/*
- * Copyright (c) 2009, 2010, 2011 Nicira Networks.
+ * Copyright (c) 2009, 2010, 2011, 2012 Nicira Networks.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#include "ofpbuf.h"
#include "packets.h"
#include "poll-loop.h"
+#include "random.h"
#include "shash.h"
#include "timeval.h"
#include "util.h"
int open_cnt;
bool destroyed;
- bool drop_frags; /* Drop all IP fragments, if true. */
struct dp_netdev_queue queues[N_QUEUES];
struct hmap flow_table; /* Flow table. */
/* Statistics. */
- long long int n_frags; /* Number of dropped IP fragments. */
long long int n_hit; /* Number of flow table matches. */
long long int n_missed; /* Number of flow table misses. */
long long int n_lost; /* Number of misses not passed to client. */
static int dp_netdev_output_userspace(struct dp_netdev *, const struct ofpbuf *,
int queue_no, const struct flow *,
uint64_t arg);
-static int dp_netdev_execute_actions(struct dp_netdev *,
- struct ofpbuf *, struct flow *,
- const struct nlattr *actions,
- size_t actions_len);
+static void dp_netdev_execute_actions(struct dp_netdev *,
+ struct ofpbuf *, struct flow *,
+ const struct nlattr *actions,
+ size_t actions_len);
static struct dpif_class dpif_dummy_class;
dp->class = class;
dp->name = xstrdup(name);
dp->open_cnt = 0;
- dp->drop_frags = false;
for (i = 0; i < N_QUEUES; i++) {
dp->queues[i].head = dp->queues[i].tail = 0;
}
{
struct dp_netdev *dp = get_dp_netdev(dpif);
stats->n_flows = hmap_count(&dp->flow_table);
- stats->n_frags = dp->n_frags;
stats->n_hit = dp->n_hit;
stats->n_missed = dp->n_missed;
stats->n_lost = dp->n_lost;
return 0;
}
-static int
-dpif_netdev_get_drop_frags(const struct dpif *dpif, bool *drop_fragsp)
-{
- struct dp_netdev *dp = get_dp_netdev(dpif);
- *drop_fragsp = dp->drop_frags;
- return 0;
-}
-
-static int
-dpif_netdev_set_drop_frags(struct dpif *dpif, bool drop_frags)
-{
- struct dp_netdev *dp = get_dp_netdev(dpif);
- dp->drop_frags = drop_frags;
- return 0;
-}
-
static int
do_add_port(struct dp_netdev *dp, const char *devname, const char *type,
uint16_t port_no)
}
static int
-dpif_netdev_port_add(struct dpif *dpif, struct netdev *netdev,
- uint16_t *port_nop)
+choose_port(struct dpif *dpif, struct netdev *netdev)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
int port_no;
+ if (dpif->dpif_class == &dpif_dummy_class) {
+ /* If the port name contains a number, try to assign that port number.
+ * This can make writing unit tests easier because port numbers are
+ * predictable. */
+ const char *p;
+
+ for (p = netdev_get_name(netdev); *p != '\0'; p++) {
+ if (isdigit((unsigned char) *p)) {
+ port_no = strtol(p, NULL, 10);
+ if (port_no > 0 && port_no < MAX_PORTS
+ && !dp->ports[port_no]) {
+ return port_no;
+ }
+ break;
+ }
+ }
+ }
+
for (port_no = 0; port_no < MAX_PORTS; port_no++) {
if (!dp->ports[port_no]) {
- *port_nop = port_no;
- return do_add_port(dp, netdev_get_name(netdev),
- netdev_get_type(netdev), port_no);
+ return port_no;
}
}
+
+ return -1;
+}
+
+static int
+dpif_netdev_port_add(struct dpif *dpif, struct netdev *netdev,
+ uint16_t *port_nop)
+{
+ struct dp_netdev *dp = get_dp_netdev(dpif);
+ int port_no;
+
+ port_no = choose_port(dpif, netdev);
+ if (port_no >= 0) {
