mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-01-19 21:42:35 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions

branches/innodb+: buf_flush_insert_sorted_into_flush_list(): Restore ut_ad()

Browse source 
assertions that were removed when introducing the flush list mutex,
and add comments that explain why these assertions are valid.

Discussed with Inaam Rana.
This commit is contained in:
mmakela 2010-04-07 18:11:18 +00:00
parent 25cc5e6b20
commit 1c6ee3c95c
1 changed files with 16 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

16
buf/buf0flu.c
View file

@ -280,6 +280,22 @@ buf_flush_insert_sorted_into_flush_list(
buf_flush_list_mutex_enter();
/* The field in_LRU_list is protected by buf_pool_mutex, which
we are not holding. However, while a block is in the flush
list, it is dirty and cannot be discarded, not from the
page_hash or from the LRU list. At most, the uncompressed
page frame of a compressed block may be discarded or created
(copying the block->page to or from a buf_page_t that is
dynamically allocated from buf_buddy_alloc()). Because those
transitions hold block->mutex and the flush list mutex (via
buf_flush_relocate_on_flush_list()), there is no possibility
of a race condition in the assertions below. */
ut_ad(block->page.in_LRU_list);
ut_ad(block->page.in_page_hash);
/* buf_buddy_block_register() will take a block in the
BUF_BLOCK_MEMORY state, not a file page. */
ut_ad(!block->page.in_zip_hash);
ut_ad(!block->page.in_flush_list);
ut_d(block->page.in_flush_list = TRUE);
block->page.oldest_modification = lsn;