[PATCH v6 00/39] mm: make VMA flag semantics explicit, eliminate VM_SPECIAL

[PATCH v6 00/39] mm: make VMA flag semantics explicit, eliminate VM_SPECIAL

[Lorenzo Stoakes (ARM)]

The VM_SPECIAL / VMA_SPECIAL_FLAGS mask conflates several unrelated
properties:

* Is this managed by core mm, or by a driver mapping MMIO, kernel-allocated
  pages, or even ordinary pages itself?
* Can it be expanded or merged?
* Is this a 'weird' case like mlock where migration might race and we
  'have' to set invalid flags to notify?
* Is it another 'weird' case where we just want to stop GUP from touching
  it?

Driver writers have often been confused about this, and who can blame them?

It also interacts badly with the eternal edgecase known as hugetlb - which
sets VMA_DONTEXPAND_BIT but doesn't also want to be treated like a
'special' flag.

Another issue is that we cannot make sensible assumptions about flag
use. It's not possible to assume VMA_IO_BIT means iommu because drivers
abuse it and mlock abuses it.

Special is also an overloaded term in mm. VDSO and VVAR mappings are also
called 'special' but they're special in a... special way.

Sometimes things are called special that are a subset of
VMA_SPECIAL_FLAGS (VMA_PFNMAP_BIT and VMA_MIXEDMAP_BIT for instance when it
comes to zapping or vm_normal_folio()).

There's a specific kind of special for THP too, which considers
PFN map, mixed map 'special' but DAX not.

It's all rather a mess.

This series brings some order to things by both limiting what drivers can
do with VMA flags and switching to using predicates that describe
behaviour, not arbitrary flags.

It establishes the invariant that mm-managed mappings may neither set
VMA_IO_BIT nor clear VMA_MAYWRITE_BIT in an mmap hook, enforcing this by
validating VMA state after every mmap and mmap_prepare hook.

It updates usbmon and sg to mmap_prepare in order to do so, adding a new
mmap action for mapping discontiguous kernel pages, and has hfi1 and the
ALSA PCM status page map their pages eagerly instead.

It also establishes the invariant that VMA_MIXEDMAP_BIT be set when mapping
kernel memory, something that is usually the case but happens not to be for
some users - specifically defio, cmt_speech, uprobes and the bpf arena, all
of which are updated to do the right thing.

It replaces VM_SPECIAL and arbitrary flag tests with predicates that say
what is actually being tested:

     vma_is_mm_managed()   Are the VMA's contents established and managed
                           by core mm, rather than a driver or other kernel
                           component?

      vma_is_mm_backed()   Is the VMA backed by memory core mm itself
                           faults in, rather than a driver or device?

  vma_is_fixed_mapping()   Is the VMA not permitted to be expanded or
                           merged?

         vma_can_merge()   Can the VMA be merged with a compatible
                           neighbour?

Remaining raw VMA_IO_BIT, VMA_PFNMAP_BIT and VMA_MIXEDMAP_BIT tests
scattered across mm are converted to predicates where they express one, and
left as explicit flag tests where they do not.

And also the opportunity is taken to eliminate THP's vma_is_special_huge()
which was an existing source of confusion.

Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
v6:
* Applied tags (thanks David!)
* Added 7/39 to split out the vm_insert_page() / vm_insert_pages()
  mmap lock assertions and VMA API changes out into their own patch,
  as per David.
* Updated 8/39 (was 7/38) to collapse the discontig single page case to
  a single one and to rename things, as per David.
* Updated 9/39 (was 8/38) to update the discontig docs to reflect the
  change.
* Updated 23/39 (was 22/38) to put expression on single line, as per David.
* Updated 25/39 (was 24/38) to mention the open-coding in the commit
  message (fixed in next commit), as per discussion with David.
* Updated 26/39 (was 25/38) to clarify the commit message around the
  removal of the VMA_IO_BIT check in vma_flags_is_mm_managed(), as per
  David.
* Updated 30/39 (was 30/38) to no longer exclude droppable mappings from
  vma_flags_is_mm_backed(), as it doesn't quite fit into the broader
  definition, as per David.
* Updated 30/39 (was 30/38) to express vma_flags_can_merge() in terms of
  vma_flags_is_mm_backed() (with an explicit hugetlb exclusion), and to
  no longer use vma_can_merge() in __thp_vma_allowable_orders().
* Updated 31/39 (was 29/38) to use vma_is_mm_backed() (with hugetlb
  exclusion) rather than vma_can_merge(), as guard regions have nothing to
  do with mergeability.
* Updated 32/39 (was 31/38) to retain an explicit droppable check in
  vma_can_userfault(), as per David.
* Updated 33/39 (was 32/38) to use vma_is_mm_backed() with an explicit
  hugetlb exclusion, rather than vma_can_merge() which makes no sense here.
* Updated 34/39 (was 33/38) to retain explicit droppable checks in
  vma_supports_mlock() and ksm_compatible(), as per David.
* Updated 34/39 (was 33/38) to use vma_is_mm_backed() in should_skip_vma(),
  based on discussion with David.
* Updated 35/39 (was 34/38) to use vma_is_mm_backed() in
  lru_gen_look_around(), as per David.
* Updated 36/39 (was 35/38) to use vma_is_mm_backed() with an explicit
  hugetlb exclusion in migrate_vma_setup(), as per David.

v5:
* Applied tags (thanks everybody!)
* Updated 3/38 to fix a bug in __mmap_new_file_vma() by zapping + closing
  the VMA before setting vma->vm_file NULL to avoid an i_mmap_writable
  imbalance. A repro was tested before/after and confirmed fixed.
* Updated 7/38 to make __map_kernel_pages_prepare() use
  mmap_assert_write_locked() rather than the weird
  assert-on-mmap_read_trylock() pattern, as per Suren.
* Also update vm_insert_page(), vm_insert_pages() to fix the same pattern
  and to use the new VMA API, and converted a BUG_ON to a
  VM_WARN_ON_ONCE().
https://lore.kernel.org/r/20261009-b4-mmap-prepare-vma-flag-sanify-v5-0-923b70125b1e@kernel.org

v4:
* Pulled what was 2/40 out of the series to send as a fix instead, as per
  Suren.
* Updated 1/38 to rename put_map() to map_put_vm_file(), as per Suren and
  Liam.
* Updated 3/38 (was 4/40) to perform the vma_close() after unmapping the
  range as per Gregory.
* Updated 3/38 (was 4/40) to rename prev_* to orig_* in the mmap hook
  validation code, as per Suren.
* Updated 4/38 (was 5/40) to clarify that the mergeable-VMA action check
  runs after mmap_action_prepare(), as per Zi.
* Updated 7/38 (was 8/40) to have discontig_kernel_map_page_range() map
  only the pages which fit in the VMA, consistent with
  discontig_kernel_map_page(), as per Suren and Zi.
* Updated 7/38 (was 8/40) to unmap and close the VMA when an mmap action
  fails under the mmap_prepare compatibility layer, so state is correctly
  released there too, as per Lance.
* Updated 8/38 (was 9/40) to document this, as per Zi.
* Updated 14/38 (was 15/40) to replace vma[_flags]_is_kernel_owned() with
  its inverse, vma[_flags]_is_mm_managed() and reword accordingly
  throughout, as per David.
* Updated 22/38 (was 23/40) to add a clarifying comment, as per Zi.
* Updated 23/38 (was 24/40) to test VMA_LOCKED_MASK before draining the
  mlock batch in try_to_unmap_one() and try_to_migrate_one(), as per Lance.
* Updated 30/38 (was 31/40) to rename vma[_flags]_is_persistent() to
  vma[_flags]_is_mm_backed() and reword its documentation around what backs
  the mapping rather than persistence, as per David.
* Updated 31/38 (was 32/40) to make the code more succinct in
  vma_move_compatible(), as per David.
* Dropped what was 40/40, vma[_flags]_can_gup(), as per David.
https://lore.kernel.org/r/20261003-b4-mmap-prepare-vma-flag-sanify-v4-0-a1f052500fd7@kernel.org

v3:
* Fixed up bug in patch 1 as reported by Mike - have to delay setting
  map->vma_flags until after action prepare, though map->vm_file needs to
  be set before for correct reference count management.
* Updated 4/40 to add a symmetric vm_end check as well as vm_start in case
  of a dangerously insane driver, as per Sashiko.
* Updated 8/40 to check if a driver did something REALLY stupid like having
  a NULL discontig_kernel_page_ops ptr, as per Sashiko.
* Updated 15/40 to trivially synchronise userland test comments.
* Updated 17/40 to correctly duplicate code to the userland VMA tests as
  per Sashiko.
https://lore.kernel.org/r/20260917-b4-mmap-prepare-vma-flag-sanify-v3-0-4583d8a23bca@kernel.org

v2:
* Rebased on mm-unstable.
* Introduced new patch to fix various mmap_prepare and file interactions
  that weren't quite right as per Sashiko. None impact anything upstream
  yet so it doesn't need to be a fix.
* Restore vma->vm_start if an mmap hook has moved it before tearing the
  VMA down, so we unmap the range we established rather than the one the
  hook invented, as per Sashiko.
* Reject a discontiguous kernel page batch of zero pages rather than
  looping forever, and bound batches by the pages remaining in the VMA,
  as per Sashiko.
* Add the missing map_kernel_discontig member to the userland VMA tests'
  copy of struct mmap_action as per Sashiko.
* Set VM_DONTEXPAND on hfi1's RCV_HDRQ, RCV_EGRBUF and RTAIL mappings, as
  dma_mmap_coherent() doesn't on the IOMMU-DMA path, as per Sashiko.
* Recompute vma->vm_page_prot in snd_pcm_mmap_status() after clearing
  VM_WRITE, as vm_insert_page() uses it immediately rather than at fault
  time, as per Sashiko.
* munlock_vma_folio() now tests VMA_LOCKED_MASK so an unmap racing the
  mlock walk still munlocks folios the walk has already counted, as per
  Sashiko.
https://lore.kernel.org/r/20260914-b4-mmap-prepare-vma-flag-sanify-v2-0-7d9781ed5361@kernel.org

v1:
https://lore.kernel.org/r/20260908-b4-mmap-prepare-vma-flag-sanify-v1-0-dacf19cce22b@kernel.org

---
Lorenzo Stoakes (ARM) (39):
      mm/vma: fix mmap_prepare file handling, remove file_doesnt_need_get
      mm/vma: introduce and use vma_[flags_]can_merge()
      mm: consistently validate VMA state after mmap[_prepare] hooks
      mm/vma: ensure mmap_prepare doesn't set actions on a mergeable vma
      mm: make map_kernel_pages_[prepare,complete] internal and unexported
      mm/vma: tidy up map kernel pages enum values
      mm: assert mmap lock held, use new flags API in vm_insert_page[s]
      mm: add mmap action for discontiguous kernel page mapping
      docs: filesystems: update mmap_prepare docs for discontig kernel pgs
      drivers/usb/mon: update to use mmap_prepare + map kernel pages
      infiniband: update hfi1 to use remap_vmalloc_range()
      selinux: reject writable opens of policy file, drop mmap shared/write check
      ALSA: pcm: use vm_insert_page() to map PCM status page
      bpf: arena: mark arena_map_mmap() mappings VM_MIXEDMAP
      mm/vma: add vma[_flags]_is_mm_managed() predicates
      mm/vma: only allow mmap to clear VMA_MAYWRITE_BIT if not mm-managed
      mm/vma: add and use vma_[flags]_is_fixed_mapping
      scsi: sg: convert mmap hook to mmap_prepare and rework
      fbdev: defio: assert FBINFO_VIRTFB, drop VM_IO, add VM_MIXEDMAP
      HSI: cmt_speech: convert mmap hook to mmap_prepare, refactor
      mm/gup: error out early on !VMA_MAYREAD_BIT VMAs
      uprobes: remove VM_IO, set VM_MIXEDMAP for mapped kernel pages
      mm/mlock: clear VMA_LOCKED_MASK over mmap callback
      mm/mlock: eliminate weird VMA_IO_BIT abuse and simplify
      mm/vma: enforce that mm-managed mappings may not set VMA_IO_BIT
      mm: remove VMA_IO_BIT check in vma[_flags]_is_mm_managed()
      mm: remove hugetlb_inline.h
      mm: rename is_vm_hugetlb_page() to vma_is_hugetlb()
      mm: drop some redundant checks around hugetlb VMAs
      mm/vma: introduce vma[_flags]_is_mm_backed()
      mm/madvise: update is_valid_guard_vma() to use vma_is_mm_backed()
      mm/uffd: use predicates for userfaultfd checks
      mm/madvise: use predicates for madvise(..., MADV_DOFORK)
      mm: eliminate VMA_SPECIAL_FLAGS usage when hugetlb explicitly tested
      mm: eliminate VMA_SPECIAL_FLAGS check in lru_gen_look_around()
      mm: avoid use of VMA_SPECIAL_FLAGS in migrate_vma_setup()
      mm: eliminate VM_SPECIAL, VMA_SPECIAL_FLAGS
      fuse: dax: do not set VM_MIXEDMAP
      mm/huge_memory: remove vma_is_special_huge()

 Documentation/filesystems/mmap_prepare.rst |  81 ++++++++++
 arch/arm64/kvm/mmu.c                       |   4 +-
 arch/powerpc/mm/book3s64/radix_tlb.c       |   6 +-
 arch/powerpc/mm/nohash/e500_hugetlbpage.c  |   2 +-
 arch/powerpc/mm/nohash/tlb.c               |   2 +-
 arch/riscv/kvm/mmu.c                       |   2 +-
 arch/riscv/mm/tlbflush.c                   |   2 +-
 arch/s390/mm/gmap_helpers.c                |   6 +-
 arch/sparc/mm/init_64.c                    |   2 +-
 arch/x86/kernel/uprobes.c                  |   2 +-
 drivers/gpu/drm/drm_gpusvm.c               |   5 +-
 drivers/hsi/clients/cmt_speech.c           |  33 ++--
 drivers/infiniband/hw/hfi1/file_ops.c      |  83 +++-------
 drivers/scsi/sg.c                          | 115 ++++++-------
 drivers/usb/mon/mon_bin.c                  |  82 ++++++----
 drivers/video/fbdev/core/fb_defio.c        |   6 +-
 drivers/video/fbdev/ssd1307fb.c            |   2 +
 fs/coredump.c                              |   2 +-
 fs/fuse/dax.c                              |   2 +-
 fs/hugetlbfs/inode.c                       |   2 +-
 fs/proc/task_mmu.c                         |   8 +-
 include/asm-generic/tlb.h                  |   4 +-
 include/linux/hugetlb.h                    |   5 +-
 include/linux/hugetlb_inline.h             |  28 ----
 include/linux/mm.h                         | 236 +++++++++++++++++++++++++--
 include/linux/mm_types.h                   |  48 +++++-
 include/linux/pagemap.h                    |   1 -
 include/linux/rmap.h                       |   2 +-
 include/linux/userfaultfd_k.h              |   1 -
 kernel/bpf/arena.c                         |   3 +-
 kernel/events/core.c                       |   2 +-
 kernel/events/uprobes.c                    |   4 +-
 kernel/sched/fair.c                        |   3 +-
 mm/folio.c                                 |   2 +-
 mm/gup.c                                   |   8 +-
 mm/huge_memory.c                           |  30 ++--
 mm/hugetlb.c                               |  14 +-
 mm/internal.h                              |  79 +++++----
 mm/ksm.c                                   |   5 +-
 mm/madvise.c                               |  31 ++--
 mm/memory.c                                | 145 ++++++++++++++---
 mm/mempolicy.c                             |   2 +-
 mm/migrate_device.c                        |  12 +-
 mm/mlock.c                                 |  51 +++---
 mm/mmap.c                                  |   2 +-
 mm/mmu_gather.c                            |   2 +-
 mm/mprotect.c                              |   5 +-
 mm/mremap.c                                |  11 +-
 mm/page_vma_mapped.c                       |   4 +-
 mm/pagewalk.c                              |   2 +-
 mm/rmap.c                                  |   8 +-
 mm/swapfile.c                              |   2 +-
 mm/userfaultfd.c                           |  46 +++---
 mm/util.c                                  |  49 ++++--
 mm/vma.c                                   | 248 ++++++++++++++++++++++-------
 mm/vma.h                                   |  31 +++-
 mm/vma_internal.h                          |   1 -
 mm/vmscan.c                                |   8 +-
 security/selinux/selinuxfs.c               |  11 +-
 sound/core/pcm_native.c                    |  38 ++---
 tools/testing/vma/include/dup.h            | 106 ++++++++++--
 tools/testing/vma/include/stubs.h          |   6 +-
 tools/testing/vma/tests/merge.c            |  10 +-
 63 files changed, 1187 insertions(+), 568 deletions(-)
---
base-commit: 6b41451631cabf9ea3b384c2a099088e1598f963
change-id: 20260721-b4-mmap-prepare-vma-flag-sanify-2100425df5aa

Cheers,
-- 
Lorenzo Stoakes (ARM) <ljs@kernel.org>

[PATCH v6 01/39] mm/vma: fix mmap_prepare file handling, remove file_doesnt_need_get

[Lorenzo Stoakes (ARM)]

The map->file_doesnt_need_get flag is confusing and the existing
implementation has holes.

Drivers are permitted to change the owning file of a mapping. If they do
so, they are required to take a reference on that file.

The mmap() operation which ultimately invokes __mmap_region() is guaranteed
to drop the refcount for the original file the mapping was made under, but
this is not true for the replaced file.

This has been addressed so far by tracking map->file_doesnt_need_get, which
is rather poorly named and unfortunately fails to correctly track whether
or not an additional put were needed in a number of cases.

Make life easier by removing this flag, and instead drop the reference for
both mmap_prepare and the deprecated mmap callback in a new function
map_put_vm_file().

Track whether this needs to be done by aligning mmap_state with
vm_area_desc and store the original file in the map->file field, keeping
the updated file in map->vm_file.

In order to have the same behaviour for both types of hooks, only drop the
reference __mmap_new_file_vma() itself took in its error path, deferring
the replaced file's reference to map_put_vm_file().

To make this work correctly, map->vm_file has to be updated before any
error handling, so update __mmap_new_file_vma() and call_mmap_prepare() to
set this field first.

Also when mmap_prepare() changes the file and is then merged, the reference
count also must be decremented, so update the logic to call
map_put_vm_file() in this case too.

Also update __compat_vma_mmap() to manually perform this step for stacked
file systems using the compatibility layer, and update
compat_set_vma_from_desc() to replace vma_set_file() with a correct
refcount/file update.

No in-tree driver is impacted by the incorrect implementation of this
currently (no driver that does this is mergeable for one), so this does not
need to be a fix.

Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/internal.h |  1 +
 mm/util.c     |  5 +++-
 mm/vma.c      | 83 +++++++++++++++++++++++++++++++++++------------------------
 mm/vma.h      |  6 +++--
 4 files changed, 59 insertions(+), 36 deletions(-)

diff --git a/mm/internal.h b/mm/internal.h
index 0dca33db068f..fe576d468af4 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -7,6 +7,7 @@
 #ifndef __MM_INTERNAL_H
 #define __MM_INTERNAL_H
 
+#include <linux/file.h>
 #include <linux/fs.h>
 #include <linux/khugepaged.h>
 #include <linux/mm.h>
diff --git a/mm/util.c b/mm/util.c
index bf0513d1d3d0..016932780925 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -1228,8 +1228,11 @@ int __compat_vma_mmap(struct vm_area_desc *desc,
 
 	/* Perform any preparatory tasks for mmap action. */
 	err = mmap_action_prepare(desc);
-	if (err)
+	if (err) {
+		if (desc->vm_file != vma->vm_file)
+			fput(desc->vm_file);
 		return err;
+	}
 	/* Update the VMA from the descriptor. */
 	compat_set_vma_from_desc(vma, desc);
 	/* Complete any specified mmap actions. */
diff --git a/mm/vma.c b/mm/vma.c
index 55917d097933..c27aea6d58d9 100644
--- a/mm/vma.c
+++ b/mm/vma.c
@@ -24,7 +24,8 @@ struct mmap_state {
 		vm_flags_t vm_flags;
 		vma_flags_t vma_flags;
 	};
-	struct file *file;
+	struct file *file;	/* mmap()-specified file. */
+	struct file *vm_file;	/* May be updated by mmap_prepare. */
 	pgprot_t page_prot;
 
 	/* User-defined fields, perhaps updated by .mmap_prepare(). */
@@ -43,8 +44,6 @@ struct mmap_state {
 
 	/* Determine if we can check KSM flags early in mmap() logic. */
 	bool check_ksm_early :1;
-	/* If .mmap_prepare changed the file, we don't need to pin. */
-	bool file_doesnt_need_get :1;
 };
 
 #define MMAP_STATE(name, mm_, vmi_, addr_, len_, pgoff_, anon_pgoff_, vma_flags_, file_) \
@@ -58,6 +57,7 @@ struct mmap_state {
 		.pglen = PHYS_PFN(len_),				\
 		.vma_flags = vma_flags_,				\
 		.file = file_,						\
+		.vm_file = file_,					\
 		.page_prot = vma_flags_to_page_prot(vma_flags_),	\
 	}
 
@@ -70,7 +70,7 @@ struct mmap_state {
 		.vma_flags = (map_)->vma_flags,				\
 		.pgoff = (map_)->pgoff,					\
 		.anon_pgoff = (map_)->anon_pgoff,			\
-		.file = (map_)->file,					\
+		.file = (map_)->vm_file,				\
 		.prev = (map_)->prev,					\
 		.middle = vma_,						\
 		.next = (vma_) ? NULL : (map_)->next,			\
@@ -2447,7 +2447,7 @@ void mm_drop_all_locks(struct mm_struct *mm)
  */
 static bool accountable_mapping(struct mmap_state *map)
 {
-	const struct file *file = map->file;
+	const struct file *file = map->vm_file;
 
 	/*
 	 * hugetlb has its own accounting separate from the core VM
@@ -2496,7 +2496,7 @@ static void vms_abort_munmap_vmas(struct vma_munmap_struct *vms,
 
 static void update_ksm_flags(struct mmap_state *map)
 {
-	map->vma_flags = ksm_vma_flags(map->mm, map->file, map->vma_flags);
+	map->vma_flags = ksm_vma_flags(map->mm, map->vm_file, map->vma_flags);
 }
 
 static void set_desc_from_map(struct vm_area_desc *desc,
@@ -2506,7 +2506,7 @@ static void set_desc_from_map(struct vm_area_desc *desc,
 	desc->end = map->end;
 
 	desc->pgoff = map->pgoff;
-	desc->vm_file = map->file;
+	desc->vm_file = map->vm_file;
 	desc->vma_flags = map->vma_flags;
 	desc->page_prot = map->page_prot;
 }
@@ -2586,6 +2586,10 @@ static int __mmap_setup(struct mmap_state *map, struct vm_area_desc *desc,
 	return 0;
 }
 
+static bool map_same_file(struct mmap_state *map)
+{
+	return map->vm_file == map->file;
+}
 
 static int __mmap_new_file_vma(struct mmap_state *map,
 			       struct vm_area_struct *vma)
@@ -2593,20 +2597,23 @@ static int __mmap_new_file_vma(struct mmap_state *map,
 	struct vma_iterator *vmi = map->vmi;
 	int error;
 
-	vma->vm_file = map->file;
-	if (!map->file_doesnt_need_get)
-		get_file(map->file);
+	vma->vm_file = map->vm_file;
+	if (map_same_file(map))
+		get_file(map->vm_file);
 
-	if (!map->file->f_op->mmap)
+	if (!map->vm_file->f_op->mmap)
 		return 0;
 
 	error = mmap_file(vma->vm_file, vma);
+	map->vm_file = vma->vm_file;
+
 	if (error) {
 		UNMAP_STATE(unmap, vmi, vma, vma->vm_start, vma->vm_end,
 			    map->prev, map->next);
-		fput(vma->vm_file);
-		vma->vm_file = NULL;
+		if (map_same_file(map))
+			fput(map->vm_file);
 
+		vma->vm_file = NULL;
 		vma_iter_set(vmi, vma->vm_end);
 		/* Undo any partial mapping done by a device driver. */
 		unmap_region(&unmap);
@@ -2623,7 +2630,6 @@ static int __mmap_new_file_vma(struct mmap_state *map,
 			!vma_flags_test(&map->vma_flags, VMA_MAYWRITE_BIT) &&
 			vma_test(vma, VMA_MAYWRITE_BIT));
 
-	map->file = vma->vm_file;
 	map->vma_flags = vma->flags;
 
 	return 0;
@@ -2631,7 +2637,7 @@ static int __mmap_new_file_vma(struct mmap_state *map,
 
 static void map_set_anon(struct mmap_state *map)
 {
-	map->file = NULL;
+	map->vm_file = NULL;
 	map->vm_ops = NULL;
 	map->pgoff = map->addr >> PAGE_SHIFT;
 }
@@ -2643,7 +2649,7 @@ static bool map_is_private(const struct mmap_state *map)
 
 static bool map_is_anon(const struct mmap_state *map)
 {
-	return map_is_private(map) && !map->file;
+	return map_is_private(map) && !map->vm_file;
 }
 
 /*
@@ -2688,7 +2694,7 @@ static int __mmap_new_vma(struct mmap_state *map, struct vm_area_struct **vmap,
 	}
 
 	/* Invoke callbacks. */
-	if (map->file)
+	if (map->vm_file)
 		error = __mmap_new_file_vma(map, vma);
 	else if (!is_anon)
 		error = shmem_zero_setup(vma);
@@ -2797,11 +2803,15 @@ static int call_mmap_prepare(struct mmap_state *map,
 	int err;
 
 	/* Invoke the hook. */
-	err = vfs_mmap_prepare(map->file, desc);
+	err = vfs_mmap_prepare(map->vm_file, desc);
 	if (err)
 		return err;
 
-	/* It's invalid for mmap_preprare hooks to clear vm_ops. */
+	/* Update first so file refcount tracked correctly. */
+	if (desc->vm_file != map->vm_file)
+		map->vm_file = desc->vm_file;
+
+	/* It's invalid for mmap_prepare hooks to clear vm_ops. */
 	if (!desc->vm_ops)
 		return -EINVAL;
 
@@ -2811,10 +2821,6 @@ static int call_mmap_prepare(struct mmap_state *map,
 
 	/* Update fields permitted to be changed. */
 	map->pgoff = desc->pgoff;
-	if (desc->vm_file != map->file) {
-		map->file_doesnt_need_get = true;
-		map->file = desc->vm_file;
-	}
 	map->vma_flags = desc->vma_flags;
 	map->page_prot = desc->page_prot;
 	/* User-defined fields. */
@@ -2826,7 +2832,7 @@ static int call_mmap_prepare(struct mmap_state *map,
 	 * anonymous mappings. Rather than allowing these mappings to be odd
 	 * outliers, simply make them truly anonymous.
 	 */
-	if (map_is_private(map) && file_is_dev_zero(map->file))
+	if (map_is_private(map) && file_is_dev_zero(map->vm_file))
 		map_set_anon(map);
 
 	return 0;
@@ -2845,7 +2851,7 @@ static void set_vma_user_defined_fields(struct vm_area_struct *vma,
  */
 static bool can_set_ksm_flags_early(struct mmap_state *map)
 {
-	struct file *file = map->file;
+	struct file *file = map->vm_file;
 
 	/* Anonymous mappings have no driver which can change them. */
 	if (!file)
@@ -2868,6 +2874,20 @@ static bool can_set_ksm_flags_early(struct mmap_state *map)
 	return false;
 }
 
+static void map_put_vm_file(struct mmap_state *map)
+{
+	/*
+	 * An error occurred or the VMA was merged.
+	 *
+	 * If the file was changed by the driver (which is required to increment
+	 * the replacement file's reference count), drop its reference count.
+	 *
+	 * On error, the caller always drops the original file regardless.
+	 */
+	if (map->vm_file && !map_same_file(map))
+		fput(map->vm_file);
+}
+
 static unsigned long __mmap_region(struct file *file, unsigned long addr,
 		unsigned long len, vma_flags_t vma_flags,
 		unsigned long pgoff, struct list_head *uf)
@@ -2922,7 +2942,10 @@ static unsigned long __mmap_region(struct file *file, unsigned long addr,
 
 	__mmap_complete(&map, vma);
 
-	if (have_mmap_prepare && allocated_new) {
+	if (!allocated_new) {
+		/* Merged, so need to drop refcount. */
+		map_put_vm_file(&map);
+	} else if (have_mmap_prepare) {
 		error = mmap_action_complete(vma, &desc.action,
 					     /*is_compat=*/false);
 		if (error)
@@ -2936,13 +2959,7 @@ static unsigned long __mmap_region(struct file *file, unsigned long addr,
 	if (map.charged)
 		vm_unacct_memory(map.charged);
 abort_munmap:
-	/*
-	 * This indicates that .mmap_prepare has set a new file, differing from
-	 * desc->vm_file. But since we're aborting the operation, only the
-	 * original file will be cleaned up. Ensure we clean up both.
-	 */
-	if (map.file_doesnt_need_get)
-		fput(map.file);
+	map_put_vm_file(&map);
 	vms_abort_munmap_vmas(&map.vms, &map.mas_detach);
 	return error;
 }
diff --git a/mm/vma.h b/mm/vma.h
index e97bd2dfa786..f15faa83f3d6 100644
--- a/mm/vma.h
+++ b/mm/vma.h
@@ -394,8 +394,10 @@ static inline void compat_set_vma_from_desc(struct vm_area_struct *vma,
 
 	/* Mutable fields. Populated with initial state. */
 	vma_set_pgoff(vma, desc->pgoff);
-	if (desc->vm_file != vma->vm_file)
-		vma_set_file(vma, desc->vm_file);
+	if (desc->vm_file != vma->vm_file) {
+		fput(vma->vm_file);
+		vma->vm_file = desc->vm_file;
+	}
 	vma->flags = desc->vma_flags;
 	vma->vm_page_prot = desc->page_prot;
 

-- 
2.55.0

Re: [PATCH v6 01/39] mm/vma: fix mmap_prepare file handling, remove file_doesnt_need_get

[sashiko-bot]

> The map->file_doesnt_need_get flag is confusing and the existing
> implementation has holes.
> 
> Drivers are permitted to change the owning file of a mapping. If they do
> so, they are required to take a reference on that file.
> 
> The mmap() operation which ultimately invokes __mmap_region() is guaranteed
> to drop the refcount for the original file the mapping was made under, but
> this is not true for the replaced file.
> 
> This has been addressed so far by tracking map->file_doesnt_need_get, which
> is rather poorly named and unfortunately fails to correctly track whether
> or not an additional put were needed in a number of cases.
> 
> Make life easier by removing this flag, and instead drop the reference for
> [ ... ]
> currently (no driver that does this is mergeable for one), so this does not
> need to be a fix.
> 
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=1

[PATCH v6 02/39] mm/vma: introduce and use vma_[flags_]can_merge()

[Lorenzo Stoakes (ARM)]

Replace the open-coded VMA_SPECIAL_FLAGS check in the VMA merge logic with
two new functions vma_flags_can_merge() and vma_can_merge() and update the
merge logic to use the former.

This abstracts the check and expresses it in terms of the desired behaviour
rather than an arbitrary and confusing VMA flag.

This also lays the groundwork for making further improvements in VMA flag
usage.

Also update the userland VMA tests to reflect the change.

No functional change intended.

Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 include/linux/mm.h              | 21 +++++++++++++++++++++
 mm/vma.c                        | 19 +++++++++++--------
 tools/testing/vma/include/dup.h |  5 +++++
 3 files changed, 37 insertions(+), 8 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 969594074fd2..68250ece4219 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1612,6 +1612,27 @@ static inline bool vma_is_shared_maywrite(const struct vm_area_struct *vma)
 	return is_shared_maywrite(&vma->flags);
 }
 
+/**
+ * vma_flags_can_merge() - Do the specified VMA flags permit the VMA to be
+ * merged with another?
+ * @flags: The VMA flags to test.
+ * Returns: true if the flags permit merging, false otherwise.
+ */
+static inline bool vma_flags_can_merge(const vma_flags_t *flags)
+{
+	return !vma_flags_test_any_mask(flags, VMA_SPECIAL_FLAGS);
+}
+
+/**
+ * vma_can_merge() - Do @vma's flags permit it to be merged with another VMA?
+ * @vma: The VMA to test.
+ * Returns: true if the flags permit merging, otherwise false.
+ */
+static inline bool vma_can_merge(const struct vm_area_struct *vma)
+{
+	return vma_flags_can_merge(&vma->flags);
+}
+
 /**
  * vma_kernel_pagesize - Default page size granularity for this VMA.
  * @vma: The user mapping.
diff --git a/mm/vma.c b/mm/vma.c
index c27aea6d58d9..8f8c554d5c07 100644
--- a/mm/vma.c
+++ b/mm/vma.c
@@ -924,13 +924,14 @@ static __must_check struct vm_area_struct *vma_merge_existing_range(
 
 	vmg->state = VMA_MERGE_NOMERGE;
 
+	if (!vma_flags_can_merge(&vmg->vma_flags))
+		return NULL;
 	/*
-	 * If a special mapping or if the range being modified is neither at the
-	 * furthermost left or right side of the VMA, then we have no chance of
-	 * merging and should abort.
+	 * If the range being modified is neither at the furthermost left or
+	 * right side of the VMA, then we have no chance of merging and should
+	 * abort.
 	 */
-	if (vma_flags_test_any_mask(&vmg->vma_flags, VMA_SPECIAL_FLAGS) ||
-	    (!left_side && !right_side))
+	if (!left_side && !right_side)
 		return NULL;
 
 	if (left_side)
@@ -1152,9 +1153,11 @@ struct vm_area_struct *vma_merge_new_range(struct vma_merge_struct *vmg)
 
 	vmg->state = VMA_MERGE_NOMERGE;
 
-	/* Special VMAs are unmergeable, also if no prev/next. */
-	if (vma_flags_test_any_mask(&vmg->vma_flags, VMA_SPECIAL_FLAGS) ||
-	    (!prev && !next))
+	if (!vma_flags_can_merge(&vmg->vma_flags))
+		return NULL;
+
+	/* VMAs with no prev/next are unmergeable. */
+	if (!prev && !next)
 		return NULL;
 
 	can_merge_left = can_vma_merge_left(vmg);
diff --git a/tools/testing/vma/include/dup.h b/tools/testing/vma/include/dup.h
index 16c09dac59d9..2fd422789717 100644
--- a/tools/testing/vma/include/dup.h
+++ b/tools/testing/vma/include/dup.h
@@ -1647,3 +1647,8 @@ static inline bool file_is_dev_zero(const struct file *file)
 {
 	return file && file->f_op == &zero_fops;
 }
+
+static inline bool vma_flags_can_merge(const vma_flags_t *flags)
+{
+	return !vma_flags_test_any_mask(flags, VMA_SPECIAL_FLAGS);
+}

-- 
2.55.0

Re: [PATCH v6 02/39] mm/vma: introduce and use vma_[flags_]can_merge()

[David Hildenbrand (Arm)]

On 10/13/26 17:04, Lorenzo Stoakes (ARM) wrote:
> Replace the open-coded VMA_SPECIAL_FLAGS check in the VMA merge logic with
> two new functions vma_flags_can_merge() and vma_can_merge() and update the
> merge logic to use the former.
> 
> This abstracts the check and expresses it in terms of the desired behaviour
> rather than an arbitrary and confusing VMA flag.
> 
> This also lays the groundwork for making further improvements in VMA flag
> usage.
> 
> Also update the userland VMA tests to reflect the change.
> 
> No functional change intended.
> 
> Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
> ---

I'd have called it "vma_mergable" etc, because "can_merge" makes me imply a
"with what".

Acked-by: David Hildenbrand (Arm) <david@kernel.org>

-- 
Cheers,

David

Re: [PATCH v6 02/39] mm/vma: introduce and use vma_[flags_]can_merge()

[sashiko-bot]

> Replace the open-coded VMA_SPECIAL_FLAGS check in the VMA merge logic with
> two new functions vma_flags_can_merge() and vma_can_merge() and update the
> merge logic to use the former.
> 
> This abstracts the check and expresses it in terms of the desired behaviour
> rather than an arbitrary and confusing VMA flag.
> 
> This also lays the groundwork for making further improvements in VMA flag
> usage.
> 
> Also update the userland VMA tests to reflect the change.
> 
> No functional change intended.
> 
> Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=2

[PATCH v6 03/39] mm: consistently validate VMA state after mmap[_prepare] hooks

[Lorenzo Stoakes (ARM)]

When the f_op->mmap_prepare or deprecated f_op->mmap hooks are invoked, the
driver might have done something crazy that is not permitted by the kernel.

Currently we check for three such cases in __mmap_new_file_vma(), but only
if the legacy f_op->mmap hook is used:

* Did sparc ADI result in invalid flags?

* Did the driver alter vma->vm_start?

* Did the driver make a file-backed mapping on a read-only file writable?

Generalise these checks for both mmap_prepare and mmap and apply to all
invocations of mmap_file(), the f_op->mmap and f_op->mmap_prepare handling
in the core VMA code and the mmap_prepare compatibility layer.

Also extend the vm_start check to vm_end also - drivers must not change the
VMA range at all.

