Date:	Fri, 23 Oct 2015 10:30:24 -0700
From:	Pravin Shelar <pshelar@...ira.com>
To:	Jiri Benc <jbenc@...hat.com>
Cc:	netdev <netdev@...r.kernel.org>
Subject: Re: [PATCH net v3] openvswitch: Fix egress tunnel info.

On Fri, Oct 23, 2015 at 5:17 AM, Jiri Benc <jbenc@...hat.com> wrote:
> On Thu, 22 Oct 2015 18:17:16 -0700, Pravin B Shelar wrote:
>> While transitioning to netdev based vport we broke OVS
>> feature which allows user to retrieve tunnel packet egress
>> information for lwtunnel devices.  Following patch fixes it
>> by introducing ndo operation to get the tunnel egress info.
>> Same ndo operation can be used for lwtunnel devices and compat
>> ovs-tnl-vport devices. So after adding such device operation
>> we can remove similar operation from ovs-vport.
>>
>> Fixes: 614732eaa12d ("openvswitch: Use regular VXLAN net_device device").
>> Signed-off-by: Pravin B Shelar <pshelar@...ira.com>
>> --
>> v2-v3:
>> - Remove unused tun_info
>> v1-v2:
>> - changed ndo operation name to ndo_fill_metadata_dst()
>> - Fix geneve stats update
>
> This looks good overall, thanks. I see some issues with the patch but
> most of it can be fixed in net-next.git. See below.
>
Thanks for the review.

> [...]
>> --- a/drivers/net/vxlan.c
>> +++ b/drivers/net/vxlan.c
>> @@ -2337,6 +2337,46 @@ static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
>>       return 0;
>>  }
>>
>> +static int egress_ipv4_tun_info(struct net_device *dev, struct sk_buff *skb,
>> +                             struct ip_tunnel_info *info,
>> +                             __be16 sport, __be16 dport)
>> +{
>> +     struct vxlan_dev *vxlan = netdev_priv(dev);
>> +     struct rtable *rt;
>> +     struct flowi4 fl4;
>> +
>> +     memset(&fl4, 0, sizeof(fl4));
>> +     fl4.flowi4_tos = RT_TOS(info->key.tos);
>> +     fl4.flowi4_mark = skb->mark;
>> +     fl4.flowi4_proto = IPPROTO_UDP;
>> +     fl4.daddr = info->key.u.ipv4.dst;
>> +
>> +     rt = ip_route_output_key(vxlan->net, &fl4);
>> +     if (IS_ERR(rt))
>> +             return PTR_ERR(rt);
>> +     ip_rt_put(rt);
>> +
>> +     info->key.u.ipv4.src = fl4.saddr;
>> +     info->key.tp_src = sport;
>> +     info->key.tp_dst = dport;
>> +     return 0;
>> +}
>
> Do you plan to address the introduced code duplication for net-next.git?
>
I see lot of refactoring scope for vxlan code even without this patch.
I am planing to address it in net-next.

>> +
>> +static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
>> +{
>> +     struct vxlan_dev *vxlan = netdev_priv(dev);
>> +     struct ip_tunnel_info *info = skb_tunnel_info(skb);
>> +     __be16 sport, dport;
>> +
>> +     sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
>> +                               vxlan->cfg.port_max, true);
>> +     dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
>> +
>> +     if (ip_tunnel_info_af(info) == AF_INET)
>> +             return egress_ipv4_tun_info(dev, skb, info, sport, dport);
>> +     return -EINVAL;
>
> What about IPv6? There's IPv6 support for metadata based vxlan in
> net.git, thus this should have IPv6 support, too. But then, this is
> currently used only by ovs which got the IPv6 support only in
> net-next.git, thus it may be enough to fix it there.
>
I did choose to implement only for ipv4 since it is pretty late for
fix so wanted to keep simple as possible. IPv6 support is not there
yet anyways.

