[v1] mm/huge_memory: fix override of entry in remove_migration_pmd

[PATCH] mm/huge_memory: fix override of entry in remove_migration_pmd

Posted by Balbir Singh 3 weeks, 3 days ago

The softleaf changes exposed a BUG in remove_rmap_pmd() where the
migration entry was being overridden when the folio is device private.

Use scope local entry for creating the device private pmde. Make the
pmde writable if the migration entry is writable by moving the check
softleaf_is_migration_write() prior to creating the device private entry.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Joshua Hahn <joshua.hahnjy@gmail.com>
Cc: Rakie Kim <rakie.kim@sk.com>
Cc: Byungchul Park <byungchul@sk.com>
Cc: Gregory Price <gourry@gourry.net>
Cc: Ying Huang <ying.huang@linux.alibaba.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: "Liam R. Howlett" <Liam.Howlett@oracle.com>
Cc: Nico Pache <npache@redhat.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: Dev Jain <dev.jain@arm.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Lyude Paul <lyude@redhat.com>
Cc: Danilo Krummrich <dakr@kernel.org>
Cc: David Airlie <airlied@gmail.com>
Cc: Simona Vetter <simona@ffwll.ch>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Mika Penttilä <mpenttil@redhat.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Francois Dugast <francois.dugast@intel.com>

Signed-off-by: Balbir Singh <balbirs@nvidia.com>
---
 mm/huge_memory.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 0184cd915f44..e4dadcb9e90b 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -4850,7 +4850,12 @@ void remove_migration_pmd(struct page_vma_mapped_walk *pvmw, struct page *new)
 	folio_get(folio);
 	pmde = folio_mk_pmd(folio, READ_ONCE(vma->vm_page_prot));

+	if (softleaf_is_migration_write(entry))
+		pmde = pmd_mkwrite(pmde, vma);
+
 	if (folio_is_device_private(folio)) {
+		swp_entry_t entry;
+
 		if (pmd_write(pmde))
 			entry = make_writable_device_private_entry(
 							page_to_pfn(new));
@@ -4862,8 +4867,6 @@ void remove_migration_pmd(struct page_vma_mapped_walk *pvmw, struct page *new)

 	if (pmd_swp_soft_dirty(*pvmw->pmd))
 		pmde = pmd_mksoft_dirty(pmde);
-	if (softleaf_is_migration_write(entry))
-		pmde = pmd_mkwrite(pmde, vma);
 	if (pmd_swp_uffd_wp(*pvmw->pmd))
 		pmde = pmd_mkuffd_wp(pmde);
 	if (!softleaf_is_migration_young(entry))
-- 
2.51.1

Re: [PATCH] mm/huge_memory: fix override of entry in remove_migration_pmd

Posted by Lorenzo Stoakes 3 weeks, 3 days ago

On Thu, Nov 13, 2025 at 04:13:52PM +1100, Balbir Singh wrote:
> The softleaf changes exposed a BUG in remove_rmap_pmd() where the
> migration entry was being overridden when the folio is device private.
>
> Use scope local entry for creating the device private pmde. Make the
> pmde writable if the migration entry is writable by moving the check
> softleaf_is_migration_write() prior to creating the device private entry.
>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: David Hildenbrand <david@redhat.com>
> Cc: Zi Yan <ziy@nvidia.com>
> Cc: Joshua Hahn <joshua.hahnjy@gmail.com>
> Cc: Rakie Kim <rakie.kim@sk.com>
> Cc: Byungchul Park <byungchul@sk.com>
> Cc: Gregory Price <gourry@gourry.net>
> Cc: Ying Huang <ying.huang@linux.alibaba.com>
> Cc: Alistair Popple <apopple@nvidia.com>
> Cc: Oscar Salvador <osalvador@suse.de>
> Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
> Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
> Cc: "Liam R. Howlett" <Liam.Howlett@oracle.com>
> Cc: Nico Pache <npache@redhat.com>
> Cc: Ryan Roberts <ryan.roberts@arm.com>
> Cc: Dev Jain <dev.jain@arm.com>
> Cc: Barry Song <baohua@kernel.org>
> Cc: Lyude Paul <lyude@redhat.com>
> Cc: Danilo Krummrich <dakr@kernel.org>
> Cc: David Airlie <airlied@gmail.com>
> Cc: Simona Vetter <simona@ffwll.ch>
> Cc: Ralph Campbell <rcampbell@nvidia.com>
> Cc: Mika Penttilä <mpenttil@redhat.com>
> Cc: Matthew Brost <matthew.brost@intel.com>
> Cc: Francois Dugast <francois.dugast@intel.com>
>
> Signed-off-by: Balbir Singh <balbirs@nvidia.com>

THe logic LGTM but we don't want to have a bisect hazard here by having the bug
introduced earlier then resolved here.

