[PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[Dev Jain]

Suppose a folio is under migration, and khugepaged is also trying to
collapse it. collapse_pte_mapped_thp() will retrieve the folio from the
page cache via filemap_lock_folio(), thus taking a reference on the folio
and sleeping on the folio lock, since the lock is held by the migration
path. Migration will then fail in
__folio_migrate_mapping -> folio_ref_freeze. Reduce the probability of
such a race happening (leading to migration failure) by bailing out
if we detect a PMD is marked with a migration entry.

This fixes the migration-shared-anon-thp testcase failure on Apple M3.

Note that, this is not a "fix" since it only reduces the chance of
interference of khugepaged with migration, wherein both the kernel
functionalities are deemed "best-effort".

Signed-off-by: Dev Jain <dev.jain@arm.com>
---

v1->v2:
 - Remove SCAN_PMD_MIGRATION, merge into SCAN_PMD_MAPPED (David, Anshuman)
 - Add a comment (Lorenzo)

v1:
 - https://lore.kernel.org/all/20250630044837.4675-1-dev.jain@arm.com/

 mm/khugepaged.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 1aa7ca67c756..3fdefc4f4984 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -941,6 +941,15 @@ static inline int check_pmd_state(pmd_t *pmd)
 
 	if (pmd_none(pmde))
 		return SCAN_PMD_NONE;
+
+	/*
+	 * The folio may be under migration when khugepaged is trying to
+	 * collapse it. Migration success or failure will eventually end
+	 * up with the PMD still pointing to a PMD-order folio, so return
+	 * SCAN_PMD_MAPPED.
+	 */
+	if (is_pmd_migration_entry(pmde))
+		return SCAN_PMD_MAPPED;
 	if (!pmd_present(pmde))
 		return SCAN_PMD_NULL;
 	if (pmd_trans_huge(pmde))
-- 
2.30.2

Re: [PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[David Hildenbrand]

On 03.07.25 07:48, Dev Jain wrote:
> Suppose a folio is under migration, and khugepaged is also trying to
> collapse it. collapse_pte_mapped_thp() will retrieve the folio from the
> page cache via filemap_lock_folio(), thus taking a reference on the folio
> and sleeping on the folio lock, since the lock is held by the migration
> path. Migration will then fail in
> __folio_migrate_mapping -> folio_ref_freeze. Reduce the probability of
> such a race happening (leading to migration failure) by bailing out
> if we detect a PMD is marked with a migration entry.
> 
> This fixes the migration-shared-anon-thp testcase failure on Apple M3.
> 
> Note that, this is not a "fix" since it only reduces the chance of
> interference of khugepaged with migration, wherein both the kernel
> functionalities are deemed "best-effort".
> 
> Signed-off-by: Dev Jain <dev.jain@arm.com>
> ---
> 
> v1->v2:
>   - Remove SCAN_PMD_MIGRATION, merge into SCAN_PMD_MAPPED (David, Anshuman)
>   - Add a comment (Lorenzo)
> 
> v1:
>   - https://lore.kernel.org/all/20250630044837.4675-1-dev.jain@arm.com/
> 
>   mm/khugepaged.c | 9 +++++++++
>   1 file changed, 9 insertions(+)
> 
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 1aa7ca67c756..3fdefc4f4984 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -941,6 +941,15 @@ static inline int check_pmd_state(pmd_t *pmd)
>   
>   	if (pmd_none(pmde))
>   		return SCAN_PMD_NONE;
> +
> +	/*
> +	 * The folio may be under migration when khugepaged is trying to
> +	 * collapse it. Migration success or failure will eventually end
> +	 * up with the PMD still pointing to a PMD-order folio, so return
> +	 * SCAN_PMD_MAPPED.

Nit: the last part (, so return ..) is obvious from the code.

I would have written

/*
  * The folio may be under migration when khugepaged is trying to
  * collapse it. Migration success or failure will eventually end
  * up with a present PMD entry again.
  */


Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers,

David / dhildenb

Re: [PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[Anshuman Khandual]

