[ovs-dev] [PATCH 3/7] vxlan-gpe-nsh: tun_nodecap mode in DPDk-netdev dataplane.
mengke
mengke.liu at intel.com
Wed Sep 30 09:47:20 UTC 2015
This patch covers VxLAN-GPE NSH tun_nodecap feature for NSH
decapsulation-reencapsulation case.
When VxLAN-GPE NSH packets are received and then resent to VxLAN-GPE NSH port.
The decapsulation and encapsulation will be implemented. However, tunnel pop
and tunnel push actions are very time-consuming when decapsulation and
encapsulation.
With this feature (options:tun_nodecap=true), tunnel port will parse the input
tunnel packets, but the tunnel header will be kept. And the tunnel header can
be modified by the set field actions. This feature can improve performance.
Signed-off-by: Ricky Li <ricky.li at intel.com>
Signed-off-by: Mengke Liu <mengke.liu at intel.com>
---
datapath/linux/compat/include/linux/openvswitch.h | 23 +++++
lib/dpif-netdev.c | 36 +++++++
lib/dpif.c | 1 +
lib/netdev-bsd.c | 1 +
lib/netdev-dpdk.c | 1 +
lib/netdev-dummy.c | 1 +
lib/netdev-linux.c | 1 +
lib/netdev-provider.h | 4 +
lib/netdev-vport.c | 111 +++++++++++++++++++---
lib/netdev.c | 24 +++++
lib/netdev.h | 7 +-
lib/odp-execute.c | 5 +
lib/odp-util.c | 96 ++++++++++++++++++-
lib/odp-util.h | 7 +-
lib/ofp-print.c | 1 +
lib/packets.c | 20 ++++
lib/packets.h | 6 ++
ofproto/ofproto-dpif-sflow.c | 4 +
ofproto/ofproto-dpif-xlate.c | 50 ++++++++--
ofproto/tunnel.c | 86 +++++++++++++++++
ofproto/tunnel.h | 4 +
tests/tunnel.at | 28 ++++++
22 files changed, 495 insertions(+), 22 deletions(-)
diff --git a/datapath/linux/compat/include/linux/openvswitch.h b/datapath/linux/compat/include/linux/openvswitch.h
index aa5dfde..b8ac152 100644
--- a/datapath/linux/compat/include/linux/openvswitch.h
+++ b/datapath/linux/compat/include/linux/openvswitch.h
@@ -645,6 +645,27 @@ struct ovs_action_push_tnl {
uint32_t tnl_type; /* For logging. */
uint8_t header[TNL_PUSH_HEADER_SIZE];
};
+
+#define OVS_POP_SPEC_ACTION_NO_DECAP 2
+
+/*
+ * struct ovs_action_pop_tnl - %OVS_ACTION_ATTR_TUNNEL_POP_SPEC
+ * @tnl_port: To identify tunnel port to pass header info.
+ * @out_port: Physical port to send encapsulated packet.
+ * @header_len: Length of the header to be pushed.
+ * @tnl_type: This is only required to format this header. Otherwise
+ * ODP layer can not parse %header.
+ * @header: Partial header for the tunnel. Specified pop action can use
+ * this header to build final header according to actual packet parameters.
+ */
+struct ovs_action_pop_tnl {
+ uint32_t tnl_port;
+ uint32_t out_port;
+ uint32_t header_len;
+ uint16_t tnl_type; /* For logging. */
+ uint16_t pop_type;
+ uint8_t header[TNL_PUSH_HEADER_SIZE];
+};
#endif
/**
@@ -712,6 +733,8 @@ enum ovs_action_attr {
#ifndef __KERNEL__
OVS_ACTION_ATTR_TUNNEL_PUSH, /* struct ovs_action_push_tnl*/
OVS_ACTION_ATTR_TUNNEL_POP, /* u32 port number. */
+ OVS_ACTION_ATTR_TUNNEL_POP_SPEC, /* struct ovs_action_pop_tnl */
+
#endif
__OVS_ACTION_ATTR_MAX, /* Nothing past this will be accepted
* from userspace. */
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index e6ba33f..07f05c4 100644
--- a/lib/dpif-netdev.c
+++ b/lib/dpif-netdev.c
@@ -3549,6 +3549,42 @@ dp_execute_cb(void *aux_, struct dp_packet **packets, int cnt,
}
break;
+ case OVS_ACTION_ATTR_TUNNEL_POP_SPEC:
+ if (*depth < MAX_RECIRC_DEPTH) {
+ const struct ovs_action_pop_tnl *data;
+ odp_port_t portno;
+
+ data = nl_attr_get(a);
+ portno = u32_to_odp(data->tnl_port);
+
+ p = dp_netdev_lookup_port(dp, portno);
+ if (p) {
+ struct dp_packet *tnl_pkt[NETDEV_MAX_BURST];
+ int err;
+
+ if (!may_steal) {
+ dp_netdev_clone_pkt_batch(tnl_pkt, packets, cnt);
+ packets = tnl_pkt;
+ }
+
+ err = netdev_pop_header_spec(p->netdev, packets, cnt, data);
+ if (!err) {
+
+ for (i = 0; i < cnt; i++) {
+ packets[i]->md.in_port.odp_port = portno;
+ }
+
+ (*depth)++;
+ dp_netdev_input(pmd, packets, cnt);
+ (*depth)--;
+ } else {
+ dp_netdev_drop_packets(tnl_pkt, cnt, !may_steal);
+ }
+ return;
+ }
+ }
+ break;
+
case OVS_ACTION_ATTR_USERSPACE:
if (!fat_rwlock_tryrdlock(&dp->upcall_rwlock)) {
const struct nlattr *userdata;
diff --git a/lib/dpif.c b/lib/dpif.c
index 9a67a24..bb2d519 100644
--- a/lib/dpif.c
+++ b/lib/dpif.c
