/*
- * Copyright (c) 2009, 2010, 2011 Nicira Networks.
+ * Copyright (c) 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#include <config.h>
#include "dpif.h"
-#include <assert.h>
#include <ctype.h>
#include <errno.h>
#include <fcntl.h>
#include "hmap.h"
#include "list.h"
#include "netdev.h"
+#include "netdev-vport.h"
#include "netlink.h"
#include "odp-util.h"
#include "ofp-print.h"
#include "ofpbuf.h"
#include "packets.h"
#include "poll-loop.h"
+#include "random.h"
#include "shash.h"
+#include "sset.h"
#include "timeval.h"
#include "util.h"
#include "vlog.h"
enum { QUEUE_MASK = MAX_QUEUE_LEN - 1 };
BUILD_ASSERT_DECL(IS_POW2(MAX_QUEUE_LEN));
+struct dp_netdev_upcall {
+ struct dpif_upcall upcall; /* Queued upcall information. */
+ struct ofpbuf buf; /* ofpbuf instance for upcall.packet. */
+};
+
struct dp_netdev_queue {
- struct dpif_upcall *upcalls[MAX_QUEUE_LEN];
+ struct dp_netdev_upcall upcalls[MAX_QUEUE_LEN];
unsigned int head, tail;
};
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. */
int port_no; /* Index into dp_netdev's 'ports'. */
struct list node; /* Element in dp_netdev's 'port_list'. */
struct netdev *netdev;
- bool internal; /* Internal port? */
+ char *type; /* Port type as requested by user. */
};
/* A flow in dp_netdev's 'flow_table'. */
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. */
- ovs_be16 tcp_ctl; /* Bitwise-OR of seen tcp_ctl values. */
+ uint8_t tcp_flags; /* Bitwise-OR of seen tcp_flags values. */
/* Actions. */
struct nlattr *actions;
struct dpif_netdev {
struct dpif dpif;
struct dp_netdev *dp;
- int listen_mask;
unsigned int dp_serial;
};
/* Maximum port MTU seen so far. */
static int max_mtu = ETH_PAYLOAD_MAX;
-static int get_port_by_number(struct dp_netdev *, uint16_t port_no,
+static int get_port_by_number(struct dp_netdev *, uint32_t port_no,
struct dp_netdev_port **portp);
static int get_port_by_name(struct dp_netdev *, const char *devname,
struct dp_netdev_port **portp);
static void dp_netdev_free(struct dp_netdev *);
static void dp_netdev_flow_flush(struct dp_netdev *);
static int do_add_port(struct dp_netdev *, const char *devname,
- const char *type, uint16_t port_no);
-static int do_del_port(struct dp_netdev *, uint16_t port_no);
+ const char *type, uint32_t port_no);
+static int do_del_port(struct dp_netdev *, uint32_t port_no);
static int dpif_netdev_open(const struct dpif_class *, const char *name,
bool create, struct dpif **);
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 struct dpif_class dpif_dummy_class;
+ const struct nlattr *userdata);
+static void dp_netdev_execute_actions(struct dp_netdev *,
+ struct ofpbuf *, struct flow *,
+ const struct nlattr *actions,
+ size_t actions_len);
static struct dpif_netdev *
dpif_netdev_cast(const struct dpif *dpif)
{
- assert(dpif->dpif_class->open == dpif_netdev_open);
+ ovs_assert(dpif->dpif_class->open == dpif_netdev_open);
return CONTAINER_OF(dpif, struct dpif_netdev, dpif);
}
return dpif_netdev_cast(dpif)->dp;
}
+static int
+dpif_netdev_enumerate(struct sset *all_dps)
+{
+ struct shash_node *node;
+
+ SHASH_FOR_EACH(node, &dp_netdevs) {
+ sset_add(all_dps, node->name);
+ }
+ return 0;
+}
+
+static bool
