/* Maximum depth of flow table recursion (due to resubmit actions) in a
* flow translation. */
#define MAX_RESUBMIT_RECURSION 64
+#define MAX_INTERNAL_RESUBMITS 1 /* Max resbmits allowed using rules in
+ internal table. */
/* Maximum number of resubmit actions in a flow translation, whether they are
* recursive or not. */
bool has_in_band; /* Bridge has in band control? */
bool forward_bpdu; /* Bridge forwards STP BPDUs? */
+ /* True if the datapath supports recirculation. */
+ bool enable_recirc;
+
/* True if the datapath supports variable-length
* OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions.
* False if the datapath supports only 8-byte (or shorter) userdata. */
* it did not arrive on a "real" port. 'ofpp_none_bundle' exists for
* when an input bundle is needed for validation (e.g., mirroring or
* OFPP_NORMAL processing). It is not connected to an 'ofproto' or have
- * any 'port' structs, so care must be taken when dealing with it.
- * The bundle's name and vlan mode are initialized in lookup_input_bundle() */
-static struct xbundle ofpp_none_bundle;
+ * any 'port' structs, so care must be taken when dealing with it. */
+static struct xbundle ofpp_none_bundle = {
+ .name = "OFPP_NONE",
+ .vlan_mode = PORT_VLAN_TRUNK
+};
/* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from
* 'priority' (the datapath's term for QoS queue) to the dscp bits which all
struct xlate_ctx *);
static void xlate_actions__(struct xlate_in *, struct xlate_out *)
OVS_REQ_RDLOCK(xlate_rwlock);
- static void xlate_normal(struct xlate_ctx *);
- static void xlate_report(struct xlate_ctx *, const char *);
- static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port,
- uint8_t table_id, bool may_packet_in);
+static void xlate_normal(struct xlate_ctx *);
+static void xlate_report(struct xlate_ctx *, const char *);
+static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port,
+ uint8_t table_id, bool may_packet_in,
+ bool honor_table_miss);
static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn);
static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid);
static void output_normal(struct xlate_ctx *, const struct xbundle *,
const struct dpif_ipfix *ipfix,
const struct netflow *netflow, enum ofp_config_flags frag,
bool forward_bpdu, bool has_in_band,
+ bool enable_recirc,
bool variable_length_userdata,
size_t max_mpls_depth)
{
xbridge->frag = frag;
xbridge->miss_rule = miss_rule;
xbridge->no_packet_in_rule = no_packet_in_rule;
+ xbridge->enable_recirc = enable_recirc;
xbridge->variable_length_userdata = variable_length_userdata;
xbridge->max_mpls_depth = max_mpls_depth;
}
: NULL;
}
-static enum stp_state
+static bool
xport_stp_learn_state(const struct xport *xport)
{
struct stp_port *sp = xport_get_stp_port(xport);
return stp_forward_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
}
+static bool
+xport_stp_listen_state(const struct xport *xport)
+{
+ struct stp_port *sp = xport_get_stp_port(xport);
+ return stp_listen_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
+}
+
/* Returns true if STP should process 'flow'. Sets fields in 'wc' that
* were used to make the determination.*/
static bool
{
struct stp_port *sp = xport_get_stp_port(xport);
struct ofpbuf payload = *packet;
- struct eth_header *eth = payload.data;
+ struct eth_header *eth = ofpbuf_data(&payload);
/* Sink packets on ports that have STP disabled when the bridge has
* STP enabled. */
}
/* Trim off padding on payload. */
