On Thursday, 16 December 2021 5:17:30 PM AEDT Peter Xu wrote:

On Thu, Dec 16, 2021 at 04:56:42PM +1100, Alistair Popple wrote:

quoted

On Monday, 15 November 2021 6:55:05 PM AEDT Peter Xu wrote:

[...]

quoted

diff --git a/mm/memory.c b/mm/memory.c
index d5966d9e24c3..e8557d43a87d 100644
--- a/mm/memory.c
+++ b/mm/memory.c

@@ -3452,6 +3452,43 @@ static vm_fault_t remove_device_exclusive_entry(struct vm_fault *vmf)
 	return 0;
 }
 
+static vm_fault_t pte_marker_clear(struct vm_fault *vmf)
+{
+	vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm, vmf->pmd,
+				       vmf->address, &vmf->ptl);
+	/*
+	 * Be careful so that we will only recover a special uffd-wp pte into a
+	 * none pte.  Otherwise it means the pte could have changed, so retry.
+	 */
+	if (is_pte_marker(*vmf->pte))
+		pte_clear(vmf->vma->vm_mm, vmf->address, vmf->pte);
+	pte_unmap_unlock(vmf->pte, vmf->ptl);
+	return 0;
+}
+
+/*
+ * This is actually a page-missing access, but with uffd-wp special pte
+ * installed.  It means this pte was wr-protected before being unmapped.
+ */
+static vm_fault_t pte_marker_handle_uffd_wp(struct vm_fault *vmf)
+{
+	/* Careful!  vmf->pte unmapped after return */
+	if (!pte_unmap_same(vmf))

Hasn't vmf->pte already been unmapped by do_swap_page() by the time we get
here?

Great catch, thanks!

It was needed before with the "swap special pte" version because that was
handled outside do_swap_page().  After the rebase I forgot to remove it.

No worries, and for what it's worth IMHO this version that handles it inside
do_swap_page() along with all the other "special" cases is much nicer.

I believe it didn't crash simply because we've got commit 2ca99358671a ("mm:
clear vmf->pte after pte_unmap_same() returns", 2021-11-06) very recently so it
just became a safe no-op, so all things will still work.

I'll drop it.

quoted

quoted

+		return 0;
+
+	/*
+	 * Just in case there're leftover special ptes even after the region
+	 * got unregistered - we can simply clear them.  We can also do that
+	 * proactively when e.g. when we do UFFDIO_UNREGISTER upon some uffd-wp
+	 * ranges, but it should be more efficient to be done lazily here.
+	 */
+	if (unlikely(!userfaultfd_wp(vmf->vma) || vma_is_anonymous(vmf->vma)))
+		return pte_marker_clear(vmf);
+
+	/* do_fault() can handle pte markers too like none pte */
+	return do_fault(vmf);
+}
+
 static vm_fault_t handle_pte_marker(struct vm_fault *vmf)
 {
 	swp_entry_t entry = pte_to_swp_entry(vmf->orig_pte);

@@ -3465,8 +3502,11 @@ static vm_fault_t handle_pte_marker(struct vm_fault *vmf)
 	if (WARN_ON_ONCE(vma_is_anonymous(vmf->vma) || !marker))
 		return VM_FAULT_SIGBUS;
 
-	/* TODO: handle pte markers */
-	return 0;
+	if (marker & PTE_MARKER_UFFD_WP)

Can we make this check `marker == PTE_MARKER_UFFD_WP`? There is currently only
one user of pte markers, and from what I can tell pte_marker_handle_uffd_wp()
wouldn't do the correct thing if other users were added because it could clear
non-uffd-wp markers. I don't think it's worth making it do the right thing now,
but a comment noting that would be helpful.

Sure thing, and yeah I agree it's trivial and shouldn't matter in real-life.

I'll change it to "marker == PTE_MARKER_UFFD_WP" as you suggested, so if
there's surprise we'll get a sigbus.

Thanks,

quoted

quoted

+		return pte_marker_handle_uffd_wp(vmf);
+
+	/* This is an unknown pte marker */
+	return VM_FAULT_SIGBUS;
 }