Re: [PATCH v1 3/6] mm/migrate: don't call folio_putback_active_hugetlb() on dst hugetlb folio

[Baolin Wang <baolin.wang@linux.alibaba.com>]

On 2025/1/13 17:50, David Hildenbrand wrote:
> On 13.01.25 08:00, Baolin Wang wrote:
>>
>>
>> On 2025/1/11 02:21, David Hildenbrand wrote:
>>> We replaced a simple put_page() by a putback_active_hugepage() call in
>>> commit 3aaa76e125c1 ("mm: migrate: hugetlb: putback destination hugepage
>>> to active list"), to set the "active" flag on the dst hugetlb folio.
>>>
>>> Nowadays, we decoupled the "active" list from the flag, by calling the
>>> flag "migratable".
>>>
>>> Calling "putback" on something that wasn't allocated is weird and not
>>> future proof, especially if we might reach that path when migration 
>>> failed
>>> and we just want to free the freshly allocated hugetlb folio.
>>>
>>> Let's simply set the "migratable" flag in move_hugetlb_state(), where we
>>> know that allocation succeeded, and use simple folio_put() to return
>>> our reference.
>>>
>>> Do we need the hugetlb_lock for setting that flag? Staring at other
>>> users of folio_set_hugetlb_migratable(), it does not look like it. After
>>> all, the dst folio should already be on the active list, and we are not
>>> modifying that list.
>>>
>>> Signed-off-by: David Hildenbrand <david@redhat.com>
>>> ---
>>>    mm/hugetlb.c | 5 +++++
>>>    mm/migrate.c | 8 ++++----
>>>    2 files changed, 9 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>> index da98d671088d0..b24ccf8ecbf38 100644
>>> --- a/mm/hugetlb.c
>>> +++ b/mm/hugetlb.c
>>> @@ -7529,6 +7529,11 @@ void move_hugetlb_state(struct folio 
>>> *old_folio, struct folio *new_folio, int re
>>>            }
>>>            spin_unlock_irq(&hugetlb_lock);
>>>        }
>>> +    /*
>>> +     * Our old folio is isolated and has "migratable" cleared until it
>>> +     * is putback. As migration succeeded, set the new folio 
>>> "migratable".
>>> +     */
>>> +    folio_set_hugetlb_migratable(new_folio);
>>>    }
>>>    static void hugetlb_unshare_pmds(struct vm_area_struct *vma,
>>> diff --git a/mm/migrate.c b/mm/migrate.c
>>> index 80887cadb2774..7e23e78f1e57b 100644
>>> --- a/mm/migrate.c
>>> +++ b/mm/migrate.c
>>> @@ -1542,14 +1542,14 @@ static int 
>>> unmap_and_move_huge_page(new_folio_t get_new_folio,
>>>            list_move_tail(&src->lru, ret);
>>>        /*
>>> -     * If migration was not successful and there's a freeing 
>>> callback, use
>>> -     * it.  Otherwise, put_page() will drop the reference grabbed 
>>> during
>>> -     * isolation.
>>> +     * If migration was not successful and there's a freeing callback,
>>> +     * return the folio to that special allocator. Otherwise, simply 
>>> drop
>>> +     * our additional reference.
>>>         */
>>>        if (put_new_folio)
>>>            put_new_folio(dst, private);
>>>        else
>>> -        folio_putback_active_hugetlb(dst);
>>> +        folio_put(dst);
>>
> 
> Hi Baolin,
> 
> thanks for the review!
> 
>> IIUC, after the changes, so the 'dst' folio might not be added into the
>> 'h->hugepage_activelist' list (if the 'dst' i:s temporarily allocated),
>> Could this cause any side effects?
> 
> Good point, so far I was under the assumption that the dst folio would
> already be in the active list.
> 
> alloc_migration_target() and friends call alloc_hugetlb_folio_nodemask().
> 
> 
> There are two cases:
> 
> 1) We call dequeue_hugetlb_folio_nodemask() -> 
> dequeue_hugetlb_folio_node_exact()
> where we add the folio to the hugepage_activelist.
> 
> 2) We call alloc_migrate_hugetlb_folio()
> 
> It indeed looks like we don't add them to the active list. So likely we 
> should do:
> 
> 
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index dca4f310617a2..c6463dd7a1fc8 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -7546,7 +7546,10 @@ void move_hugetlb_state(struct folio *old_folio, 
> struct folio *new_folio, int re
>           * Our old folio is isolated and has "migratable" cleared until it
>           * is putback. As migration succeeded, set the new folio 
> "migratable".
>           */
> +       spin_lock_irq(&hugetlb_lock);
>          folio_set_hugetlb_migratable(new_folio);
> +       list_move_tail(&new_folio->lru, 
> &(folio_hstate(new_folio))->hugepage_activelist);
> +       spin_unlock_irq(&hugetlb_lock);
>   }
> 
>   static void hugetlb_unshare_pmds(struct vm_area_struct *vma,
> 
> 
> move_hugetlb_state() also takes care of that "temporary" handling.

LGTM. With the changes:
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>


> (I wonder if in case 1) it was a problem that the folio was already on the
> active list)

Seems harmless, and putback_active_hugepage() did the same before.