@@ -1100,6 +1100,7 @@ dpif_execute_helper_cb(void *aux_, struct dp_packet **packets, int cnt,
case OVS_ACTION_ATTR_OUTPUT:
case OVS_ACTION_ATTR_TUNNEL_PUSH:
case OVS_ACTION_ATTR_TUNNEL_POP:
+ case OVS_ACTION_ATTR_TUNNEL_POP_SPEC:
case OVS_ACTION_ATTR_USERSPACE:
case OVS_ACTION_ATTR_RECIRC: {
struct dpif_execute execute;
diff --git a/lib/netdev-bsd.c b/lib/netdev-bsd.c
index 60e5615..001e888 100644
--- a/lib/netdev-bsd.c
+++ b/lib/netdev-bsd.c
@@ -1560,6 +1560,7 @@ netdev_bsd_update_flags(struct netdev *netdev_, enum netdev_flags off,
NULL, /* build header */ \
NULL, /* push header */ \
NULL, /* pop header */ \
+ NULL, /* pop specific header */ \
NULL, /* get_numa_id */ \
NULL, /* set_multiq */ \
\
diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
index b72a33b..4947013 100644
--- a/lib/netdev-dpdk.c
+++ b/lib/netdev-dpdk.c
@@ -2050,6 +2050,7 @@ unlock_dpdk:
NULL, /* build header */ \
NULL, /* push header */ \
NULL, /* pop header */ \
+ NULL, /* pop specific header */ \
netdev_dpdk_get_numa_id, /* get_numa_id */ \
MULTIQ, /* set_multiq */ \
\
diff --git a/lib/netdev-dummy.c b/lib/netdev-dummy.c
index 76815c2..2f44901 100644
--- a/lib/netdev-dummy.c
+++ b/lib/netdev-dummy.c
@@ -1098,6 +1098,7 @@ static const struct netdev_class dummy_class = {
NULL, /* build header */
NULL, /* push header */
NULL, /* pop header */
+ NULL, /* pop specific header */
NULL, /* get_numa_id */
NULL, /* set_multiq */
diff --git a/lib/netdev-linux.c b/lib/netdev-linux.c
index 584e804..79f47f3 100644
--- a/lib/netdev-linux.c
+++ b/lib/netdev-linux.c
@@ -2790,6 +2790,7 @@ netdev_linux_update_flags(struct netdev *netdev_, enum netdev_flags off,
NULL, /* build header */ \
NULL, /* push header */ \
NULL, /* pop header */ \
+ NULL, /* pop specific header */ \
NULL, /* get_numa_id */ \
NULL, /* set_multiq */ \
\
diff --git a/lib/netdev-provider.h b/lib/netdev-provider.h
index a33bb3b..e163376 100644
--- a/lib/netdev-provider.h
+++ b/lib/netdev-provider.h
@@ -271,6 +271,10 @@ struct netdev_class {
* for further processing. */
int (*pop_header)(struct dp_packet *packet);
+ /* Pop tunnel header from packet with specific actions */
+ int (*pop_header_spec)(struct dp_packet *packet,
+ const struct ovs_action_pop_tnl *data);
+
/* Returns the id of the numa node the 'netdev' is on. If there is no
* such info, returns NETDEV_NUMA_UNSPEC. */
int (*get_numa_id)(const struct netdev *netdev);
diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c
index a0a4da2..d926c00 100644
--- a/lib/netdev-vport.c
+++ b/lib/netdev-vport.c
@@ -561,7 +561,13 @@ set_tunnel_config(struct netdev *dev_, const struct smap *args)
} else {
tnl_cfg.ip_dst = in_addr.s_addr;
}
- } else if (!strcmp(node->key, "local_ip")) {
+ } else if (!strcmp(node->key, "tun_nodecap")) {
+ if (!strcmp(node->value, "true")) {
+ tnl_cfg.tun_nodecap = true;
+ } else {
+ tnl_cfg.tun_nodecap = false;
+ }
+ }else if (!strcmp(node->key, "local_ip")) {
struct in_addr in_addr;
if (!strcmp(node->value, "flow")) {
tnl_cfg.ip_src_flow = true;
@@ -878,6 +884,10 @@ get_tunnel_config(const struct netdev *dev, struct smap *args)
smap_add(args, "df_default", "false");
}
+ if (tnl_cfg.tun_nodecap) {
+ smap_add(args, "tun_nodecap", "true");
+ }
+
if (tnl_cfg.in_nsp_flow && tnl_cfg.out_nsp_flow) {
smap_add(args, "nsp", "flow");
} else if (tnl_cfg.in_nsp_present && tnl_cfg.out_nsp_present
@@ -1536,6 +1546,80 @@ netdev_vxlan_pop_header(struct dp_packet *packet)
}
static int
+vxlan_extract_md_no_decap(struct dp_packet *packet)
+{
+ struct pkt_metadata *md = &packet->md;
+ struct flow_tnl *tnl = &md->tunnel;
+ struct ip_header *nh;
+ struct udp_header *udp;
+
+ memset(md, 0, sizeof *md);
+ if (VXLAN_HLEN > dp_packet_size(packet)) {
+ return EINVAL;;
+ }
+
+ udp = ip_extract_tnl_md(packet, tnl);
+ if (!udp) {
+ return EINVAL;;
+ }
+
+ /* TTL decrement */
+ nh = dp_packet_l3(packet);
+ if(nh->ip_ttl){
+ nh->ip_ttl = nh->ip_ttl - 1;
+ }
+
+ if (ntohs(udp->udp_dst) == VXGPE_DST_PORT) {
+
+ struct vxgpehdr *vxg = (struct vxgpehdr *) (udp + 1);
+
+ if (get_16aligned_be32(&vxg->vx_vni) & htonl(0xff)) {
+ VLOG_WARN_RL(&err_rl, "invalid vxlan-gpe vni=%#x\n",
+ ntohl(get_16aligned_be32(&vxg->vx_vni)));