On 03/07/25 2:55 PM, David Hildenbrand wrote:
> On 03.07.25 07:48, Dev Jain wrote:
>> Suppose a folio is under migration, and khugepaged is also trying to
>> collapse it. collapse_pte_mapped_thp() will retrieve the folio from the
>> page cache via filemap_lock_folio(), thus taking a reference on the folio
>> and sleeping on the folio lock, since the lock is held by the migration
>> path. Migration will then fail in
>> __folio_migrate_mapping -> folio_ref_freeze. Reduce the probability of
>> such a race happening (leading to migration failure) by bailing out
>> if we detect a PMD is marked with a migration entry.
>>
>> This fixes the migration-shared-anon-thp testcase failure on Apple M3.
>>
>> Note that, this is not a "fix" since it only reduces the chance of
>> interference of khugepaged with migration, wherein both the kernel
>> functionalities are deemed "best-effort".
>>
>> Signed-off-by: Dev Jain <dev.jain@arm.com>
>> ---
>>
>> v1->v2:
>>   - Remove SCAN_PMD_MIGRATION, merge into SCAN_PMD_MAPPED (David, Anshuman)
>>   - Add a comment (Lorenzo)
>>
>> v1:
>>   - https://lore.kernel.org/all/20250630044837.4675-1-dev.jain@arm.com/
>>
>>   mm/khugepaged.c | 9 +++++++++
>>   1 file changed, 9 insertions(+)
>>
>> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
>> index 1aa7ca67c756..3fdefc4f4984 100644
>> --- a/mm/khugepaged.c
>> +++ b/mm/khugepaged.c
>> @@ -941,6 +941,15 @@ static inline int check_pmd_state(pmd_t *pmd)
>>         if (pmd_none(pmde))
>>           return SCAN_PMD_NONE;
>> +
>> +    /*
>> +     * The folio may be under migration when khugepaged is trying to
>> +     * collapse it. Migration success or failure will eventually end
>> +     * up with the PMD still pointing to a PMD-order folio, so return
>> +     * SCAN_PMD_MAPPED.
> 
> Nit: the last part (, so return ..) is obvious from the code.
> 
> I would have written
> 
> /*
>  * The folio may be under migration when khugepaged is trying to
>  * collapse it. Migration success or failure will eventually end
>  * up with a present PMD entry again.
>  */

+1

> 
> 
> Acked-by: David Hildenbrand <david@redhat.com>
>

Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>

Re: [PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[Oscar Salvador]

On Thu, Jul 03, 2025 at 11:18:23AM +0530, Dev Jain wrote:
> Suppose a folio is under migration, and khugepaged is also trying to
> collapse it. collapse_pte_mapped_thp() will retrieve the folio from the
> page cache via filemap_lock_folio(), thus taking a reference on the folio
> and sleeping on the folio lock, since the lock is held by the migration
> path. Migration will then fail in
> __folio_migrate_mapping -> folio_ref_freeze. Reduce the probability of
> such a race happening (leading to migration failure) by bailing out
> if we detect a PMD is marked with a migration entry.
> 
> This fixes the migration-shared-anon-thp testcase failure on Apple M3.
> 
> Note that, this is not a "fix" since it only reduces the chance of
> interference of khugepaged with migration, wherein both the kernel
> functionalities are deemed "best-effort".
> 
> Signed-off-by: Dev Jain <dev.jain@arm.com>

David's comment refering to a 'present PMD entry' seems more clear to
me, but

Acked-by: Oscar Salvador <osalvador@suse.de>

 

-- 
Oscar Salvador
SUSE Labs

Re: [PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[Dev Jain]

On 03/07/25 2:55 pm, David Hildenbrand wrote:
> On 03.07.25 07:48, Dev Jain wrote:
>> Suppose a folio is under migration, and khugepaged is also trying to
>> collapse it. collapse_pte_mapped_thp() will retrieve the folio from the
>> page cache via filemap_lock_folio(), thus taking a reference on the 
>> folio
>> and sleeping on the folio lock, since the lock is held by the migration
>> path. Migration will then fail in
>> __folio_migrate_mapping -> folio_ref_freeze. Reduce the probability of
>> such a race happening (leading to migration failure) by bailing out
>> if we detect a PMD is marked with a migration entry.
>>
>> This fixes the migration-shared-anon-thp testcase failure on Apple M3.
>>
>> Note that, this is not a "fix" since it only reduces the chance of
>> interference of khugepaged with migration, wherein both the kernel
>> functionalities are deemed "best-effort".
>>
>> Signed-off-by: Dev Jain <dev.jain@arm.com>
>> ---
>>
>> v1->v2:
>>   - Remove SCAN_PMD_MIGRATION, merge into SCAN_PMD_MAPPED (David, 
>> Anshuman)
>>   - Add a comment (Lorenzo)
>>
>> v1:
>>   - https://lore.kernel.org/all/20250630044837.4675-1-dev.jain@arm.com/
>>
>>   mm/khugepaged.c | 9 +++++++++
>>   1 file changed, 9 insertions(+)
>>
>> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
>> index 1aa7ca67c756..3fdefc4f4984 100644
>> --- a/mm/khugepaged.c
>> +++ b/mm/khugepaged.c
>> @@ -941,6 +941,15 @@ static inline int check_pmd_state(pmd_t *pmd)
>>         if (pmd_none(pmde))
>>           return SCAN_PMD_NONE;
>> +
>> +    /*
>> +     * The folio may be under migration when khugepaged is trying to
>> +     * collapse it. Migration success or failure will eventually end
>> +     * up with the PMD still pointing to a PMD-order folio, so return
>> +     * SCAN_PMD_MAPPED.
>
> Nit: the last part (, so return ..) is obvious from the code.
>
> I would have written
>
> /*
>  * The folio may be under migration when khugepaged is trying to
>  * collapse it. Migration success or failure will eventually end
>  * up with a present PMD entry again.
>  */
>
Thanks for the suggestion, but

