vmscan,memcg: reintroduce sc->may_swap
KOSAKI Motohiro [Tue, 21 Apr 2009 19:24:57 +0000 (12:24 -0700)]
Commit a6dc60f8975ad96d162915e07703a4439c80dcf0 ("vmscan: rename
sc.may_swap to may_unmap") removed the may_swap flag, but memcg had used
it as a flag for "we need to use swap?", as the name indicate.

And in the current implementation, memcg cannot reclaim mapped file
caches when mem+swap hits the limit.

re-introduce may_swap flag and handle it at get_scan_ratio().  This
patch doesn't influence any scan_control users other than memcg.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/vmscan.c

index 99155b7..eac9577 100644 (file)
@@ -63,6 +63,9 @@ struct scan_control {
        /* Can mapped pages be reclaimed? */
        int may_unmap;
 
+       /* Can pages be swapped as part of reclaim? */
+       int may_swap;
+
        /* This context's SWAP_CLUSTER_MAX. If freeing memory for
         * suspend, we effectively ignore SWAP_CLUSTER_MAX.
         * In this context, it doesn't matter that we scan the
@@ -1380,7 +1383,7 @@ static void get_scan_ratio(struct zone *zone, struct scan_control *sc,
        struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(zone, sc);
 
        /* If we have no swap space, do not bother scanning anon pages. */
-       if (nr_swap_pages <= 0) {
+       if (!sc->may_swap || (nr_swap_pages <= 0)) {
                percent[0] = 0;
                percent[1] = 100;
                return;
@@ -1697,6 +1700,7 @@ unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
                .may_writepage = !laptop_mode,
                .swap_cluster_max = SWAP_CLUSTER_MAX,
                .may_unmap = 1,
+               .may_swap = 1,
                .swappiness = vm_swappiness,
                .order = order,
                .mem_cgroup = NULL,
@@ -1717,6 +1721,7 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem_cont,
        struct scan_control sc = {
                .may_writepage = !laptop_mode,
                .may_unmap = 1,
+               .may_swap = !noswap,
                .swap_cluster_max = SWAP_CLUSTER_MAX,
                .swappiness = swappiness,
                .order = 0,
@@ -1726,9 +1731,6 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem_cont,
        };
        struct zonelist *zonelist;
 
-       if (noswap)
-               sc.may_unmap = 0;
-
        sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
                        (GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
        zonelist = NODE_DATA(numa_node_id())->node_zonelists;
@@ -1767,6 +1769,7 @@ static unsigned long balance_pgdat(pg_data_t *pgdat, int order)
        struct scan_control sc = {
                .gfp_mask = GFP_KERNEL,
                .may_unmap = 1,
+               .may_swap = 1,
                .swap_cluster_max = SWAP_CLUSTER_MAX,
                .swappiness = vm_swappiness,
                .order = order,
@@ -2298,6 +2301,7 @@ static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
        struct scan_control sc = {
                .may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
                .may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
+               .may_swap = 1,
                .swap_cluster_max = max_t(unsigned long, nr_pages,
                                        SWAP_CLUSTER_MAX),
                .gfp_mask = gfp_mask,