+ *port_nop = port_no;
+ return do_add_port(dp, netdev_get_name(netdev),
+ netdev_get_type(netdev), port_no);
+ }
return EFBIG;
}
struct dpif_port *dpif_port)
{
dpif_port->name = xstrdup(netdev_get_name(port->netdev));
- dpif_port->type = xstrdup(port->internal ? "internal" : "system");
+ dpif_port->type = xstrdup(port->internal ? "internal"
+ : netdev_get_type(port->netdev));
dpif_port->port_no = port->port_no;
}
free(state->name);
state->name = xstrdup(netdev_get_name(port->netdev));
dpif_port->name = state->name;
- dpif_port->type = port->internal ? "internal" : "system";
+ dpif_port->type = (char *) (port->internal ? "internal"
+ : netdev_get_type(port->netdev));
dpif_port->port_no = port->port_no;
state->port_no = port_no + 1;
return 0;
return 0;
}
-static int
-dpif_netdev_validate_actions(const struct nlattr *actions,
- size_t actions_len, bool *mutates)
-{
- const struct nlattr *a;
- unsigned int left;
-
- *mutates = false;
- NL_ATTR_FOR_EACH (a, left, actions, actions_len) {
- uint16_t type = nl_attr_type(a);
- int len = odp_action_len(type);
-
- if (len != nl_attr_get_size(a)) {
- return EINVAL;
- }
-
- switch (type) {
- case OVS_ACTION_ATTR_OUTPUT:
- if (nl_attr_get_u32(a) >= MAX_PORTS) {
- return EINVAL;
- }
- break;
-
- case OVS_ACTION_ATTR_USERSPACE:
- break;
-
- case OVS_ACTION_ATTR_PUSH_VLAN:
- *mutates = true;
- if (nl_attr_get_be16(a) & htons(VLAN_CFI)) {
- return EINVAL;
- }
- break;
-
- case OVS_ACTION_ATTR_SET_NW_TOS:
- *mutates = true;
- if (nl_attr_get_u8(a) & IP_ECN_MASK) {
- return EINVAL;
- }
- break;
-
- case OVS_ACTION_ATTR_POP_VLAN:
- case OVS_ACTION_ATTR_SET_DL_SRC:
- case OVS_ACTION_ATTR_SET_DL_DST:
- case OVS_ACTION_ATTR_SET_NW_SRC:
- case OVS_ACTION_ATTR_SET_NW_DST:
- case OVS_ACTION_ATTR_SET_TP_SRC:
- case OVS_ACTION_ATTR_SET_TP_DST:
- *mutates = true;
- break;
-
- case OVS_ACTION_ATTR_SET_TUNNEL:
- case OVS_ACTION_ATTR_SET_PRIORITY:
- case OVS_ACTION_ATTR_POP_PRIORITY:
- default:
- return EOPNOTSUPP;
- }
- }
- return 0;
-}
-
static int
set_flow_actions(struct dp_netdev_flow *flow,
const struct nlattr *actions, size_t actions_len)
{
- bool mutates;
- int error;
-
- error = dpif_netdev_validate_actions(actions, actions_len, &mutates);
- if (error) {
- return error;
- }
-
flow->actions = xrealloc(flow->actions, actions_len);
flow->actions_len = actions_len;
memcpy(flow->actions, actions, actions_len);
{
struct dp_netdev *dp = get_dp_netdev(dpif);
struct ofpbuf copy;
- bool mutates;
struct flow key;
int error;
return EINVAL;
}
- error = dpif_netdev_validate_actions(actions, actions_len, &mutates);
- if (error) {
- return error;
- }
+ /* Make a deep copy of 'packet', because we might modify its data. */
+ ofpbuf_init(©, DP_NETDEV_HEADROOM + packet->size);
+ ofpbuf_reserve(©, DP_NETDEV_HEADROOM);
+ ofpbuf_put(©, packet->data, packet->size);
- if (mutates) {
- /* We need a deep copy of 'packet' since we're going to modify its
- * data. */
- ofpbuf_init(©, DP_NETDEV_HEADROOM + packet->size);
- ofpbuf_reserve(©, DP_NETDEV_HEADROOM);
- ofpbuf_put(©, packet->data, packet->size);
- } else {
- /* We still need a shallow copy of 'packet', even though we won't
- * modify its data, because flow_extract() modifies packet->l2, etc.
- * We could probably get away with modifying those but it's more polite
- * if we don't. */
- copy = *packet;
- }
-
- flow_extract(©, 0, -1, &key);
+ flow_extract(©, 0, 0, -1, &key);
error = dpif_netdev_flow_from_nlattrs(key_attrs, key_len, &key);
if (!error) {
- error = dp_netdev_execute_actions(dp, ©, &key,
- actions, actions_len);
- }
- if (mutates) {
- ofpbuf_uninit(©);
+ dp_netdev_execute_actions(dp, ©, &key,
+ actions, actions_len);
}
+
+ ofpbuf_uninit(©);
return error;
}
return 0;
}
+static int
+dpif_netdev_queue_to_priority(const struct dpif *dpif OVS_UNUSED,
+ uint32_t queue_id, uint32_t *priority)
+{
+ *priority = queue_id;
+ return 0;
+}
+
static struct dp_netdev_queue *
find_nonempty_queue(struct dpif *dpif)
{
if (packet->size < ETH_HEADER_LEN) {
return;
}
- if (flow_extract(packet, 0, port->port_no, &key) && dp->drop_frags) {
- dp->n_frags++;
- return;
- }
-
+ flow_extract(packet, 0, 0, port->port_no, &key);
flow = dp_netdev_lookup_flow(dp, &key);
if (flow) {
dp_netdev_flow_used(flow, &key, packet);
}
static void
-dp_netdev_pop_vlan(struct ofpbuf *packet)
-{
- struct vlan_eth_header *veh = packet->l2;
- if (packet->size >= sizeof *veh
- && veh->veth_type == htons(ETH_TYPE_VLAN)) {
- struct eth_header tmp;
-
- memcpy(tmp.eth_dst, veh->veth_dst, ETH_ADDR_LEN);
- memcpy(tmp.eth_src, veh->veth_src, ETH_ADDR_LEN);
- tmp.eth_type = veh->veth_next_type;
-
- ofpbuf_pull(packet, VLAN_HEADER_LEN);
- packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN;
- memcpy(packet->data, &tmp, sizeof tmp);
- }
-}
-
-static void
-dp_netdev_set_dl_src(struct ofpbuf *packet, const uint8_t dl_addr[ETH_ADDR_LEN])
-{
- struct eth_header *eh = packet->l2;
- memcpy(eh->eth_src, dl_addr, sizeof eh->eth_src);
-}
-
-static void
-dp_netdev_set_dl_dst(struct ofpbuf *packet, const uint8_t dl_addr[ETH_ADDR_LEN])
+dp_netdev_set_dl(struct ofpbuf *packet, const struct ovs_key_ethernet *eth_key)
{
struct eth_header *eh = packet->l2;
- memcpy(eh->eth_dst, dl_addr, sizeof eh->eth_dst);
-}
-static bool
-is_ip(const struct ofpbuf *packet, const struct flow *key)
-{
- return key->dl_type == htons(ETH_TYPE_IP) && packet->l4;
-}
-
-static void
-dp_netdev_set_nw_addr(struct ofpbuf *packet, const struct flow *key,
- const struct nlattr *a)
-{
- if (is_ip(packet, key)) {
- struct ip_header *nh = packet->l3;
- ovs_be32 ip = nl_attr_get_be32(a);
- uint16_t type = nl_attr_type(a);
- ovs_be32 *field;
-
- field = type == OVS_ACTION_ATTR_SET_NW_SRC ? &nh->ip_src : &nh->ip_dst;
- if (key->nw_proto == IPPROTO_TCP && packet->l7) {
- struct tcp_header *th = packet->l4;
- th->tcp_csum = recalc_csum32(th->tcp_csum, *field, ip);
- } else if (key->nw_proto == IPPROTO_UDP && packet->l7) {
- struct udp_header *uh = packet->l4;
- if (uh->udp_csum) {
- uh->udp_csum = recalc_csum32(uh->udp_csum, *field, ip);
- if (!uh->udp_csum) {
- uh->udp_csum = htons(0xffff);
- }
- }
- }
- nh->ip_csum = recalc_csum32(nh->ip_csum, *field, ip);
- *field = ip;
- }
-}
-
-static void
-dp_netdev_set_nw_tos(struct ofpbuf *packet, const struct flow *key,
- uint8_t nw_tos)
-{
- if (is_ip(packet, key)) {
- struct ip_header *nh = packet->l3;
- uint8_t *field = &nh->ip_tos;
-
- /* Set the DSCP bits and preserve the ECN bits. */
- uint8_t new = nw_tos | (nh->ip_tos & IP_ECN_MASK);
-
- nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t)*field),
- htons((uint16_t) new));
- *field = new;
- }
-}
-
-static void
-dp_netdev_set_tp_port(struct ofpbuf *packet, const struct flow *key,
- const struct nlattr *a)
-{
- if (is_ip(packet, key)) {
- uint16_t type = nl_attr_type(a);
- ovs_be16 port = nl_attr_get_be16(a);
- ovs_be16 *field;
-
- if (key->nw_proto == IPPROTO_TCP && packet->l7) {
- struct tcp_header *th = packet->l4;
- field = (type == OVS_ACTION_ATTR_SET_TP_SRC
- ? &th->tcp_src : &th->tcp_dst);
- th->tcp_csum = recalc_csum16(th->tcp_csum, *field, port);
- *field = port;
- } else if (key->nw_proto == IPPROTO_UDP && packet->l7) {
- struct udp_header *uh = packet->l4;
- field = (type == OVS_ACTION_ATTR_SET_TP_SRC
- ? &uh->udp_src : &uh->udp_dst);
- uh->udp_csum = recalc_csum16(uh->udp_csum, *field, port);
- *field = port;
- } else {
- return;
- }
- }
+ memcpy(eh->eth_src, eth_key->eth_src, sizeof eh->eth_src);
+ memcpy(eh->eth_dst, eth_key->eth_dst, sizeof eh->eth_dst);
}
static void
return 0;
}
-static int
+static void
+dp_netdev_sample(struct dp_netdev *dp,
+ struct ofpbuf *packet, struct flow *key,
+ const struct nlattr *action)
+{
+ const struct nlattr *subactions = NULL;
+ const struct nlattr *a;
+ size_t left;
+
+ NL_NESTED_FOR_EACH_UNSAFE (a, left, action) {
+ int type = nl_attr_type(a);
+
+ switch ((enum ovs_sample_attr) type) {
+ case OVS_SAMPLE_ATTR_PROBABILITY:
+ if (random_uint32() >= nl_attr_get_u32(a)) {
+ return;
+ }
+ break;
+
+ case OVS_SAMPLE_ATTR_ACTIONS:
+ subactions = a;
+ break;
+
+ case OVS_SAMPLE_ATTR_UNSPEC:
+ case __OVS_SAMPLE_ATTR_MAX:
+ default:
+ NOT_REACHED();
+ }
+ }
+
+ dp_netdev_execute_actions(dp, packet, key, nl_attr_get(subactions),
+ nl_attr_get_size(subactions));
+}
+
+static void
+dp_netdev_action_userspace(struct dp_netdev *dp,
+ struct ofpbuf *packet, struct flow *key,
+ const struct nlattr *a)
+{
+ const struct nlattr *userdata_attr;
+ uint64_t userdata;
+
+ userdata_attr = nl_attr_find_nested(a, OVS_USERSPACE_ATTR_USERDATA);
+ userdata = userdata_attr ? nl_attr_get_u64(userdata_attr) : 0;
+ dp_netdev_output_userspace(dp, packet, DPIF_UC_ACTION, key, userdata);
+}
+
+static void
+execute_set_action(struct ofpbuf *packet, const struct nlattr *a)
+{
+ enum ovs_key_attr type = nl_attr_type(a);
+ const struct ovs_key_ipv4 *ipv4_key;
+ const struct ovs_key_tcp *tcp_key;
+ const struct ovs_key_udp *udp_key;
+
+ switch (type) {
+ case OVS_KEY_ATTR_TUN_ID:
+ case OVS_KEY_ATTR_PRIORITY:
+ case OVS_KEY_ATTR_IPV6:
+ /* not implemented */
+ break;
+
+ case OVS_KEY_ATTR_ETHERNET:
+ dp_netdev_set_dl(packet,
+ nl_attr_get_unspec(a, sizeof(struct ovs_key_ethernet)));
+ break;
+
+ case OVS_KEY_ATTR_IPV4:
+ ipv4_key = nl_attr_get_unspec(a, sizeof(struct ovs_key_ipv4));
+ packet_set_ipv4(packet, ipv4_key->ipv4_src, ipv4_key->ipv4_dst,
+ ipv4_key->ipv4_tos, ipv4_key->ipv4_ttl);
+ break;
+
+ case OVS_KEY_ATTR_TCP:
+ tcp_key = nl_attr_get_unspec(a, sizeof(struct ovs_key_tcp));
+ packet_set_tcp_port(packet, tcp_key->tcp_src, tcp_key->tcp_dst);
+ break;
+
+ case OVS_KEY_ATTR_UDP:
+ udp_key = nl_attr_get_unspec(a, sizeof(struct ovs_key_udp));
+ packet_set_udp_port(packet, udp_key->udp_src, udp_key->udp_dst);
+ break;
+
+ case OVS_KEY_ATTR_UNSPEC:
+ case OVS_KEY_ATTR_ENCAP:
+ case OVS_KEY_ATTR_ETHERTYPE:
+ case OVS_KEY_ATTR_IN_PORT:
+ case OVS_KEY_ATTR_VLAN:
+ case OVS_KEY_ATTR_ICMP:
+ case OVS_KEY_ATTR_ICMPV6:
+ case OVS_KEY_ATTR_ARP:
+ case OVS_KEY_ATTR_ND:
+ case __OVS_KEY_ATTR_MAX:
+ default:
+ NOT_REACHED();
+ }
+}
+
+static void
dp_netdev_execute_actions(struct dp_netdev *dp,
struct ofpbuf *packet, struct flow *key,
const struct nlattr *actions,
unsigned int left;
NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
- switch (nl_attr_type(a)) {
+ const struct ovs_action_push_vlan *vlan;
+ int type = nl_attr_type(a);
+
+ switch ((enum ovs_action_attr) type) {
case OVS_ACTION_ATTR_OUTPUT:
dp_netdev_output_port(dp, packet, nl_attr_get_u32(a));
break;
case OVS_ACTION_ATTR_USERSPACE:
- dp_netdev_output_userspace(dp, packet, DPIF_UC_ACTION,
- key, nl_attr_get_u64(a));
+ dp_netdev_action_userspace(dp, packet, key, a);
break;
case OVS_ACTION_ATTR_PUSH_VLAN:
- eth_push_vlan(packet, nl_attr_get_be16(a));
+ vlan = nl_attr_get(a);
+ eth_push_vlan(packet, vlan->vlan_tci);
break;
case OVS_ACTION_ATTR_POP_VLAN:
- dp_netdev_pop_vlan(packet);
- break;
-
- case OVS_ACTION_ATTR_SET_DL_SRC:
- dp_netdev_set_dl_src(packet, nl_attr_get_unspec(a, ETH_ADDR_LEN));
+ eth_pop_vlan(packet);
break;
- case OVS_ACTION_ATTR_SET_DL_DST:
- dp_netdev_set_dl_dst(packet, nl_attr_get_unspec(a, ETH_ADDR_LEN));
+ case OVS_ACTION_ATTR_SET:
+ execute_set_action(packet, nl_attr_get(a));
break;
- case OVS_ACTION_ATTR_SET_NW_SRC:
- case OVS_ACTION_ATTR_SET_NW_DST:
- dp_netdev_set_nw_addr(packet, key, a);
+ case OVS_ACTION_ATTR_SAMPLE:
+ dp_netdev_sample(dp, packet, key, a);
break;
- case OVS_ACTION_ATTR_SET_NW_TOS:
- dp_netdev_set_nw_tos(packet, key, nl_attr_get_u8(a));
- break;
-
- case OVS_ACTION_ATTR_SET_TP_SRC:
- case OVS_ACTION_ATTR_SET_TP_DST:
- dp_netdev_set_tp_port(packet, key, a);
- break;
+ case OVS_ACTION_ATTR_UNSPEC:
+ case __OVS_ACTION_ATTR_MAX:
+ NOT_REACHED();
}
}
- return 0;
}
const struct dpif_class dpif_netdev_class = {
dpif_netdev_run,
dpif_netdev_wait,
dpif_netdev_get_stats,
- dpif_netdev_get_drop_frags,
- dpif_netdev_set_drop_frags,
dpif_netdev_port_add,
dpif_netdev_port_del,
dpif_netdev_port_query_by_number,
dpif_netdev_port_query_by_name,
dpif_netdev_get_max_ports,
+ NULL, /* port_get_pid */
dpif_netdev_port_dump_start,
dpif_netdev_port_dump_next,
dpif_netdev_port_dump_done,
dpif_netdev_flow_dump_next,
dpif_netdev_flow_dump_done,
dpif_netdev_execute,
+ NULL, /* operate */
dpif_netdev_recv_get_mask,
dpif_netdev_recv_set_mask,
- NULL, /* queue_to_priority */
+ dpif_netdev_queue_to_priority,
dpif_netdev_recv,
dpif_netdev_recv_wait,
dpif_netdev_recv_purge,
git://git.onelab.eu / sliver-openvswitch.git / blobdiff