On 22 Apr 2026, at 22:43, Lance Yang wrote:
> On Fri, Apr 17, 2026 at 10:44:18PM -0400, Zi Yan wrote:
>> collapse_file() requires FSes supporting large folio with at least
>> PMD_ORDER, so replace the READ_ONLY_THP_FOR_FS check with that.
>> MADV_COLLAPSE ignores shmem huge config, so exclude the check for shmem.
>>
>> While at it, replace VM_BUG_ON with VM_WARN_ON_ONCE.
>>
>> Add a helper function mapping_pmd_thp_support() for FSes supporting large
>> folio with at least PMD_ORDER.
>>
>> Signed-off-by: Zi Yan <[email protected]>
>> ---
>> include/linux/pagemap.h | 10 ++++++++++
>> mm/khugepaged.c | 5 +++--
>> 2 files changed, 13 insertions(+), 2 deletions(-)
>>
>> diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
>> index ec442af3f886..c3cb1ec982cd 100644
>> --- a/include/linux/pagemap.h
>> +++ b/include/linux/pagemap.h
>> @@ -524,6 +524,16 @@ static inline bool mapping_large_folio_support(const
>> struct address_space *mappi
>> return mapping_max_folio_order(mapping) > 0;
>> }
>>
>> +static inline bool mapping_pmd_thp_support(const struct address_space
>> *mapping)
>> +{
>> + /* AS_FOLIO_ORDER is only reasonable for pagecache folios */
>> + VM_WARN_ONCE((unsigned long)mapping & FOLIO_MAPPING_ANON,
>> + "Anonymous mapping always supports PMD THP");
>
> Nit: afraid not, at least when running on architectures without PMD leaf
> entries ...
>
> Maybe better to say this helper is only meaningful for pagecache-backed
> mappings. Anonymous mappings should not reach here.
Good suggestion. Will fix it.
>
>> +
>> + return mapping_max_folio_order(mapping) >= PMD_ORDER;
>> +}
>> +
>> +
>> /* Return the maximum folio size for this pagecache mapping, in bytes. */
>> static inline size_t mapping_max_folio_size(const struct address_space
>> *mapping)
>> {
>> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
>> index b8452dbdb043..3eb5d982d3d3 100644
>> --- a/mm/khugepaged.c
>> +++ b/mm/khugepaged.c
>> @@ -1892,8 +1892,9 @@ static enum scan_result collapse_file(struct mm_struct
>> *mm, unsigned long addr,
>> int nr_none = 0;
>> bool is_shmem = shmem_file(file);
>>
>> - VM_BUG_ON(!IS_ENABLED(CONFIG_READ_ONLY_THP_FOR_FS) && !is_shmem);
>> - VM_BUG_ON(start & (HPAGE_PMD_NR - 1));
>> + /* MADV_COLLAPSE ignores shmem huge config, so do not check shmem */
>> + VM_WARN_ON_ONCE(!is_shmem && !mapping_pmd_thp_support(mapping));
>
> With [1], can we drop !is_shmem here as well? shmem would then always
> call mapping_set_large_folios(inode->i_mapping):
>
> ---8<---
> diff --git a/mm/shmem.c b/mm/shmem.c
> index 4ecefe02881d..dafbea53b22d 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -3087,10 +3087,7 @@ static struct inode *__shmem_get_inode(struct
> mnt_idmap *idmap,
> cache_no_acl(inode);
> if (sbinfo->noswap)
> mapping_set_unevictable(inode->i_mapping);
> -
> - /* Don't consider 'deny' for emergencies and 'force' for testing */
> - if (sbinfo->huge)
> - mapping_set_large_folios(inode->i_mapping);
> + mapping_set_large_folios(inode->i_mapping);
>
> switch (mode & S_IFMT) {
> default:
> --
>
> But we can do that in a follow-up, once the revert lands :)
Right. That would make this patchset depend on Baolin’s fix. A follow-up
patch is easier for managing these patches. I will add a TODO in the
comment, so that we will not forget. Thank you for the suggestion.
>
> [1]
> https://lore.kernel.org/linux-mm/b2c7deee259a94b0d00a7c320d8d24d2c421f761.1776908112.git.baolin.w...@linux.alibaba.com/
Best Regards,
Yan, Zi
