Memory controller: make charging gfp mask aware

Nick Piggin pointed out that swap cache and page cache addition routines
could be called from non GFP_KERNEL contexts.  This patch makes the
charging routine aware of the gfp context.  Charging might fail if the
cgroup is over it's limit, in which case a suitable error is returned.

This patch was tested on a Powerpc box.  I am still looking at being able
to test the path, through which allocations happen in non GFP_KERNEL
contexts.

[kamezawa.hiroyu@jp.fujitsu.com: problem with ZONE_MOVABLE]
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: 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>

7 files changed, 31 insertions, 25 deletions

diff --git a/mm/filemap.c b/mm/filemap.c
index 8ae171c..63040d5 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -464,7 +464,7 @@ int add_to_page_cache(struct page *page, struct address_space *mapping,
 
 	if (error == 0) {
 
-		error = mem_cgroup_cache_charge(page, current->mm);
+		error = mem_cgroup_cache_charge(page, current->mm, gfp_mask);
 		if (error)
 			goto out;
 
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index ff7cac6..ac87744 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -261,7 +261,8 @@ unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan,
  * 0 if the charge was successful
  * < 0 if the cgroup is over its limit
  */
-int mem_cgroup_charge(struct page *page, struct mm_struct *mm)
+int mem_cgroup_charge(struct page *page, struct mm_struct *mm,
+				gfp_t gfp_mask)
 {
 	struct mem_cgroup *mem;
 	struct page_cgroup *pc, *race_pc;
@@ -293,7 +294,7 @@ retry:
 
 	unlock_page_cgroup(page);
 
-	pc = kzalloc(sizeof(struct page_cgroup), GFP_KERNEL);
+	pc = kzalloc(sizeof(struct page_cgroup), gfp_mask);
 	if (pc == NULL)
 		goto err;
 
@@ -320,7 +321,14 @@ retry:
 	 * the cgroup limit.
 	 */
 	while (res_counter_charge(&mem->res, PAGE_SIZE)) {
-		if (try_to_free_mem_cgroup_pages(mem))
+		bool is_atomic = gfp_mask & GFP_ATOMIC;
+		/*
+		 * We cannot reclaim under GFP_ATOMIC, fail the charge
+		 */
+		if (is_atomic)
+			goto noreclaim;
+
+		if (try_to_free_mem_cgroup_pages(mem, gfp_mask))
 			continue;
 
 		/*
@@ -344,9 +352,10 @@ retry:
 			congestion_wait(WRITE, HZ/10);
 			continue;
 		}
-
+noreclaim:
 		css_put(&mem->css);
-		mem_cgroup_out_of_memory(mem, GFP_KERNEL);
+		if (!is_atomic)
+			mem_cgroup_out_of_memory(mem, GFP_KERNEL);
 		goto free_pc;
 	}
 
@@ -385,7 +394,8 @@ err:
 /*
  * See if the cached pages should be charged at all?
  */
-int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm)
+int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
+				gfp_t gfp_mask)
 {
 	struct mem_cgroup *mem;
 	if (!mm)
@@ -393,7 +403,7 @@ int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm)
 
 	mem = rcu_dereference(mm->mem_cgroup);
 	if (mem->control_type == MEM_CGROUP_TYPE_ALL)
-		return mem_cgroup_charge(page, mm);
+		return mem_cgroup_charge(page, mm, gfp_mask);
 	else
 		return 0;
 }
diff --git a/mm/memory.c b/mm/memory.c
index 0ba224e..153a54b 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1147,7 +1147,7 @@ static int insert_page(struct mm_struct *mm, unsigned long addr, struct page *pa
 	pte_t *pte;
 	spinlock_t *ptl;
 
-	retval = mem_cgroup_charge(page, mm);
+	retval = mem_cgroup_charge(page, mm, GFP_KERNEL);
 	if (retval)
 		goto out;
 
@@ -1650,7 +1650,7 @@ gotten:
 	cow_user_page(new_page, old_page, address, vma);
 	__SetPageUptodate(new_page);
 
-	if (mem_cgroup_charge(new_page, mm))
+	if (mem_cgroup_charge(new_page, mm, GFP_KERNEL))
 		goto oom_free_new;
 
