Michal Hocko [off-list ref] writes:

On Mon 11-06-12 15:10:20, Aneesh Kumar K.V wrote:

quoted

Michal Hocko [off-list ref] writes:

[...]

quoted

quoted

quoted

+static int hugetlb_cgroup_move_parent(int idx, struct cgroup *cgroup,
+				      struct page *page)

deserves a comment about the locking (needs to be called with
hugetlb_lock).

will do

quoted

quoted

+{
+	int csize;
+	struct res_counter *counter;
+	struct res_counter *fail_res;
+	struct hugetlb_cgroup *page_hcg;
+	struct hugetlb_cgroup *h_cg   = hugetlb_cgroup_from_cgroup(cgroup);
+	struct hugetlb_cgroup *parent = parent_hugetlb_cgroup(cgroup);
+
+	if (!get_page_unless_zero(page))
+		goto out;
+
+	page_hcg = hugetlb_cgroup_from_page(page);
+	/*
+	 * We can have pages in active list without any cgroup
+	 * ie, hugepage with less than 3 pages. We can safely
+	 * ignore those pages.
+	 */
+	if (!page_hcg || page_hcg != h_cg)
+		goto err_out;

How can we have page_hcg != NULL && page_hcg != h_cg?

pages belonging to other cgroup ?

OK, I've forgot that you are iterating over all active huge pages in
hugetlb_cgroup_pre_destroy. What prevents you from doing the filtering
in the caller? 
I am also wondering why you need to play with the page reference
counting here. You are under hugetlb_lock so the page cannot disappear
in the meantime or am I missing something?

That is correct. Updated the patch and also added the below comment to
the function.
+
+/*
+ * Should be called with hugetlb_lock held.
+ * Since we are holding hugetlb_lock, pages cannot get moved from
+ * active list or uncharged from the cgroup, So no need to get
+ * page reference and test for page active here.
+ */

-aneesh

--
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>