- if (payload.size > ntohs(eth->eth_type) + ETH_HEADER_LEN) {
- payload.size = ntohs(eth->eth_type) + ETH_HEADER_LEN;
+ if (ofpbuf_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) {
+ ofpbuf_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN);
}
if (ofpbuf_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) {
- stp_received_bpdu(sp, payload.data, payload.size);
+ stp_received_bpdu(sp, ofpbuf_data(&payload), ofpbuf_size(&payload));
}
}
/* Special-case OFPP_NONE, which a controller may use as the ingress
* port for traffic that it is sourcing. */
if (in_port == OFPP_NONE) {
- ofpp_none_bundle.name = "OFPP_NONE";
- ofpp_none_bundle.vlan_mode = PORT_VLAN_TRUNK;
return &ofpp_none_bundle;
}
/* Partially configured bundle with no slaves. Drop the packet. */
return;
} else if (!out_xbundle->bond) {
+ ctx->xout->use_recirc = false;
xport = CONTAINER_OF(list_front(&out_xbundle->xports), struct xport,
bundle_node);
} else {
struct ofport_dpif *ofport;
+ struct xlate_recirc *xr = &ctx->xout->recirc;
+ struct flow_wildcards *wc = &ctx->xout->wc;
+
+ if (ctx->xbridge->enable_recirc) {
+ ctx->xout->use_recirc = bond_may_recirc(
+ out_xbundle->bond, &xr->recirc_id, &xr->hash_bias);
- ofport = bond_choose_output_slave(out_xbundle->bond, &ctx->xin->flow,
- &ctx->xout->wc, vid);
+ if (ctx->xout->use_recirc) {
+ /* Only TCP mode uses recirculation. */
+ xr->hash_alg = OVS_RECIRC_HASH_ALG_L4;
+ bond_update_post_recirc_rules(out_xbundle->bond, false);
+
+ /* Recirculation does not require unmasking hash fields. */
+ wc = NULL;
+ }
+ }
+
+ ofport = bond_choose_output_slave(out_xbundle->bond,
+ &ctx->xin->flow, wc, vid);
xport = xport_lookup(ofport);
if (!xport) {
actions_offset = nl_msg_start_nested(odp_actions, OVS_SAMPLE_ATTR_ACTIONS);
odp_port = ofp_port_to_odp_port(xbridge, flow->in_port.ofp_port);
- pid = dpif_port_get_pid(xbridge->dpif, odp_port);
+ pid = dpif_port_get_pid(xbridge->dpif, odp_port, 0);
cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, odp_actions);
nl_msg_end_nested(odp_actions, actions_offset);
/* If 'struct flow' gets additional metadata, we'll need to zero it out
* before traversing a patch port. */
- BUILD_ASSERT_DECL(FLOW_WC_SEQ == 24);
+ BUILD_ASSERT_DECL(FLOW_WC_SEQ == 25);
if (!xport) {
xlate_report(ctx, "Nonexistent output port");
} else if (xport->config & OFPUTIL_PC_NO_FWD) {
xlate_report(ctx, "OFPPC_NO_FWD set, skipping output");
return;
- } else if (check_stp && !xport_stp_forward_state(xport)) {
- xlate_report(ctx, "STP not in forwarding state, skipping output");
- return;
+ } else if (check_stp) {
+ if (eth_addr_equals(ctx->base_flow.dl_dst, eth_addr_stp)) {
+ if (!xport_stp_listen_state(xport)) {
+ xlate_report(ctx, "STP not in listening state, "
+ "skipping bpdu output");
+ return;
+ }
+ } else if (!xport_stp_forward_state(xport)) {
+ xlate_report(ctx, "STP not in forwarding state, "
+ "skipping output");
+ return;
+ }
}
if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) {
ctx->xout->slow |= special;
} else if (may_receive(peer, ctx)) {
if (xport_stp_forward_state(peer)) {
- xlate_table_action(ctx, flow->in_port.ofp_port, 0, true);
+ xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true);
} else {
/* Forwarding is disabled by STP. Let OFPP_NORMAL and the
* learning action look at the packet, then drop it. */
struct flow old_base_flow = ctx->base_flow;
- size_t old_size = ctx->xout->odp_actions.size;
+ size_t old_size = ofpbuf_size(&ctx->xout->odp_actions);
mirror_mask_t old_mirrors = ctx->xout->mirrors;
