Skip to contentSkip to conversation

mm/migrate_device: Clear stale mapping after freeing swapcache

PATCH Superseded Arvind Yadav 5 participants 10 messages Jul 24 – Jul 27

Subsystems (2)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 mm/migrate_device: Clear stale mapping after freeing swapcache

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

v2:
- Refresh the mapping using folio_mapping(), as suggested by Zi Yan.

Fixes df263d9a7dff ("mm/migrate_device: try to handle swapcache pages")
Cc Andrew Morton <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Matthew Brost <[email protected]>
Cc Joshua Hahn <[email protected]>
Cc Zi Yan <[email protected]>
Cc Rakie Kim <[email protected]>
Cc Byungchul Park <[email protected]>
Cc Gregory Price <[email protected]>
Cc Ying Huang <[email protected]>
Cc Alistair Popple <[email protected]>
---
 mm/migrate_device.c | 6 ++++++
 1 file changed, 6 insertions(+)
mm/migrate_device.c +6 -0
index 554754eb26ff..ae895758b72c 100644
@@ -1204,6 +1204,12 @@ static void __migrate_device_pages(unsigned long *src_pfns,
 					src_pfns[i] &= ~MIGRATE_PFN_MIGRATE;
 					goto next;
 				}
+
+				/*
+				 * folio_free_swap() removed the folio from the swap
+				 * cache. Refresh the saved mapping before migration.
+				 */
+				mapping = folio_mapping(folio);
 			}
 		} else if (folio_is_zone_device(newfolio)) {
 			/*
Signature · 1 line
2.43.0
reviewer Reviewed-by on v2 1/1
#1 lore raw
Quoted 27 lines · Arvind Yadav

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

v2:
- Refresh the mapping using folio_mapping(), as suggested by Zi Yan.

Fixes df263d9a7dff ("mm/migrate_device: try to handle swapcache pages")
Cc Andrew Morton <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Matthew Brost <[email protected]>
Cc Joshua Hahn <[email protected]>
Cc Zi Yan <[email protected]>
Cc Rakie Kim <[email protected]>
Cc Byungchul Park <[email protected]>
Cc Gregory Price <[email protected]>
Cc Ying Huang <[email protected]>
Cc Alistair Popple <[email protected]>
Quoted 4 lines
---
 mm/migrate_device.c | 6 ++++++
 1 file changed, 6 insertions(+)

LGTM. Thanks.

Signature · 2 lines
Best Regards,
Yan, Zi
maintainer
#2 lore raw
In reply to Arvind Yadav
Quoted 4 lines · On Fri, 24 Jul 2026 13:57:02 +0530 Arvind Yadav <[email protected]> wrote:

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Quoted 8 lines

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

Thanks. AI review might have found an issue with this. And one possible pre-existing issue in the code which Alistair and Balbir worked on.

	https://sashiko.dev/#/patchset/[email protected]
#3 lore raw
Quoted 15 lines · Andrew Morton
Quoted 4 lines · On Fri, 24 Jul 2026 13:57:02 +0530 Arvind Yadav <[email protected]> wrote:

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Quoted 8 lines

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

Quoted 5 lines

Thanks. AI review might have found an issue with this. And one possible pre-existing issue in the code which Alistair and Balbir worked on.

	https://sashiko.dev/#/patchset/[email protected]

Yeah, this might need another careful look.

Signature · 3 lines
Cheers,

David
reviewer
#4 lore raw
Quoted 16 lines · David Hildenbrand (Arm)
Quoted 15 lines · Andrew Morton
Quoted 12 lines · On Fri, 24 Jul 2026 13:57:02 +0530 Arvind Yadav <[email protected]> wrote:

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

Quoted 7 lines
Quoted 5 lines

Thanks. AI review might have found an issue with this. And one possible pre-existing issue in the code which Alistair and Balbir worked on.

	https://sashiko.dev/#/patchset/[email protected]

Yeah, this might need another careful look.

It seems that the pre-existing issue can be fixed by resetting nr to 1 after split is successful. It should also complete this patch. Something like this:

mm/migrate_device.c +5 -0
index 18d097c388530..4a77b6c86ae4f 100644
@@ -1193,6 +1193,11 @@ static void __migrate_device_pages(unsigned long *src_pfns,
 							 MIGRATE_PFN_COMPOUND);
 					goto next;
 				}
+				/*
+				 * reset nr so that only first after-split folio
+				 * is processed below
+				 */
+				nr = 1;
 			} else if ((src_pfns[i] & MIGRATE_PFN_MIGRATE) &&
 				(dst_pfns[i] & MIGRATE_PFN_COMPOUND) &&
 				!(src_pfns[i] & MIGRATE_PFN_COMPOUND)) {
Signature · 2 lines
Best Regards,
Yan, Zi
#5 lore raw
Quoted 38 lines · Zi Yan
Quoted 16 lines · David Hildenbrand (Arm)
Quoted 15 lines · Andrew Morton
Quoted 12 lines · On Fri, 24 Jul 2026 13:57:02 +0530 Arvind Yadav <[email protected]> wrote:

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

Quoted 7 lines
Quoted 5 lines

Thanks. AI review might have found an issue with this. And one possible pre-existing issue in the code which Alistair and Balbir worked on.

	https://sashiko.dev/#/patchset/[email protected]

Yeah, this might need another careful look.

