Files
kernel/include/linux
Hugh Dickins 4c21e2f244 [PATCH] mm: split page table lock
Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with
a many-threaded application which concurrently initializes different parts of
a large anonymous area.

This patch corrects that, by using a separate spinlock per page table page, to
guard the page table entries in that page, instead of using the mm's single
page_table_lock.  (But even then, page_table_lock is still used to guard page
table allocation, and anon_vma allocation.)

In this implementation, the spinlock is tucked inside the struct page of the
page table page: with a BUILD_BUG_ON in case it overflows - which it would in
the case of 32-bit PA-RISC with spinlock debugging enabled.

Splitting the lock is not quite for free: another cacheline access.  Ideally,
I suppose we would use split ptlock only for multi-threaded processes on
multi-cpu machines; but deciding that dynamically would have its own costs.
So for now enable it by config, at some number of cpus - since the Kconfig
language doesn't support inequalities, let preprocessor compare that with
NR_CPUS.  But I don't think it's worth being user-configurable: for good
testing of both split and unsplit configs, split now at 4 cpus, and perhaps
change that to 8 later.

There is a benefit even for singly threaded processes: kswapd can be attacking
one part of the mm while another part is busy faulting.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-29 21:40:42 -07:00
..
2005-09-13 08:22:31 -07:00
2005-10-18 23:19:47 -07:00
2005-10-17 17:03:57 -07:00
2005-10-22 14:27:05 -04:00
2005-10-28 08:16:49 -07:00
2005-09-07 16:57:21 -07:00
2005-10-04 13:22:01 -07:00
2005-10-28 08:16:47 -07:00
2005-09-10 10:06:21 -07:00
2005-09-10 10:16:27 -07:00
2005-09-05 00:06:09 -07:00
2005-09-07 16:57:21 -07:00
2005-10-29 02:23:58 -02:00
2005-09-13 08:22:32 -07:00
2005-09-09 13:57:55 -07:00
2005-09-13 08:22:32 -07:00
2005-10-28 08:16:47 -07:00
2005-10-28 10:13:16 -07:00
2005-10-28 08:16:46 -07:00
2005-09-05 09:14:13 -07:00
2005-10-28 10:13:16 -07:00
2005-10-29 19:32:20 +01:00
2005-10-28 08:16:47 -07:00
2005-08-29 16:01:32 -07:00
2005-08-29 16:01:32 -07:00
2005-09-19 15:41:28 -07:00
2005-08-29 16:01:32 -07:00
2005-09-07 16:57:48 -07:00
2005-10-03 14:13:38 -07:00
2005-10-28 08:16:47 -07:00
2005-09-17 11:50:02 -07:00
2005-10-08 15:00:57 -07:00
2005-10-28 08:16:47 -07:00
2005-10-22 14:27:05 -04:00
2005-10-28 08:16:47 -07:00
2005-09-14 08:46:21 -04:00
2005-10-29 21:40:42 -07:00
2005-10-28 08:16:51 -07:00
2005-08-29 16:01:32 -07:00
2005-08-29 16:01:32 -07:00
2005-10-28 08:16:51 -07:00
2005-09-07 16:57:24 -07:00
2005-09-13 08:22:31 -07:00
2005-10-28 08:16:47 -07:00
2005-08-30 22:48:51 -05:00
2005-09-09 13:57:54 -07:00
2005-10-28 08:16:47 -07:00
2005-09-07 16:57:18 -07:00
2005-10-29 21:40:41 -07:00
2005-10-29 21:40:39 -07:00
2005-10-28 08:16:47 -07:00
2005-10-28 08:16:47 -07:00
2005-08-29 16:01:32 -07:00
2005-09-10 10:06:21 -07:00
2005-09-10 10:06:21 -07:00
2005-10-28 08:16:49 -07:00
2005-10-28 08:16:47 -07:00
2005-09-05 00:05:45 -07:00
2005-10-28 08:16:47 -07:00
2005-09-18 00:18:32 -07:00
2005-09-10 10:06:36 -07:00
2005-09-07 16:57:34 -07:00
2005-09-09 13:57:58 -07:00
2005-10-28 08:16:46 -07:00
2005-10-28 08:16:49 -07:00
2005-10-29 21:40:35 -07:00