diff mbox

[ovs-dev,V11,18/33] netdev-tc-offloads: Implement netdev flow put using tc interface

Message ID 1497366235-20892-19-git-send-email-roid@mellanox.com
State Accepted
Headers show

Commit Message

Roi Dayan June 13, 2017, 3:03 p.m. UTC
From: Paul Blakey <paulb@mellanox.com>

Currently only tunnel offload is supported.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>
---
 lib/netdev-tc-offloads.c | 403 +++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 393 insertions(+), 10 deletions(-)

Comments

Simon Horman June 15, 2017, 10:13 a.m. UTC | #1
On Tue, Jun 13, 2017 at 06:03:40PM +0300, Roi Dayan wrote:
> From: Paul Blakey <paulb@mellanox.com>
> 
> Currently only tunnel offload is supported.
> 
> Signed-off-by: Paul Blakey <paulb@mellanox.com>
> Reviewed-by: Roi Dayan <roid@mellanox.com>
> Reviewed-by: Simon Horman <simon.horman@netronome.com>
> Acked-by: Flavio Leitner <fbl@sysclose.org>

I have applied this to master with the following trivial modification.

> +static int
> +test_key_and_mask(struct match *match) {

I changed the above to:

static int
test_key_and_mask(struct match *match)
{
diff mbox

Patch

diff --git a/lib/netdev-tc-offloads.c b/lib/netdev-tc-offloads.c
index a9e35e0..8dbccb1 100644
--- a/lib/netdev-tc-offloads.c
+++ b/lib/netdev-tc-offloads.c
@@ -89,7 +89,7 @@  del_ufid_tc_mapping(const ovs_u128 *ufid)
 
 /* Add ufid entry to ufid_tc hashmap.
  * If entry exists already it will be replaced. */
-static void OVS_UNUSED
+static void
 add_ufid_tc_mapping(const ovs_u128 *ufid, int prio, int handle,
                     struct netdev *netdev, int ifindex)
 {
@@ -120,7 +120,7 @@  add_ufid_tc_mapping(const ovs_u128 *ufid, int prio, int handle,
  * Returns handle if successful and fill prio and netdev for that ufid.
  * Otherwise returns 0.
  */
-static int OVS_UNUSED
+static int
 get_ufid_tc_mapping(const ovs_u128 *ufid, int *prio, struct netdev **netdev)
 {
     size_t ufid_hash = hash_bytes(ufid, sizeof *ufid, 0);
@@ -183,7 +183,7 @@  struct prio_map_data {
  *
  * Return prio on success or 0 if we are out of prios.
  */
-static uint16_t OVS_UNUSED
+static uint16_t
 get_prio_for_tc_flower(struct tc_flower *flower)
 {
     static struct hmap prios = HMAP_INITIALIZER(&prios);
@@ -441,16 +441,399 @@  netdev_tc_flow_dump_next(struct netdev_flow_dump *dump,
     return false;
 }
 
