-/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc.
+/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016 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 "ofproto/ofproto-dpif-sflow.h"
#include "ofproto/ofproto-dpif.h"
#include "ofproto/ofproto-provider.h"
+#include "packets.h"
#include "ovs-router.h"
#include "tnl-ports.h"
#include "tunnel.h"
} group;
struct {
char br_name[IFNAMSIZ];
- ovs_be32 d_ip;
+ struct in6_addr d_ipv6;
} tnl_neigh_cache;
} u;
};
static struct xbridge *xbridge_lookup(struct xlate_cfg *,
const struct ofproto_dpif *);
+static struct xbridge *xbridge_lookup_by_uuid(struct xlate_cfg *,
+ const struct uuid *);
static struct xbundle *xbundle_lookup(struct xlate_cfg *,
const struct ofbundle *);
static struct xport *xport_lookup(struct xlate_cfg *,
return NULL;
}
+static struct xbridge *
+xbridge_lookup_by_uuid(struct xlate_cfg *xcfg, const struct uuid *uuid)
+{
+ struct xbridge *xbridge;
+
+ HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) {
+ if (uuid_equals(ofproto_dpif_get_uuid(xbridge->ofproto), uuid)) {
+ return xbridge;
+ }
+ }
+ return NULL;
+}
+
static struct xbundle *
xbundle_lookup(struct xlate_cfg *xcfg, const struct ofbundle *ofbundle)
{
static int
tnl_route_lookup_flow(const struct flow *oflow,
- ovs_be32 *ip, struct xport **out_port)
+ struct in6_addr *ip, struct xport **out_port)
{
char out_dev[IFNAMSIZ];
struct xbridge *xbridge;
struct xlate_cfg *xcfg;
- ovs_be32 gw;
+ struct in6_addr gw;
+ struct in6_addr dst;
- if (!ovs_router_lookup4(oflow->tunnel.ip_dst, out_dev, &gw)) {
+ dst = flow_tnl_dst(&oflow->tunnel);
+ if (!ovs_router_lookup(&dst, out_dev, &gw)) {
return -ENOENT;
}
- if (gw) {
+ if (ipv6_addr_is_set(&gw) &&
+ (!IN6_IS_ADDR_V4MAPPED(&gw) || in6_addr_get_mapped_ipv4(&gw))) {
*ip = gw;
} else {
- *ip = oflow->tunnel.ip_dst;
+ *ip = dst;
}
xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp);
ctx->recurse, ctx->resubmits, packet);
}
+static void
+tnl_send_nd_request(struct xlate_ctx *ctx, const struct xport *out_dev,
+ const struct eth_addr eth_src,
+ struct in6_addr * ipv6_src, struct in6_addr * ipv6_dst)
+{
+ struct dp_packet packet;
+
+ dp_packet_init(&packet, 0);
+ compose_nd(&packet, eth_src, ipv6_src, ipv6_dst);
+ compose_table_xlate(ctx, out_dev, &packet);
+ dp_packet_uninit(&packet);
+}
+
static void
tnl_send_arp_request(struct xlate_ctx *ctx, const struct xport *out_dev,
const struct eth_addr eth_src,
{
struct ovs_action_push_tnl tnl_push_data;
struct xport *out_dev = NULL;
- ovs_be32 s_ip, d_ip = 0;
+ ovs_be32 s_ip = 0, d_ip = 0;
+ struct in6_addr s_ip6 = in6addr_any;
+ struct in6_addr d_ip6 = in6addr_any;
struct eth_addr smac;
struct eth_addr dmac;
int err;
+ char buf_sip6[INET6_ADDRSTRLEN];
+ char buf_dip6[INET6_ADDRSTRLEN];
- err = tnl_route_lookup_flow(flow, &d_ip, &out_dev);
+ err = tnl_route_lookup_flow(flow, &d_ip6, &out_dev);
if (err) {