- xlate_table_action(ctx, flow->in_port.ofp_port, 0, true);
+ xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true);
ctx->xout->mirrors = old_mirrors;
ctx->base_flow = old_base_flow;
- ctx->xout->odp_actions.size = old_size;
+ ofpbuf_set_size(&ctx->xout->odp_actions, old_size);
}
}
flow_nw_tos = flow->nw_tos;
if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) {
- wc->masks.nw_tos |= IP_ECN_MASK;
+ wc->masks.nw_tos |= IP_DSCP_MASK;
flow->nw_tos &= ~IP_DSCP_MASK;
flow->nw_tos |= dscp;
}
ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow,
&ctx->xout->odp_actions,
&ctx->xout->wc);
- nl_msg_put_odp_port(&ctx->xout->odp_actions, OVS_ACTION_ATTR_OUTPUT,
- out_port);
+
+ if (ctx->xout->use_recirc) {
+ struct ovs_action_recirc *act_recirc;
+ struct xlate_recirc *xr = &ctx->xout->recirc;
+
+ act_recirc = nl_msg_put_unspec_uninit(&ctx->xout->odp_actions,
+ OVS_ACTION_ATTR_RECIRC, sizeof *act_recirc);
+ act_recirc->recirc_id = xr->recirc_id;
+ act_recirc->hash_alg = xr->hash_alg;
+ act_recirc->hash_bias = xr->hash_bias;
+ } else {
+ nl_msg_put_odp_port(&ctx->xout->odp_actions, OVS_ACTION_ATTR_OUTPUT,
+ out_port);
+ }
ctx->sflow_odp_port = odp_port;
ctx->sflow_n_outputs++;
ctx->rule = rule;
actions = rule_dpif_get_actions(rule);
do_xlate_actions(actions->ofpacts, actions->ofpacts_len, ctx);
- rule_actions_unref(actions);
ctx->rule = old_rule;
ctx->recurse--;
}
{
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
- if (ctx->recurse >= MAX_RESUBMIT_RECURSION) {
+ if (ctx->recurse >= MAX_RESUBMIT_RECURSION + MAX_INTERNAL_RESUBMITS) {
VLOG_ERR_RL(&rl, "resubmit actions recursed over %d times",
MAX_RESUBMIT_RECURSION);
- } else if (ctx->resubmits >= MAX_RESUBMITS) {
+ } else if (ctx->resubmits >= MAX_RESUBMITS + MAX_INTERNAL_RESUBMITS) {
VLOG_ERR_RL(&rl, "over %d resubmit actions", MAX_RESUBMITS);
- } else if (ctx->xout->odp_actions.size > UINT16_MAX) {
+ } else if (ofpbuf_size(&ctx->xout->odp_actions) > UINT16_MAX) {
VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of actions");
- } else if (ctx->stack.size >= 65536) {
+ } else if (ofpbuf_size(&ctx->stack) >= 65536) {
VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of stack");
} else {
return true;
}
static void
-xlate_table_action(struct xlate_ctx *ctx,
- ofp_port_t in_port, uint8_t table_id, bool may_packet_in)
+xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id,
+ bool may_packet_in, bool honor_table_miss)
{
if (xlate_resubmit_resource_check(ctx)) {
ofp_port_t old_in_port = ctx->xin->flow.in_port.ofp_port;
bool skip_wildcards = ctx->xin->skip_wildcards;
uint8_t old_table_id = ctx->table_id;
struct rule_dpif *rule;
+ enum rule_dpif_lookup_verdict verdict;
+ enum ofputil_port_config config = 0;
ctx->table_id = table_id;
* original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will
* have surprising behavior). */
ctx->xin->flow.in_port.ofp_port = in_port;
- rule_dpif_lookup_in_table(ctx->xbridge->ofproto, &ctx->xin->flow,
- !skip_wildcards ? &ctx->xout->wc : NULL,
- table_id, &rule);
+ verdict = rule_dpif_lookup_from_table(ctx->xbridge->ofproto,
+ &ctx->xin->flow,
+ !skip_wildcards
+ ? &ctx->xout->wc : NULL,
+ honor_table_miss,
+ &ctx->table_id, &rule);
ctx->xin->flow.in_port.ofp_port = old_in_port;
if (ctx->xin->resubmit_hook) {
ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse);
}
- if (!rule && may_packet_in) {
- struct xport *xport;
-
- /* XXX
- * check if table configuration flags
- * OFPTC11_TABLE_MISS_CONTROLLER, default.
