hugetlb: fix hugetlb cgroup refcounting during mremap
authorBui Quang Minh <minhquangbui99@gmail.com>
Sat, 20 Nov 2021 00:43:40 +0000 (16:43 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 20 Nov 2021 18:35:54 +0000 (10:35 -0800)
When hugetlb_vm_op_open() is called during copy_vma(), we may take the
reference to resv_map->css.  Later, when clearing the reservation
pointer of old_vma after transferring it to new_vma, we forget to drop
the reference to resv_map->css.  This leads to a reference leak of css.

Fixes this by adding a check to drop reservation css reference in
clear_vma_resv_huge_pages()

Link: https://lkml.kernel.org/r/20211113154412.91134-1-minhquangbui99@gmail.com
Fixes: 550a7d60bd5e35 ("mm, hugepages: add mremap() support for hugepage backed vma")
Signed-off-by: Bui Quang Minh <minhquangbui99@gmail.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/hugetlb_cgroup.h
mm/hugetlb.c

index c137396129db632786e6b5720646226b7814addd..ba025ae278827a2b26df2f682acf7ec32d0784cc 100644 (file)
@@ -128,6 +128,13 @@ static inline void resv_map_dup_hugetlb_cgroup_uncharge_info(
                css_get(resv_map->css);
 }
 
+static inline void resv_map_put_hugetlb_cgroup_uncharge_info(
+                                               struct resv_map *resv_map)
+{
+       if (resv_map->css)
+               css_put(resv_map->css);
+}
+
 extern int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
                                        struct hugetlb_cgroup **ptr);
 extern int hugetlb_cgroup_charge_cgroup_rsvd(int idx, unsigned long nr_pages,
@@ -211,6 +218,11 @@ static inline void resv_map_dup_hugetlb_cgroup_uncharge_info(
 {
 }
 
+static inline void resv_map_put_hugetlb_cgroup_uncharge_info(
+                                               struct resv_map *resv_map)
+{
+}
+
 static inline int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
                                               struct hugetlb_cgroup **ptr)
 {
index e09159c957e367f9adb37f9cc648d2cc39ff67ac..3a2479003ddf2f8b136035802544b9c612b021c5 100644 (file)
@@ -1037,8 +1037,10 @@ void clear_vma_resv_huge_pages(struct vm_area_struct *vma)
         */
        struct resv_map *reservations = vma_resv_map(vma);
 
-       if (reservations && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
+       if (reservations && is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
+               resv_map_put_hugetlb_cgroup_uncharge_info(reservations);
                kref_put(&reservations->refs, resv_map_release);
+       }
 
        reset_vma_resv_huge_pages(vma);
 }