+ return EINVAL;;
+ }
+
+ tnl->tp_src = udp->udp_src;
+ tnl->tp_dst = udp->udp_dst;
+
+ if (vxg->p == 0x01 && vxg->proto == VXG_P_NSH) {
+ struct nshhdr *nsh = (struct nshhdr *) (vxg + 1);
+
+ tnl->nsp = nsh->b.b2 << 8;
+ tnl->nsi = nsh->b.svc_idx;
+ tnl->nshc1 = nsh->c.nshc1;
+ tnl->nshc2 = nsh->c.nshc2;
+ tnl->nshc3 = nsh->c.nshc3;
+ tnl->nshc4 = nsh->c.nshc4;
+ tnl->flags |= FLOW_TNL_F_NSP;
+ tnl->flags |= FLOW_TNL_F_NSI;
+ tnl->flags |= FLOW_TNL_F_NSH_C1 | FLOW_TNL_F_NSH_C2 | \
+ FLOW_TNL_F_NSH_C3 | FLOW_TNL_F_NSH_C4;
+ tnl->nsh_flags = NSH_TNL_F_NODECAP;
+ } else {
+ VLOG_WARN("Unsupported vxlan GPE + NSH format!");
+ return EINVAL;;
+ }
+
+ }
+
+ return 0;
+}
+
+
+static int
+netdev_vxlan_pop_header_spec(struct dp_packet *packet,
+ const struct ovs_action_pop_tnl *data)
+{
+ if (data->pop_type == OVS_POP_SPEC_ACTION_NO_DECAP) {
+ return vxlan_extract_md_no_decap(packet);
+ }
+
+ return EINVAL;
+}
+
+
+static int
netdev_vxlan_build_header(const struct netdev *netdev,
struct ovs_action_push_tnl *data,
const struct flow *tnl_flow)
@@ -1772,7 +1856,8 @@ netdev_vport_range(struct unixctl_conn *conn, int argc,
#define VPORT_FUNCTIONS(GET_CONFIG, SET_CONFIG, \
GET_TUNNEL_CONFIG, GET_STATUS, \
BUILD_HEADER, \
- PUSH_HEADER, POP_HEADER) \
+ PUSH_HEADER, POP_HEADER, \
+ POP_HEADER_SPEC) \
NULL, \
netdev_vport_run, \
netdev_vport_wait, \
@@ -1787,6 +1872,7 @@ netdev_vport_range(struct unixctl_conn *conn, int argc,
BUILD_HEADER, \
PUSH_HEADER, \
POP_HEADER, \
+ POP_HEADER_SPEC, \
NULL, /* get_numa_id */ \
NULL, /* set_multiq */ \
\
@@ -1840,13 +1926,13 @@ netdev_vport_range(struct unixctl_conn *conn, int argc,
-#define TUNNEL_CLASS(NAME, DPIF_PORT, BUILD_HEADER, PUSH_HEADER, POP_HEADER) \
+#define TUNNEL_CLASS(NAME, DPIF_PORT, BUILD_HEADER, PUSH_HEADER, POP_HEADER,POP_HEADER_SPEC) \
{ DPIF_PORT, \
{ NAME, VPORT_FUNCTIONS(get_tunnel_config, \
set_tunnel_config, \
get_netdev_tunnel_config, \
tunnel_get_status, \
- BUILD_HEADER, PUSH_HEADER, POP_HEADER) }}
+ BUILD_HEADER, PUSH_HEADER, POP_HEADER, POP_HEADER_SPEC) }}
void
netdev_vport_tunnel_register(void)
@@ -1855,17 +1941,18 @@ netdev_vport_tunnel_register(void)
* a port number to the end if one is necessary. */
static const struct vport_class vport_classes[] = {
TUNNEL_CLASS("geneve", "genev_sys", netdev_geneve_build_header,
- push_udp_header,
- netdev_geneve_pop_header),
+ push_udp_header,
+ netdev_geneve_pop_header, NULL),
TUNNEL_CLASS("gre", "gre_sys", netdev_gre_build_header,
netdev_gre_push_header,
- netdev_gre_pop_header),
- TUNNEL_CLASS("ipsec_gre", "gre_sys", NULL, NULL, NULL),
+ netdev_gre_pop_header,NULL),
+ TUNNEL_CLASS("ipsec_gre", "gre_sys", NULL, NULL, NULL,NULL),
TUNNEL_CLASS("vxlan", "vxlan_sys", netdev_vxlan_build_header,
netdev_vxlan_push_header,
- netdev_vxlan_pop_header),
- TUNNEL_CLASS("lisp", "lisp_sys", NULL, NULL, NULL),
- TUNNEL_CLASS("stt", "stt_sys", NULL, NULL, NULL),
+ netdev_vxlan_pop_header,
+ netdev_vxlan_pop_header_spec),
+ TUNNEL_CLASS("lisp", "lisp_sys", NULL, NULL, NULL,NULL),
+ TUNNEL_CLASS("stt", "stt_sys", NULL, NULL, NULL,NULL),
};
static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
@@ -1891,6 +1978,6 @@ netdev_vport_patch_register(void)
{ "patch", VPORT_FUNCTIONS(get_patch_config,
set_patch_config,
NULL,
- NULL, NULL, NULL, NULL) }};
+ NULL, NULL, NULL, NULL, NULL) }};
netdev_register_provider(&patch_class.netdev_class);
}
diff --git a/lib/netdev.c b/lib/netdev.c
index e3b70b1..884bc0e 100644
--- a/lib/netdev.c
+++ b/lib/netdev.c
@@ -796,6 +796,30 @@ netdev_push_header(const struct netdev *netdev,
return 0;
}
+
+int
+netdev_pop_header_spec(struct netdev *netdev,
+ struct dp_packet **buffers, int cnt,
+ const struct ovs_action_pop_tnl *data)
+{
+ int i;
+
+ if (!netdev->netdev_class->pop_header_spec) {
+ return -EINVAL;
+ }
+
+ for (i = 0; i < cnt; i++) {
+ int err;
+
+ err = netdev->netdev_class->pop_header_spec(buffers[i], data);
+ if (err) {
+ dp_packet_clear(buffers[i]);
+ }
+ }
+
+ return 0;
+}
+
/* Registers with the poll loop to wake up from the next call to poll_block()
* when the packet transmission queue has sufficient room to transmit a packet
* with netdev_send().