+static int
+parse_put_flow_set_action(struct tc_flower *flower, const struct nlattr *set,
+                          size_t set_len)
+{
+    static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
+    const struct nlattr *set_attr;
+    size_t set_left;
+
+    NL_ATTR_FOR_EACH_UNSAFE(set_attr, set_left, set, set_len) {
+        if (nl_attr_type(set_attr) == OVS_KEY_ATTR_TUNNEL) {
+            const struct nlattr *tunnel = nl_attr_get(set_attr);
+            const size_t tunnel_len = nl_attr_get_size(set_attr);
+            const struct nlattr *tun_attr;
+            size_t tun_left;
+
+            flower->set.set = true;
+            NL_ATTR_FOR_EACH_UNSAFE(tun_attr, tun_left, tunnel, tunnel_len) {
+                switch (nl_attr_type(tun_attr)) {
+                case OVS_TUNNEL_KEY_ATTR_ID: {
+                    flower->set.id = nl_attr_get_be64(tun_attr);
+                }
+                break;
+                case OVS_TUNNEL_KEY_ATTR_IPV4_SRC: {
+                    flower->set.ipv4.ipv4_src = nl_attr_get_be32(tun_attr);
+                }
+                break;
+                case OVS_TUNNEL_KEY_ATTR_IPV4_DST: {
+                    flower->set.ipv4.ipv4_dst = nl_attr_get_be32(tun_attr);
+                }
+                break;
+                case OVS_TUNNEL_KEY_ATTR_IPV6_SRC: {
+                    flower->set.ipv6.ipv6_src =
+                        nl_attr_get_in6_addr(tun_attr);
+                }
+                break;
+                case OVS_TUNNEL_KEY_ATTR_IPV6_DST: {
+                    flower->set.ipv6.ipv6_dst =
+                        nl_attr_get_in6_addr(tun_attr);
+                }
+                break;
+                case OVS_TUNNEL_KEY_ATTR_TP_SRC: {
+                    flower->set.tp_src = nl_attr_get_be16(tun_attr);
+                }
+                break;
+                case OVS_TUNNEL_KEY_ATTR_TP_DST: {
+                    flower->set.tp_dst = nl_attr_get_be16(tun_attr);
+                }
+                break;
+                }
+            }
+        } else {
+            VLOG_DBG_RL(&rl, "unsupported set action type: %d",
+                        nl_attr_type(set_attr));
+            return EOPNOTSUPP;
+        }
+    }
+    return 0;
+}
+
+static int
+test_key_and_mask(struct match *match) {
+    static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
+    const struct flow *key = &match->flow;
+    struct flow *mask = &match->wc.masks;
+
+    if (mask->pkt_mark) {
+        VLOG_DBG_RL(&rl, "offloading attribute pkt_mark isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->recirc_id && key->recirc_id) {
+        VLOG_DBG_RL(&rl, "offloading attribute recirc_id isn't supported");
+        return EOPNOTSUPP;
+    }
+    mask->recirc_id = 0;
+
+    if (mask->dp_hash) {
+        VLOG_DBG_RL(&rl, "offloading attribute dp_hash isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->conj_id) {
+        VLOG_DBG_RL(&rl, "offloading attribute conj_id isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->skb_priority) {
+        VLOG_DBG_RL(&rl, "offloading attribute skb_priority isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->actset_output) {
+        VLOG_DBG_RL(&rl,
+                    "offloading attribute actset_output isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->ct_state) {
+        VLOG_DBG_RL(&rl, "offloading attribute ct_state isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->ct_zone) {
+        VLOG_DBG_RL(&rl, "offloading attribute ct_zone isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->ct_mark) {
