Given my concerns around exactly what is going on in the Windows code,
I'm now afraid to mess with an all-platforms change to fdatasync as the
preferred default; if we do that it should probably just be in HEAD not
the back branches. So I've come around to the idea that Marti's
proposal of a PLATFORM_DEFAULT_SYNC_METHOD symbol is the best way.
(One reason for adopting that rather than some other way is that it
seems quite likely we'll end up needing it for Windows.)
I haven't touched the documentation yet, but attached is a proposed
code patch against HEAD. This forces the default to fdatasync on Linux,
and makes some cosmetic cleanups around the HAVE_FSYNC_WRITETHROUGH_ONLY
confusion.
regards, tom lane
diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c
index fd5ec78..4f7dc39 100644
*** a/src/backend/storage/file/fd.c
--- b/src/backend/storage/file/fd.c
*************** static bool looks_like_temp_rel_name(con
*** 260,271 ****
int
pg_fsync(int fd)
{
! #ifndef HAVE_FSYNC_WRITETHROUGH_ONLY
! if (sync_method != SYNC_METHOD_FSYNC_WRITETHROUGH)
! return pg_fsync_no_writethrough(fd);
else
#endif
! return pg_fsync_writethrough(fd);
}
--- 260,272 ----
int
pg_fsync(int fd)
{
! /* #if is to skip the sync_method test if there's no need for it */
! #if defined(HAVE_FSYNC_WRITETHROUGH) && !defined(FSYNC_WRITETHROUGH_IS_FSYNC)
! if (sync_method == SYNC_METHOD_FSYNC_WRITETHROUGH)
! return pg_fsync_writethrough(fd);
else
#endif
! return pg_fsync_no_writethrough(fd);
}
diff --git a/src/include/access/xlogdefs.h b/src/include/access/xlogdefs.h
index 18b214e..072096d 100644
*** a/src/include/access/xlogdefs.h
--- b/src/include/access/xlogdefs.h
*************** typedef uint32 TimeLineID;
*** 123,134 ****
#endif
#endif
! #if defined(OPEN_DATASYNC_FLAG)
#define DEFAULT_SYNC_METHOD SYNC_METHOD_OPEN_DSYNC
#elif defined(HAVE_FDATASYNC)
#define DEFAULT_SYNC_METHOD SYNC_METHOD_FDATASYNC
- #elif defined(HAVE_FSYNC_WRITETHROUGH_ONLY)
- #define DEFAULT_SYNC_METHOD SYNC_METHOD_FSYNC_WRITETHROUGH
#else
#define DEFAULT_SYNC_METHOD SYNC_METHOD_FSYNC
#endif
--- 123,134 ----
#endif
#endif
! #if defined(PLATFORM_DEFAULT_SYNC_METHOD)
! #define DEFAULT_SYNC_METHOD PLATFORM_DEFAULT_SYNC_METHOD
! #elif defined(OPEN_DATASYNC_FLAG)
#define DEFAULT_SYNC_METHOD SYNC_METHOD_OPEN_DSYNC
#elif defined(HAVE_FDATASYNC)
#define DEFAULT_SYNC_METHOD SYNC_METHOD_FDATASYNC
#else
#define DEFAULT_SYNC_METHOD SYNC_METHOD_FSYNC
#endif
diff --git a/src/include/port/linux.h b/src/include/port/linux.h
index b9498b2..bcaa42d 100644
*** a/src/include/port/linux.h
--- b/src/include/port/linux.h
***************
*** 12,14 ****
--- 12,22 ----
* to have a kernel version test here.
*/
#define HAVE_LINUX_EIDRM_BUG
+
+ /*
+ * Set the default wal_sync_method to fdatasync. With recent Linux versions,
+ * xlogdefs.h's normal rules will prefer open_datasync, which (a) doesn't
+ * perform better and (b) causes outright failures on ext4 data=journal
+ * filesystems, because those don't support O_DIRECT.
+ */
+ #define PLATFORM_DEFAULT_SYNC_METHOD SYNC_METHOD_FDATASYNC
diff --git a/src/include/port/win32.h b/src/include/port/win32.h
index 3417ab5..9c2ae4d 100644
*** a/src/include/port/win32.h
--- b/src/include/port/win32.h
***************
*** 34,47 ****
/* Must be here to avoid conflicting with prototype in windows.h */
#define mkdir(a,b) mkdir(a)
- #define HAVE_FSYNC_WRITETHROUGH
- #define HAVE_FSYNC_WRITETHROUGH_ONLY
#define ftruncate(a,b) chsize(a,b)
/*
! * Even though we don't support 'fsync' as a wal_sync_method,
! * we do fsync() a few other places where _commit() is just fine.
*/
! #define fsync(fd) _commit(fd)
#define USES_WINSOCK
--- 34,51 ----
/* Must be here to avoid conflicting with prototype in windows.h */
#define mkdir(a,b) mkdir(a)
#define ftruncate(a,b) chsize(a,b)
+
+ /* Windows doesn't have fsync() as such, use _commit() */
+ #define fsync(fd) _commit(fd)
+
/*
! * For historical reasons, we allow setting wal_sync_method to
! * fsync_writethrough on Windows, even though it's really identical to fsync
! * (both code paths wind up at _commit()).
*/
! #define HAVE_FSYNC_WRITETHROUGH
! #define FSYNC_WRITETHROUGH_IS_FSYNC
#define USES_WINSOCK
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