PMD pointing to PMD-order folio necessarily implies present PMD entry,

but the converse is not true? For example it may point to a PTE table.

So I wanted to make it explicitly clear that the transient entry will be

converted to a PMD-THP in the end.


>
> Acked-by: David Hildenbrand <david@redhat.com>
Thanks!

Re: [PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[David Hildenbrand]

On 03.07.25 11:52, Dev Jain wrote:
> 
> On 03/07/25 2:55 pm, David Hildenbrand wrote:
>> On 03.07.25 07:48, Dev Jain wrote:
>>> Suppose a folio is under migration, and khugepaged is also trying to
>>> collapse it. collapse_pte_mapped_thp() will retrieve the folio from the
>>> page cache via filemap_lock_folio(), thus taking a reference on the
>>> folio
>>> and sleeping on the folio lock, since the lock is held by the migration
>>> path. Migration will then fail in
>>> __folio_migrate_mapping -> folio_ref_freeze. Reduce the probability of
>>> such a race happening (leading to migration failure) by bailing out
>>> if we detect a PMD is marked with a migration entry.
>>>
>>> This fixes the migration-shared-anon-thp testcase failure on Apple M3.
>>>
>>> Note that, this is not a "fix" since it only reduces the chance of
>>> interference of khugepaged with migration, wherein both the kernel
>>> functionalities are deemed "best-effort".
>>>
>>> Signed-off-by: Dev Jain <dev.jain@arm.com>
>>> ---
>>>
>>> v1->v2:
>>>    - Remove SCAN_PMD_MIGRATION, merge into SCAN_PMD_MAPPED (David,
>>> Anshuman)
>>>    - Add a comment (Lorenzo)
>>>
>>> v1:
>>>    - https://lore.kernel.org/all/20250630044837.4675-1-dev.jain@arm.com/
>>>
>>>    mm/khugepaged.c | 9 +++++++++
>>>    1 file changed, 9 insertions(+)
>>>
>>> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
>>> index 1aa7ca67c756..3fdefc4f4984 100644
>>> --- a/mm/khugepaged.c
>>> +++ b/mm/khugepaged.c
>>> @@ -941,6 +941,15 @@ static inline int check_pmd_state(pmd_t *pmd)
>>>          if (pmd_none(pmde))
>>>            return SCAN_PMD_NONE;
>>> +
>>> +    /*
>>> +     * The folio may be under migration when khugepaged is trying to
>>> +     * collapse it. Migration success or failure will eventually end
>>> +     * up with the PMD still pointing to a PMD-order folio, so return
>>> +     * SCAN_PMD_MAPPED.
>>
>> Nit: the last part (, so return ..) is obvious from the code.
>>
>> I would have written
>>
>> /*
>>   * The folio may be under migration when khugepaged is trying to
>>   * collapse it. Migration success or failure will eventually end
>>   * up with a present PMD entry again.
>>   */
>>
> Thanks for the suggestion, but
> 
> PMD pointing to PMD-order folio necessarily implies present PMD entry,
 > > but the converse is not true? For example it may point to a PTE table.

I see, talking about orders is confusing though.

"with a present PMD mapping a folio again."

-- 
Cheers,

David / dhildenb

Re: [PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[Dev Jain]

