]> git.ipfire.org Git - thirdparty/kernel/linux.git/commitdiff
kernfs: Drop kernfs_rwsem while invoking lookup_positive_unlocked().
authorSebastian Andrzej Siewior <bigeasy@linutronix.de>
Tue, 18 Feb 2025 16:39:38 +0000 (17:39 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 19 Feb 2025 16:07:41 +0000 (17:07 +0100)
syzbot reported two warnings:
- kernfs_node::name was accessed outside of a RCU section so it created
  warning. The kernfs_rwsem was held so it was okay but it wasn't seen.

- While kernfs_rwsem was held invoked lookup_positive_unlocked()->
  kernfs_dop_revalidate() which acquired kernfs_rwsem.

kernfs_rwsem was both acquired as a read lock so it can be acquired
twice. However if a writer acquires the lock after the first reader then
neither the writer nor the second reader can obtain the lock so it
deadlocks.

The reason for the lock is to ensure that kernfs_node::name remain
stable during lookup_positive_unlocked()'s invocation. The function can
not be invoked within a RCU section because it may sleep.

Make a temporary copy of the kernfs_node::name under the lock so
GFP_KERNEL can be used and use this instead.

Reported-by: syzbot+ecccecbc636b455f9084@syzkaller.appspotmail.com
Fixes: 5b2fabf7fe8f ("kernfs: Acquire kernfs_rwsem in kernfs_node_dentry().")
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20250218163938.xmvjlJ0K@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/kernfs/mount.c

index d1f512b7bf8678c4af0964af51525d7a416c3edd..f1cea282aae325daef8500ec07168a08aba9d347 100644 (file)
@@ -220,12 +220,19 @@ struct dentry *kernfs_node_dentry(struct kernfs_node *kn,
                return dentry;
 
        root = kernfs_root(kn);
-       guard(rwsem_read)(&root->kernfs_rwsem);
-
-       knparent = find_next_ancestor(kn, NULL);
-       if (WARN_ON(!knparent)) {
-               dput(dentry);
+       /*
+        * As long as kn is valid, its parent can not vanish. This is cgroup's
+        * kn so it not have its parent replaced. Therefore it is safe to use
+        * the ancestor node outside of the RCU or locked section.
+        */
+       if (WARN_ON_ONCE(!(root->flags & KERNFS_ROOT_INVARIANT_PARENT)))
                return ERR_PTR(-EINVAL);
+       scoped_guard(rcu) {
+               knparent = find_next_ancestor(kn, NULL);
+               if (WARN_ON(!knparent)) {
+                       dput(dentry);
+                       return ERR_PTR(-EINVAL);
+               }
        }
 
        do {
@@ -235,14 +242,22 @@ struct dentry *kernfs_node_dentry(struct kernfs_node *kn,
 
                if (kn == knparent)
                        return dentry;
-               kntmp = find_next_ancestor(kn, knparent);
-               if (WARN_ON(!kntmp)) {
+
+               scoped_guard(rwsem_read, &root->kernfs_rwsem) {
+                       kntmp = find_next_ancestor(kn, knparent);
+                       if (WARN_ON(!kntmp)) {
+                               dput(dentry);
+                               return ERR_PTR(-EINVAL);
+                       }
+                       name = kstrdup(kernfs_rcu_name(kntmp), GFP_KERNEL);
+               }
+               if (!name) {
                        dput(dentry);
-                       return ERR_PTR(-EINVAL);
+                       return ERR_PTR(-ENOMEM);
                }
-               name = rcu_dereference(kntmp->name);
                dtmp = lookup_positive_unlocked(name, dentry, strlen(name));
                dput(dentry);
+               kfree(name);
                if (IS_ERR(dtmp))
                        return dtmp;
                knparent = kntmp;