Skip to contentSkip to conversation

mm-rmap-allocate-anon_vma_chain-objects-unlocked-when-possible.patch removed from -mm tree

merged mm-stable Andrew Morton 1 participant 1 message Jan 27

lore
Share & tools
lore

Andrew Morton

author tree notice Applied
#0 lore raw

The quilt patch titled
Subject: mm/rmap: allocate anon_vma_chain objects unlocked when possible
has been removed from the -mm tree. Its filename was
mm-rmap-allocate-anon_vma_chain-objects-unlocked-when-possible.patch

This patch was dropped because it was merged into the mm-stable branch of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm

------------------------------------------------------
From: Lorenzo Stoakes <[email protected]>
Subject: mm/rmap: allocate anon_vma_chain objects unlocked when possible
Date: Sun, 18 Jan 2026 14:50:44 +0000

There is no reason to allocate the anon_vma_chain under the anon_vma write lock when cloning - we can in fact assign these to the destination VMA safely as we hold the exclusive mmap lock and therefore preclude anybody else accessing these fields.

We only need take the anon_vma write lock when we link rbtree edges from the anon_vma to the newly established AVCs.

This also allows us to eliminate the weird GFP_NOWAIT, GFP_KERNEL dance introduced in commit dd34739c03f2 ("mm: avoid anon_vma_chain allocation under anon_vma lock"), further simplifying this logic.

This should reduce lock anon_vma contention, and clarifies exactly where the anon_vma lock is required.

We cannot adjust __anon_vma_prepare() in the same way as this is only protected by VMA read lock, so we have to perform the allocation here under the anon_vma write lock and page_table_lock (to protect against racing threads), and we wish to retain the lock ordering.

With this change we can simplify cleanup_partial_anon_vmas() even further
- since we allocate AVC's without any lock taken and do not insert anything into the interval tree until after the allocations are tried, we can remove all logic pertaining to this and just free up AVC's only.

---

 mm/rmap.c |   81 +++++++++++++++++++++++-----------------------------
 1 file changed, 36 insertions(+), 45 deletions(-)
mm/rmap.c renamed from mm/rmap.c~mm-rmap-allocate-anon_vma_chain-objects-unlocked-when-possible +36 -45
@@ -146,14 +146,13 @@ static void anon_vma_chain_free(struct a
 	kmem_cache_free(anon_vma_chain_cachep, anon_vma_chain);
 }
 
-static void anon_vma_chain_link(struct vm_area_struct *vma,
-				struct anon_vma_chain *avc,
-				struct anon_vma *anon_vma)
+static void anon_vma_chain_assign(struct vm_area_struct *vma,
+				  struct anon_vma_chain *avc,
+				  struct anon_vma *anon_vma)
 {
 	avc->vma = vma;
 	avc->anon_vma = anon_vma;
 	list_add(&avc->same_vma, &vma->anon_vma_chain);
-	anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
 }
 
 /**
@@ -210,7 +209,8 @@ int __anon_vma_prepare(struct vm_area_st
 	spin_lock(&mm->page_table_lock);
 	if (likely(!vma->anon_vma)) {
 		vma->anon_vma = anon_vma;
-		anon_vma_chain_link(vma, avc, anon_vma);
+		anon_vma_chain_assign(vma, avc, anon_vma);
+		anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
 		anon_vma->num_active_vmas++;
 		allocated = NULL;
 		avc = NULL;
@@ -291,21 +291,33 @@ int anon_vma_clone(struct vm_area_struct
 
 	check_anon_vma_clone(dst, src);
 
-	/* All anon_vma's share the same root. */
+	/*
+	 * Allocate AVCs. We don't need an anon_vma lock for this as we
+	 * are not updating the anon_vma rbtree nor are we changing
+	 * anon_vma statistics.
+	 *
+	 * Either src, dst have the same mm for which we hold an exclusive mmap
+	 * write lock, or we are forking and we hold it on src->vm_mm and dst is
+	 * not yet accessible to other threads so there's no possibliity of the
+	 * unlinked AVC's being observed yet.
+	 */
+	list_for_each_entry(pavc, &src->anon_vma_chain, same_vma) {
+		avc = anon_vma_chain_alloc(GFP_KERNEL);
+		if (!avc)
+			goto enomem_failure;
+
+		anon_vma_chain_assign(dst, avc, pavc->anon_vma);
+	}
+
+	/*
+	 * Now link the anon_vma's back to the newly inserted AVCs.
+	 * Note that all anon_vma's share the same root.
+	 */
 	anon_vma_lock_write(src->anon_vma);
-	list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
-		struct anon_vma *anon_vma;
+	list_for_each_entry_reverse(avc, &dst->anon_vma_chain, same_vma) {
+		struct anon_vma *anon_vma = avc->anon_vma;
 
