[PATCH 1/2] shmem: fix init_page_accessed use to stop !PageLRU bug

[PATCH 1/2] shmem: fix init_page_accessed use to stop !PageLRU bug

[Hugh Dickins]

Under shmem swapping load, I sometimes hit the VM_BUG_ON_PAGE(!PageLRU)
in isolate_lru_pages() at mm/vmscan.c:1281!

Commit 2457aec63745 ("mm: non-atomically mark page accessed during page
cache allocation where possible") looks like interrupted work-in-progress.

mm/filemap.c's call to init_page_accessed() is fine, but not mm/shmem.c's
- shmem_write_begin() is clearly wrong to use it after shmem_getpage(),
when the page is always visible in radix_tree, and often already on LRU.

Revert change to shmem_write_begin(), and use init_page_accessed() or
mark_page_accessed() appropriately for SGP_WRITE in shmem_getpage_gfp().

SGP_WRITE also covers shmem_symlink(), which did not mark_page_accessed()
before; but since many other filesystems use [__]page_symlink(), which did
and does mark the page accessed, consider this as rectifying an oversight.

Signed-off-by: Hugh Dickins <hughd@google.com>
---

 mm/shmem.c |   15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

--- 3.16-rc3/mm/shmem.c	2014-06-29 15:22:10.592003936 -0700
+++ linux/mm/shmem.c	2014-06-30 12:15:52.204093217 -0700
@@ -1029,6 +1029,9 @@ repeat:
 		goto failed;
 	}
 
+	if (page && sgp == SGP_WRITE)
+		mark_page_accessed(page);
+
 	/* fallocated page? */
 	if (page && !PageUptodate(page)) {
 		if (sgp != SGP_READ)
@@ -1110,6 +1113,9 @@ repeat:
 		shmem_recalc_inode(inode);
 		spin_unlock(&info->lock);
 
+		if (sgp == SGP_WRITE)
+			mark_page_accessed(page);
+
 		delete_from_swap_cache(page);
 		set_page_dirty(page);
 		swap_free(swap);
@@ -1136,6 +1142,9 @@ repeat:
 
 		__SetPageSwapBacked(page);
 		__set_page_locked(page);
+		if (sgp == SGP_WRITE)
+			init_page_accessed(page);
+
 		error = mem_cgroup_charge_file(page, current->mm,
 						gfp & GFP_RECLAIM_MASK);
 		if (error)
@@ -1412,13 +1421,9 @@ shmem_write_begin(struct file *file, str
 			loff_t pos, unsigned len, unsigned flags,
 			struct page **pagep, void **fsdata)
 {
-	int ret;
 	struct inode *inode = mapping->host;
 	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
-	ret = shmem_getpage(inode, index, pagep, SGP_WRITE, NULL);
-	if (ret == 0 && *pagep)
-		init_page_accessed(*pagep);
-	return ret;
+	return shmem_getpage(inode, index, pagep, SGP_WRITE, NULL);
 }
 
 static int

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 2/2] mm: replace init_page_accessed by __SetPageReferenced

[Hugh Dickins]

Do we really need an exported alias for __SetPageReferenced()?
Its callers better know what they're doing, in which case the page
would not be already marked referenced.  Kill init_page_accessed(),
just __SetPageReferenced() inline.

Signed-off-by: Hugh Dickins <hughd@google.com>
---

 include/linux/swap.h |    1 -
 mm/filemap.c         |    4 ++--
 mm/shmem.c           |    2 +-
 mm/swap.c            |   14 +++-----------
 4 files changed, 6 insertions(+), 15 deletions(-)

--- 3.16-rc3+/include/linux/swap.h	2014-06-16 00:28:54.916076526 -0700
+++ linux/include/linux/swap.h	2014-06-30 12:55:35.216149853 -0700
@@ -311,7 +311,6 @@ extern void lru_add_page_tail(struct pag
 			 struct lruvec *lruvec, struct list_head *head);
 extern void activate_page(struct page *);
 extern void mark_page_accessed(struct page *);
-extern void init_page_accessed(struct page *page);
 extern void lru_add_drain(void);
 extern void lru_add_drain_cpu(int cpu);
 extern void lru_add_drain_all(void);
--- 3.16-rc3+/mm/filemap.c	2014-06-16 00:28:55.100076530 -0700
+++ linux/mm/filemap.c	2014-06-30 12:55:35.216149853 -0700
@@ -1100,9 +1100,9 @@ no_page:
 		if (WARN_ON_ONCE(!(fgp_flags & FGP_LOCK)))
 			fgp_flags |= FGP_LOCK;
 