+dpif_netdev_class_is_dummy(const struct dpif_class *class)
+{
+ return class != &dpif_netdev_class;
+}
+
+static bool
+dpif_netdev_class_is_planetlab(const struct dpif_class *class)
+{
+ return class == &dpif_planetlab_class;
+}
+
+static const char *
+dpif_netdev_port_open_type(const struct dpif_class *class, const char *type)
+{
+ return strcmp(type, "internal") ? type
+ : dpif_netdev_class_is_planetlab(class) ? "pltap"
+ : dpif_netdev_class_is_dummy(class) ? "dummy"
+ : "tap";
+}
+
static struct dpif *
create_dpif_netdev(struct dp_netdev *dp)
{
dpif = xmalloc(sizeof *dpif);
dpif_init(&dpif->dpif, dp->class, dp->name, netflow_id >> 8, netflow_id);
dpif->dp = dp;
- dpif->listen_mask = 0;
dpif->dp_serial = dp->serial;
return &dpif->dpif;
}
+static int
+choose_port(struct dp_netdev *dp, const char *name)
+{
+ int port_no;
+
+ if (dp->class != &dpif_netdev_class &&
+ dp->class != &dpif_planetlab_class) {
+ const char *p;
+ int start_no = 0;
+
+ /* If the port name begins with "br", start the number search at
+ * 100 to make writing tests easier. */
+ if (!strncmp(name, "br", 2)) {
+ start_no = 100;
+ }
+
+ /* 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. */
+ for (p = name; *p != '\0'; p++) {
+ if (isdigit((unsigned char) *p)) {
+ port_no = start_no + strtol(p, NULL, 10);
+ if (port_no > 0 && port_no < MAX_PORTS
+ && !dp->ports[port_no]) {
+ return port_no;
+ }
+ break;
+ }
+ }
+ }
+
+ for (port_no = 1; port_no < MAX_PORTS; port_no++) {
+ if (!dp->ports[port_no]) {
+ return port_no;
+ }
+ }
+
+ return -1;
+}
+
static int
create_dp_netdev(const char *name, const struct dpif_class *class,
struct dp_netdev **dpp)
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", OVSP_LOCAL);
if (error) {
dp_netdev_free(dp);
if (error) {
return error;
}
- assert(dp != NULL);
+ ovs_assert(dp != NULL);
}
} else {
if (dp->class != class) {
struct dp_netdev_queue *q = &dp->queues[i];
while (q->tail != q->head) {
- struct dpif_upcall *upcall = q->upcalls[q->tail++ & QUEUE_MASK];
-
- ofpbuf_delete(upcall->packet);
- free(upcall);
+ struct dp_netdev_upcall *u = &q->upcalls[q->tail++ & QUEUE_MASK];
+ ofpbuf_uninit(&u->buf);
}
}
}
dpif_netdev_close(struct dpif *dpif)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
- assert(dp->open_cnt > 0);
+ ovs_assert(dp->open_cnt > 0);
if (--dp->open_cnt == 0 && dp->destroyed) {
shash_find_and_delete(&dp_netdevs, dp->name);
dp_netdev_free(dp);
{
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)
+ uint32_t port_no)
{
struct dp_netdev_port *port;
struct netdev *netdev;
- bool internal;
+ const char *open_type;
int mtu;
int error;
/* XXX reject devices already in some dp_netdev. */
- if (type[0] == '\0' || !strcmp(type, "system") || !strcmp(type, "dummy")) {
- internal = false;
- } else if (!strcmp(type, "internal")) {
- internal = true;
- } else {
- VLOG_WARN("%s: unsupported port type %s", devname, type);
- return EINVAL;
- }
/* Open and validate network device. */
- if (dp->class == &dpif_dummy_class) {
- type = "dummy";
- } else if (internal) {
- type = "tap";
- }
-
- error = netdev_open(devname, type, &netdev);
+ open_type = dpif_netdev_port_open_type(dp->class, type);
+ error = netdev_open(devname, open_type, &netdev);
