mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2026-05-16 20:07:13 +02:00
Code Issues Projects Releases Packages Wiki Activity

Fix Bug #54538 - use of exclusive innodb dictionary lock limits performance.

Browse source 
This patch doesn't get rid of the need to acquire the dict_sys->mutex but
reduces the need to keep the mutex locked for the duration of the query
to fsp_get_available_space_in_free_extents() from ha_innobase::info().

rb://390.
This commit is contained in:
Sunny Bains 2010-08-20 17:49:43 +10:00
commit 54fd56c5c3
5 changed files with 160 additions and 33 deletions
Download patch file Download diff file Expand all files Collapse all files

49
storage/innobase/fsp/fsp0fsp.c
View file

@ -2842,12 +2842,61 @@ fsp_get_available_space_in_free_extents(
ut_ad(!mutex_own(&kernel_mutex));
/* The convoluted mutex acquire is to overcome latching order
issues: The problem is that the fil_mutex is at a lower level
than the tablespace latch and the buffer pool mutex. We have to
first prevent any operations on the file system by acquiring the
dictionary mutex. Then acquire the tablespace latch to obey the
latching order and then release the dictionary mutex. That way we
ensure that the tablespace instance can't be freed while we are
examining its contents (see fil_space_free()).
However, there is one further complication, we release the fil_mutex
when we need to invalidate the the pages in the buffer pool and we
reacquire the fil_mutex when deleting and freeing the tablespace
instance in fil0fil.c. Here we need to account for that situation
too. */
dict_mutex_enter_for_mysql();
/* At this stage there is no guarantee that the tablespace even
exists in the cache. */
if (fil_tablespace_deleted_or_being_deleted_in_mem(space, -1)) {
dict_mutex_exit_for_mysql();
return(ULLINT_UNDEFINED);
}
mtr_start(&mtr);
latch = fil_space_get_latch(space);
/* This should ensure that the tablespace instance can't be freed
by another thread. However, the tablespace pages can still be freed
from the buffer pool. We need to check for that again. */
mtr_x_lock(latch, &mtr);
dict_mutex_exit_for_mysql();
/* At this point it is possible for the tablespace to be deleted and
its pages removed from the buffer pool. We need to check for that
situation. However, the tablespace instance can't be deleted because
our latching above should ensure that. */
if (fil_tablespace_is_being_deleted(space)) {
mtr_commit(&mtr);
return(ULLINT_UNDEFINED);
}
/* From here on even if the user has dropped the tablespace, the
pages _must_ still exist in the buffer pool and the tablespace
instance _must be in the file system hash table. */
space_header = fsp_get_space_header(space, &mtr);
size = mtr_read_ulint(space_header + FSP_SIZE, MLOG_4BYTES, &mtr);