diff --git a/lib/netdev.h b/lib/netdev.h
index 4dadf1c..b30c932 100644
--- a/lib/netdev.h
+++ b/lib/netdev.h
@@ -72,6 +72,7 @@ struct in6_addr;
struct smap;
struct sset;
struct ovs_action_push_tnl;
+struct ovs_action_pop_tnl;
/* Network device statistics.
*
@@ -181,6 +182,8 @@ struct netdev_tunnel_config {
bool out_nshc4_flow;
ovs_be32 out_nshc4; /* outgoing NSH context c4 */
+ bool tun_nodecap;
+
};
void netdev_run(void);
@@ -241,7 +244,9 @@ int netdev_push_header(const struct netdev *netdev,
const struct ovs_action_push_tnl *data);
int netdev_pop_header(struct netdev *netdev, struct dp_packet **buffers,
int cnt);
-
+int netdev_pop_header_spec(struct netdev *netdev,
+ struct dp_packet **buffers, int cnt,
+ const struct ovs_action_pop_tnl *data);
/* Hardware address. */
int netdev_set_etheraddr(struct netdev *, const struct eth_addr mac);
int netdev_get_etheraddr(const struct netdev *, struct eth_addr *mac);
diff --git a/lib/odp-execute.c b/lib/odp-execute.c
index 54a43cd..2ecdabf 100644
--- a/lib/odp-execute.c
+++ b/lib/odp-execute.c
@@ -238,6 +238,9 @@ odp_execute_set_action(struct dp_packet *packet, const struct nlattr *a)
case OVS_KEY_ATTR_TUNNEL:
odp_set_tunnel_action(a, &md->tunnel);
+ if (md->tunnel.nsh_flags & NSH_TNL_F_NODECAP) {
+ packet_set_nsh(packet, &md->tunnel);
+ }
break;
case OVS_KEY_ATTR_SKB_MARK:
@@ -474,6 +477,7 @@ requires_datapath_assistance(const struct nlattr *a)
case OVS_ACTION_ATTR_OUTPUT:
case OVS_ACTION_ATTR_TUNNEL_PUSH:
case OVS_ACTION_ATTR_TUNNEL_POP:
+ case OVS_ACTION_ATTR_TUNNEL_POP_SPEC:
case OVS_ACTION_ATTR_USERSPACE:
case OVS_ACTION_ATTR_RECIRC:
return true;
@@ -609,6 +613,7 @@ odp_execute_actions(void *dp, struct dp_packet **packets, int cnt, bool steal,
case OVS_ACTION_ATTR_OUTPUT:
case OVS_ACTION_ATTR_TUNNEL_PUSH:
case OVS_ACTION_ATTR_TUNNEL_POP:
+ case OVS_ACTION_ATTR_TUNNEL_POP_SPEC:
case OVS_ACTION_ATTR_USERSPACE:
case OVS_ACTION_ATTR_RECIRC:
case OVS_ACTION_ATTR_UNSPEC:
diff --git a/lib/odp-util.c b/lib/odp-util.c
index 1696f77..190117f 100644
--- a/lib/odp-util.c
+++ b/lib/odp-util.c
@@ -103,6 +103,7 @@ odp_action_len(uint16_t type)
case OVS_ACTION_ATTR_OUTPUT: return sizeof(uint32_t);
case OVS_ACTION_ATTR_TUNNEL_PUSH: return ATTR_LEN_VARIABLE;
case OVS_ACTION_ATTR_TUNNEL_POP: return sizeof(uint32_t);
+ case OVS_ACTION_ATTR_TUNNEL_POP_SPEC: return ATTR_LEN_VARIABLE;
case OVS_ACTION_ATTR_USERSPACE: return ATTR_LEN_VARIABLE;
case OVS_ACTION_ATTR_PUSH_VLAN: return sizeof(struct ovs_action_push_vlan);
case OVS_ACTION_ATTR_POP_VLAN: return 0;
@@ -551,6 +552,20 @@ format_odp_tnl_push_header(struct ds *ds, struct ovs_action_push_tnl *data)
}
static void
+format_odp_tnl_pop_spec_action(struct ds *ds, const struct nlattr *attr)
+{
+ struct ovs_action_pop_tnl *data;
+
+ data = (struct ovs_action_pop_tnl *) nl_attr_get(attr);
+
+ ds_put_format(ds, "tnl_pop_spec(tnl_port(%"PRIu32"),", data->tnl_port);
+ if (data->pop_type == OVS_POP_SPEC_ACTION_NO_DECAP) {
+ ds_put_format(ds, "pop_type=%"PRIu16")",
+ OVS_POP_SPEC_ACTION_NO_DECAP);
+ }
+}
+
+static void
format_odp_tnl_push_action(struct ds *ds, const struct nlattr *attr)
{
struct ovs_action_push_tnl *data;
@@ -586,6 +601,9 @@ format_odp_action(struct ds *ds, const struct nlattr *a)
case OVS_ACTION_ATTR_TUNNEL_POP:
ds_put_format(ds, "tnl_pop(%"PRIu32")", nl_attr_get_u32(a));
break;
+ case OVS_ACTION_ATTR_TUNNEL_POP_SPEC:
+ format_odp_tnl_pop_spec_action(ds, a);
+ break;
case OVS_ACTION_ATTR_TUNNEL_PUSH:
format_odp_tnl_push_action(ds, a);
break;
@@ -1029,6 +1047,35 @@ ovs_parse_tnl_push(const char *s, struct ovs_action_push_tnl *data)
}
static int
+ovs_parse_tnl_pop_spec(const char *s, struct ovs_action_pop_tnl *data)
