[PATCH] mm/khugepaged: don't map folios that are not uptodate in MADV_COLLAPSE

[PATCH] mm/khugepaged: don't map folios that are not uptodate in MADV_COLLAPSE

[David Hildenbrand (Arm)]

madvise(MADV_COLLAPSE) can map PMD-sized pagecache folios that are not
uptodate. For example, shmem fallocate() can put such folios into the
page cache.

collapse_file() already refuses to proceed with folios that are not
uptodate. Let's just do the same in try_collapse_pte_mapped_thp() and skip
such folios, leaving handling to the page faults.

Note that collapse_file() effectively clears shmem folios through
shmem_get_folio(); let's keep this fix simple and simply let the page
fault path deal with that.

Fixes: 34488399fa08 ("mm/madvise: add file and shmem support to MADV_COLLAPSE")
Reported-by: Chris Mason <mason@kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
---
Once reviewed, I will pick this up in a timely manner.
---
 mm/khugepaged.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 8d9f05a39afb3..db135b3c44ef1 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -1900,6 +1900,12 @@ static enum scan_result try_collapse_pte_mapped_thp(struct mm_struct *mm, unsign
 		goto drop_folio;
 	}
 
+	/* Leave any uninitialized folios to the page fault path. */
+	if (!folio_test_uptodate(folio)) {
+		result = SCAN_FAIL;
+		goto drop_folio;
+	}
+
 	result = find_pmd_or_thp_or_none(mm, haddr, &pmd);
 	switch (result) {
 	case SCAN_SUCCEED:

---

base-commit: f44fd601f72e09e44c667b9234c9cd469856679c

change-id: 20261105-set_huge_pmd_uptodate-9b9820307f30

--

Cheers,

David

Re: [PATCH] mm/khugepaged: don't map folios that are not uptodate in MADV_COLLAPSE

[Lorenzo Stoakes (ARM)]

On Thu, Nov 05, 2026 at 07:16:44PM +0100, David Hildenbrand (Arm) wrote:
> madvise(MADV_COLLAPSE) can map PMD-sized pagecache folios that are not
> uptodate. For example, shmem fallocate() can put such folios into the
> page cache.
>
> collapse_file() already refuses to proceed with folios that are not
> uptodate. Let's just do the same in try_collapse_pte_mapped_thp() and skip
> such folios, leaving handling to the page faults.
>
> Note that collapse_file() effectively clears shmem folios through
> shmem_get_folio(); let's keep this fix simple and simply let the page
> fault path deal with that.
>
> Fixes: 34488399fa08 ("mm/madvise: add file and shmem support to MADV_COLLAPSE")
> Reported-by: Chris Mason <mason@kernel.org>
> Cc: stable@vger.kernel.org
> Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>

Looks reasonable to me, there's no reason we should be dealing with non-uptodate
folios there.

Reviewed-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

> ---
> Once reviewed, I will pick this up in a timely manner.
> ---
>  mm/khugepaged.c | 6 ++++++
>  1 file changed, 6 insertions(+)
>
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 8d9f05a39afb3..db135b3c44ef1 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -1900,6 +1900,12 @@ static enum scan_result try_collapse_pte_mapped_thp(struct mm_struct *mm, unsign
>  		goto drop_folio;
>  	}
>
> +	/* Leave any uninitialized folios to the page fault path. */
> +	if (!folio_test_uptodate(folio)) {
> +		result = SCAN_FAIL;
> +		goto drop_folio;
> +	}
> +
>  	result = find_pmd_or_thp_or_none(mm, haddr, &pmd);
>  	switch (result) {
>  	case SCAN_SUCCEED:
>
> ---
>
> base-commit: f44fd601f72e09e44c667b9234c9cd469856679c
>
> change-id: 20261105-set_huge_pmd_uptodate-9b9820307f30
>
> --
>
> Cheers,
>
> David
>

--
Cheers, Lorenzo

Re: [PATCH] mm/khugepaged: don't map folios that are not uptodate in MADV_COLLAPSE

[Hugh Dickins]

On Thu, 5 Nov 2026, David Hildenbrand (Arm) wrote:

> madvise(MADV_COLLAPSE) can map PMD-sized pagecache folios that are not
> uptodate. For example, shmem fallocate() can put such folios into the
> page cache.
> 
> collapse_file() already refuses to proceed with folios that are not
> uptodate. Let's just do the same in try_collapse_pte_mapped_thp() and skip
> such folios, leaving handling to the page faults.
> 
> Note that collapse_file() effectively clears shmem folios through
> shmem_get_folio(); let's keep this fix simple and simply let the page
> fault path deal with that.
> 
> Fixes: 34488399fa08 ("mm/madvise: add file and shmem support to MADV_COLLAPSE")
> Reported-by: Chris Mason <mason@kernel.org>
> Cc: stable@vger.kernel.org
> Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>

Acked-by: Hugh Dickins <hughd@google.com>

Valuable catch, thanks. I was worried that "try_collapse_pte_mapped_thp"
implies there's already a PTE mapping the non-uptodate folio, which
would already be wrong. But we've been a little too free with the
SCAN_PTE_MAPPED_HUGEPAGE result code, so there's no such implication
(and even if there had been a good PTE, I doubt there's enough locking
to stop racing fallocates putting a non-uptodate folio there meanwhile).

