[11/13] fuse: fix UAF in rcu pathwalks

->permission(), ->get_link() and ->inode_get_acl() might dereference
->s_fs_info (and, in case of ->permission(), ->s_fs_info->fc->user_ns
as well) when called from rcu pathwalk.

Freeing ->s_fs_info->fc is rcu-delayed; we need to make freeing ->s_fs_info
and dropping ->user_ns rcu-delayed too.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
---
 fs/fuse/cuse.c   |  3 +--
 fs/fuse/fuse_i.h |  1 +
 fs/fuse/inode.c  | 15 +++++++++++----
 3 files changed, 13 insertions(+), 6 deletions(-)

On Sun, Feb 04, 2024 at 02:17:37AM +0000, Al Viro wrote:
> ->permission(), ->get_link() and ->inode_get_acl() might dereference
> ->s_fs_info (and, in case of ->permission(), ->s_fs_info->fc->user_ns
> as well) when called from rcu pathwalk.
> 
> Freeing ->s_fs_info->fc is rcu-delayed; we need to make freeing ->s_fs_info
> and dropping ->user_ns rcu-delayed too.
> 
> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
> ---

Reviewed-by: Christian Brauner <brauner@kernel.org>

On Mon, 5 Feb 2024 at 13:31, Christian Brauner <brauner@kernel.org> wrote:
>
> On Sun, Feb 04, 2024 at 02:17:37AM +0000, Al Viro wrote:
> > ->permission(), ->get_link() and ->inode_get_acl() might dereference
> > ->s_fs_info (and, in case of ->permission(), ->s_fs_info->fc->user_ns
> > as well) when called from rcu pathwalk.
> >
> > Freeing ->s_fs_info->fc is rcu-delayed; we need to make freeing ->s_fs_info
> > and dropping ->user_ns rcu-delayed too.
> >
> > Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
> > ---
>
> Reviewed-by: Christian Brauner <brauner@kernel.org>

Acked-by: Miklos Szeredi <mszeredi@redhat.com>

On Mon, Feb 5, 2024 at 3:52 PM Miklos Szeredi <miklos@szeredi.hu> wrote:
>
> On Mon, 5 Feb 2024 at 13:31, Christian Brauner <brauner@kernel.org> wrote:
> >
> > On Sun, Feb 04, 2024 at 02:17:37AM +0000, Al Viro wrote:
> > > ->permission(), ->get_link() and ->inode_get_acl() might dereference
> > > ->s_fs_info (and, in case of ->permission(), ->s_fs_info->fc->user_ns
> > > as well) when called from rcu pathwalk.
> > >
> > > Freeing ->s_fs_info->fc is rcu-delayed; we need to make freeing ->s_fs_info
> > > and dropping ->user_ns rcu-delayed too.
> > >
> > > Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
> > > ---
> >
> > Reviewed-by: Christian Brauner <brauner@kernel.org>
>
> Acked-by: Miklos Szeredi <mszeredi@redhat.com>
>

Miklos,

FYI, this is now merged and conflicts with:

dc076c73b9f9 ("fuse: implement ioctls to manage backing files")

from fuse/for-next:

--- a/fs/fuse/inode.c
+++ b/fs/fuse/inode.c
@@@ -1373,7 -1398,9 +1405,13 @@@ EXPORT_SYMBOL_GPL(fuse_send_init)
  void fuse_free_conn(struct fuse_conn *fc)
  {
        WARN_ON(!list_empty(&fc->devices));
++<<<<<<< HEAD
 +      kfree(fc);
++=======
+       if (IS_ENABLED(CONFIG_FUSE_PASSTHROUGH))
+               fuse_backing_files_free(fc);
+       kfree_rcu(fc, rcu);
++>>>>>>> fuse/for-next
  }
  EXPORT_SYMBOL_GPL(fuse_free_conn);

Note that fuse_backing_files_free() calls
fuse_backing_id_free() => fuse_backing_free() => kfree_rcu()

Should we move fuse_backing_files_free() into
fuse_conn_put() above fuse_dax_conn_free()?

That will avoid the merge conflict and still be correct. no?

Thanks,
Amir.

On Mon, 4 Mar 2024 at 15:36, Amir Goldstein <amir73il@gmail.com> wrote:

> Note that fuse_backing_files_free() calls
> fuse_backing_id_free() => fuse_backing_free() => kfree_rcu()
>
> Should we move fuse_backing_files_free() into
> fuse_conn_put() above fuse_dax_conn_free()?
>
> That will avoid the merge conflict and still be correct. no?

Looks like a good cleanup.

Force-pushed to fuse.git#for-next.

Thanks,
Miklos

On Tue, Mar 5, 2024 at 2:43 PM Miklos Szeredi <miklos@szeredi.hu> wrote:
>
> On Mon, 4 Mar 2024 at 15:36, Amir Goldstein <amir73il@gmail.com> wrote:
>
> > Note that fuse_backing_files_free() calls
> > fuse_backing_id_free() => fuse_backing_free() => kfree_rcu()
> >
> > Should we move fuse_backing_files_free() into
> > fuse_conn_put() above fuse_dax_conn_free()?
> >
> > That will avoid the merge conflict and still be correct. no?
>
> Looks like a good cleanup.
>
> Force-pushed to fuse.git#for-next.
>

FYI, the version that you pushed will generate a minor conflict with

                }
-               fc->release(fc);
+               call_rcu(&fc->rcu, delayed_release);
        }
 }
 EXPORT_SYMBOL_GPL(fuse_conn_put);

If you move fuse_backing_files_free() to the start of the function,
I think merge conflict will be avoided:

 void fuse_conn_put(struct fuse_conn *fc)
 {
        if (refcount_dec_and_test(&fc->count)) {
                struct fuse_iqueue *fiq = &fc->iq;
                struct fuse_sync_bucket *bucket;

+               if (IS_ENABLED(CONFIG_FUSE_PASSTHROUGH))
+                       fuse_backing_files_free(fc);
                if (IS_ENABLED(CONFIG_FUSE_DAX))
                        fuse_dax_conn_free(fc);


Thanks,
Amir.

On Wed, 6 Mar 2024 at 11:18, Amir Goldstein <amir73il@gmail.com> wrote:

> If you move fuse_backing_files_free() to the start of the function,
> I think merge conflict will be avoided:

Yeah, but I don't think it's worth messing with this just to avoid a conflict.

Thanks,
Miklos