if (error) {
return error;
}
/* XXX reject non-Ethernet devices */
error = netdev_listen(netdev);
- if (error) {
+ if (error
+ && !(error == EOPNOTSUPP && dpif_netdev_class_is_dummy(dp->class))) {
VLOG_ERR("%s: cannot receive packets on this network device (%s)",
devname, strerror(errno));
netdev_close(netdev);
port = xmalloc(sizeof *port);
port->port_no = port_no;
port->netdev = netdev;
- port->internal = internal;
+ port->type = xstrdup(type);
error = netdev_get_mtu(netdev, &mtu);
if (!error) {
static int
dpif_netdev_port_add(struct dpif *dpif, struct netdev *netdev,
- uint16_t *port_nop)
+ uint32_t *port_nop)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
int port_no;
- 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);
+ if (*port_nop != UINT32_MAX) {
+ if (*port_nop >= MAX_PORTS) {
+ return EFBIG;
+ } else if (dp->ports[*port_nop]) {
+ return EBUSY;
}
+ port_no = *port_nop;
+ } else {
+ port_no = choose_port(dp, netdev_vport_get_dpif_port(netdev));
+ }
+ if (port_no >= 0) {
+ *port_nop = port_no;
+ return do_add_port(dp, netdev_vport_get_dpif_port(netdev),
+ netdev_get_type(netdev), port_no);
}
return EFBIG;
}
static int
-dpif_netdev_port_del(struct dpif *dpif, uint16_t port_no)
+dpif_netdev_port_del(struct dpif *dpif, uint32_t port_no)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
return port_no == OVSP_LOCAL ? EINVAL : do_del_port(dp, port_no);
}
static bool
-is_valid_port_number(uint16_t port_no)
+is_valid_port_number(uint32_t port_no)
{
return port_no < MAX_PORTS;
}
static int
get_port_by_number(struct dp_netdev *dp,
- uint16_t port_no, struct dp_netdev_port **portp)
+ uint32_t port_no, struct dp_netdev_port **portp)
{
if (!is_valid_port_number(port_no)) {
*portp = NULL;
struct dp_netdev_port *port;
LIST_FOR_EACH (port, node, &dp->port_list) {
- if (!strcmp(netdev_get_name(port->netdev), devname)) {
+ if (!strcmp(netdev_vport_get_dpif_port(port->netdev), devname)) {
*portp = port;
return 0;
}
}
static int
-do_del_port(struct dp_netdev *dp, uint16_t port_no)
+do_del_port(struct dp_netdev *dp, uint32_t port_no)
{
struct dp_netdev_port *port;
char *name;
dp->ports[port->port_no] = NULL;
dp->serial++;
- name = xstrdup(netdev_get_name(port->netdev));
+ name = xstrdup(netdev_vport_get_dpif_port(port->netdev));
netdev_close(port->netdev);
+ free(port->type);
free(name);
free(port);
answer_port_query(const struct dp_netdev_port *port,
struct dpif_port *dpif_port)
{
- dpif_port->name = xstrdup(netdev_get_name(port->netdev));
- dpif_port->type = xstrdup(port->internal ? "internal" : "system");
+ dpif_port->name = xstrdup(netdev_vport_get_dpif_port(port->netdev));
+ dpif_port->type = xstrdup(port->type);
dpif_port->port_no = port->port_no;
}
static int
-dpif_netdev_port_query_by_number(const struct dpif *dpif, uint16_t port_no,
+dpif_netdev_port_query_by_number(const struct dpif *dpif, uint32_t port_no,
struct dpif_port *dpif_port)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
int error;
error = get_port_by_number(dp, port_no, &port);
- if (!error) {
+ if (!error && dpif_port) {
answer_port_query(port, dpif_port);
}
return error;
int error;
error = get_port_by_name(dp, devname, &port);
- if (!error) {