We also WARN_ON_ONCE() on these conditions as they are things that should
simply not occur in the kernel and it's important to call it out when it
does.

We invoke mmap_prepare_validate() after mmap_action_prepare(), as mmap
actions often manipulate state in the descriptor thus providing the final
state the VMA will be derived from.

Also call mmap_validate_vma_flags() in insert_vm_struct() to ensure that
special regions which are inserted (such as a VDSO or VVAR) also satisfy
the sanity checks.

This way every VMA established through an mmap hook, whether via mmap() or
the compatibility layer, or inserted via insert_vm_struct(), has been
validated. brk() VMAs never pass through a driver hook and so need no such
check.

While we're here, also fixup a couple disjoint blocks of #ifdef CONFIG_MMU.

Finally, update the VMA userland tests to reflect the change.

Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/internal.h                     |  50 +++++++++++-------
 mm/util.c                         |  19 +++++--
 mm/vma.c                          | 106 ++++++++++++++++++++++++++++++++------
 mm/vma.h                          |  25 +++++++--
 tools/testing/vma/include/dup.h   |  10 ++++
 tools/testing/vma/include/stubs.h |   4 ++
 6 files changed, 172 insertions(+), 42 deletions(-)

diff --git a/mm/internal.h b/mm/internal.h
index fe576d468af4..061a87f9c3fb 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -213,6 +213,24 @@ static inline void *folio_raw_mapping(const struct folio *folio)
 	return (void *)(mapping & ~FOLIO_MAPPING_FLAGS);
 }
 
+/*
+ * If the VMA has a close hook then close it, and since closing it might leave
+ * it in an inconsistent state which makes the use of any hooks suspect, clear
+ * them down by installing dummy empty hooks.
+ */
+static inline void vma_close(struct vm_area_struct *vma)
+{
+	if (vma->vm_ops && vma->vm_ops->close) {
+		vma->vm_ops->close(vma);
+
+		/*
+		 * The mapping is in an inconsistent state, and no further hooks
+		 * may be invoked upon it.
+		 */
+		vma->vm_ops = &vma_dummy_vm_ops;
+	}
+}
+
 /*
  * This is a file-backed mapping, and is about to be memory mapped - invoke its
  * mmap hook and safely handle error conditions. On error, VMA hooks will be
@@ -225,8 +243,12 @@ static inline void *folio_raw_mapping(const struct folio *folio)
  */
 static inline int mmap_file(struct file *file, struct vm_area_struct *vma)
 {
-	int err = vfs_mmap(file, vma);
+	const unsigned long orig_start = vma->vm_start;
+	const unsigned long orig_end = vma->vm_end;
+	const vma_flags_t orig_flags = vma->flags;
+	int err;
 
+	err = vfs_mmap(file, vma);
 	/*
 	 * Either we tried to call the file hook for mmap() and an error arose
 	 * or a driver set vma->vm_ops = NULL intending there to be no VMA
@@ -239,26 +261,16 @@ static inline int mmap_file(struct file *file, struct vm_area_struct *vma)
 	 */
 	if (unlikely(err || !vma->vm_ops))
 		vma->vm_ops = &vma_dummy_vm_ops;
+	if (unlikely(err))
+		return err;
 
-	return err;
-}
-
-/*
- * If the VMA has a close hook then close it, and since closing it might leave
- * it in an inconsistent state which makes the use of any hooks suspect, clear
- * them down by installing dummy empty hooks.
- */
-static inline void vma_close(struct vm_area_struct *vma)
-{
-	if (vma->vm_ops && vma->vm_ops->close) {
-		vma->vm_ops->close(vma);
-
-		/*
-		 * The mapping is in an inconsistent state, and no further hooks
-		 * may be invoked upon it.
-		 */
-		vma->vm_ops = &vma_dummy_vm_ops;
+	err = mmap_hook_validate(orig_start, orig_end, &orig_flags, vma);
+	if (unlikely(err)) {
+		vma->vm_start = orig_start;
+		vma->vm_end = orig_end;
 	}
+
+	return err;
 }
 
 /* unmap_vmas is in mm/memory.c */
diff --git a/mm/util.c b/mm/util.c
index 016932780925..d809e08c7b4d 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -1224,19 +1224,28 @@ EXPORT_SYMBOL(compat_set_desc_from_vma);
 int __compat_vma_mmap(struct vm_area_desc *desc,
 		      struct vm_area_struct *vma)
 {
+	struct vm_area_desc orig_desc;
 	int err;
 
+	/* Derive state prior to mmap_prepare hook. */
+	compat_set_desc_from_vma(&orig_desc, desc->file, vma);
 	/* Perform any preparatory tasks for mmap action. */
 	err = mmap_action_prepare(desc);
-	if (err) {
-		if (desc->vm_file != vma->vm_file)
-			fput(desc->vm_file);
-		return err;
-	}
+	if (err)
+		goto err_put;
+	/* Check the caller did nothing crazy. */
+	err = mmap_prepare_validate(&orig_desc, desc);
+	if (err)
+		goto err_put;
 	/* Update the VMA from the descriptor. */
 	compat_set_vma_from_desc(vma, desc);
 	/* Complete any specified mmap actions. */
 	return mmap_action_complete(vma, &desc->action, /*is_compat=*/true);
+
+err_put:
+	if (desc->vm_file != vma->vm_file)
+		fput(desc->vm_file);
+	return err;
 }
 EXPORT_SYMBOL(__compat_vma_mmap);
 
diff --git a/mm/vma.c b/mm/vma.c
index 8f8c554d5c07..cedc24dcdf3c 100644
--- a/mm/vma.c
+++ b/mm/vma.c
@@ -2613,6 +2613,12 @@ static int __mmap_new_file_vma(struct mmap_state *map,
 	if (error) {
 		UNMAP_STATE(unmap, vmi, vma, vma->vm_start, vma->vm_end,
 			    map->prev, map->next);
+
+		if (vma->vm_ops->close) {
+			zap_vma(vma);
+			/* Safe, because VMA zapped. */
+			vma_close(vma);
+		}
 		if (map_same_file(map))
 			fput(map->vm_file);
 
@@ -2623,16 +2629,6 @@ static int __mmap_new_file_vma(struct mmap_state *map,
 		return error;
 	}
 
-	/* Drivers cannot alter the address of the VMA. */
-	WARN_ON_ONCE(map->addr != vma->vm_start);
-	/*
-	 * Drivers should not permit writability when previously it was
-	 * disallowed.
-	 */
-	VM_WARN_ON_ONCE(!vma_flags_same_pair(&map->vma_flags, &vma->flags) &&
-			!vma_flags_test(&map->vma_flags, VMA_MAYWRITE_BIT) &&
-			vma_test(vma, VMA_MAYWRITE_BIT));
-
 	map->vma_flags = vma->flags;
 
 	return 0;
@@ -2710,11 +2706,6 @@ static int __mmap_new_vma(struct mmap_state *map, struct vm_area_struct **vmap,
 		vma->flags = map->vma_flags;
 	}
 
-#ifdef CONFIG_SPARC64
-	/* TODO: Fix SPARC ADI! */
-	WARN_ON_ONCE(!arch_validate_flags(map->vm_flags));
-#endif
-
 	/* Lock the VMA since it is modified after insertion into VMA tree */
 	vma_start_write(vma);
 	vma_iter_store_new(vmi, vma);
@@ -2777,6 +2768,80 @@ static void __mmap_complete(struct mmap_state *map, struct vm_area_struct *vma)
 	vma_set_page_prot(vma);
 }
 
+/* Check to ensure that the VMA flags of a newly mapped VMA are sane. */
+static int mmap_validate_vma_flags(const vma_flags_t *flags)
+{
+#ifdef CONFIG_SPARC64
+	const vm_flags_t legacy_flags = vma_flags_to_legacy(*flags);
+
+	/* TODO: Fix SPARC ADI! */
+	if (WARN_ON_ONCE(!arch_validate_flags(legacy_flags)))
+		return -EINVAL;
+#endif
+
+	return 0;
+}
+
+/* Check to ensure a driver hasn't done something crazy. */
+static int mmap_validate(unsigned long orig_start, unsigned long orig_end,
+			 unsigned long curr_start, unsigned long curr_end,
+			 const vma_flags_t *orig_flags,
+			 const vma_flags_t *curr_flags)
+{
+	bool was_maywrite, is_maywrite;
+
+	/* Drivers cannot alter the range of the VMA. */
+	if (WARN_ON_ONCE(orig_start != curr_start || orig_end != curr_end))
+		return -EINVAL;
+
+	was_maywrite = vma_flags_test(orig_flags, VMA_MAYWRITE_BIT);
+	is_maywrite = vma_flags_test(curr_flags, VMA_MAYWRITE_BIT);
+
+	/* A driver may not make a previously unwritable mapping writable. */
+	if (WARN_ON_ONCE(!was_maywrite && is_maywrite))
+		return -EINVAL;
+
+	return mmap_validate_vma_flags(curr_flags);
+}
+
+/**
+ * mmap_prepare_validate() - Ensure the driver hasn't violated invariants in its
+ * f_op->mmap_prepare hook.
+ * @orig_desc: The VMA descriptor prior to the mmap_prepare hook being called.
+ * @desc: The VMA descriptor after the mmap_prepare hook has been called.
+ *
+ * Returns: 0 on success, otherwise an error.
+ */
+int mmap_prepare_validate(const struct vm_area_desc *orig_desc,
+			  const struct vm_area_desc *desc)
+{
+	return mmap_validate(orig_desc->start, orig_desc->end,
+			     desc->start, desc->end,
+			     &orig_desc->vma_flags, &desc->vma_flags);
+}
+
+/**
+ * mmap_hook_validate() - Ensure the driver hasn't violated invariants in
+ * its f_op->mmap hook.
+ * @orig_start: The start of the mapping prior to the mmap hook.
+ * @orig_end: The end of the mapping prior to the mmap hook.
+ * @orig_flags: The VMA flags set for the VMA prior to the mmap hook.
+ * @vma: The VMA after the hook has been applied.
+ *
+ * Returns: 0 on success, otherwise an error.
+ */
+int mmap_hook_validate(unsigned long orig_start, unsigned long orig_end,
+		       const vma_flags_t *orig_flags,
+		       const struct vm_area_struct *vma)
+{
+	const unsigned long start = vma->vm_start;
+	const unsigned long end = vma->vm_end;
+	const vma_flags_t *flags = &vma->flags;
+
+	return mmap_validate(orig_start, orig_end, start, end, orig_flags,
+			     flags);
+}
+
 static int call_action_prepare(struct mmap_state *map,
 			       struct vm_area_desc *desc)
 {
@@ -2803,6 +2868,7 @@ static int call_action_prepare(struct mmap_state *map,
 static int call_mmap_prepare(struct mmap_state *map,
 		struct vm_area_desc *desc)
 {
+	const struct vm_area_desc orig_desc = *desc;
 	int err;
 
 	/* Invoke the hook. */
@@ -2822,6 +2888,11 @@ static int call_mmap_prepare(struct mmap_state *map,
 	if (err)
 		return err;
 
+	/* Check the caller did nothing crazy. */
+	err = mmap_prepare_validate(&orig_desc, desc);
+	if (err)
+		return err;
+
 	/* Update fields permitted to be changed. */
 	map->pgoff = desc->pgoff;
 	map->vma_flags = desc->vma_flags;
@@ -3457,10 +3528,15 @@ int __vm_munmap(unsigned long start, size_t len, bool unlock)
 int insert_vm_struct(struct mm_struct *mm, struct vm_area_struct *vma)
 {
 	unsigned long charged = vma_pages(vma);
+	int err;
 
 	if (find_vma_intersection(mm, vma->vm_start, vma->vm_end))
 		return -ENOMEM;
 
+	err = mmap_validate_vma_flags(&vma->flags);
+	if (err)
+		return err;
+
 	if (vma_test(vma, VMA_ACCOUNT_BIT) &&
 	     security_vm_enough_memory_mm(mm, charged))
 		return -ENOMEM;
diff --git a/mm/vma.h b/mm/vma.h
index f15faa83f3d6..89f831080bd7 100644
--- a/mm/vma.h
+++ b/mm/vma.h
@@ -782,14 +782,19 @@ struct vm_area_struct *vm_area_alloc(struct mm_struct *mm);
 struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig);
 void vm_area_free(struct vm_area_struct *vma);
 
-/* vma_exec.c */
 #ifdef CONFIG_MMU
+int mmap_prepare_validate(const struct vm_area_desc *orig_desc,
+			  const struct vm_area_desc *desc);
+
+int mmap_hook_validate(unsigned long orig_start, unsigned long orig_end,
+		       const vma_flags_t *orig_flags,
+		       const struct vm_area_struct *vma);
+
+/* vma_exec.c */
 int create_init_stack_vma(struct mm_struct *mm, struct vm_area_struct **vmap,
 			  unsigned long *top_mem_p);
 int relocate_vma_down(struct vm_area_struct *vma, unsigned long shift);
-#endif
 
-#ifdef CONFIG_MMU
 /*
  * Denies creating a writable executable mapping or gaining executable permissions.
  *
@@ -838,6 +843,20 @@ static inline bool map_deny_write_exec(const vma_flags_t *old,
 
 	return false;
 }
+#else
+static inline int mmap_prepare_validate(const struct vm_area_desc *orig_desc,
+					const struct vm_area_desc *desc)
+{
+	return 0;
+}
+
+static inline int mmap_hook_validate(unsigned long orig_start,
+				     unsigned long orig_end,
+				     const vma_flags_t *orig_flags,
+				     const struct vm_area_struct *vma)
+{
+	return 0;
+}
 #endif
 
 struct vm_area_struct *__install_special_mapping(struct mm_struct *mm,
diff --git a/tools/testing/vma/include/dup.h b/tools/testing/vma/include/dup.h
index 2fd422789717..077df510a311 100644
--- a/tools/testing/vma/include/dup.h
+++ b/tools/testing/vma/include/dup.h
@@ -1359,13 +1359,23 @@ static inline int vfs_mmap_prepare(struct file *file, struct vm_area_desc *desc)
 	return file->f_op->mmap_prepare(desc);
 }
 
+int mmap_prepare_validate(const struct vm_area_desc *orig_desc,
+			  const struct vm_area_desc *desc);
+
 static inline int __compat_vma_mmap(struct vm_area_desc *desc,
 		struct vm_area_struct *vma)
 {
+	struct vm_area_desc orig_desc;
 	int err;
 
+	/* Derive state prior to mmap_prepare hook. */
+	compat_set_desc_from_vma(&orig_desc, desc->file, vma);
 	/* Perform any preparatory tasks for mmap action. */
 	err = mmap_action_prepare(desc);
+	if (err)
+		return err;
+	/* Check the caller did nothing crazy. */
+	err = mmap_prepare_validate(&orig_desc, desc);
 	if (err)
 		return err;
 	/* Update the VMA from the descriptor. */
diff --git a/tools/testing/vma/include/stubs.h b/tools/testing/vma/include/stubs.h
index d6136e19a8af..b35279a0ec78 100644
--- a/tools/testing/vma/include/stubs.h
+++ b/tools/testing/vma/include/stubs.h
@@ -329,6 +329,10 @@ static inline void vma_close(struct vm_area_struct *vma)
 {
 }
 
+static inline void zap_vma(struct vm_area_struct *vma)
+{
+}
+
 static inline int mmap_file(struct file *file, struct vm_area_struct *vma)
 {
 	return 0;

-- 
2.55.0

Re: [PATCH v6 03/39] mm: consistently validate VMA state after mmap[_prepare] hooks

[sashiko-bot]

> When the f_op->mmap_prepare or deprecated f_op->mmap hooks are invoked, the
> driver might have done something crazy that is not permitted by the kernel.
> 
> Currently we check for three such cases in __mmap_new_file_vma(), but only
> if the legacy f_op->mmap hook is used:
> 
> * Did sparc ADI result in invalid flags?
> 
> * Did the driver alter vma->vm_start?
> 
> * Did the driver make a file-backed mapping on a read-only file writable?
> 
> Generalise these checks for both mmap_prepare and mmap and apply to all
> invocations of mmap_file(), the f_op->mmap and f_op->mmap_prepare handling
> in the core VMA code and the mmap_prepare compatibility layer.
> [ ... ]
> 
> Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=3

[PATCH v6 04/39] mm/vma: ensure mmap_prepare doesn't set actions on a mergeable vma

[Lorenzo Stoakes (ARM)]

When a user requests an mmap_action be performed in mmap_prepare, this
involves populating the VMA range with data.

However, if the VMA is mergeable, it might then mistakenly be merged with
another VMA without having populated the range.

Every mmap action currently available sets VMA flags such that the VMA
cannot be merged.

However, to ensure that no future mmap action falls foul of this, assert
that this is the case upon mmap_prepare validation.

Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/vma.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/mm/vma.c b/mm/vma.c
index cedc24dcdf3c..c067d6a4c826 100644
--- a/mm/vma.c
+++ b/mm/vma.c
@@ -2815,6 +2815,18 @@ static int mmap_validate(unsigned long orig_start, unsigned long orig_end,
 int mmap_prepare_validate(const struct vm_area_desc *orig_desc,
 			  const struct vm_area_desc *desc)
 {
+	/*
+	 * It is not valid to execute mmap actions for VMAs which can be merged,
+	 * as any such merge would leave portions of the mapping incorrectly
+	 * unmapped.
+	 *
+	 * This is checked after mmap_action_prepare(), as it may update the VMA
+	 * flags and therefore whether the VMA can be merged.
+	 */
+	if (vma_flags_can_merge(&desc->vma_flags) &&
+	    WARN_ON_ONCE(desc->action.type != MMAP_NOTHING))
+		return -EINVAL;
+
 	return mmap_validate(orig_desc->start, orig_desc->end,
 			     desc->start, desc->end,
 			     &orig_desc->vma_flags, &desc->vma_flags);

-- 
2.55.0

Re: [PATCH v6 04/39] mm/vma: ensure mmap_prepare doesn't set actions on a mergeable vma

[sashiko-bot]

> When a user requests an mmap_action be performed in mmap_prepare, this
> involves populating the VMA range with data.
> 
> However, if the VMA is mergeable, it might then mistakenly be merged with
> another VMA without having populated the range.
> 
> Every mmap action currently available sets VMA flags such that the VMA
> cannot be merged.
> 
> However, to ensure that no future mmap action falls foul of this, assert
> that this is the case upon mmap_prepare validation.
> 
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=4

[PATCH v6 05/39] mm: make map_kernel_pages_[prepare,complete] internal and unexported

[Lorenzo Stoakes (ARM)]

There's no reason to export the symbols for these functions which are only
called from internal mm logic, additionally there's no reason for them to
be declared in mm.h.

This patch therefore removes the exports and moves the declarations to
mm/internal.h.

No functional change intended.

Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 include/linux/mm.h | 3 ---
 mm/internal.h      | 3 +++
 mm/memory.c        | 2 --
 3 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 68250ece4219..0c17f0932b76 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -4768,9 +4768,6 @@ int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
 			struct page **pages, unsigned long *num);
-int map_kernel_pages_prepare(struct vm_area_desc *desc);
-int map_kernel_pages_complete(struct vm_area_struct *vma,
-			      struct mmap_action *action);
 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
 				unsigned long num);
 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
diff --git a/mm/internal.h b/mm/internal.h
index 061a87f9c3fb..fc4e249e2343 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -1512,6 +1512,9 @@ int remap_pfn_range_prepare(struct vm_area_desc *desc);
 int remap_pfn_range_complete(struct vm_area_struct *vma,
 			     struct mmap_action *action);
 int simple_ioremap_prepare(struct vm_area_desc *desc);
+int map_kernel_pages_prepare(struct vm_area_desc *desc);
+int map_kernel_pages_complete(struct vm_area_struct *vma,
+			      struct mmap_action *action);
 
 static inline int io_remap_pfn_range_prepare(struct vm_area_desc *desc)
 {
diff --git a/mm/memory.c b/mm/memory.c
index 926276d41920..448342883e9d 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2628,7 +2628,6 @@ int map_kernel_pages_prepare(struct vm_area_desc *desc)
 
 	return 0;
 }
-EXPORT_SYMBOL(map_kernel_pages_prepare);
 
 int map_kernel_pages_complete(struct vm_area_struct *vma,
 			      struct mmap_action *action)
@@ -2640,7 +2639,6 @@ int map_kernel_pages_complete(struct vm_area_struct *vma,
 			    action->map_kernel.pages,
 			    &nr_pages, vma->vm_page_prot);
 }
-EXPORT_SYMBOL(map_kernel_pages_complete);
 
 /**
  * vm_insert_page - insert single page into user vma

-- 
2.55.0

Re: [PATCH v6 05/39] mm: make map_kernel_pages_[prepare,complete] internal and unexported

[sashiko-bot]

> There's no reason to export the symbols for these functions which are only
> called from internal mm logic, additionally there's no reason for them to
> be declared in mm.h.
> 
> This patch therefore removes the exports and moves the declarations to
> mm/internal.h.
> 
> No functional change intended.
> 
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=5

[PATCH v6 06/39] mm/vma: tidy up map kernel pages enum values

[Lorenzo Stoakes (ARM)]

MMAP_MAP_KERNEL_PAGES is a mouthful, discard the MAP_ as that's implied by
MMAP.

Also while we're here delete useless comments for mmap actions whose names
clearly indicate what they are for.

Also update the userland VMA tests to reflect this change.

No functional change intended.

Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 include/linux/mm.h              | 2 +-
 include/linux/mm_types.h        | 8 ++++----
 mm/util.c                       | 8 ++++----
 tools/testing/vma/include/dup.h | 8 ++++----
 4 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 0c17f0932b76..a1f2d375cf7d 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -4623,7 +4623,7 @@ static inline void mmap_action_map_kernel_pages(struct vm_area_desc *desc,
 {
 	struct mmap_action *action = &desc->action;
 
-	action->type = MMAP_MAP_KERNEL_PAGES;
+	action->type = MMAP_KERNEL_PAGES;
 	action->map_kernel.start = start;
 	action->map_kernel.pages = pages;
 	action->map_kernel.nr_pages = nr_pages;
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 5413bd10fff2..9ca2ea3664bc 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -815,11 +815,11 @@ struct pfnmap_track_ctx {
 
 /* What action should be taken after an .mmap_prepare call is complete? */
 enum mmap_action_type {
-	MMAP_NOTHING,		/* Mapping is complete, no further action. */
-	MMAP_REMAP_PFN,		/* Remap PFN range. */
-	MMAP_IO_REMAP_PFN,	/* I/O remap PFN range. */
+	MMAP_NOTHING,
+	MMAP_REMAP_PFN,
+	MMAP_IO_REMAP_PFN,
 	MMAP_SIMPLE_IO_REMAP,	/* I/O remap with guardrails. */
-	MMAP_MAP_KERNEL_PAGES,	/* Map kernel page range from array. */
+	MMAP_KERNEL_PAGES,	/* Map kernel page range from array. */
 };
 
 /*
diff --git a/mm/util.c b/mm/util.c
index d809e08c7b4d..1660910e2a22 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -1467,7 +1467,7 @@ int mmap_action_prepare(struct vm_area_desc *desc)
 		return io_remap_pfn_range_prepare(desc);
 	case MMAP_SIMPLE_IO_REMAP:
 		return simple_ioremap_prepare(desc);
-	case MMAP_MAP_KERNEL_PAGES:
+	case MMAP_KERNEL_PAGES:
 		return map_kernel_pages_prepare(desc);
 	}
 
@@ -1498,7 +1498,7 @@ int mmap_action_complete(struct vm_area_struct *vma,
 	case MMAP_REMAP_PFN:
 		err = remap_pfn_range_complete(vma, action);
 		break;
-	case MMAP_MAP_KERNEL_PAGES:
+	case MMAP_KERNEL_PAGES:
 		err = map_kernel_pages_complete(vma, action);
 		break;
 	case MMAP_IO_REMAP_PFN:
@@ -1521,7 +1521,7 @@ int mmap_action_prepare(struct vm_area_desc *desc)
 	case MMAP_REMAP_PFN:
 	case MMAP_IO_REMAP_PFN:
 	case MMAP_SIMPLE_IO_REMAP:
-	case MMAP_MAP_KERNEL_PAGES:
+	case MMAP_KERNEL_PAGES:
 		WARN_ON_ONCE(1); /* nommu cannot handle these. */
 		break;
 	}
@@ -1542,7 +1542,7 @@ int mmap_action_complete(struct vm_area_struct *vma,
 	case MMAP_REMAP_PFN:
 	case MMAP_IO_REMAP_PFN:
 	case MMAP_SIMPLE_IO_REMAP:
-	case MMAP_MAP_KERNEL_PAGES:
+	case MMAP_KERNEL_PAGES:
 		WARN_ON_ONCE(1); /* nommu cannot handle this. */
 
 		err = -EINVAL;
diff --git a/tools/testing/vma/include/dup.h b/tools/testing/vma/include/dup.h
index 077df510a311..023ad6092e9c 100644
--- a/tools/testing/vma/include/dup.h
+++ b/tools/testing/vma/include/dup.h
@@ -454,11 +454,11 @@ static __always_inline bool vma_flags_empty(const vma_flags_t *flags)
 
 /* What action should be taken after an .mmap_prepare call is complete? */
 enum mmap_action_type {
-	MMAP_NOTHING,		/* Mapping is complete, no further action. */
-	MMAP_REMAP_PFN,		/* Remap PFN range. */
-	MMAP_IO_REMAP_PFN,	/* I/O remap PFN range. */
+	MMAP_NOTHING,
+	MMAP_REMAP_PFN,
+	MMAP_IO_REMAP_PFN,
 	MMAP_SIMPLE_IO_REMAP,	/* I/O remap with guardrails. */
-	MMAP_MAP_KERNEL_PAGES,	/* Map kernel page range from an array. */
+	MMAP_KERNEL_PAGES,	/* Map kernel page range from array. */
 };
 
 /*

-- 
2.55.0

Re: [PATCH v6 06/39] mm/vma: tidy up map kernel pages enum values

[sashiko-bot]

> MMAP_MAP_KERNEL_PAGES is a mouthful, discard the MAP_ as that's implied by
> MMAP.
> 
> Also while we're here delete useless comments for mmap actions whose names
> clearly indicate what they are for.
> 
> Also update the userland VMA tests to reflect this change.
> 
> No functional change intended.
> 
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=6

[PATCH v6 07/39] mm: assert mmap lock held, use new flags API in vm_insert_page[s]

[Lorenzo Stoakes (ARM)]

The existing logic asserts that an mmap write lock is held, bizarrely, by
asserting that an mmap_read_trylock() fails (which is not the same thing).

Assert the mmap write lock here and switch to using the new VMA flags API
while here.

Note that the vma_start_write() is explicitly broken out from
vm_flags_set() here in order to use vma_set_flags() safely.

Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/memory.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/mm/memory.c b/mm/memory.c
index 448342883e9d..a35cca09eb71 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2599,10 +2599,11 @@ int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
 
 	if (!range_in_vma(vma, addr, end))
 		return -EFAULT;
-	if (!(vma->vm_flags & VM_MIXEDMAP)) {
-		VM_WARN_ON_ONCE(mmap_read_trylock(vma->vm_mm));
-		VM_WARN_ON_ONCE(vma->vm_flags & VM_PFNMAP);
-		vm_flags_set(vma, VM_MIXEDMAP);
+	if (!vma_test(vma, VMA_MIXEDMAP_BIT)) {
+		mmap_assert_write_locked(vma->vm_mm);
+		VM_WARN_ON_ONCE(vma_test(vma, VMA_PFNMAP_BIT));
+		vma_start_write(vma);
+		vma_set_flags(vma, VMA_MIXEDMAP_BIT);
 	}
 	/* Defer page refcount checking till we're about to map that page. */
 	return insert_pages(vma, addr, pages, num, vma->vm_page_prot);
@@ -2675,10 +2676,11 @@ int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
 {
 	if (addr < vma->vm_start || addr >= vma->vm_end)
 		return -EFAULT;
-	if (!(vma->vm_flags & VM_MIXEDMAP)) {
-		BUG_ON(mmap_read_trylock(vma->vm_mm));
-		BUG_ON(vma->vm_flags & VM_PFNMAP);
-		vm_flags_set(vma, VM_MIXEDMAP);
+	if (!vma_test(vma, VMA_MIXEDMAP_BIT)) {
+		mmap_assert_write_locked(vma->vm_mm);
+		VM_WARN_ON_ONCE(vma_test(vma, VMA_PFNMAP_BIT));
+		vma_start_write(vma);
+		vma_set_flags(vma, VMA_MIXEDMAP_BIT);
 	}
 	return insert_page(vma, addr, page, vma->vm_page_prot, false);
 }

-- 
2.55.0

Re: [PATCH v6 07/39] mm: assert mmap lock held, use new flags API in vm_insert_page[s]

[David Hildenbrand (Arm)]

On 10/13/26 17:04, Lorenzo Stoakes (ARM) wrote:
> The existing logic asserts that an mmap write lock is held, bizarrely, by
> asserting that an mmap_read_trylock() fails (which is not the same thing).
> 
> Assert the mmap write lock here and switch to using the new VMA flags API
> while here.
> 
> Note that the vma_start_write() is explicitly broken out from
> vm_flags_set() here in order to use vma_set_flags() safely.

Which is a lot clearer.

Acked-by: David Hildenbrand (Arm) <david@kernel.org>

-- 
Cheers,

David

Re: [PATCH v6 07/39] mm: assert mmap lock held, use new flags API in vm_insert_page[s]

[sashiko-bot]

> The existing logic asserts that an mmap write lock is held, bizarrely, by
> asserting that an mmap_read_trylock() fails (which is not the same thing).
> 
> Assert the mmap write lock here and switch to using the new VMA flags API
> while here.
> 
> Note that the vma_start_write() is explicitly broken out from
> vm_flags_set() here in order to use vma_set_flags() safely.
> 
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=7

[PATCH v6 08/39] mm: add mmap action for discontiguous kernel page mapping

[Lorenzo Stoakes (ARM)]

The existing kernel page mapping mmap actions allow for partial and full
mapping of an array of struct page pointers.

However some drivers require the mapping of discontiguous ranges. Permit
this by providing discontig_kernel_page_ops which allows a driver to
specify how the operation should begin and how batches of pages should be
retrieved.

It uses the minimum exposed interface to do so, providing address, page
offset and both vm_private_data state and a local private state object.

ops->init can establish state for the operation, and ops->get outputs the
pages to map and their count. Should an error arise the core unmaps the
VMA, invoking vm_ops->close, which is therefore where any state established
by ops->init is released.

Also update the compatibility layer to do this too via
compat_mmap_action_abort().

Batches may not exceed the VMA, but may map less than its full range in
case the driver wishes to allow the user to map an area larger than the
available data.

To use it, users invoke mmap_action_map_discontig_kernel_pages() with
initial local private state and a set of operations.

Users can then use one of the provided helper functions to perform an
action:

* discontig_kernel_map_abort() - Abort and leave the mapping as it has
  been accumulated so far.
* discontig_kernel_map_page() - Map a single page, or a compound page given
  its head page, as 1..N contiguous pages.
* discontig_kernel_map_pages() - Map a struct page ** array of a specified
  count, mapping only those pages which fit in the VMA.

The userland VMA tests are updated accordingly.

Acked-by: Zi Yan <ziy@nvidia.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 include/linux/mm.h              |  43 ++++++++++++++++
 include/linux/mm_types.h        |  42 +++++++++++++++-
 mm/internal.h                   |   3 ++
 mm/memory.c                     | 106 ++++++++++++++++++++++++++++++++++++++--
 mm/util.c                       |  27 ++++++++--
 tools/testing/vma/include/dup.h |  11 ++++-
 6 files changed, 218 insertions(+), 14 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index a1f2d375cf7d..438ab06a4a91 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -4647,10 +4647,53 @@ static inline void mmap_action_map_kernel_pages_full(struct vm_area_desc *desc,
 				     vma_desc_pages(desc));
 }
 
+static inline
+void mmap_action_map_discontig_kernel_pages(struct vm_area_desc *desc,
+		void *init_private, const struct discontig_kernel_page_ops *ops)
+{
+	struct mmap_action *action = &desc->action;
+
+	action->type = MMAP_DISCONTIG_KERNEL_PAGES;
+	action->map_kernel_discontig.init_private = init_private;
+	action->map_kernel_discontig.ops = ops;
+}
+
 int mmap_action_prepare(struct vm_area_desc *desc);
 int mmap_action_complete(struct vm_area_struct *vma,
 			 struct mmap_action *action, bool is_compat);
 
+static inline void
+discontig_kernel_map_abort(struct discontig_kernel_page_state *state)
+{
+	state->action = DISCONTIG_KERNEL_PAGE_ABORT;
+}
+
+static inline void
+discontig_kernel_map_page(struct discontig_kernel_page_state *state,
+		struct page *page)
+{
+	struct folio *folio = page_folio(page);
+	unsigned long nr_pages = 1;
+
+	if (folio_test_large(folio)) {
+		VM_WARN_ON_ONCE(page != folio_page(folio, 0));
+		nr_pages = folio_nr_pages(folio);
+	}
+
+	state->action = DISCONTIG_KERNEL_PAGE_MAP_CONTIG_PAGES;
+	state->__page = page;
+	state->__nr_pages = min(nr_pages, state->nr_pages_remain);
+}
+
+static inline void
+discontig_kernel_map_pages(struct discontig_kernel_page_state *state,
+		struct page **page_arr, unsigned long nr_pages)
+{
+	state->action = DISCONTIG_KERNEL_PAGE_MAP_PAGES;
+	state->__page_arr = page_arr;
+	state->__nr_pages = min(nr_pages, state->nr_pages_remain);
+}
+
 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
 				unsigned long vm_start, unsigned long vm_end)
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 9ca2ea3664bc..07efb336bacb 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -818,8 +818,42 @@ enum mmap_action_type {
 	MMAP_NOTHING,
 	MMAP_REMAP_PFN,
 	MMAP_IO_REMAP_PFN,
-	MMAP_SIMPLE_IO_REMAP,	/* I/O remap with guardrails. */
-	MMAP_KERNEL_PAGES,	/* Map kernel page range from array. */
+	MMAP_SIMPLE_IO_REMAP,		/* I/O remap with guardrails. */
+	MMAP_KERNEL_PAGES,		/* Map kernel page range from array. */
+	MMAP_DISCONTIG_KERNEL_PAGES,	/* Map kernel discontig page range. */
+};
+
+enum discontig_kernel_page_action {
+	DISCONTIG_KERNEL_PAGE_ABORT,
+	DISCONTIG_KERNEL_PAGE_MAP_PAGES,	/* Array of individual pages. */
+	DISCONTIG_KERNEL_PAGE_MAP_CONTIG_PAGES,	/* 1..N contiguous pages. */
+};
+
+struct discontig_kernel_page_state {
+	/* Map state. */
+	const unsigned long start;	/* Start address of VMA. */
+	const unsigned long end;	/* End address of VMA. */
+	unsigned long addr;		/* The current address to be mapped. */
+	pgoff_t pgoff;			/* The current pgoff to be mapped. */
+	unsigned long nr_pages_mapped;	/* The number of pages mapped. */
+	unsigned long nr_pages_remain;	/* The number of pages remaining. */
+
+	/* User-defined state. */
+	void *vm_private_data;		/* VMA private data. */
+	void *private;			/* Mapping private data. */
+
+	/* Users should not touch these, use discontig_kernel_map_*() helpers. */
+	enum discontig_kernel_page_action action;
+	union {
+		struct page *__page;
+		struct page **__page_arr;
+	};
+	unsigned long __nr_pages;
+};
+
+struct discontig_kernel_page_ops {
+	int (*init)(void *vm_private_data, void **private);
+	int (*get)(struct discontig_kernel_page_state *state);
 };
 
 /*
@@ -844,6 +878,10 @@ struct mmap_action {
 			unsigned long nr_pages;
 			pgoff_t pgoff;
 		} map_kernel;
+		struct {
+			void *init_private;
+			const struct discontig_kernel_page_ops *ops;
+		} map_kernel_discontig;
 	};
 	enum mmap_action_type type;
 
diff --git a/mm/internal.h b/mm/internal.h
index fc4e249e2343..557e629db8c5 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -1515,6 +1515,9 @@ int simple_ioremap_prepare(struct vm_area_desc *desc);
 int map_kernel_pages_prepare(struct vm_area_desc *desc);
 int map_kernel_pages_complete(struct vm_area_struct *vma,
 			      struct mmap_action *action);
+int map_discontig_kernel_pages_prepare(struct vm_area_desc *desc);
+int map_discontig_kernel_pages_complete(struct vm_area_struct *vma,
+					struct mmap_action *action);
 
 static inline int io_remap_pfn_range_prepare(struct vm_area_desc *desc)
 {
diff --git a/mm/memory.c b/mm/memory.c
index a35cca09eb71..1a6bb0aef8cf 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2610,17 +2610,23 @@ int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
 }
 EXPORT_SYMBOL(vm_insert_pages);
 
