The test is unstable because 'UPDATE t SET b = 100' latches a page and
waits for 'upd_cont' signal in lock_trx_handle_wait_enter sync point, then
purge requests RW_X_LATCH on the same page, and then 'SELECT * FROM t
WHERE a = 10 FOR UPDATE' requests RW_S_LATCH, waiting for RW_X_LATCH
requested by purge. 'UPDATE t SET b = 100' can't release page latch as
it waits for upd_cont signal, which must be emitted after 'SELECT * FROM
t WHERE a = 10 FOR UPDATE' acquired RW_S_LATCH. So we have a deadlock,
which is resolved by finishing the debug sync point wait by timeout, and
the 'UPDATE t SET b = 100' releases it's record locks rolling back the
transaction, and 'SELECT * FROM t WHERE a = 10 FOR UPDATE' is finished
successfully instead of finishing by lock wait timeout.
The fix is to forbid purging during the test by opening read view in a
separate connection before the first insert into the table.
Besides, 'lock_wait_end' syncpoint is not needed, as it enough to wait
the end of the SELECT execution to let the UPDATE to continue.
* MDEV-29426 Fix memory leak in CONNECT JSON/BSON
If information is being gathered on JSON/BSON tables we leak the heap
allocated to the table class. With this fix we close and cleanup
instead, just as we do for XML tables.
The reason why mysql/mysql-server@8020cfac20
split the files was some unit tests that never existed in the
MariaDB Server code base. The storage/innobase/unittest/ works just fine
with this file.
This is reverting part of 2e814d4702
which applied InnoDB changes from MySQL 5.7.9.
The futex system calls were introduced in Linux 2.6.0,
which was released in December 2003. It should be safe to assume
that the system calls are always available on the Linux kernels
that MariaDB Server 10.3 would run on.
Let us use the normal platform-specific preprocessor symbols
__linux__, __sun__, _AIX instead of some homebrew ones.
The preprocessor symbol UNIV_HPUX must have lost its meaning
by f6deb00a56 (note: the symbol
UNIV_HPUX10 is being checked for, but only UNIV_HPUX is defined).
log_phys_t::apply(): When parsing an INSERT_HEAP_DYNAMIC record,
allow ll==rlen to hold for the last part. A secondary index record
may inherit all preceding bytes from the infimum pseudo-record.
For INSERT_HEAP_REDUNDANT, some header bytes will always be present
because the header will never be copied from the page infimum.
We will tolerate ll==rlen also in that case to be consistent with
the parsing of INSERT_HEAP_DYNAMIC.
To make it possible to update the HELP text of UPDATE without
changing this test every time, the test is changed to look
for a faked help topic. A non-existant help topic generates
different metadata and wasn't suitable as a test.
Requested by Ian Gilfillan.
Reviewed by Dmitry Shulga
Recording test results according to MDEV-29446 changes:
mysql-test/suite/galera/r/galera_rsu_wsrep_desync.result
mysql-test/suite/galera/r/galera_sync_wait_show.result
mysql-test/suite/galera/r/lp1376747-4.result
mysql-test/suite/s3/replication_partition.result
mysql-test/suite/s3/replication_stmt.result
this test loads sql_errlog plugin. then in a second connection
it triggers an error, this locks the plugin in that thd.
then the plugin is uninstalled in the default connection.
but that doesn't unload the plugin, as it's still locked. it'll
auto-unload after the foo connection is closed. without an explicit
disconnect it is closed after mysqltest exits and the post-test check
might still see sql_errlog not fully unoaded.
check_audit_mask(mysql_global_audit_mask, event_class_mask) is tested in
mysql_audit_general_log() and then assert in mysql_audit_acquire_plugins()
verifies that the condition still holds.
But this code path is not protected by LOCK_audit_mask, so
mysql_global_audit_mask can change its value between the if() and the
assert. That is, the assert is invalid and will fire if the
audit plugin is unloaded concurrently with mysql_audit_general_log().
Nothing bad will happen in this case though, we'll just do a useless
loop over all remaining installed audit plugins.
That is, the fix is simply to remove the assert.
- Commit c8948b0d0d introduced `get_one_variable()` - updating missing argument.
- Remove caller setting of empty string in `rpl_filter`, since underlying functions will do the same
(commit 9584cbe7fc introduced).
Reviewed by: <brandon.nesterenko@mariadb.com>
The issue manifests due to a bug in mysql_routine_grant. This was a side
effect of e46eea8660 which fixed the problem of not giving appropriate error
message (ER_NONEXISTING_PROC_GRANT) when a routine grant existed due to role
inheritance.
When granting a routine privilege, it is possible to have a GRANT_NAME
entry already created from an inherited role, but with it's init_privs
set to 0.
In this case we must not create a *new* grant entry, but we must edit
this grant entry to set its init_privs.
