On Wed, 11 Jul 2018 16:41:35 +0100
Edward Cree <ec...@solarflare.com> wrote:
> On 11/07/18 16:01, Jesper Dangaard Brouer wrote:
> > In commit 5fa12739a53d ("net: ipv4: listify ip_rcv_finish") calling
> > dst_input(skb) was split-out. The ip_sublist_rcv_finish() just calls
> > dst_input(skb) in a loop.
> > The problem is that ip_sublist_rcv_finish() forgot to remove the SKB
> > from the list before invoking dst_input(). Further more we need to
> > clear skb->next as other parts of the network stack use another kind
> > of SKB lists for xmit_more (see dev_hard_start_xmit).
> > A crash occurs if e.g. dst_input() invoke ip_forward(), which calls
> > dst_output()/ip_output() that eventually calls __dev_queue_xmit() +
> > sch_direct_xmit(), and a crash occurs in validate_xmit_skb_list().
> > This patch only fixes the crash, but there is a huge potential for
> > a performance boost if we can pass an SKB-list through to ip_forward.
> > Fixes: 5fa12739a53d ("net: ipv4: listify ip_rcv_finish")
> > Signed-off-by: Jesper Dangaard Brouer <bro...@redhat.com>
> Acked-by: Edward Cree <ec...@solarflare.com>
> But it feels weird and asymmetric to only NULL skb->next (not ->prev), and
> to have to do this by hand rather than e.g. being able to use
> list_del_init(&skb->list). Hopefully this can be revisited once
> sch_direct_xmit() has been changed to use the list_head rather than SKB
> special lists.
I cannot use list_del_init(&skb->list) it would also break.
This is a fix, and this code should be revisited.
The reason I used the list_del() + skb->next=NULL, combo, is to keep as
much as possible of the list-poisoning, e.g. 'prev' will be LIST_POISON2.
Jesper Dangaard Brouer
MSc.CS, Principal Kernel Engineer at Red Hat