- * OFPTC11_TABLE_MISS_CONTINUE,
- * OFPTC11_TABLE_MISS_DROP
- * When OF1.0, OFPTC11_TABLE_MISS_CONTINUE is used. What to do? */
- xport = get_ofp_port(ctx->xbridge, ctx->xin->flow.in_port.ofp_port);
- choose_miss_rule(xport ? xport->config : 0,
- ctx->xbridge->miss_rule,
- ctx->xbridge->no_packet_in_rule, &rule);
+ switch (verdict) {
+ case RULE_DPIF_LOOKUP_VERDICT_MATCH:
+ goto match;
+ case RULE_DPIF_LOOKUP_VERDICT_CONTROLLER:
+ if (may_packet_in) {
+ struct xport *xport;
+
+ xport = get_ofp_port(ctx->xbridge,
+ ctx->xin->flow.in_port.ofp_port);
+ config = xport ? xport->config : 0;
+ break;
+ }
+ /* Fall through to drop */
+ case RULE_DPIF_LOOKUP_VERDICT_DROP:
+ config = OFPUTIL_PC_NO_PACKET_IN;
+ break;
+ case RULE_DPIF_LOOKUP_VERDICT_DEFAULT:
+ if (!ofproto_dpif_wants_packet_in_on_miss(ctx->xbridge->ofproto)) {
+ config = OFPUTIL_PC_NO_PACKET_IN;
+ }
+ break;
+ default:
+ OVS_NOT_REACHED();
}
+
+ choose_miss_rule(config, ctx->xbridge->miss_rule,
+ ctx->xbridge->no_packet_in_rule, &rule);
+
+match:
if (rule) {
xlate_recursively(ctx, rule);
rule_dpif_unref(rule);
ofpacts_execute_action_set(&action_list, &action_set);
ctx->recurse++;
- do_xlate_actions(action_list.data, action_list.size, ctx);
+ do_xlate_actions(ofpbuf_data(&action_list), ofpbuf_size(&action_list), ctx);
ctx->recurse--;
ofpbuf_uninit(&action_set);
const struct ofputil_bucket *bucket;
uint32_t basis;
- basis = hash_bytes(ctx->xin->flow.dl_dst, sizeof ctx->xin->flow.dl_dst, 0);
+ basis = hash_mac(ctx->xin->flow.dl_dst, 0, 0);
bucket = group_best_live_bucket(ctx, group, basis);
if (bucket) {
memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
{
ofp_port_t in_port;
uint8_t table_id;
+ bool may_packet_in = false;
+ bool honor_table_miss = false;
+
+ if (ctx->rule && rule_dpif_is_internal(ctx->rule)) {
+ /* Still allow missed packets to be sent to the controller
+ * if resubmitting from an internal table. */
+ may_packet_in = true;
+ honor_table_miss = true;
+ }
in_port = resubmit->in_port;
if (in_port == OFPP_IN_PORT) {
table_id = ctx->table_id;
}
- xlate_table_action(ctx, in_port, table_id, false);
+ xlate_table_action(ctx, in_port, table_id, may_packet_in,
+ honor_table_miss);
}
static void
&ctx->xout->odp_actions,
&ctx->xout->wc);
- odp_execute_actions(NULL, packet, &md, ctx->xout->odp_actions.data,
- ctx->xout->odp_actions.size, NULL);
+ odp_execute_actions(NULL, packet, false, &md,
+ ofpbuf_data(&ctx->xout->odp_actions),
+ ofpbuf_size(&ctx->xout->odp_actions), NULL);
pin = xmalloc(sizeof *pin);
- pin->up.packet_len = packet->size;
+ pin->up.packet_len = ofpbuf_size(packet);
pin->up.packet = ofpbuf_steal_data(packet);
pin->up.reason = reason;
pin->up.table_id = ctx->table_id;
pin->controller_id = controller_id;
pin->send_len = len;
- pin->generated_by_table_miss = (ctx->rule
- && rule_dpif_is_table_miss(ctx->rule));
+ /* 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);
ofpbuf_delete(packet);
