| Message ID | 20150417010216.GA22415@gondor.apana.org.au |
|---|---|
| State | Superseded, archived |
| Delegated to: | David Miller |
| Headers | show |
On Fri, Apr 17, 2015 at 09:02:16AM +0800, Herbert Xu wrote: > @@ -105,7 +105,7 @@ static void ri_tasklet(unsigned long dev) > if (from & AT_EGRESS) { > dev_queue_xmit(skb); > } else if (from & AT_INGRESS) { > - skb_pull(skb, skb->dev->hard_header_len); > + skb_pull(skb, _dev->hard_header_len); ... > if (!(at & AT_EGRESS)) { > if (m->tcfm_ok_push) > - skb_push(skb2, skb2->dev->hard_header_len); > + skb_push(skb2, dev->hard_header_len); seems the cleaner fix will be to push skb->mac_len instead? -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, Apr 16, 2015 at 06:34:02PM -0700, Alexei Starovoitov wrote: > On Fri, Apr 17, 2015 at 09:02:16AM +0800, Herbert Xu wrote: > > @@ -105,7 +105,7 @@ static void ri_tasklet(unsigned long dev) > > if (from & AT_EGRESS) { > > dev_queue_xmit(skb); > > } else if (from & AT_INGRESS) { > > - skb_pull(skb, skb->dev->hard_header_len); > > + skb_pull(skb, _dev->hard_header_len); > ... > > if (!(at & AT_EGRESS)) { > > if (m->tcfm_ok_push) > > - skb_push(skb2, skb2->dev->hard_header_len); > > + skb_push(skb2, dev->hard_header_len); > > seems the cleaner fix will be to push skb->mac_len instead? No skb->mac_len is the same as skb2->dev->hard_header_len. Cheers,
On Fri, Apr 17, 2015 at 10:15:01AM +0800, Herbert Xu wrote: > > seems the cleaner fix will be to push skb->mac_len instead? > > No skb->mac_len is the same as skb2->dev->hard_header_len. hmm. please help me understand the problem then. In the commit log you mentioned that your vlan dev and ifb have unequal hard header length. I think that can only happen if your master dev used to create vlan, didn't have vlan offload, so vlandev->hhl = 18 and ifb->hhl = 14. Then when tagged packet arrives on master device we call skb_vlan_untag() and skb->mac_len becomes 14, then vlan_do_receive() will do another_round, ingress qdisc will trigger and act_mirred eventually will be called. So existing act_mirred will be pushing 18 bytes, whereas only 14 bytes are valid and the lowest 4 have junk. By 'fixing it' using ifb's 14 byte the patch is only masking the problem, no? -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/net/ifb.c b/drivers/net/ifb.c index 34f846b..1256d26 100644 --- a/drivers/net/ifb.c +++ b/drivers/net/ifb.c @@ -105,7 +105,7 @@ static void ri_tasklet(unsigned long dev) if (from & AT_EGRESS) { dev_queue_xmit(skb); } else if (from & AT_INGRESS) { - skb_pull(skb, skb->dev->hard_header_len); + skb_pull(skb, _dev->hard_header_len); netif_receive_skb(skb); } else BUG(); diff --git a/net/sched/act_mirred.c b/net/sched/act_mirred.c index 5953517..44e4d50 100644 --- a/net/sched/act_mirred.c +++ b/net/sched/act_mirred.c @@ -151,13 +151,25 @@ static int tcf_mirred(struct sk_buff *skb, const struct tc_action *a, } at = G_TC_AT(skb->tc_verd); + + if (!(at & AT_EGRESS) && m->tcfm_ok_push && + (skb_headroom(skb) < dev->hard_header_len || + skb->dev->hard_header_len < dev->hard_header_len)) { + net_notice_ratelimited( + "tc mirred: Insufficient headroom from %s to %s: " + "%d %d\n", + skb->dev->name, dev->name, + skb_headroom(skb), skb->dev->hard_header_len); + goto out; + } + skb2 = skb_act_clone(skb, GFP_ATOMIC, m->tcf_action); if (skb2 == NULL) goto out; if (!(at & AT_EGRESS)) { if (m->tcfm_ok_push) - skb_push(skb2, skb2->dev->hard_header_len); + skb_push(skb2, dev->hard_header_len); } /* mirror is always swallowed */
When you redirect a VLAN device to an ifb device, you end up with crap within the ifb device because they have unequal hard header lengths and the redirected packet contains four extra bytes at the start which then gets interpreted as part of the MAC address. This patch fixes this by only pushing on the hard header length of ifb. This assumes that the original packet actually has enough header for that so checks have been added to that effect. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>