On Tue, May 8, 2018 at 11:59 PM, Jason Wang <jasow...@redhat.com> wrote:
> We used to initialize ptr_ring during TUNSETIFF, this is because its
> size depends on the tx_queue_len of netdevice. And we try to clean it
> up when socket were detached from netdevice. A race were spotted when
> trying to do uninit during a read which will lead a use after free for
> pointer ring. Solving this by always initialize a zero size ptr_ring
> in open() and do resizing during TUNSETIFF, and then we can safely do
> cleanup during close(). With this, there's no need for the workaround
> that was introduced by commit 4df0bfc79904 ("tun: fix a memory leak
> for tfile->tx_array").
>
Ah, I didn't know ptr_ring_init(0) could work... Nice patch!
Except one thing below.
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index ef33950..298cb96 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -681,15 +681,6 @@ static void tun_queue_purge(struct tun_file *tfile)
> skb_queue_purge(&tfile->sk.sk_error_queue);
> }
>
> -static void tun_cleanup_tx_ring(struct tun_file *tfile)
> -{
> - if (tfile->tx_ring.queue) {
> - ptr_ring_cleanup(&tfile->tx_ring, tun_ptr_free);
> - xdp_rxq_info_unreg(&tfile->xdp_rxq);
> - memset(&tfile->tx_ring, 0, sizeof(tfile->tx_ring));
> - }
> -}
I don't think you can totally remove ptr_ring_cleanup(), it should be
called unconditionally with your ptr_ring_init(0) trick, right?