 	/*
@@ -2052,7 +2052,7 @@ static int do_swap_page(struct mm_struct *mm, struct vm_area_struct *vma,
 		count_vm_event(PGMAJFAULT);
 	}
 
-	if (mem_cgroup_charge(page, mm)) {
+	if (mem_cgroup_charge(page, mm, GFP_KERNEL)) {
 		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
 		ret = VM_FAULT_OOM;
 		goto out;
@@ -2139,7 +2139,7 @@ static int do_anonymous_page(struct mm_struct *mm, struct vm_area_struct *vma,
 		goto oom;
 	__SetPageUptodate(page);
 
-	if (mem_cgroup_charge(page, mm))
+	if (mem_cgroup_charge(page, mm, GFP_KERNEL))
 		goto oom_free_page;
 
 	entry = mk_pte(page, vma->vm_page_prot);
@@ -2277,7 +2277,7 @@ static int __do_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 
 	}
 
-	if (mem_cgroup_charge(page, mm)) {
+	if (mem_cgroup_charge(page, mm, GFP_KERNEL)) {
 		ret = VM_FAULT_OOM;
 		goto out;
 	}
diff --git a/mm/migrate.c b/mm/migrate.c
index 417bbda..7637941 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -153,7 +153,7 @@ static void remove_migration_pte(struct vm_area_struct *vma,
  		return;
  	}
 
-	if (mem_cgroup_charge(new, mm)) {
+	if (mem_cgroup_charge(new, mm, GFP_KERNEL)) {
 		pte_unmap(ptep);
 		return;
 	}
diff --git a/mm/swap_state.c b/mm/swap_state.c
index 8825886..581b609 100644
--- a/mm/swap_state.c
+++ b/mm/swap_state.c
@@ -78,7 +78,7 @@ int add_to_swap_cache(struct page *page, swp_entry_t entry, gfp_t gfp_mask)
 	error = radix_tree_preload(gfp_mask);
 	if (!error) {
 
-		error = mem_cgroup_cache_charge(page, current->mm);
+		error = mem_cgroup_cache_charge(page, current->mm, gfp_mask);
 		if (error)
 			goto out;
 
diff --git a/mm/swapfile.c b/mm/swapfile.c
index fddc4cc..35e00c3 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -510,7 +510,7 @@ unsigned int count_swap_pages(int type, int free)
 static int unuse_pte(struct vm_area_struct *vma, pte_t *pte,
 		unsigned long addr, swp_entry_t entry, struct page *page)
 {
-	if (mem_cgroup_charge(page, vma->vm_mm))
+	if (mem_cgroup_charge(page, vma->vm_mm, GFP_KERNEL))
 		return -ENOMEM;
 
 	inc_mm_counter(vma->vm_mm, anon_rss);
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 215f6a7..b7d868c 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1337,16 +1337,11 @@ unsigned long try_to_free_pages(struct zone **zones, int order, gfp_t gfp_mask)
 
 #ifdef CONFIG_CGROUP_MEM_CONT
 
-#ifdef CONFIG_HIGHMEM
-#define ZONE_USERPAGES ZONE_HIGHMEM
-#else
-#define ZONE_USERPAGES ZONE_NORMAL
-#endif
-
-unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem_cont)
+unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem_cont,
+						gfp_t gfp_mask)
 {
 	struct scan_control sc = {
-		.gfp_mask = GFP_KERNEL,
+		.gfp_mask = gfp_mask,
 		.may_writepage = !laptop_mode,
 		.may_swap = 1,
 		.swap_cluster_max = SWAP_CLUSTER_MAX,
@@ -1357,9 +1352,10 @@ unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem_cont)
 	};
 	int node;
 	struct zone **zones;
+	int target_zone = gfp_zone(GFP_HIGHUSER_MOVABLE);
 
 	for_each_online_node(node) {
-		zones = NODE_DATA(node)->node_zonelists[ZONE_USERPAGES].zones;
+		zones = NODE_DATA(node)->node_zonelists[target_zone].zones;
 		if (do_try_to_free_pages(zones, sc.gfp_mask, &sc))
 			return 1;
 	}