@@ -902,7 +902,7 @@ static long hugetlbfs_fallocate(struct file *file, int mode, loff_t offset,
* to keep reservation accounting consistent.
*/
hugetlb_set_vma_policy(&pseudo_vma, inode, index);
- folio = alloc_hugetlb_folio(&pseudo_vma, addr, 0);
+ folio = alloc_hugetlb_folio(&pseudo_vma, addr, 0, true);
hugetlb_drop_vma_policy(&pseudo_vma);
if (IS_ERR(folio)) {
mutex_unlock(&hugetlb_fault_mutex_table[hash]);
@@ -713,7 +713,8 @@ struct huge_bootmem_page {
int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list);
struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
- unsigned long addr, int avoid_reserve);
+ unsigned long addr, int avoid_reserve,
+ bool restore_reserve_on_memcg_failure);
struct folio *alloc_hugetlb_folio_nodemask(struct hstate *h, int preferred_nid,
nodemask_t *nmask, gfp_t gfp_mask);
struct folio *alloc_hugetlb_folio_vma(struct hstate *h, struct vm_area_struct *vma,
@@ -1016,7 +1017,8 @@ static inline int isolate_or_dissolve_huge_page(struct page *page,
static inline struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
unsigned long addr,
- int avoid_reserve)
+ int avoid_reserve,
+ bool restore_reserve_on_memcg_failure)
{
return NULL;
}
@@ -677,6 +677,8 @@ static inline int mem_cgroup_charge(struct folio *folio, struct mm_struct *mm,
return __mem_cgroup_charge(folio, mm, gfp);
}
+int mem_cgroup_hugetlb_charge_folio(struct folio *folio, gfp_t gfp);
+
int mem_cgroup_swapin_charge_folio(struct folio *folio, struct mm_struct *mm,
gfp_t gfp, swp_entry_t entry);
void mem_cgroup_swapin_uncharge_swap(swp_entry_t entry);
@@ -1251,6 +1253,12 @@ static inline int mem_cgroup_charge(struct folio *folio,
return 0;
}
+static inline int mem_cgroup_hugetlb_charge_folio(struct folio *folio,
+ gfp_t gfp)
+{
+ return 0;
+}
+
static inline int mem_cgroup_swapin_charge_folio(struct folio *folio,
struct mm_struct *mm, gfp_t gfp, swp_entry_t entry)
{
@@ -1902,6 +1902,7 @@ void free_huge_folio(struct folio *folio)
pages_per_huge_page(h), folio);
hugetlb_cgroup_uncharge_folio_rsvd(hstate_index(h),
pages_per_huge_page(h), folio);
+ mem_cgroup_uncharge(folio);
if (restore_reserve)
h->resv_huge_pages++;
@@ -3004,7 +3005,8 @@ int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list)
}
struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
- unsigned long addr, int avoid_reserve)
+ unsigned long addr, int avoid_reserve,
+ bool restore_reserve_on_memcg_failure)
{
struct hugepage_subpool *spool = subpool_vma(vma);
struct hstate *h = hstate_vma(vma);
@@ -3119,6 +3121,15 @@ struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
hugetlb_cgroup_uncharge_folio_rsvd(hstate_index(h),
pages_per_huge_page(h), folio);
}
+
+ /* undo allocation if memory controller disallows it. */
+ if (mem_cgroup_hugetlb_charge_folio(folio, GFP_KERNEL)) {
+ if (restore_reserve_on_memcg_failure)
+ restore_reserve_on_error(h, vma, addr, folio);
+ folio_put(folio);
+ return ERR_PTR(-ENOMEM);
+ }
+
return folio;
out_uncharge_cgroup:
@@ -5179,7 +5190,7 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
spin_unlock(src_ptl);
spin_unlock(dst_ptl);
/* Do not use reserve as it's private owned */
- new_folio = alloc_hugetlb_folio(dst_vma, addr, 1);
+ new_folio = alloc_hugetlb_folio(dst_vma, addr, 1, false);
if (IS_ERR(new_folio)) {
folio_put(pte_folio);
ret = PTR_ERR(new_folio);
@@ -5656,7 +5667,7 @@ static vm_fault_t hugetlb_wp(struct mm_struct *mm, struct vm_area_struct *vma,
* be acquired again before returning to the caller, as expected.