On 03/07/25 3:39 pm, David Hildenbrand wrote:
> On 03.07.25 11:52, Dev Jain wrote:
>>
>> On 03/07/25 2:55 pm, David Hildenbrand wrote:
>>> On 03.07.25 07:48, Dev Jain wrote:
>>>> Suppose a folio is under migration, and khugepaged is also trying to
>>>> collapse it. collapse_pte_mapped_thp() will retrieve the folio from 
>>>> the
>>>> page cache via filemap_lock_folio(), thus taking a reference on the
>>>> folio
>>>> and sleeping on the folio lock, since the lock is held by the 
>>>> migration
>>>> path. Migration will then fail in
>>>> __folio_migrate_mapping -> folio_ref_freeze. Reduce the probability of
>>>> such a race happening (leading to migration failure) by bailing out
>>>> if we detect a PMD is marked with a migration entry.
>>>>
>>>> This fixes the migration-shared-anon-thp testcase failure on Apple M3.
>>>>
>>>> Note that, this is not a "fix" since it only reduces the chance of
>>>> interference of khugepaged with migration, wherein both the kernel
>>>> functionalities are deemed "best-effort".
>>>>
>>>> Signed-off-by: Dev Jain <dev.jain@arm.com>
>>>> ---
>>>>
>>>> v1->v2:
>>>>    - Remove SCAN_PMD_MIGRATION, merge into SCAN_PMD_MAPPED (David,
>>>> Anshuman)
>>>>    - Add a comment (Lorenzo)
>>>>
>>>> v1:
>>>>    - 
>>>> https://lore.kernel.org/all/20250630044837.4675-1-dev.jain@arm.com/
>>>>
>>>>    mm/khugepaged.c | 9 +++++++++
>>>>    1 file changed, 9 insertions(+)
>>>>
>>>> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
>>>> index 1aa7ca67c756..3fdefc4f4984 100644
>>>> --- a/mm/khugepaged.c
>>>> +++ b/mm/khugepaged.c
>>>> @@ -941,6 +941,15 @@ static inline int check_pmd_state(pmd_t *pmd)
>>>>          if (pmd_none(pmde))
>>>>            return SCAN_PMD_NONE;
>>>> +
>>>> +    /*
>>>> +     * The folio may be under migration when khugepaged is trying to
>>>> +     * collapse it. Migration success or failure will eventually end
>>>> +     * up with the PMD still pointing to a PMD-order folio, so return
>>>> +     * SCAN_PMD_MAPPED.
>>>
>>> Nit: the last part (, so return ..) is obvious from the code.
>>>
>>> I would have written
>>>
>>> /*
>>>   * The folio may be under migration when khugepaged is trying to
>>>   * collapse it. Migration success or failure will eventually end
>>>   * up with a present PMD entry again.
>>>   */
>>>
>> Thanks for the suggestion, but
>>
>> PMD pointing to PMD-order folio necessarily implies present PMD entry,
> > > but the converse is not true? For example it may point to a PTE 
> table.
>
> I see, talking about orders is confusing though.
>
> "with a present PMD mapping a folio again."

Sure. Shall respin.

Re: [PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[Zi Yan]

On 3 Jul 2025, at 1:48, Dev Jain wrote:

> Suppose a folio is under migration, and khugepaged is also trying to
> collapse it. collapse_pte_mapped_thp() will retrieve the folio from the
> page cache via filemap_lock_folio(), thus taking a reference on the folio
> and sleeping on the folio lock, since the lock is held by the migration
> path. Migration will then fail in
> __folio_migrate_mapping -> folio_ref_freeze. Reduce the probability of
> such a race happening (leading to migration failure) by bailing out
> if we detect a PMD is marked with a migration entry.
>
> This fixes the migration-shared-anon-thp testcase failure on Apple M3.
>
> Note that, this is not a "fix" since it only reduces the chance of
> interference of khugepaged with migration, wherein both the kernel
> functionalities are deemed "best-effort".
>
> Signed-off-by: Dev Jain <dev.jain@arm.com>
> ---
>
> v1->v2:
>  - Remove SCAN_PMD_MIGRATION, merge into SCAN_PMD_MAPPED (David, Anshuman)
>  - Add a comment (Lorenzo)
>
> v1:
>  - https://lore.kernel.org/all/20250630044837.4675-1-dev.jain@arm.com/
>
>  mm/khugepaged.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
>

Reviewed-by: Zi Yan <ziy@nvidia.com>

Best Regards,
Yan, Zi

Re: [PATCH v2] khugepaged: Reduce race probability between migration and khugepaged

[Baolin Wang]

On 2025/7/3 13:48, Dev Jain wrote:
> Suppose a folio is under migration, and khugepaged is also trying to
> collapse it. collapse_pte_mapped_thp() will retrieve the folio from the
> page cache via filemap_lock_folio(), thus taking a reference on the folio
> and sleeping on the folio lock, since the lock is held by the migration
> path. Migration will then fail in
> __folio_migrate_mapping -> folio_ref_freeze. Reduce the probability of
> such a race happening (leading to migration failure) by bailing out
> if we detect a PMD is marked with a migration entry.
> 
> This fixes the migration-shared-anon-thp testcase failure on Apple M3.
> 
> Note that, this is not a "fix" since it only reduces the chance of
> interference of khugepaged with migration, wherein both the kernel
> functionalities are deemed "best-effort".
> 
> Signed-off-by: Dev Jain <dev.jain@arm.com>

With David's comments addressed, LGTM.

Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>