+        VLOG_DBG_RL(&rl, "offloading attribute ct_mark isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->packet_type && key->packet_type) {
+        VLOG_DBG_RL(&rl, "offloading attribute packet_type isn't supported");
+        return EOPNOTSUPP;
+    }
+    mask->packet_type = 0;
+
+    if (!ovs_u128_is_zero(mask->ct_label)) {
+        VLOG_DBG_RL(&rl, "offloading attribute ct_label isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    for (int i = 0; i < FLOW_N_REGS; i++) {
+        if (mask->regs[i]) {
+            VLOG_DBG_RL(&rl,
+                        "offloading attribute regs[%d] isn't supported", i);
+            return EOPNOTSUPP;
+        }
+    }
+
+    if (mask->metadata) {
+        VLOG_DBG_RL(&rl, "offloading attribute metadata isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->nw_tos) {
+        VLOG_DBG_RL(&rl, "offloading attribute nw_tos isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->nw_ttl) {
+        VLOG_DBG_RL(&rl, "offloading attribute nw_ttl isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    if (mask->nw_frag) {
+        VLOG_DBG_RL(&rl, "offloading attribute nw_frag isn't supported");
+        return EOPNOTSUPP;
+    }
+
+    for (int i = 0; i < FLOW_MAX_MPLS_LABELS; i++) {
+        if (mask->mpls_lse[i]) {
+            VLOG_DBG_RL(&rl, "offloading attribute mpls_lse isn't supported");
+            return EOPNOTSUPP;
+        }
+    }
+
+    if (key->dl_type == htons(ETH_TYPE_IP) &&
+        key->nw_proto == IPPROTO_ICMP) {
+        if (mask->tp_src) {
+            VLOG_DBG_RL(&rl,
+                        "offloading attribute icmp_type isn't supported");
+            return EOPNOTSUPP;
+        }
+        if (mask->tp_dst) {
+            VLOG_DBG_RL(&rl,
+                        "offloading attribute icmp_code isn't supported");
+            return EOPNOTSUPP;
+        }
+    } else if (key->dl_type == htons(ETH_TYPE_IP) &&
+               key->nw_proto == IPPROTO_IGMP) {
+        if (mask->tp_src) {
+            VLOG_DBG_RL(&rl,
+                        "offloading attribute igmp_type isn't supported");
+            return EOPNOTSUPP;
+        }
+        if (mask->tp_dst) {
+            VLOG_DBG_RL(&rl,
+                        "offloading attribute igmp_code isn't supported");
+            return EOPNOTSUPP;
+        }
+    } else if (key->dl_type == htons(ETH_TYPE_IPV6) &&
+               key->nw_proto == IPPROTO_ICMPV6) {
+        if (mask->tp_src) {
+            VLOG_DBG_RL(&rl,
+                        "offloading attribute icmp_type isn't supported");
+            return EOPNOTSUPP;
+        }
+        if (mask->tp_dst) {
+            VLOG_DBG_RL(&rl,
+                        "offloading attribute icmp_code isn't supported");
+            return EOPNOTSUPP;
+        }
+    }
+    if (is_ip_any(key) && key->nw_proto == IPPROTO_TCP && mask->tcp_flags) {
+        if (mask->tcp_flags) {
+            VLOG_DBG_RL(&rl,
+                        "offloading attribute tcp_flags isn't supported");
+            return EOPNOTSUPP;
+        }
+    }
+
+    if (!is_all_zeros(mask, sizeof *mask)) {
+        VLOG_DBG_RL(&rl, "offloading isn't supported, unknown attribute");
+        return EOPNOTSUPP;
+    }
+
+    return 0;
+}
+
 int
-netdev_tc_flow_put(struct netdev *netdev OVS_UNUSED,
-                   struct match *match OVS_UNUSED,
-                   struct nlattr *actions OVS_UNUSED,
-                   size_t actions_len OVS_UNUSED,
