[PATCH v2] mm/memfd_luo: don't use memfd_pin_folios()

Pratyush Yadav pratyush at kernel.org
Tue Nov 10 09:38:06 PST 2026


On Wed, Nov 04 2026, Mike Rapoport wrote:

>> memfd_luo takes a snapshot of the folios in the file's mapping. To make
>> sure the snapshot does not go out of sync, it pins the folios in the
>> file via memfd_pin_folios(). The pinning also migrates the folios out of
>> MIGRATE_MOVABLE to make sure doesn't land in KHO boot memory or in a CMA
>> region.
>> 
>> memfd_pin_folios() does not lock the folios. So it can race with
>> shmem_writeout(). A folio returned by memfd_pin_folios() might already
>> be in swap cache. This happens to work but is fragile. memfd_luo looks
>> at folio flags and index on a folio that isn't in the memfd's page
>> cache.
>> 
>> In addition, memfd_pin_folios() needs the caller to allocate an array of
>> folios. memfd_luo does not really need that since it will vmalloc
>> folios_ser and store the information there anyway. The folios array is
>> unnecessary overhead.
>> 
>> Finally, memfd_pin_folios() fills in holes in the memfd. This makes
>> preserving sparse memfds very expensive in both compute and memory.
>> While this patch doesn't yet stop filling in the holes, it sets up the
>> infrastructure to make it simpler.
>> 
>> Fix all this by walking the page cache directly using shmem_get_folio()
>> with SGP_CACHE. This returns a locked folio with an elevated refcount.
>> memfd_luo holds that reference until the memfd is unpreserved. This
>> blocks compaction, migration, etc.
>> 
>> There is still a need to migrate folios out of MIGRATE_MOVABLE page
>> blocks. This is because those might be CMA or KHO scratch. memfd_luo
>> will hold a longterm reference to the folios so it will block migration
>> in that pageblock. Introduce memfd_luo_migrate_folio() to do this.
>> 
>> Also factor out folio unpreservation logic into memfd_luo_put_folios()
>> and use it from memfd_luo_unpreserve_folios() and from error paths in
>> memfd_luo_preserve_folios().
>> 
>> Since shmem_get_folio() always returns uptodate folios, drop the folio
>> zeroing logic.
>> 
>> This change also improves preservation performance, shown in the table
>> below.
>> 
>> |------------------+----------+------------+-------|
>> | memfd            | baseline | this patch | delta |
>> |------------------+----------+------------+-------|
>> | 4K pages, 32 GiB | 748 ms   | 600 ms     |  -20% |
>> | THP, 4 GiB       | 0.59 ms  | 0.59 ms    |    0% |
>> | THP, 16 GiB      | 2.99 ms  | 2.61 ms    |  -13% |
>> | THP, 32 GiB      | 4.83 ms  | 5.21 ms    |   +8% |
>> | sparse, 32 GiB   | 8695 ms  | 7269 ms    |  -16% |
>> |------------------+----------+------------+-------|
>> 
>> Acked-by: Pasha Tatashin <pasha.tatashin at soleen.com>
>> Signed-off-by: Pratyush Yadav (Google) <pratyush at kernel.org>
>
> With a few nits below
>
> Reviewed-by: Mike Rapoport (Microsoft) <rppt at kernel.org>

Thanks!

>
>> +/*
>> + * Move a folio out of ZONE_MOVABLE or CMA, so it can be held until the next
>
> "held until next kernel" - here and below - I feel that something is off
> here, but can't pinpoint what exactly.

Hmm, perhaps this:

    Preservation holds a longterm reference on the folio. This would
    block migration in ZONE_MOVABLE or CMA. Move it out.

?

>
>> + * kernel. Called with a reference on the unlocked folio, which is always
>> + * dropped. The caller must look up the folio again afterwards.
>
> I'm having hard time parsing "called with a reference" sentence :)

What about this:

     Must be called with folio unlocked. Drops the reference to folio.
     The caller must look it up again.

?

>
>> +/*
>> + * Look up the folio at @index, allocating it if needed, and make sure it can be
>> + * held until the next kernel. Folios that are not longterm pinnable are migrated
>> + * out first, retrying a few times if someone else has them isolated.
>> + *
>> + * On success, returns 0 with *@foliop locked and holding a reference.
>> + */
>> +static int memfd_luo_get_folio(struct inode *inode, pgoff_t index,
>> +			       struct folio **foliop)
>> +{
>> +	int err, busy_retries = 0;
>> +	struct folio *folio;
>> +
>> +	for (;;) {
>
> Won't
> 	for (retry = 0; retry < MEMFD_LUO_MIGRATE_RETRIES; retry++) {
>
> look better and the body simpler?

Right, good idea.

>
>> +		err = shmem_get_folio(inode, index, 0, &folio, SGP_CACHE);
>> +		if (err)
>> +			return err;
>> +
>> +		if (likely(folio_is_longterm_pinnable(folio)))
>> +			break;
>> +
>> +		folio_unlock(folio);
>> +		err = memfd_luo_migrate_folio(folio);
>> +		/* Someone else might have it isolated, retry a bit. */
>> +		if (err == -EBUSY && ++busy_retries < MEMFD_LUO_MIGRATE_RETRIES)
>> +			continue;
>> +		if (err)
>> +			return err;
>> +	}
>> +
>> +	*foliop = folio;
>> +	return 0;
>> +}
>
> .

-- 
Regards,
Pratyush Yadav



More information about the kexec mailing list