+ if (!error && dpif_port) {
answer_port_query(port, dpif_port);
}
return error;
struct dp_netdev_port *port = dp->ports[port_no];
if (port) {
free(state->name);
- state->name = xstrdup(netdev_get_name(port->netdev));
+ state->name = xstrdup(netdev_vport_get_dpif_port(port->netdev));
dpif_port->name = state->name;
- dpif_port->type = port->internal ? "internal" : "system";
+ dpif_port->type = port->type;
dpif_port->port_no = port->port_no;
state->port_no = port_no + 1;
return 0;
stats->n_packets = flow->packet_count;
stats->n_bytes = flow->byte_count;
stats->used = flow->used;
- stats->tcp_flags = TCP_FLAGS(flow->tcp_ctl);
+ stats->tcp_flags = flow->tcp_flags;
}
static int
dpif_netdev_flow_from_nlattrs(const struct nlattr *key, uint32_t key_len,
struct flow *flow)
{
- if (odp_flow_key_to_flow(key, key_len, flow)) {
+ if (odp_flow_key_to_flow(key, key_len, flow) != ODP_FIT_PERFECT) {
/* This should not happen: it indicates that odp_flow_key_from_flow()
* and odp_flow_key_to_flow() disagree on the acceptable form of a
* flow. Log the problem as an error, with enough details to enable
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);
}
static int
-add_flow(struct dpif *dpif, const struct flow *key,
- const struct nlattr *actions, size_t actions_len)
+dp_netdev_flow_add(struct dp_netdev *dp, const struct flow *key,
+ const struct nlattr *actions, size_t actions_len)
{
- struct dp_netdev *dp = get_dp_netdev(dpif);
struct dp_netdev_flow *flow;
int error;
flow->used = 0;
flow->packet_count = 0;
flow->byte_count = 0;
- flow->tcp_ctl = 0;
+ flow->tcp_flags = 0;
}
static int
-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 dpif_flow_stats *stats)
+dpif_netdev_flow_put(struct dpif *dpif, const struct dpif_flow_put *put)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
struct dp_netdev_flow *flow;
struct flow key;
int error;
- error = dpif_netdev_flow_from_nlattrs(nl_key, nl_key_len, &key);
+ error = dpif_netdev_flow_from_nlattrs(put->key, put->key_len, &key);
if (error) {
return error;
}
flow = dp_netdev_lookup_flow(dp, &key);
if (!flow) {
- if (flags & DPIF_FP_CREATE) {
+ if (put->flags & DPIF_FP_CREATE) {
if (hmap_count(&dp->flow_table) < MAX_FLOWS) {
- if (stats) {
- memset(stats, 0, sizeof *stats);
+ if (put->stats) {
+ memset(put->stats, 0, sizeof *put->stats);
}
- return add_flow(dpif, &key, actions, actions_len);
+ return dp_netdev_flow_add(dp, &key, put->actions,
+ put->actions_len);
} else {
return EFBIG;
}
return ENOENT;
}
} else {
- if (flags & DPIF_FP_MODIFY) {
- int error = set_flow_actions(flow, actions, actions_len);
+ if (put->flags & DPIF_FP_MODIFY) {
+ int error = set_flow_actions(flow, put->actions, put->actions_len);
if (!error) {
- if (stats) {
- get_dpif_flow_stats(flow, stats);
+ if (put->stats) {
+ get_dpif_flow_stats(flow, put->stats);
}
- if (flags & DPIF_FP_ZERO_STATS) {
+ if (put->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 dpif_flow_stats *stats)
+dpif_netdev_flow_del(struct dpif *dpif, const struct dpif_flow_del *del)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
struct dp_netdev_flow *flow;
struct flow key;
int error;
- error = dpif_netdev_flow_from_nlattrs(nl_key, nl_key_len, &key);