xlate_report(ctx, "native tunnel routing failed");
return err;
}
- xlate_report(ctx, "tunneling to "IP_FMT" via %s",
- IP_ARGS(d_ip), netdev_get_name(out_dev->netdev));
+
+ xlate_report(ctx, "tunneling to %s via %s",
+ ipv6_string_mapped(buf_dip6, &d_ip6),
+ netdev_get_name(out_dev->netdev));
/* Use mac addr of bridge port of the peer. */
err = netdev_get_etheraddr(out_dev->netdev, &smac);
return err;
}
- err = netdev_get_in4(out_dev->netdev, (struct in_addr *) &s_ip, NULL);
- if (err) {
- xlate_report(ctx, "tunnel output device lacks IPv4 address");
- return err;
+ d_ip = in6_addr_get_mapped_ipv4(&d_ip6);
+ if (d_ip) {
+ err = netdev_get_in4(out_dev->netdev, (struct in_addr *) &s_ip, NULL);
+ if (err) {
+ xlate_report(ctx, "tunnel output device lacks IPv4 address");
+ return err;
+ }
+ in6_addr_set_mapped_ipv4(&s_ip6, s_ip);
+ } else {
+ err = netdev_get_in6(out_dev->netdev, &s_ip6);
+ if (err) {
+ xlate_report(ctx, "tunnel output device lacks IPv6 address");
+ return err;
+ }
}
- err = tnl_arp_lookup(out_dev->xbridge->name, d_ip, &dmac);
+ err = tnl_neigh_lookup(out_dev->xbridge->name, &d_ip6, &dmac);
if (err) {
- xlate_report(ctx, "ARP cache miss for "IP_FMT" on bridge %s, "
- "sending ARP request",
- IP_ARGS(d_ip), out_dev->xbridge->name);
- tnl_send_arp_request(ctx, out_dev, smac, s_ip, d_ip);
+ xlate_report(ctx, "neighbor cache miss for %s on bridge %s, "
+ "sending %s request",
+ buf_dip6, out_dev->xbridge->name, d_ip ? "ARP" : "ND");
+ if (d_ip) {
+ tnl_send_arp_request(ctx, out_dev, smac, s_ip, d_ip);
+ } else {
+ tnl_send_nd_request(ctx, out_dev, smac, &s_ip6, &d_ip6);
+ }
return err;
}
+
if (ctx->xin->xcache) {
struct xc_entry *entry;
entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TNL_NEIGH);
ovs_strlcpy(entry->u.tnl_neigh_cache.br_name, out_dev->xbridge->name,
sizeof entry->u.tnl_neigh_cache.br_name);
- entry->u.tnl_neigh_cache.d_ip = d_ip;
+ entry->u.tnl_neigh_cache.d_ipv6 = d_ip6;
}
- xlate_report(ctx, "tunneling from "ETH_ADDR_FMT" "IP_FMT
- " to "ETH_ADDR_FMT" "IP_FMT,
- ETH_ADDR_ARGS(smac), IP_ARGS(s_ip),
- ETH_ADDR_ARGS(dmac), IP_ARGS(d_ip));
+ xlate_report(ctx, "tunneling from "ETH_ADDR_FMT" %s"
+ " to "ETH_ADDR_FMT" %s",
+ ETH_ADDR_ARGS(smac), ipv6_string_mapped(buf_sip6, &s_ip6),
+ ETH_ADDR_ARGS(dmac), buf_dip6);
+
err = tnl_port_build_header(xport->ofport, flow,
- dmac, smac, s_ip, &tnl_push_data);
+ dmac, smac, &s_ip6, &tnl_push_data);
if (err) {
return err;
}
/* If 'struct flow' gets additional metadata, we'll need to zero it out
* before traversing a patch port. */
- BUILD_ASSERT_DECL(FLOW_WC_SEQ == 34);
+ BUILD_ASSERT_DECL(FLOW_WC_SEQ == 35);
memset(&flow_tnl, 0, sizeof flow_tnl);
if (!xport) {
}
if (xport->is_tunnel) {
- ovs_be32 dst;
+ struct in6_addr dst;
/* Save tunnel metadata so that changes made due to
* the Logical (tunnel) Port are not visible for any further
* matches, while explicit set actions on tunnel metadata are.