}
break;
case OFPP_TABLE:
xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
- 0, may_packet_in);
+ 0, may_packet_in, true);
break;
case OFPP_NORMAL:
xlate_normal(ctx);
ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub);
ofpacts_execute_action_set(&action_list, &ctx->action_set);
- do_xlate_actions(action_list.data, action_list.size, ctx);
+ do_xlate_actions(ofpbuf_data(&action_list), ofpbuf_size(&action_list), ctx);
ofpbuf_uninit(&action_list);
}
ovs_assert(ctx->table_id < ogt->table_id);
xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
- ogt->table_id, true);
+ ogt->table_id, true, true);
break;
}
ofpbuf_use_stub(&dst->odp_actions, dst->odp_actions_stub,
sizeof dst->odp_actions_stub);
- ofpbuf_put(&dst->odp_actions, src->odp_actions.data,
- src->odp_actions.size);
+ ofpbuf_put(&dst->odp_actions, ofpbuf_data(&src->odp_actions),
+ ofpbuf_size(&src->odp_actions));
}
\f
static struct skb_priority_to_dscp *
const struct nlattr *a;
unsigned int left;
- NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->xout->odp_actions.data,
- ctx->xout->odp_actions.size) {
+ NL_ATTR_FOR_EACH_UNSAFE (a, left, ofpbuf_data(&ctx->xout->odp_actions),
+ ofpbuf_size(&ctx->xout->odp_actions)) {
if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT
&& nl_attr_get_odp_port(a) == local_odp_port) {
return true;
ctx.exit = false;
if (!xin->ofpacts && !ctx.rule) {
- rule_dpif_lookup(ctx.xbridge->ofproto, flow,
- !xin->skip_wildcards ? wc : NULL, &rule);
+ ctx.table_id = rule_dpif_lookup(ctx.xbridge->ofproto, flow,
+ !xin->skip_wildcards ? wc : NULL,
+ &rule);
if (ctx.xin->resubmit_stats) {
rule_dpif_credit_stats(rule, ctx.xin->resubmit_stats);
}
ctx.rule = rule;
}
xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule);
+ xout->use_recirc = false;
if (xin->ofpacts) {
ofpacts = xin->ofpacts;
add_sflow_action(&ctx);
add_ipfix_action(&ctx);
- sample_actions_len = ctx.xout->odp_actions.size;
+ sample_actions_len = ofpbuf_size(&ctx.xout->odp_actions);
if (tnl_may_send && (!in_port || may_receive(in_port, &ctx))) {
do_xlate_actions(ofpacts, ofpacts_len, &ctx);
/* We've let OFPP_NORMAL and the learning action look at the
* packet, so drop it now if forwarding is disabled. */
if (in_port && !xport_stp_forward_state(in_port)) {
- ctx.xout->odp_actions.size = sample_actions_len;
+ ofpbuf_set_size(&ctx.xout->odp_actions, sample_actions_len);
}
}
- if (ctx.action_set.size) {
+ if (ofpbuf_size(&ctx.action_set)) {
xlate_action_set(&ctx);
}
}
}
- if (nl_attr_oversized(ctx.xout->odp_actions.size)) {
+ if (nl_attr_oversized(ofpbuf_size(&ctx.xout->odp_actions))) {
/* These datapath actions are too big for a Netlink attribute, so we
* can't hand them to the kernel directly. dpif_execute() can execute
* them one by one with help, so just mark the result as SLOW_ACTION to
}
out:
- rule_actions_unref(actions);
rule_dpif_unref(rule);
}
git://git.onelab.eu / sliver-openvswitch.git / blobdiff