+ error = dpif_netdev_flow_from_nlattrs(del->key, del->key_len, &key);
if (error) {
return error;
}
flow = dp_netdev_lookup_flow(dp, &key);
if (flow) {
- if (stats) {
- get_dpif_flow_stats(flow, stats);
+ if (del->stats) {
+ get_dpif_flow_stats(flow, del->stats);
}
dp_netdev_free_flow(dp, flow);
return 0;
struct ofpbuf buf;
ofpbuf_use_stack(&buf, &state->keybuf, sizeof state->keybuf);
- odp_flow_key_from_flow(&buf, &flow->key);
+ odp_flow_key_from_flow(&buf, &flow->key, flow->key.in_port);
*key = buf.data;
*key_len = buf.size;
}
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)
+dpif_netdev_execute(struct dpif *dpif, const struct dpif_execute *execute)
{
struct dp_netdev *dp = get_dp_netdev(dpif);
struct ofpbuf copy;
- bool mutates;
struct flow key;
int error;
- if (packet->size < ETH_HEADER_LEN || packet->size > UINT16_MAX) {
+ if (execute->packet->size < ETH_HEADER_LEN ||
+ execute->packet->size > UINT16_MAX) {
return EINVAL;
}
- error = dpif_netdev_validate_actions(actions, actions_len, &mutates);
- if (error) {
- return error;
- }
-
- 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;
- }
+ /* Make a deep copy of 'packet', because we might modify its data. */
+ ofpbuf_init(©, DP_NETDEV_HEADROOM + execute->packet->size);
+ ofpbuf_reserve(©, DP_NETDEV_HEADROOM);
+ ofpbuf_put(©, execute->packet->data, execute->packet->size);
- flow_extract(©, 0, -1, &key);
- error = dpif_netdev_flow_from_nlattrs(key_attrs, key_len, &key);
+ flow_extract(©, 0, 0, NULL, -1, &key);
+ error = dpif_netdev_flow_from_nlattrs(execute->key, execute->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,
+ execute->actions, execute->actions_len);
}
+
+ ofpbuf_uninit(©);
return error;
}
static int
-dpif_netdev_recv_get_mask(const struct dpif *dpif, int *listen_mask)
+dpif_netdev_recv_set(struct dpif *dpif OVS_UNUSED, bool enable OVS_UNUSED)
{
- struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif);
- *listen_mask = dpif_netdev->listen_mask;
return 0;
}
static int
-dpif_netdev_recv_set_mask(struct dpif *dpif, int listen_mask)
+dpif_netdev_queue_to_priority(const struct dpif *dpif OVS_UNUSED,
+ uint32_t queue_id, uint32_t *priority)
{
- struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif);
- dpif_netdev->listen_mask = listen_mask;
+ *priority = queue_id;
return 0;
}
static struct dp_netdev_queue *
find_nonempty_queue(struct dpif *dpif)
{
- struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif);
struct dp_netdev *dp = get_dp_netdev(dpif);
- int mask = dpif_netdev->listen_mask;
int i;
for (i = 0; i < N_QUEUES; i++) {
struct dp_netdev_queue *q = &dp->queues[i];
- if (q->head != q->tail && mask & (1u << i)) {
+ if (q->head != q->tail) {
return q;
}
}
}
static int
-dpif_netdev_recv(struct dpif *dpif, struct dpif_upcall *upcall)
+dpif_netdev_recv(struct dpif *dpif, struct dpif_upcall *upcall,
+ struct ofpbuf *buf)
{
struct dp_netdev_queue *q = find_nonempty_queue(dpif);
if (q) {
- struct dpif_upcall *u = q->upcalls[q->tail++ & QUEUE_MASK];
- *upcall = *u;
- free(u);
+ struct dp_netdev_upcall *u = &q->upcalls[q->tail++ & QUEUE_MASK];
+
+ *upcall = u->upcall;
+ upcall->packet = buf;
+
+ ofpbuf_uninit(buf);
+ *buf = u->buf;
return 0;
} else {
}
\f
