[PATCH v2] mm/hugetlb: initialize hugetlb_usage in mm_init

[PATCH v2] mm/hugetlb: initialize hugetlb_usage in mm_init

[Liu Zixian]

After fork, the child process will get incorrect (2x) hugetlb_usage.
If a process uses 5 2MB hugetlb pages in an anonymous mapping,
	HugetlbPages:	   10240 kB
and then forks, the child will show,
	HugetlbPages:	   20480 kB
The reason for double the amount is because hugetlb_usage will be
copied from the parent and then increased when we copy page tables
from parent to child. Child will have 2x actual usage.

Fix this by adding hugetlb_count_init in mm_init.

Fixes: 5d317b2b6536 ("mm: hugetlb: proc: add HugetlbPages field to
/proc/PID/status")
Signed-off-by: Liu Zixian <liuzixian4@huawei.com>
---
v2:
1. Create two hugetlb_count_init in hugetlb.h instead of using #ifdef
  in fork.c
2. Add an example to clearify this issue.
---
 include/linux/hugetlb.h | 9 +++++++++
 kernel/fork.c           | 1 +
 2 files changed, 10 insertions(+)

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index f7ca1a387..1faebe1cd 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -858,6 +858,11 @@ static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
 
 void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm);
 
+static inline void hugetlb_count_init(struct mm_struct *mm)
+{
+	atomic_long_set(&mm->hugetlb_usage, 0);
+}
+
 static inline void hugetlb_count_add(long l, struct mm_struct *mm)
 {
 	atomic_long_add(l, &mm->hugetlb_usage);
@@ -1042,6 +1047,10 @@ static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
 	return &mm->page_table_lock;
 }
 
+static inline void hugetlb_count_init(struct mm_struct *mm)
+{
+}
+
 static inline void hugetlb_report_usage(struct seq_file *f, struct mm_struct *m)
 {
 }
diff --git a/kernel/fork.c b/kernel/fork.c
index bc94b2cc5..0dbc96ade 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1050,6 +1050,7 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
 	mm->pmd_huge_pte = NULL;
 #endif
 	mm_init_uprobes_state(mm);
+	hugetlb_count_init(mm);
 
 	if (current->mm) {
 		mm->flags = current->mm->flags & MMF_INIT_MASK;
-- 
2.18.1

Re: [PATCH v2] mm/hugetlb: initialize hugetlb_usage in mm_init

[HORIGUCHI NAOYA(堀口 直也)]

On Thu, Aug 26, 2021 at 03:17:42PM +0800, Liu Zixian wrote:
> After fork, the child process will get incorrect (2x) hugetlb_usage.
> If a process uses 5 2MB hugetlb pages in an anonymous mapping,
> 	HugetlbPages:	   10240 kB
> and then forks, the child will show,
> 	HugetlbPages:	   20480 kB
> The reason for double the amount is because hugetlb_usage will be
> copied from the parent and then increased when we copy page tables
> from parent to child. Child will have 2x actual usage.
> 
> Fix this by adding hugetlb_count_init in mm_init.
> 
> Fixes: 5d317b2b6536 ("mm: hugetlb: proc: add HugetlbPages field to
> /proc/PID/status")
> Signed-off-by: Liu Zixian <liuzixian4@huawei.com>

Thank you for the fix.

Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>

Re: [PATCH v2] mm/hugetlb: initialize hugetlb_usage in mm_init

[Mike Kravetz]

On 8/26/21 12:17 AM, Liu Zixian wrote:
> After fork, the child process will get incorrect (2x) hugetlb_usage.
> If a process uses 5 2MB hugetlb pages in an anonymous mapping,
> 	HugetlbPages:	   10240 kB
> and then forks, the child will show,
> 	HugetlbPages:	   20480 kB
> The reason for double the amount is because hugetlb_usage will be
> copied from the parent and then increased when we copy page tables
> from parent to child. Child will have 2x actual usage.
> 
> Fix this by adding hugetlb_count_init in mm_init.
> 
> Fixes: 5d317b2b6536 ("mm: hugetlb: proc: add HugetlbPages field to
> /proc/PID/status")
> Signed-off-by: Liu Zixian <liuzixian4@huawei.com>
> ---
> v2:
> 1. Create two hugetlb_count_init in hugetlb.h instead of using #ifdef
>   in fork.c
> 2. Add an example to clearify this issue.
> ---
>  include/linux/hugetlb.h | 9 +++++++++
>  kernel/fork.c           | 1 +
>  2 files changed, 10 insertions(+)

Thanks,

Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>

Andrew, can you add a CC stable?
-- 
Mike Kravetz