xlate_report(ctx, "Tunneling decided against output");
goto out; /* restore flow_nw_tos */
}
- dst = in6_addr_get_mapped_ipv4(&ctx->orig_tunnel_ipv6_dst);
- if (flow->tunnel.ip_dst == dst) {
+ dst = flow_tnl_dst(&flow->tunnel);
+ if (ipv6_addr_equals(&dst, &ctx->orig_tunnel_ipv6_dst)) {
xlate_report(ctx, "Not tunneling to our own address");
goto out; /* restore flow_nw_tos */
}
enum ofp_packet_in_reason reason,
uint16_t controller_id)
{
- struct ofproto_packet_in *pin;
struct dp_packet *packet;
ctx->xout->slow |= SLOW_CONTROLLER;
+ xlate_commit_actions(ctx);
if (!ctx->xin->packet) {
return;
}
packet = dp_packet_clone(ctx->xin->packet);
- xlate_commit_actions(ctx);
-
odp_execute_actions(NULL, &packet, 1, false,
ctx->odp_actions->data, ctx->odp_actions->size, NULL);
- pin = xmalloc(sizeof *pin);
- pin->up.packet_len = dp_packet_size(packet);
- pin->up.packet = dp_packet_steal_data(packet);
- pin->up.reason = reason;
- pin->up.table_id = ctx->table_id;
- pin->up.cookie = ctx->rule_cookie;
-
- flow_get_metadata(&ctx->xin->flow, &pin->up.flow_metadata);
+ /* A packet sent by an action in a table-miss rule is considered an
+ * explicit table miss. OpenFlow before 1.3 doesn't have that concept so
+ * it will get translated back to OFPR_ACTION for those versions. */
+ if (reason == OFPR_ACTION
+ && ctx->rule && rule_dpif_is_table_miss(ctx->rule)) {
+ reason = OFPR_EXPLICIT_MISS;
+ }
+
+ size_t packet_len = dp_packet_size(packet);
+
+ struct ofproto_async_msg *am = xmalloc(sizeof *am);
+ *am = (struct ofproto_async_msg) {
+ .controller_id = controller_id,
+ .oam = OAM_PACKET_IN,
+ .pin = {
+ .up = {
+ .packet = dp_packet_steal_data(packet),
+ .len = packet_len,
+ .reason = reason,
+ .table_id = ctx->table_id,
+ .cookie = ctx->rule_cookie,
+ },
+ .max_len = len,
+ },
+ };
+ flow_get_metadata(&ctx->xin->flow, &am->pin.up.flow_metadata);
- pin->controller_id = controller_id;
- pin->send_len = len;
- /* If a rule is a table-miss rule then this is
- * a table-miss handled by a table-miss rule.
- *
- * Else, if rule is internal and has a controller action,
- * the later being implied by the rule being processed here,
- * then this is a table-miss handled without a table-miss rule.
- *
- * Otherwise this is not a table-miss. */
- pin->miss_type = OFPROTO_PACKET_IN_NO_MISS;
- if (ctx->rule) {
- if (rule_dpif_is_table_miss(ctx->rule)) {
- pin->miss_type = OFPROTO_PACKET_IN_MISS_FLOW;
- } else if (rule_dpif_is_internal(ctx->rule)) {
- pin->miss_type = OFPROTO_PACKET_IN_MISS_WITHOUT_FLOW;
- }
- }
- ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, pin);
+ ofproto_dpif_send_async_msg(ctx->xbridge->ofproto, am);
dp_packet_delete(packet);
}
struct recirc_state state = {
.table_id = table,
- .ofproto = ctx->xbridge->ofproto,
+ .ofproto_uuid = *ofproto_dpif_get_uuid(ctx->xbridge->ofproto),
.metadata = md,
- .stack = &ctx->stack,
+ .stack = ctx->stack.data,
+ .n_stack = ctx->stack.size / sizeof(union mf_subvalue),
.mirrors = ctx->mirrors,
.conntracked = ctx->conntracked,
+ .ofpacts = ((struct ofpact *) ctx->action_set.data
+ + ctx->recirc_action_offset / sizeof(struct ofpact)),
+ .ofpacts_len = ctx->action_set.size - ctx->recirc_action_offset,
+ .action_set = ctx->action_set.data,
.action_set_len = ctx->recirc_action_offset,
- .ofpacts_len = ctx->action_set.size,
- .ofpacts = ctx->action_set.data,
};
/* Allocate a unique recirc id for the given metadata state in the
}
static void
-xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact *a)
+xlate_write_actions__(struct xlate_ctx *ctx,
+ const struct ofpact *ofpacts, size_t ofpacts_len)
{
- const struct ofpact_nest *on = ofpact_get_WRITE_ACTIONS(a);
- size_t on_len = ofpact_nest_get_action_len(on);
- const struct ofpact *inner;
-
/* Maintain actset_output depending on the contents of the action set:
*
* - OFPP_UNSET, if there is no "output" action.
* - OFPP_UNSET, if there is a "group" action.