-                   const ovs_u128 *ufid OVS_UNUSED,
-                   struct offload_info *info OVS_UNUSED,
+netdev_tc_flow_put(struct netdev *netdev, struct match *match,
+                   struct nlattr *actions, size_t actions_len,
+                   const ovs_u128 *ufid, struct offload_info *info,
                    struct dpif_flow_stats *stats OVS_UNUSED)
 {
-    return EOPNOTSUPP;
+    static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
+    struct tc_flower flower;
+    const struct flow *key = &match->flow;
+    struct flow *mask = &match->wc.masks;
+    const struct flow_tnl *tnl = &match->flow.tunnel;
+    struct nlattr *nla;
+    size_t left;
+    int prio = 0;
+    int handle;
+    int ifindex;
+    int err;
+
+    ifindex = netdev_get_ifindex(netdev);
+    if (ifindex < 0) {
+        VLOG_ERR_RL(&error_rl, "failed to get ifindex for %s: %s",
+                    netdev_get_name(netdev), ovs_strerror(-ifindex));
+        return -ifindex;
+    }
+
+    memset(&flower, 0, sizeof flower);
+
+    if (tnl->tun_id) {
+        VLOG_DBG_RL(&rl,
+                    "tunnel: id %#" PRIx64 " src " IP_FMT
+                    " dst " IP_FMT " tp_src %d tp_dst %d",
+                    ntohll(tnl->tun_id),
+                    IP_ARGS(tnl->ip_src), IP_ARGS(tnl->ip_dst),
+                    ntohs(tnl->tp_src), ntohs(tnl->tp_dst));
+        flower.tunnel.id = tnl->tun_id;
+        flower.tunnel.ipv4.ipv4_src = tnl->ip_src;
+        flower.tunnel.ipv4.ipv4_dst = tnl->ip_dst;
+        flower.tunnel.ipv6.ipv6_src = tnl->ipv6_src;
+        flower.tunnel.ipv6.ipv6_dst = tnl->ipv6_dst;
+        flower.tunnel.tp_src = tnl->tp_src;
+        flower.tunnel.tp_dst = tnl->tp_dst;
+        flower.tunnel.tunnel = true;
+
+        memset(&mask->tunnel, 0, sizeof mask->tunnel);
+    }
+
+    flower.key.eth_type = key->dl_type;
+    flower.mask.eth_type = mask->dl_type;
+
+    if (mask->vlans[0].tci) {
+        ovs_be16 vid_mask = mask->vlans[0].tci & htons(VLAN_VID_MASK);
+        ovs_be16 pcp_mask = mask->vlans[0].tci & htons(VLAN_PCP_MASK);
+        ovs_be16 cfi = mask->vlans[0].tci & htons(VLAN_CFI);
+
+        if (cfi && key->vlans[0].tci & htons(VLAN_CFI)
+            && (!vid_mask || vid_mask == htons(VLAN_VID_MASK))
+            && (!pcp_mask || pcp_mask == htons(VLAN_PCP_MASK))
+            && (vid_mask || pcp_mask)) {
+            if (vid_mask) {
+                flower.key.vlan_id = vlan_tci_to_vid(key->vlans[0].tci);
+                VLOG_DBG_RL(&rl, "vlan_id: %d\n", flower.key.vlan_id);
+            }
+            if (pcp_mask) {
+                flower.key.vlan_prio = vlan_tci_to_pcp(key->vlans[0].tci);
+                VLOG_DBG_RL(&rl, "vlan_prio: %d\n", flower.key.vlan_prio);
+            }
+            flower.key.encap_eth_type = flower.key.eth_type;
+            flower.key.eth_type = htons(ETH_TYPE_VLAN);
+        } else if (mask->vlans[0].tci == htons(0xffff) &&
+                   ntohs(key->vlans[0].tci) == 0) {
+            /* exact && no vlan */
+        } else {
+            /* partial mask */
+            return EOPNOTSUPP;
+        }
+    } else if (mask->vlans[1].tci) {
+        return EOPNOTSUPP;
+    }
+    memset(mask->vlans, 0, sizeof mask->vlans);
+
+    flower.key.dst_mac = key->dl_dst;
+    flower.mask.dst_mac = mask->dl_dst;
+    flower.key.src_mac = key->dl_src;
+    flower.mask.src_mac = mask->dl_src;
+    memset(&mask->dl_dst, 0, sizeof mask->dl_dst);
+    memset(&mask->dl_src, 0, sizeof mask->dl_src);
