On Sun, Oct 20, 2019 at 7:12 AM Martin Varghese
<martinvargheseno...@gmail.com> wrote:
>
> From: Martin Varghese <martin.vargh...@nokia.com>
>
> The openvswitch was supporting a MPLS label depth of 1 in the ingress
> direction though the userspace OVS supports a max depth of 3 labels.
> This change enables openvswitch module to support a max depth of
> 3 labels in the ingress.
>
> Signed-off-by: Martin Varghese <martin.vargh...@nokia.com>
> ---
> Changes in v2
> - Moved MPLS count validation from datapath to configuration.
> - Fixed set mpls function.
>
This patch looks pretty close now.
> net/openvswitch/actions.c | 2 +-
> net/openvswitch/flow.c | 20 ++++++++++-----
> net/openvswitch/flow.h | 9 ++++---
> net/openvswitch/flow_netlink.c | 57
> +++++++++++++++++++++++++++++++++---------
> 4 files changed, 66 insertions(+), 22 deletions(-)
>
...
> diff --git a/net/openvswitch/flow_netlink.c b/net/openvswitch/flow_netlink.c
> index d7559c6..21de061 100644
> --- a/net/openvswitch/flow_netlink.c
> +++ b/net/openvswitch/flow_netlink.c
> @@ -424,7 +424,7 @@ size_t ovs_key_attr_size(void)
> [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
> [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
> .next = ovs_tunnel_key_lens, },
> - [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
> + [OVS_KEY_ATTR_MPLS] = { .len = OVS_ATTR_VARIABLE },
> [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
> [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
> [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
> @@ -1628,10 +1628,25 @@ static int ovs_key_from_nlattrs(struct net *net,
> struct sw_flow_match *match,
>
> if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
> const struct ovs_key_mpls *mpls_key;
> + u32 hdr_len;
> + u32 label_count, label_count_mask, i;
>
> mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
> - SW_FLOW_KEY_PUT(match, mpls.top_lse,
> - mpls_key->mpls_lse, is_mask);
> + hdr_len = nla_len(a[OVS_KEY_ATTR_MPLS]);
> + label_count = hdr_len / sizeof(struct ovs_key_mpls);
> +
> + if (label_count == 0 || label_count > MPLS_LABEL_DEPTH ||
> + hdr_len % sizeof(struct ovs_key_mpls))
> + return -EINVAL;
> +
> + label_count_mask = GENMASK(label_count - 1, 0);
> +
> + for (i = 0 ; i < label_count; i++)
> + SW_FLOW_KEY_PUT(match, mpls.lse[i],
> + mpls_key[i].mpls_lse, is_mask);
> +
> + SW_FLOW_KEY_PUT(match, mpls.num_labels_mask,
> + label_count_mask, is_mask);
>
> attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
> }
> @@ -2114,13 +2129,18 @@ static int __ovs_nla_put_key(const struct sw_flow_key
> *swkey,
> ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
> ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
> } else if (eth_p_mpls(swkey->eth.type)) {
> + u8 i, num_labels;
> struct ovs_key_mpls *mpls_key;
>
> - nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
> + num_labels = hweight_long(output->mpls.num_labels_mask);
> + nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS,
> + num_labels * sizeof(*mpls_key));
> if (!nla)
> goto nla_put_failure;
> +
> mpls_key = nla_data(nla);
> - mpls_key->mpls_lse = output->mpls.top_lse;
> + for (i = 0; i < num_labels; i++)
> + mpls_key[i].mpls_lse = output->mpls.lse[i];
> }
>
> if ((swkey->eth.type == htons(ETH_P_IP) ||
> @@ -2957,6 +2977,10 @@ static int __ovs_nla_copy_actions(struct net *net,
> const struct nlattr *attr,
> u8 mac_proto = ovs_key_mac_proto(key);
> const struct nlattr *a;
> int rem, err;
> + u32 mpls_label_count = 0;
> +
> + if (eth_p_mpls(eth_type))
> + mpls_label_count = hweight_long(key->mpls.num_labels_mask);
>
The MPLS push and pop action could be part of nested actions in
sample, so the count needs to be global count across such nested
actions. have a look at validate_and_copy_sample().
> nla_for_each_nested(a, attr, rem) {
> /* Expected argument lengths, (u32)-1 for variable length. */
> @@ -3065,25 +3089,34 @@ static int __ovs_nla_copy_actions(struct net *net,
> const struct nlattr *attr,
> !eth_p_mpls(eth_type)))
> return -EINVAL;
> eth_type = mpls->mpls_ethertype;
> + mpls_label_count++;
> break;
> }
>
> - case OVS_ACTION_ATTR_POP_MPLS:
> + case OVS_ACTION_ATTR_POP_MPLS: {
> + __be16 proto;
> if (vlan_tci & htons(VLAN_CFI_MASK) ||
> !eth_p_mpls(eth_type))
> return -EINVAL;
>
> - /* Disallow subsequent L2.5+ set and mpls_pop actions
> - * as there is no check here to ensure that the new
> - * eth_type is valid and thus set actions could
> - * write off the end of the packet or otherwise
> - * corrupt it.
> + /* Disallow subsequent L2.5+ set actions as there is
> + * no check here to ensure that the new eth type is
> + * valid and thus set actions could write off the
> + * end of the packet or otherwise corrupt it.
> *
> * Support for these actions is planned using packet
> * recirculation.
> */
This comment needs updated.
> - eth_type = htons(0);
> + proto = nla_get_be16(a);
> + mpls_label_count--;
> +
> + if (!eth_p_mpls(proto) || !mpls_label_count)
> + eth_type = htons(0);
> + else
> + eth_type = proto;
> +
> break;
> + }
>
> case OVS_ACTION_ATTR_SET:
> err = validate_set(a, key, sfa,
> --
> 1.8.3.1
>