+{
+ struct eth_header *eth;
+ struct nshhdr *nsh;
+ uint32_t tnl_type = 0, header_len = 0;
+ uint16_t dl_type;
+ ovs_be32 nsp, nshc1, nshc2, nshc3, nshc4;
+ uint8_t nsi;
+ int n = 0;
+ if (!ovs_scan_len(s, &n, "tnl_pop_spec(tnl_port(%"SCNi32"),",
+ &data->tnl_port)) {
+ return -EINVAL;
+ }
+
+ if (!ovs_scan_len(s, &n, "pop_type=%"SCNi16")",
+ &data->pop_type)) {
+ return -EINVAL;
+ }
+
+ if (data->pop_type == OVS_POP_SPEC_ACTION_NO_DECAP) {
+ return n;
+ } else {
+ return -EINVAL;
+ }
+
+ return n;
+}
+
+static int
parse_odp_action(const char *s, const struct simap *port_names,
struct ofpbuf *actions)
{
@@ -1186,6 +1233,22 @@ parse_odp_action(const char *s, const struct simap *port_names,
}
{
+ struct ovs_action_pop_tnl data;
+ int n;
+
+ if (ovs_scan(s, "tnl_pop_spec(tnl_port(%"SCNi32"),", &(data.tnl_port))) {
+ memset(&data, 0, sizeof data);
+ n = ovs_parse_tnl_pop_spec(s, &data);
+ if (n > 0) {
+ odp_put_tnl_pop_spec_action(actions, &data);
+ return n;
+ } else if (n < 0) {
+ return n;
+ }
+ }
+ }
+
+ {
struct ovs_action_push_tnl data;
int n;
@@ -1461,7 +1524,9 @@ enum odp_key_fitness
odp_tun_key_from_attr(const struct nlattr *attr, bool udpif,
struct flow_tnl *tun)
{
+ uint8_t nsh_flags=tun->nsh_flags;
memset(tun, 0, sizeof *tun);
+ tun->nsh_flags=nsh_flags;
return odp_tun_key_from_attr__(attr, NULL, 0, NULL, tun, udpif);
}
@@ -4564,7 +4629,16 @@ odp_put_tnl_push_action(struct ofpbuf *odp_actions,
nl_msg_put_unspec(odp_actions, OVS_ACTION_ATTR_TUNNEL_PUSH, data, size);
}
-
+void
+odp_put_tnl_pop_spec_action(struct ofpbuf *odp_actions,
+ struct ovs_action_pop_tnl *data)
+{
+ int size = offsetof(struct ovs_action_pop_tnl, header);
+
+ size += data->header_len;
+ nl_msg_put_unspec(odp_actions, OVS_ACTION_ATTR_TUNNEL_POP_SPEC, data, size);
+}
+
/* The commit_odp_actions() function and its helpers. */
static void
@@ -4617,6 +4691,26 @@ commit_odp_tunnel_action(const struct flow *flow, struct flow *base,
}
}
+/* If any of the flow key data that ODP actions can modify are different in
+ * 'base->tunnel' and 'flow->tunnel', appends a set_tunnel ODP action to
+ * 'odp_actions' that change the flow tunneling information in key from
+ * 'base->tunnel' into 'flow->tunnel', and then changes 'base->tunnel' in the
+ * same way. In other words, operates the same as commit_odp_actions(), but
+ * only on tunneling information. */
+void
+commit_odp_tunnel_set_action(const struct flow_tnl *tunnel, struct flow_tnl *base,
+ struct ofpbuf *odp_actions)
+{
+ /* A valid IPV4_TUNNEL must have non-zero ip_dst. */
+ if (tunnel->ip_dst) {
+
+ if (!memcmp(tunnel, base, sizeof *tunnel)) {
+ return;
+ }
+ odp_put_tunnel_action(tunnel, odp_actions);
+ }
+}
+
static bool
commit(enum ovs_key_attr attr, bool use_masked_set,
const void *key, void *base, void *mask, size_t size,
diff --git a/lib/odp-util.h b/lib/odp-util.h
index 9f8e741..b30c246 100644
--- a/lib/odp-util.h
+++ b/lib/odp-util.h
@@ -245,6 +245,10 @@ const char *odp_key_fitness_to_string(enum odp_key_fitness);
void commit_odp_tunnel_action(const struct flow *, struct flow *base,
struct ofpbuf *odp_actions);
+void commit_odp_tunnel_set_action(const struct flow_tnl *tunnel,
+ struct flow_tnl *base,
+ struct ofpbuf *odp_actions);
+
void commit_masked_set_action(struct ofpbuf *odp_actions,
enum ovs_key_attr key_type, const void *key,
const void *mask, size_t key_size);
@@ -308,7 +312,8 @@ size_t odp_put_userspace_action(uint32_t pid,
struct ofpbuf *odp_actions);
void odp_put_tunnel_action(const struct flow_tnl *tunnel,
struct ofpbuf *odp_actions);
-
void odp_put_tnl_push_action(struct ofpbuf *odp_actions,
struct ovs_action_push_tnl *data);