I think better would be to do the fix up in the original series as a fix-patch,
then maybe point at this one as a conflict resolution for my series?

In any case, LGTM so:

Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>

> ---
>  mm/huge_memory.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 0184cd915f44..e4dadcb9e90b 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -4850,7 +4850,12 @@ void remove_migration_pmd(struct page_vma_mapped_walk *pvmw, struct page *new)
>  	folio_get(folio);
>  	pmde = folio_mk_pmd(folio, READ_ONCE(vma->vm_page_prot));
>
> +	if (softleaf_is_migration_write(entry))
> +		pmde = pmd_mkwrite(pmde, vma);
> +
>  	if (folio_is_device_private(folio)) {
> +		swp_entry_t entry;
> +
>  		if (pmd_write(pmde))
>  			entry = make_writable_device_private_entry(
>  							page_to_pfn(new));
> @@ -4862,8 +4867,6 @@ void remove_migration_pmd(struct page_vma_mapped_walk *pvmw, struct page *new)
>
>  	if (pmd_swp_soft_dirty(*pvmw->pmd))
>  		pmde = pmd_mksoft_dirty(pmde);
> -	if (softleaf_is_migration_write(entry))
> -		pmde = pmd_mkwrite(pmde, vma);
>  	if (pmd_swp_uffd_wp(*pvmw->pmd))
>  		pmde = pmd_mkuffd_wp(pmde);
>  	if (!softleaf_is_migration_young(entry))
> --
> 2.51.1
>

Re: [PATCH] mm/huge_memory: fix override of entry in remove_migration_pmd

Posted by David Hildenbrand (Red Hat) 3 weeks, 3 days ago

On 13.11.25 12:56, Lorenzo Stoakes wrote:
> On Thu, Nov 13, 2025 at 04:13:52PM +1100, Balbir Singh wrote:
>> The softleaf changes exposed a BUG in remove_rmap_pmd() where the
>> migration entry was being overridden when the folio is device private.
>>
>> Use scope local entry for creating the device private pmde. Make the
>> pmde writable if the migration entry is writable by moving the check
>> softleaf_is_migration_write() prior to creating the device private entry.
>>
>> Cc: Andrew Morton <akpm@linux-foundation.org>
>> Cc: David Hildenbrand <david@redhat.com>
>> Cc: Zi Yan <ziy@nvidia.com>
>> Cc: Joshua Hahn <joshua.hahnjy@gmail.com>
>> Cc: Rakie Kim <rakie.kim@sk.com>
>> Cc: Byungchul Park <byungchul@sk.com>
>> Cc: Gregory Price <gourry@gourry.net>
>> Cc: Ying Huang <ying.huang@linux.alibaba.com>
>> Cc: Alistair Popple <apopple@nvidia.com>
>> Cc: Oscar Salvador <osalvador@suse.de>
>> Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
>> Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
>> Cc: "Liam R. Howlett" <Liam.Howlett@oracle.com>
>> Cc: Nico Pache <npache@redhat.com>
>> Cc: Ryan Roberts <ryan.roberts@arm.com>
>> Cc: Dev Jain <dev.jain@arm.com>
>> Cc: Barry Song <baohua@kernel.org>
>> Cc: Lyude Paul <lyude@redhat.com>
>> Cc: Danilo Krummrich <dakr@kernel.org>
>> Cc: David Airlie <airlied@gmail.com>
>> Cc: Simona Vetter <simona@ffwll.ch>
>> Cc: Ralph Campbell <rcampbell@nvidia.com>
>> Cc: Mika Penttilä <mpenttil@redhat.com>
>> Cc: Matthew Brost <matthew.brost@intel.com>
>> Cc: Francois Dugast <francois.dugast@intel.com>
>>
>> Signed-off-by: Balbir Singh <balbirs@nvidia.com>
> 
> THe logic LGTM but we don't want to have a bisect hazard here by having the bug
> introduced earlier then resolved here.

Exactly.

