RE: [PATCH net] hyperv: Fix vlan_proto setting innetvsc_recv_callback()

From: Haiyang Zhang
Date: Fri Jun 14 2013 - 12:25:24 EST




> -----Original Message-----
> From: Olaf Hering [mailto:olaf@xxxxxxxxx]
> Sent: Friday, June 14, 2013 11:29 AM
> To: Haiyang Zhang
> Cc: davem@xxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx; KY Srinivasan;
> jasowang@xxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> devel@xxxxxxxxxxxxxxxxxxxxxx
> Subject: Re: [PATCH net] hyperv: Fix vlan_proto setting in
> netvsc_recv_callback()
>
> On Tue, May 28, Haiyang Zhang wrote:
>
> > Since the recent addition of 8021AD, we need to set the new field
> > vlan_proto in sk_buff. Otherwise, it will trigger BUG() call in
> vlan_proto_idx().
>
> > @@ -284,7 +285,7 @@ int netvsc_recv_callback(struct hv_device
> > *device_obj,
> >
> > skb->protocol = eth_type_trans(skb, net);
> > skb->ip_summed = CHECKSUM_NONE;
> > - skb->vlan_tci = packet->vlan_tci;
> > + __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), packet-
> >vlan_tci);
>
>
> Just curious:
>
> Doesnt that change behaviour in the sense that __vlan_hwaccel_put_tag()
> now always adds VLAN_TAG_PRESENT to skb->vlan_tci, while before that
> change packet->vlan_tci may or may not had this flag?
>
> In other words, should a variant of this patch be backported if it was a bug
> not not use __vlan_hwaccel_put_tag right from the start?
>

Thanks for looking at this. Actually, the previous code was correct, because I
add the VLAN_TAG_PRESENT if the packet comes from vlan in function
rndis_filter_receive_data():
if (vlan) {
pkt->vlan_tci = VLAN_TAG_PRESENT | vlan->vlanid |
(vlan->pri << VLAN_PRIO_SHIFT);
} else {
pkt->vlan_tci = 0;
}

But, the new code should be updated below, so the VLAN_TAG_PRESENT won't be set
when it's not from a vlan:
If (VLAN_TAG_PRESENT & packet->vlan_tci)
__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), packet- vlan_tci);

I will submit a patch soon.

Thanks,
- Haiyang

¢éì®&Þ~º&¶¬–+-±éÝ¥Šw®žË±Êâmébžìdz¹Þ)í…æèw*jg¬±¨¶‰šŽŠÝj/êäz¹ÞŠà2ŠÞ¨è­Ú&¢)ß«a¶Úþø®G«éh®æj:+v‰¨Šwè†Ù>Wš±êÞiÛaxPjØm¶Ÿÿà -»+ƒùdš_