+void odp_put_tnl_pop_spec_action(struct ofpbuf *odp_actions,
+ struct ovs_action_pop_tnl *data);
#endif /* odp-util.h */
diff --git a/lib/ofp-print.c b/lib/ofp-print.c
index d0c94ce..0684405 100644
--- a/lib/ofp-print.c
+++ b/lib/ofp-print.c
@@ -63,6 +63,7 @@ ofp_packet_to_string(const void *data, size_t len)
struct flow flow;
size_t l4_size;
+ buf.md.tunnel.nsh_flags=0;
dp_packet_use_const(&buf, data, len);
flow_extract(&buf, &flow);
flow_format(&ds, &flow);
diff --git a/lib/packets.c b/lib/packets.c
index a4d7854..d69d006 100644
--- a/lib/packets.c
+++ b/lib/packets.c
@@ -926,6 +926,26 @@ packet_set_nd(struct dp_packet *packet, const ovs_be32 target[4],
}
}
+void
+packet_set_nsh(struct dp_packet *packet, struct flow_tnl *tun_key)
+{
+ struct eth_header *eth;
+ struct nshhdr *nsh;
+
+ eth = (struct eth_header *) dp_packet_data(packet);
+ struct ip_header *ip = (struct ip_header *) (eth + 1);
+ struct udp_header *udp = (struct udp_header *) (ip + 1);
+ struct vxgpehdr *vxg = (struct vxgpehdr *) (udp + 1);
+
+ nsh = (struct nshhdr *) (vxg + 1);
+ nsh->b.b2 = tun_key->nsp >> 8;
+ nsh->b.svc_idx = tun_key->nsi;
+ nsh->c.nshc1 = tun_key->nshc1;
+ nsh->c.nshc2 = tun_key->nshc2;
+ nsh->c.nshc3 = tun_key->nshc3;
+ nsh->c.nshc4 = tun_key->nshc4;
+}
+
const char *
packet_tcp_flag_to_string(uint32_t flag)
{
diff --git a/lib/packets.h b/lib/packets.h
index 7f9ab98..87c955a 100644
--- a/lib/packets.h
+++ b/lib/packets.h
@@ -33,6 +33,7 @@
struct dp_packet;
struct ds;
+/* Tunnel information used in flow key and metadata. */
struct flow_tnl {
ovs_be32 ip_dst;
ovs_be32 ip_src;
@@ -44,12 +45,14 @@ struct flow_tnl {
ovs_be16 tp_dst;
ovs_be16 gbp_id;
uint8_t gbp_flags;
+ uint8_t nsh_flags;
uint8_t nsi;
ovs_be32 nsp;
ovs_be32 nshc1;
ovs_be32 nshc2;
ovs_be32 nshc3;
ovs_be32 nshc4;
+ uint8_t pad1[7]; /* Pad to 64 bits. */
struct tun_metadata metadata;
};
@@ -80,6 +83,8 @@ struct flow_tnl {
#define FLOW_TNL_F_NSH_C3 (1 << 9)
#define FLOW_TNL_F_NSH_C4 (1 << 10)
+#define NSH_TNL_F_NODECAP (1 << 1)
+
/* Returns an offset to 'src' covering all the meaningful fields in 'src'. */
static inline size_t
flow_tnl_size(const struct flow_tnl *src)
@@ -1055,6 +1060,7 @@ void packet_set_udp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
void packet_set_sctp_port(struct dp_packet *, ovs_be16 src, ovs_be16 dst);
void packet_set_nd(struct dp_packet *, const ovs_be32 target[4],
const struct eth_addr sll, const struct eth_addr tll);
+void packet_set_nsh(struct dp_packet *packet, struct flow_tnl *tun_key);
void packet_format_tcp_flags(struct ds *, uint16_t);
const char *packet_tcp_flag_to_string(uint32_t flag);
diff --git a/ofproto/ofproto-dpif-sflow.c b/ofproto/ofproto-dpif-sflow.c
index d479997..da8a9d1 100644
--- a/ofproto/ofproto-dpif-sflow.c
+++ b/ofproto/ofproto-dpif-sflow.c
@@ -1119,6 +1119,10 @@ dpif_sflow_read_actions(const struct flow *flow,
sflow_actions->tunnel_err = true;
break;
+ case OVS_ACTION_ATTR_TUNNEL_POP_SPEC:
+ sflow_actions->tunnel_err = true;
+ break;
+
case OVS_ACTION_ATTR_TUNNEL_PUSH:
/* XXX: This actions appears to come with it's own
* OUTPUT action, so should it be regarded as having
diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c
index 9c64c24..71e255e 100644
--- a/ofproto/ofproto-dpif-xlate.c
+++ b/ofproto/ofproto-dpif-xlate.c
@@ -171,7 +171,7 @@ struct xlate_ctx {
* which might lead to an infinite loop. This could happen easily
* if a tunnel is marked as 'ip_remote=flow', and the flow does not
* actually set the tun_dst field. */
- ovs_be32 orig_tunnel_ip_dst;
+ struct flow_tnl orig_tunnel;
/* Stack for the push and pop actions. Each stack element is of type
* "union mf_subvalue". */