-		/* Init accessed so avoit atomic mark_page_accessed later */
+		/* Init accessed so avoid atomic mark_page_accessed later */
 		if (fgp_flags & FGP_ACCESSED)
-			init_page_accessed(page);
+			__SetPageReferenced(page);
 
 		err = add_to_page_cache_lru(page, mapping, offset, radix_gfp_mask);
 		if (unlikely(err)) {
--- 3.16-rc3+/mm/shmem.c	2014-06-30 12:15:52.204093217 -0700
+++ linux/mm/shmem.c	2014-06-30 12:55:35.216149853 -0700
@@ -1143,7 +1143,7 @@ repeat:
 		__SetPageSwapBacked(page);
 		__set_page_locked(page);
 		if (sgp == SGP_WRITE)
-			init_page_accessed(page);
+			__SetPageReferenced(page);
 
 		error = mem_cgroup_charge_file(page, current->mm,
 						gfp & GFP_RECLAIM_MASK);
--- 3.16-rc3+/mm/swap.c	2014-06-16 00:28:55.132076531 -0700
+++ linux/mm/swap.c	2014-06-30 12:55:35.216149853 -0700
@@ -589,6 +589,9 @@ static void __lru_cache_activate_page(st
  * inactive,unreferenced	->	inactive,referenced
  * inactive,referenced		->	active,unreferenced
  * active,unreferenced		->	active,referenced
+ *
+ * When a newly allocated page is not yet visible, so safe for non-atomic ops,
+ * __SetPageReferenced(page) may be substituted for mark_page_accessed(page).
  */
 void mark_page_accessed(struct page *page)
 {
@@ -614,17 +617,6 @@ void mark_page_accessed(struct page *pag
 }
 EXPORT_SYMBOL(mark_page_accessed);
 
-/*
- * Used to mark_page_accessed(page) that is not visible yet and when it is
- * still safe to use non-atomic ops
- */
-void init_page_accessed(struct page *page)
-{
-	if (!PageReferenced(page))
-		__SetPageReferenced(page);
-}
-EXPORT_SYMBOL(init_page_accessed);
-
 static void __lru_cache_add(struct page *page)
 {
 	struct pagevec *pvec = &get_cpu_var(lru_add_pvec);

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 2/2] mm: replace init_page_accessed by __SetPageReferenced

[Mel Gorman]

On Mon, Jun 30, 2014 at 02:09:49PM -0700, Hugh Dickins wrote:
> Do we really need an exported alias for __SetPageReferenced()?
> Its callers better know what they're doing, in which case the page
> would not be already marked referenced.  Kill init_page_accessed(),
> just __SetPageReferenced() inline.
> 
> Signed-off-by: Hugh Dickins <hughd@google.com>

Ok, fair enough. The context it was written in was that callers should not
need to know the internals of what mark_page_accessed does. Initially I
thought there might be filesystem users that really should not know the
internals but that is not necessary obviously. I still feel that
init_page_accessed shows the intent more clearly and you're certainly
right that the checking PageReferenced is redundant. I don't object to
the patch but I don't think it's obviously better either other than it
avoids the temptation of anyone using __SetPageReferenced incorrectly.

Acked-by: Mel Gorman <mgorman@suse.de>

-- 
Mel Gorman
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 1/2] shmem: fix init_page_accessed use to stop !PageLRU bug

[Mel Gorman]

On Mon, Jun 30, 2014 at 02:08:11PM -0700, Hugh Dickins wrote:
> Under shmem swapping load, I sometimes hit the VM_BUG_ON_PAGE(!PageLRU)
> in isolate_lru_pages() at mm/vmscan.c:1281!
> 
> Commit 2457aec63745 ("mm: non-atomically mark page accessed during page
> cache allocation where possible") looks like interrupted work-in-progress.
> 
> mm/filemap.c's call to init_page_accessed() is fine, but not mm/shmem.c's
> - shmem_write_begin() is clearly wrong to use it after shmem_getpage(),
> when the page is always visible in radix_tree, and often already on LRU.
> 
> Revert change to shmem_write_begin(), and use init_page_accessed() or
> mark_page_accessed() appropriately for SGP_WRITE in shmem_getpage_gfp().
> 
> SGP_WRITE also covers shmem_symlink(), which did not mark_page_accessed()
> before; but since many other filesystems use [__]page_symlink(), which did
> and does mark the page accessed, consider this as rectifying an oversight.
> 
> Signed-off-by: Hugh Dickins <hughd@google.com>

Acked-by: Mel Gorman <mgorman@suse.de>

-- 
Mel Gorman
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>