-- 
Cheers

David

Re: [PATCH] mm/huge_memory: fix override of entry in remove_migration_pmd

Posted by Balbir Singh 3 weeks, 3 days ago

On 11/13/25 23:07, David Hildenbrand (Red Hat) wrote:
> On 13.11.25 12:56, Lorenzo Stoakes wrote:
>> On Thu, Nov 13, 2025 at 04:13:52PM +1100, Balbir Singh wrote:
>>> The softleaf changes exposed a BUG in remove_rmap_pmd() where the
>>> migration entry was being overridden when the folio is device private.
>>>
>>> Use scope local entry for creating the device private pmde. Make the
>>> pmde writable if the migration entry is writable by moving the check
>>> softleaf_is_migration_write() prior to creating the device private entry.
>>>
>>> Cc: Andrew Morton <akpm@linux-foundation.org>
>>> Cc: David Hildenbrand <david@redhat.com>
>>> Cc: Zi Yan <ziy@nvidia.com>
>>> Cc: Joshua Hahn <joshua.hahnjy@gmail.com>
>>> Cc: Rakie Kim <rakie.kim@sk.com>
>>> Cc: Byungchul Park <byungchul@sk.com>
>>> Cc: Gregory Price <gourry@gourry.net>
>>> Cc: Ying Huang <ying.huang@linux.alibaba.com>
>>> Cc: Alistair Popple <apopple@nvidia.com>
>>> Cc: Oscar Salvador <osalvador@suse.de>
>>> Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
>>> Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
>>> Cc: "Liam R. Howlett" <Liam.Howlett@oracle.com>
>>> Cc: Nico Pache <npache@redhat.com>
>>> Cc: Ryan Roberts <ryan.roberts@arm.com>
>>> Cc: Dev Jain <dev.jain@arm.com>
>>> Cc: Barry Song <baohua@kernel.org>
>>> Cc: Lyude Paul <lyude@redhat.com>
>>> Cc: Danilo Krummrich <dakr@kernel.org>
>>> Cc: David Airlie <airlied@gmail.com>
>>> Cc: Simona Vetter <simona@ffwll.ch>
>>> Cc: Ralph Campbell <rcampbell@nvidia.com>
>>> Cc: Mika Penttilä <mpenttil@redhat.com>
>>> Cc: Matthew Brost <matthew.brost@intel.com>
>>> Cc: Francois Dugast <francois.dugast@intel.com>
>>>
>>> Signed-off-by: Balbir Singh <balbirs@nvidia.com>
>>
>> THe logic LGTM but we don't want to have a bisect hazard here by having the bug
>> introduced earlier then resolved here.
> 
> Exactly.
> 

The hazard is a VM_WARN_ON() that checks for is_migration_entry() in the
call to softleaf_is_migration_young(). I am happy to go down that path,
let me send out the fixups, it felt weird to break the rebase, but I
am sure Andrew has a better way of dealing with this stuff.

Balbir

Re: [PATCH] mm/huge_memory: fix override of entry in remove_migration_pmd

Posted by Andrew Morton 3 weeks, 3 days ago

On Fri, 14 Nov 2025 07:55:29 +1100 Balbir Singh <balbirs@nvidia.com> wrote:

> >> THe logic LGTM but we don't want to have a bisect hazard here by having the bug
> >> introduced earlier then resolved here.
> > 
> > Exactly.
> > 
> 
> The hazard is a VM_WARN_ON() that checks for is_migration_entry() in the
> call to softleaf_is_migration_young(). I am happy to go down that path,
> let me send out the fixups, it felt weird to break the rebase, but I
> am sure Andrew has a better way of dealing with this stuff.

Yes, please tell us which patch introduced the issue then send a fix
for that patch.  We can figure out the softleaf changes later.  (Which
this patch provides anyway).

Also, confusing that the changelog says "The softleaf changes exposed a
BUG in remove_rmap_pmd()" but it's remove_migration_pmd() that gets
altered.  Please expand and clarify?

Re: [PATCH] mm/huge_memory: fix override of entry in remove_migration_pmd

Posted by Balbir Singh 3 weeks, 3 days ago

On 11/14/25 10:08, Andrew Morton wrote:
> On Fri, 14 Nov 2025 07:55:29 +1100 Balbir Singh <balbirs@nvidia.com> wrote:
> 
>>>> THe logic LGTM but we don't want to have a bisect hazard here by having the bug
>>>> introduced earlier then resolved here.
>>>
>>> Exactly.
>>>
>>
>> The hazard is a VM_WARN_ON() that checks for is_migration_entry() in the
>> call to softleaf_is_migration_young(). I am happy to go down that path,
>> let me send out the fixups, it felt weird to break the rebase, but I
>> am sure Andrew has a better way of dealing with this stuff.
> 
> Yes, please tell us which patch introduced the issue then send a fix
> for that patch.  We can figure out the softleaf changes later.  (Which
> this patch provides anyway).
> 
> Also, confusing that the changelog says "The softleaf changes exposed a
> BUG in remove_rmap_pmd()" but it's remove_migration_pmd() that gets
> altered.  Please expand and clarify?

Sorry, it's remove_migration_pmd() that is effected. I've updated
the changelog and sent out the patches. I've sent the fix to the
original patch and the fixup along with it, hopefully I got it
right :)

Balbir