static void
-dp_netdev_flow_used(struct dp_netdev_flow *flow, struct flow *key,
- const struct ofpbuf *packet)
+dp_netdev_flow_used(struct dp_netdev_flow *flow, const struct ofpbuf *packet)
{
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) {
- struct tcp_header *th = packet->l4;
- flow->tcp_ctl |= th->tcp_ctl;
- }
+ flow->tcp_flags |= packet_get_tcp_flags(packet, &flow->key);
}
static void
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, NULL, port->port_no, &key);
flow = dp_netdev_lookup_flow(dp, &key);
if (flow) {
- dp_netdev_flow_used(flow, &key, packet);
+ dp_netdev_flow_used(flow, packet);
dp_netdev_execute_actions(dp, packet, &key,
flow->actions, flow->actions_len);
dp->n_hit++;
} else {
dp->n_missed++;
- dp_netdev_output_userspace(dp, packet, DPIF_UC_MISS, &key, 0);
+ dp_netdev_output_userspace(dp, packet, DPIF_UC_MISS, &key, NULL);
}
}
} 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));
+ netdev_vport_get_dpif_port(port->netdev),
+ strerror(error));
}
}
ofpbuf_uninit(&packet);
}
static void
-dp_netdev_pop_vlan(struct ofpbuf *packet)
+dp_netdev_set_dl(struct ofpbuf *packet, const struct ovs_key_ethernet *eth_key)
{
- 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;
+ struct eth_header *eh = packet->l2;
- ofpbuf_pull(packet, VLAN_HEADER_LEN);
- packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN;
- memcpy(packet->data, &tmp, sizeof tmp);
- }
+ 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_src(struct ofpbuf *packet, const uint8_t dl_addr[ETH_ADDR_LEN])
+dp_netdev_output_port(struct dp_netdev *dp, struct ofpbuf *packet,
+ uint32_t out_port)
{
- struct eth_header *eh = packet->l2;
- memcpy(eh->eth_src, dl_addr, sizeof eh->eth_src);
+ struct dp_netdev_port *p = dp->ports[out_port];
+ if (p) {
+ netdev_send(p->netdev, packet);
+ }
}
-static void
-dp_netdev_set_dl_dst(struct ofpbuf *packet, const uint8_t dl_addr[ETH_ADDR_LEN])
+static int
+dp_netdev_output_userspace(struct dp_netdev *dp, const struct ofpbuf *packet,
+ int queue_no, const struct flow *flow,
+ const struct nlattr *userdata)
{
- struct eth_header *eh = packet->l2;
- memcpy(eh->eth_dst, dl_addr, sizeof eh->eth_dst);
-}
+ struct dp_netdev_queue *q = &dp->queues[queue_no];
+ if (q->head - q->tail < MAX_QUEUE_LEN) {
+ struct dp_netdev_upcall *u = &q->upcalls[q->head++ & QUEUE_MASK];
+ struct dpif_upcall *upcall = &u->upcall;
+ struct ofpbuf *buf = &u->buf;
+ size_t buf_size;
+
+ upcall->type = queue_no;
+
+ /* Allocate buffer big enough for everything. */
+ buf_size = ODPUTIL_FLOW_KEY_BYTES + 2 + packet->size;
+ if (userdata) {
+ buf_size += NLA_ALIGN(userdata->nla_len);
+ }
+ ofpbuf_init(buf, buf_size);
-static bool
-is_ip(const struct ofpbuf *packet, const struct flow *key)
-{
- return key->dl_type == htons(ETH_TYPE_IP) && packet->l4;
-}
+ /* Put ODP flow. */
+ odp_flow_key_from_flow(buf, flow, flow->in_port);
+ upcall->key = buf->data;
+ upcall->key_len = buf->size;
-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);
- }
- }
+ /* Put userdata. */
+ if (userdata) {
+ upcall->userdata = ofpbuf_put(buf, userdata,
+ NLA_ALIGN(userdata->nla_len));
}
- nh->ip_csum = recalc_csum32(nh->ip_csum, *field, ip);
- *field = ip;
+
+ /* Put packet.