-		avc = anon_vma_chain_alloc(GFP_NOWAIT);
-		if (unlikely(!avc)) {
-			anon_vma_unlock_write(src->anon_vma);
-			avc = anon_vma_chain_alloc(GFP_KERNEL);
-			if (!avc)
-				goto enomem_failure;
-			anon_vma_lock_write(src->anon_vma);
-		}
-		anon_vma = pavc->anon_vma;
-		anon_vma_chain_link(dst, avc, anon_vma);
+		anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
 
 		/*
 		 * Reuse existing anon_vma if it has no vma and only one
@@ -321,7 +333,6 @@ int anon_vma_clone(struct vm_area_struct
 	}
 	if (dst->anon_vma)
 		dst->anon_vma->num_active_vmas++;
-
 	anon_vma_unlock_write(src->anon_vma);
 	return 0;
 
@@ -391,8 +402,10 @@ int anon_vma_fork(struct vm_area_struct
 	get_anon_vma(anon_vma->root);
 	/* Mark this anon_vma as the one where our new (COWed) pages go. */
 	vma->anon_vma = anon_vma;
+	anon_vma_chain_assign(vma, avc, anon_vma);
+	/* Now let rmap see it. */
 	anon_vma_lock_write(anon_vma);
-	anon_vma_chain_link(vma, avc, anon_vma);
+	anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
 	anon_vma->parent->num_children++;
 	anon_vma_unlock_write(anon_vma);
 
@@ -403,40 +416,18 @@ int anon_vma_fork(struct vm_area_struct
  * In the unfortunate case of anon_vma_clone() failing to allocate memory we
  * have to clean things up.
  *
- * On clone we hold the exclusive mmap write lock, so we can't race
- * unlink_anon_vmas(). Since we're cloning, we know we can't have empty
- * anon_vma's, since existing anon_vma's are what we're cloning from.
- *
- * So this function needs only traverse the anon_vma_chain and free each
- * allocated anon_vma_chain.
+ * Since we allocate anon_vma_chain's before we insert them into the interval
+ * trees, we simply have to free up the AVC's and remove the entries from the
+ * VMA's anon_vma_chain.
  */
 static void cleanup_partial_anon_vmas(struct vm_area_struct *vma)
 {
 	struct anon_vma_chain *avc, *next;
-	struct anon_vma *root = NULL;
-
-	/*
-	 * We exclude everybody else from being able to modify anon_vma's
-	 * underneath us.
-	 */
-	mmap_assert_locked(vma->vm_mm);
 
 	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
-		struct anon_vma *anon_vma = avc->anon_vma;
-
-		/* All anon_vma's share the same root. */
-		if (!root) {
-			root = anon_vma->root;
-			anon_vma_lock_write(root);
-		}
-
-		anon_vma_interval_tree_remove(avc, &anon_vma->rb_root);
 		list_del(&avc->same_vma);
 		anon_vma_chain_free(avc);
 	}
-
-	if (root)
-		anon_vma_unlock_write(root);
 }
 
 /**

_

Patches currently in -mm which might be from [email protected] are

selftests-mm-remove-virtual_address_range-test.patch mm-vma-rename-vma_lock_offset-to-vm_refcnt_exclude_readers_flag.patch mm-vma-document-possible-vma-vm_refcnt-values-and-reference-comment.patch mm-vma-rename-is_vma_write_only-separate-out-shared-refcount-put.patch mm-vma-adduse-vma-lockdep-acquire-release-defines.patch mm-vma-de-duplicate-__vma_enter_locked-error-path.patch mm-vma-clean-up-__vma_enter-exit_locked.patch mm-vma-introduce-helper-struct-thread-through-exclusive-lock-fns.patch mm-vma-improve-and-document-__is_vma_write_locked.patch mm-vma-improve-and-document-__is_vma_write_locked-fix.patch mm-vma-update-vma_assert_locked-to-use-lockdep.patch mm-vma-update-vma_assert_locked-to-use-lockdep-fix.patch mm-vma-add-and-use-vma_assert_stabilised.patch mm-vma-remove-__private-sparse-decoration-from-vma_flags_t.patch mm-rename-vma_flag_test-set_atomic-to-vma_test-set_atomic_flag.patch mm-add-mk_vma_flags-bitmap-flag-macro-helper.patch tools-bitmap-add-missing-bitmap_.patch mm-add-basic-vma-flag-operation-helper-functions.patch mm-update-hugetlbfs-to-use-vma-flags-on-mmap_prepare.patch mm-update-secretmem-to-use-vma-flags-on-mmap_prepare.patch mm-update-shmem__file_-functions-to-use-vma_flags_t.patch mm-update-all-remaining-mmap_prepare-users-to-use-vma_flags_t.patch mm-make-vm_area_desc-utilise-vma_flags_t-only.patch tools-testing-vma-separate-vma-userland-tests-into-separate-files.patch tools-testing-vma-separate-out-vma_internalh-into-logical-headers.patch tools-testing-vma-add-vma-userland-tests-for-vma-flag-functions.patch