[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>

4 files changed, 13 insertions, 12 deletions

diff --git a/fs/afs/file.c b/fs/afs/file.c
index 0d576987ec6..4975c9c193d 100644
--- a/fs/afs/file.c
+++ b/fs/afs/file.c
@@ -291,8 +291,8 @@ static int afs_file_releasepage(struct page *page, gfp_t gfp_flags)
                 cachefs_uncache_page(vnode->cache, page);
 #endif
 
-                pageio = (struct cachefs_page *) page->private;
-                page->private = 0;
+                pageio = (struct cachefs_page *) page_private(page);
+                set_page_private(page, 0);
                 ClearPagePrivate(page);
 
                 if (pageio)
diff --git a/fs/buffer.c b/fs/buffer.c
index b1667986442..2066e4cb700 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -96,7 +96,7 @@ static void
 __clear_page_buffers(struct page *page)
 {
         ClearPagePrivate(page);
-        page->private = 0;
+        set_page_private(page, 0);
         page_cache_release(page);
 }
 
diff --git a/fs/jfs/jfs_metapage.c b/fs/jfs/jfs_metapage.c
index 26091a5f88d..8a53981f9f2 100644
--- a/fs/jfs/jfs_metapage.c
+++ b/fs/jfs/jfs_metapage.c
@@ -86,7 +86,7 @@ struct meta_anchor {
         atomic_t io_count;
         struct metapage *mp[MPS_PER_PAGE];
 };
-#define mp_anchor(page) ((struct meta_anchor *)page->private)
+#define mp_anchor(page) ((struct meta_anchor *)page_private(page))
 
 static inline struct metapage *page_to_mp(struct page *page, uint offset)
 {
@@ -108,7 +108,7 @@ static inline int insert_metapage(struct page *page, struct metapage *mp)
                 if (!a)
                         return -ENOMEM;
                 memset(a, 0, sizeof(struct meta_anchor));
-                page->private = (unsigned long)a;
+                set_page_private(page, (unsigned long)a);
                 SetPagePrivate(page);
                 kmap(page);
         }
@@ -136,7 +136,7 @@ static inline void remove_metapage(struct page *page, struct metapage *mp)
         a->mp[index] = NULL;
         if (--a->mp_count == 0) {
                 kfree(a);
-                page->private = 0;
+                set_page_private(page, 0);
                 ClearPagePrivate(page);
                 kunmap(page);
         }
@@ -156,13 +156,13 @@ static inline void dec_io(struct page *page, void (*handler) (struct page *))
 #else
 static inline struct metapage *page_to_mp(struct page *page, uint offset)
 {
-        return PagePrivate(page) ? (struct metapage *)page->private : NULL;
+        return PagePrivate(page) ? (struct metapage *)page_private(page) : NULL;
 }
 
 static inline int insert_metapage(struct page *page, struct metapage *mp)
 {
         if (mp) {
-                page->private = (unsigned long)mp;
+                set_page_private(page, (unsigned long)mp);
                 SetPagePrivate(page);
                 kmap(page);
         }
@@ -171,7 +171,7 @@ static inline int insert_metapage(struct page *page, struct metapage *mp)
 
 static inline void remove_metapage(struct page *page, struct metapage *mp)
 {
-        page->private = 0;
+        set_page_private(page, 0);
         ClearPagePrivate(page);
         kunmap(page);
 }
diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c
index ba4767c04ad..4cd46abe843 100644
--- a/fs/xfs/linux-2.6/xfs_buf.c
+++ b/fs/xfs/linux-2.6/xfs_buf.c
@@ -181,8 +181,9 @@ set_page_region(
         size_t          offset,
         size_t          length)
 {
-        page->private |= page_region_mask(offset, length);
-        if (page->private == ~0UL)
+        set_page_private(page,
+                page_private(page) | page_region_mask(offset, length));
+        if (page_private(page) == ~0UL)
                 SetPageUptodate(page);
 }
 
@@ -194,7 +195,7 @@ test_page_region(
 {
         unsigned long   mask = page_region_mask(offset, length);
 
-        return (mask && (page->private & mask) == mask);
+        return (mask && (page_private(page) & mask) == mask);
 }
 
 /*