+ *
+ * We adjust 'data' and 'size' in 'buf' so that only the packet itself
+ * is visible in 'upcall->packet'. The ODP flow and (if present)
+ * userdata become part of the headroom. */
+ ofpbuf_put_zeros(buf, 2);
+ buf->data = ofpbuf_put(buf, packet->data, packet->size);
+ buf->size = packet->size;
+ upcall->packet = buf;
+
+ return 0;
+ } else {
+ dp->n_lost++;
+ return ENOBUFS;
}
}
static void
-dp_netdev_set_nw_tos(struct ofpbuf *packet, const struct flow *key,
- uint8_t nw_tos)
+dp_netdev_sample(struct dp_netdev *dp,
+ struct ofpbuf *packet, struct flow *key,
+ const struct nlattr *action)
{
- if (is_ip(packet, key)) {
- struct ip_header *nh = packet->l3;
- uint8_t *field = &nh->ip_tos;
+ const struct nlattr *subactions = NULL;
+ const struct nlattr *a;
+ size_t left;
- /* Set the DSCP bits and preserve the ECN bits. */
- uint8_t new = nw_tos | (nh->ip_tos & IP_ECN_MASK);
+ NL_NESTED_FOR_EACH_UNSAFE (a, left, action) {
+ int type = nl_attr_type(a);
- nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t)*field),
- htons((uint16_t) new));
- *field = new;
- }
-}
+ switch ((enum ovs_sample_attr) type) {
+ case OVS_SAMPLE_ATTR_PROBABILITY:
+ if (random_uint32() >= nl_attr_get_u32(a)) {
+ return;
+ }
+ break;
-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;
+ 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_output_port(struct dp_netdev *dp, struct ofpbuf *packet,
- uint16_t out_port)
+dp_netdev_action_userspace(struct dp_netdev *dp,
+ struct ofpbuf *packet, struct flow *key,
+ const struct nlattr *a)
{
- struct dp_netdev_port *p = dp->ports[out_port];
- if (p) {
- netdev_send(p->netdev, packet);
- }
-}
+ const struct nlattr *userdata;
-static int
-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];
- struct dpif_upcall *upcall;
- struct ofpbuf *buf;
- size_t key_len;
+ userdata = nl_attr_find_nested(a, OVS_USERSPACE_ATTR_USERDATA);
+ dp_netdev_output_userspace(dp, packet, DPIF_UC_ACTION, key, userdata);
+}
- if (q->head - q->tail >= MAX_QUEUE_LEN) {
- dp->n_lost++;
- return ENOBUFS;
+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_ipv6 *ipv6_key;
+ const struct ovs_key_tcp *tcp_key;
+ const struct ovs_key_udp *udp_key;
+
+ switch (type) {
+ case OVS_KEY_ATTR_PRIORITY:
+ case OVS_KEY_ATTR_SKB_MARK:
+ case OVS_KEY_ATTR_TUNNEL:
+ /* 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_IPV6:
+ ipv6_key = nl_attr_get_unspec(a, sizeof(struct ovs_key_ipv6));
+ packet_set_ipv6(packet, ipv6_key->ipv6_proto, ipv6_key->ipv6_src,
+ ipv6_key->ipv6_dst, ipv6_key->ipv6_tclass,
+ ipv6_key->ipv6_label, ipv6_key->ipv6_hlimit);
+ 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_MPLS:
+ set_mpls_lse(packet, nl_attr_get_be32(a));
+ 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();
}
-
- buf = ofpbuf_new(ODPUTIL_FLOW_KEY_BYTES + 2 + packet->size);
- odp_flow_key_from_flow(buf, flow);
- key_len = buf->size;
- ofpbuf_pull(buf, key_len);
- ofpbuf_reserve(buf, 2);
- ofpbuf_put(buf, packet->data, packet->size);
-
- upcall = xzalloc(sizeof *upcall);
- upcall->type = queue_no;
- upcall->packet = buf;
- upcall->key = buf->base;
- upcall->key_len = key_len;
- upcall->userdata = arg;
-
- q->upcalls[q->head++ & QUEUE_MASK] = upcall;
-
