INSERT has ON DUPLICATE
KEY UPDATE clause, so it can determine whenever it is needed to
require UPDATE_ACL by itself. Also calling of mysql_insert() is
simplified.
sql/sql_prepare.cc:
insert_precheck() don't need "update" parameter any longer since
now we set lex->duplicates to DUP_UPDATE if INSERT has ON DUPLICATE
KEY UPDATE clause, so it can determine whenever it is needed to
require UPDATE_ACL by itself. Also calling of mysql_insert() is
simplified.
sql/sql_yacc.yy:
It is better to set Lex->duplicates= DUP_UPDATE right in parser if we
have INSERT with ON DUPLICATE KEY UPDATE clause, rather doing this later.
'SHOW GRANTS' syntax is added
'SHOW GRANTS FOR CURRENT_USER' syntax is added
'SHOW GRANTS FOR CURRENT_USER()' syntax is added
CURRENT_USER without parens in expressions(SELECT CURRENT_USER;)
mysql-test/r/grant2.result:
WL1368: SHOW GRANTS FOR CURRENT USER
mysql-test/r/grant_cache.result:
WL1368: SHOW GRANTS FOR CURRENT USER
mysql-test/t/grant2.test:
WL1368: SHOW GRANTS FOR CURRENT USER
mysql-test/t/grant_cache.test:
WL1368: SHOW GRANTS FOR CURRENT USER
sql/lex.h:
WL1368: SHOW GRANTS FOR CURRENT USER
sql/sql_yacc.yy:
WL1368: SHOW GRANTS FOR CURRENT USER