@@ -2787,6 +2787,37 @@ build_tunnel_send(struct xlate_ctx *ctx, const struct xport *xport,
return 0;
}
+static int
+build_tunnel_pop(const struct xlate_ctx *ctx, odp_port_t tunnel_odp_port, struct flow *flow)
+{
+ const struct netdev_tunnel_config * cfg;
+
+ cfg = tnl_port_cfg(tunnel_odp_port, flow);
+
+ if (cfg) {
+ if (cfg->tun_nodecap) {
+ struct ovs_action_pop_tnl tnl_pop_data;
+ memset(&tnl_pop_data, 0, sizeof tnl_pop_data);
+
+ tnl_pop_data.tnl_port = odp_to_u32(tunnel_odp_port);
+ tnl_pop_data.pop_type = OVS_POP_SPEC_ACTION_NO_DECAP;
+ odp_put_tnl_pop_spec_action(ctx->odp_actions, &tnl_pop_data);
+
+ } else {
+ nl_msg_put_odp_port(ctx->odp_actions,
+ OVS_ACTION_ATTR_TUNNEL_POP,
+ tunnel_odp_port);
+ }
+
+ } else {
+ nl_msg_put_odp_port(ctx->odp_actions,
+ OVS_ACTION_ATTR_TUNNEL_POP,
+ tunnel_odp_port);
+ }
+
+ return 0;
+}
+
static void
xlate_commit_actions(struct xlate_ctx *ctx)
{
@@ -2970,7 +3001,7 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port,
xlate_report(ctx, "Tunneling decided against output");
goto out; /* restore flow_nw_tos */
}
- if (flow->tunnel.ip_dst == ctx->orig_tunnel_ip_dst) {
+ if (flow->tunnel.ip_dst == ctx->orig_tunnel.ip_dst) {
xlate_report(ctx, "Not tunneling to our own address");
goto out; /* restore flow_nw_tos */
}
@@ -3011,6 +3042,10 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port,
if (out_port != ODPP_NONE) {
xlate_commit_actions(ctx);
+ if (flow->tunnel.nsh_flags & NSH_TNL_F_NODECAP) {
+ commit_odp_tunnel_set_action(&flow->tunnel, &ctx->orig_tunnel,
+ ctx->odp_actions);
+ }
if (xr) {
struct ovs_action_hash *act_hash;
@@ -3040,10 +3075,11 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port,
odp_tnl_port = tnl_port_map_lookup(flow, wc);
}
- if (odp_tnl_port != ODPP_NONE) {
- nl_msg_put_odp_port(ctx->odp_actions,
- OVS_ACTION_ATTR_TUNNEL_POP,
- odp_tnl_port);
+ if (odp_tnl_port != ODPP_NONE &&
+ !(flow->tunnel.nsh_flags & NSH_TNL_F_NODECAP)) {
+ flow_tnl = flow->tunnel;
+ build_tunnel_pop(ctx, odp_tnl_port, flow);
+ flow->tunnel = flow_tnl;
} else {
/* Tunnel push-pop action is not compatible with
* IPFIX action. */
@@ -4791,7 +4827,7 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout)
.xin = xin,
.xout = xout,
.base_flow = *flow,
- .orig_tunnel_ip_dst = flow->tunnel.ip_dst,
+ .orig_tunnel = flow->tunnel,
.xbridge = xbridge,
.stack = OFPBUF_STUB_INITIALIZER(stack_stub),
.rule = xin->rule,
diff --git a/ofproto/tunnel.c b/ofproto/tunnel.c
index 52e28fb..4606fb6 100644
--- a/ofproto/tunnel.c
+++ b/ofproto/tunnel.c
@@ -553,6 +553,55 @@ out:
return out_port;
}
+const struct netdev_tunnel_config *
+tnl_port_cfg(odp_port_t odp_port, struct flow *flow) OVS_EXCLUDED(rwlock)
+{
+ const struct netdev_tunnel_config *cfg = NULL;
+ struct tnl_port *tnl_port;
+
+ fat_rwlock_rdlock(&rwlock);
+ tnl_port = tnl_find_odp_port(odp_port);
+ if (!tnl_port) {
+ goto out;
+ }
+
+ cfg = netdev_get_tunnel_config(tnl_port->netdev);
+ ovs_assert(cfg);
+
+ if (!cfg->out_nsp_flow) {
+ flow->tunnel.nsp = cfg->out_nsp;
+ }
+ if (!cfg->out_nsi_flow) {
+ flow->tunnel.nsi = cfg->out_nsi;
+ }
+ if (!cfg->out_nshc1_flow) {
+ flow->tunnel.nshc1 = cfg->out_nshc1;
+ }
+ if (!cfg->out_nshc2_flow) {
+ flow->tunnel.nshc2 = cfg->out_nshc2;
+ }
+ if (!cfg->out_nshc3_flow) {
+ flow->tunnel.nshc3 = cfg->out_nshc3;
+ }
+ if (!cfg->out_nshc4_flow) {
+ flow->tunnel.nshc4 = cfg->out_nshc4;
+ }
+
+ flow->tunnel.flags = (cfg->dont_fragment ? FLOW_TNL_F_DONT_FRAGMENT : 0)
+ | (cfg->csum ? FLOW_TNL_F_CSUM : 0)
+ | (cfg->out_nsp_present ? FLOW_TNL_F_NSP : 0)
+ | (cfg->out_nsi_present ? FLOW_TNL_F_NSI : 0)
+ | (cfg->out_nshc1_present ? FLOW_TNL_F_NSH_C1 : 0)
+ | (cfg->out_nshc2_present ? FLOW_TNL_F_NSH_C2 : 0)
