Get rid of WALBufMappingLock
authorAlexander Korotkov <akorotkov@postgresql.org>
Mon, 17 Feb 2025 02:19:01 +0000 (04:19 +0200)
committerAlexander Korotkov <akorotkov@postgresql.org>
Mon, 17 Feb 2025 02:25:29 +0000 (04:25 +0200)
commit6a2275b8953a4462d44daf001bdd60b3d48f0946
treeaf1ae24663b9ebe674e77e981b48ea101ee4e596
parentfbc0fe9a2efeb6172f93ad74a9656fdac5de2507
Get rid of WALBufMappingLock

Allow multiple backends to initialize WAL buffers concurrently.  This way
`MemSet((char *) NewPage, 0, XLOG_BLCKSZ);` can run in parallel without
taking a single LWLock in exclusive mode.

The new algorithm works as follows:
 * reserve a page for initialization using XLogCtl->InitializeReserved,
 * ensure the page is written out,
 * once the page is initialized, try to advance XLogCtl->InitializedUpTo and
   signal to waiters using XLogCtl->InitializedUpToCondVar condition
   variable,
 * repeat previous steps until we reserve initialization up to the target
   WAL position,
 * wait until concurrent initialization finishes using a
   XLogCtl->InitializedUpToCondVar.

Now, multiple backends can, in parallel, concurrently reserve pages,
initialize them, and advance XLogCtl->InitializedUpTo to point to the latest
initialized page.

Author: Yura Sokolov <y.sokolov@postgrespro.ru>
Co-authored-by: Alexander Korotkov <aekorotkov@gmail.com>
Reviewed-by: Pavel Borisov <pashkin.elfe@gmail.com>
src/backend/access/transam/xlog.c
src/backend/utils/activity/wait_event_names.txt
src/include/storage/lwlocklist.h