*/
if (!ctx->action_set_has_group) {
- OFPACT_FOR_EACH (inner, on->actions, on_len) {
- if (inner->type == OFPACT_OUTPUT) {
- ctx->xin->flow.actset_output = ofpact_get_OUTPUT(inner)->port;
- } else if (inner->type == OFPACT_GROUP) {
+ const struct ofpact *a;
+ OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) {
+ if (a->type == OFPACT_OUTPUT) {
+ ctx->xin->flow.actset_output = ofpact_get_OUTPUT(a)->port;
+ } else if (a->type == OFPACT_GROUP) {
ctx->xin->flow.actset_output = OFPP_UNSET;
ctx->action_set_has_group = true;
break;
}
}
- ofpbuf_put(&ctx->action_set, on->actions, on_len);
- ofpact_pad(&ctx->action_set);
+ ofpbuf_put(&ctx->action_set, ofpacts, ofpacts_len);
+}
+
+static void
+xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact_nest *a)
+{
+ xlate_write_actions__(ctx, a->actions, ofpact_nest_get_action_len(a));
}
static void
/* Copy remaining actions to the action_set to be executed after recirculation.
* UNROLL_XLATE action is inserted, if not already done so, before actions that
- * may generate PACKET_INs from the current table and without matching another
- * rule. */
+ * may depend on the current table ID or flow cookie. */
static void
recirc_unroll_actions(const struct ofpact *ofpacts, size_t ofpacts_len,
struct xlate_ctx *ctx)
OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) {
switch (a->type) {
- /* May generate PACKET INs. */
case OFPACT_OUTPUT_REG:
case OFPACT_GROUP:
case OFPACT_OUTPUT:
case OFPACT_CONTROLLER:
case OFPACT_DEC_MPLS_TTL:
case OFPACT_DEC_TTL:
+ /* These actions may generate asynchronous messages, which include
+ * table ID and flow cookie information. */
recirc_put_unroll_xlate(ctx);
break;
- /* These may not generate PACKET INs. */
+ case OFPACT_RESUBMIT:
+ if (ofpact_get_RESUBMIT(a)->table_id == 0xff) {
+ /* This resubmit action is relative to the current table, so we
+ * need to track what table that is.*/
+ recirc_put_unroll_xlate(ctx);
+ }
+ break;
+
case OFPACT_SET_TUNNEL:
case OFPACT_REG_MOVE:
case OFPACT_SET_FIELD:
case OFPACT_STACK_POP:
case OFPACT_LEARN:
case OFPACT_WRITE_METADATA:
- case OFPACT_RESUBMIT: /* May indirectly generate PACKET INs, */
- case OFPACT_GOTO_TABLE: /* but from a different table and rule. */
+ case OFPACT_GOTO_TABLE:
case OFPACT_ENQUEUE:
case OFPACT_SET_VLAN_VID:
case OFPACT_SET_VLAN_PCP:
case OFPACT_DEBUG_RECIRC:
case OFPACT_CT:
case OFPACT_NAT:
+ /* These may not generate PACKET INs. */
break;
- /* These need not be copied for restoration. */
case OFPACT_NOTE:
case OFPACT_CONJUNCTION:
+ /* These need not be copied for restoration. */
continue;
}
/* Copy the action over. */
break;
case OFPACT_RESUBMIT:
+ /* Recirculation complicates resubmit. There are two cases:
+ *
+ * - If mpls_pop has been executed, then the flow table lookup
+ * as part of resubmit might depend on fields that can only
+ * be obtained via recirculation, so the resubmit itself
+ * triggers recirculation and we need to make sure that the
+ * resubmit is executed again after recirculation.
+ * Therefore, in this case we trigger recirculation and let
+ * the code following this "switch" append the resubmit to
+ * the post-recirculation actions.
+ *
+ * - Otherwise, some action in the flow entry found by resubmit
+ * might trigger recirculation. If that happens, then we do
+ * not want to execute the resubmit again after
+ * recirculation, so we want to skip back to the head of the
+ * loop to avoid that, only adding any actions that follow
+ * the resubmit to the post-recirculation actions.
