net: bond: skip vlan header when do layer 3+4 hash policy

From: liujian <liujian56@huawei.com>

When the hash policy is BOND_XMIT_POLICY_LAYER34 mode and skb protocol 
is 802.1q VLAN, the policy will be degenerated to LAYER2 mode; Now, 
change it to get the next layer protocol to ensure that it worked in 
BOND_XMIT_POLICY_LAYER34 mode.

Signed-off-by: liujian <liujian56@huawei.com>
---
 drivers/net/bonding/bond_main.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

On 31/03/18 12:14, liujian56@huawei.com wrote:
> From: liujian <liujian56@huawei.com>
> 
> When the hash policy is BOND_XMIT_POLICY_LAYER34 mode and skb protocol
> is 802.1q VLAN, the policy will be degenerated to LAYER2 mode; Now,
> change it to get the next layer protocol to ensure that it worked in
> BOND_XMIT_POLICY_LAYER34 mode.
> 
> Signed-off-by: liujian <liujian56@huawei.com>
> ---
>   drivers/net/bonding/bond_main.c | 11 ++++++++---
>   1 file changed, 8 insertions(+), 3 deletions(-)


Nak
Use BOND_XMIT_POLICY_ENCAP34 (encap3+4), that was one of the main reasons it was added.

Hi liujian,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on net-next/master]
[also build test WARNING on v4.16-rc7 next-20180329]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/liujian56-huawei-com/net-bond-skip-vlan-header-when-do-layer-3-4-hash-policy/20180401-045327
reproduce:
        # apt-get install sparse
        make ARCH=x86_64 allmodconfig
        make C=1 CF=-D__CHECK_ENDIAN__


sparse warnings: (new ones prefixed by >>)

   drivers/net/bonding/bond_main.c:2343:26: sparse: restricted __be16 degrades to integer
   drivers/net/bonding/bond_main.c:2349:20: sparse: restricted __be16 degrades to integer
   drivers/net/bonding/bond_main.c:2396:40: sparse: incorrect type in assignment (different base types) @@    expected restricted __be16 [usertype] vlan_proto @@    got e] vlan_proto @@
   drivers/net/bonding/bond_main.c:2396:40:    expected restricted __be16 [usertype] vlan_proto
   drivers/net/bonding/bond_main.c:2396:40:    got int
>> drivers/net/bonding/bond_main.c:3217:18: sparse: incorrect type in assignment (different base types) @@    expected int [signed] skbproto @@    got restricted __be1int [signed] skbproto @@
   drivers/net/bonding/bond_main.c:3217:18:    expected int [signed] skbproto
   drivers/net/bonding/bond_main.c:3217:18:    got restricted __be16 [usertype] protocol
>> drivers/net/bonding/bond_main.c:3219:26: sparse: incorrect type in assignment (different base types) @@    expected int [signed] skbproto @@    got restint [signed] skbproto @@
   drivers/net/bonding/bond_main.c:3219:26:    expected int [signed] skbproto
   drivers/net/bonding/bond_main.c:3219:26:    got restricted __be16
   drivers/net/bonding/bond_main.c:3221:25: sparse: restricted __be16 degrades to integer
   drivers/net/bonding/bond_main.c:3229:32: sparse: restricted __be16 degrades to integer
   drivers/net/bonding/bond_main.c:3199:60: sparse: restricted __be16 degrades to integer
   drivers/net/bonding/bond_main.c:3199:60: sparse: restricted __be16 degrades to integer

vim +3217 drivers/net/bonding/bond_main.c

  3202	
  3203	/* Extract the appropriate headers based on bond's xmit policy */
  3204	static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb,
  3205				      struct flow_keys *fk)
  3206	{
  3207		const struct ipv6hdr *iph6;
  3208		const struct iphdr *iph;
  3209		int noff, proto = -1, skbproto;
  3210	
  3211		if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23)
  3212			return skb_flow_dissect_flow_keys(skb, fk, 0);
  3213	
  3214		fk->ports.ports = 0;
  3215		noff = skb_network_offset(skb);
  3216	
> 3217		skbproto = skb->protocol;
  3218		if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34)
> 3219			skbproto = vlan_get_protocol(skb);
  3220	
  3221		if (skbproto == htons(ETH_P_IP)) {
  3222			if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph))))
  3223				return false;
  3224			iph = ip_hdr(skb);
  3225			iph_to_flow_copy_v4addrs(fk, iph);
  3226			noff += iph->ihl << 2;
  3227			if (!ip_is_fragment(iph))
  3228				proto = iph->protocol;
  3229		} else if (skbproto == htons(ETH_P_IPV6)) {
  3230			if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6))))
  3231				return false;
  3232			iph6 = ipv6_hdr(skb);
  3233			iph_to_flow_copy_v6addrs(fk, iph6);
  3234			noff += sizeof(*iph6);
  3235			proto = iph6->nexthdr;
  3236		} else {
  3237			return false;
  3238		}
  3239		if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0)
  3240			fk->ports.ports = skb_flow_get_ports(skb, noff, proto);
  3241	
  3242		return true;
  3243	}
  3244	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

> -----Original Message-----
> From: Nikolay Aleksandrov [mailto:nikolay@cumulusnetworks.com]
> Sent: Saturday, March 31, 2018 5:23 PM
> To: liujian (CE) <liujian56@huawei.com>; davem@davemloft.net;
> j.vosburgh@gmail.com; vfalico@gmail.com; andy@greyhouse.net
> Cc: netdev@vger.kernel.org; weiyongjun (A) <weiyongjun1@huawei.com>
> Subject: Re: [PATCH] net: bond: skip vlan header when do layer 3+4 hash policy
> 
> On 31/03/18 12:14, liujian56@huawei.com wrote:
> > From: liujian <liujian56@huawei.com>
> >
> > When the hash policy is BOND_XMIT_POLICY_LAYER34 mode and skb
> protocol
> > is 802.1q VLAN, the policy will be degenerated to LAYER2 mode; Now,
> > change it to get the next layer protocol to ensure that it worked in
> > BOND_XMIT_POLICY_LAYER34 mode.
> >
> > Signed-off-by: liujian <liujian56@huawei.com>
> > ---
> >   drivers/net/bonding/bond_main.c | 11 ++++++++---
> >   1 file changed, 8 insertions(+), 3 deletions(-)
> 
> 
> Nak
> Use BOND_XMIT_POLICY_ENCAP34 (encap3+4), that was one of the main
> reasons it was added.

Got it, thank you~