+static void __map_kernel_pages_prepare(struct vm_area_desc *desc)
+{
+	if (vma_desc_test(desc, VMA_MIXEDMAP_BIT))
+		return;
+
+	mmap_assert_write_locked(desc->mm);
+	VM_WARN_ON_ONCE(vma_desc_test(desc, VMA_PFNMAP_BIT));
+	vma_desc_set_flags(desc, VMA_MIXEDMAP_BIT);
+}
+
 int map_kernel_pages_prepare(struct vm_area_desc *desc)
 {
 	const struct mmap_action *action = &desc->action;
 	const unsigned long addr = action->map_kernel.start;
 	unsigned long nr_pages, end;
 
-	if (!vma_desc_test(desc, VMA_MIXEDMAP_BIT)) {
-		VM_WARN_ON_ONCE(mmap_read_trylock(desc->mm));
-		VM_WARN_ON_ONCE(vma_desc_test(desc, VMA_PFNMAP_BIT));
-		vma_desc_set_flags(desc, VMA_MIXEDMAP_BIT);
-	}
+	__map_kernel_pages_prepare(desc);
 
 	nr_pages = action->map_kernel.nr_pages;
 	end = addr + PAGE_SIZE * nr_pages;
@@ -2641,6 +2647,96 @@ int map_kernel_pages_complete(struct vm_area_struct *vma,
 			    &nr_pages, vma->vm_page_prot);
 }
 
+int map_discontig_kernel_pages_prepare(struct vm_area_desc *desc)
+{
+	const struct mmap_action *action = &desc->action;
+	const struct discontig_kernel_page_ops *ops =
+		action->map_kernel_discontig.ops;
+
+	/* At minimum need to be able to get pages. */
+	if (WARN_ON_ONCE(!ops || !ops->get))
+		return -EINVAL;
+
+	__map_kernel_pages_prepare(desc);
+	return 0;
+}
+
+static int apply_discontig_action(struct vm_area_struct *vma,
+		struct discontig_kernel_page_state *state)
+{
+	unsigned long nr_pages = state->__nr_pages;
+	unsigned long addr = state->addr;
+	unsigned long i;
+
+	if (state->action == DISCONTIG_KERNEL_PAGE_MAP_PAGES)
+		return insert_pages(vma, addr, state->__page_arr,
+				&nr_pages, vma->vm_page_prot);
+
+	VM_WARN_ON_ONCE(state->action != DISCONTIG_KERNEL_PAGE_MAP_CONTIG_PAGES);
+	/* Contiguous pages - have to iterate through each one. */
+	for (i = 0; i < nr_pages; i++, addr += PAGE_SIZE) {
+		struct page *page = state->__page + i;
+		int err;
+
+		err = insert_page(vma, addr, page, vma->vm_page_prot,
+				/*mkwrite=*/false);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+int map_discontig_kernel_pages_complete(struct vm_area_struct *vma,
+					struct mmap_action *action)
+{
+	const struct discontig_kernel_page_ops *ops =
+		action->map_kernel_discontig.ops;
+	struct discontig_kernel_page_state state = {
+		.start = vma->vm_start,
+		.end = vma->vm_end,
+		.addr = vma->vm_start,
+		.pgoff = vma->vm_pgoff,
+		.nr_pages_mapped = 0,
+		.nr_pages_remain = vma_pages(vma),
+		.vm_private_data = vma->vm_private_data,
+		.private = action->map_kernel_discontig.init_private,
+	};
+	int err = 0;
+
+	if (ops->init)
+		err = ops->init(vma->vm_private_data, &state.private);
+	if (err)
+		return err;
+
+	do {
+		unsigned long end, pgoff_end;
+		unsigned long nr_pages;
+
+		/* Default to abort. */
+		state.action = DISCONTIG_KERNEL_PAGE_ABORT;
+		err = ops->get(&state);
+		if (err || state.action == DISCONTIG_KERNEL_PAGE_ABORT)
+			return err;
+		nr_pages = state.__nr_pages;
+
+		if (!nr_pages || nr_pages > state.nr_pages_remain)
+			return -EINVAL;
+		end = state.addr + PAGE_SIZE * nr_pages;
+		pgoff_end = state.pgoff + nr_pages;
+
+		err = apply_discontig_action(vma, &state);
+		if (err)
+			return err;
+
+		state.addr = end;
+		state.pgoff = pgoff_end;
+		state.nr_pages_mapped += nr_pages;
+		state.nr_pages_remain -= nr_pages;
+	} while (state.addr < vma->vm_end);
+
+	return 0;
+}
+
 /**
  * vm_insert_page - insert single page into user vma
  * @vma: user vma to map to
diff --git a/mm/util.c b/mm/util.c
index 1660910e2a22..5b2e1ab4aeba 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -1401,6 +1401,15 @@ static int call_vma_mapped(struct vm_area_struct *vma)
 	return 0;
 }
 
+/* An mmap action failed under the compatibility layer - unmap and close. */
+static void compat_mmap_action_abort(struct vm_area_struct *vma)
+{
+#ifdef CONFIG_MMU
+	zap_vma_range(vma, vma->vm_start, vma_pages(vma) << PAGE_SHIFT);
+#endif
+	vma_close(vma);
+}
+
 static int mmap_action_finish(struct vm_area_struct *vma,
 			      struct mmap_action *action, int err,
 			      bool is_compat)
@@ -1412,12 +1421,13 @@ static int mmap_action_finish(struct vm_area_struct *vma,
 
 	/* do_munmap() might take rmap lock, so release if held. */
 	maybe_rmap_unlock_action(vma, action);
-	/*
-	 * If this is invoked from the compatibility layer, post-mmap() hook
-	 * logic will handle cleanup for us.
-	 */
-	if (!err || is_compat)
+	if (!err)
+		return 0;
+
+	if (is_compat) {
+		compat_mmap_action_abort(vma);
 		return err;
+	}
 
 	/*
 	 * If an error occurs, unmap the VMA altogether and return an error. We
@@ -1469,6 +1479,8 @@ int mmap_action_prepare(struct vm_area_desc *desc)
 		return simple_ioremap_prepare(desc);
 	case MMAP_KERNEL_PAGES:
 		return map_kernel_pages_prepare(desc);
+	case MMAP_DISCONTIG_KERNEL_PAGES:
+		return map_discontig_kernel_pages_prepare(desc);
 	}
 
 	WARN_ON_ONCE(1);
@@ -1501,6 +1513,9 @@ int mmap_action_complete(struct vm_area_struct *vma,
 	case MMAP_KERNEL_PAGES:
 		err = map_kernel_pages_complete(vma, action);
 		break;
+	case MMAP_DISCONTIG_KERNEL_PAGES:
+		err = map_discontig_kernel_pages_complete(vma, action);
+		break;
 	case MMAP_IO_REMAP_PFN:
 	case MMAP_SIMPLE_IO_REMAP:
 		/* Should have been delegated. */
@@ -1522,6 +1537,7 @@ int mmap_action_prepare(struct vm_area_desc *desc)
 	case MMAP_IO_REMAP_PFN:
 	case MMAP_SIMPLE_IO_REMAP:
 	case MMAP_KERNEL_PAGES:
+	case MMAP_DISCONTIG_KERNEL_PAGES:
 		WARN_ON_ONCE(1); /* nommu cannot handle these. */
 		break;
 	}
@@ -1543,6 +1559,7 @@ int mmap_action_complete(struct vm_area_struct *vma,
 	case MMAP_IO_REMAP_PFN:
 	case MMAP_SIMPLE_IO_REMAP:
 	case MMAP_KERNEL_PAGES:
+	case MMAP_DISCONTIG_KERNEL_PAGES:
 		WARN_ON_ONCE(1); /* nommu cannot handle this. */
 
 		err = -EINVAL;
diff --git a/tools/testing/vma/include/dup.h b/tools/testing/vma/include/dup.h
index 023ad6092e9c..a6a8e53c4b38 100644
--- a/tools/testing/vma/include/dup.h
+++ b/tools/testing/vma/include/dup.h
@@ -457,14 +457,17 @@ enum mmap_action_type {
 	MMAP_NOTHING,
 	MMAP_REMAP_PFN,
 	MMAP_IO_REMAP_PFN,
-	MMAP_SIMPLE_IO_REMAP,	/* I/O remap with guardrails. */
-	MMAP_KERNEL_PAGES,	/* Map kernel page range from array. */
+	MMAP_SIMPLE_IO_REMAP,		/* I/O remap with guardrails. */
+	MMAP_KERNEL_PAGES,		/* Map kernel page range from array. */
+	MMAP_DISCONTIG_KERNEL_PAGES,	/* Map kernel discontig page range. */
 };
 
 /*
  * Describes an action an mmap_prepare hook can instruct to be taken to complete
  * the mapping of a VMA. Specified in vm_area_desc.
  */
+struct discontig_kernel_page_ops;
+
 struct mmap_action {
 	union {
 		struct {
@@ -483,6 +486,10 @@ struct mmap_action {
 			unsigned long nr_pages;
 			pgoff_t pgoff;
 		} map_kernel;
+		struct {
+			void *init_private;
+			const struct discontig_kernel_page_ops *ops;
+		} map_kernel_discontig;
 	};
 	enum mmap_action_type type;
 

-- 
2.55.0

Re: [PATCH v6 08/39] mm: add mmap action for discontiguous kernel page mapping

[David Hildenbrand (Arm)]

On 10/13/26 17:04, Lorenzo Stoakes (ARM) wrote:
> The existing kernel page mapping mmap actions allow for partial and full
> mapping of an array of struct page pointers.
> 
> However some drivers require the mapping of discontiguous ranges. Permit
> this by providing discontig_kernel_page_ops which allows a driver to
> specify how the operation should begin and how batches of pages should be
> retrieved.
> 
> It uses the minimum exposed interface to do so, providing address, page
> offset and both vm_private_data state and a local private state object.
> 
> ops->init can establish state for the operation, and ops->get outputs the
> pages to map and their count. Should an error arise the core unmaps the
> VMA, invoking vm_ops->close, which is therefore where any state established
> by ops->init is released.
> 
> Also update the compatibility layer to do this too via
> compat_mmap_action_abort().
> 
> Batches may not exceed the VMA, but may map less than its full range in
> case the driver wishes to allow the user to map an area larger than the
> available data.
> 
> To use it, users invoke mmap_action_map_discontig_kernel_pages() with
> initial local private state and a set of operations.
> 
> Users can then use one of the provided helper functions to perform an
> action:
> 
> * discontig_kernel_map_abort() - Abort and leave the mapping as it has
>   been accumulated so far.
> * discontig_kernel_map_page() - Map a single page, or a compound page given
>   its head page, as 1..N contiguous pages.
> * discontig_kernel_map_pages() - Map a struct page ** array of a specified
>   count, mapping only those pages which fit in the VMA.
> 
> The userland VMA tests are updated accordingly.
> 
> Acked-by: Zi Yan <ziy@nvidia.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
> ---


[...]

> +
> +int map_discontig_kernel_pages_complete(struct vm_area_struct *vma,
> +					struct mmap_action *action)

... indentation ;)

Acked-by: David Hildenbrand (Arm) <david@kernel.org>

-- 
Cheers,

David

Re: [PATCH v6 08/39] mm: add mmap action for discontiguous kernel page mapping

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] mm: mmap_action_finish bypasses error_override for compat tasks
--

commit 21d3d0c977ac349611c733fc3c288ae67e5f71a1
Author: Lorenzo Stoakes (ARM) <ljs@kernel.org>

mm: add mmap action for discontiguous kernel page mapping

This commit introduces discontig_kernel_page_ops and a new mmap action type
to allow drivers to map discontiguous arrays of kernel pages into a VMA.
It updates the compatibility layer via compat_mmap_action_abort() to properly
handle VMA teardown on mmap action failures.

> diff --git a/mm/util.c b/mm/util.c
> --- a/mm/util.c
> +++ b/mm/util.c

[ ... ]

> @@ -1412,12 +1421,13 @@ static int mmap_action_finish(struct vm_area_struct *vma,
>  
>  	/* do_munmap() might take rmap lock, so release if held. */
>  	maybe_rmap_unlock_action(vma, action);
> -	/*
> -	 * If this is invoked from the compatibility layer, post-mmap() hook
> -	 * logic will handle cleanup for us.
> -	 */
> -	if (!err || is_compat)
> +	if (!err)
> +		return 0;
> +
> +	if (is_compat) {
> +		compat_mmap_action_abort(vma);
>  		return err;

[Severity: Medium]
Does this bypass the error_override logic for compat tasks?

When an mmap action fails, the mmap_action struct provides an error_override
field to allow drivers to replace the internal error code with a user-facing
one. Later in this function, non-compat tasks return 
action->error_override ?: err.

By returning err directly here when is_compat is true, it seems 32-bit compat
tasks might receive different error codes than 64-bit tasks, potentially
breaking ABI consistency.

> +	}
>  
>  	/*
>  	 * If an error occurs, unmap the VMA altogether and return an error. We

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=8

[PATCH v6 09/39] docs: filesystems: update mmap_prepare docs for discontig kernel pgs

[Lorenzo Stoakes (ARM)]

Describe the newly introduced discontiguous kernel page mapping mechanism,
detailing how to use it sensibly and how the API looks.

Explicitly detail the various discontiguous actions available and how to
use them.

Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 Documentation/filesystems/mmap_prepare.rst | 81 ++++++++++++++++++++++++++++++
 1 file changed, 81 insertions(+)

diff --git a/Documentation/filesystems/mmap_prepare.rst b/Documentation/filesystems/mmap_prepare.rst
index 82c99c95ad85..b5afbb33e5c8 100644
--- a/Documentation/filesystems/mmap_prepare.rst
+++ b/Documentation/filesystems/mmap_prepare.rst
@@ -164,5 +164,86 @@ pointer. These are:
   sufficient entries in the page array to cover the entire range of the
   described VMA.
 
+* mmap_action_map_discontig_kernel_pages() - Maps a discontiguous range of
+  `struct page` pointers over the VMA. They must span from the start of the VMA,
+  but may terminate prior to the end (leaving the remainder unmapped).
+
 **NOTE:** The ``action`` field should never normally be manipulated directly,
 rather you ought to use one of these helpers.
+
+Discontiguous Actions
+=====================
+
+Some actions can be performed across discontiguous ranges.
+
+Map kernel pages
+----------------
+
+To map kernel pages discontiguously, you must provide hooks using ``struct
+discontig_kernel_page_ops``:
+
+.. code-block:: C
+
+    struct discontig_kernel_page_ops {
+        int (*init)(void *vm_private_data, void **private);
+        int (*get)(struct discontig_kernel_page_state *state);
+    };
+
+The ``init`` hook is optional and allows state to be established before the
+operation starts, for instance taking a reference count. Nothing is invoked
+after the operation, so ``init`` must not leave locks held, and state that must
+be released once the mapping goes away should be released in
+``vm_ops->close``.
+
+The ``init`` hook, if provided, is invoked prior to the operation starting. It
+may update what is pointed to by ``vm_private_data`` and/or ``private``. If an
+error is returned, then the operation is aborted. The ``private`` field can be
+reassigned.
+
+**NOTE:** The operation may sleep between invocations of ``get``, so locks
+needed to stabilise state must be taken and released within each hook.
+
+The ``get`` handler is the key means through which the operation is
+executed. The current state of the operation is provided through ``struct
+discontig_kernel_page_state``:
+
+.. code-block:: C
+
+    struct discontig_kernel_page_state {
+        /* Map state. */
+        unsigned long start;            /* Start address of VMA. */
+        unsigned long end;              /* End address of VMA. */
+        unsigned long addr;             /* The current address to be mapped. */
+        pgoff_t pgoff;                  /* The current pgoff to be mapped. */
+        unsigned long nr_pages_mapped;  /* The number of pages mapped. */
+        unsigned long nr_pages_remain;  /* The number of pages remaining. */
+
+        /* User-defined state. */
+        void *vm_private_data;          /* VMA private data. */
+        void *private;                  /* Mapping private data. */
+
+        /* Users should not touch these, use discontig_kernel_map_*() helpers. */
+        ... internal fields ...
+    };
+
+With ``private`` being an additional user-controllable state variable,
+initialised via ``mmap_action_map_discontig_kernel_pages()``, and
+``vm_private_data`` being equal to the ``desc->private_data`` field set in
+the ``mmap_prepare()`` hook.
+
+In the ``get`` hook, the user must choose how to map kernel pages:
+
+* ``discontig_kernel_map_abort()`` - Call this to abort the operation, whatever
+  has been mapped so far will be retained, the rest of the mapping will SIGBUS
+  if accessed.
+* ``discontig_kernel_map_page()`` - Maps a single page, correctly handling
+  compound pages (if the compound page is bigger than the remaining pages in the
+  VMA, then only those pages that fit will be mapped). For a compound page, the
+  head page must be passed.
+* ``discontig_kernel_map_pages()`` - Map an array of pages of a specified
+  size. As with a compound page, if more pages are specified than remain in
+  the VMA then only those that fit will be mapped.
+
+If an error arises after ``init`` succeeded, the core unmaps the VMA, invoking
+``vm_ops->close`` if set, which is therefore the place to release any state
+that ``init`` established.

-- 
2.55.0

Re: [PATCH v6 09/39] docs: filesystems: update mmap_prepare docs for discontig kernel pgs

[David Hildenbrand (Arm)]

On 10/13/26 17:04, Lorenzo Stoakes (ARM) wrote:
> Describe the newly introduced discontiguous kernel page mapping mechanism,
> detailing how to use it sensibly and how the API looks.
> 
> Explicitly detail the various discontiguous actions available and how to
> use them.
> 
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
> ---
>  Documentation/filesystems/mmap_prepare.rst | 81 ++++++++++++++++++++++++++++++
>  1 file changed, 81 insertions(+)
> 
> diff --git a/Documentation/filesystems/mmap_prepare.rst b/Documentation/filesystems/mmap_prepare.rst
> index 82c99c95ad85..b5afbb33e5c8 100644
> --- a/Documentation/filesystems/mmap_prepare.rst
> +++ b/Documentation/filesystems/mmap_prepare.rst
> @@ -164,5 +164,86 @@ pointer. These are:
>    sufficient entries in the page array to cover the entire range of the
>    described VMA.
>  
> +* mmap_action_map_discontig_kernel_pages() - Maps a discontiguous range of
> +  `struct page` pointers over the VMA. They must span from the start of the VMA,
> +  but may terminate prior to the end (leaving the remainder unmapped).
> +
>  **NOTE:** The ``action`` field should never normally be manipulated directly,
>  rather you ought to use one of these helpers.
> +
> +Discontiguous Actions
> +=====================
> +
> +Some actions can be performed across discontiguous ranges.
> +
> +Map kernel pages
> +----------------
> +
> +To map kernel pages discontiguously, you must provide hooks using ``struct
> +discontig_kernel_page_ops``:
> +
> +.. code-block:: C
> +
> +    struct discontig_kernel_page_ops {
> +        int (*init)(void *vm_private_data, void **private);
> +        int (*get)(struct discontig_kernel_page_state *state);
> +    };
> +
> +The ``init`` hook is optional and allows state to be established before the
> +operation starts, for instance taking a reference count. Nothing is invoked
> +after the operation, so ``init`` must not leave locks held, and state that must
> +be released once the mapping goes away should be released in
> +``vm_ops->close``.
> +
> +The ``init`` hook, if provided, is invoked prior to the operation starting. It
> +may update what is pointed to by ``vm_private_data`` and/or ``private``. If an
> +error is returned, then the operation is aborted. The ``private`` field can be
> +reassigned.
> +
> +**NOTE:** The operation may sleep between invocations of ``get``, so locks
> +needed to stabilise state must be taken and released within each hook.
> +
> +The ``get`` handler is the key means through which the operation is
> +executed. The current state of the operation is provided through ``struct
> +discontig_kernel_page_state``:
> +
> +.. code-block:: C
> +
> +    struct discontig_kernel_page_state {
> +        /* Map state. */
> +        unsigned long start;            /* Start address of VMA. */
> +        unsigned long end;              /* End address of VMA. */
> +        unsigned long addr;             /* The current address to be mapped. */
> +        pgoff_t pgoff;                  /* The current pgoff to be mapped. */
> +        unsigned long nr_pages_mapped;  /* The number of pages mapped. */
> +        unsigned long nr_pages_remain;  /* The number of pages remaining. */
> +
> +        /* User-defined state. */
> +        void *vm_private_data;          /* VMA private data. */
> +        void *private;                  /* Mapping private data. */
> +
> +        /* Users should not touch these, use discontig_kernel_map_*() helpers. */
> +        ... internal fields ...
> +    };
> +
> +With ``private`` being an additional user-controllable state variable,
> +initialised via ``mmap_action_map_discontig_kernel_pages()``, and
> +``vm_private_data`` being equal to the ``desc->private_data`` field set in
> +the ``mmap_prepare()`` hook.
> +
> +In the ``get`` hook, the user must choose how to map kernel pages:
> +
> +* ``discontig_kernel_map_abort()`` - Call this to abort the operation, whatever
> +  has been mapped so far will be retained, the rest of the mapping will SIGBUS
> +  if accessed.
> +* ``discontig_kernel_map_page()`` - Maps a single page, correctly handling

"single page" can be misleading with compound pages.

"Map an entire (compound) page" or sth like that would be clearer.

Acked-by: David Hildenbrand (Arm) <david@kernel.org>

-- 
Cheers,

David

Re: [PATCH v6 09/39] docs: filesystems: update mmap_prepare docs for discontig kernel pgs

[sashiko-bot]

> Describe the newly introduced discontiguous kernel page mapping mechanism,
> detailing how to use it sensibly and how the API looks.
> 
> Explicitly detail the various discontiguous actions available and how to
> use them.
> 
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=9

[PATCH v6 10/39] drivers/usb/mon: update to use mmap_prepare + map kernel pages

[Lorenzo Stoakes (ARM)]

Replace the deprecated .mmap hook with its replacement .mmap_prepare. As
part of this change, additionally take the approach of mapping pages upon
mmap rather than providing a fault handler.

The page span cannot be mutated when an mmap mapping is in place, so this
is safe to do in advance (the MON_IOCT_RING_SIZE ioctl operation exits
-EBUSY if it's attempted, gated by the rp->mmap_active reference count).

Utilise the newly introduced mmap_action_map_discontig_kernel_pages() to do
this, which allows for iteration over pages in mon_bin_discontig_get().

mon_bin_discontig_init() increments the rp->mmap_active reference count to
stabilise page spans. Should an error arise the core unmaps the VMA and
mon_bin_vma_close() drops the reference again.

The vm_ops->close hook implemented in mon_bin_vma_close() will ensure
correct reference count arithmetic upon unmap (with mon_bin_vma_open()
accounting for splitting).

The existing semantics are all retained, including not mapping past the
range of available pages, with a SIGBUS being raised in a userland process
that attempts to access past this point.

Ultimately insert_page() is invoked to insert each page, which increments
the reference count on each mapped page. This mimics what was being done
previously, only we pre-map the entire range rather than doing so on
demand.

The existing fault handler did nothing that required demand paging, and was
presumably implemented this way for historical reasons.

One behavioural difference: pages are no longer faulted in on demand, so a
page discarded with MADV_DONTNEED is not repopulated and a subsequent
access raises SIGBUS, as with other pre-populated kernel mappings.

Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 drivers/usb/mon/mon_bin.c | 82 ++++++++++++++++++++++++++++++-----------------
 1 file changed, 53 insertions(+), 29 deletions(-)

diff --git a/drivers/usb/mon/mon_bin.c b/drivers/usb/mon/mon_bin.c
index 687f6a8981f3..9d00b21a8153 100644
--- a/drivers/usb/mon/mon_bin.c
+++ b/drivers/usb/mon/mon_bin.c
@@ -1219,6 +1219,15 @@ mon_bin_poll(struct file *file, struct poll_table_struct *wait)
 	return mask;
 }
 
+static void __mon_bin_vma_open(struct mon_reader_bin *rp)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&rp->b_lock, flags);
+	rp->mmap_active++;
+	spin_unlock_irqrestore(&rp->b_lock, flags);
+}
+
 /*
  * open and close: just keep track of how many times the device is
  * mapped, to use the proper memory allocation function.
@@ -1226,64 +1235,79 @@ mon_bin_poll(struct file *file, struct poll_table_struct *wait)
 static void mon_bin_vma_open(struct vm_area_struct *vma)
 {
 	struct mon_reader_bin *rp = vma->vm_private_data;
-	unsigned long flags;
 
-	spin_lock_irqsave(&rp->b_lock, flags);
-	rp->mmap_active++;
-	spin_unlock_irqrestore(&rp->b_lock, flags);
+	__mon_bin_vma_open(rp);
 }
 
-static void mon_bin_vma_close(struct vm_area_struct *vma)
+static void __mon_bin_vma_close(struct mon_reader_bin *rp)
 {
 	unsigned long flags;
 
-	struct mon_reader_bin *rp = vma->vm_private_data;
 	spin_lock_irqsave(&rp->b_lock, flags);
 	rp->mmap_active--;
 	spin_unlock_irqrestore(&rp->b_lock, flags);
 }
 
-/*
- * Map ring pages to user space.
- */
-static vm_fault_t mon_bin_vma_fault(struct vm_fault *vmf)
+static void mon_bin_vma_close(struct vm_area_struct *vma)
 {
-	struct mon_reader_bin *rp = vmf->vma->vm_private_data;
+	struct mon_reader_bin *rp = vma->vm_private_data;
+
+	__mon_bin_vma_close(rp);
+}
+
+static const struct vm_operations_struct mon_bin_vm_ops = {
+	.open =     mon_bin_vma_open,
+	.close =    mon_bin_vma_close,
+};
+
+static int mon_bin_discontig_init(void *vm_private_data, void **private)
+{
+	struct mon_reader_bin *rp = vm_private_data;
+
+	/* Dropped by mon_bin_vma_close() on unmap, including on error. */
+	__mon_bin_vma_open(rp);
+	return 0;
+}
+
+static int mon_bin_discontig_get(struct discontig_kernel_page_state *state)
+{
+	struct mon_reader_bin *rp = state->vm_private_data;
 	unsigned long offset, chunk_idx;
-	struct page *pageptr;
 	unsigned long flags;
 
 	spin_lock_irqsave(&rp->b_lock, flags);
-	offset = vmf->pgoff << PAGE_SHIFT;
+
+	offset = state->pgoff << PAGE_SHIFT;
 	if (offset >= rp->b_size) {
 		spin_unlock_irqrestore(&rp->b_lock, flags);
-		return VM_FAULT_SIGBUS;
+		discontig_kernel_map_abort(state);
+		return 0;
 	}
 	chunk_idx = offset / CHUNK_SIZE;
-	pageptr = rp->b_vec[chunk_idx].pg;
-	get_page(pageptr);
-	vmf->page = pageptr;
+	discontig_kernel_map_page(state, rp->b_vec[chunk_idx].pg);
+
 	spin_unlock_irqrestore(&rp->b_lock, flags);
 	return 0;
 }
 
-static const struct vm_operations_struct mon_bin_vm_ops = {
-	.open =     mon_bin_vma_open,
-	.close =    mon_bin_vma_close,
-	.fault =    mon_bin_vma_fault,
+static const struct discontig_kernel_page_ops mon_discontig_ops = {
+	.init = mon_bin_discontig_init,
+	.get = mon_bin_discontig_get,
 };
 
-static int mon_bin_mmap(struct file *filp, struct vm_area_struct *vma)
+static int mon_bin_mmap_prepare(struct vm_area_desc *desc)
 {
-	/* don't do anything here: "fault" will set up page table entries */
-	vma->vm_ops = &mon_bin_vm_ops;
+	const struct file *filp = desc->file;
 
-	if (vma->vm_flags & VM_WRITE)
+	if (vma_desc_test(desc, VMA_WRITE_BIT))
 		return -EPERM;
 
-	vm_flags_mod(vma, VM_DONTEXPAND | VM_DONTDUMP, VM_MAYWRITE);
-	vma->vm_private_data = filp->private_data;
-	mon_bin_vma_open(vma);
+	desc->vm_ops = &mon_bin_vm_ops;
+	vma_desc_clear_flags(desc, VMA_MAYWRITE_BIT);
+	vma_desc_set_flags(desc, VMA_DONTEXPAND_BIT, VMA_DONTDUMP_BIT);
+	desc->private_data = filp->private_data;
+
+	mmap_action_map_discontig_kernel_pages(desc, NULL, &mon_discontig_ops);
 	return 0;
 }
 
@@ -1298,7 +1322,7 @@ static const struct file_operations mon_fops_binary = {
 	.compat_ioctl =	mon_bin_compat_ioctl,
 #endif
 	.release =	mon_bin_release,
-	.mmap =		mon_bin_mmap,
+	.mmap_prepare = mon_bin_mmap_prepare,
 };
 
 static int mon_bin_wait_event(struct file *file, struct mon_reader_bin *rp)

-- 
2.55.0

Re: [PATCH v6 10/39] drivers/usb/mon: update to use mmap_prepare + map kernel pages

[sashiko-bot]

> Replace the deprecated .mmap hook with its replacement .mmap_prepare. As
> part of this change, additionally take the approach of mapping pages upon
> mmap rather than providing a fault handler.
> 
> The page span cannot be mutated when an mmap mapping is in place, so this
> is safe to do in advance (the MON_IOCT_RING_SIZE ioctl operation exits
> -EBUSY if it's attempted, gated by the rp->mmap_active reference count).
> 
> Utilise the newly introduced mmap_action_map_discontig_kernel_pages() to do
> this, which allows for iteration over pages in mon_bin_discontig_get().
> 
> mon_bin_discontig_init() increments the rp->mmap_active reference count to
> stabilise page spans. Should an error arise the core unmaps the VMA and
> mon_bin_vma_close() drops the reference again.
> 
> [ ... ]
> access raises SIGBUS, as with other pre-populated kernel mappings.
> 
> Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=10

[PATCH v6 11/39] infiniband: update hfi1 to use remap_vmalloc_range()

[Lorenzo Stoakes (ARM)]

In cases which map chip memory from vmalloc()'d ranges, the hfi1 infiniband
drivers currently installs a fault handler, and then smuggles the kernel
virtual address of this range in vma->vm_pgoff.

This is exposing KASLR-sensitive internal kernel state in the VMA, and is
entirely unnecessary.

Instead, use remap_vmalloc_range() to remap the VMA to the span, and
eliminate the fault handler altogether.

remap_vmalloc_range() checks that the VMA does not extend beyond the
vmalloc area, and the driver already requires the VMA to exactly match the
span of the memory being mapped, so this has no impact.

The memory is all preallocated so not having a fault handler has no impact
either, other than pre-mapping the ranges which is beneficial.

We also remove the VM_IO flag as it's not appropriate here, and the
VM_DONTEXPAND flag as remap_vmalloc_range() will set it (and also mark the
range correctly as a mixed map).

We also update the vmalloc paths to place the virtual kernel address in
memvirt, rather than overloading the physical address memaddr. We predicate
the vmalloc handling on the vmalloc flag before we check memvirt for the
virtual address-derived PFN remap path, so this works fine.

remap_vmalloc_range() requires that the vmalloc()'d areas were all
allocated using vmalloc_user() - each of cq->comps,
uctxt->subctxt_rcvegrbuf, uctxt->subctxt_rcvhdr_base,
uctxt->subctxt_uregbase and dd->events were allocated this way, so that
requirement is satisfied.

We also remove VM_IO and VM_DONTEXPAND from the STATUS command, as these
are both set on remap.

Finally, we remove VM_DONTEXPAND from the PIO_BUFS, PIO_BUFS_SOP and UREGS
commands, as these are also all set on remap. PIO_CRED retains it, as
dma_mmap_coherent() may map via vm_insert_page() on the IOMMU-DMA path,
which sets only VM_MIXEDMAP. The RCV_HDRQ, RCV_EGRBUF and RTAIL commands
also map via dma_mmap_coherent() and never set VM_DONTEXPAND, so set it for
them for the same reason.

Note that we retain expected behaviour throughout - the vmalloc remapped
ranges set VM_MIXEDMAP | VM_DONTDUMP | VM_DONTEXPAND for each range.

VM_IO was never appropriate as the ranges are explicitly not MMIO, and the
reference to the v3.7 VM_RESERVED semantics map on to VM_MIXEDMAP |
VM_DONTDUMP | VM_DONTEXPAND correctly - no core dump, unmergeable, no
normal vm page for purposes of reclaim/migration/etc.

There is a change in behaviour in that pages mapped using
remap_vmalloc_range() will now have normal GUP-able pages, however this
should have no impact as there is no reason not to allow this.

Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 drivers/infiniband/hw/hfi1/file_ops.c | 83 +++++++++++------------------------
 1 file changed, 25 insertions(+), 58 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/file_ops.c b/drivers/infiniband/hw/hfi1/file_ops.c
index dc548e6802e2..b02d1f1dbb27 100644
--- a/drivers/infiniband/hw/hfi1/file_ops.c
+++ b/drivers/infiniband/hw/hfi1/file_ops.c
@@ -70,7 +70,6 @@ static int set_ctxt_pkey(struct hfi1_ctxtdata *uctxt, unsigned long arg);
 static int ctxt_reset(struct hfi1_ctxtdata *uctxt);
 static int manage_rcvq(struct hfi1_ctxtdata *uctxt, u16 subctxt,
 		       unsigned long arg);
-static vm_fault_t vma_fault(struct vm_fault *vmf);
 static long hfi1_file_ioctl(struct file *fp, unsigned int cmd,
 			    unsigned long arg);
 
@@ -85,10 +84,6 @@ static const struct file_operations hfi1_file_ops = {
 	.llseek = noop_llseek,
 };
 
-static const struct vm_operations_struct vm_ops = {
-	.fault = vma_fault,
-};
-
 /*
  * Types of memories mapped into user processes' space
  */
@@ -304,13 +299,13 @@ static ssize_t hfi1_write_iter(struct kiocb *kiocb, struct iov_iter *from)
 	return reqs;
 }
 
-static inline void mmap_cdbg(u16 ctxt, u8 subctxt, u8 type, u8 mapio, u8 vmf,
+static inline void mmap_cdbg(u16 ctxt, u8 subctxt, u8 type, u8 mapio, u8 is_vmalloc,
 			     u64 memaddr, void *memvirt, dma_addr_t memdma,
 			     ssize_t memlen, struct vm_area_struct *vma)
 {
 	hfi1_cdbg(PROC,
-		  "%u:%u type:%u io/vf/dma:%d/%d/%d, addr:0x%llx, len:%lu(%lu), flags:0x%lx",
-		  ctxt, subctxt, type, mapio, vmf, !!memdma,
+		  "%u:%u type:%u io/vmalloc/dma:%d/%d/%d, addr:0x%llx, len:%lu(%lu), flags:0x%lx",
+		  ctxt, subctxt, type, mapio, is_vmalloc, !!memdma,
 		  memaddr ?: (u64)memvirt, memlen,
 		  vma->vm_end - vma->vm_start, vma->vm_flags);
 }
@@ -325,7 +320,7 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 		memaddr = 0;
 	void *memvirt = NULL;
 	dma_addr_t memdma = 0;
-	u8 subctxt, mapio = 0, vmf = 0, type;
+	u8 subctxt, mapio = 0, is_vmalloc = 0, type;
 	ssize_t memlen = 0;
 	int ret = 0;
 	u16 ctxt;
@@ -347,7 +342,7 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 	/*
 	 * vm_pgoff is used as a buffer selector cookie.  Always mmap from
 	 * the beginning.
-	 */ 
+	 */
 	vma->vm_pgoff = 0;
 	flags = vma->vm_flags;
 
@@ -366,7 +361,7 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 		 */
 		memlen = PAGE_ALIGN(uctxt->sc->credits * PIO_BLOCK_SIZE);
 		flags &= ~VM_MAYREAD;
-		flags |= VM_DONTCOPY | VM_DONTEXPAND;
+		flags |= VM_DONTCOPY;
 		vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
 		mapio = 1;
 		break;
@@ -401,6 +396,7 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 		memlen = rcvhdrq_size(uctxt);
 		memvirt = uctxt->rcvhdrq;
 		memdma = uctxt->rcvhdrq_dma;
+		flags |= VM_DONTEXPAND;
 		break;
 	case RCV_EGRBUF: {
 		unsigned long vm_start_save;
@@ -422,7 +418,7 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 			ret = -EPERM;
 			goto done;
 		}
-		vm_flags_clear(vma, VM_MAYWRITE);
+		vm_flags_mod(vma, VM_DONTEXPAND, VM_MAYWRITE);
 		/*
 		 * Mmap multiple separate allocations into a single vma.  From
 		 * here, dma_mmap_coherent() calls dma_direct_mmap(), which
@@ -438,7 +434,7 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 			memvirt = uctxt->egrbufs.buffers[i].addr;
 			memdma = uctxt->egrbufs.buffers[i].dma;
 			vma->vm_end += memlen;
-			mmap_cdbg(ctxt, subctxt, type, mapio, vmf, memaddr,
+			mmap_cdbg(ctxt, subctxt, type, mapio, is_vmalloc, memaddr,
 				  memvirt, memdma, memlen, vma);
 			ret = dma_mmap_coherent(&dd->pcidev->dev, vma,
 						memvirt, memdma, memlen);
@@ -467,7 +463,7 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 		 * user registers.
 		 */
 		memlen = PAGE_SIZE;
-		flags |= VM_DONTCOPY | VM_DONTEXPAND;
+		flags |= VM_DONTCOPY;
 		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
 		mapio = 1;
 		break;
@@ -476,15 +472,10 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 		 * Use the page where this context's flags are. User level
 		 * knows where it's own bitmap is within the page.
 		 */
-		memaddr = (unsigned long)
-			(dd->events + uctxt_offset(uctxt)) & PAGE_MASK;
+		memvirt = dd->events + uctxt_offset(uctxt);
+		memvirt = (void *)(((uintptr_t)memvirt) & PAGE_MASK);
 		memlen = PAGE_SIZE;
-		/*
-		 * v3.7 removes VM_RESERVED but the effect is kept by
-		 * using VM_IO.
-		 */
-		flags |= VM_IO | VM_DONTEXPAND;
-		vmf = 1;
+		is_vmalloc = 1;
 		break;
 	case STATUS:
 		if (flags & VM_WRITE) {
@@ -493,7 +484,6 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 		}
 		memaddr = kvirt_to_phys((void *)dd->status);
 		memlen = PAGE_SIZE;
-		flags |= VM_IO | VM_DONTEXPAND;
 		break;
 	case RTAIL:
 		if (!HFI1_CAP_IS_USET(DMA_RTAIL)) {
@@ -512,25 +502,23 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 		memvirt = (void *)hfi1_rcvhdrtail_kvaddr(uctxt);
 		memdma = uctxt->rcvhdrqtailaddr_dma;
 		flags &= ~VM_MAYWRITE;
+		flags |= VM_DONTEXPAND;
 		break;
 	case SUBCTXT_UREGS:
-		memaddr = (u64)uctxt->subctxt_uregbase;
+		memvirt = uctxt->subctxt_uregbase;
 		memlen = PAGE_SIZE;
-		flags |= VM_IO | VM_DONTEXPAND;
-		vmf = 1;
+		is_vmalloc = 1;
 		break;
 	case SUBCTXT_RCV_HDRQ:
-		memaddr = (u64)uctxt->subctxt_rcvhdr_base;
+		memvirt = uctxt->subctxt_rcvhdr_base;
 		memlen = rcvhdrq_size(uctxt) * uctxt->subctxt_cnt;
-		flags |= VM_IO | VM_DONTEXPAND;
-		vmf = 1;
+		is_vmalloc = 1;
 		break;
 	case SUBCTXT_EGRBUF:
-		memaddr = (u64)uctxt->subctxt_rcvegrbuf;
+		memvirt = uctxt->subctxt_rcvegrbuf;
 		memlen = uctxt->egrbufs.size * uctxt->subctxt_cnt;
-		flags |= VM_IO | VM_DONTEXPAND;
 		flags &= ~VM_MAYWRITE;
-		vmf = 1;
+		is_vmalloc = 1;
 		break;
 	case SDMA_COMP: {
 		struct hfi1_user_sdma_comp_q *cq = fd->cq;
@@ -539,10 +527,9 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 			ret = -EFAULT;
 			goto done;
 		}
-		memaddr = (u64)cq->comps;
+		memvirt = cq->comps;
 		memlen = PAGE_ALIGN(sizeof(*cq->comps) * cq->nentries);
-		flags |= VM_IO | VM_DONTEXPAND;
-		vmf = 1;
+		is_vmalloc = 1;
 		break;
 	}
 	default:
@@ -559,12 +546,10 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 	}
 
 	vm_flags_reset(vma, flags);
-	mmap_cdbg(ctxt, subctxt, type, mapio, vmf, memaddr, memvirt, memdma, 
+	mmap_cdbg(ctxt, subctxt, type, mapio, is_vmalloc, memaddr, memvirt, memdma,
 		  memlen, vma);
-	if (vmf) {
-		vma->vm_pgoff = PFN_DOWN(memaddr);
-		vma->vm_ops = &vm_ops;
-		ret = 0;
+	if (is_vmalloc) {
+		ret = remap_vmalloc_range(vma, memvirt, 0);
 	} else if (memdma) {
 		ret = dma_mmap_coherent(&dd->pcidev->dev, vma,
 					memvirt, memdma, memlen);
@@ -588,24 +573,6 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 	return ret;
 }
 
-/*
- * Local (non-chip) user memory is not mapped right away but as it is
- * accessed by the user-level code.
- */
-static vm_fault_t vma_fault(struct vm_fault *vmf)
-{
-	struct page *page;
-
-	page = vmalloc_to_page((void *)(vmf->pgoff << PAGE_SHIFT));
-	if (!page)
-		return VM_FAULT_SIGBUS;
-
-	get_page(page);
-	vmf->page = page;
-
-	return 0;
-}
-
 static __poll_t hfi1_poll(struct file *fp, struct poll_table_struct *pt)
 {
 	struct hfi1_ctxtdata *uctxt;

-- 
2.55.0

Re: [PATCH v6 11/39] infiniband: update hfi1 to use remap_vmalloc_range()

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] mm: VM_WARN_ON_ONCE triggered by vm_flags_reset() in mmap hook
--

commit a356d47e4beede9d3e032c777130976d06dd5dd8
Author: Lorenzo Stoakes (ARM) <ljs@kernel.org>

    infiniband: update hfi1 to use remap_vmalloc_range()
    
    In cases which map chip memory from vmalloc()'d ranges, the hfi1 infiniband
    drivers currently installs a fault handler, and then smuggles the kernel
    virtual address of this range in vma->vm_pgoff.

> diff --git a/drivers/infiniband/hw/hfi1/file_ops.c b/drivers/infiniband/hw/hfi1/file_ops.c
> index dc548e6802e24..b02d1f1dbb27b 100644
> --- a/drivers/infiniband/hw/hfi1/file_ops.c
> +++ b/drivers/infiniband/hw/hfi1/file_ops.c
[ ... ]
> @@ -422,7 +418,7 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
>  			ret = -EPERM;
>  			goto done;
>  		}
> -		vm_flags_clear(vma, VM_MAYWRITE);
> +		vm_flags_mod(vma, VM_DONTEXPAND, VM_MAYWRITE);
>  		/*
>  		 * Mmap multiple separate allocations into a single vma.  From
>  		 * here, dma_mmap_coherent() calls dma_direct_mmap(), which
[ ... ]
> @@ -559,12 +546,10 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
>  	}
>  
>  	vm_flags_reset(vma, flags);
> -	mmap_cdbg(ctxt, subctxt, type, mapio, vmf, memaddr, memvirt, memdma, 
> +	mmap_cdbg(ctxt, subctxt, type, mapio, is_vmalloc, memaddr, memvirt, memdma,
>  		  memlen, vma);
> -	if (vmf) {
> -		vma->vm_pgoff = PFN_DOWN(memaddr);
> -		vma->vm_ops = &vm_ops;
> -		ret = 0;
> +	if (is_vmalloc) {
> +		ret = remap_vmalloc_range(vma, memvirt, 0);

[Severity: Medium]
Will this trigger a lockdep warning due to the VMA not being write-locked?

In hfi1_file_mmap(), modifying the VMA flags using vm_flags_reset() (and
the newly introduced vm_flags_mod() earlier in this function) enforces 
write-lock assertions via vma_assert_write_locked(). 

Because the VMA is not yet inserted into the tree and is not write-locked 
during the mmap callback, its vm_lock_seq is uninitialized (UINT_MAX) and 
will not match the mm lock sequence. This triggers a VM_WARN_ON_ONCE_VMA.

Should vm_flags_init() be used instead?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=11

[PATCH v6 12/39] selinux: reject writable opens of policy file, drop mmap shared/write check

[Lorenzo Stoakes (ARM)]

The policy file has no write method and is exposed read-only (S_IRUGO in
selinux_files[]), yet sel_open_policy() performs no open mode check, so a
CAP_DAC_OVERRIDE caller can open it O_RDWR. Reject FMODE_WRITE at open, as
kernfs does.

The file can then never be mapped with FMODE_WRITE, so do_mmap() always
clears VM_MAYWRITE and VM_SHARED for MAP_SHARED mappings and the VM_SHARED
check in sel_mmap_policy() cannot be reached. Remove it.

This also stops sel_mmap_policy() clearing VM_MAYWRITE on a mapping that is
neither a PFN map nor a mixed map, ahead of the core enforcing that only
such mappings may do so.

Acked-by: Stephen Smalley <stephen.smalley.work@gmail.com>
Reviewed-by: Jann Horn <jannh@google.com>
Acked-by: Paul Moore <paul@paul-moore.com>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 security/selinux/selinuxfs.c | 11 +++--------
 1 file changed, 3 insertions(+), 8 deletions(-)

diff --git a/security/selinux/selinuxfs.c b/security/selinux/selinuxfs.c
index c7d91476971c..545a6f89f9e7 100644
--- a/security/selinux/selinuxfs.c
+++ b/security/selinux/selinuxfs.c
@@ -340,6 +340,9 @@ static int sel_open_policy(struct inode *inode, struct file *filp)
 	struct policy_load_memory *plm = NULL;
 	int rc;
 
+	if (filp->f_mode & FMODE_WRITE)
+		return -EACCES;
+
 	rc = avc_has_perm(current_sid(), SECINITSID_SECURITY,
 			  SECCLASS_SECURITY, SECURITY__READ_POLICY, NULL);
 	if (rc)
@@ -424,14 +427,6 @@ static const struct vm_operations_struct sel_mmap_policy_ops = {
 
 static int sel_mmap_policy(struct file *filp, struct vm_area_struct *vma)
 {
-	if (vma->vm_flags & VM_SHARED) {
-		/* do not allow mprotect to make mapping writable */
-		vm_flags_clear(vma, VM_MAYWRITE);
-
-		if (vma->vm_flags & VM_WRITE)
-			return -EACCES;
-	}
-
 	vm_flags_set(vma, VM_DONTEXPAND | VM_DONTDUMP);
 	vma->vm_ops = &sel_mmap_policy_ops;
 

-- 
2.55.0

Re: [PATCH v6 12/39] selinux: reject writable opens of policy file, drop mmap shared/write check

[sashiko-bot]

> The policy file has no write method and is exposed read-only (S_IRUGO in
> selinux_files[]), yet sel_open_policy() performs no open mode check, so a
> CAP_DAC_OVERRIDE caller can open it O_RDWR. Reject FMODE_WRITE at open, as
> kernfs does.
> 
> The file can then never be mapped with FMODE_WRITE, so do_mmap() always
> clears VM_MAYWRITE and VM_SHARED for MAP_SHARED mappings and the VM_SHARED
> check in sel_mmap_policy() cannot be reached. Remove it.
> 
> This also stops sel_mmap_policy() clearing VM_MAYWRITE on a mapping that is
> neither a PFN map nor a mixed map, ahead of the core enforcing that only
> such mappings may do so.
> 
> Acked-by: Stephen Smalley <stephen.smalley.work@gmail.com>
> Reviewed-by: Jann Horn <jannh@google.com>
> Acked-by: Paul Moore <paul@paul-moore.com>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=12

[PATCH v6 13/39] ALSA: pcm: use vm_insert_page() to map PCM status page

[Lorenzo Stoakes (ARM)]

There's no need to keep a fault handler around for this, instead map on
mmap.

While we're here, rename area to vma to be consistent.

This correctly makes the mapping a mixed map mapping.

This works towards establishing the invariant that only PFN mapped or mixed
map mappings may clear the VM_MAYWRITE flag. The status page mapping clears
VM_MAYWRITE, so it must not be mm-managed; the control page mapping remains
writable and is left fault-based.

The assumption is made that the struct pcm_mmap_status structure is at most
a page in size, which is asserted as a build bug.

This is safe to assume, as the size of the structure is 56 bytes at most.

Acked-by: Takashi Iwai <tiwai@suse.de>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 sound/core/pcm_native.c | 38 +++++++++++++-------------------------
 1 file changed, 13 insertions(+), 25 deletions(-)

diff --git a/sound/core/pcm_native.c b/sound/core/pcm_native.c
index 62324282fcae..c98fb3df14f3 100644
--- a/sound/core/pcm_native.c
+++ b/sound/core/pcm_native.c
@@ -3760,39 +3760,27 @@ static __poll_t snd_pcm_poll(struct file *file, poll_table *wait)
 /*
  * mmap status record
  */
-static vm_fault_t snd_pcm_mmap_status_fault(struct vm_fault *vmf)
+static int snd_pcm_mmap_status(struct snd_pcm_substream *substream, struct file *file,
+			       struct vm_area_struct *vma)
 {
-	struct snd_pcm_substream *substream = vmf->vma->vm_private_data;
+	const unsigned long size = vma->vm_end - vma->vm_start;
 	struct snd_pcm_runtime *runtime;
-	
-	if (substream == NULL)
-		return VM_FAULT_SIGBUS;
-	runtime = substream->runtime;
-	vmf->page = virt_to_page(runtime->status);
-	get_page(vmf->page);
-	return 0;
-}
+	struct page *page;
 
-static const struct vm_operations_struct snd_pcm_vm_ops_status =
-{
-	.fault =	snd_pcm_mmap_status_fault,
-};
+	BUILD_BUG_ON(sizeof(struct snd_pcm_mmap_status) > PAGE_SIZE);
 
-static int snd_pcm_mmap_status(struct snd_pcm_substream *substream, struct file *file,
-			       struct vm_area_struct *area)
-{
-	long size;
-	if (!(area->vm_flags & VM_READ))
+	if (!(vma->vm_flags & VM_READ))
 		return -EINVAL;
-	size = area->vm_end - area->vm_start;
-	if (size != PAGE_ALIGN(sizeof(struct snd_pcm_mmap_status)))
+	if (size != PAGE_SIZE)
 		return -EINVAL;
-	area->vm_ops = &snd_pcm_vm_ops_status;
-	area->vm_private_data = substream;
-	vm_flags_mod(area, VM_DONTEXPAND | VM_DONTDUMP,
+
+	vm_flags_mod(vma, VM_DONTEXPAND | VM_DONTDUMP,
 		     VM_WRITE | VM_MAYWRITE);
+	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
 
-	return 0;
+	runtime = substream->runtime;
+	page = virt_to_page(runtime->status);
+	return vm_insert_page(vma, vma->vm_start, page);
 }
 
 /*

-- 
2.55.0

Re: [PATCH v6 13/39] ALSA: pcm: use vm_insert_page() to map PCM status page

[sashiko-bot]

> There's no need to keep a fault handler around for this, instead map on
> mmap.
> 
> While we're here, rename area to vma to be consistent.
> 
> This correctly makes the mapping a mixed map mapping.
> 
> This works towards establishing the invariant that only PFN mapped or mixed
> map mappings may clear the VM_MAYWRITE flag. The status page mapping clears
> VM_MAYWRITE, so it must not be mm-managed; the control page mapping remains
> writable and is left fault-based.
> 
> The assumption is made that the struct pcm_mmap_status structure is at most
> a page in size, which is asserted as a build bug.
> 
> This is safe to assume, as the size of the structure is 56 bytes at most.
> 
> Acked-by: Takashi Iwai <tiwai@suse.de>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=13

[PATCH v6 14/39] bpf: arena: mark arena_map_mmap() mappings VM_MIXEDMAP

[Lorenzo Stoakes (ARM)]

The bpf_map->ops->map_mmap callback invoked by bpf_map_mmap() can be set to
one of ringbuf_map_mmap_kern(), ringbuf_map_mmap_user(), array_map_mmap()
or arena_map_mmap().

It is convention in mm to mark mappings whose pages the kernel manages
itself with VM_MIXEDMAP, so core mm knows not to treat them as ordinary
page cache or anonymous memory.

The map_mmap callbacks ringbuf_map_mmap_kern() and ringbuf_map_mmap_user()
use remap_vmalloc_range(), which ultimately invokes vm_insert_page() and so
marks the ranges VM_MIXEDMAP, and array_map_mmap() sets VM_MIXEDMAP
explicitly.

However, the exception to this is arena_map_mmap(), which doesn't set the
flag.

This patch corrects this and updates the comment to reflect it.

The pages are refcounted and vm_normal_page() finds them regardless of the
flag, and VM_DONTEXPAND remains set (marking the memory as VM_SPECIAL and
thus unmergeable). The one effect is that NUMA balancing now skips these
VMAs, as it already does for the other bpf map mappings, which is the
reason array_map_mmap() gives for setting the flag.

The intent of this patch is to be able to establish the invariant that only
PFN-mapped or mixed map ranges may clear the VM_MAYWRITE flag, as is done
in bpf_map_mmap().

Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 kernel/bpf/arena.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/bpf/arena.c b/kernel/bpf/arena.c
index 7b6847200b43..b69fe5e34339 100644
--- a/kernel/bpf/arena.c
+++ b/kernel/bpf/arena.c
@@ -620,8 +620,9 @@ static int arena_map_mmap(struct bpf_map *map, struct vm_area_struct *vma)
 	 * clears VM_MAYEXEC. Set VM_DONTEXPAND to avoid potential change
 	 * of user_vm_start. Set VM_DONTCOPY to prevent arena VMA from
 	 * being copied into the child process on fork.
+	 * This is a kernel page so set VM_MIXEDMAP.
 	 */
-	vm_flags_set(vma, VM_DONTEXPAND | VM_DONTCOPY);
+	vm_flags_set(vma, VM_MIXEDMAP | VM_DONTEXPAND | VM_DONTCOPY);
 	vma->vm_ops = &arena_vm_ops;
 	return 0;
 }

-- 
2.55.0

Re: [PATCH v6 14/39] bpf: arena: mark arena_map_mmap() mappings VM_MIXEDMAP

[sashiko-bot]

> The bpf_map->ops->map_mmap callback invoked by bpf_map_mmap() can be set to
> one of ringbuf_map_mmap_kern(), ringbuf_map_mmap_user(), array_map_mmap()
> or arena_map_mmap().
> 
> It is convention in mm to mark mappings whose pages the kernel manages
> itself with VM_MIXEDMAP, so core mm knows not to treat them as ordinary
> page cache or anonymous memory.
> 
> The map_mmap callbacks ringbuf_map_mmap_kern() and ringbuf_map_mmap_user()
> use remap_vmalloc_range(), which ultimately invokes vm_insert_page() and so
> marks the ranges VM_MIXEDMAP, and array_map_mmap() sets VM_MIXEDMAP
> explicitly.
> 
> However, the exception to this is arena_map_mmap(), which doesn't set the
> flag.
> [ ... ]
> 
> Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=14

[PATCH v6 15/39] mm/vma: add vma[_flags]_is_mm_managed() predicates

[Lorenzo Stoakes (ARM)]

Rather than referring to VMA flags with uncertain meaning, add a new
predicate that explicitly describes what absence of the VMA_PFNMAP_BIT and
VMA_MIXEDMAP_BIT flags means, and then refer to that function for
determining VMA mergeability.

If neither of these flags are set, that means the contents of the mapping
are managed by core mm.

Otherwise they are instead owned by some other kernel component, usually a
driver: the memory may be MMIO, kernel-allocated pages or even ordinary
pages the driver maps itself, but core mm must not populate, reclaim,
migrate, copy-on-write or merge the range itself.

We initially also treat VMA_IO_BIT as implying a mapping is not mm-managed,
as by implication it cannot be. (mlock() also sets VMA_IO_BIT transiently
on ordinary VMAs while locking them, which is addressed later in this
series.)

However the intent is to in future remove this, as no mapping should be
marked as an I/O mapping without also being marked with VMA_PFNMAP_BIT.

This forms the basis of further work intended to improve how we express VMA
properties such as this.

Also update the VMA userland tests to reflect the change.

No functional change intended.

Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 include/linux/mm.h              | 53 ++++++++++++++++++++++++++++++++++++++++-
 tools/testing/vma/include/dup.h | 29 +++++++++++++++++++++-
 2 files changed, 80 insertions(+), 2 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 438ab06a4a91..df8a5a1eb9c3 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1612,6 +1612,41 @@ static inline bool vma_is_shared_maywrite(const struct vm_area_struct *vma)
 	return is_shared_maywrite(&vma->flags);
 }
 
+/**
+ * vma_flags_is_mm_managed() - Do the specified VMA flags indicate that the
+ * contents of the VMA are managed by core mm?
+ * @flags: The VMA flags to test.
+ *
+ * Such a mapping is one established by a simple mmap() or brk() with no driver
+ * or other kernel component involved.
+ *
+ * They are rmappable, and core mm services rmap operations including reclaim,
+ * as well as population, CoW and merge/split (if eligible).
+ *
+ * Mappings which fail this test are those where this is not the case -
+ * e.g. MMIO, user-mapped kernel pages, udmabuf (pfnmap shmem), etc.
+ *
+ * Returns: true if the flags indicate an mm-managed mapping.
+ */
+static inline bool vma_flags_is_mm_managed(const vma_flags_t *flags)
+{
+	return !vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT,
+				   VMA_IO_BIT);
+}
+
+/**
+ * vma_is_mm_managed() - Are the contents of @vma managed by core mm?
+ * @vma: The VMA to test.
+ *
+ * See vma_flags_is_mm_managed() for a description of this property.
+ *
+ * Returns: true if the VMA is mm-managed.
+ */
+static inline bool vma_is_mm_managed(const struct vm_area_struct *vma)
+{
+	return vma_flags_is_mm_managed(&vma->flags);
+}
+
 /**
  * vma_flags_can_merge() - Do the specified VMA flags permit the VMA to be
  * merged with another?
@@ -1620,7 +1655,23 @@ static inline bool vma_is_shared_maywrite(const struct vm_area_struct *vma)
  */
 static inline bool vma_flags_can_merge(const vma_flags_t *flags)
 {
-	return !vma_flags_test_any_mask(flags, VMA_SPECIAL_FLAGS);
+	/*
+	 * VMA merging assumes that a VMA's flags and fields completely describe
+	 * its state.
+	 *
+	 * However, mappings which are not mm-managed may have established state
+	 * upon mapping not embodied in any attribute of the VMA.
+	 *
+	 * Additionally, private (CoW) PFN maps encode the source PFN of the
+	 * range in vma->vm_pgoff, which may otherwise cause spurious merges.
+	 */
+	if (!vma_flags_is_mm_managed(flags))
+		return false;
+	/* VMA explicitly marked as being unmergeable. */
+	if (vma_flags_test(flags, VMA_DONTEXPAND_BIT))
+		return false;
+
+	return true;
 }
 
 /**
diff --git a/tools/testing/vma/include/dup.h b/tools/testing/vma/include/dup.h
index a6a8e53c4b38..89c645a6ad60 100644
--- a/tools/testing/vma/include/dup.h
+++ b/tools/testing/vma/include/dup.h
@@ -1665,7 +1665,34 @@ static inline bool file_is_dev_zero(const struct file *file)
 	return file && file->f_op == &zero_fops;
 }
 
+static inline bool vma_flags_is_mm_managed(const vma_flags_t *flags)
+{
+	return !vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT,
+				   VMA_IO_BIT);
+}
+
+static inline bool vma_is_mm_managed(const struct vm_area_struct *vma)
+{
+	return vma_flags_is_mm_managed(&vma->flags);
+}
+
 static inline bool vma_flags_can_merge(const vma_flags_t *flags)
 {
-	return !vma_flags_test_any_mask(flags, VMA_SPECIAL_FLAGS);
+	/*
+	 * VMA merging assumes that a VMA's flags and fields completely describe
+	 * its state.
+	 *
+	 * However, mappings which are not mm-managed may have established state
+	 * upon mapping not embodied in any attribute of the VMA.
+	 *
+	 * Additionally, private (CoW) PFN maps encode the source PFN of the
+	 * range in vma->vm_pgoff, which may otherwise cause spurious merges.
+	 */
+	if (!vma_flags_is_mm_managed(flags))
+		return false;
+	/* VMA explicitly marked as being unmergeable. */
+	if (vma_flags_test(flags, VMA_DONTEXPAND_BIT))
+		return false;
+
+	return true;
 }

-- 
2.55.0

Re: [PATCH v6 15/39] mm/vma: add vma[_flags]_is_mm_managed() predicates

[sashiko-bot]

> Rather than referring to VMA flags with uncertain meaning, add a new
> predicate that explicitly describes what absence of the VMA_PFNMAP_BIT and
> VMA_MIXEDMAP_BIT flags means, and then refer to that function for
> determining VMA mergeability.
> 
> If neither of these flags are set, that means the contents of the mapping
> are managed by core mm.
> 
> Otherwise they are instead owned by some other kernel component, usually a
> driver: the memory may be MMIO, kernel-allocated pages or even ordinary
> pages the driver maps itself, but core mm must not populate, reclaim,
> migrate, copy-on-write or merge the range itself.
> 
> We initially also treat VMA_IO_BIT as implying a mapping is not mm-managed,
> as by implication it cannot be. (mlock() also sets VMA_IO_BIT transiently
> [ ... ]
> 
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=15

[PATCH v6 16/39] mm/vma: only allow mmap to clear VMA_MAYWRITE_BIT if not mm-managed

[Lorenzo Stoakes (ARM)]

For ordinary files the only way the VMA_MAYWRITE_BIT flag is cleared is if
the underlying file is itself read-only.

This means that mprotect() cannot mark a shared mapping of a read-only file
as read/write, as doing so would violate the read only attribute, and
permit writes.

In general, we do not want file systems to be able to do this for
read/write files.

Doing so would violate fundamental user expectation of file attributes and
likely break userspace.

However, drivers pose a tricky problem here - the /dev/xxx file may be
read/write but provide access to a resource which is fundamentally
read-only.

Therefore we must allow drivers to be able to clear VMA_MAYWRITE_BIT.

To achieve both of these things, forbid this for mm-managed mappings as
identified by vma_flags_is_mm_managed().

This constrains this ability to drivers which own the mapping's contents,
whether memory-mapped I/O, kernel-allocated pages, or ordinary pages they
map themselves, and so define its semantics.

Every in-tree mmap hook which clears VMA_MAYWRITE_BIT, some twenty sites
across drivers, filesystems and bpf, establishes a mapping which is not
mm-managed,
with usbmon and the ALSA PCM status page converted earlier in this series
to do so.

Note that drivers may, if they do not gate on VMA_SHARED_BIT, be able to
disable MAP_PRIVATE-file-backed mapping CoW semantics.

This is perhaps not always intended, but we retain this capacity to
maintain existing behaviour.

As no driver which clears VMA_MAYWRITE_BIT establishes an mm-managed
mapping, no functional change is intended.

Acked-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/vma.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/mm/vma.c b/mm/vma.c
index c067d6a4c826..39ceb03023e2 100644
--- a/mm/vma.c
+++ b/mm/vma.c
@@ -2801,6 +2801,11 @@ static int mmap_validate(unsigned long orig_start, unsigned long orig_end,
 	if (WARN_ON_ONCE(!was_maywrite && is_maywrite))
 		return -EINVAL;
 
+	/* mm-managed mappings may not clear VMA_MAYWRITE_BIT. */
+	if (vma_flags_is_mm_managed(curr_flags) &&
+	    WARN_ON_ONCE(was_maywrite && !is_maywrite))
+		return -EINVAL;
+
 	return mmap_validate_vma_flags(curr_flags);
 }
 

-- 
2.55.0

Re: [PATCH v6 16/39] mm/vma: only allow mmap to clear VMA_MAYWRITE_BIT if not mm-managed

[sashiko-bot]

> For ordinary files the only way the VMA_MAYWRITE_BIT flag is cleared is if
> the underlying file is itself read-only.
> 
> This means that mprotect() cannot mark a shared mapping of a read-only file
> as read/write, as doing so would violate the read only attribute, and
> permit writes.
> 
> In general, we do not want file systems to be able to do this for
> read/write files.
> 
> Doing so would violate fundamental user expectation of file attributes and
> likely break userspace.
> 
> However, drivers pose a tricky problem here - the /dev/xxx file may be
> read/write but provide access to a resource which is fundamentally
> [ ... ]
> mapping, no functional change is intended.
> 
> Acked-by: Zi Yan <ziy@nvidia.com>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=16

[PATCH v6 17/39] mm/vma: add and use vma_[flags]_is_fixed_mapping

[Lorenzo Stoakes (ARM)]

This determines whether a VMA cannot be expanded or merged because what
they mapped was determined to be a set size at mmap time.

This typically refers to mappings not managed by core mm, however
VMA_DONTEXPAND_BIT
is not reliably set alongside VMA_PFNMAP_BIT or VMA_MIXEDMAP_BIT, so we
must explicitly test for this for now.

We also explicitly test for VMA_PFNMAP_BIT as VMA_DONTEXPAND_BIT may not be
set for VMA_PFNMAP_BIT's despite the one implying the other.

Use this predicate in vma_flags_can_merge() and in check_prep_vma() in the
mremap logic testing to see if mremap() can expand the VMA. The criteria
for khugepaged and MADV_COLLAPSE eligibility in
__thp_vma_allowable_orders() are precisely those for mergeability, so use
vma_can_merge() there (with an expanded comment).

This obviates the need for the VM_NO_KHUGEPAGED mask, so remove it.

Hugetlb VMAs remain excluded from khugepaged as hugetlbfs always sets
VMA_DONTEXPAND_BIT.

Also update the userland VMA tests to reflect the change.

No functional change intended.

Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 include/linux/mm.h              | 39 +++++++++++++++++++++++++++++++++++----
 mm/huge_memory.c                | 11 +++++++----
 mm/mremap.c                     |  5 ++---
 tools/testing/vma/include/dup.h | 16 +++++++++++++++-
 4 files changed, 59 insertions(+), 12 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index df8a5a1eb9c3..2ecf39bc857b 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -600,9 +600,6 @@ enum {
 #define VMA_REMAP_FLAGS mk_vma_flags(VMA_IO_BIT, VMA_PFNMAP_BIT,	\
 				     VMA_DONTEXPAND_BIT, VMA_DONTDUMP_BIT)
 
-/* This mask prevents VMA from being scanned with khugepaged */
-#define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
-
 /* This mask defines which mm->def_flags a process can inherit its parent */
 #define VM_INIT_DEF_MASK	VM_NOHUGEPAGE
 
@@ -1647,6 +1644,40 @@ static inline bool vma_is_mm_managed(const struct vm_area_struct *vma)
 	return vma_flags_is_mm_managed(&vma->flags);
 }
 
+/**
+ * vma_flags_is_fixed_mapping() - Do the specified VMA flags indicate that this
+ * is a fixed mapping that cannot be expanded or merged?
+ * @flags: The VMA flags to test.
+ *
+ * Fixed mappings are those whose size is set at the point of mmap (for
+ * instance, a mapping of a fixed range of memory established by a driver),
+ * and thus cannot be expanded or merged.
+ *
+ * Returns: true if the flags indicate a fixed mapping.
+ */
+static inline bool vma_flags_is_fixed_mapping(const vma_flags_t *flags)
+{
+	/*
+	 * VMA_PFNMAP_BIT should imply VMA_DONTEXPAND_BIT, but some callers set
+	 * only the former.
+	 */
+	return vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_DONTEXPAND_BIT);
+}
+
+/**
+ * vma_is_fixed_mapping() - Is this VMA a fixed mapping that cannot be
+ * expanded or merged?
+ * @vma: The VMA to test.
+ *
+ * See vma_flags_is_fixed_mapping() for a description of this property.
+ *
+ * Returns: true if the VMA maps a fixed mapping.
+ */
+static inline bool vma_is_fixed_mapping(const struct vm_area_struct *vma)
+{
+	return vma_flags_is_fixed_mapping(&vma->flags);
+}
+
 /**
  * vma_flags_can_merge() - Do the specified VMA flags permit the VMA to be
  * merged with another?
@@ -1668,7 +1699,7 @@ static inline bool vma_flags_can_merge(const vma_flags_t *flags)
 	if (!vma_flags_is_mm_managed(flags))
 		return false;
 	/* VMA explicitly marked as being unmergeable. */
-	if (vma_flags_test(flags, VMA_DONTEXPAND_BIT))
+	if (vma_flags_is_fixed_mapping(flags))
 		return false;
 
 	return true;
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 4cd917f77f3f..87e8bf8029ab 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -212,11 +212,14 @@ unsigned long __thp_vma_allowable_orders(struct vm_area_struct *vma,
 		return in_pf ? orders : 0;
 
 	/*
-	 * khugepaged special VMA and hugetlb VMA.
-	 * Must be checked after dax since some dax mappings may have
-	 * VM_MIXEDMAP set.
+	 * khugepaged moves data from VMAs once collapsed, after they have been
+	 * faulted in, relying on refaulting for file-backed memory.
+	 *
+	 * Mappings not managed by core mm cannot be reliably reconstructed from
+	 * page faults, and fixed mappings (including hugetlb) may well be
+	 * mm-managed - precisely the mappings which cannot be merged.
 	 */
-	if (!in_pf && !smaps && (vm_flags & VM_NO_KHUGEPAGED))
+	if (!in_pf && !smaps && !vma_can_merge(vma))
 		return 0;
 
 	/*
diff --git a/mm/mremap.c b/mm/mremap.c
index 7c368440fafe..ed19b47c2caf 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -1788,8 +1788,7 @@ static int check_prep_vma(struct vma_remap_struct *vrm)
 		return -EINVAL;
 	}
 
-	if ((vrm->flags & MREMAP_DONTUNMAP) &&
-	    vma_test_any(vma, VMA_DONTEXPAND_BIT, VMA_PFNMAP_BIT))
+	if ((vrm->flags & MREMAP_DONTUNMAP) && vma_is_fixed_mapping(vma))
 		return -EINVAL;
 
 	/*
@@ -1827,7 +1826,7 @@ static int check_prep_vma(struct vma_remap_struct *vrm)
 	if (pgoff + (new_len >> PAGE_SHIFT) < pgoff)
 		return -EINVAL;
 
-	if (vma_test_any(vma, VMA_DONTEXPAND_BIT, VMA_PFNMAP_BIT))
+	if (vma_is_fixed_mapping(vma))
 		return -EFAULT;
 
 	if (!mlock_future_ok(mm, vma_test(vma, VMA_LOCKED_BIT), vrm->delta))
diff --git a/tools/testing/vma/include/dup.h b/tools/testing/vma/include/dup.h
index 89c645a6ad60..442ea9eb098d 100644
--- a/tools/testing/vma/include/dup.h
+++ b/tools/testing/vma/include/dup.h
@@ -1676,6 +1676,20 @@ static inline bool vma_is_mm_managed(const struct vm_area_struct *vma)
 	return vma_flags_is_mm_managed(&vma->flags);
 }
 
+static inline bool vma_flags_is_fixed_mapping(const vma_flags_t *flags)
+{
+	/*
+	 * VMA_PFNMAP_BIT should imply VMA_DONTEXPAND_BIT, but some callers set
+	 * only the former.
+	 */
+	return vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_DONTEXPAND_BIT);
+}
+
+static inline bool vma_is_fixed_mapping(const struct vm_area_struct *vma)
+{
+	return vma_flags_is_fixed_mapping(&vma->flags);
+}
+
 static inline bool vma_flags_can_merge(const vma_flags_t *flags)
 {
 	/*
@@ -1691,7 +1705,7 @@ static inline bool vma_flags_can_merge(const vma_flags_t *flags)
 	if (!vma_flags_is_mm_managed(flags))
 		return false;
 	/* VMA explicitly marked as being unmergeable. */
-	if (vma_flags_test(flags, VMA_DONTEXPAND_BIT))
+	if (vma_flags_is_fixed_mapping(flags))
 		return false;
 
 	return true;

-- 
2.55.0

Re: [PATCH v6 17/39] mm/vma: add and use vma_[flags]_is_fixed_mapping

[sashiko-bot]

> This determines whether a VMA cannot be expanded or merged because what
> they mapped was determined to be a set size at mmap time.
> 
> This typically refers to mappings not managed by core mm, however
> VMA_DONTEXPAND_BIT
> is not reliably set alongside VMA_PFNMAP_BIT or VMA_MIXEDMAP_BIT, so we
> must explicitly test for this for now.
> 
> We also explicitly test for VMA_PFNMAP_BIT as VMA_DONTEXPAND_BIT may not be
> set for VMA_PFNMAP_BIT's despite the one implying the other.
> 
> Use this predicate in vma_flags_can_merge() and in check_prep_vma() in the
> mremap logic testing to see if mremap() can expand the VMA. The criteria
> for khugepaged and MADV_COLLAPSE eligibility in
> __thp_vma_allowable_orders() are precisely those for mergeability, so use
> vma_can_merge() there (with an expanded comment).
> 
> This obviates the need for the VM_NO_KHUGEPAGED mask, so remove it.
> 
> Hugetlb VMAs remain excluded from khugepaged as hugetlbfs always sets
> VMA_DONTEXPAND_BIT.
> 
> Also update the userland VMA tests to reflect the change.
> 
> No functional change intended.
> 
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=17

[PATCH v6 18/39] scsi: sg: convert mmap hook to mmap_prepare and rework

[Lorenzo Stoakes (ARM)]

Move from the deprecated mmap hook to the new mmap_prepare hook.

We are mapping kernel pages here, so use the discontiguous kernel mapping
mmap action to do so.

Unwind the rather confusing loop and instead map as many pages as we can at
one time.

Note that we do not need to pay attention to rsv_schp->k_use_sg here, as
the pages are populated for the length of the buffer at
rsv_schp->page_order granularity as compound pages.