It seems that the pre-existing issue can be fixed by resetting nr to 1 after split is successful. It should also complete this patch. Something like this:

mm/migrate_device.c +0 -0
index 18d097c388530..4a77b6c86ae4f 100644
@@ -1193,6 +1193,11 @@ static void __migrate_device_pages(unsigned long *src_pfns,
 							 MIGRATE_PFN_COMPOUND);

mm/migrate_device.c static void __migrate_device_pages(unsigned long *src_pfns, · line 1203

Quoted 10 lines
mm/migrate_device.c +5 -0
 					goto next;
 				}
+				/*
+				 * reset nr so that only first after-split folio
+				 * is processed below
+				 */
+				nr = 1;
 			} else if ((src_pfns[i] & MIGRATE_PFN_MIGRATE) &&
 				(dst_pfns[i] & MIGRATE_PFN_COMPOUND) &&
 				!(src_pfns[i] & MIGRATE_PFN_COMPOUND)) {

Hmm.. I don't this error condition possible, migrate_vma_split_unmapped_folio() will VM_WARN_ON non anonymous folios, but the design contract is for anonymous folios only. The enforcement comes from the callers of migrate_vma_pages() and migrate_device_pages(). Also __folio_freeze_and_split_unmapped() checks if the folio has a swapcache and mapping associated with it, prior to split. I think this is a false positive

Balbir

reviewer
#6 lore raw
Quoted 45 lines · Balbir Singh
Quoted 44 lines · Zi Yan
Quoted 16 lines · David Hildenbrand (Arm)
Quoted 15 lines · Andrew Morton
Quoted 12 lines · On Fri, 24 Jul 2026 13:57:02 +0530 Arvind Yadav <[email protected]> wrote:

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

Quoted 7 lines
Quoted 5 lines

Thanks. AI review might have found an issue with this. And one possible pre-existing issue in the code which Alistair and Balbir worked on.

	https://sashiko.dev/#/patchset/[email protected]

Yeah, this might need another careful look.

It seems that the pre-existing issue can be fixed by resetting nr to 1 after split is successful. It should also complete this patch. Something like this:

mm/migrate_device.c +5 -0
index 18d097c388530..4a77b6c86ae4f 100644
@@ -1193,6 +1193,11 @@ static void __migrate_device_pages(unsigned long *src_pfns,
 							 MIGRATE_PFN_COMPOUND);
 					goto next;
 				}
+				/*
+				 * reset nr so that only first after-split folio
+				 * is processed below
+				 */
+				nr = 1;
 			} else if ((src_pfns[i] & MIGRATE_PFN_MIGRATE) &&
Quoted 8 lines

mm/migrate_device.c static void __migrate_device_pages(unsigned long *src_pfns, · line 1203

Quoted 4 lines
mm/migrate_device.c +0 -0
 				(dst_pfns[i] & MIGRATE_PFN_COMPOUND) &&
 				!(src_pfns[i] & MIGRATE_PFN_COMPOUND)) {

Hmm.. I don't this error condition possible, migrate_vma_split_unmapped_folio() will VM_WARN_ON non anonymous folios, but the design contract is for anonymous folios only. The enforcement comes from the callers of migrate_vma_pages() and

folio_test_anon() returns true for an anon folio in swapcache. So migrate_vma_split_unmapped_folio()'s VM_WARN_ON() does not prevent anon folios in swapcache.

Quoted 4 lines

migrate_device_pages(). Also __folio_freeze_and_split_unmapped() checks if the folio has a swapcache and mapping associated with it, prior to split. I think this is a false positive

NULL is passed as mapping to __folio_freeze_and_split_unmapped() by migrate_vma_split_unmapped_folio(), so the VM_WARN_ON_ONCE() there will not warn this.

None of the above arguments is valid.

One thing prevents large anon folios in swapcache from reaching to __migrate_device_pages() is migrate_vma_check_page(). When folio_mapping() is not NULL, extra pin count is only 1 + folio_has_private(), which happens to exclude large anon folios in swapcache.

Regardless, adding nr = 1 here still makes sense, since why should the code below process the old nr pages after split is successful? It might be an optimization for current large anon folio only case, but it is more like an issue in the future.