*/
spin_unlock(ptl);
- new_folio = alloc_hugetlb_folio(vma, haddr, outside_reserve);
+ new_folio = alloc_hugetlb_folio(vma, haddr, outside_reserve, true);
if (IS_ERR(new_folio)) {
/*
@@ -5930,7 +5941,7 @@ static vm_fault_t hugetlb_no_page(struct mm_struct *mm,
VM_UFFD_MISSING);
}
- folio = alloc_hugetlb_folio(vma, haddr, 0);
+ folio = alloc_hugetlb_folio(vma, haddr, 0, true);
if (IS_ERR(folio)) {
/*
* Returning error will result in faulting task being
@@ -6352,7 +6363,7 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
goto out;
}
- folio = alloc_hugetlb_folio(dst_vma, dst_addr, 0);
+ folio = alloc_hugetlb_folio(dst_vma, dst_addr, 0, true);
if (IS_ERR(folio)) {
ret = -ENOMEM;
goto out;
@@ -6394,7 +6405,7 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
goto out;
}
- folio = alloc_hugetlb_folio(dst_vma, dst_addr, 0);
+ folio = alloc_hugetlb_folio(dst_vma, dst_addr, 0, false);
if (IS_ERR(folio)) {
folio_put(*foliop);
ret = -ENOMEM;
@@ -7050,6 +7050,46 @@ int __mem_cgroup_charge(struct folio *folio, struct mm_struct *mm, gfp_t gfp)
return ret;
}
+static struct mem_cgroup *get_mem_cgroup_from_current(void)
+{
+ struct mem_cgroup *memcg;
+
+again:
+ rcu_read_lock();
+ memcg = mem_cgroup_from_task(current);
+ if (!css_tryget(&memcg->css)) {
+ rcu_read_unlock();
+ goto again;
+ }
+ rcu_read_unlock();
+ return memcg;
+}
+
+/**
+ * mem_cgroup_hugetlb_charge_folio - Charge a newly allocated hugetlb folio.
+ * @folio: folio to charge.
+ * @gfp: reclaim mode
+ *
+ * This function charges an allocated hugetlbf folio to the memcg of the
+ * current task.
+ *
+ * Returns 0 on success. Otherwise, an error code is returned.
+ */
+int mem_cgroup_hugetlb_charge_folio(struct folio *folio, gfp_t gfp)
+{
+ struct mem_cgroup *memcg;
+ int ret;
+
+ if (mem_cgroup_disabled())
+ return 0;
+
+ memcg = get_mem_cgroup_from_current();
+ ret = charge_memcg(folio, memcg, gfp);
+ mem_cgroup_put(memcg);
+
+ return ret;
+}
+
/**
* mem_cgroup_swapin_charge_folio - Charge a newly allocated folio for swapin.
* @folio: folio to charge.
Currently, hugetlb memory usage is not acounted for in the memory controller, which could lead to memory overprotection for cgroups with hugetlb-backed memory. This has been observed in our production system. This patch rectifies this issue by charging the memcg when the hugetlb folio is allocated, and uncharging when the folio is freed (analogous to the hugetlb controller). Signed-off-by: Nhat Pham <nphamcs@gmail.com> --- fs/hugetlbfs/inode.c | 2 +- include/linux/hugetlb.h | 6 ++++-- include/linux/memcontrol.h | 8 ++++++++ mm/hugetlb.c | 23 ++++++++++++++++------ mm/memcontrol.c | 40 ++++++++++++++++++++++++++++++++++++++ 5 files changed, 70 insertions(+), 9 deletions(-)