+ | (cfg->out_nshc3_present ? FLOW_TNL_F_NSH_C3 : 0)
+ | (cfg->out_nshc4_present ? FLOW_TNL_F_NSH_C4 : 0)
+ | (cfg->out_key_present ? FLOW_TNL_F_KEY : 0);
+
+out:
+ fat_rwlock_unlock(&rwlock);
+ return cfg;
+}
+
static uint32_t
tnl_hash(struct tnl_match *match)
{
@@ -590,6 +639,43 @@ tnl_find_exact(struct tnl_match *match, struct hmap *map)
return NULL;
}
+static struct tnl_port *
+tnl_find_exact_odp_port(odp_port_t odp_port, struct hmap *map)
+ OVS_REQ_RDLOCK(rwlock)
+{
+ if (map) {
+ struct tnl_port *tnl_port;
+
+ HMAP_FOR_EACH (tnl_port, match_node, map) {
+ if (tnl_port->match.odp_port == odp_port) {
+ return tnl_port;
+ }
+ }
+ }
+ return NULL;
+}
+
+static struct tnl_port *
+tnl_find_odp_port(odp_port_t odp_port) OVS_REQ_RDLOCK(rwlock)
+{
+ int i;
+
+ for (i = 0; i < N_MATCH_TYPES; i++) {
+ struct hmap *map = tnl_match_maps[i];
+
+ if (map) {
+ struct tnl_port *tnl_port;
+
+ tnl_port = tnl_find_exact_odp_port(odp_port, map);
+ if (tnl_port) {
+ return tnl_port;
+ }
+ }
+ }
+
+ return NULL;
+}
+
/* Returns the tnl_port that is the best match for the tunnel data in 'flow',
* or NULL if no tnl_port matches 'flow'. */
static struct tnl_port *
diff --git a/ofproto/tunnel.h b/ofproto/tunnel.h
index 3e704fb..2b608ce 100644
--- a/ofproto/tunnel.h
+++ b/ofproto/tunnel.h
@@ -26,6 +26,7 @@
* header information from the kernel. */
struct ovs_action_push_tnl;
+struct ovs_action_pop_tnl;
struct ofport_dpif;
struct netdev;
@@ -43,6 +44,9 @@ bool tnl_process_ecn(struct flow *);
odp_port_t tnl_port_send(const struct ofport_dpif *, struct flow *,
struct flow_wildcards *wc);
+const struct netdev_tunnel_config *
+tnl_port_cfg(odp_port_t odp_port, struct flow *flow);
+
/* Returns true if 'flow' should be submitted to tnl_port_receive(). */
static inline bool
tnl_port_should_receive(const struct flow *flow)
diff --git a/tests/tunnel.at b/tests/tunnel.at
index 5ec5e6c..851afdc 100644
--- a/tests/tunnel.at
+++ b/tests/tunnel.at
@@ -645,6 +645,34 @@ AT_CHECK([tail -1 stdout], [0],
OVS_VSWITCHD_STOP(["/The Open vSwitch kernel module is probably not loaded/d"])
AT_CLEANUP
+AT_SETUP([tunnel - VXLAN-GPE NSH tun_nodecap - user space])
+OVS_VSWITCHD_START([add-port br0 p0 -- set Interface p0 type=dummy ofport_request=1 other-config:hwaddr=aa:55:aa:55:00:00])
+AT_CHECK([ovs-vsctl add-br int-br -- set bridge int-br datapath_type=dummy], [0])
+AT_CHECK([ovs-vsctl add-port int-br p1 -- set Interface p1 type=vxlan options:key=flow \
+ options:remote_ip=1.1.1.1 options:tun_nodecap=true options:nsi=flow options:nsp=flow options:nshc1=flow options:in_key=flow options:dst_port=4790 ofport_request=2])
+AT_CHECK([ovs-vsctl add-port int-br p2 -- set Interface p2 type=dummy ofport_request=3])
+
+AT_CHECK([ovs-appctl netdev-dummy/ip4addr br0 2.2.2.22/24], [0], [OK
+])
+AT_CHECK([ovs-appctl ovs/route/add 1.1.1.1/24 br0], [0], [OK
+])
+
+AT_CHECK([ovs-ofctl add-flow br0 action=normal])
+
+AT_CHECK([ovs-appctl tnl/ports/show |sort], [0], [dnl
+Listening ports:
+vxlan_sys_4790 (4790)
+])
+
+dnl remote_ip p0
+AT_CHECK([ovs-appctl ofproto/trace ovs-dummy 'in_port(1),eth(src=50:54:00:00:00:05,dst=aa:55:aa:55:00:00),eth_type(0x0800),ipv4(src=1.1.1.1,dst=2.2.2.22,proto=17,tos=0,ttl=64,frag=no),udp(src=8,dst=4790)'], [0], [stdout])
+AT_CHECK([tail -1 stdout], [0],
+ [Datapath actions: tnl_pop_spec(tnl_port(4790),pop_type=2)
+])
+
+OVS_VSWITCHD_STOP(["/The Open vSwitch kernel module is probably not loaded/d"])
+AT_CLEANUP
+
AT_SETUP([tunnel - Geneve metadata])
OVS_VSWITCHD_START([add-port br0 p1 -- set Interface p1 type=geneve \
options:remote_ip=1.1.1.1 ofport_request=1 \
--
1.9.3
More information about the dev
mailing list