Signature · 2 lines
Best Regards,
Yan, Zi
#7 lore raw
Quoted 55 lines · Zi Yan
Quoted 45 lines · Balbir Singh
Quoted 44 lines · Zi Yan
Quoted 16 lines · David Hildenbrand (Arm)
Quoted 15 lines · Andrew Morton
Quoted 12 lines · On Fri, 24 Jul 2026 13:57:02 +0530 Arvind Yadav <[email protected]> wrote:

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

Quoted 7 lines
Quoted 5 lines

Thanks. AI review might have found an issue with this. And one possible pre-existing issue in the code which Alistair and Balbir worked on.

	https://sashiko.dev/#/patchset/[email protected]

Yeah, this might need another careful look.

It seems that the pre-existing issue can be fixed by resetting nr to 1 after split is successful. It should also complete this patch. Something like this:

mm/migrate_device.c +5 -0
index 18d097c388530..4a77b6c86ae4f 100644
@@ -1193,6 +1193,11 @@ static void __migrate_device_pages(unsigned long *src_pfns,
 							 MIGRATE_PFN_COMPOUND);
 					goto next;
 				}
+				/*
+				 * reset nr so that only first after-split folio
+				 * is processed below
+				 */
+				nr = 1;
 			} else if ((src_pfns[i] & MIGRATE_PFN_MIGRATE) &&
Quoted 8 lines

mm/migrate_device.c static void __migrate_device_pages(unsigned long *src_pfns, · line 1203

Quoted 4 lines
mm/migrate_device.c +0 -0
 				(dst_pfns[i] & MIGRATE_PFN_COMPOUND) &&
 				!(src_pfns[i] & MIGRATE_PFN_COMPOUND)) {

Hmm.. I don't this error condition possible, migrate_vma_split_unmapped_folio() will VM_WARN_ON non anonymous folios, but the design contract is for anonymous folios only. The enforcement comes from the callers of migrate_vma_pages() and

Quoted 4 lines

folio_test_anon() returns true for an anon folio in swapcache. So migrate_vma_split_unmapped_folio()'s VM_WARN_ON() does not prevent anon folios in swapcache.

Agreed. Rhe warning is in folio_split_unmapped() it only works fornon-anon folios. So it doesn't exclude the swapcache case.

And the !folio_test_anon(folio) || !folio_free_swap(folio) actually handles a swapcache anon folio by dropping the swap, which is precisely where the stale mapping originates. So you're right, my argument there doesn't hold.

Quoted 5 lines
Quoted 4 lines

migrate_device_pages(). Also __folio_freeze_and_split_unmapped() checks if the folio has a swapcache and mapping associated with it, prior to split. I think this is a false positive

Quoted 7 lines

NULL is passed as mapping to __folio_freeze_and_split_unmapped() by migrate_vma_split_unmapped_folio(), so the VM_WARN_ON_ONCE() there will not warn this.

None of the above arguments is valid.

Agreed, mapping isn't passed all the way. The split still resolves the swapcache correctly from the current folio rather than mapping - in __folio_freeze_and_split_unmapped(), ci = swap_cluster_get_and_lock(folio) guarded by folio_test_swapcache(folio), and each after-split folio is fixed up via __swap_cache_replace_folio(ci, ...)

Quoted 6 lines

One thing prevents large anon folios in swapcache from reaching to __migrate_device_pages() is migrate_vma_check_page(). When folio_mapping() is not NULL, extra pin count is only 1 + folio_has_private(), which happens to exclude large anon folios in swapcache.

Yes, agreed - that's the reason why it's safe, and it kicks in during migrate_vma_unmap() before any of the above. So the condition is not reachable today for large folios; it's a false positive in that sense.

Quoted 4 lines

Regardless, adding nr = 1 here still makes sense, since why should the code below process the old nr pages after split is successful? It might be an optimization for current large anon folio only case, but it is more like an issue in the future.

Agreed, and we could add it in as a correctness check. After a successful split the inner loop still runs folio_migrate_mapping(mapping, ...) over all nr after-split folios, but folio_free_swap() was only called on the first one - so folios [i+1 .. i+nr) would be migrated as swapcache folios against a single cached mapping.

Thanks for the correction!
Balbir

#8 lore raw
In reply to Zi Yan
Quoted 70 lines · Zi Yan
Quoted 41 lines · Balbir Singh
Quoted 40 lines · Zi Yan
Quoted 15 lines · David Hildenbrand (Arm)
Quoted 14 lines · Andrew Morton
Quoted 12 lines · On Fri, 24 Jul 2026 13:57:02 +0530 Arvind Yadav <[email protected]> wrote:

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

Quoted 6 lines
Quoted 5 lines

Thanks. AI review might have found an issue with this. And one possible pre-existing issue in the code which Alistair and Balbir worked on.

	https://sashiko.dev/#/patchset/[email protected]

Yeah, this might need another careful look.

