As recently discussed, Tejun Heo, the cgroup maintainer, tries to
remove ->pre_destroy() and cgroup will never return -EBUSY at rmdir().
To do that, in memcg, handling case of use_hierarchy==false is a problem.
We move memcg's charges to its parent at rmdir(). If use_hierarchy==true,
it's already accounted in the parent, no problem. If use_hierarchy==false,
we cannot guarantee we can move all charges to the parent.
This patch changes the behavior to move all charges to root_mem_cgroup
if use_hierarchy=false. It seems this matches semantics of use_hierarchy==false,which means parent and child has no hierarchical relationship.
With this, we can remove -ENOMEM error check at pre_destroy(), called at rmdir.
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
---
Documentation/cgroups/memory.txt | 6 ++++--
mm/memcontrol.c | 38 ++++++++++++--------------------------
2 files changed, 16 insertions(+), 28 deletions(-)
diff --git a/Documentation/cgroups/memory.txt b/Documentation/cgroups/memory.txt
index 84d4f00..f717f6a 100644
--- a/Documentation/cgroups/memory.txt
+++ b/Documentation/cgroups/memory.txt
@@ -377,8 +377,10 @@ cgroup might have some charge associated with it, even though all
tasks have migrated away from it. (because we charge against pages, not
against tasks.)
-Such charges are freed or moved to their parent. At moving, both of RSS
-and CACHES are moved to parent.
+Such charges are freed or moved to their parent if use_hierarchy==true.
+If use_hierarchy==false, charges are moved to root memory cgroup.
+
+At moving, both of RSS and CACHES are moved to parent.
rmdir() may return -EBUSY if freeing/moving fails. See 5.1 also.
Charges recorded in swap information is not updated at removal of cgroup.
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 3215880..8246418 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -2662,15 +2662,14 @@ static int mem_cgroup_move_parent(struct page *page,
struct mem_cgroup *child,
gfp_t gfp_mask)
{
- struct cgroup *cg = child->css.cgroup;
- struct cgroup *pcg = cg->parent;
struct mem_cgroup *parent;
unsigned int nr_pages;
unsigned long uninitialized_var(flags);
+ bool need_cancel = false;
int ret;
/* Is ROOT ? */
- if (!pcg)
+ if (mem_cgroup_is_root(child))
return -EINVAL;
ret = -EBUSY;
@@ -2680,33 +2679,23 @@ static int mem_cgroup_move_parent(struct page *page,
goto put;
nr_pages = hpage_nr_pages(page);
- parent = mem_cgroup_from_cont(pcg);
-
- if (!parent->use_hierarchy) {
- ret = __mem_cgroup_try_charge(NULL, gfp_mask,
- nr_pages, &parent, false);
- if (ret)
- goto put_back;
+ parent = parent_mem_cgroup(child);
+ if (!parent) {
+ parent = root_mem_cgroup;
+ need_cancel = true;
}
if (nr_pages > 1)
flags = compound_lock_irqsave(page);
- if (!parent->use_hierarchy) {
- ret = mem_cgroup_move_account(page, nr_pages, pc,
- child, parent, true);
- if (ret)
- __mem_cgroup_cancel_charge(parent, nr_pages);
- } else {
- ret = mem_cgroup_move_account(page, nr_pages, pc,
- child, parent, false);
- if (!ret)
- __mem_cgroup_move_charge_parent(child, nr_pages);
- }
+ ret = mem_cgroup_move_account(page, nr_pages, pc, child, parent,
+ need_cancel);
+ if (!need_cancel && !ret)
+ __mem_cgroup_move_charge_parent(child, nr_pages);
if (nr_pages > 1)
compound_unlock_irqrestore(page, flags);
-put_back:
+
putback_lru_page(page);
put:
put_page(page);
@@ -3677,7 +3666,7 @@ static int mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
pc = lookup_page_cgroup(page);
ret = mem_cgroup_move_parent(page, pc, memcg, GFP_KERNEL);
- if (ret == -ENOMEM || ret == -EINTR)
+ if (ret == -EINTR)
break;
if (ret == -EBUSY || ret == -EINVAL) {
@@ -3738,9 +3727,6 @@ move_account:
}
mem_cgroup_end_move(memcg);
memcg_oom_recover(memcg);
- /* it seems parent cgroup doesn't have enough mem */
- if (ret == -ENOMEM)
- goto try_to_free;
cond_resched();
/* "ret" should also be checked to ensure all lists are empty. */
} while (memcg->res.usage > 0 || ret);
--
1.7.4.1
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>