- return 0;
}
-static int
+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)) {
+ 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));
+ case OVS_ACTION_ATTR_PUSH_VLAN: {
+ const struct ovs_action_push_vlan *vlan = nl_attr_get(a);
+ eth_push_vlan(packet, vlan->vlan_tci);
break;
+ }
case OVS_ACTION_ATTR_POP_VLAN:
- dp_netdev_pop_vlan(packet);
+ eth_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));
+ case OVS_ACTION_ATTR_PUSH_MPLS: {
+ const struct ovs_action_push_mpls *mpls = nl_attr_get(a);
+ push_mpls(packet, mpls->mpls_ethertype, mpls->mpls_lse);
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_POP_MPLS:
+ pop_mpls(packet, nl_attr_get_be16(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_SET:
+ execute_set_action(packet, nl_attr_get(a));
break;
- case OVS_ACTION_ATTR_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 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;
}
+#define DPIF_NETDEV_CLASS_FUNCTIONS \
+ dpif_netdev_enumerate, \
+ dpif_netdev_port_open_type, \
+ dpif_netdev_open, \
+ dpif_netdev_close, \
+ dpif_netdev_destroy, \
+ dpif_netdev_run, \
+ dpif_netdev_wait, \
+ dpif_netdev_get_stats, \
+ 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_port_poll, \
+ dpif_netdev_port_poll_wait, \
+ dpif_netdev_flow_get, \
+ dpif_netdev_flow_put, \
+ dpif_netdev_flow_del, \
+ dpif_netdev_flow_flush, \
+ dpif_netdev_flow_dump_start, \
+ dpif_netdev_flow_dump_next, \
+ dpif_netdev_flow_dump_done, \
+ dpif_netdev_execute, \
+ NULL, /* operate */ \
+ dpif_netdev_recv_set, \
+ dpif_netdev_queue_to_priority, \
+ dpif_netdev_recv, \
+ dpif_netdev_recv_wait, \
+ dpif_netdev_recv_purge, \
+
const struct dpif_class dpif_netdev_class = {
"netdev",
- NULL, /* enumerate */
- dpif_netdev_open,
- dpif_netdev_close,
- 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,
- dpif_netdev_port_dump_start,
- dpif_netdev_port_dump_next,
- dpif_netdev_port_dump_done,
- dpif_netdev_port_poll,
- dpif_netdev_port_poll_wait,
- dpif_netdev_flow_get,
- dpif_netdev_flow_put,
- dpif_netdev_flow_del,
- dpif_netdev_flow_flush,
- dpif_netdev_flow_dump_start,
- dpif_netdev_flow_dump_next,
- dpif_netdev_flow_dump_done,
- dpif_netdev_execute,
- dpif_netdev_recv_get_mask,
- dpif_netdev_recv_set_mask,
- NULL, /* queue_to_priority */
- dpif_netdev_recv,
- dpif_netdev_recv_wait,
- dpif_netdev_recv_purge,
+ DPIF_NETDEV_CLASS_FUNCTIONS
};
+const struct dpif_class dpif_planetlab_class = {
+ "planetlab",
+ DPIF_NETDEV_CLASS_FUNCTIONS
+};
+
+static void
+dpif_dummy_register__(const char *type)
+{
+ struct dpif_class *class;
+
+ class = xmalloc(sizeof *class);
+ *class = dpif_netdev_class;
+ class->type = xstrdup(type);
+ dp_register_provider(class);
+}
+
void
-dpif_dummy_register(void)
+dpif_dummy_register(bool override)
{
- if (!dpif_dummy_class.type) {
- dpif_dummy_class = dpif_netdev_class;
- dpif_dummy_class.type = "dummy";
- dp_register_provider(&dpif_dummy_class);
+ if (override) {
+ struct sset types;
+ const char *type;
+
+ sset_init(&types);
+ dp_enumerate_types(&types);
+ SSET_FOR_EACH (type, &types) {
+ if (!dp_unregister_provider(type)) {
+ dpif_dummy_register__(type);
+ }
+ }
+ sset_destroy(&types);
}
+
+ dpif_dummy_register__("dummy");
}
+