It seems that the pre-existing issue can be fixed by resetting nr to 1 after split is successful. It should also complete this patch. Something like this:

mm/migrate_device.c +5 -0
index 18d097c388530..4a77b6c86ae4f 100644
@@ -1193,6 +1193,11 @@ static void __migrate_device_pages(unsigned long *src_pfns,
  							 MIGRATE_PFN_COMPOUND);
  					goto next;
  				}
+				/*
+				 * reset nr so that only first after-split folio
+				 * is processed below
+				 */
+				nr = 1;
Quoted 8 lines

mm/migrate_device.c static void __migrate_device_pages(unsigned long *src_pfns, · line 1203

Quoted 5 lines
mm/migrate_device.c +0 -0
  			} else if ((src_pfns[i] & MIGRATE_PFN_MIGRATE) &&
  				(dst_pfns[i] & MIGRATE_PFN_COMPOUND) &&
  				!(src_pfns[i] & MIGRATE_PFN_COMPOUND)) {

Hmm.. I don't this error condition possible, migrate_vma_split_unmapped_folio() will VM_WARN_ON non anonymous folios, but the design contract is for anonymous folios only. The enforcement comes from the callers of migrate_vma_pages() and

folio_test_anon() returns true for an anon folio in swapcache. So migrate_vma_split_unmapped_folio()'s VM_WARN_ON() does not prevent anon folios in swapcache.

Quoted 4 lines

migrate_device_pages(). Also __folio_freeze_and_split_unmapped() checks if the folio has a swapcache and mapping associated with it, prior to split. I think this is a false positive

NULL is passed as mapping to __folio_freeze_and_split_unmapped() by migrate_vma_split_unmapped_folio(), so the VM_WARN_ON_ONCE() there will not warn this.

None of the above arguments is valid.

One thing prevents large anon folios in swapcache from reaching to __migrate_device_pages() is migrate_vma_check_page(). When folio_mapping() is not NULL, extra pin count is only 1 + folio_has_private(), which happens to exclude large anon folios in swapcache.

Quoted 4 lines

Regardless, adding nr = 1 here still makes sense, since why should the code below process the old nr pages after split is successful? It might be an optimization for current large anon folio only case, but it is more like an issue in the future.

Thanks Zi Yan, this makes sense.
After the split, each page is a separate order-0 folio so setting nr = 1 lets each folio go through folio_free_swap() and mapping lookup independently. I will fold this into v3.

Regards,
Arvind

Quoted 2 lines
Reviewed-by on v2 1/1
#9 lore raw
In reply to Arvind Yadav
Quoted 38 lines · Arvind Yadav

__migrate_device_pages() reads the folio mapping before calling folio_free_swap(). When folio_free_swap() succeeds, the folio is removed from the swap cache, but the saved mapping still points to swap_space.

Passing the stale mapping to folio_migrate_mapping() makes it take the mapped-folio path after the swapcache reference has been dropped. This can cause an invalid swap_space lock access followed by a folio reference count BUG.

Refresh the saved mapping after folio_free_swap() so the current folio state is used during migration.

v2:
- Refresh the mapping using folio_mapping(), as suggested by Zi Yan.

Fixes df263d9a7dff ("mm/migrate_device: try to handle swapcache pages")
Cc Andrew Morton <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Matthew Brost <[email protected]>
Cc Joshua Hahn <[email protected]>
Cc Zi Yan <[email protected]>
Cc Rakie Kim <[email protected]>
Cc Byungchul Park <[email protected]>
Cc Gregory Price <[email protected]>
Cc Ying Huang <[email protected]>
Cc Alistair Popple <[email protected]>
---
 mm/migrate_device.c | 6 ++++++
 1 file changed, 6 insertions(+)
mm/migrate_device.c +0 -0
index 554754eb26ff..ae895758b72c 100644
@@ -1204,6 +1204,12 @@ static void __migrate_device_pages(unsigned long *src_pfns,
 					src_pfns[i] &= ~MIGRATE_PFN_MIGRATE;
 					goto next;

mm/migrate_device.c static void __migrate_device_pages(unsigned long *src_pfns, · line 1215

Quoted 10 lines
mm/migrate_device.c +6 -0
 				}
+
+				/*
+				 * folio_free_swap() removed the folio from the swap
+				 * cache. Refresh the saved mapping before migration.
+				 */
+				mapping = folio_mapping(folio);
 			}
 		} else if (folio_is_zone_device(newfolio)) {
 			/*

LGTM