> [...]
>> --- a/include/net/dst_metadata.h
>> +++ b/include/net/dst_metadata.h
> [...]
>> +static inline struct ip_tunnel_info *skb_tunnel_info_unclone(struct sk_buff *skb)
>> +{
>> +     struct metadata_dst *dst;
>> +
>> +     dst = tun_dst_unclone(skb);
>> +     if (IS_ERR(dst))
>> +             return NULL;
>> +
>> +     return &dst->u.tun_info;
>> +}
>
> This doesn't do what the name suggests and is, actually, ovs specific.
> The ip_tunnel_info can be provided as a part of lwtstate and this
> function should handle that case, too. This is not a problem for
> net.git, as the function just returns EINVAL in such case, but should
> be addressed for net-next.git. As ovs is currently the only user, I'd
> be also fine with just a comment stating that, so it's clear for future
> users of this function that it needs to be extended before it can be
> used out of ovs.
>

I considered lwstate, but I am reluctant to add this complexity
without a usecase.

> [...]
>> --- a/net/core/dev.c
>> +++ b/net/core/dev.c
>> @@ -99,6 +99,7 @@
>>  #include <linux/rtnetlink.h>
>>  #include <linux/stat.h>
>>  #include <net/dst.h>
>> +#include <net/dst_metadata.h>
>>  #include <net/pkt_sched.h>
>>  #include <net/checksum.h>
>>  #include <net/xfrm.h>
>> @@ -682,6 +683,32 @@ int dev_get_iflink(const struct net_device *dev)
>>  EXPORT_SYMBOL(dev_get_iflink);
>>
>>  /**
>> + *   dev_fill_metadata_dst - Retrieve tunnel egress information.
>> + *   @dev: targeted interface
>> + *   @skb: The packet.
>> + *
>> + *   For better visibility of tunnel traffic OVS needs to retrieve
>> + *   egress tunnel information for a packet. Following API allows
>> + *   user to get this info.
>> + */
>> +int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
>> +{
>> +     struct ip_tunnel_info *info;
>> +
>> +     if (!dev->netdev_ops  || !dev->netdev_ops->ndo_fill_metadata_dst)
>> +             return -EINVAL;
>> +
>> +     info = skb_tunnel_info_unclone(skb);
>> +     if (!info)
>> +             return -ENOMEM;
>
> ENOMEM is a wrong error code to return. skb_tunnel_info_unclone should
> return the error code returned by tun_dst_unclone, in particular the
> EINVAL case which will be much more common than the ENOMEM case.
>
I agree in general it is true, but this is only called in OVS case. In
that context ENOMEM is common error case than any other case. But I
see your point, I will send patch to fix the return code.


>> +     if (unlikely(!(info->mode & IP_TUNNEL_INFO_TX)))
>> +             return -EINVAL;
>
> It would be much better to check the mode before copying the metadata.
>
This is pretty rare case, Thats why I would rather keep the code
simple and not to call skb_tunnel_info() and then
skb_tunnel_info_unclone() to optimize this case.

> [...]
>> --- a/net/openvswitch/flow_netlink.c
>> +++ b/net/openvswitch/flow_netlink.c
> [...]
>> @@ -749,13 +749,12 @@ static int ipv4_tun_to_nlattr(struct sk_buff *skb,
>>       return 0;
>>  }
>>
>> -int ovs_nla_put_egress_tunnel_key(struct sk_buff *skb,
>> -                               const struct ip_tunnel_info *egress_tun_info,
>> -                               const void *egress_tun_opts)
>> +int ovs_nla_put_tunnel_info(struct sk_buff *skb,
>> +                         struct ip_tunnel_info *tun_info)
>>  {
>> -     return __ipv4_tun_to_nlattr(skb, &egress_tun_info->key,
>> -                                 egress_tun_opts,
>> -                                 egress_tun_info->options_len);
>> +     return __ipv4_tun_to_nlattr(skb, &tun_info->key,
>> +                                 ip_tunnel_info_opts(tun_info),
>> +                                 tun_info->options_len);
>>  }
>
> This should at least check whether the tun_info is indeed IPv4. Actual
> IPv6 support for this function can be added to net-next.git.
>
net tree only supports IPv4 tunnels. I am not sure value of this
check, specially since we need differ changes on net-next.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists