#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. */
/* Ports. */
- int n_ports;
struct dp_netdev_port *ports[MAX_PORTS];
struct list port_list;
unsigned int serial;
struct flow key;
/* Statistics. */
- struct timespec used; /* Last used time. */
+ long long int used; /* Last used time, in monotonic msecs. */
long long int packet_count; /* Number of packets matched. */
long long int byte_count; /* Number of bytes matched. */
- uint16_t tcp_ctl; /* Bitwise-OR of seen tcp_ctl values. */
+ ovs_be16 tcp_ctl; /* Bitwise-OR of seen tcp_ctl values. */
/* Actions. */
struct nlattr *actions;
static int do_del_port(struct dp_netdev *, uint16_t port_no);
static int dpif_netdev_open(const struct dpif_class *, const char *name,
bool create, struct dpif **);
-static int dp_netdev_output_control(struct dp_netdev *, const struct ofpbuf *,
+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;
}
hmap_init(&dp->flow_table);
list_init(&dp->port_list);
- error = do_add_port(dp, name, "internal", ODPP_LOCAL);
+ error = do_add_port(dp, name, "internal", OVSP_LOCAL);
if (error) {
dp_netdev_free(dp);
return error;
static void
dp_netdev_free(struct dp_netdev *dp)
{
+ struct dp_netdev_port *port, *next;
+
dp_netdev_flow_flush(dp);
- while (dp->n_ports > 0) {
- struct dp_netdev_port *port = CONTAINER_OF(
- dp->port_list.next, struct dp_netdev_port, node);
+ LIST_FOR_EACH_SAFE (port, next, node, &dp->port_list) {
do_del_port(dp, port->port_no);
}
dp_netdev_purge_queues(dp);
}
static int
-dpif_netdev_get_stats(const struct dpif *dpif, struct odp_stats *stats)
+dpif_netdev_get_stats(const struct dpif *dpif, struct dpif_dp_stats *stats)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
- memset(stats, 0, sizeof *stats);
- stats->n_frags = dp->n_frags;
+ stats->n_flows = hmap_count(&dp->flow_table);
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)
{
struct dp_netdev_port *port;
- struct netdev_options netdev_options;
struct netdev *netdev;
bool internal;
int mtu;
int error;
/* XXX reject devices already in some dp_netdev. */
- if (type[0] == '\0' || !strcmp(type, "system")) {
+ if (type[0] == '\0' || !strcmp(type, "system") || !strcmp(type, "dummy")) {
internal = false;
} else if (!strcmp(type, "internal")) {
internal = true;
}
/* Open and validate network device. */
- memset(&netdev_options, 0, sizeof netdev_options);
- netdev_options.name = devname;
- netdev_options.ethertype = NETDEV_ETH_TYPE_ANY;
if (dp->class == &dpif_dummy_class) {
- netdev_options.type = "dummy";
+ type = "dummy";
} else if (internal) {
- netdev_options.type = "tap";
+ type = "tap";
}
- error = netdev_open(&netdev_options, &netdev);
+ error = netdev_open(devname, type, &netdev);
if (error) {
return error;
}
/* XXX reject loopback devices */
/* XXX reject non-Ethernet devices */
+ error = netdev_listen(netdev);
+ if (error) {
+ VLOG_ERR("%s: cannot receive packets on this network device (%s)",
+ devname, strerror(errno));
+ netdev_close(netdev);
+ return error;
+ }
+
error = netdev_turn_flags_on(netdev, NETDEV_PROMISC, false);
if (error) {
netdev_close(netdev);
port->netdev = netdev;
port->internal = internal;
- netdev_get_mtu(netdev, &mtu);
- if (mtu > max_mtu) {
+ error = netdev_get_mtu(netdev, &mtu);
+ if (!error) {
max_mtu = mtu;
}
list_push_back(&dp->port_list, &port->node);
dp->ports[port_no] = port;
- dp->n_ports++;
dp->serial++;
return 0;
dpif_netdev_port_del(struct dpif *dpif, uint16_t port_no)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
- return port_no == ODPP_LOCAL ? EINVAL : do_del_port(dp, port_no);
+ return port_no == OVSP_LOCAL ? EINVAL : do_del_port(dp, port_no);
}
static bool
list_remove(&port->node);
dp->ports[port->port_no] = NULL;
- dp->n_ports--;
dp->serial++;
name = xstrdup(netdev_get_name(port->netdev));
}
static void
-get_odp_flow_stats(struct dp_netdev_flow *flow, struct odp_flow_stats *stats)
+get_dpif_flow_stats(struct dp_netdev_flow *flow, struct dpif_flow_stats *stats)
{
stats->n_packets = flow->packet_count;
stats->n_bytes = flow->byte_count;
- stats->used_sec = flow->used.tv_sec;
- stats->used_nsec = flow->used.tv_nsec;
+ stats->used = flow->used;
stats->tcp_flags = TCP_FLAGS(flow->tcp_ctl);
- stats->reserved = 0;
}
static int
return EINVAL;
}
+ if (flow->in_port < OFPP_MAX
+ ? flow->in_port >= MAX_PORTS
+ : flow->in_port != OFPP_LOCAL && flow->in_port != OFPP_NONE) {
+ return EINVAL;
+ }
+
return 0;
}
static int
-dpif_netdev_flow_get(const struct dpif *dpif, int flags,
+dpif_netdev_flow_get(const struct dpif *dpif,
const struct nlattr *nl_key, size_t nl_key_len,
- struct ofpbuf **actionsp, struct odp_flow_stats *stats)
+ struct ofpbuf **actionsp, struct dpif_flow_stats *stats)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
struct dp_netdev_flow *flow;
}
if (stats) {
- get_odp_flow_stats(flow, stats);
+ get_dpif_flow_stats(flow, stats);
}
if (actionsp) {
*actionsp = ofpbuf_clone_data(flow->actions, flow->actions_len);
}
- if (flags & ODPFF_ZERO_TCP_FLAGS) {
- flow->tcp_ctl = 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 ODPAT_OUTPUT:
- if (nl_attr_get_u32(a) >= MAX_PORTS) {
- return EINVAL;
- }
- break;
-
- case ODPAT_CONTROLLER:
- case ODPAT_DROP_SPOOFED_ARP:
- break;
-
- case ODPAT_SET_DL_TCI:
- *mutates = true;
- if (nl_attr_get_be16(a) & htons(VLAN_CFI)) {
- return EINVAL;
- }
- break;
-
- case ODPAT_SET_NW_TOS:
- *mutates = true;
- if (nl_attr_get_u8(a) & IP_ECN_MASK) {
- return EINVAL;
- }
- break;
-
- case ODPAT_STRIP_VLAN:
- case ODPAT_SET_DL_SRC:
- case ODPAT_SET_DL_DST:
- case ODPAT_SET_NW_SRC:
- case ODPAT_SET_NW_DST:
- case ODPAT_SET_TP_SRC:
- case ODPAT_SET_TP_DST:
- *mutates = true;
- break;
-
- case ODPAT_SET_TUNNEL:
- case ODPAT_SET_PRIORITY:
- case ODPAT_POP_PRIORITY:
- default:
- return EOPNOTSUPP;
- }
- }
return 0;
}
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);
static void
clear_stats(struct dp_netdev_flow *flow)
{
- flow->used.tv_sec = 0;
- flow->used.tv_nsec = 0;
+ flow->used = 0;
flow->packet_count = 0;
flow->byte_count = 0;
flow->tcp_ctl = 0;
}
static int
-dpif_netdev_flow_put(struct dpif *dpif, int flags,
+dpif_netdev_flow_put(struct dpif *dpif, enum dpif_flow_put_flags flags,
const struct nlattr *nl_key, size_t nl_key_len,
const struct nlattr *actions, size_t actions_len,
- struct odp_flow_stats *stats)
+ struct dpif_flow_stats *stats)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
struct dp_netdev_flow *flow;
flow = dp_netdev_lookup_flow(dp, &key);
if (!flow) {
- if (flags & ODPPF_CREATE) {
+ if (flags & DPIF_FP_CREATE) {
if (hmap_count(&dp->flow_table) < MAX_FLOWS) {
if (stats) {
memset(stats, 0, sizeof *stats);
return ENOENT;
}
} else {
- if (flags & ODPPF_MODIFY) {
+ if (flags & DPIF_FP_MODIFY) {
int error = set_flow_actions(flow, actions, actions_len);
if (!error) {
if (stats) {
- get_odp_flow_stats(flow, stats);
+ get_dpif_flow_stats(flow, stats);
}
- if (flags & ODPPF_ZERO_STATS) {
+ if (flags & DPIF_FP_ZERO_STATS) {
clear_stats(flow);
}
}
static int
dpif_netdev_flow_del(struct dpif *dpif,
const struct nlattr *nl_key, size_t nl_key_len,
- struct odp_flow_stats *stats)
+ struct dpif_flow_stats *stats)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
struct dp_netdev_flow *flow;
flow = dp_netdev_lookup_flow(dp, &key);
if (flow) {
if (stats) {
- get_odp_flow_stats(flow, stats);
+ get_dpif_flow_stats(flow, stats);
}
dp_netdev_free_flow(dp, flow);
return 0;
uint32_t bucket;
uint32_t offset;
struct nlattr *actions;
- uint32_t keybuf[ODPUTIL_FLOW_KEY_U32S];
- struct odp_flow_stats stats;
+ struct odputil_keybuf keybuf;
+ struct dpif_flow_stats stats;
};
static int
dpif_netdev_flow_dump_next(const struct dpif *dpif, void *state_,
const struct nlattr **key, size_t *key_len,
const struct nlattr **actions, size_t *actions_len,
- const struct odp_flow_stats **stats)
+ const struct dpif_flow_stats **stats)
{
struct dp_netdev_flow_state *state = state_;
struct dp_netdev *dp = get_dp_netdev(dpif);
if (key) {
struct ofpbuf buf;
- ofpbuf_use_stack(&buf, state->keybuf, sizeof state->keybuf);
+ ofpbuf_use_stack(&buf, &state->keybuf, sizeof state->keybuf);
odp_flow_key_from_flow(&buf, &flow->key);
- assert(buf.base == state->keybuf);
*key = buf.data;
*key_len = buf.size;
}
if (stats) {
- get_odp_flow_stats(flow, &state->stats);
+ get_dpif_flow_stats(flow, &state->stats);
*stats = &state->stats;
}
static int
dpif_netdev_execute(struct dpif *dpif,
+ const struct nlattr *key_attrs, size_t key_len,
const struct nlattr *actions, size_t actions_len,
const struct ofpbuf *packet)
{
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);
- error = dp_netdev_execute_actions(dp, ©, &key, actions, actions_len);
- if (mutates) {
- ofpbuf_uninit(©);
+ error = dpif_netdev_flow_from_nlattrs(key_attrs, key_len, &key);
+ if (!error) {
+ dp_netdev_execute_actions(dp, ©, &key,
+ actions, actions_len);
}
+
+ ofpbuf_uninit(©);
return error;
}
dpif_netdev_recv_set_mask(struct dpif *dpif, int listen_mask)
{
struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif);
- if (!(listen_mask & ~ODPL_ALL)) {
- dpif_netdev->listen_mask = listen_mask;
- return 0;
- } else {
- return EINVAL;
- }
+ dpif_netdev->listen_mask = listen_mask;
+ return 0;
}
static struct dp_netdev_queue *
dp_netdev_flow_used(struct dp_netdev_flow *flow, struct flow *key,
const struct ofpbuf *packet)
{
- time_timespec(&flow->used);
+ flow->used = time_msec();
flow->packet_count++;
flow->byte_count += packet->size;
if (key->dl_type == htons(ETH_TYPE_IP) && key->nw_proto == IPPROTO_TCP) {
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, port->port_no, &key);
flow = dp_netdev_lookup_flow(dp, &key);
if (flow) {
dp_netdev_flow_used(flow, &key, packet);
dp->n_hit++;
} else {
dp->n_missed++;
- dp_netdev_output_control(dp, packet, _ODPL_MISS_NR, &key, 0);
+ dp_netdev_output_userspace(dp, packet, DPIF_UC_MISS, &key, 0);
}
}
static void
-dp_netdev_run(void)
+dpif_netdev_run(struct dpif *dpif)
{
- struct shash_node *node;
+ struct dp_netdev *dp = get_dp_netdev(dpif);
+ struct dp_netdev_port *port;
struct ofpbuf packet;
ofpbuf_init(&packet, DP_NETDEV_HEADROOM + VLAN_ETH_HEADER_LEN + max_mtu);
- SHASH_FOR_EACH (node, &dp_netdevs) {
- struct dp_netdev *dp = node->data;
- struct dp_netdev_port *port;
-
- LIST_FOR_EACH (port, node, &dp->port_list) {
- int error;
- /* Reset packet contents. */
- ofpbuf_clear(&packet);
- ofpbuf_reserve(&packet, DP_NETDEV_HEADROOM);
-
- error = netdev_recv(port->netdev, &packet);
- if (!error) {
- dp_netdev_port_input(dp, port, &packet);
- } else if (error != EAGAIN && error != EOPNOTSUPP) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_ERR_RL(&rl, "error receiving data from %s: %s",
- netdev_get_name(port->netdev), strerror(error));
- }
+ LIST_FOR_EACH (port, node, &dp->port_list) {
+ int error;
+
+ /* Reset packet contents. */
+ ofpbuf_clear(&packet);
+ ofpbuf_reserve(&packet, DP_NETDEV_HEADROOM);
+
+ error = netdev_recv(port->netdev, &packet);
+ if (!error) {
+ dp_netdev_port_input(dp, port, &packet);
+ } else if (error != EAGAIN && error != EOPNOTSUPP) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_ERR_RL(&rl, "error receiving data from %s: %s",
+ netdev_get_name(port->netdev), strerror(error));
}
}
ofpbuf_uninit(&packet);
}
static void
-dp_netdev_wait(void)
+dpif_netdev_wait(struct dpif *dpif)
{
- struct shash_node *node;
-
- SHASH_FOR_EACH (node, &dp_netdevs) {
- struct dp_netdev *dp = node->data;
- struct dp_netdev_port *port;
-
- LIST_FOR_EACH (port, node, &dp->port_list) {
- netdev_recv_wait(port->netdev);
- }
- }
-}
-
-
-/* Modify the TCI field of 'packet'. If a VLAN tag is present, its TCI field
- * is replaced by 'tci'. If a VLAN tag is not present, one is added with the
- * TCI field set to 'tci'.
- */
-static void
-dp_netdev_set_dl_tci(struct ofpbuf *packet, uint16_t tci)
-{
- struct vlan_eth_header *veh;
- struct eth_header *eh;
+ struct dp_netdev *dp = get_dp_netdev(dpif);
+ struct dp_netdev_port *port;
- eh = packet->l2;
- if (packet->size >= sizeof(struct vlan_eth_header)
- && eh->eth_type == htons(ETH_TYPE_VLAN)) {
- veh = packet->l2;
- veh->veth_tci = tci;
- } else {
- /* Insert new 802.1Q header. */
- struct vlan_eth_header tmp;
- memcpy(tmp.veth_dst, eh->eth_dst, ETH_ADDR_LEN);
- memcpy(tmp.veth_src, eh->eth_src, ETH_ADDR_LEN);
- tmp.veth_type = htons(ETH_TYPE_VLAN);
- tmp.veth_tci = tci;
- tmp.veth_next_type = eh->eth_type;
-
- veh = ofpbuf_push_uninit(packet, VLAN_HEADER_LEN);
- memcpy(veh, &tmp, sizeof tmp);
- packet->l2 = (char*)packet->l2 - VLAN_HEADER_LEN;
+ LIST_FOR_EACH (port, node, &dp->port_list) {
+ netdev_recv_wait(port->netdev);
}
}
static void
-dp_netdev_strip_vlan(struct ofpbuf *packet)
+dp_netdev_pop_vlan(struct ofpbuf *packet)
{
struct vlan_eth_header *veh = packet->l2;
if (packet->size >= sizeof *veh
}
static void
-dp_netdev_set_dl_src(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_src, dl_addr, sizeof eh->eth_src);
+
+ 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
-dp_netdev_set_dl_dst(struct ofpbuf *packet, const uint8_t dl_addr[ETH_ADDR_LEN])
+dp_netdev_set_ip_addr(struct ofpbuf *packet, ovs_be32 *addr, ovs_be32 new_addr)
{
- struct eth_header *eh = packet->l2;
- memcpy(eh->eth_dst, dl_addr, sizeof eh->eth_dst);
+ struct ip_header *nh = packet->l3;
+
+ if (nh->ip_proto == IPPROTO_TCP && packet->l7) {
+ struct tcp_header *th = packet->l4;
+ th->tcp_csum = recalc_csum32(th->tcp_csum, *addr, new_addr);
+ } else if (nh->ip_proto == IPPROTO_UDP && packet->l7) {
+ struct udp_header *uh = packet->l4;
+ if (uh->udp_csum) {
+ uh->udp_csum = recalc_csum32(uh->udp_csum, *addr, new_addr);
+ if (!uh->udp_csum) {
+ uh->udp_csum = htons(0xffff);
+ }
+ }
+ }
+ nh->ip_csum = recalc_csum32(nh->ip_csum, *addr, new_addr);
+ *addr = new_addr;
}
-static bool
-is_ip(const struct ofpbuf *packet, const struct flow *key)
+static void
+dp_netdev_set_ip_tos(struct ip_header *nh, uint8_t new_tos)
{
- return key->dl_type == htons(ETH_TYPE_IP) && packet->l4;
+ uint8_t *field = &nh->ip_tos;
+
+ /* Set the DSCP bits and preserve the ECN bits. */
+ uint8_t new = new_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_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);
- uint32_t *field;
-
- field = type == ODPAT_SET_NW_SRC ? &nh->ip_src : &nh->ip_dst;
- if (key->nw_proto == IP_TYPE_TCP && packet->l7) {
- struct tcp_header *th = packet->l4;
- th->tcp_csum = recalc_csum32(th->tcp_csum, *field, ip);
- } else if (key->nw_proto == IP_TYPE_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 = 0xffff;
- }
- }
- }
- nh->ip_csum = recalc_csum32(nh->ip_csum, *field, ip);
- *field = ip;
+dp_netdev_set_ipv4(struct ofpbuf *packet, const struct ovs_key_ipv4 *ipv4_key)
+{
+ struct ip_header *nh = packet->l3;
+
+ if (nh->ip_src != ipv4_key->ipv4_src) {
+ dp_netdev_set_ip_addr(packet, &nh->ip_src, ipv4_key->ipv4_src);
+ }
+ if (nh->ip_dst != ipv4_key->ipv4_dst) {
+ dp_netdev_set_ip_addr(packet, &nh->ip_dst, ipv4_key->ipv4_dst);
+ }
+ if (nh->ip_tos != ipv4_key->ipv4_tos) {
+ dp_netdev_set_ip_tos(nh, ipv4_key->ipv4_tos);
}
}
static void
-dp_netdev_set_nw_tos(struct ofpbuf *packet, const struct flow *key,
- uint8_t nw_tos)
+dp_netdev_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum)
{
- if (is_ip(packet, key)) {
- struct ip_header *nh = packet->l3;
- uint8_t *field = &nh->ip_tos;
+ *csum = recalc_csum16(*csum, *port, new_port);
+ *port = new_port;
+}
- /* Set the DSCP bits and preserve the ECN bits. */
- uint8_t new = nw_tos | (nh->ip_tos & IP_ECN_MASK);
+static void
+dp_netdev_set_tcp_port(struct ofpbuf *packet, const struct ovs_key_tcp *tcp_key)
+{
+ struct tcp_header *th = packet->l4;
- nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t)*field),
- htons((uint16_t) new));
- *field = new;
+ if (th->tcp_src != tcp_key->tcp_src) {
+ dp_netdev_set_port(&th->tcp_src, tcp_key->tcp_src, &th->tcp_csum);
+ }
+ if (th->tcp_dst != tcp_key->tcp_dst) {
+ dp_netdev_set_port(&th->tcp_dst, tcp_key->tcp_dst, &th->tcp_csum);
}
}
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);
- uint16_t *field;
-
- if (key->nw_proto == IPPROTO_TCP && packet->l7) {
- struct tcp_header *th = packet->l4;
- field = type == ODPAT_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 == ODPAT_SET_TP_SRC ? &uh->udp_src : &uh->udp_dst;
- uh->udp_csum = recalc_csum16(uh->udp_csum, *field, port);
- *field = port;
- } else {
- return;
- }
+dp_netdev_set_udp_port(struct ofpbuf *packet, const struct ovs_key_udp *udp_key)
+{
+ struct udp_header *uh = packet->l4;
+
+ if (uh->udp_src != udp_key->udp_src) {
+ dp_netdev_set_port(&uh->udp_src, udp_key->udp_src, &uh->udp_csum);
+ }
+ if (uh->udp_dst != udp_key->udp_dst) {
+ dp_netdev_set_port(&uh->udp_dst, udp_key->udp_dst, &uh->udp_csum);
}
}
}
static int
-dp_netdev_output_control(struct dp_netdev *dp, const struct ofpbuf *packet,
+dp_netdev_output_userspace(struct dp_netdev *dp, const struct ofpbuf *packet,
int queue_no, const struct flow *flow, uint64_t arg)
{
struct dp_netdev_queue *q = &dp->queues[queue_no];
upcall->key_len = key_len;
upcall->userdata = arg;
- q->upcalls[++q->head & QUEUE_MASK] = upcall;
+ q->upcalls[q->head++ & QUEUE_MASK] = upcall;
return 0;
}
-/* Returns true if 'packet' is an invalid Ethernet+IPv4 ARP packet: one with
- * screwy or truncated header fields or one whose inner and outer Ethernet
- * address differ. */
-static bool
-dp_netdev_is_spoofed_arp(struct ofpbuf *packet, const struct flow *key)
+static void
+dp_netdev_sample(struct dp_netdev *dp,
+ struct ofpbuf *packet, struct flow *key,
+ const struct nlattr *action)
{
- struct arp_eth_header *arp;
- struct eth_header *eth;
- ptrdiff_t l3_size;
+ const struct nlattr *subactions = NULL;
+ const struct nlattr *a;
+ size_t left;
- if (key->dl_type != htons(ETH_TYPE_ARP)) {
- return false;
- }
+ NL_NESTED_FOR_EACH_UNSAFE (a, left, action) {
+ int type = nl_attr_type(a);
- l3_size = (char *) ofpbuf_end(packet) - (char *) packet->l3;
- if (l3_size < sizeof(struct arp_eth_header)) {
- return true;
+ 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();
+ }
}
- eth = packet->l2;
- arp = packet->l3;
- return (arp->ar_hrd != htons(ARP_HRD_ETHERNET)
- || arp->ar_pro != htons(ARP_PRO_IP)
- || arp->ar_hln != ETH_HEADER_LEN
- || arp->ar_pln != 4
- || !eth_addr_equals(arp->ar_sha, eth->eth_src));
+ dp_netdev_execute_actions(dp, packet, key, nl_attr_get(subactions),
+ nl_attr_get_size(subactions));
}
-static int
+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);
+ switch (type) {
+ case OVS_KEY_ATTR_TUN_ID:
+ 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:
+ dp_netdev_set_ipv4(packet,
+ nl_attr_get_unspec(a, sizeof(struct ovs_key_ipv4)));
+ break;
+
+ case OVS_KEY_ATTR_TCP:
+ dp_netdev_set_tcp_port(packet,
+ nl_attr_get_unspec(a, sizeof(struct ovs_key_tcp)));
+ break;
+
+ case OVS_KEY_ATTR_UDP:
+ dp_netdev_set_udp_port(packet,
+ nl_attr_get_unspec(a, sizeof(struct ovs_key_udp)));
+ break;
+
+ case OVS_KEY_ATTR_UNSPEC:
+ case OVS_KEY_ATTR_ETHERTYPE:
+ case OVS_KEY_ATTR_IPV6:
+ case OVS_KEY_ATTR_IN_PORT:
+ case OVS_KEY_ATTR_8021Q:
+ 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)) {
- case ODPAT_OUTPUT:
- dp_netdev_output_port(dp, packet, nl_attr_get_u32(a));
- break;
+ const struct nlattr *nested;
+ const struct ovs_key_8021q *q_key;
+ int type = nl_attr_type(a);
- case ODPAT_CONTROLLER:
- dp_netdev_output_control(dp, packet, _ODPL_ACTION_NR,
- key, nl_attr_get_u64(a));
- break;
-
- case ODPAT_SET_DL_TCI:
- dp_netdev_set_dl_tci(packet, nl_attr_get_be16(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 ODPAT_STRIP_VLAN:
- dp_netdev_strip_vlan(packet);
+ case OVS_ACTION_ATTR_USERSPACE:
+ dp_netdev_action_userspace(dp, packet, key, a);
break;
- case ODPAT_SET_DL_SRC:
- dp_netdev_set_dl_src(packet, nl_attr_get_unspec(a, ETH_ADDR_LEN));
+ case OVS_ACTION_ATTR_PUSH:
+ nested = nl_attr_get(a);
+ assert(nl_attr_type(nested) == OVS_KEY_ATTR_8021Q);
+ q_key = nl_attr_get_unspec(nested, sizeof(*q_key));
+ eth_push_vlan(packet, q_key->q_tci);
break;
- case ODPAT_SET_DL_DST:
- dp_netdev_set_dl_dst(packet, nl_attr_get_unspec(a, ETH_ADDR_LEN));
+ case OVS_ACTION_ATTR_POP:
+ assert(nl_attr_get_u16(a) == OVS_KEY_ATTR_8021Q);
+ dp_netdev_pop_vlan(packet);
break;
- case ODPAT_SET_NW_SRC:
- case ODPAT_SET_NW_DST:
- dp_netdev_set_nw_addr(packet, key, a);
+ case OVS_ACTION_ATTR_SET:
+ execute_set_action(packet, nl_attr_get(a));
break;
- case ODPAT_SET_NW_TOS:
- dp_netdev_set_nw_tos(packet, key, nl_attr_get_u8(a));
+ case OVS_ACTION_ATTR_SAMPLE:
+ dp_netdev_sample(dp, packet, key, a);
break;
- case ODPAT_SET_TP_SRC:
- case ODPAT_SET_TP_DST:
- dp_netdev_set_tp_port(packet, key, a);
+ case OVS_ACTION_ATTR_SET_PRIORITY:
+ case OVS_ACTION_ATTR_POP_PRIORITY:
+ /* not implemented */
break;
- case ODPAT_DROP_SPOOFED_ARP:
- if (dp_netdev_is_spoofed_arp(packet, key)) {
- return 0;
- }
+ case OVS_ACTION_ATTR_UNSPEC:
+ case __OVS_ACTION_ATTR_MAX:
+ NOT_REACHED();
}
}
- return 0;
}
const struct dpif_class dpif_netdev_class = {
"netdev",
- dp_netdev_run,
- dp_netdev_wait,
NULL, /* enumerate */
dpif_netdev_open,
dpif_netdev_close,
- NULL, /* get_all_names */
dpif_netdev_destroy,
+ 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, /* get_sflow_probability */
- NULL, /* set_sflow_probability */
NULL, /* queue_to_priority */
dpif_netdev_recv,
dpif_netdev_recv_wait,
git://git.onelab.eu / sliver-openvswitch.git / blobdiff