netdev - Re: [PATCH net V2] tuntap: synchronize through tfiles array instead of tun->numqueues

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAM_iQpVGdduQGdkBn2a+8=VTuZcoTxBdve6+uDHACcDrdtL=Og@mail.gmail.com>
Date:   Tue, 7 May 2019 07:41:32 -0700
From:   Cong Wang <xiyou.wangcong@...il.com>
To:     Jason Wang <jasowang@...hat.com>
Cc:     Linux Kernel Network Developers <netdev@...r.kernel.org>,
        "Michael S. Tsirkin" <mst@...hat.com>,
        YueHaibing <yuehaibing@...wei.com>,
        "weiyongjun (A)" <weiyongjun1@...wei.com>,
        Eric Dumazet <eric.dumazet@...il.com>
Subject: Re: [PATCH net V2] tuntap: synchronize through tfiles array instead
 of tun->numqueues

On Mon, May 6, 2019 at 11:19 PM Jason Wang <jasowang@...hat.com> wrote:
>
>
> On 2019/5/7 下午12:54, Cong Wang wrote:
> > On Mon, May 6, 2019 at 9:03 PM Jason Wang <jasowang@...hat.com> wrote:
> >> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> >> index e9ca1c0..32a0b23 100644
> >> --- a/drivers/net/tun.c
> >> +++ b/drivers/net/tun.c
> >> @@ -700,6 +700,8 @@ static void __tun_detach(struct tun_file *tfile, bool clean)
> >>                                     tun->tfiles[tun->numqueues - 1]);
> >>                  ntfile = rtnl_dereference(tun->tfiles[index]);
> >>                  ntfile->queue_index = index;
> >> +               rcu_assign_pointer(tun->tfiles[tun->numqueues - 1],
> >> +                                  NULL);
> >>
> > How does this work? Existing readers could still read this
> > tun->tfiles[tun->numqueues - 1] before you NULL it. And,
> > _if_ the following sock_put() is the one frees it, you still miss
> > a RCU grace period.
> >
> >                  if (clean) {
> >                          RCU_INIT_POINTER(tfile->tun, NULL);
> >                          sock_put(&tfile->sk);
> >
> >
> > Thanks.
>
>
> My understanding is the socket will never be freed for this sock_put().
> We just drop an extra reference count we held when the socket was
> attached to the netdevice (there's a sock_hold() in tun_attach()). The
> real free should happen at another sock_put() in the end of this function.

So you are saying readers will never read this sock after free, then
what are you fixing with this patch? Nothing, right?

As I said, reading a stale tun->numqueues is fine, you just keep
believing it is a problem.