Re: KASAN: use-after-free Read in __dev_queue_xmit

[Eric Biggers]

On Wed, Jan 03, 2018 at 10:53:14PM -0800, Eric Dumazet wrote:
> On Wed, 2018-01-03 at 21:13 -0800, Eric Dumazet wrote:
> > Note: all commands must start from beginning of the line in the email body.
> > 
> > I guess skb_probe_transport_header() should be hardened to reject malicious
> > packets given by user space, instead of being gentle.
> 
> Although bug triggered for this particular repro is in flow dissector
> :/
> 
> I will test :
> 
> diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
> index 15ce300637650e17fcab7e378b20fe7972686d46..544bddf08e13c7f6e47aadc737244c9ba5af56b2 100644
> --- a/net/core/flow_dissector.c
> +++ b/net/core/flow_dissector.c
> @@ -976,8 +976,8 @@ bool __skb_flow_dissect(const struct sk_buff *skb,
>  out_good:
>         ret = true;
>  
> -       key_control->thoff = (u16)nhoff;
>  out:
> +       key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
>         key_basic->n_proto = proto;
>         key_basic->ip_proto = ip_proto;
>  
> @@ -985,7 +985,6 @@ bool __skb_flow_dissect(const struct sk_buff *skb,
>  
>  out_bad:
>         ret = false;
> -       key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
>         goto out;
>  }
>  EXPORT_SYMBOL(__skb_flow_dissect);

Fix for this was commit d0c081b49137cd:

#syz fix: flow_dissector: properly cap thoff field

But a crash with the same signature is still occurring, so it should eventually
get reported again.  C reproducer is here, it works on Linus' tree (commit
036db8bd963): https://syzkaller.appspot.com/text?tag=ReproC&x=105b1ae7800000

- Eric