The discontiguous kernel page mapping logic handles the compound pages for
us.

sfp->mmap_called keeps the buffer stable for us. As before it is never
cleared, so a failed mmap also leaves it set.

We also remove some useless vma, vma->vm_file NULL checks - these will
always be non-NULL if you reached the mmap hook logic.

We retain log output for consistency, but change what's output on page
mapping to indicate that sg_discontig_get() does the work now.

Note that we drop the VMA_IO_BIT flag for the VMA here. It was never
necessary as we invoke alloc_pages() which gives us refcounted folios that
are fine for GUP to access (VMA_IO_BIT would prevent that).

Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 drivers/scsi/sg.c | 115 ++++++++++++++++++++++++------------------------------
 1 file changed, 51 insertions(+), 64 deletions(-)

diff --git a/drivers/scsi/sg.c b/drivers/scsi/sg.c
index 5408f002e6c0..3f9e08725602 100644
--- a/drivers/scsi/sg.c
+++ b/drivers/scsi/sg.c
@@ -1212,85 +1212,72 @@ sg_fasync(int fd, struct file *filp, int mode)
 	return fasync_helper(fd, filp, mode, &sfp->async_qp);
 }
 
-static vm_fault_t
-sg_vma_fault(struct vm_fault *vmf)
+static int sg_discontig_init(void *vm_private_data, void **private)
 {
-	struct vm_area_struct *vma = vmf->vma;
-	Sg_fd *sfp;
-	unsigned long offset, len, sa;
-	Sg_scatter_hold *rsv_schp;
-	int k, length;
-
-	if ((NULL == vma) || (!(sfp = (Sg_fd *) vma->vm_private_data)))
-		return VM_FAULT_SIGBUS;
-	rsv_schp = &sfp->reserve;
-	offset = vmf->pgoff << PAGE_SHIFT;
-	if (offset >= rsv_schp->bufflen)
-		return VM_FAULT_SIGBUS;
-	SCSI_LOG_TIMEOUT(3, sg_printk(KERN_INFO, sfp->parentdp,
-				      "sg_vma_fault: offset=%lu, scatg=%d\n",
-				      offset, rsv_schp->k_use_sg));
-	sa = vma->vm_start;
-	length = 1 << (PAGE_SHIFT + rsv_schp->page_order);
-	for (k = 0; k < rsv_schp->k_use_sg && sa < vma->vm_end; k++) {
-		len = vma->vm_end - sa;
-		len = (len < length) ? len : length;
-		if (offset < len) {
-			struct page *page = rsv_schp->pages[k] + (offset >> PAGE_SHIFT);
-			get_page(page);	/* increment page count */
-			vmf->page = page;
-			return 0; /* success */
-		}
-		sa += len;
-		offset -= len;
+	const unsigned long req_sz = (unsigned long)*private;
+	Sg_fd *sfp = vm_private_data;
+	Sg_scatter_hold *rsv_schp = &sfp->reserve;
+	int err = 0;
+
+	mutex_lock(&sfp->f_mutex);
+	if (req_sz > rsv_schp->bufflen) {
+		err = -ENOMEM;	/* cannot map more than reserved buffer */
+		goto out;
+	}
+	sfp->mmap_called = 1; /* Prevents changes to buffer size. */
+out:
+	mutex_unlock(&sfp->f_mutex);
+	return err;
+}
+
+static int
+sg_discontig_get(struct discontig_kernel_page_state *state)
+{
+	Sg_fd *sfp = state->vm_private_data;
+	Sg_scatter_hold *rsv_schp = &sfp->reserve;
+	const unsigned int order = rsv_schp->page_order;
+	const pgoff_t nr_pages = state->nr_pages_mapped;
+
+	if (nr_pages >= (rsv_schp->bufflen >> PAGE_SHIFT)) {
+		discontig_kernel_map_abort(state);
+		return 0;
 	}
 
-	return VM_FAULT_SIGBUS;
+	SCSI_LOG_TIMEOUT(3, sg_printk(KERN_INFO, sfp->parentdp,
+				      "%s: offset=%lu, scatg=%d\n", __func__,
+				      nr_pages << PAGE_SHIFT, rsv_schp->k_use_sg));
+
+	discontig_kernel_map_page(state, rsv_schp->pages[nr_pages >> order]);
+	return 0;
 }
 
-static const struct vm_operations_struct sg_mmap_vm_ops = {
-	.fault = sg_vma_fault,
+static const struct discontig_kernel_page_ops sg_discontig_ops = {
+	.init = sg_discontig_init,
+	.get = sg_discontig_get,
 };
 
 static int
-sg_mmap(struct file *filp, struct vm_area_struct *vma)
+sg_mmap_prepare(struct vm_area_desc *desc)
 {
-	Sg_fd *sfp;
-	unsigned long req_sz, len, sa;
-	Sg_scatter_hold *rsv_schp;
-	int k, length;
-	int ret = 0;
+	Sg_fd *sfp = desc->file->private_data;
+	const unsigned long req_sz = vma_desc_size(desc);
 
-	if ((!filp) || (!vma) || (!(sfp = (Sg_fd *) filp->private_data)))
+	if (!sfp)
 		return -ENXIO;
-	req_sz = vma->vm_end - vma->vm_start;
+
 	SCSI_LOG_TIMEOUT(3, sg_printk(KERN_INFO, sfp->parentdp,
 				      "sg_mmap starting, vm_start=%p, len=%d\n",
-				      (void *) vma->vm_start, (int) req_sz));
-	if (vma->vm_pgoff)
+				      (void *) desc->start, (int) req_sz));
+
+	if (desc->pgoff)
 		return -EINVAL;	/* want no offset */
-	rsv_schp = &sfp->reserve;
-	mutex_lock(&sfp->f_mutex);
-	if (req_sz > rsv_schp->bufflen) {
-		ret = -ENOMEM;	/* cannot map more than reserved buffer */
-		goto out;
-	}
 
-	sa = vma->vm_start;
-	length = 1 << (PAGE_SHIFT + rsv_schp->page_order);
-	for (k = 0; k < rsv_schp->k_use_sg && sa < vma->vm_end; k++) {
-		len = vma->vm_end - sa;
-		len = (len < length) ? len : length;
-		sa += len;
-	}
+	vma_desc_set_flags(desc, VMA_DONTEXPAND_BIT, VMA_DONTDUMP_BIT);
+	desc->private_data = sfp;
 
-	sfp->mmap_called = 1;
-	vm_flags_set(vma, VM_IO | VM_DONTEXPAND | VM_DONTDUMP);
-	vma->vm_private_data = sfp;
-	vma->vm_ops = &sg_mmap_vm_ops;
-out:
-	mutex_unlock(&sfp->f_mutex);
-	return ret;
+	mmap_action_map_discontig_kernel_pages(desc, (void *)req_sz,
+					       &sg_discontig_ops);
+	return 0;
 }
 
 static void
@@ -1415,7 +1402,7 @@ static const struct file_operations sg_fops = {
 	.unlocked_ioctl = sg_ioctl,
 	.compat_ioctl = compat_ptr_ioctl,
 	.open = sg_open,
-	.mmap = sg_mmap,
+	.mmap_prepare = sg_mmap_prepare,
 	.release = sg_release,
 	.fasync = sg_fasync,
 };

-- 
2.55.0

Re: [PATCH v6 18/39] scsi: sg: convert mmap hook to mmap_prepare and rework

[sashiko-bot]

> Move from the deprecated mmap hook to the new mmap_prepare hook.
> 
> We are mapping kernel pages here, so use the discontiguous kernel mapping
> mmap action to do so.
> 
> Unwind the rather confusing loop and instead map as many pages as we can at
> one time.
> 
> Note that we do not need to pay attention to rsv_schp->k_use_sg here, as
> the pages are populated for the length of the buffer at
> rsv_schp->page_order granularity as compound pages.
> 
> The discontiguous kernel page mapping logic handles the compound pages for
> us.
> 
> sfp->mmap_called keeps the buffer stable for us. As before it is never
> cleared, so a failed mmap also leaves it set.
> 
> We also remove some useless vma, vma->vm_file NULL checks - these will
> always be non-NULL if you reached the mmap hook logic.
> 
> We retain log output for consistency, but change what's output on page
> mapping to indicate that sg_discontig_get() does the work now.
> 
> Note that we drop the VMA_IO_BIT flag for the VMA here. It was never
> necessary as we invoke alloc_pages() which gives us refcounted folios that
> are fine for GUP to access (VMA_IO_BIT would prevent that).
> 
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=18

[PATCH v6 19/39] fbdev: defio: assert FBINFO_VIRTFB, drop VM_IO, add VM_MIXEDMAP

[Lorenzo Stoakes (ARM)]

Currently all drivers which use defio allocate system memory. All of them
also set FBINFO_VIRTFB, other than ssd1307fb, however this driver allocates
system RAM, so simply failed to set this flag when it ought to.

This patch sets FBINFO_VIRTFB on ssd1307fb probe, then drops setting VM_IO
in fb_deferred_io_mmap() and instead requires FBINFO_VIRTFB to be set,
erroring out with a kernel warning if not.

The logic requires a page from the driver and since commit 1ecbc7dd2902
("fbdev/deferred-io: Always call get_page() for framebuffer pages") has
always required it to be refcounted, so this was implicitly already the
case.

Finally this patch sets VM_MIXEDMAP, as the logic is mapping
kernel-allocated memory so this is appropriate.

Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 drivers/video/fbdev/core/fb_defio.c | 6 +++---
 drivers/video/fbdev/ssd1307fb.c     | 2 ++
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/video/fbdev/core/fb_defio.c b/drivers/video/fbdev/core/fb_defio.c
index fd00b86e1ae6..fb359ecc3966 100644
--- a/drivers/video/fbdev/core/fb_defio.c
+++ b/drivers/video/fbdev/core/fb_defio.c
@@ -366,13 +366,13 @@ int fb_deferred_io_mmap(struct fb_info *info, struct vm_area_struct *vma)
 {
 	vma->vm_page_prot = pgprot_decrypted(vma->vm_page_prot);
 
+	if (WARN_ON_ONCE(!(info->flags & FBINFO_VIRTFB)))
+		return -EINVAL;
 	if (!try_module_get(THIS_MODULE))
 		return -EINVAL;
 
 	vma->vm_ops = &fb_deferred_io_vm_ops;
-	vm_flags_set(vma, VM_DONTEXPAND | VM_DONTDUMP);
-	if (!(info->flags & FBINFO_VIRTFB))
-		vm_flags_set(vma, VM_IO);
+	vm_flags_set(vma, VM_MIXEDMAP | VM_DONTEXPAND | VM_DONTDUMP);
 	vma->vm_private_data = info->fbdefio_state;
 
 	fb_deferred_io_state_get(info->fbdefio_state); /* released in vma->vm_ops->close() */
diff --git a/drivers/video/fbdev/ssd1307fb.c b/drivers/video/fbdev/ssd1307fb.c
index c4fdecafd856..958514a35433 100644
--- a/drivers/video/fbdev/ssd1307fb.c
+++ b/drivers/video/fbdev/ssd1307fb.c
@@ -763,6 +763,8 @@ static int ssd1307fb_probe(struct i2c_client *client)
 	info->fix.smem_start = __pa(vmem);
 	info->fix.smem_len = vmem_size;
 
+	info->flags = FBINFO_VIRTFB;
+
 	fb_deferred_io_init(info);
 
 	i2c_set_clientdata(client, info);

-- 
2.55.0

Re: [PATCH v6 19/39] fbdev: defio: assert FBINFO_VIRTFB, drop VM_IO, add VM_MIXEDMAP

[sashiko-bot]

> Currently all drivers which use defio allocate system memory. All of them
> also set FBINFO_VIRTFB, other than ssd1307fb, however this driver allocates
> system RAM, so simply failed to set this flag when it ought to.
> 
> This patch sets FBINFO_VIRTFB on ssd1307fb probe, then drops setting VM_IO
> in fb_deferred_io_mmap() and instead requires FBINFO_VIRTFB to be set,
> erroring out with a kernel warning if not.
> 
> The logic requires a page from the driver and since commit 1ecbc7dd2902
> ("fbdev/deferred-io: Always call get_page() for framebuffer pages") has
> always required it to be refcounted, so this was implicitly already the
> case.
> 
> Finally this patch sets VM_MIXEDMAP, as the logic is mapping
> kernel-allocated memory so this is appropriate.
> 
> Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=19

[PATCH v6 20/39] HSI: cmt_speech: convert mmap hook to mmap_prepare, refactor

[Lorenzo Stoakes (ARM)]

Use the mmap_prepare in favour of the deprecated mmap hook as part of the
work to convert one to another.

Since this is simply a refcounted kernel page that has been allocated, it
should not be marked VM_IO and should be inserted using the kernel page
insertion mechanism, so convert it to do this instead.

Use the VMA descriptor's private data field as a scratch buffer to store
the page in - this stays valid throughout the kernel page mapping
operation.

Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 drivers/hsi/clients/cmt_speech.c | 33 +++++++++------------------------
 1 file changed, 9 insertions(+), 24 deletions(-)

diff --git a/drivers/hsi/clients/cmt_speech.c b/drivers/hsi/clients/cmt_speech.c
index 7226677ebde7..801697b74d4f 100644
--- a/drivers/hsi/clients/cmt_speech.c
+++ b/drivers/hsi/clients/cmt_speech.c
@@ -1084,22 +1084,6 @@ static void cs_hsi_stop(struct cs_hsi_iface *hi)
 	kfree(hi);
 }
 
-static vm_fault_t cs_char_vma_fault(struct vm_fault *vmf)
-{
-	struct cs_char *csdata = vmf->vma->vm_private_data;
-	struct page *page;
-
-	page = virt_to_page((void *)csdata->mmap_base);
-	get_page(page);
-	vmf->page = page;
-
-	return 0;
-}
-
-static const struct vm_operations_struct cs_char_vm_ops = {
-	.fault	= cs_char_vma_fault,
-};
-
 static int cs_char_fasync(int fd, struct file *file, int on)
 {
 	struct cs_char *csdata = file->private_data;
@@ -1256,18 +1240,19 @@ static long cs_char_ioctl(struct file *file, unsigned int cmd,
 	return r;
 }
 
-static int cs_char_mmap(struct file *file, struct vm_area_struct *vma)
+static int cs_char_mmap_prepare(struct vm_area_desc *desc)
 {
-	if (vma->vm_end < vma->vm_start)
-		return -EINVAL;
+	struct file *file = desc->file;
+	struct cs_char *csdata = file->private_data;
+	struct page **pages = (struct page **)&desc->private_data;
 
-	if (vma_pages(vma) != 1)
+	if (vma_desc_pages(desc) != 1)
 		return -EINVAL;
 
-	vm_flags_set(vma, VM_IO | VM_DONTDUMP | VM_DONTEXPAND);
-	vma->vm_ops = &cs_char_vm_ops;
-	vma->vm_private_data = file->private_data;
+	vma_desc_set_flags(desc, VMA_DONTDUMP_BIT, VMA_DONTEXPAND_BIT);
 
+	*pages = virt_to_page((void *)csdata->mmap_base);
+	mmap_action_map_kernel_pages_full(desc, pages);
 	return 0;
 }
 
@@ -1353,7 +1338,7 @@ static const struct file_operations cs_char_fops = {
 	.write		= cs_char_write,
 	.poll		= cs_char_poll,
 	.unlocked_ioctl	= cs_char_ioctl,
-	.mmap		= cs_char_mmap,
+	.mmap_prepare	= cs_char_mmap_prepare,
 	.open		= cs_char_open,
 	.release	= cs_char_release,
 	.fasync		= cs_char_fasync,

-- 
2.55.0

Re: [PATCH v6 20/39] HSI: cmt_speech: convert mmap hook to mmap_prepare, refactor

[sashiko-bot]

> Use the mmap_prepare in favour of the deprecated mmap hook as part of the
> work to convert one to another.
> 
> Since this is simply a refcounted kernel page that has been allocated, it
> should not be marked VM_IO and should be inserted using the kernel page
> insertion mechanism, so convert it to do this instead.
> 
> Use the VMA descriptor's private data field as a scratch buffer to store
> the page in - this stays valid throughout the kernel page mapping
> operation.
> 
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=20

[PATCH v6 21/39] mm/gup: error out early on !VMA_MAYREAD_BIT VMAs

[Lorenzo Stoakes (ARM)]

When populating a VMA range via the aptly named populate_vma_page_range()
an unreadable VMA will always eventually fail with -EFAULT.

That a VMA is accessible is always checked, however VMA_MAYREAD_BIT is not.

All user mappings always have VMA_MAYREAD_BIT set, so this check only
impacts kernel mappings.

It is implemented specifically to disallow population of uprobes XOL
mappings which are exec-only.

A nasty interaction with these mappings may occur if they are mlocked, so
actively disallow this early.

This allows a subsequent commit to remove the VM_IO check in
__mm_populate() which otherwise requires non-MMIO mappings to be wrongly
flagged simply as a workaround.

Acked-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/gup.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/mm/gup.c b/mm/gup.c
index c2dfcb4744bc..e6310a7cc05b 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -1836,6 +1836,10 @@ long populate_vma_page_range(struct vm_area_struct *vma,
 	if (!vma_is_accessible(vma))
 		return -EFAULT;
 
+	/* Unreadable VMAs also cannot be faulted in. */
+	if (!vma_test(vma, VMA_MAYREAD_BIT))
+		return -EFAULT;
+
 	gup_flags = FOLL_TOUCH;
 	/*
 	 * We want to touch writable mappings with a write fault in order

-- 
2.55.0

Re: [PATCH v6 21/39] mm/gup: error out early on !VMA_MAYREAD_BIT VMAs

[sashiko-bot]

> When populating a VMA range via the aptly named populate_vma_page_range()
> an unreadable VMA will always eventually fail with -EFAULT.
> 
> That a VMA is accessible is always checked, however VMA_MAYREAD_BIT is not.
> 
> All user mappings always have VMA_MAYREAD_BIT set, so this check only
> impacts kernel mappings.
> 
> It is implemented specifically to disallow population of uprobes XOL
> mappings which are exec-only.
> 
> A nasty interaction with these mappings may occur if they are mlocked, so
> actively disallow this early.
> 
> This allows a subsequent commit to remove the VM_IO check in
> __mm_populate() which otherwise requires non-MMIO mappings to be wrongly
> flagged simply as a workaround.
> 
> Acked-by: Zi Yan <ziy@nvidia.com>
> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> Acked-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013-b4-mmap-prepare-vma-flag-sanify-v6-0-1aa3c9b139f9@kernel.org?part=21

[PATCH v6 22/39] uprobes: remove VM_IO, set VM_MIXEDMAP for mapped kernel pages

[Lorenzo Stoakes (ARM)]

These are not MMIO pages so VMA_IO_BIT is an inappropriate flag to set.

Instead, set them VMA_MIXEDMAP_BIT as they are kernel mappings and this is
the appropriate flag to set for those.

This provides the semantics required - no VMA merging is permitted, but
does not prevent GUP.

However this has no meaningful impact as these are refcounted and thus can
be GUPed.

A previous commit already prevented __mm_populate() from being invoked on
XOL areas, which VMA_IO_BIT was previously relied upon to do, so that is no
longer required.

Both VMAs set a VMA name, so always_dump_vma() returns true before
vma_dump_size() reaches its VMA_IO_BIT check, and thus there is no change
in core dump behaviour.

Change this for both the core xol_add_vma() function and the x86-specific
get_uprobe_trampoline() function.

Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 arch/x86/kernel/uprobes.c | 2 +-
 kernel/events/uprobes.c   | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c
index 65a2de82ecd2..0f60c0d076b6 100644
--- a/arch/x86/kernel/uprobes.c
+++ b/arch/x86/kernel/uprobes.c
@@ -715,7 +715,7 @@ static struct vm_area_struct *get_uprobe_trampoline(struct mm_struct *mm, unsign
 
 	*new_mapping = true;
 	return _install_special_mapping(mm, vaddr, PAGE_SIZE,
-				VM_READ|VM_EXEC|VM_MAYEXEC|VM_MAYREAD|VM_IO,
+				VM_READ|VM_EXEC|VM_MAYEXEC|VM_MAYREAD|VM_MIXEDMAP,
 				&tramp_mapping);
 }
 
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index 7709ea882477..b89cc5cee002 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -1726,8 +1726,8 @@ static int xol_add_vma(struct mm_struct *mm, struct xol_area *area)
 	}
 
 	vma = _install_special_mapping(mm, area->vaddr, PAGE_SIZE,
-				VM_EXEC|VM_MAYEXEC|VM_DONTCOPY|VM_IO|
-				VM_SEALED_SYSMAP,
+				VM_EXEC|VM_MAYEXEC|VM_DONTCOPY|
+				VM_MIXEDMAP|VM_SEALED_SYSMAP,
 				&xol_mapping);
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);

-- 
2.55.0

[PATCH v6 23/39] mm/mlock: clear VMA_LOCKED_MASK over mmap callback

[Lorenzo Stoakes (ARM)]

Currently there's a confusing mess around VMA_LOCKED_BIT and
VMA_LOCKONFAULT_BIT.

It is permitted for drivers to set any flags they like, with the VMA
already possessing lock flags.

This results in the absurd situation of a VMA possessing both
VMA_SPECIAL_FLAGS and VMA_LOCKED_MASK flags, which is not permitted.

This has resulted in mlock_vma_folio() having a very silly check for this
scenario to work around it.

There is no need for this - just clear the flags before invoking the hook
and reinstate them afterwards if they are required.

Nothing relies upon this being set during the mmap operation.

mmap_prepare is unaffected by this so requires no fix.

Reviewed-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/internal.h |  9 +--------
 mm/vma.c      | 15 +++++++++++++++
 2 files changed, 16 insertions(+), 8 deletions(-)

diff --git a/mm/internal.h b/mm/internal.h
index 557e629db8c5..4ae08fedffff 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -970,14 +970,7 @@ void mlock_folio(struct folio *folio);
 static inline void mlock_vma_folio(struct folio *folio,
 				struct vm_area_struct *vma)
 {
-	/*
-	 * The VM_SPECIAL check here serves two purposes.
-	 * 1) VM_IO check prevents migration from double-counting during mlock.
-	 * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED
-	 *    is never left set on a VM_SPECIAL vma, there is an interval while
-	 *    file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may
-	 *    still be set while VM_SPECIAL bits are added: so ignore it then.
-	 */
+	/* The VM_IO check prevents migration from double-counting during mlock. */
 	if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED))
 		mlock_folio(folio);
 }
diff --git a/mm/vma.c b/mm/vma.c
index 39ceb03023e2..6cfda4575997 100644
--- a/mm/vma.c
+++ b/mm/vma.c
@@ -2607,6 +2607,11 @@ static int __mmap_new_file_vma(struct mmap_state *map,
 	if (!map->vm_file->f_op->mmap)
 		return 0;
 
+	/*
+	 * Driver-specified flags may make the lock flags invalid, so clear
+	 * VMA_LOCKED_MASK and reinstate it afterwards if appropriate.
+	 */
+	vma_clear_flags_mask(vma, VMA_LOCKED_MASK);
 	error = mmap_file(vma->vm_file, vma);
 	map->vm_file = vma->vm_file;
 
@@ -2629,6 +2634,16 @@ static int __mmap_new_file_vma(struct mmap_state *map,
 		return error;
 	}
 
+	/*
+	 * If the VMA is still eligible for mlock(), reinstate any original
+	 * VMA_LOCKED_BIT and/or VMA_LOCKONFAULT_BIT flags.
+	 */
+	if (vma_supports_mlock(vma)) {
+		const vma_flags_t mask = vma_flags_and_mask(&map->vma_flags, VMA_LOCKED_MASK);
+
+		vma_set_flags_mask(vma, mask);
+	}
+
 	map->vma_flags = vma->flags;
 
 	return 0;

-- 
2.55.0

[PATCH v6 24/39] mm/mlock: eliminate weird VMA_IO_BIT abuse and simplify

[Lorenzo Stoakes (ARM)]

When performing mlock() or munlock() otherwise normal VMAs have VMA_IO_BIT
solely to fix a race with migration which might otherwise double-count
mlock VMAs.

This is unnecessary - at the point of applying folio mlock state, whether
setting or clearing PG_mlocked, we know whether or not we are locking.

Solve this in two ways - thread a boolean through the page table walk
indicating whether a lock or unlock is being performed, and run a locking
walk with VMA_LOCKONFAULT_BIT set and VMA_LOCKED_BIT cleared.

This state never occurs otherwise, as VMA_LOCKONFAULT_BIT always implies
VMA_LOCKED_BIT. These are also always cleared together.

Then, update folio_add_lru_vma() and mlock_folio() to check only for
VMA_LOCKED_BIT, and update try_to_unmap_one() to check for VMA_LOCKED_MASK
instead.

Also remove the useless invocation of allow_mlock_munlock() which simply
returns true if unlocking and instead rename it to allow_mlock() and only
call it when locking.

With the other mlock abuse of VMA_IO_BIT addressed, update
mlock_vma_folio() and folio_add_lru_vma() to simply test for
VMA_LOCKED_BIT.

The munlock_vma_folio() function tests VMA_LOCKED_MASK instead, as an
unmap racing with the locking walk must still munlock folios the walk has
already counted.

Additionally, in order to ensure a racing unmap/migration releases the
reference munlock_folio() took (whose put was batched), check
VMA_LOCKED_MASK before calling mlock_drain_local() in try_to_unmap_one()
and try_to_migrate_one().

While here, also replace some deprecated VMA flag predicates.

Reviewed-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/folio.c    |  2 +-
 mm/internal.h | 10 +++++++---
 mm/mlock.c    | 51 +++++++++++++++++++--------------------------------
 mm/rmap.c     |  8 +++++---
 4 files changed, 32 insertions(+), 39 deletions(-)

diff --git a/mm/folio.c b/mm/folio.c
index 47a437e0f7fd..35e242b48870 100644
--- a/mm/folio.c
+++ b/mm/folio.c
@@ -505,7 +505,7 @@ void folio_add_lru_vma(struct folio *folio, struct vm_area_struct *vma)
 {
 	VM_BUG_ON_FOLIO(folio_test_lru(folio), folio);
 
-	if (unlikely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) == VM_LOCKED))
+	if (vma_test(vma, VMA_LOCKED_BIT))
 		mlock_new_folio(folio);
 	else
 		folio_add_lru(folio);
diff --git a/mm/internal.h b/mm/internal.h
index 4ae08fedffff..ea8fa94ebe1b 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -970,8 +970,7 @@ void mlock_folio(struct folio *folio);
 static inline void mlock_vma_folio(struct folio *folio,
 				struct vm_area_struct *vma)
 {
-	/* The VM_IO check prevents migration from double-counting during mlock. */
-	if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED))
+	if (vma_test(vma, VMA_LOCKED_BIT))
 		mlock_folio(folio);
 }
 
@@ -988,7 +987,12 @@ static inline void munlock_vma_folio(struct folio *folio,
 	 * always munlock the folio and page reclaim will correct it
 	 * if it's wrong.
 	 */
-	if (unlikely(vma->vm_flags & VM_LOCKED))
+	/*
+	 * VMA_LOCKONFAULT_BIT alone marks an mlock walk in progress, see
+	 * mlock_vma_pages_range(). An unmap racing with the walk must still
+	 * munlock folios the walk has already counted.
+	 */
+	if (unlikely(vma_test_any_mask(vma, VMA_LOCKED_MASK)))
 		munlock_folio(folio);
 }
 
diff --git a/mm/mlock.c b/mm/mlock.c
index 39215a3eab1f..4235a1518fc9 100644
--- a/mm/mlock.c
+++ b/mm/mlock.c
@@ -316,22 +316,10 @@ static inline unsigned int folio_mlock_step(struct folio *folio,
 	return folio_pte_batch(folio, pte, ptent, count);
 }
 
-static inline bool allow_mlock_munlock(struct folio *folio,
+static inline bool allow_mlock(struct folio *folio,
 		struct vm_area_struct *vma, unsigned long start,
 		unsigned long end, unsigned int step)
 {
-	/*
-	 * For unlock, allow munlock large folio which is partially
-	 * mapped to VMA. As it's possible that large folio is
-	 * mlocked and VMA is split later.
-	 *
-	 * During memory pressure, such kind of large folio can
-	 * be split. And the pages are not in VM_LOCKed VMA
-	 * can be reclaimed.
-	 */
-	if (!vma_test(vma, VMA_LOCKED_BIT))
-		return true;
-
 	/* folio_within_range() cannot take KSM, but any small folio is OK */
 	if (!folio_test_large(folio))
 		return true;
@@ -352,6 +340,7 @@ static int mlock_pte_range(pmd_t *pmd, unsigned long addr,
 
 {
 	struct vm_area_struct *vma = walk->vma;
+	const bool lock = walk->private;
 	spinlock_t *ptl;
 	pte_t *start_pte, *pte;
 	pte_t ptent;
@@ -368,7 +357,7 @@ static int mlock_pte_range(pmd_t *pmd, unsigned long addr,
 		folio = pmd_folio(*pmd);
 		if (folio_is_zone_device(folio))
 			goto out;
-		if (vma_test(vma, VMA_LOCKED_BIT))
+		if (lock)
 			mlock_folio(folio);
 		else
 			munlock_folio(folio);
@@ -390,10 +379,10 @@ static int mlock_pte_range(pmd_t *pmd, unsigned long addr,
 			continue;
 
 		step = folio_mlock_step(folio, pte, addr, end);
-		if (!allow_mlock_munlock(folio, vma, start, end, step))
+		if (lock && !allow_mlock(folio, vma, start, end, step))
 			goto next_entry;
 
-		if (vma_test(vma, VMA_LOCKED_BIT))
+		if (lock)
 			mlock_folio(folio);
 		else
 			munlock_folio(folio);
@@ -428,31 +417,29 @@ static void mlock_vma_pages_range(struct vm_area_struct *vma,
 		.pmd_entry = mlock_pte_range,
 		.walk_lock = PGWALK_WRLOCK_VERIFY,
 	};
+	const bool lock = vma_flags_test(new_vma_flags, VMA_LOCKED_BIT);
+	vma_flags_t walk_flags = *new_vma_flags;
 
 	/*
-	 * There is a slight chance that concurrent page migration,
-	 * or page reclaim finding a page of this now-VMA_LOCKED_BIT vma,
-	 * will call mlock_vma_folio() and raise page's mlock_count:
-	 * double counting, leaving the page unevictable indefinitely.
-	 * Communicate this danger to mlock_vma_folio() with VMA_IO_BIT,
-	 * which is a VMA_SPECIAL_FLAGS flag not allowed on VMA_LOCKED_BIT vmas.
-	 * mmap_lock is held in write mode here, so this weird
-	 * combination should not be visible to other mmap_lock users;
-	 * but WRITE_ONCE so rmap walkers must see VMA_IO_BIT if VMA_LOCKED_BIT.
+	 * LOCKONFAULT without LOCKED never otherwise occurs: it marks a walk in
+	 * progress so that rmap-side callers, which test VMA_LOCKED_BIT, do not
+	 * count folios, while try_to_unmap_one(), which tests VMA_LOCKED_MASK,
+	 * still refuses to unmap them.
 	 */
-	if (vma_flags_test(new_vma_flags, VMA_LOCKED_BIT))
-		vma_flags_set(new_vma_flags, VMA_IO_BIT);
+	if (lock) {
+		vma_flags_clear(&walk_flags, VMA_LOCKED_BIT);
+		vma_flags_set(&walk_flags, VMA_LOCKONFAULT_BIT);
+	}
+
 	vma_start_write(vma);
-	vma_flags_reset_once(vma, new_vma_flags);
+	vma_flags_reset_once(vma, &walk_flags);
 
 	lru_add_drain();
-	walk_page_range_vma(vma, start, end, &mlock_walk_ops, NULL);
+	walk_page_range_vma(vma, start, end, &mlock_walk_ops, (void *)lock);
 	lru_add_drain();
 
-	if (vma_flags_test(new_vma_flags, VMA_IO_BIT)) {
-		vma_flags_clear(new_vma_flags, VMA_IO_BIT);
+	if (lock)
 		vma_flags_reset_once(vma, new_vma_flags);
-	}
 }
 
 /*
diff --git a/mm/rmap.c b/mm/rmap.c
index 5332c52909be..b4a4b3929044 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -2239,9 +2239,11 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
 
 		/*
 		 * If the folio is in an mlock()d vma, we must not swap it out.
+		 * VMA_LOCKONFAULT_BIT alone marks an mlock walk in progress, see
+		 * mlock_vma_pages_range().
 		 */
 		if (!(flags & TTU_IGNORE_MLOCK) &&
-		    (vma->vm_flags & VM_LOCKED)) {
+		    vma_test_any_mask(vma, VMA_LOCKED_MASK)) {
 			ptes++;
 
 			/*
@@ -2395,7 +2397,7 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
 		}
 finish_unmap:
 		folio_remove_rmap_ptes(folio, page, nr_pages, vma);
-		if (vma->vm_flags & VM_LOCKED)
+		if (vma_test_any_mask(vma, VMA_LOCKED_MASK))
 			mlock_drain_local();
 		folio_put_refs(folio, nr_pages);
 
@@ -2770,7 +2772,7 @@ static bool try_to_migrate_one(struct folio *folio, struct vm_area_struct *vma,
 			hugetlb_remove_rmap(folio);
 		else
 			folio_remove_rmap_pte(folio, subpage, vma);
-		if (vma->vm_flags & VM_LOCKED)
+		if (vma_test_any_mask(vma, VMA_LOCKED_MASK))
 			mlock_drain_local();
 		folio_put(folio);
 	}

-- 
2.55.0

[PATCH v6 25/39] mm/vma: enforce that mm-managed mappings may not set VMA_IO_BIT

[Lorenzo Stoakes (ARM)]

It makes no sense for a mapping whose contents the kernel does not own to
specify that the range is MMIO.

Prior to this patch, all in-tree drivers which did so have been updated
such that they are marked as not being mm-managed. The check WARNs and
fails the mmap for any out-of-tree driver that still sets VMA_IO_BIT
without a kernel mapping.

Here VMA_PFNMAP_BIT, VMA_MIXEDMAP are tested individually rather than using
vma_flags_is_mm_managed().

This is because vma_flags_is_mm_managed() still checks for VMA_IO_BIT.

The very next commit in this series removes this, and the check is then
converted to use vma_flags_is_mm_managed().

No functional change intended for in-tree code.

Reviewed-by: Zi Yan <ziy@nvidia.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/vma.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/mm/vma.c b/mm/vma.c
index 6cfda4575997..affe464a7c92 100644
--- a/mm/vma.c
+++ b/mm/vma.c
@@ -2794,6 +2794,12 @@ static int mmap_validate_vma_flags(const vma_flags_t *flags)
 		return -EINVAL;
 #endif
 
+	if (!vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT)) {
+		/* mm-managed mappings may not set VMA_IO_BIT. */
+		if (WARN_ON_ONCE(vma_flags_test(flags, VMA_IO_BIT)))
+			return -EINVAL;
+	}
+
 	return 0;
 }
 

-- 
2.55.0

Re: [PATCH v6 25/39] mm/vma: enforce that mm-managed mappings may not set VMA_IO_BIT

[David Hildenbrand (Arm)]

On 10/13/26 17:05, Lorenzo Stoakes (ARM) wrote:
> It makes no sense for a mapping whose contents the kernel does not own to
> specify that the range is MMIO.
> 
> Prior to this patch, all in-tree drivers which did so have been updated
> such that they are marked as not being mm-managed. The check WARNs and
> fails the mmap for any out-of-tree driver that still sets VMA_IO_BIT
> without a kernel mapping.
> 
> Here VMA_PFNMAP_BIT, VMA_MIXEDMAP are tested individually rather than using
> vma_flags_is_mm_managed().
> 
> This is because vma_flags_is_mm_managed() still checks for VMA_IO_BIT.
> 
> The very next commit in this series removes this, and the check is then
> converted to use vma_flags_is_mm_managed().
> 
> No functional change intended for in-tree code.
> 
> Reviewed-by: Zi Yan <ziy@nvidia.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>

Acked-by: David Hildenbrand (Arm) <david@kernel.org>

-- 
Cheers,

David

[PATCH v6 26/39] mm: remove VMA_IO_BIT check in vma[_flags]_is_mm_managed()

[Lorenzo Stoakes (ARM)]

We have now made it such that every driver which sets VMA_IO_BIT marks it
as not mm-managed.

However, vma_flags_is_mm_managed() still checks for VMA_IO_BIT and, if
set, treats the mapping as not mm-managed. This was previously required to
handle drivers which set VMA_IO_BIT but none of the other flags which
identify a mapping as not mm-managed.

We have now established that a driver which sets VMA_IO_BIT must also set
one of those flags, so the VMA_IO_BIT check is redundant.

So remove the VMA_IO_BIT check in vma_flags_is_mm_managed() and update
mmap_validate_vma_flags() to use vma_flags_is_mm_managed() rather than
open-coding the VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT check.

This change means that vma[_flags]_can_merge() doesn't check VMA_IO_BIT any
longer (which is now redundant) as it calls vma_flags_is_mm_managed().

