Re: [PATCH] mm: account pglazyfreed exactly

[Andrew Morton]

On Thu,  3 Dec 2015 21:51:04 +0900 Minchan Kim <minchan@xxxxxxxxxx> wrote:

> If anon pages are zapped by unmapping between page_mapped check
> and try_to_unmap in shrink_page_list, they could be !PG_dirty
> although thre are not MADV_FREEed pages so that VM accoutns it
> as pglazyfreed wrongly.
> 
> To fix, this patch counts the number of lazyfree ptes in
> try_to_unmap_one and try_to_unmap returns SWAP_LZFREE only if
> the count is not zero, page is !PG_dirty and SWAP_SUCCESS.

A few tiny things...

diff -puN mm/rmap.c~mm-support-madvisemadv_free-fix-2-fix mm/rmap.c
--- a/mm/rmap.c~mm-support-madvisemadv_free-fix-2-fix
+++ a/mm/rmap.c
@@ -1605,7 +1605,7 @@ int try_to_unmap(struct page *page, enum
 
 	struct rmap_walk_control rwc = {
 		.rmap_one = try_to_unmap_one,
-		.arg = (void *)&rp,
+		.arg = &rp,
 		.done = page_not_mapped,
 		.anon_lock = page_lock_anon_vma_read,
 	};
@@ -1651,7 +1651,6 @@ int try_to_unmap(struct page *page, enum
 int try_to_munlock(struct page *page)
 {
 	int ret;
-
 	struct rmap_private rp = {
 		.flags = TTU_MUNLOCK,
 		.lazyfreed = 0,
@@ -1659,7 +1658,7 @@ int try_to_munlock(struct page *page)
 
 	struct rmap_walk_control rwc = {
 		.rmap_one = try_to_unmap_one,
-		.arg = (void *)&rp,
+		.arg = &rp,
 		.done = page_not_mapped,
 		.anon_lock = page_lock_anon_vma_read,
 
diff -puN mm/vmscan.c~mm-support-madvisemadv_free-fix-2-fix mm/vmscan.c
_

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/