+ */
+ if (ctx->was_mpls) {
+ ctx_trigger_recirculation(ctx);
+ break;
+ }
xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a));
- break;
+ continue;
case OFPACT_SET_TUNNEL:
flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id);
break;
case OFPACT_WRITE_ACTIONS:
- xlate_write_actions(ctx, a);
+ xlate_write_actions(ctx, ofpact_get_WRITE_ACTIONS(a));
break;
case OFPACT_WRITE_METADATA:
case OFPACT_GOTO_TABLE: {
struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a);
- /* Allow ctx->table_id == TBL_INTERNAL, which will be greater
- * than ogt->table_id. This is to allow goto_table actions that
- * triggered recirculation: ctx->table_id will be TBL_INTERNAL
- * after recirculation. */
- ovs_assert(ctx->table_id == TBL_INTERNAL
- || ctx->table_id < ogt->table_id);
+ ovs_assert(ctx->table_id < ogt->table_id);
+
xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
ogt->table_id, true, true);
break;
xin->odp_actions = odp_actions;
/* Do recirc lookup. */
- xin->recirc = flow->recirc_id
- ? recirc_id_node_find(flow->recirc_id)
- : NULL;
+ xin->recirc = NULL;
+ if (flow->recirc_id) {
+ const struct recirc_id_node *node
+ = recirc_id_node_find(flow->recirc_id);
+ if (node) {
+ xin->recirc = &node->state;
+ }
+ }
}
void
{
*xout = (struct xlate_out) {
.slow = 0,
- .fail_open = false,
.recircs = RECIRC_REFS_EMPTY_INITIALIZER,
};
.xin = xin,
.xout = xout,
.base_flow = *flow,
+ .orig_tunnel_ipv6_dst = flow_tnl_dst(&flow->tunnel),
.xbridge = xbridge,
.stack = OFPBUF_STUB_INITIALIZER(stack_stub),
.rule = xin->rule,
.action_set_has_group = false,
.action_set = OFPBUF_STUB_INITIALIZER(action_set_stub),
};
- in6_addr_set_mapped_ipv4(&ctx.orig_tunnel_ipv6_dst, flow->tunnel.ip_dst);
/* 'base_flow' reflects the packet as it came in, but we need it to reflect
* the packet as the datapath will treat it for output actions:
COVERAGE_INC(xlate_actions);
if (xin->recirc) {
- const struct recirc_state *state = &xin->recirc->state;
+ const struct recirc_state *state = xin->recirc;
xlate_report(&ctx, "Restoring state post-recirculation:");
}
/* Set the bridge for post-recirculation processing if needed. */
- if (ctx.xbridge->ofproto != state->ofproto) {
+ if (!uuid_equals(ofproto_dpif_get_uuid(ctx.xbridge->ofproto),
+ &state->ofproto_uuid)) {
struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp);
const struct xbridge *new_bridge
- = xbridge_lookup(xcfg, state->ofproto);
+ = xbridge_lookup_by_uuid(xcfg, &state->ofproto_uuid);
if (OVS_UNLIKELY(!new_bridge)) {
/* Drop the packet if the bridge cannot be found. */
/* Restore stack, if any. */
if (state->stack) {
- ofpbuf_put(&ctx.stack, state->stack->data, state->stack->size);
+ ofpbuf_put(&ctx.stack, state->stack,
+ state->n_stack * sizeof *state->stack);
}
/* Restore mirror state. */
/* Restore action set, if any. */
if (state->action_set_len) {
- const struct ofpact *a;
-
xlate_report_actions(&ctx, "- Restoring action set",
- state->ofpacts, state->action_set_len);
-
- ofpbuf_put(&ctx.action_set, state->ofpacts, state->action_set_len);
+ state->action_set, state->action_set_len);
- OFPACT_FOR_EACH(a, state->ofpacts, state->action_set_len) {
- if (a->type == OFPACT_GROUP) {
- ctx.action_set_has_group = true;
- break;
- }
- }
+ flow->actset_output = OFPP_UNSET;
+ xlate_write_actions__(&ctx, state->action_set,
+ state->action_set_len);
}
/* Restore recirculation actions. If there are no actions, processing
* will start with a lookup in the table set above. */
- if (state->ofpacts_len > state->action_set_len) {
- xin->ofpacts_len = state->ofpacts_len - state->action_set_len;
- xin->ofpacts = state->ofpacts +
- state->action_set_len / sizeof *state->ofpacts;
-
+ xin->ofpacts = state->ofpacts;
+ xin->ofpacts_len = state->ofpacts_len;
+ if (state->ofpacts_len) {
xlate_report_actions(&ctx, "- Restoring actions",
xin->ofpacts, xin->ofpacts_len);
}
ctx.xin->resubmit_hook(ctx.xin, ctx.rule, 0);
}
}
- xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule);
/* Get the proximate input port of the packet. (If xin->recirc,
* flow->in_port is the ultimate input port of the packet.) */
break;
case XC_TNL_NEIGH:
/* Lookup neighbor to avoid timeout. */
- tnl_arp_lookup(entry->u.tnl_neigh_cache.br_name,
- entry->u.tnl_neigh_cache.d_ip, &dmac);
+ tnl_neigh_lookup(entry->u.tnl_neigh_cache.br_name,
+ &entry->u.tnl_neigh_cache.d_ipv6, &dmac);
break;
default:
OVS_NOT_REACHED();
projects / cascardo / ovs.git / blobdiff