Now that the predicate means precisely VMA_PFNMAP_BIT or VMA_MIXEDMAP_BIT,
also use it at the other sites which open-code that pair, so the intent is
stated rather than the flags, with no functional change:

zap_special_vma_range() only zaps mappings which are not mm-managed, as
drivers use it to tear down ranges they established themselves.

The mprotect() arch PFN modification check applies to mappings not managed
by core mm, which may map PFNs without struct pages.

NUMA balancing skips VM_MIXEDMAP mappings having already excluded VM_IO
and VM_PFNMAP mappings via vma_migratable(), so it skips exactly the
mappings not managed by core mm - say so.

Finally, update the VMA userland merge 'special' flag tests to no longer
assert that VMA_IO_BIT prevents merge as VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT
now suffices.

Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 include/linux/mm.h              |  3 +--
 kernel/sched/fair.c             |  2 +-
 mm/memory.c                     |  7 ++++---
 mm/mprotect.c                   |  3 +--
 mm/vma.c                        |  2 +-
 tools/testing/vma/include/dup.h |  3 +--
 tools/testing/vma/tests/merge.c | 10 ++--------
 7 files changed, 11 insertions(+), 19 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 2ecf39bc857b..1006d8567e06 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1627,8 +1627,7 @@ static inline bool vma_is_shared_maywrite(const struct vm_area_struct *vma)
  */
 static inline bool vma_flags_is_mm_managed(const vma_flags_t *flags)
 {
-	return !vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT,
-				   VMA_IO_BIT);
+	return !vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT);
 }
 
 /**
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 8dff37059faf..a07a658cd99f 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4212,7 +4212,7 @@ static void task_numa_work(struct callback_head *work)
 
 	for (; vma; vma = vma_next(&vmi)) {
 		if (!vma_migratable(vma) || !vma_policy_mof(vma) ||
-			is_vm_hugetlb_page(vma) || (vma->vm_flags & VM_MIXEDMAP)) {
+			is_vm_hugetlb_page(vma) || !vma_is_mm_managed(vma)) {
 			trace_sched_skip_vma_numa(mm, vma, NUMAB_SKIP_UNSUITABLE);
 			continue;
 		}
diff --git a/mm/memory.c b/mm/memory.c
index 1a6bb0aef8cf..142a8ef13f6a 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2343,19 +2343,20 @@ void zap_vma_range(struct vm_area_struct *vma, unsigned long address,
 }
 
 /**
- * zap_special_vma_range - zap all page table entries in a special vma range
+ * zap_special_vma_range - zap all page table entries in a VMA which is not
+ * mm-managed
  * @vma: the vma covering the range to zap
  * @address: starting address of the range to zap
  * @size: number of bytes to zap
  *
  * This function does nothing when the provided address range is not fully
- * contained in @vma, or when the @vma is not VM_PFNMAP or VM_MIXEDMAP.
+ * contained in @vma, or when @vma is mm-managed.
  */
 void zap_special_vma_range(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size)
 {
 	if (!range_in_vma(vma, address, address + size) ||
-	   !(vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)))
+	   vma_is_mm_managed(vma))
 		return;
 
 	zap_vma_range(vma, address, size);
diff --git a/mm/mprotect.c b/mm/mprotect.c
index 2888ee638d87..706c679d4281 100644
--- a/mm/mprotect.c
+++ b/mm/mprotect.c
@@ -783,8 +783,7 @@ mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
 	 * uncommon case, so doesn't need to be very optimized.
 	 */
 	if (arch_has_pfn_modify_check() &&
-	    vma_flags_test_any(&old_vma_flags, VMA_PFNMAP_BIT,
-			       VMA_MIXEDMAP_BIT) &&
+	    !vma_flags_is_mm_managed(&old_vma_flags) &&
 	    !vma_flags_test_any_mask(&new_vma_flags, VMA_ACCESS_FLAGS)) {
 		pgprot_t new_pgprot = vm_get_page_prot(newflags);
 
diff --git a/mm/vma.c b/mm/vma.c
index affe464a7c92..5df02665d1ed 100644
--- a/mm/vma.c
+++ b/mm/vma.c
@@ -2794,7 +2794,7 @@ static int mmap_validate_vma_flags(const vma_flags_t *flags)
 		return -EINVAL;
 #endif
 
-	if (!vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT)) {
+	if (vma_flags_is_mm_managed(flags)) {
 		/* mm-managed mappings may not set VMA_IO_BIT. */
 		if (WARN_ON_ONCE(vma_flags_test(flags, VMA_IO_BIT)))
 			return -EINVAL;
diff --git a/tools/testing/vma/include/dup.h b/tools/testing/vma/include/dup.h
index 442ea9eb098d..94bf818ad0e1 100644
--- a/tools/testing/vma/include/dup.h
+++ b/tools/testing/vma/include/dup.h
@@ -1667,8 +1667,7 @@ static inline bool file_is_dev_zero(const struct file *file)
 
 static inline bool vma_flags_is_mm_managed(const vma_flags_t *flags)
 {
-	return !vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT,
-				   VMA_IO_BIT);
+	return !vma_flags_test_any(flags, VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT);
 }
 
 static inline bool vma_is_mm_managed(const struct vm_area_struct *vma)
diff --git a/tools/testing/vma/tests/merge.c b/tools/testing/vma/tests/merge.c
index acaab282939c..b26f1a66a170 100644
--- a/tools/testing/vma/tests/merge.c
+++ b/tools/testing/vma/tests/merge.c
@@ -496,17 +496,11 @@ static bool test_vma_merge_special_flags(void)
 		.mm = &mm,
 		.vmi = &vmi,
 	};
-	vma_flag_t special_flags[] = { VMA_IO_BIT, VMA_DONTEXPAND_BIT,
+	vma_flag_t special_flags[] = { VMA_DONTEXPAND_BIT,
 		VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT };
-	vma_flags_t all_special_flags = EMPTY_VMA_FLAGS;
 	int i;
 	struct vm_area_struct *vma_left, *vma;
 
-	/* Make sure there aren't new VM_SPECIAL flags. */
-	for (i = 0; i < ARRAY_SIZE(special_flags); i++)
-		vma_flags_set(&all_special_flags, special_flags[i]);
-	ASSERT_FLAGS_SAME_MASK(&all_special_flags, VMA_SPECIAL_FLAGS);
-
 	/*
 	 * 01234
 	 * AAA
@@ -520,7 +514,7 @@ static bool test_vma_merge_special_flags(void)
 	 * 01234
 	 * AAA*
 	 *
-	 * This should merge if not for the VM_SPECIAL flag.
+	 * This should merge if not for the 'special' flag.
 	 */
 	vmg_set_range(&vmg, 0x3000, 0x4000, 3, vma_flags);
 	for (i = 0; i < ARRAY_SIZE(special_flags); i++) {

-- 
2.55.0

[PATCH v6 27/39] mm: remove hugetlb_inline.h

[Lorenzo Stoakes (ARM)]

This header really makes little sense - every place it is included mm.h is
also included, and the header itself includes mm.h, so it does nothing to
reduce header size.

It also oddly does an #ifdef around checking VMA_HUGETLB_BIT, however
VMA_HUGETLB_BIT is unconditionally available, and will never be set if
hugetlb is not enabled.

Simply remove the header, eliminate the odd ifdeffery and place the
predicates in mm.h.

The naming of these predicates is odd, but to keep changes separate, we
will address this in a separate patch.

The file was never put into MAINTAINERS so there's no change required
there.

No functional change intended.

Reviewed-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 drivers/gpu/drm/drm_gpusvm.c   |  2 +-
 include/asm-generic/tlb.h      |  2 +-
 include/linux/hugetlb.h        |  1 -
 include/linux/hugetlb_inline.h | 28 ----------------------------
 include/linux/mm.h             | 11 +++++++++++
 include/linux/pagemap.h        |  1 -
 include/linux/userfaultfd_k.h  |  1 -
 kernel/sched/fair.c            |  1 -
 mm/vma_internal.h              |  1 -
 9 files changed, 13 insertions(+), 35 deletions(-)

diff --git a/drivers/gpu/drm/drm_gpusvm.c b/drivers/gpu/drm/drm_gpusvm.c
index a93eee7ddb9e..793dacec2100 100644
--- a/drivers/gpu/drm/drm_gpusvm.c
+++ b/drivers/gpu/drm/drm_gpusvm.c
@@ -9,9 +9,9 @@
 #include <linux/dma-mapping.h>
 #include <linux/export.h>
 #include <linux/hmm.h>
-#include <linux/hugetlb_inline.h>
 #include <linux/memremap.h>
 #include <linux/mm_types.h>
+#include <linux/mm.h>
 #include <linux/slab.h>
 
 #include <drm/drm_device.h>
diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h
index 044dabc1fe9c..8b8123e5b7d7 100644
--- a/include/asm-generic/tlb.h
+++ b/include/asm-generic/tlb.h
@@ -11,9 +11,9 @@
 #ifndef _ASM_GENERIC__TLB_H
 #define _ASM_GENERIC__TLB_H
 
+#include <linux/mm.h>
 #include <linux/mmu_notifier.h>
 #include <linux/swap.h>
-#include <linux/hugetlb_inline.h>
 #include <asm/tlbflush.h>
 #include <asm/cacheflush.h>
 
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 80a5a03e9cee..d7e6563cef75 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -7,7 +7,6 @@
 #include <linux/mm_types.h>
 #include <linux/mmdebug.h>
 #include <linux/fs.h>
-#include <linux/hugetlb_inline.h>
 #include <linux/cgroup.h>
 #include <linux/page_ref.h>
 #include <linux/list.h>
diff --git a/include/linux/hugetlb_inline.h b/include/linux/hugetlb_inline.h
deleted file mode 100644
index 5c29cd3223a1..000000000000
--- a/include/linux/hugetlb_inline.h
+++ /dev/null
@@ -1,28 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-#ifndef _LINUX_HUGETLB_INLINE_H
-#define _LINUX_HUGETLB_INLINE_H
-
-#include <linux/mm.h>
-
-#ifdef CONFIG_HUGETLB_PAGE
-
-static inline bool is_vma_hugetlb_flags(const vma_flags_t *flags)
-{
-	return vma_flags_test(flags, VMA_HUGETLB_BIT);
-}
-
-#else
-
-static inline bool is_vma_hugetlb_flags(const vma_flags_t *flags)
-{
-	return false;
-}
-
-#endif
-
-static inline bool is_vm_hugetlb_page(const struct vm_area_struct *vma)
-{
-	return is_vma_hugetlb_flags(&vma->flags);
-}
-
-#endif
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 1006d8567e06..2c41974704a9 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1609,6 +1609,17 @@ static inline bool vma_is_shared_maywrite(const struct vm_area_struct *vma)
 	return is_shared_maywrite(&vma->flags);
 }
 
+static inline bool is_vma_hugetlb_flags(const vma_flags_t *flags)
+{
+	return IS_ENABLED(CONFIG_HUGETLB_PAGE) &&
+	       vma_flags_test(flags, VMA_HUGETLB_BIT);
+}
+
+static inline bool is_vm_hugetlb_page(const struct vm_area_struct *vma)
+{
+	return is_vma_hugetlb_flags(&vma->flags);
+}
+
 /**
  * vma_flags_is_mm_managed() - Do the specified VMA flags indicate that the
  * contents of the VMA are managed by core mm?
diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 939f3a5e973f..d7d8b312466c 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -14,7 +14,6 @@
 #include <linux/gfp.h>
 #include <linux/bitops.h>
 #include <linux/hardirq.h> /* for in_interrupt() */
-#include <linux/hugetlb_inline.h>
 
 struct folio_batch;
 
diff --git a/include/linux/userfaultfd_k.h b/include/linux/userfaultfd_k.h
index a4351cffc60c..a14b8a9ffb7b 100644
--- a/include/linux/userfaultfd_k.h
+++ b/include/linux/userfaultfd_k.h
@@ -18,7 +18,6 @@
 #include <linux/swap.h>
 #include <linux/leafops.h>
 #include <asm-generic/pgtable_uffd.h>
-#include <linux/hugetlb_inline.h>
 
 /* The set of all possible UFFD-related VM flags. */
 #define __VM_UFFD_FLAGS (VM_UFFD_MISSING | VM_UFFD_MINOR | \
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index a07a658cd99f..9eb12d0b6f1b 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -22,7 +22,6 @@
  */
 #include <linux/energy_model.h>
 #include <linux/mmap_lock.h>
-#include <linux/hugetlb_inline.h>
 #include <linux/jiffies.h>
 #include <linux/mm_api.h>
 #include <linux/highmem.h>
diff --git a/mm/vma_internal.h b/mm/vma_internal.h
index 4d300e7bbaf4..4f73f0a4db79 100644
--- a/mm/vma_internal.h
+++ b/mm/vma_internal.h
@@ -18,7 +18,6 @@
 #include <linux/fs.h>
 #include <linux/huge_mm.h>
 #include <linux/hugetlb.h>
-#include <linux/hugetlb_inline.h>
 #include <linux/kernel.h>
 #include <linux/ksm.h>
 #include <linux/khugepaged.h>

-- 
2.55.0

[PATCH v6 28/39] mm: rename is_vm_hugetlb_page() to vma_is_hugetlb()

[Lorenzo Stoakes (ARM)]

The is_vm_hugetlb_page() predicate is badly named - the mapping can span
more than a page and it is inconsistent with other VMA predicates that
typically are prefixed by vma_.

Rename to vma_is_hugetlb() for consistency, and while we're here update
some VM_BUG_ON_VMA() to VM_WARN_ON_ONCE_VMA() as to avoid unnecessary
oopses.

No functional change intended.

Acked-by: Marc Zyngier <maz@kernel.org>
Acked-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Acked-by: Anup Patel <anup@brainfault.org>
Acked-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 arch/arm64/kvm/mmu.c                      |  4 ++--
 arch/powerpc/mm/book3s64/radix_tlb.c      |  6 +++---
 arch/powerpc/mm/nohash/e500_hugetlbpage.c |  2 +-
 arch/powerpc/mm/nohash/tlb.c              |  2 +-
 arch/riscv/kvm/mmu.c                      |  2 +-
 arch/riscv/mm/tlbflush.c                  |  2 +-
 arch/s390/mm/gmap_helpers.c               |  6 +++---
 arch/sparc/mm/init_64.c                   |  2 +-
 drivers/gpu/drm/drm_gpusvm.c              |  2 +-
 fs/coredump.c                             |  2 +-
 fs/hugetlbfs/inode.c                      |  2 +-
 fs/proc/task_mmu.c                        |  8 ++++----
 include/asm-generic/tlb.h                 |  2 +-
 include/linux/hugetlb.h                   |  4 ++--
 include/linux/mm.h                        | 19 ++++++++++++++++---
 include/linux/rmap.h                      |  2 +-
 kernel/events/core.c                      |  2 +-
 kernel/sched/fair.c                       |  2 +-
 mm/gup.c                                  |  4 ++--
 mm/huge_memory.c                          |  2 +-
 mm/hugetlb.c                              | 14 +++++++-------
 mm/internal.h                             |  2 +-
 mm/madvise.c                              |  4 ++--
 mm/memory.c                               | 12 ++++++------
 mm/mempolicy.c                            |  2 +-
 mm/migrate_device.c                       |  2 +-
 mm/mmap.c                                 |  2 +-
 mm/mmu_gather.c                           |  2 +-
 mm/mprotect.c                             |  2 +-
 mm/mremap.c                               |  6 +++---
 mm/page_vma_mapped.c                      |  4 ++--
 mm/pagewalk.c                             |  2 +-
 mm/swapfile.c                             |  2 +-
 mm/userfaultfd.c                          | 26 +++++++++++++-------------
 mm/vma.c                                  |  8 ++++----
 mm/vmscan.c                               |  2 +-
 tools/testing/vma/include/stubs.h         |  2 +-
 37 files changed, 92 insertions(+), 79 deletions(-)

diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index 9ba86450fe4a..a7968f8d24bf 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -1463,13 +1463,13 @@ static int get_vma_page_shift(struct vm_area_struct *vma, unsigned long hva)
 {
 	unsigned long pa;
 
-	if (is_vm_hugetlb_page(vma) && !(vma->vm_flags & VM_PFNMAP))
+	if (vma_is_hugetlb(vma) && !(vma->vm_flags & VM_PFNMAP))
 		return huge_page_shift(hstate_vma(vma));
 
 	if (!(vma->vm_flags & VM_PFNMAP))
 		return PAGE_SHIFT;
 
-	VM_BUG_ON(is_vm_hugetlb_page(vma));
+	VM_BUG_ON(vma_is_hugetlb(vma));
 
 	pa = (vma->vm_pgoff << PAGE_SHIFT) + (hva - vma->vm_start);
 
diff --git a/arch/powerpc/mm/book3s64/radix_tlb.c b/arch/powerpc/mm/book3s64/radix_tlb.c
index 7de5760164a9..b4603a98224b 100644
--- a/arch/powerpc/mm/book3s64/radix_tlb.c
+++ b/arch/powerpc/mm/book3s64/radix_tlb.c
@@ -627,7 +627,7 @@ void radix__local_flush_tlb_page(struct vm_area_struct *vma, unsigned long vmadd
 {
 #ifdef CONFIG_HUGETLB_PAGE
 	/* need the return fix for nohash.c */
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		return radix__local_flush_hugetlb_page(vma, vmaddr);
 #endif
 	radix__local_flush_tlb_page_psize(vma->vm_mm, vmaddr, mmu_virtual_psize);
@@ -945,7 +945,7 @@ void radix__flush_tlb_page_psize(struct mm_struct *mm, unsigned long vmaddr,
 void radix__flush_tlb_page(struct vm_area_struct *vma, unsigned long vmaddr)
 {
 #ifdef CONFIG_HUGETLB_PAGE
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		return radix__flush_hugetlb_page(vma, vmaddr);
 #endif
 	radix__flush_tlb_page_psize(vma->vm_mm, vmaddr, mmu_virtual_psize);
@@ -1113,7 +1113,7 @@ void radix__flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
 
 {
 #ifdef CONFIG_HUGETLB_PAGE
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		return radix__flush_hugetlb_tlb_range(vma, start, end);
 #endif
 
diff --git a/arch/powerpc/mm/nohash/e500_hugetlbpage.c b/arch/powerpc/mm/nohash/e500_hugetlbpage.c
index a134d28a0e4d..b87623f04be5 100644
--- a/arch/powerpc/mm/nohash/e500_hugetlbpage.c
+++ b/arch/powerpc/mm/nohash/e500_hugetlbpage.c
@@ -180,7 +180,7 @@ book3e_hugetlb_preload(struct vm_area_struct *vma, unsigned long ea, pte_t pte)
  */
 void __update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t *ptep)
 {
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		book3e_hugetlb_preload(vma, address, *ptep);
 }
 
diff --git a/arch/powerpc/mm/nohash/tlb.c b/arch/powerpc/mm/nohash/tlb.c
index 0a650742f3a0..07a2db16c2b1 100644
--- a/arch/powerpc/mm/nohash/tlb.c
+++ b/arch/powerpc/mm/nohash/tlb.c
@@ -278,7 +278,7 @@ void __flush_tlb_page(struct mm_struct *mm, unsigned long vmaddr,
 void flush_tlb_page(struct vm_area_struct *vma, unsigned long vmaddr)
 {
 #ifdef CONFIG_HUGETLB_PAGE
-	if (vma && is_vm_hugetlb_page(vma))
+	if (vma && vma_is_hugetlb(vma))
 		flush_hugetlb_page(vma, vmaddr);
 #endif
 
diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
index 6035b5ec9503..5c5c77f98bf0 100644
--- a/arch/riscv/kvm/mmu.c
+++ b/arch/riscv/kvm/mmu.c
@@ -664,7 +664,7 @@ int kvm_riscv_mmu_map(struct kvm_vcpu *vcpu, struct kvm_memory_slot *memslot,
 		return -EFAULT;
 	}
 
-	is_hugetlb = is_vm_hugetlb_page(vma);
+	is_hugetlb = vma_is_hugetlb(vma);
 	if (is_hugetlb)
 		vma_pageshift = huge_page_shift(hstate_vma(vma));
 	else
diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c
index 962db300a166..a74a7d5258aa 100644
--- a/arch/riscv/mm/tlbflush.c
+++ b/arch/riscv/mm/tlbflush.c
@@ -149,7 +149,7 @@ void flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
 {
 	unsigned long stride_size;
 
-	if (!is_vm_hugetlb_page(vma)) {
+	if (!vma_is_hugetlb(vma)) {
 		stride_size = PAGE_SIZE;
 	} else {
 		stride_size = huge_page_size(hstate_vma(vma));
diff --git a/arch/s390/mm/gmap_helpers.c b/arch/s390/mm/gmap_helpers.c
index ff63ffb1dbd2..3f6783b93e67 100644
--- a/arch/s390/mm/gmap_helpers.c
+++ b/arch/s390/mm/gmap_helpers.c
@@ -102,7 +102,7 @@ __context_unsafe(/* pte_unmap_unlock() not instrumented */)
 
 	/* Find the vm address for the guest address */
 	vma = vma_lookup(mm, vmaddr);
-	if (!vma || is_vm_hugetlb_page(vma))
+	if (!vma || vma_is_hugetlb(vma))
 		return;
 
 	/* Get pointer to the page table entry */
@@ -139,7 +139,7 @@ void gmap_helper_discard(struct mm_struct *mm, unsigned long vmaddr, unsigned lo
 		vma = find_vma_intersection(mm, vmaddr, end);
 		if (!vma)
 			return;
-		if (!is_vm_hugetlb_page(vma))
+		if (!vma_is_hugetlb(vma))
 			zap_vma_range(vma, vmaddr, min(end, vma->vm_end) - vmaddr);
 		vmaddr = vma->vm_end;
 	}
@@ -247,7 +247,7 @@ static int __gmap_helper_unshare_zeropages(struct mm_struct *mm)
 		 * proof to catch unexpected zeropages in other mappings and
 		 * fail.
 		 */
-		if ((vma->vm_flags & VM_PFNMAP) || is_vm_hugetlb_page(vma))
+		if ((vma->vm_flags & VM_PFNMAP) || vma_is_hugetlb(vma))
 			continue;
 		addr = vma->vm_start;
 
diff --git a/arch/sparc/mm/init_64.c b/arch/sparc/mm/init_64.c
index 103db4683b16..9bbccb5d23a8 100644
--- a/arch/sparc/mm/init_64.c
+++ b/arch/sparc/mm/init_64.c
@@ -413,7 +413,7 @@ void update_mmu_cache_range(struct vm_fault *vmf, struct vm_area_struct *vma,
 	if (mm->context.hugetlb_pte_count || mm->context.thp_pte_count) {
 		unsigned long hugepage_size = PAGE_SIZE;
 
-		if (is_vm_hugetlb_page(vma))
+		if (vma_is_hugetlb(vma))
 			hugepage_size = huge_page_size(hstate_vma(vma));
 
 		if (hugepage_size >= PUD_SIZE) {
diff --git a/drivers/gpu/drm/drm_gpusvm.c b/drivers/gpu/drm/drm_gpusvm.c
index 793dacec2100..a1d4989b0b61 100644
--- a/drivers/gpu/drm/drm_gpusvm.c
+++ b/drivers/gpu/drm/drm_gpusvm.c
@@ -1142,7 +1142,7 @@ drm_gpusvm_range_find_or_insert(struct drm_gpusvm *gpusvm,
 	 * have to change.
 	 */
 	migrate_devmem = ctx->devmem_possible &&
-		vma_is_anonymous(vas) && !is_vm_hugetlb_page(vas);
+		vma_is_anonymous(vas) && !vma_is_hugetlb(vas);
 
 	chunk_size = drm_gpusvm_range_chunk_size(gpusvm, notifier, vas,
 						 fault_addr, gpuva_start,
diff --git a/fs/coredump.c b/fs/coredump.c
index ac3cd74808c6..fb21fb6703dd 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -1608,7 +1608,7 @@ static unsigned long vma_dump_size(struct vm_area_struct *vma,
 	}
 
 	/* Hugetlb memory check */
-	if (is_vm_hugetlb_page(vma)) {
+	if (vma_is_hugetlb(vma)) {
 		if ((vma->vm_flags & VM_SHARED) && FILTER(HUGETLB_SHARED))
 			goto whole;
 		if (!(vma->vm_flags & VM_SHARED) && FILTER(HUGETLB_PRIVATE))
diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
index 7611a8470ea2..ba7097d5720c 100644
--- a/fs/hugetlbfs/inode.c
+++ b/fs/hugetlbfs/inode.c
@@ -108,7 +108,7 @@ static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma)
 	 * vma address alignment (but not the pgoff alignment) has
 	 * already been checked by prepare_hugepage_range.  If you add
 	 * any error returns here, do so after setting VM_HUGETLB, so
-	 * is_vm_hugetlb_page tests below unmap_region go the right
+	 * vma_is_hugetlb tests below unmap_region go the right
 	 * way when do_mmap unwinds (may be important on powerpc
 	 * and ia64).
 	 */
diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index e671b4fd8ded..565e6446bd31 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -3015,7 +3015,7 @@ static int pagemap_scan_pte_hole(unsigned long addr, unsigned long end,
 	 * hugetlb differs, see pagemap_hugetlb_category().
 	 */
 	categories = p->cur_vma_category;
-	if (userfaultfd_wp(vma) && !is_vm_hugetlb_page(vma))
+	if (userfaultfd_wp(vma) && !vma_is_hugetlb(vma))
 		categories |= PAGE_IS_WRITTEN;
 
 	if (!pagemap_scan_is_interesting_page(categories, p))
@@ -3028,7 +3028,7 @@ static int pagemap_scan_pte_hole(unsigned long addr, unsigned long end,
 	if (~p->arg.flags & PM_SCAN_WP_MATCHING)
 		return ret;
 
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		err = pagemap_scan_hugetlb_hole_wp(vma, addr, end);
 	else
 		err = uffd_wp_range(vma, addr, end - addr, true);
@@ -3470,7 +3470,7 @@ static int show_numa_map(struct seq_file *m, void *v)
 		seq_puts(m, " stack");
 	}
 
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		seq_puts(m, " huge");
 
 	/* Skip walking pages if gate VMA */
@@ -3499,7 +3499,7 @@ static int show_numa_map(struct seq_file *m, void *v)
 	if (md->swapcache)
 		seq_printf(m, " swapcache=%lu", md->swapcache);
 
-	if (md->active < md->pages && !is_vm_hugetlb_page(vma))
+	if (md->active < md->pages && !vma_is_hugetlb(vma))
 		seq_printf(m, " active=%lu", md->active);
 
 	if (md->writeback)
diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h
index 8b8123e5b7d7..48d47b34cc77 100644
--- a/include/asm-generic/tlb.h
+++ b/include/asm-generic/tlb.h
@@ -438,7 +438,7 @@ tlb_update_vma_flags(struct mmu_gather *tlb, struct vm_area_struct *vma)
 	 * We rely on tlb_end_vma() to issue a flush, such that when we reset
 	 * these values the batch is empty.
 	 */
-	tlb->vma_huge = is_vm_hugetlb_page(vma);
+	tlb->vma_huge = vma_is_hugetlb(vma);
 	tlb->vma_exec = !!(vma->vm_flags & VM_EXEC);
 
 	/*
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index d7e6563cef75..24727ece20fe 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -251,14 +251,14 @@ extern void __hugetlb_zap_end(struct vm_area_struct *vma,
 static inline void hugetlb_zap_begin(struct vm_area_struct *vma,
 				     unsigned long *start, unsigned long *end)
 {
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		__hugetlb_zap_begin(vma, start, end);
 }
 
 static inline void hugetlb_zap_end(struct vm_area_struct *vma,
 				   struct zap_details *details)
 {
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		__hugetlb_zap_end(vma, details);
 }
 
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 2c41974704a9..70fcb7e18b50 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1609,15 +1609,28 @@ static inline bool vma_is_shared_maywrite(const struct vm_area_struct *vma)
 	return is_shared_maywrite(&vma->flags);
 }
 
-static inline bool is_vma_hugetlb_flags(const vma_flags_t *flags)
+/**
+ * vma_flags_is_hugetlb() - Do the specified VMA flags indicate that the
+ * VMA is a hugetlb mapping?
+ * @flags: The VMA flags to test.
+ *
+ * Returns: true if the flags indicate a hugetlb mapping, false otherwise.
+ */
+static inline bool vma_flags_is_hugetlb(const vma_flags_t *flags)
 {
 	return IS_ENABLED(CONFIG_HUGETLB_PAGE) &&
 	       vma_flags_test(flags, VMA_HUGETLB_BIT);
 }
 
-static inline bool is_vm_hugetlb_page(const struct vm_area_struct *vma)
+/**
+ * vma_is_hugetlb() - Is @vma a hugetlb mapping?
+ * @vma: The VMA to test.
+ *
+ * Returns: true if @vma is a hugetlb mapping, false otherwise.
+ */
+static inline bool vma_is_hugetlb(const struct vm_area_struct *vma)
 {
-	return is_vma_hugetlb_flags(&vma->flags);
+	return vma_flags_is_hugetlb(&vma->flags);
 }
 
 /**
diff --git a/include/linux/rmap.h b/include/linux/rmap.h
index 0b332770abee..74cca0e3c726 100644
--- a/include/linux/rmap.h
+++ b/include/linux/rmap.h
@@ -888,7 +888,7 @@ struct page_vma_mapped_walk {
 static inline void page_vma_mapped_walk_done(struct page_vma_mapped_walk *pvmw)
 {
 	/* HugeTLB pte is set to the relevant page table entry without pte_mapped. */
-	if (pvmw->pte && !is_vm_hugetlb_page(pvmw->vma))
+	if (pvmw->pte && !vma_is_hugetlb(pvmw->vma))
 		pte_unmap(pvmw->pte);
 	if (pvmw->ptl)
 		spin_unlock(pvmw->ptl);
diff --git a/kernel/events/core.c b/kernel/events/core.c
index a6c8e38a3110..8ca8a6842924 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -9808,7 +9808,7 @@ static void perf_event_mmap_event(struct perf_mmap_event *mmap_event)
 
 	if (vma->vm_flags & VM_LOCKED)
 		flags |= MAP_LOCKED;
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		flags |= MAP_HUGETLB;
 
 	if (file) {
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 9eb12d0b6f1b..72c057d78ebd 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4211,7 +4211,7 @@ static void task_numa_work(struct callback_head *work)
 
 	for (; vma; vma = vma_next(&vmi)) {
 		if (!vma_migratable(vma) || !vma_policy_mof(vma) ||
-			is_vm_hugetlb_page(vma) || !vma_is_mm_managed(vma)) {
+			vma_is_hugetlb(vma) || !vma_is_mm_managed(vma)) {
 			trace_sched_skip_vma_numa(mm, vma, NUMAB_SKIP_UNSUITABLE);
 			continue;
 		}
diff --git a/mm/gup.c b/mm/gup.c
index e6310a7cc05b..f166acf794e3 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -621,7 +621,7 @@ static struct page *no_page_table(struct vm_area_struct *vma,
 	 * But we can only make this optimization where a hole would surely
 	 * be zero-filled if handle_mm_fault() actually did handle it.
 	 */
-	if (is_vm_hugetlb_page(vma)) {
+	if (vma_is_hugetlb(vma)) {
 		struct hstate *h = hstate_vma(vma);
 
 		if (!hugetlbfs_pagecache_present(h, vma, address))
@@ -1213,7 +1213,7 @@ static int check_vma_flags(struct vm_area_struct *vma, unsigned long gup_flags)
 	if ((gup_flags & FOLL_LONGTERM) && vma_is_fsdax(vma))
 		return -EOPNOTSUPP;
 
-	if ((gup_flags & FOLL_SPLIT_PMD) && is_vm_hugetlb_page(vma))
+	if ((gup_flags & FOLL_SPLIT_PMD) && vma_is_hugetlb(vma))
 		return -EOPNOTSUPP;
 
 	if (vma_is_secretmem(vma))
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 87e8bf8029ab..83dd63b04568 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -4769,7 +4769,7 @@ static inline bool vma_not_suitable_for_thp_split(struct vm_area_struct *vma)
 		return true;
 	if (vma_test(vma, VMA_IO_BIT))
 		return true;
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		return true;
 
 	return false;
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index d3a0650ff690..817f57f13b09 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -1147,7 +1147,7 @@ static inline struct resv_map *inode_resv_map(struct inode *inode)
 
 static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
 {
-	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
+	VM_WARN_ON_ONCE_VMA(!vma_is_hugetlb(vma), vma);
 	if (vma->vm_flags & VM_MAYSHARE) {
 		struct address_space *mapping = vma->vm_file->f_mapping;
 		struct inode *inode = mapping->host;
@@ -1162,7 +1162,7 @@ static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
 
 static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
 {
-	VM_WARN_ON_ONCE_VMA(!is_vm_hugetlb_page(vma), vma);
+	VM_WARN_ON_ONCE_VMA(!vma_is_hugetlb(vma), vma);
 	VM_WARN_ON_ONCE_VMA(vma_test(vma, VMA_MAYSHARE_BIT), vma);
 
 	set_vma_private_data(vma, (unsigned long)map);
@@ -1170,7 +1170,7 @@ static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
 
 static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
 {
-	VM_WARN_ON_ONCE_VMA(!is_vm_hugetlb_page(vma), vma);
+	VM_WARN_ON_ONCE_VMA(!vma_is_hugetlb(vma), vma);
 	VM_WARN_ON_ONCE_VMA(vma_test(vma, VMA_MAYSHARE_BIT), vma);
 
 	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
@@ -1178,7 +1178,7 @@ static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
 
 static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
 {
-	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
+	VM_WARN_ON_ONCE_VMA(!vma_is_hugetlb(vma), vma);
 
 	return (get_vma_private_data(vma) & flag) != 0;
 }
@@ -1192,7 +1192,7 @@ bool __vma_private_lock(struct vm_area_struct *vma)
 
 void hugetlb_dup_vma_private(struct vm_area_struct *vma)
 {
-	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
+	VM_WARN_ON_ONCE_VMA(!vma_is_hugetlb(vma), vma);
 	/*
 	 * Clear vm_private_data
 	 * - For shared mappings this is a per-vma semaphore that may be
@@ -5279,7 +5279,7 @@ void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
 	unsigned long last_addr_mask;
 
 	i_mmap_assert_write_locked(vma->vm_file->f_mapping);
-	WARN_ON(!is_vm_hugetlb_page(vma));
+	WARN_ON(!vma_is_hugetlb(vma));
 	BUG_ON(start & ~huge_page_mask(h));
 	BUG_ON(end & ~huge_page_mask(h));
 
@@ -7505,6 +7505,6 @@ void hugetlb_unshare_all_pmds(struct vm_area_struct *vma)
  */
 void fixup_hugetlb_reservations(struct vm_area_struct *vma)
 {
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		clear_vma_resv_huge_pages(vma);
 }
diff --git a/mm/internal.h b/mm/internal.h
index ea8fa94ebe1b..b5f7be8f65d8 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -1116,7 +1116,7 @@ static inline bool vma_supports_mlock(const struct vm_area_struct *vma)
 		return false;
 	if (vma_test_single_mask(vma, VMA_DROPPABLE))
 		return false;
-	if (vma_is_dax(vma) || is_vm_hugetlb_page(vma))
+	if (vma_is_dax(vma) || vma_is_hugetlb(vma))
 		return false;
 	return vma != get_gate_vma(current->mm);
 }
diff --git a/mm/madvise.c b/mm/madvise.c
index fbb72ab49aa6..467601a8525b 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -881,7 +881,7 @@ bool madvise_dontneed_free_valid_vma(struct madvise_behavior *madv_behavior)
 	int behavior = madv_behavior->behavior;
 	struct madvise_behavior_range *range = &madv_behavior->range;
 
-	if (!is_vm_hugetlb_page(vma)) {
+	if (!vma_is_hugetlb(vma)) {
 		unsigned int forbidden = VM_PFNMAP;
 
 		if (behavior != MADV_DONTNEED_LOCKED)
@@ -1578,7 +1578,7 @@ static int madvise_vma_behavior(struct madvise_behavior *madv_behavior)
 		new_flags |= VM_DONTDUMP;
 		break;
 	case MADV_DODUMP:
-		if ((!is_vm_hugetlb_page(vma) && (new_flags & VM_SPECIAL)) ||
+		if ((!vma_is_hugetlb(vma) && (new_flags & VM_SPECIAL)) ||
 		    (new_flags & VM_DROPPABLE))
 			return -EINVAL;
 		new_flags &= ~VM_DONTDUMP;
diff --git a/mm/memory.c b/mm/memory.c
index 142a8ef13f6a..3290776bf2ca 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1564,7 +1564,7 @@ copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma)
 	if (!vma_needs_copy(dst_vma, src_vma))
 		return 0;
 
-	if (is_vm_hugetlb_page(src_vma))
+	if (vma_is_hugetlb(src_vma))
 		return copy_hugetlb_page_range(dst_mm, src_mm, dst_vma, src_vma);
 
 	/*
@@ -2178,7 +2178,7 @@ static void __zap_vma_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
 	if (vma->vm_file && !reaping)
 		uprobe_munmap(vma, start, end);
 