Note that this case was already covered by MDEV-29458, however due to a
forgotten "flush privileges;" the actual code path never got hit.
Remove the flush privilege command as it was never intended to be there
in the first place.
There was an issue in updating in-memory role datastructures when
propagating role grants.
The issue is that changing a particular role's privilege (on any
privilege level, global, database, etc.)
was done such that it overwrote the entire set of bits for that
particular level of privileges.
For example:
grant select on *.* to r1 -> sets the access bits to r1 to select,
regardless of what bits were present for role r1 (inherited from any
other roles).
Before this fix, the rights of role r1 were propagated to any roles r1
was granted to, however the propagated rights did *not* include the
complete rights r1 inherited from its own grants.
For example:
grant r2 to r1;
grant select on *.* to r2;
grant insert on *.* to r1; # This command completely disregards the
# select privilege from r2.
In order to correct this, ensure that before rights are propagated
onwards, that the current's role rights have been updated from its
grants.
Additionally, the patch exposed a flaw in the DROP ROLE code.
When deleting a role we removed all its previous grants, but what
remained was the actual links of roles granted to the dropped role.
Having these links present when propagating grants meant that we would
have leftover ACL_xxx entries.
Ensure that the links are removed before propagating grants.
There was a bug in the ACL internal data structures GRANT_TABLE and
GRANT_COLUMN. The semantics are: GRANT_TABLE::init_cols and
GRANT_COLUMN::init_privs represent the bits that correspond to the
privilege bits stored in the physical tables. The other struct members
GRANT_TABLE::cols and GRANT_COLUMN::privs represent the actual access
bits, as they may be modified through role grants.
The error in logic was mixing the two fields and thus we ended up
storing the logical access bits in the physical tables, instead of the
physical (init_xxx) bits.
This caused subsequent DBUG_ASSERT failures when dropping the involved
roles.
xb_read_delta_metadata(): For ROW_FORMAT=COMPRESSED tables, initialize
the info.zip_size with the physical page size and let info.page_size
remain the logical page size, like xb_delta_open_matching_space()
expects it to be ever since
commit 0a1c3477bf (MDEV-18493).
Recording test results according to MDEV-29446 changes:
mysql-test/suite/galera/r/MDEV-25494.result
mysql-test/suite/galera/r/galera_ctas.result
mysql-test/suite/galera/r/galera_schema.result
mysql-test/suite/galera_3nodes/r/galera_wsrep_schema.result
mysql-test/suite/galera_sr/r/galera_sr_create_drop.result
btr_lift_page_up(): If the leaf page only contains a hidden metadata
record for MDEV-11369 instant ADD COLUMN, convert the table to the
canonical format like we are supposed to do whenever the table
becomes empty.
lock_place_prdt_page_lock(): Do not place locks on temporary tables.
Temporary tables can only be accessed from one connection, so
it does not make any sense to acquire any transactional locks on them.
The shutdown time assert was caused by untimely deactivation of
the binlog background thread and related structs destruction.
It could specifically occur when a transaction is replication unsafe
and has to be completed with a ROLLBACK event in binlog.
This gets fixed with the binlog background thread stop relocation
to a point and user transactions have been completed.
A test case is added to binlog.binlog_checkpoint which
also receives as a bonus a minor correction to reactivate a MDEV-4322 test
case that originally required a shutdown phase (that ceased to do).
In commit 8f8ba75855 (MDEV-27234)
the data dictionary recovery was changed to use READ COMMITTED
so that table-rebuild operations (OPTIMIZE TABLE, TRUNCATE TABLE,
some forms of ALTER TABLE) would be recovered correctly.
However, for operations that avoid a table rebuild thanks to
being able to instantly ADD, DROP or reorder columns, recovery
must use the READ UNCOMMITTED isolation level so that changes to
the hidden metadata record can be rolled back.
We will detect instant operations by detecting uncommitted changes
to SYS_COLUMNS in case there is no uncommitted change of SYS_TABLES.ID
for the table. In any table-rebuilding DDL operation, the SYS_TABLES.ID
(and likely also the table name) will be updated.
As part of rolling back the instant ALTER TABLE operation, after the
operation on the hidden metadata record has been rolled back, a rollback
of an INSERT into SYS_COLUMNS in row_undo_ins_remove_clust_rec() will
invoke trx_t::evict_table() to discard the READ UNCOMMITTED definition
of the table. After that, subsequent recovery steps will load and use
the correct table definition.
Reviewed by: Thirunarayanan Balathandayuthapani
Tested by: Matthias Leich
Use suspend thread syncpoint instead of include/wait_condition.inc to
make sure DELETE created waiting lock before the next UPDATE begins
locking.
This is backport of commit 0fa4dd0747
from 10.6.
