On Tue, Sep 6, 2022 at 7:48 AM Laurent Dufour [off-list ref] wrote:

Le 01/09/2022 à 19:34, Suren Baghdasaryan a écrit :

quoted

Protect VMA from concurrent page fault handler while performing VMA
protection policy changes.

Signed-off-by: Suren Baghdasaryan <surenb@google.com>
---
 mm/mempolicy.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index b73d3248d976..6be1e5c75556 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c

@@ -383,8 +383,10 @@ void mpol_rebind_mm(struct mm_struct *mm, nodemask_t *new)
      struct vm_area_struct *vma;

      mmap_write_lock(mm);
-     for (vma = mm->mmap; vma; vma = vma->vm_next)
+     for (vma = mm->mmap; vma; vma = vma->vm_next) {
+             vma_mark_locked(vma);
              mpol_rebind_policy(vma->vm_policy, new);
+     }
      mmap_write_unlock(mm);
 }

@@ -632,6 +634,7 @@ unsigned long change_prot_numa(struct vm_area_struct *vma,
      struct mmu_gather tlb;
      int nr_updated;

+     vma_mark_locked(vma);

If I understand that corretly, the VMA itself is not impacted, only the
PMDs/PTEs, and they are protected using the page table locks.

Am I missing something?

I thought we would not want pages faulting in the VMA for which we are
changing the protection. However I think what you are saying is that
page table locks would already provide a more granular synchronization
with page fault handlers, which makes sense to me. Sounds like we can
skip locking the VMA here as well. Nice!

quoted

      tlb_gather_mmu(&tlb, vma->vm_mm);

      nr_updated = change_protection(&tlb, vma, addr, end, PAGE_NONE,

@@ -765,6 +768,7 @@ static int vma_replace_policy(struct vm_area_struct *vma,
      if (IS_ERR(new))
              return PTR_ERR(new);

+     vma_mark_locked(vma);
      if (vma->vm_ops && vma->vm_ops->set_policy) {
              err = vma->vm_ops->set_policy(vma, new);
              if (err)