-	if (unlikely(is_vm_hugetlb_page(vma))) {
+	if (unlikely(vma_is_hugetlb(vma))) {
 		zap_flags_t zap_flags = details ? details->zap_flags : 0;
 
 		VM_WARN_ON_ONCE(reaping);
@@ -2313,7 +2313,7 @@ void zap_vma_range_batched(struct mmu_gather *tlb,
 	 */
 	__zap_vma_range(tlb, vma, address, end, details);
 	mmu_notifier_invalidate_range_end(&range);
-	if (is_vm_hugetlb_page(vma)) {
+	if (vma_is_hugetlb(vma)) {
 		/*
 		 * flush tlb and free resources before hugetlb_zap_end(), to
 		 * avoid concurrent page faults' allocation failure.
@@ -6934,7 +6934,7 @@ vm_fault_t handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
 
 	lru_gen_enter_fault(vma);
 
-	if (unlikely(is_vm_hugetlb_page(vma)))
+	if (unlikely(vma_is_hugetlb(vma)))
 		ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
 	else
 		ret = __handle_mm_fault(vma, address, flags);
@@ -7804,12 +7804,12 @@ void ptlock_free(struct ptdesc *ptdesc)
 
 void vma_pgtable_walk_begin(struct vm_area_struct *vma)
 {
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		hugetlb_vma_lock_read(vma);
 }
 
 void vma_pgtable_walk_end(struct vm_area_struct *vma)
 {
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		hugetlb_vma_unlock_read(vma);
 }
diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 8fc8a975657e..044ffb4f4128 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -2023,7 +2023,7 @@ bool vma_migratable(struct vm_area_struct *vma)
 	if (vma_is_dax(vma))
 		return false;
 
-	if (is_vm_hugetlb_page(vma) &&
+	if (vma_is_hugetlb(vma) &&
 		!hugepage_migration_supported(hstate_vma(vma)))
 		return false;
 
diff --git a/mm/migrate_device.c b/mm/migrate_device.c
index 0c437004329d..c38cbaaef5a4 100644
--- a/mm/migrate_device.c
+++ b/mm/migrate_device.c
@@ -743,7 +743,7 @@ int migrate_vma_setup(struct migrate_vma *args)
 
 	args->start &= PAGE_MASK;
 	args->end &= PAGE_MASK;
-	if (!args->vma || is_vm_hugetlb_page(args->vma) ||
+	if (!args->vma || vma_is_hugetlb(args->vma) ||
 	    (args->vma->vm_flags & VM_SPECIAL) || vma_is_dax(args->vma))
 		return -EINVAL;
 	if (nr_pages <= 0)
diff --git a/mm/mmap.c b/mm/mmap.c
index 4bf26b0f1e6e..98449f364af1 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -1786,7 +1786,7 @@ __latent_entropy int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
 		/*
 		 * Copy/update hugetlb private vma information.
 		 */
-		if (is_vm_hugetlb_page(tmp))
+		if (vma_is_hugetlb(tmp))
 			hugetlb_dup_vma_private(tmp);
 
 		/*
diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c
index 2a72a9686773..9f353f0e2ef4 100644
--- a/mm/mmu_gather.c
+++ b/mm/mmu_gather.c
@@ -480,7 +480,7 @@ void tlb_gather_mmu_vma(struct mmu_gather *tlb, struct vm_area_struct *vma)
 {
 	tlb_gather_mmu(tlb, vma->vm_mm);
 	tlb_update_vma_flags(tlb, vma);
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		/* All entries have the same size. */
 		tlb_change_page_size(tlb, huge_page_size(hstate_vma(vma)));
 }
diff --git a/mm/mprotect.c b/mm/mprotect.c
index 706c679d4281..bef7b742d101 100644
--- a/mm/mprotect.c
+++ b/mm/mprotect.c
@@ -717,7 +717,7 @@ long change_protection(struct mmu_gather *tlb,
 	    (cp_flags & MM_CP_UFFD_RWP))
 		newprot = PAGE_NONE;
 
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		pages = hugetlb_change_protection(vma, start, end, newprot,
 						  cp_flags);
 	else
diff --git a/mm/mremap.c b/mm/mremap.c
index ed19b47c2caf..1122282a1d6a 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -812,7 +812,7 @@ unsigned long move_page_tables(struct pagetable_move_control *pmc)
 	if (!pmc->len_in)
 		return 0;
 
-	if (is_vm_hugetlb_page(pmc->old))
+	if (vma_is_hugetlb(pmc->old))
 		return move_hugetlb_page_tables(pmc->old, pmc->new, pmc->old_addr,
 						pmc->new_addr, pmc->len_in);
 
@@ -1735,7 +1735,7 @@ static bool vma_multi_allowed(struct vm_area_struct *vma)
 	/* Known good. */
 	if (vma_is_shmem(vma))
 		return true;
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		return true;
 	if (file->f_op->get_unmapped_area == thp_get_unmapped_area)
 		return true;
@@ -1758,7 +1758,7 @@ static int check_prep_vma(struct vma_remap_struct *vrm)
 		return -EPERM;
 
 	/* Align to hugetlb page size, if required. */
-	if (is_vm_hugetlb_page(vma) && !align_hugetlb(vrm))
+	if (vma_is_hugetlb(vma) && !align_hugetlb(vrm))
 		return -EINVAL;
 
 	vrm_set_delta(vrm);
diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
index 28e306fdb3a5..8408aee7571b 100644
--- a/mm/page_vma_mapped.c
+++ b/mm/page_vma_mapped.c
@@ -109,7 +109,7 @@ static bool check_pte(struct page_vma_mapped_walk *pvmw, unsigned long pte_nr)
 	unsigned long pfn;
 	pte_t ptent;
 
-	if (is_vm_hugetlb_page(pvmw->vma))
+	if (vma_is_hugetlb(pvmw->vma))
 		ptent = huge_ptep_get(pvmw->vma->vm_mm, pvmw->address,
 				      pvmw->pte);
 	else
@@ -206,7 +206,7 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
 	if (pvmw->pmd && !pvmw->pte)
 		return not_found(pvmw);
 
-	if (unlikely(is_vm_hugetlb_page(vma))) {
+	if (unlikely(vma_is_hugetlb(vma))) {
 		struct hstate *hstate = hstate_vma(vma);
 		unsigned long size = huge_page_size(hstate);
 		/* The only possible mapping was handled on last iteration */
diff --git a/mm/pagewalk.c b/mm/pagewalk.c
index 7411702a37f5..e6493bbe6919 100644
--- a/mm/pagewalk.c
+++ b/mm/pagewalk.c
@@ -408,7 +408,7 @@ static int __walk_page_range(unsigned long start, unsigned long end,
 	int err = 0;
 	struct vm_area_struct *vma = walk->vma;
 	const struct mm_walk_ops *ops = walk->ops;
-	bool is_hugetlb = is_vm_hugetlb_page(vma);
+	bool is_hugetlb = vma_is_hugetlb(vma);
 
 	/* We do not support hugetlb PTE installation. */
 	if (ops->install_pte && is_hugetlb)
diff --git a/mm/swapfile.c b/mm/swapfile.c
index 2cd0d0ba966c..c1c5fbb3c909 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -2707,7 +2707,7 @@ static int unuse_mm(struct mm_struct *mm, unsigned int type)
 	if (check_stable_address_space(mm))
 		goto unlock;
 	for_each_vma(vmi, vma) {
-		if (vma->anon_vma && !is_vm_hugetlb_page(vma)) {
+		if (vma->anon_vma && !vma_is_hugetlb(vma)) {
 			ret = unuse_vma(vma, type);
 			if (ret)
 				break;
diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index 79cc7b546f13..949017e60608 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -237,7 +237,7 @@ static int mfill_get_vma(struct mfill_state *state)
 	if ((flags & MFILL_ATOMIC_WP) && !(dst_vma->vm_flags & VM_UFFD_WP))
 		goto out_unlock;
 
-	if (is_vm_hugetlb_page(dst_vma))
+	if (vma_is_hugetlb(dst_vma))
 		return 0;
 
 	ops = vma_uffd_ops(dst_vma);
@@ -804,7 +804,7 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
 		}
 
 		err = -ENOENT;
-		if (!is_vm_hugetlb_page(dst_vma))
+		if (!vma_is_hugetlb(dst_vma))
 			goto out_unlock_vma;
 
 		err = -EINVAL;
@@ -967,7 +967,7 @@ static __always_inline ssize_t mfill_atomic(struct userfaultfd_ctx *ctx,
 	/*
 	 * If this is a HUGETLB vma, pass off to appropriate routine
 	 */
-	if (is_vm_hugetlb_page(state.vma))
+	if (vma_is_hugetlb(state.vma))
 		return  mfill_atomic_hugetlb(ctx, state.vma, dst_start,
 					     src_start, len, flags);
 
@@ -1114,7 +1114,7 @@ static int mwriteprotect_range(struct userfaultfd_ctx *ctx, unsigned long start,
 			break;
 		}
 
-		if (is_vm_hugetlb_page(dst_vma)) {
+		if (vma_is_hugetlb(dst_vma)) {
 			err = -EINVAL;
 			page_mask = vma_kernel_pagesize(dst_vma) - 1;
 			if ((start & page_mask) || (len & page_mask))
@@ -1172,7 +1172,7 @@ int mrwprotect_range(struct userfaultfd_ctx *ctx, unsigned long start,
 		if (!userfaultfd_rwp(dst_vma))
 			return -ENOENT;
 
-		if (is_vm_hugetlb_page(dst_vma)) {
+		if (vma_is_hugetlb(dst_vma)) {
 			unsigned long page_mask;
 
 			page_mask = vma_kernel_pagesize(dst_vma) - 1;
@@ -2149,7 +2149,7 @@ static bool vma_can_userfault(struct vm_area_struct *vma, vm_flags_t vm_flags,
 	if (vma->vm_flags & (VM_DROPPABLE | VM_SHADOW_STACK))
 		return false;
 
-	if (!is_vm_hugetlb_page(vma) && (vma->vm_flags & VM_SPECIAL))
+	if (!vma_is_hugetlb(vma) && (vma->vm_flags & VM_SPECIAL))
 		return false;
 
 	vm_flags &= __VM_UFFD_FLAGS;
@@ -2319,7 +2319,7 @@ static int userfaultfd_register_range(struct userfaultfd_ctx *ctx,
 		 */
 		userfaultfd_set_ctx(vma, ctx, vm_flags);
 
-		if (is_vm_hugetlb_page(vma) && uffd_disable_huge_pmd_share(vma))
+		if (vma_is_hugetlb(vma) && uffd_disable_huge_pmd_share(vma))
 			hugetlb_unshare_all_pmds(vma);
 
 skip:
@@ -2895,7 +2895,7 @@ vm_fault_t handle_userfault(struct vm_fault *vmf, unsigned long reason)
 	 * (sleepable) vma lock can modify the current task state, that
 	 * must be before explicitly calling set_current_state().
 	 */
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		hugetlb_vma_lock_read(vma);
 
 	spin_lock_irq(&ctx->fault_pending_wqh.lock);
@@ -2912,7 +2912,7 @@ vm_fault_t handle_userfault(struct vm_fault *vmf, unsigned long reason)
 	set_current_state(blocking_state);
 	spin_unlock_irq(&ctx->fault_pending_wqh.lock);
 
-	if (is_vm_hugetlb_page(vma)) {
+	if (vma_is_hugetlb(vma)) {
 		must_wait = userfaultfd_huge_must_wait(ctx, vmf, reason);
 		hugetlb_vma_unlock_read(vma);
 	} else {
@@ -3744,7 +3744,7 @@ static int userfaultfd_register(struct userfaultfd_ctx *ctx,
 	 * If the first vma contains huge pages, make sure start address
 	 * is aligned to huge page size.
 	 */
-	if (is_vm_hugetlb_page(vma)) {
+	if (vma_is_hugetlb(vma)) {
 		unsigned long vma_hpagesize = vma_kernel_pagesize(vma);
 
 		if (start & (vma_hpagesize - 1))
@@ -3795,7 +3795,7 @@ static int userfaultfd_register(struct userfaultfd_ctx *ctx,
 		 * If this vma contains ending address, and huge pages
 		 * check alignment.
 		 */
-		if (is_vm_hugetlb_page(cur) && end <= cur->vm_end &&
+		if (vma_is_hugetlb(cur) && end <= cur->vm_end &&
 		    end > cur->vm_start) {
 			unsigned long vma_hpagesize = vma_kernel_pagesize(cur);
 
@@ -3831,7 +3831,7 @@ static int userfaultfd_register(struct userfaultfd_ctx *ctx,
 		/*
 		 * Note vmas containing huge pages
 		 */
-		if (is_vm_hugetlb_page(cur))
+		if (vma_is_hugetlb(cur))
 			basic_ioctls = true;
 
 		found = true;
@@ -3917,7 +3917,7 @@ static int userfaultfd_unregister(struct userfaultfd_ctx *ctx,
 	 * If the first vma contains huge pages, make sure start address
 	 * is aligned to huge page size.
 	 */
-	if (is_vm_hugetlb_page(vma)) {
+	if (vma_is_hugetlb(vma)) {
 		unsigned long vma_hpagesize = vma_kernel_pagesize(vma);
 
 		if (start & (vma_hpagesize - 1))
diff --git a/mm/vma.c b/mm/vma.c
index 5df02665d1ed..b2bb444c43d7 100644
--- a/mm/vma.c
+++ b/mm/vma.c
@@ -599,7 +599,7 @@ __split_vma(struct vma_iterator *vmi, struct vm_area_struct *vma,
 	 * boundary.
 	 */
 	vma_adjust_trans_huge(vma, vma->vm_start, addr, NULL);
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		hugetlb_split(vma, addr);
 
 	if (new_below) {
@@ -2236,7 +2236,7 @@ bool vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot)
 	 * Do we need to track softdirty? hugetlb does not support softdirty
 	 * tracking yet.
 	 */
-	if (vma_soft_dirty_enabled(vma) && !is_vm_hugetlb_page(vma))
+	if (vma_soft_dirty_enabled(vma) && !vma_is_hugetlb(vma))
 		return true;
 
 	/* Do we need write faults for uffd-wp tracking? */
@@ -2355,7 +2355,7 @@ int mm_take_all_locks(struct mm_struct *mm)
 		if (signal_pending(current))
 			goto out_unlock;
 		if (vma->vm_file && vma->vm_file->f_mapping &&
-				is_vm_hugetlb_page(vma))
+				vma_is_hugetlb(vma))
 			vm_lock_mapping(mm, vma->vm_file->f_mapping);
 	}
 
@@ -2364,7 +2364,7 @@ int mm_take_all_locks(struct mm_struct *mm)
 		if (signal_pending(current))
 			goto out_unlock;
 		if (vma->vm_file && vma->vm_file->f_mapping &&
-				!is_vm_hugetlb_page(vma))
+				!vma_is_hugetlb(vma))
 			vm_lock_mapping(mm, vma->vm_file->f_mapping);
 	}
 
diff --git a/mm/vmscan.c b/mm/vmscan.c
index aaceed4759ee..12ed3388ee41 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -3471,7 +3471,7 @@ static int should_skip_vma(unsigned long start, unsigned long end, struct mm_wal
 	if (!vma_is_accessible(vma))
 		return true;
 
-	if (is_vm_hugetlb_page(vma))
+	if (vma_is_hugetlb(vma))
 		return true;
 
 	if (!vma_has_recency(vma))
diff --git a/tools/testing/vma/include/stubs.h b/tools/testing/vma/include/stubs.h
index b35279a0ec78..83ff23226ddf 100644
--- a/tools/testing/vma/include/stubs.h
+++ b/tools/testing/vma/include/stubs.h
@@ -193,7 +193,7 @@ static inline bool mapping_can_writeback(struct address_space *mapping)
 	return true;
 }
 
-static inline bool is_vm_hugetlb_page(struct vm_area_struct *vma)
+static inline bool vma_is_hugetlb(struct vm_area_struct *vma)
 {
 	return false;
 }

-- 
2.55.0

[PATCH v6 29/39] mm: drop some redundant checks around hugetlb VMAs

[Lorenzo Stoakes (ARM)]

Adjust code which inadvertently perform redundant checks on hugetlb VMAs
and clean them up:

* hugetlb VMAs have VMA_DONTEXPAND_BIT set so a VMA_SPECIAL_FLAGS check
  suffices. (migrate_vma_setup() regains an explicit hugetlb test later in
  the series, once VMA_SPECIAL_FLAGS is removed.)

* hugetlb VMAs unconditionally set vma->vm_ops, so they are never
  anonymous.

* hugetlb VMAs do not set VMA_PFNMAP_BIT so checking for this is redundant.

While we're here also drop a VM_BUG_ON() which the simplified check above
makes unreachable, and use the new VMA flag API.

No functional change intended.

Acked-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 arch/arm64/kvm/mmu.c         | 4 +---
 drivers/gpu/drm/drm_gpusvm.c | 3 +--
 mm/migrate_device.c          | 4 ++--
 3 files changed, 4 insertions(+), 7 deletions(-)

diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index a7968f8d24bf..3c1240ffc38d 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -1463,14 +1463,12 @@ static int get_vma_page_shift(struct vm_area_struct *vma, unsigned long hva)
 {
 	unsigned long pa;
 
-	if (vma_is_hugetlb(vma) && !(vma->vm_flags & VM_PFNMAP))
+	if (vma_is_hugetlb(vma))
 		return huge_page_shift(hstate_vma(vma));
 
 	if (!(vma->vm_flags & VM_PFNMAP))
 		return PAGE_SHIFT;
 
-	VM_BUG_ON(vma_is_hugetlb(vma));
-
 	pa = (vma->vm_pgoff << PAGE_SHIFT) + (hva - vma->vm_start);
 
 #ifndef __PAGETABLE_PMD_FOLDED
diff --git a/drivers/gpu/drm/drm_gpusvm.c b/drivers/gpu/drm/drm_gpusvm.c
index a1d4989b0b61..fab34fea99c2 100644
--- a/drivers/gpu/drm/drm_gpusvm.c
+++ b/drivers/gpu/drm/drm_gpusvm.c
@@ -1141,8 +1141,7 @@ drm_gpusvm_range_find_or_insert(struct drm_gpusvm *gpusvm,
 	 * limitations. If/when migrate_vma_* add more support, this logic will
 	 * have to change.
 	 */
-	migrate_devmem = ctx->devmem_possible &&
-		vma_is_anonymous(vas) && !vma_is_hugetlb(vas);
+	migrate_devmem = ctx->devmem_possible && vma_is_anonymous(vas);
 
 	chunk_size = drm_gpusvm_range_chunk_size(gpusvm, notifier, vas,
 						 fault_addr, gpuva_start,
diff --git a/mm/migrate_device.c b/mm/migrate_device.c
index c38cbaaef5a4..b9c453c28795 100644
--- a/mm/migrate_device.c
+++ b/mm/migrate_device.c
@@ -743,8 +743,8 @@ int migrate_vma_setup(struct migrate_vma *args)
 
 	args->start &= PAGE_MASK;
 	args->end &= PAGE_MASK;
-	if (!args->vma || vma_is_hugetlb(args->vma) ||
-	    (args->vma->vm_flags & VM_SPECIAL) || vma_is_dax(args->vma))
+	if (!args->vma || vma_test_any_mask(args->vma, VMA_SPECIAL_FLAGS) ||
+	    vma_is_dax(args->vma))
 		return -EINVAL;
 	if (nr_pages <= 0)
 		return -EINVAL;

-- 
2.55.0

[PATCH v6 30/39] mm/vma: introduce vma[_flags]_is_mm_backed()

[Lorenzo Stoakes (ARM)]

Introduce vma[_flags]_is_mm_backed() for the purposes of identifying
mappings which are mm-managed and whose contents were placed there by core
mm's own fault path - rather than a custom ->fault handler.

Mappings not managed by core mm do not fall into this category, as they are
populated by their owner, and nor of course does memory-mapped I/O.

We exclude fixed mappings as these have a layout core mm cannot assume
ordinary fault semantics over.

hugetlb mappings are fixed mappings, but core mm owns them entirely, so
they are explicitly carved out as mm-backed, as the MADV_DODUMP check
already does.

Use this functionality to update the madvise MADV_DODUMP check to test for
this rather than open-coding it.

Also update vma_flags_can_merge() so it is written in terms of
vma_flags_is_mm_backed() - all mm-backed mappings can be merged except
hugetlb mappings, whose flags and fields do not completely describe them.

Also update __thp_vma_allowable_orders() to use vma_is_mm_backed() rather
than testing for mergeability, which is irrelevant to this function.

This replaces the VM_SPECIAL check which means it no longer checks for
VMA_IO_BIT, however this is safe as we have established the invariant that
mm-managed mappings may not set VMA_IO_BIT, so we implicitly include
these.

No functional change intended.

Acked-by: Zi Yan <ziy@nvidia.com>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 include/linux/mm.h              | 66 +++++++++++++++++++++++++++++++++++------
 mm/huge_memory.c                |  7 ++---
 mm/madvise.c                    |  7 +++--
 tools/testing/vma/include/dup.h | 44 +++++++++++++++++++++------
 4 files changed, 100 insertions(+), 24 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 70fcb7e18b50..e8d2b0e847b3 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1701,6 +1701,54 @@ static inline bool vma_is_fixed_mapping(const struct vm_area_struct *vma)
 	return vma_flags_is_fixed_mapping(&vma->flags);
 }
 
+/**
+ * vma_flags_is_mm_backed() - Do the specified VMA flags imply the mapping is
+ * backed by core mm?
+ * @flags: The VMA flags to test.
+ *
+ * An mm-backed mapping is one managed by core mm (see
+ * vma_flags_is_mm_managed()) whose contents are established by core mm's own
+ * fault path rather than a custom ->fault handler.
+ *
+ * Mappings not managed by core mm are populated by their owner, and fixed
+ * mappings may be mm-managed but have their contents established by a custom
+ * ->fault handler, so neither are mm-backed. hugetlb mappings are fixed, but
+ * core mm owns them entirely, so they are.
+ *
+ * These are the only mappings which can sensibly be merged, locked, dumped or
+ * be subject to uffd faulting, as all of these assume core mm and core mm
+ * alone manages their contents.
+ *
+ * Returns: true if the flags imply an mm-backed mapping, otherwise false.
+ */
+static inline bool vma_flags_is_mm_backed(const vma_flags_t *flags)
+{
+	/* hugetlb is a fixed mapping, but core mm owns it entirely. */
+	if (vma_flags_is_hugetlb(flags))
+		return true;
+	/*
+	 * Mappings not managed by core mm are populated by their owner.
+	 *
+	 * Fixed mappings may be mm-managed, but their contents are established
+	 * by a custom ->fault handler, so core mm cannot assume ordinary fault
+	 * semantics over them.
+	 */
+	return vma_flags_is_mm_managed(flags) && !vma_flags_is_fixed_mapping(flags);
+}
+
+/**
+ * vma_is_mm_backed() - Is @vma backed by core mm?
+ * @vma: The VMA to test.
+ *
+ * See vma_flags_is_mm_backed() for details.
+ *
+ * Returns: true if the VMA is mm-backed, otherwise false.
+ */
+static inline bool vma_is_mm_backed(const struct vm_area_struct *vma)
+{
+	return vma_flags_is_mm_backed(&vma->flags);
+}
+
 /**
  * vma_flags_can_merge() - Do the specified VMA flags permit the VMA to be
  * merged with another?
@@ -1711,21 +1759,21 @@ static inline bool vma_flags_can_merge(const vma_flags_t *flags)
 {
 	/*
 	 * VMA merging assumes that a VMA's flags and fields completely describe
-	 * its state.
+	 * its state, which is only the case for mm-backed mappings.
 	 *
-	 * However, mappings which are not mm-managed may have established state
-	 * upon mapping not embodied in any attribute of the VMA.
+	 * Mappings which are not mm-managed may have established state upon
+	 * mapping not embodied in any attribute of the VMA.
 	 *
 	 * Additionally, private (CoW) PFN maps encode the source PFN of the
 	 * range in vma->vm_pgoff, which may otherwise cause spurious merges.
 	 */
-	if (!vma_flags_is_mm_managed(flags))
-		return false;
-	/* VMA explicitly marked as being unmergeable. */
-	if (vma_flags_is_fixed_mapping(flags))
+	if (!vma_flags_is_mm_backed(flags))
 		return false;
-
-	return true;
+	/*
+	 * hugetlb mappings contain per-VMA state, so their flags and fields do
+	 * not completely describe them.
+	 */
+	return !vma_flags_is_hugetlb(flags);
 }
 
 /**
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 83dd63b04568..5b295a7979cd 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -215,11 +215,10 @@ unsigned long __thp_vma_allowable_orders(struct vm_area_struct *vma,
 	 * khugepaged moves data from VMAs once collapsed, after they have been
 	 * faulted in, relying on refaulting for file-backed memory.
 	 *
-	 * Mappings not managed by core mm cannot be reliably reconstructed from
-	 * page faults, and fixed mappings (including hugetlb) may well be
-	 * mm-managed - precisely the mappings which cannot be merged.
+	 * Only mm-backed mappings can be reliably reconstructed from page
+	 * faults, and hugetlb mappings cannot be collapsed.
 	 */
-	if (!in_pf && !smaps && !vma_can_merge(vma))
+	if (!in_pf && !smaps && (!vma_is_mm_backed(vma) || vma_is_hugetlb(vma)))
 		return 0;
 
 	/*
diff --git a/mm/madvise.c b/mm/madvise.c
index 467601a8525b..3d25202db5cc 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -1578,8 +1578,11 @@ static int madvise_vma_behavior(struct madvise_behavior *madv_behavior)
 		new_flags |= VM_DONTDUMP;
 		break;
 	case MADV_DODUMP:
-		if ((!vma_is_hugetlb(vma) && (new_flags & VM_SPECIAL)) ||
-		    (new_flags & VM_DROPPABLE))
+		/* Only mm-backed memory can be meaningfully dumped. */
+		if (!vma_is_mm_backed(vma))
+			return -EINVAL;
+		/* Core mm may discard droppable memory at any time. */
+		if (new_flags & VM_DROPPABLE)
 			return -EINVAL;
 		new_flags &= ~VM_DONTDUMP;
 		break;
diff --git a/tools/testing/vma/include/dup.h b/tools/testing/vma/include/dup.h
index 94bf818ad0e1..19ce10faa9f3 100644
--- a/tools/testing/vma/include/dup.h
+++ b/tools/testing/vma/include/dup.h
@@ -1689,23 +1689,49 @@ static inline bool vma_is_fixed_mapping(const struct vm_area_struct *vma)
 	return vma_flags_is_fixed_mapping(&vma->flags);
 }
 
+static inline bool vma_flags_is_hugetlb(const vma_flags_t *flags)
+{
+	return IS_ENABLED(CONFIG_HUGETLB_PAGE) &&
+	       vma_flags_test(flags, VMA_HUGETLB_BIT);
+}
+
+static inline bool vma_flags_is_mm_backed(const vma_flags_t *flags)
+{
+	/* hugetlb is a fixed mapping, but core mm owns it entirely. */
+	if (vma_flags_is_hugetlb(flags))
+		return true;
+	/*
+	 * Mappings not managed by core mm are populated by their owner.
+	 *
+	 * Fixed mappings may be mm-managed, but their contents are established
+	 * by a custom ->fault handler, so core mm cannot assume ordinary fault
+	 * semantics over them.
+	 */
+	return vma_flags_is_mm_managed(flags) && !vma_flags_is_fixed_mapping(flags);
+}
+
+static inline bool vma_is_mm_backed(const struct vm_area_struct *vma)
+{
+	return vma_flags_is_mm_backed(&vma->flags);
+}
+
 static inline bool vma_flags_can_merge(const vma_flags_t *flags)
 {
 	/*
 	 * VMA merging assumes that a VMA's flags and fields completely describe
-	 * its state.
+	 * its state, which is only the case for mm-backed mappings.
 	 *
-	 * However, mappings which are not mm-managed may have established state
-	 * upon mapping not embodied in any attribute of the VMA.
+	 * Mappings which are not mm-managed may have established state upon
+	 * mapping not embodied in any attribute of the VMA.
 	 *
 	 * Additionally, private (CoW) PFN maps encode the source PFN of the
 	 * range in vma->vm_pgoff, which may otherwise cause spurious merges.
 	 */
-	if (!vma_flags_is_mm_managed(flags))
-		return false;
-	/* VMA explicitly marked as being unmergeable. */
-	if (vma_flags_is_fixed_mapping(flags))
+	if (!vma_flags_is_mm_backed(flags))
 		return false;
-
-	return true;
+	/*
+	 * hugetlb mappings contain per-VMA state, so their flags and fields do
+	 * not completely describe them.
+	 */
+	return !vma_flags_is_hugetlb(flags);
 }

-- 
2.55.0

Re: [PATCH v6 30/39] mm/vma: introduce vma[_flags]_is_mm_backed()

[David Hildenbrand (Arm)]

On 10/13/26 17:05, Lorenzo Stoakes (ARM) wrote:
> Introduce vma[_flags]_is_mm_backed() for the purposes of identifying
> mappings which are mm-managed and whose contents were placed there by core
> mm's own fault path - rather than a custom ->fault handler.
> 
> Mappings not managed by core mm do not fall into this category, as they are
> populated by their owner, and nor of course does memory-mapped I/O.
> 
> We exclude fixed mappings as these have a layout core mm cannot assume
> ordinary fault semantics over.
> 
> hugetlb mappings are fixed mappings, but core mm owns them entirely, so
> they are explicitly carved out as mm-backed, as the MADV_DODUMP check
> already does.
> 
> Use this functionality to update the madvise MADV_DODUMP check to test for
> this rather than open-coding it.
> 
> Also update vma_flags_can_merge() so it is written in terms of
> vma_flags_is_mm_backed() - all mm-backed mappings can be merged except
> hugetlb mappings, whose flags and fields do not completely describe them.
> 
> Also update __thp_vma_allowable_orders() to use vma_is_mm_backed() rather
> than testing for mergeability, which is irrelevant to this function.
> 
> This replaces the VM_SPECIAL check which means it no longer checks for
> VMA_IO_BIT, however this is safe as we have established the invariant that
> mm-managed mappings may not set VMA_IO_BIT, so we implicitly include
> these.
> 
> No functional change intended.
> 
> Acked-by: Zi Yan <ziy@nvidia.com>
> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
> ---

Acked-by: David Hildenbrand (Arm) <david@kernel.org>

-- 
Cheers,

David

[PATCH v6 31/39] mm/madvise: update is_valid_guard_vma() to use vma_is_mm_backed()

[Lorenzo Stoakes (ARM)]

We currently disallow the installation of lightweight guard regions in VMAs
whose flags intersect VMA_SPECIAL_FLAGS or VMA_HUGETLB_BIT, or
VMA_LOCKED_BIT unless allow_locked is set.

hugetlb VMAs set VMA_DONTEXPAND_BIT so this was already redundant,
VMA_SPECIAL_FLAGS already sufficed.

However, now that VMA_IO_BIT is only set if VMA_PFNMAP or VMA_MIXEDMAP_BIT
is set, this check collapses to being the equivalent of
!vma_is_mm_backed(), excluding hugetlb mappings in which guard markers
cannot be installed.

Update is_valid_guard_vma() to reflect this.

No functional change intended.

Acked-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/madvise.c | 20 +++++++++++++-------
 1 file changed, 13 insertions(+), 7 deletions(-)

diff --git a/mm/madvise.c b/mm/madvise.c
index 3d25202db5cc..7e14b3fd92f6 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -1221,19 +1221,25 @@ static long madvise_remove(struct madvise_behavior *madv_behavior)
 	return error;
 }
 
-static bool is_valid_guard_vma(struct vm_area_struct *vma, bool allow_locked)
+static bool is_valid_guard_vma(const struct vm_area_struct *vma,
+			       bool allow_locked)
 {
-	vm_flags_t disallowed = VM_SPECIAL | VM_HUGETLB;
-
 	/*
-	 * A user could lock after setting a guard range but that's fine, as
+	 * A user could lock after setting a guard range but that's fine as
 	 * they'd not be able to fault in. The issue arises when we try to zap
 	 * existing locked VMAs. We don't want to do that.
 	 */
-	if (!allow_locked)
-		disallowed |= VM_LOCKED;
+	if (!allow_locked && vma_test(vma, VMA_LOCKED_BIT))
+		return false;
+	/*
+	 * Guard regions require a VMA whose page tables are managed solely by
+	 * core mm, so the VMA must be mm-backed (excluding hugetlb, in which
+	 * guard markers cannot be installed).
+	 */
+	if (!vma_is_mm_backed(vma) || vma_is_hugetlb(vma))
+		return false;
 
-	return !(vma->vm_flags & disallowed);
+	return true;
 }
 
 static bool is_guard_pte_marker(pte_t ptent)

-- 
2.55.0

[PATCH v6 32/39] mm/uffd: use predicates for userfaultfd checks

[Lorenzo Stoakes (ARM)]

Rather than directly checking VMA flags, use the newly introduced
vma_is_mm_managed() and vma_is_mm_backed() helpers in userfaultfd
when assessing VMA suitability for userfaultfd and UFFDIO_MOVE.

The droppable check is kept explicit, as userfaultfd cannot function
correctly with droppable mappings.

Update vma_move_compatible() so it's expressed in terms of VMA
characteristics rather than arbitrary flags.

Additionally, update the use of the deprecated VMA flag API when checking
VMA_SHADOW_STACK_BIT.

A VMA_IO_BIT check is no longer required but that is fine as a hard
invariant has been established that mm-managed mappings may not set
VMA_IO_BIT so the check is now redundant.

Acked-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/userfaultfd.c | 22 ++++++++++++++++------
 1 file changed, 16 insertions(+), 6 deletions(-)

diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index 949017e60608..b3076d4a622a 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -1754,10 +1754,16 @@ static inline bool move_splits_huge_pmd(unsigned long dst_addr,
 }
 #endif
 
-static inline bool vma_move_compatible(struct vm_area_struct *vma)
+static inline bool vma_move_compatible(const struct vm_area_struct *vma)
 {
-	return !(vma->vm_flags & (VM_PFNMAP | VM_IO |  VM_HUGETLB |
-				  VM_MIXEDMAP | VM_SHADOW_STACK));
+	/* uffd is generally incompatible with mappings not managed by core mm. */
+	if (!vma_is_mm_managed(vma))
+		return false;
+	/* The shadow stack should not be written to by userspace. */
+	if (vma_test_single_mask(vma, VMA_SHADOW_STACK))
+		return false;
+	/* hugetlb mappings cannot be safely moved. */
+	return !vma_is_hugetlb(vma);
 }
 
 static int validate_move_areas(struct userfaultfd_ctx *ctx,
@@ -2146,10 +2152,14 @@ static bool vma_can_userfault(struct vm_area_struct *vma, vm_flags_t vm_flags,
 {
 	const struct vm_uffd_ops *ops = vma_uffd_ops(vma);
 
-	if (vma->vm_flags & (VM_DROPPABLE | VM_SHADOW_STACK))
+	/* Only mm-backed memory can have its faults handled by userspace. */
+	if (!vma_is_mm_backed(vma))
 		return false;
-
-	if (!vma_is_hugetlb(vma) && (vma->vm_flags & VM_SPECIAL))
+	/* Core mm may discard droppable memory at any time. */
+	if (vma_test_single_mask(vma, VMA_DROPPABLE))
+		return false;
+	/* The shadow stack should not be written to by userspace. */
+	if (vma_test_single_mask(vma, VMA_SHADOW_STACK))
 		return false;
 
 	vm_flags &= __VM_UFFD_FLAGS;

-- 
2.55.0

[PATCH v6 33/39] mm/madvise: use predicates for madvise(..., MADV_DOFORK)

[Lorenzo Stoakes (ARM)]

Make it clear what we're blocking in MADV_DOFORK. Previously we simply
disallowed VM_SPECIAL i.e. mappings not managed by core mm, fixed mappings
and VMA_IO_BIT.

Now the invariant is established that mm-managed mappings may not set
VMA_IO_BIT, the VMA_IO_BIT check is redundant.

The rest is equivalent to testing for a fixed or non-mm-managed mapping,
i.e. a mapping which is not mm-backed, or hugetlb, which is specifically
excluded.

This was established by commit 0b2758f48f22 ("Require (reasonably) normal
mappings for MADV_DOFORK") containing my hands-down favourite call out of
all time.

Express the same thing using the predicates - only mm-backed mappings may
have their copy-on-fork behaviour changed, retaining the eternal exception
to all rules, hugetlb.

No functional change intended.

Acked-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/madvise.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/madvise.c b/mm/madvise.c
index 7e14b3fd92f6..e050a21e4db1 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -1565,7 +1565,7 @@ static int madvise_vma_behavior(struct madvise_behavior *madv_behavior)
 		new_flags |= VM_DONTCOPY;
 		break;
 	case MADV_DOFORK:
-		if (new_flags & VM_SPECIAL)
+		if (!vma_is_mm_backed(vma) || vma_is_hugetlb(vma))
 			return -EINVAL;
 		new_flags &= ~VM_DONTCOPY;
 		break;

-- 
2.55.0

[PATCH v6 34/39] mm: eliminate VMA_SPECIAL_FLAGS usage when hugetlb explicitly tested

[Lorenzo Stoakes (ARM)]

It is now an invariant that VMA_IO_BIT is not set by mm-managed mappings,
so each existing VMA_SPECIAL_FLAGS test need only test for
VMA_DONTEXPAND_BIT, VMA_PFNMAP_BIT and VMA_MIXEDMAP_BIT.

This is precisely a test for a fixed or non-mm-managed mapping.

Update a number of callsites which already explicitly handle hugetlb
mappings.

vma_supports_mlock(), ksm_compatible() and should_skip_vma() explicitly
check for non-mm-managed or fixed mappings, i.e. for !vma_is_mm_backed(),
with hugetlb having already been excluded separately.

So update those to check for mm-backed (and add comments for the retained
droppable checks).

No functional change intended.

Reviewed-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/internal.h | 3 ++-
 mm/ksm.c      | 5 +++--
 mm/vmscan.c   | 2 +-
 3 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/mm/internal.h b/mm/internal.h
index b5f7be8f65d8..03f2bb4c30b7 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -1112,8 +1112,9 @@ static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
 
 static inline bool vma_supports_mlock(const struct vm_area_struct *vma)
 {
-	if (vma_test_any_mask(vma, VMA_SPECIAL_FLAGS))
+	if (!vma_is_mm_backed(vma))
 		return false;
+	/* Core mm may discard droppable memory at any time. */
 	if (vma_test_single_mask(vma, VMA_DROPPABLE))
 		return false;
 	if (vma_is_dax(vma) || vma_is_hugetlb(vma))
diff --git a/mm/ksm.c b/mm/ksm.c
index 624f37975e12..df4c5199f625 100644
--- a/mm/ksm.c
+++ b/mm/ksm.c
@@ -747,9 +747,10 @@ static bool ksm_compatible(const struct file *file, vma_flags_t vma_flags)
 	if (vma_flags_test_any(&vma_flags, VMA_SHARED_BIT, VMA_MAYSHARE_BIT,
 			       VMA_HUGETLB_BIT))
 		return false;
-	if (vma_flags_test_single_mask(&vma_flags, VMA_DROPPABLE))
+	if (!vma_flags_is_mm_backed(&vma_flags))
 		return false;
-	if (vma_flags_test_any_mask(&vma_flags, VMA_SPECIAL_FLAGS))
+	/* Core mm may discard droppable memory at any time. */
+	if (vma_flags_test_single_mask(&vma_flags, VMA_DROPPABLE))
 		return false;
 	if (file_is_dax(file))
 		return false;
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 12ed3388ee41..aecb157bc379 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -3477,7 +3477,7 @@ static int should_skip_vma(unsigned long start, unsigned long end, struct mm_wal
 	if (!vma_has_recency(vma))
 		return true;
 
-	if (vma->vm_flags & (VM_LOCKED | VM_SPECIAL))
+	if (vma_test(vma, VMA_LOCKED_BIT) || !vma_is_mm_backed(vma))
 		return true;
 
 	if (vma == get_gate_vma(vma->vm_mm))

-- 
2.55.0

[PATCH v6 35/39] mm: eliminate VMA_SPECIAL_FLAGS check in lru_gen_look_around()

[Lorenzo Stoakes (ARM)]

A fixed or non-mm-managed mapping is one which sets VMA_PFNMAP_BIT,
VMA_MIXEDMAP_BIT or VMA_DONTEXPAND_BIT, which is precisely what
VMA_SPECIAL_FLAGS tests for other than VMA_IO_BIT, which is safe to drop as
mm-managed mappings may not set it.

This is equivalent to testing for an mm-backed mapping, except for hugetlb.

However, hugetlb folios are never placed on the LRU and
lru_gen_look_around() is only called from LRU shrinkers on reclaim so this
is fine.

Using vma_is_mm_backed() rather than checking VMA_SPECIAL_FLAGS makes the
check self-documenting and helps eliminate confusion around 'special
flags'.

No functional change intended.

Reviewed-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/vmscan.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index aecb157bc379..178a5de6c439 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -4417,8 +4417,8 @@ bool lru_gen_look_around(struct page_vma_mapped_walk *pvmw, unsigned int nr)
 	if (spin_is_contended(pvmw->ptl))
 		return true;
 
-	/* exclude special VMAs containing anon pages from COW */
-	if (vma->vm_flags & VM_SPECIAL)
+	/* exclude non-mm-backed VMAs containing anon pages from COW */
+	if (!vma_is_mm_backed(vma))
 		return true;
 
 	/* avoid taking the LRU lock under the PTL when possible */

-- 
2.55.0

[PATCH v6 36/39] mm: avoid use of VMA_SPECIAL_FLAGS in migrate_vma_setup()

[Lorenzo Stoakes (ARM)]

Now we have the expressive vma_is_mm_backed() predicate, use it to
determine whether to proceed with migration. This drops the VMA_IO_BIT
test, which is safe as mm-managed mappings may not set it.

hugetlb mappings are mm-backed but remain explicitly excluded, as the
migrate_vma page walk has no hugetlb support.

No functional change intended.

Reviewed-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/migrate_device.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/mm/migrate_device.c b/mm/migrate_device.c
index b9c453c28795..9e19fadfa65c 100644
--- a/mm/migrate_device.c
+++ b/mm/migrate_device.c
@@ -739,19 +739,21 @@ static void migrate_vma_unmap(struct migrate_vma *migrate)
  */
 int migrate_vma_setup(struct migrate_vma *args)
 {
+	const struct vm_area_struct *vma = args->vma;
 	long nr_pages = (args->end - args->start) >> PAGE_SHIFT;
 
 	args->start &= PAGE_MASK;
 	args->end &= PAGE_MASK;
-	if (!args->vma || vma_test_any_mask(args->vma, VMA_SPECIAL_FLAGS) ||
-	    vma_is_dax(args->vma))
+	if (!vma)
+		return -EINVAL;
+	/* The migrate_vma page walk has no hugetlb support. */
+	if (!vma_is_mm_backed(vma) || vma_is_hugetlb(vma) || vma_is_dax(vma))
 		return -EINVAL;
 	if (nr_pages <= 0)
 		return -EINVAL;
-	if (args->start < args->vma->vm_start ||
-	    args->start >= args->vma->vm_end)
+	if (args->start < vma->vm_start || args->start >= vma->vm_end)
 		return -EINVAL;
-	if (args->end <= args->vma->vm_start || args->end > args->vma->vm_end)
+	if (args->end <= vma->vm_start || args->end > vma->vm_end)
 		return -EINVAL;
 	if (!args->src || !args->dst)
 		return -EINVAL;

-- 
2.55.0

[PATCH v6 37/39] mm: eliminate VM_SPECIAL, VMA_SPECIAL_FLAGS

[Lorenzo Stoakes (ARM)]

Every user of the VM_SPECIAL or VMA_SPECIAL_FLAGS has now been converted to
predicates which explicitly express what is actually being checked for
rather than the nebulous concept of possessing 'special' VMA flags.

In any case 'special' is not so special a term of art in mm - it includes
VDSO/VVAR mappings, special in the sense of vm_normal_folio() and probably
other cases too.

Therefore make things less special by eliminating these now unused flags.

No functional change intended.

Acked-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 include/linux/mm.h              | 8 --------
 tools/testing/vma/include/dup.h | 8 --------
 2 files changed, 16 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index e8d2b0e847b3..c5644732ff0f 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -576,14 +576,6 @@ enum {
 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
 #define VMA_ACCESS_FLAGS mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, VMA_EXEC_BIT)
 
-/*
- * Special vmas that are non-mergable, non-mlock()able.
- */
-
-#define VMA_SPECIAL_FLAGS mk_vma_flags(VMA_IO_BIT, VMA_DONTEXPAND_BIT, \
-				       VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT)
-#define VM_SPECIAL vma_flags_to_legacy(VMA_SPECIAL_FLAGS)
-
 /*
  * Physically remapped pages are special. Tell the
  * rest of the world about it:
diff --git a/tools/testing/vma/include/dup.h b/tools/testing/vma/include/dup.h
index 19ce10faa9f3..953895832901 100644
--- a/tools/testing/vma/include/dup.h
+++ b/tools/testing/vma/include/dup.h
@@ -352,14 +352,6 @@ enum {
 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
 #define VMA_ACCESS_FLAGS mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, VMA_EXEC_BIT)
 
-/*
- * Special vmas that are non-mergable, non-mlock()able.
- */
-#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
-
-#define VMA_SPECIAL_FLAGS mk_vma_flags(VMA_IO_BIT, VMA_DONTEXPAND_BIT, \
-				       VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT)
-
 #define VMA_REMAP_FLAGS mk_vma_flags(VMA_IO_BIT, VMA_PFNMAP_BIT,	\
 				     VMA_DONTEXPAND_BIT, VMA_DONTDUMP_BIT)
 

-- 
2.55.0

[PATCH v6 38/39] fuse: dax: do not set VM_MIXEDMAP

[Lorenzo Stoakes (ARM)]

Commit e1fb4a086495 ("dax: remove VM_MIXEDMAP for fsdax and device dax")
prevented fsdax and device-dax from setting VM_MIXEDMAP, as DAX no longer
relies on it to direct core mm paths.

The fuse DAX implementation, added later, copied the old pattern and still
sets it.

Fuse DAX maps pages the same way fsdax does, via dax_iomap_fault() and
ultimately vmf_insert_page_mkwrite() and vmf_insert_folio_pmd(), which
insert ordinary refcounted pages and so do not require VM_MIXEDMAP.

Setting it only serves to mark the mapping as not mm-managed, making fuse
DAX the sole DAX implementation whose mappings are unmergeable, cannot be
mlock()'d, eagerly copy page tables on fork and reject MADV_DOFORK and
MADV_DODUMP.

It also requires vma_is_special_huge() in mm/huge_memory.c to carve DAX out
of its check for mappings not managed by core mm explicitly.

There is no reason for fuse DAX to keep on using this flag so drop it.

Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 fs/fuse/dax.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/fuse/dax.c b/fs/fuse/dax.c
index 85cdf0199bc0..a5994f1c637d 100644
--- a/fs/fuse/dax.c
+++ b/fs/fuse/dax.c
@@ -826,7 +826,7 @@ int fuse_dax_mmap(struct file *file, struct vm_area_struct *vma)
 {
 	file_accessed(file);
 	vma->vm_ops = &fuse_dax_vm_ops;
-	vm_flags_set(vma, VM_MIXEDMAP | VM_HUGEPAGE);
+	vma_set_flags(vma, VMA_HUGEPAGE_BIT);
 	return 0;
 }
 

-- 
2.55.0

[PATCH v6 39/39] mm/huge_memory: remove vma_is_special_huge()

[Lorenzo Stoakes (ARM)]

vma_is_special_huge() tests whether either the VMA_PFNMAP_BIT or
VMA_MIXEDMAP_BIT is set (i.e. whether the VMA is not mm-managed),
but with a DAX carve-out.

DAX however no longer sets VMA_MIXEDMAP_BIT, so this carve-out is no longer
required.

Therefore test for !vma_is_mm_managed() instead and also drop the
VMA_IO_BIT check, as it is now redundant since it is enforced that
mm-managed mappings may not set VMA_IO_BIT.

This also eliminates another overloaded use of 'special' within mm.

No functional change intended.

Acked-by: Zi Yan <ziy@nvidia.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
---
 mm/huge_memory.c | 18 ++++--------------
 1 file changed, 4 insertions(+), 14 deletions(-)

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 5b295a7979cd..73575145db7c 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -110,14 +110,6 @@ static inline bool file_thp_enabled(const struct vm_area_struct *vma)
 	return S_ISREG(inode->i_mode);
 }
 
-/* If returns true, we are unable to access the VMA's folios. */
-static bool vma_is_special_huge(const struct vm_area_struct *vma)
-{
-	if (vma_is_dax(vma))
-		return false;
-	return vma_test_any(vma, VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT);
-}
-
 static bool vma_file_bypass_thp_tuneables(const struct vm_area_struct *vma,
 		enum tva_type type)
 {
@@ -192,7 +184,7 @@ unsigned long __thp_vma_allowable_orders(struct vm_area_struct *vma,
 	/* Check the intersection of requested and supported orders. */
 	if (vma_is_anonymous(vma))
 		supported_orders = THP_ORDERS_ALL_ANON;
-	else if (vma_is_dax(vma) || vma_is_special_huge(vma))
+	else if (vma_is_dax(vma) || !vma_is_mm_managed(vma))
 		supported_orders = THP_ORDERS_ALL_SPECIAL_DAX;
 	else
 		supported_orders = THP_ORDERS_ALL_FILE_DEFAULT;
@@ -3065,7 +3057,7 @@ int zap_huge_pud(struct mmu_gather *tlb, struct vm_area_struct *vma,
 	orig_pud = pudp_huge_get_and_clear_full(vma, addr, pud, tlb->fullmm);
 	arch_check_zapped_pud(vma, orig_pud);
 	tlb_remove_pud_tlb_entry(tlb, pud, addr);
-	if (vma_is_special_huge(vma)) {
+	if (!vma_is_mm_managed(vma)) {
 		spin_unlock(ptl);
 		/* No zero page support yet */
 	} else {
@@ -3221,7 +3213,7 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd,
 		 */
 		if (arch_needs_pgtable_deposit())
 			zap_deposited_table(mm, pmd);
-		if (vma_is_special_huge(vma))
+		if (!vma_is_mm_managed(vma))
 			return;
 		if (unlikely(pmd_is_migration_entry(old_pmd))) {
 			const softleaf_t old_entry = softleaf_from_pmd(old_pmd);
@@ -4764,9 +4756,7 @@ static inline bool vma_not_suitable_for_thp_split(struct vm_area_struct *vma)
 {
 	if (vma_is_dax(vma))
 		return true;
-	if (vma_is_special_huge(vma))
-		return true;
-	if (vma_test(vma, VMA_IO_BIT))
+	if (!vma_is_mm_managed(vma))
 		return true;
 	if (vma_is_hugetlb(vma))
 		return true;

-- 
2.55.0

Re: [PATCH v6 00/39] mm: make VMA flag semantics explicit, eliminate VM_SPECIAL

[Andrew Morton]

On Tue, 13 Oct 2026 16:04:40 +0100 "Lorenzo Stoakes (ARM)" <ljs@kernel.org> wrote:

> The VM_SPECIAL / VMA_SPECIAL_FLAGS mask conflates several unrelated
> properties:
> 
> * Is this managed by core mm, or by a driver mapping MMIO, kernel-allocated
>   pages, or even ordinary pages itself?
> * Can it be expanded or merged?
> * Is this a 'weird' case like mlock where migration might race and we
>   'have' to set invalid flags to notify?
> * Is it another 'weird' case where we just want to stop GUP from touching
>   it?
> 
> Driver writers have often been confused about this, and who can blame them?
> 
> It also interacts badly with the eternal edgecase known as hugetlb - which
> sets VMA_DONTEXPAND_BIT but doesn't also want to be treated like a
> 'special' flag.
> 
> Another issue is that we cannot make sensible assumptions about flag
> use. It's not possible to assume VMA_IO_BIT means iommu because drivers
> abuse it and mlock abuses it.
> 
> Special is also an overloaded term in mm. VDSO and VVAR mappings are also
> called 'special' but they're special in a... special way.
> 
> Sometimes things are called special that are a subset of
> VMA_SPECIAL_FLAGS (VMA_PFNMAP_BIT and VMA_MIXEDMAP_BIT for instance when it
> comes to zapping or vm_normal_folio()).
> 
> There's a specific kind of special for THP too, which considers
> PFN map, mixed map 'special' but DAX not.
> 
> It's all rather a mess.
> 
> This series brings some order to things by both limiting what drivers can
> do with VMA flags and switching to using predicates that describe
> behaviour, not arbitrary flags.

Thanks, I've updated mm-unstable to this version.

> v6:
> * Applied tags (thanks David!)
> * Added 7/39 to split out the vm_insert_page() / vm_insert_pages()
>   mmap lock assertions and VMA API changes out into their own patch,
>   as per David.
> * Updated 8/39 (was 7/38) to collapse the discontig single page case to
>   a single one and to rename things, as per David.
> * Updated 9/39 (was 8/38) to update the discontig docs to reflect the
>   change.
> * Updated 23/39 (was 22/38) to put expression on single line, as per David.
> * Updated 25/39 (was 24/38) to mention the open-coding in the commit
>   message (fixed in next commit), as per discussion with David.
> * Updated 26/39 (was 25/38) to clarify the commit message around the
>   removal of the VMA_IO_BIT check in vma_flags_is_mm_managed(), as per
>   David.
> * Updated 30/39 (was 30/38) to no longer exclude droppable mappings from
>   vma_flags_is_mm_backed(), as it doesn't quite fit into the broader
>   definition, as per David.
> * Updated 30/39 (was 30/38) to express vma_flags_can_merge() in terms of
>   vma_flags_is_mm_backed() (with an explicit hugetlb exclusion), and to
>   no longer use vma_can_merge() in __thp_vma_allowable_orders().
> * Updated 31/39 (was 29/38) to use vma_is_mm_backed() (with hugetlb
>   exclusion) rather than vma_can_merge(), as guard regions have nothing to
>   do with mergeability.
> * Updated 32/39 (was 31/38) to retain an explicit droppable check in
>   vma_can_userfault(), as per David.
> * Updated 33/39 (was 32/38) to use vma_is_mm_backed() with an explicit
>   hugetlb exclusion, rather than vma_can_merge() which makes no sense here.
> * Updated 34/39 (was 33/38) to retain explicit droppable checks in
>   vma_supports_mlock() and ksm_compatible(), as per David.
> * Updated 34/39 (was 33/38) to use vma_is_mm_backed() in should_skip_vma(),
>   based on discussion with David.
> * Updated 35/39 (was 34/38) to use vma_is_mm_backed() in
>   lru_gen_look_around(), as per David.
> * Updated 36/39 (was 35/38) to use vma_is_mm_backed() with an explicit
>   hugetlb exclusion in migrate_vma_setup(), as per David.

Here's how v6 altered mm.git.


 Documentation/filesystems/mmap_prepare.rst |    2 
 include/linux/mm.h                         |  120 +++++++++----------
 include/linux/mm_types.h                   |    6 
 mm/huge_memory.c                           |    7 -
 mm/internal.h                              |    3 
 mm/ksm.c                                   |    3 
 mm/madvise.c                               |   11 +
 mm/memory.c                                |   16 +-
 mm/migrate_device.c                        |    4 
 mm/userfaultfd.c                           |    3 
 mm/vma.c                                   |    4 
 mm/vmscan.c                                |    7 -
 tools/testing/vma/include/dup.h            |   44 +++++-
 13 files changed, 128 insertions(+), 102 deletions(-)

--- a/Documentation/filesystems/mmap_prepare.rst~b
+++ a/Documentation/filesystems/mmap_prepare.rst
@@ -240,7 +240,7 @@ In the ``get`` hook, the user must choos
   compound pages (if the compound page is bigger than the remaining pages in the
   VMA, then only those pages that fit will be mapped). For a compound page, the
   head page must be passed.
-* ``discontig_kernel_map_page_range()`` - Map an array of pages of a specified
+* ``discontig_kernel_map_pages()`` - Map an array of pages of a specified
   size. As with a compound page, if more pages are specified than remain in
   the VMA then only those that fit will be mapped.
 
--- a/include/linux/mm.h~b
+++ a/include/linux/mm.h
@@ -1695,57 +1695,22 @@ static inline bool vma_is_fixed_mapping(
 }
 
 /**
- * vma_flags_can_merge() - Do the specified VMA flags permit the VMA to be
- * merged with another?
- * @flags: The VMA flags to test.
- * Returns: true if the flags permit merging, false otherwise.
- */
-static inline bool vma_flags_can_merge(const vma_flags_t *flags)
-{
-	/*
-	 * VMA merging assumes that a VMA's flags and fields completely describe
-	 * its state.
-	 *
-	 * However, mappings which are not mm-managed may have established state
-	 * upon mapping not embodied in any attribute of the VMA.
-	 *
-	 * Additionally, private (CoW) PFN maps encode the source PFN of the
-	 * range in vma->vm_pgoff, which may otherwise cause spurious merges.
-	 */
-	if (!vma_flags_is_mm_managed(flags))
-		return false;
-	/* VMA explicitly marked as being unmergeable. */
-	if (vma_flags_is_fixed_mapping(flags))
-		return false;
-
-	return true;
-}
-
-/**
- * vma_can_merge() - Do @vma's flags permit it to be merged with another VMA?
- * @vma: The VMA to test.
- * Returns: true if the flags permit merging, otherwise false.
- */
-static inline bool vma_can_merge(const struct vm_area_struct *vma)
-{
-	return vma_flags_can_merge(&vma->flags);
-}
-
-/**
  * vma_flags_is_mm_backed() - Do the specified VMA flags imply the mapping is
  * backed by core mm?
  * @flags: The VMA flags to test.
  *
  * An mm-backed mapping is one managed by core mm (see
- * vma_flags_is_mm_managed()) for which two further things are true:
+ * vma_flags_is_mm_managed()) whose contents are established by core mm's own
+ * fault path rather than a custom ->fault handler.
  *
- * 1. Everything mapped was placed there by core mm's fault path (no custom
- *    ->fault).
- * 2. What is mapped stays there until it is zapped or unmapped by the user.
- *
- * These mappings are the only ones which can be sensibly locked, merged, dumped
- * or be subject to uffd faulting, as all of these assume core mm and core mm
- * alone manages these, and that the mappings will remain there.
+ * Mappings not managed by core mm are populated by their owner, and fixed
+ * mappings may be mm-managed but have their contents established by a custom
+ * ->fault handler, so neither are mm-backed. hugetlb mappings are fixed, but
+ * core mm owns them entirely, so they are.
+ *
+ * These are the only mappings which can sensibly be merged, locked, dumped or
+ * be subject to uffd faulting, as all of these assume core mm and core mm
+ * alone manages their contents.
  *
  * Returns: true if the flags imply an mm-backed mapping, otherwise false.
  */
@@ -1761,11 +1726,7 @@ static inline bool vma_flags_is_mm_backe
 	 * by a custom ->fault handler, so core mm cannot assume ordinary fault
 	 * semantics over them.
 	 */
-	if (!vma_flags_is_mm_managed(flags) ||
-	    vma_flags_is_fixed_mapping(flags))
-		return false;
-	/* Core mm may discard droppable memory at any time. */
-	return !vma_flags_test_single_mask(flags, VMA_DROPPABLE);
+	return vma_flags_is_mm_managed(flags) && !vma_flags_is_fixed_mapping(flags);
 }
 
 /**
@@ -1782,6 +1743,43 @@ static inline bool vma_is_mm_backed(cons
 }
 
 /**
+ * vma_flags_can_merge() - Do the specified VMA flags permit the VMA to be
+ * merged with another?
+ * @flags: The VMA flags to test.
+ * Returns: true if the flags permit merging, false otherwise.
+ */
+static inline bool vma_flags_can_merge(const vma_flags_t *flags)
+{
+	/*
+	 * VMA merging assumes that a VMA's flags and fields completely describe
+	 * its state, which is only the case for mm-backed mappings.
+	 *
+	 * Mappings which are not mm-managed may have established state upon
+	 * mapping not embodied in any attribute of the VMA.
+	 *
+	 * Additionally, private (CoW) PFN maps encode the source PFN of the
+	 * range in vma->vm_pgoff, which may otherwise cause spurious merges.
+	 */
+	if (!vma_flags_is_mm_backed(flags))
+		return false;
+	/*
+	 * hugetlb mappings contain per-VMA state, so their flags and fields do
+	 * not completely describe them.
+	 */
+	return !vma_flags_is_hugetlb(flags);
+}
+
+/**
+ * vma_can_merge() - Do @vma's flags permit it to be merged with another VMA?
+ * @vma: The VMA to test.
+ * Returns: true if the flags permit merging, otherwise false.
+ */
+static inline bool vma_can_merge(const struct vm_area_struct *vma)
+{
+	return vma_flags_can_merge(&vma->flags);
+}
+
+/**
  * vma_kernel_pagesize - Default page size granularity for this VMA.
  * @vma: The user mapping.
  *
@@ -4824,28 +4822,26 @@ discontig_kernel_map_abort(struct discon
 
 static inline void
 discontig_kernel_map_page(struct discontig_kernel_page_state *state,
-			  struct page *page)
+		struct page *page)
 {
 	struct folio *folio = page_folio(page);
+	unsigned long nr_pages = 1;
 
 	if (folio_test_large(folio)) {
 		VM_WARN_ON_ONCE(page != folio_page(folio, 0));
-		state->action = DISCONTIG_KERNEL_PAGE_MAP_COMPOUND_PAGE;
-		state->__folio = folio;
-		state->__nr_pages = min(state->nr_pages_remain,
-					folio_nr_pages(folio));
-	} else {
-		state->action = DISCONTIG_KERNEL_PAGE_MAP_PAGE;
-		state->__page = page;
-		state->__nr_pages = 1;
+		nr_pages = folio_nr_pages(folio);
 	}
+
+	state->action = DISCONTIG_KERNEL_PAGE_MAP_CONTIG_PAGES;
+	state->__page = page;
+	state->__nr_pages = min(nr_pages, state->nr_pages_remain);
 }
 
 static inline void
-discontig_kernel_map_page_range(struct discontig_kernel_page_state *state,
-				struct page **page_arr, unsigned long nr_pages)
+discontig_kernel_map_pages(struct discontig_kernel_page_state *state,
+		struct page **page_arr, unsigned long nr_pages)
 {
-	state->action = DISCONTIG_KERNEL_PAGE_MAP_PAGE_RANGE;
+	state->action = DISCONTIG_KERNEL_PAGE_MAP_PAGES;
 	state->__page_arr = page_arr;
 	state->__nr_pages = min(nr_pages, state->nr_pages_remain);
 }
--- a/include/linux/mm_types.h~b
+++ a/include/linux/mm_types.h
@@ -825,9 +825,8 @@ enum mmap_action_type {
 
 enum discontig_kernel_page_action {
 	DISCONTIG_KERNEL_PAGE_ABORT,
-	DISCONTIG_KERNEL_PAGE_MAP_PAGE,
-	DISCONTIG_KERNEL_PAGE_MAP_COMPOUND_PAGE,
-	DISCONTIG_KERNEL_PAGE_MAP_PAGE_RANGE,
+	DISCONTIG_KERNEL_PAGE_MAP_PAGES,	/* Array of individual pages. */
+	DISCONTIG_KERNEL_PAGE_MAP_CONTIG_PAGES,	/* 1..N contiguous pages. */
 };
 
 struct discontig_kernel_page_state {
@@ -847,7 +846,6 @@ struct discontig_kernel_page_state {
 	enum discontig_kernel_page_action action;
 	union {
 		struct page *__page;
-		struct folio *__folio;
 		struct page **__page_arr;
 	};
 	unsigned long __nr_pages;
--- a/mm/huge_memory.c~b
+++ a/mm/huge_memory.c
@@ -207,11 +207,10 @@ unsigned long __thp_vma_allowable_orders
 	 * khugepaged moves data from VMAs once collapsed, after they have been
 	 * faulted in, relying on refaulting for file-backed memory.
 	 *
-	 * Mappings not managed by core mm cannot be reliably reconstructed from
-	 * page faults, and fixed mappings (including hugetlb) may well be
-	 * mm-managed - precisely the mappings which cannot be merged.
+	 * Only mm-backed mappings can be reliably reconstructed from page
+	 * faults, and hugetlb mappings cannot be collapsed.
 	 */
-	if (!in_pf && !smaps && !vma_can_merge(vma))
+	if (!in_pf && !smaps && (!vma_is_mm_backed(vma) || vma_is_hugetlb(vma)))
 		return 0;
 
 	/*
--- a/mm/internal.h~b
+++ a/mm/internal.h
@@ -1114,6 +1114,9 @@ static inline bool vma_supports_mlock(co
 {
 	if (!vma_is_mm_backed(vma))
 		return false;
+	/* Core mm may discard droppable memory at any time. */
+	if (vma_test_single_mask(vma, VMA_DROPPABLE))
+		return false;
 	if (vma_is_dax(vma) || vma_is_hugetlb(vma))
 		return false;
 	return vma != get_gate_vma(current->mm);
--- a/mm/ksm.c~b
+++ a/mm/ksm.c
@@ -749,6 +749,9 @@ static bool ksm_compatible(const struct
 		return false;
 	if (!vma_flags_is_mm_backed(&vma_flags))
 		return false;
+	/* Core mm may discard droppable memory at any time. */
+	if (vma_flags_test_single_mask(&vma_flags, VMA_DROPPABLE))
+		return false;
 	if (file_is_dax(file))
 		return false;
 #ifdef VM_SAO
--- a/mm/madvise.c~b
+++ a/mm/madvise.c
@@ -1069,10 +1069,10 @@ static bool is_valid_guard_vma(const str
 		return false;
 	/*
 	 * Guard regions require a VMA whose page tables are managed solely by
-	 * the core, which is also what merging requires, so disallow any flags
-	 * that would prevent a merge.
+	 * core mm, so the VMA must be mm-backed (excluding hugetlb, in which
+	 * guard markers cannot be installed).
 	 */
-	if (!vma_can_merge(vma))
+	if (!vma_is_mm_backed(vma) || vma_is_hugetlb(vma))
 		return false;
 
 	return true;
@@ -1402,7 +1402,7 @@ static int madvise_vma_behavior(struct m
 		new_flags |= VM_DONTCOPY;
 		break;
 	case MADV_DOFORK:
-		if (!vma_can_merge(vma))
+		if (!vma_is_mm_backed(vma) || vma_is_hugetlb(vma))
 			return -EINVAL;
 		new_flags &= ~VM_DONTCOPY;
 		break;
@@ -1424,6 +1424,9 @@ static int madvise_vma_behavior(struct m
 		/* Only mm-backed memory can be meaningfully dumped. */
 		if (!vma_is_mm_backed(vma))
 			return -EINVAL;
+		/* Core mm may discard droppable memory at any time. */
+		if (new_flags & VM_DROPPABLE)
+			return -EINVAL;
 		new_flags &= ~VM_DONTDUMP;
 		break;
 	case MADV_MERGEABLE:
--- a/mm/memory.c~b
+++ a/mm/memory.c
@@ -2674,26 +2674,24 @@ int map_discontig_kernel_pages_prepare(s
 }
 
 static int apply_discontig_action(struct vm_area_struct *vma,
-				  struct discontig_kernel_page_state *state)
+		struct discontig_kernel_page_state *state)
 {
 	unsigned long nr_pages = state->__nr_pages;
 	unsigned long addr = state->addr;
 	unsigned long i;
 
-	if (state->action == DISCONTIG_KERNEL_PAGE_MAP_PAGE)
-		return insert_page(vma, addr, state->__page,
-				   vma->vm_page_prot, /*mkwrite=*/false);
-	if (state->action == DISCONTIG_KERNEL_PAGE_MAP_PAGE_RANGE)
+	if (state->action == DISCONTIG_KERNEL_PAGE_MAP_PAGES)
 		return insert_pages(vma, addr, state->__page_arr,
-				    &nr_pages, vma->vm_page_prot);
+				&nr_pages, vma->vm_page_prot);
 
-	/* Compound folio - have to iterate through each page. */
+	VM_WARN_ON_ONCE(state->action != DISCONTIG_KERNEL_PAGE_MAP_CONTIG_PAGES);
+	/* Contiguous pages - have to iterate through each one. */
 	for (i = 0; i < nr_pages; i++, addr += PAGE_SIZE) {
-		struct page *page = folio_page(state->__folio, i);
+		struct page *page = state->__page + i;
 		int err;
 
 		err = insert_page(vma, addr, page, vma->vm_page_prot,
-				  /*mkwrite=*/false);
+				/*mkwrite=*/false);
 		if (err)
 			return err;
 	}
--- a/mm/migrate_device.c~b
+++ a/mm/migrate_device.c
@@ -746,8 +746,8 @@ int migrate_vma_setup(struct migrate_vma
 	args->end &= PAGE_MASK;
 	if (!vma)
 		return -EINVAL;
-	if (!vma_is_mm_managed(vma) || vma_is_fixed_mapping(vma) ||
-	    vma_is_dax(vma))
+	/* The migrate_vma page walk has no hugetlb support. */
+	if (!vma_is_mm_backed(vma) || vma_is_hugetlb(vma) || vma_is_dax(vma))
 		return -EINVAL;
 	if (nr_pages <= 0)
 		return -EINVAL;
--- a/mm/userfaultfd.c~b
+++ a/mm/userfaultfd.c
@@ -2156,6 +2156,9 @@ static bool vma_can_userfault(struct vm_
 	/* Only mm-backed memory can have its faults handled by userspace. */
 	if (!vma_is_mm_backed(vma))
 		return false;
+	/* Core mm may discard droppable memory at any time. */
+	if (vma_test_single_mask(vma, VMA_DROPPABLE))
+		return false;
 	/* The shadow stack should not be written to by userspace. */
 	if (vma_test_single_mask(vma, VMA_SHADOW_STACK))
 		return false;
--- a/mm/vma.c~b
+++ a/mm/vma.c
@@ -2654,9 +2654,7 @@ static int __mmap_new_file_vma(struct mm
 	 * VMA_LOCKED_BIT and/or VMA_LOCKONFAULT_BIT flags.
 	 */
 	if (vma_supports_mlock(vma)) {
-		const vma_flags_t mask =
-			vma_flags_and_mask(&map->vma_flags,
-					   VMA_LOCKED_MASK);
+		const vma_flags_t mask = vma_flags_and_mask(&map->vma_flags, VMA_LOCKED_MASK);
 
 		vma_set_flags_mask(vma, mask);
 	}
--- a/mm/vmscan.c~b
+++ a/mm/vmscan.c
@@ -3521,8 +3521,7 @@ static int should_skip_vma(unsigned long
 	if (!vma_has_recency(vma))
 		return true;
 
-	if (vma_test(vma, VMA_LOCKED_BIT) || !vma_is_mm_managed(vma) ||
-	    vma_is_fixed_mapping(vma))
+	if (vma_test(vma, VMA_LOCKED_BIT) || !vma_is_mm_backed(vma))
 		return true;
 
 	if (vma == get_gate_vma(vma->vm_mm))
@@ -4462,8 +4461,8 @@ bool lru_gen_look_around(struct page_vma
 	if (spin_is_contended(pvmw->ptl))
 		return true;
 
-	/* exclude fixed and non-mm-managed VMAs containing anon pages from COW */
-	if (!vma_is_mm_managed(vma) || vma_is_fixed_mapping(vma))
+	/* exclude non-mm-backed VMAs containing anon pages from COW */
+	if (!vma_is_mm_backed(vma))
 		return true;
 
 	/* avoid taking the LRU lock under the PTL when possible */
--- a/tools/testing/vma/include/dup.h~b
+++ a/tools/testing/vma/include/dup.h
@@ -1681,23 +1681,49 @@ static inline bool vma_is_fixed_mapping(
 	return vma_flags_is_fixed_mapping(&vma->flags);
 }
 
+static inline bool vma_flags_is_hugetlb(const vma_flags_t *flags)
+{
+	return IS_ENABLED(CONFIG_HUGETLB_PAGE) &&
+	       vma_flags_test(flags, VMA_HUGETLB_BIT);
+}
+
+static inline bool vma_flags_is_mm_backed(const vma_flags_t *flags)
+{
+	/* hugetlb is a fixed mapping, but core mm owns it entirely. */
+	if (vma_flags_is_hugetlb(flags))
+		return true;
+	/*
+	 * Mappings not managed by core mm are populated by their owner.
+	 *
+	 * Fixed mappings may be mm-managed, but their contents are established
+	 * by a custom ->fault handler, so core mm cannot assume ordinary fault
+	 * semantics over them.
+	 */
+	return vma_flags_is_mm_managed(flags) && !vma_flags_is_fixed_mapping(flags);
+}
+
+static inline bool vma_is_mm_backed(const struct vm_area_struct *vma)
+{
+	return vma_flags_is_mm_backed(&vma->flags);
+}
+
 static inline bool vma_flags_can_merge(const vma_flags_t *flags)
 {
 	/*
 	 * VMA merging assumes that a VMA's flags and fields completely describe
-	 * its state.
+	 * its state, which is only the case for mm-backed mappings.
 	 *
-	 * However, mappings which are not mm-managed may have established state
-	 * upon mapping not embodied in any attribute of the VMA.
+	 * Mappings which are not mm-managed may have established state upon
+	 * mapping not embodied in any attribute of the VMA.
 	 *
 	 * Additionally, private (CoW) PFN maps encode the source PFN of the
 	 * range in vma->vm_pgoff, which may otherwise cause spurious merges.
 	 */
-	if (!vma_flags_is_mm_managed(flags))
+	if (!vma_flags_is_mm_backed(flags))
 		return false;
-	/* VMA explicitly marked as being unmergeable. */
-	if (vma_flags_is_fixed_mapping(flags))
-		return false;
-
-	return true;
+	/*
+	 * hugetlb mappings contain per-VMA state, so their flags and fields do
+	 * not completely describe them.
+	 */
+	return !vma_flags_is_hugetlb(flags);
 }
_