+    mask->dl_type = 0;
+    mask->in_port.odp_port = 0;
+
+    if (is_ip_any(key)) {
+        flower.key.ip_proto = key->nw_proto;
+        flower.mask.ip_proto = mask->nw_proto;
+
+        if (key->nw_proto == IPPROTO_TCP || key->nw_proto == IPPROTO_UDP) {
+            flower.key.dst_port = key->tp_dst;
+            flower.mask.dst_port = mask->tp_dst;
+            flower.key.src_port = key->tp_src;
+            flower.mask.src_port = mask->tp_src;
+            mask->tp_src = 0;
+            mask->tp_dst = 0;
+        }
+
+        mask->nw_frag = 0;
+        mask->nw_tos = 0;
+        mask->nw_proto = 0;
+
+        if (key->dl_type == htons(ETH_P_IP)) {
+            flower.key.ipv4.ipv4_src = key->nw_src;
+            flower.mask.ipv4.ipv4_src = mask->nw_src;
+            flower.key.ipv4.ipv4_dst = key->nw_dst;
+            flower.mask.ipv4.ipv4_dst = mask->nw_dst;
+            mask->nw_src = 0;
+            mask->nw_dst = 0;
+        } else if (key->dl_type == htons(ETH_P_IPV6)) {
+            flower.key.ipv6.ipv6_src = key->ipv6_src;
+            flower.mask.ipv6.ipv6_src = mask->ipv6_src;
+            flower.key.ipv6.ipv6_dst = key->ipv6_dst;
+            flower.mask.ipv6.ipv6_dst = mask->ipv6_dst;
+            memset(&mask->ipv6_src, 0, sizeof mask->ipv6_src);
+            memset(&mask->ipv6_dst, 0, sizeof mask->ipv6_dst);
+        }
+    }
+
+    err = test_key_and_mask(match);
+    if (err) {
+        return err;
+    }
+
+    NL_ATTR_FOR_EACH(nla, left, actions, actions_len) {
+        if (nl_attr_type(nla) == OVS_ACTION_ATTR_OUTPUT) {
+            odp_port_t port = nl_attr_get_odp_port(nla);
+            struct netdev *outdev = netdev_ports_get(port,
+                                                     info->port_hmap_obj);
+
+            flower.ifindex_out = netdev_get_ifindex(outdev);
+            flower.set.tp_dst = info->tp_dst_port;
+            netdev_close(outdev);
+        } else if (nl_attr_type(nla) == OVS_ACTION_ATTR_PUSH_VLAN) {
+            const struct ovs_action_push_vlan *vlan_push = nl_attr_get(nla);
+
+            flower.vlan_push_id = vlan_tci_to_vid(vlan_push->vlan_tci);
+            flower.vlan_push_prio = vlan_tci_to_pcp(vlan_push->vlan_tci);
+        } else if (nl_attr_type(nla) == OVS_ACTION_ATTR_POP_VLAN) {
+            flower.vlan_pop = 1;
+        } else if (nl_attr_type(nla) == OVS_ACTION_ATTR_SET) {
+            const struct nlattr *set = nl_attr_get(nla);
+            const size_t set_len = nl_attr_get_size(nla);
+
+            err = parse_put_flow_set_action(&flower, set, set_len);
+            if (err) {
+                return err;
+            }
+        } else {
+            VLOG_DBG_RL(&rl, "unsupported put action type: %d",
+                        nl_attr_type(nla));
+            return EOPNOTSUPP;
+        }
+    }
+
+    handle = get_ufid_tc_mapping(ufid, &prio, NULL);
+    if (handle && prio) {
+        VLOG_DBG_RL(&rl, "updating old handle: %d prio: %d", handle, prio);
+        tc_del_filter(ifindex, prio, handle);
+    }
+
+    if (!prio) {
+        prio = get_prio_for_tc_flower(&flower);
+        if (prio == 0) {
+            VLOG_ERR_RL(&rl, "couldn't get tc prio: %s", ovs_strerror(ENOSPC));
+            return ENOSPC;
+        }
+    }
+
+    flower.act_cookie.data = ufid;
+    flower.act_cookie.len = sizeof *ufid;
+
+    err = tc_replace_flower(ifindex, prio, handle, &flower);
+    if (!err) {
+        add_ufid_tc_mapping(ufid, flower.prio, flower.handle, netdev, ifindex);
+    }
+
+    return err;
 }
 
 int