I was going to quibble with the Fixes, thinking the problem was there
earlier: but no, I agree with you, that's when install_pmd came in.

> ---
> Once reviewed, I will pick this up in a timely manner.
> ---
>  mm/khugepaged.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 8d9f05a39afb3..db135b3c44ef1 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -1900,6 +1900,12 @@ static enum scan_result try_collapse_pte_mapped_thp(struct mm_struct *mm, unsign
>  		goto drop_folio;
>  	}
>  
> +	/* Leave any uninitialized folios to the page fault path. */
> +	if (!folio_test_uptodate(folio)) {
> +		result = SCAN_FAIL;
> +		goto drop_folio;
> +	}
> +
>  	result = find_pmd_or_thp_or_none(mm, haddr, &pmd);
>  	switch (result) {
>  	case SCAN_SUCCEED:
> 
> ---
> 
> base-commit: f44fd601f72e09e44c667b9234c9cd469856679c
> 
> change-id: 20261105-set_huge_pmd_uptodate-9b9820307f30
> 
> --
> 
> Cheers,
> 
> David

Re: [PATCH] mm/khugepaged: don't map folios that are not uptodate in MADV_COLLAPSE

[Zi Yan]

On Thu Nov 5, 2026 at 1:16 PM EST, David Hildenbrand (Arm) wrote:
> madvise(MADV_COLLAPSE) can map PMD-sized pagecache folios that are not
> uptodate. For example, shmem fallocate() can put such folios into the
> page cache.
>
> collapse_file() already refuses to proceed with folios that are not
> uptodate. Let's just do the same in try_collapse_pte_mapped_thp() and skip
> such folios, leaving handling to the page faults.
>
> Note that collapse_file() effectively clears shmem folios through
> shmem_get_folio(); let's keep this fix simple and simply let the page
> fault path deal with that.
>
> Fixes: 34488399fa08 ("mm/madvise: add file and shmem support to MADV_COLLAPSE")
> Reported-by: Chris Mason <mason@kernel.org>
> Cc: stable@vger.kernel.org
> Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
> ---
> Once reviewed, I will pick this up in a timely manner.
> ---
>  mm/khugepaged.c | 6 ++++++
>  1 file changed, 6 insertions(+)
>
Makes sense.

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

-- 
Best Regards,
Yan, Zi

Re: [PATCH] mm/khugepaged: don't map folios that are not uptodate in MADV_COLLAPSE

[Zach O'Keefe]

On Thu, Nov 5, 2026 at 12:09 PM Zi Yan <ziy@nvidia.com> wrote:
>
> On Thu Nov 5, 2026 at 1:16 PM EST, David Hildenbrand (Arm) wrote:
> > madvise(MADV_COLLAPSE) can map PMD-sized pagecache folios that are not
> > uptodate. For example, shmem fallocate() can put such folios into the
> > page cache.
> >
> > collapse_file() already refuses to proceed with folios that are not
> > uptodate. Let's just do the same in try_collapse_pte_mapped_thp() and skip
> > such folios, leaving handling to the page faults.
> >
> > Note that collapse_file() effectively clears shmem folios through
> > shmem_get_folio(); let's keep this fix simple and simply let the page
> > fault path deal with that.
> >
> > Fixes: 34488399fa08 ("mm/madvise: add file and shmem support to MADV_COLLAPSE")
> > Reported-by: Chris Mason <mason@kernel.org>
> > Cc: stable@vger.kernel.org
> > Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
> > ---
> > Once reviewed, I will pick this up in a timely manner.
> > ---
> >  mm/khugepaged.c | 6 ++++++
> >  1 file changed, 6 insertions(+)
> >
> Makes sense.
>
> Acked-by: Zi Yan <ziy@nvidia.com>

Ah, good catch..

Agree with Hugh that the "pte_mapped" naming is a bit of a misnomer,
though it was how I was thinking of things at the time.

Thanks David / Chris.

Reviewed-by: Zach O'Keefe <zokeefe@google.com>


> --
> Best Regards,
> Yan, Zi
>