#include "learn.h"
#include "list.h"
#include "mac-learning.h"
+#include "mcast-snooping.h"
#include "meta-flow.h"
#include "multipath.h"
#include "netdev-vport.h"
#include "ofproto/ofproto-dpif-sflow.h"
#include "ofproto/ofproto-dpif.h"
#include "ofproto/ofproto-provider.h"
+#include "packet-dpif.h"
#include "tunnel.h"
#include "vlog.h"
COVERAGE_DEFINE(xlate_actions);
COVERAGE_DEFINE(xlate_actions_oversize);
+COVERAGE_DEFINE(xlate_actions_too_many_output);
COVERAGE_DEFINE(xlate_actions_mpls_overflow);
VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate);
#define MAX_INTERNAL_RESUBMITS 1 /* Max resbmits allowed using rules in
internal table. */
+/* Timeout for internal rules created to handle recirculation */
+#define RECIRC_TIMEOUT 60
+
/* Maximum number of resubmit actions in a flow translation, whether they are
* recursive or not. */
#define MAX_RESUBMITS (MAX_RESUBMIT_RECURSION * MAX_RESUBMIT_RECURSION)
char *name; /* Name used in log messages. */
struct dpif *dpif; /* Datapath interface. */
struct mac_learning *ml; /* Mac learning handle. */
+ struct mcast_snooping *ms; /* Multicast Snooping handle. */
struct mbridge *mbridge; /* Mirroring. */
struct dpif_sflow *sflow; /* SFlow handle, or null. */
struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */
struct netflow *netflow; /* Netflow handle, or null. */
struct stp *stp; /* STP or null if disabled. */
+ struct rstp *rstp; /* RSTP or null if disabled. */
/* Special rules installed by ofproto-dpif. */
struct rule_dpif *miss_rule;
/* Number of MPLS label stack entries that the datapath supports
* in matches. */
size_t max_mpls_depth;
+
+ /* True if the datapath supports masked data in OVS_ACTION_ATTR_SET
+ * actions. */
+ bool masked_set_action;
};
struct xbundle {
enum ofputil_port_config config; /* OpenFlow port configuration. */
enum ofputil_port_state state; /* OpenFlow port state. */
int stp_port_no; /* STP port number or -1 if not in use. */
+ struct rstp_port *rstp_port; /* RSTP port or null. */
struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */
struct xlate_recirc recirc; /* Information used for generating
* recirculation actions */
+ /* True if a packet was but is no longer MPLS (due to an MPLS pop action).
+ * This is a trigger for recirculation in cases where translating an action
+ * or looking up a flow requires access to the fields of the packet after
+ * the MPLS label stack that was originally present. */
+ bool was_mpls;
+
/* OpenFlow 1.1+ action set.
*
* 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS.
uint16_t vid;
} bond;
struct {
- struct rule_dpif *rule;
+ struct ofproto_dpif *ofproto;
+ struct ofputil_flow_mod *fm;
+ struct ofpbuf *ofpacts;
} learn;
struct {
struct ofproto_dpif *ofproto;
static struct skb_priority_to_dscp *get_skb_priority(const struct xport *,
uint32_t skb_priority);
static void clear_skb_priorities(struct xport *);
+static size_t count_skb_priorities(const struct xport *);
static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority,
uint8_t *dscp);
static void xlate_xbridge_init(struct xlate_cfg *, struct xbridge *);
static void xlate_xbundle_init(struct xlate_cfg *, struct xbundle *);
static void xlate_xport_init(struct xlate_cfg *, struct xport *);
-static void xlate_xbridge_set(struct xbridge *xbridge,
- struct dpif *dpif,
+static void xlate_xbridge_set(struct xbridge *, struct dpif *,
struct rule_dpif *miss_rule,
struct rule_dpif *no_packet_in_rule,
- const struct mac_learning *ml, struct stp *stp,
- const struct mbridge *mbridge,
- const struct dpif_sflow *sflow,
- const struct dpif_ipfix *ipfix,
- const struct netflow *netflow,
- enum ofp_config_flags frag,
+ const struct mac_learning *, struct stp *,
+ struct rstp *, const struct mcast_snooping *,
+ const struct mbridge *,
+ const struct dpif_sflow *,
+ const struct dpif_ipfix *,
+ const struct netflow *, enum ofp_config_flags,
bool forward_bpdu, bool has_in_band,
bool enable_recirc,
bool variable_length_userdata,
- size_t max_mpls_depth);
+ size_t max_mpls_depth,
+ bool masked_set_action);
static void xlate_xbundle_set(struct xbundle *xbundle,
enum port_vlan_mode vlan_mode, int vlan,
unsigned long *trunks, bool use_priority_tags,
static void xlate_xport_set(struct xport *xport, odp_port_t odp_port,
const struct netdev *netdev, const struct cfm *cfm,
const struct bfd *bfd, int stp_port_no,
+ const struct rstp_port *rstp_port,
enum ofputil_port_config config,
enum ofputil_port_state state, bool is_tunnel,
bool may_enable);
struct rule_dpif *miss_rule,
struct rule_dpif *no_packet_in_rule,
const struct mac_learning *ml, struct stp *stp,
+ struct rstp *rstp, const struct mcast_snooping *ms,
const struct mbridge *mbridge,
const struct dpif_sflow *sflow,
const struct dpif_ipfix *ipfix,
bool forward_bpdu, bool has_in_band,
bool enable_recirc,
bool variable_length_userdata,
- size_t max_mpls_depth)
+ size_t max_mpls_depth,
+ bool masked_set_action)
{
if (xbridge->ml != ml) {
mac_learning_unref(xbridge->ml);
xbridge->ml = mac_learning_ref(ml);
}
+ if (xbridge->ms != ms) {
+ mcast_snooping_unref(xbridge->ms);
+ xbridge->ms = mcast_snooping_ref(ms);
+ }
+
if (xbridge->mbridge != mbridge) {
mbridge_unref(xbridge->mbridge);
xbridge->mbridge = mbridge_ref(mbridge);
xbridge->stp = stp_ref(stp);
}
+ if (xbridge->rstp != rstp) {
+ rstp_unref(xbridge->rstp);
+ xbridge->rstp = rstp_ref(rstp);
+ }
+
if (xbridge->netflow != netflow) {
netflow_unref(xbridge->netflow);
xbridge->netflow = netflow_ref(netflow);
xbridge->enable_recirc = enable_recirc;
xbridge->variable_length_userdata = variable_length_userdata;
xbridge->max_mpls_depth = max_mpls_depth;
+ xbridge->masked_set_action = masked_set_action;
}
static void
xlate_xport_set(struct xport *xport, odp_port_t odp_port,
const struct netdev *netdev, const struct cfm *cfm,
const struct bfd *bfd, int stp_port_no,
+ const struct rstp_port* rstp_port,
enum ofputil_port_config config, enum ofputil_port_state state,
bool is_tunnel, bool may_enable)
{
xport->may_enable = may_enable;
xport->odp_port = odp_port;
+ if (xport->rstp_port != rstp_port) {
+ rstp_port_unref(xport->rstp_port);
+ xport->rstp_port = rstp_port_ref(rstp_port);
+ }
+
if (xport->cfm != cfm) {
cfm_unref(xport->cfm);
xport->cfm = cfm_ref(cfm);
xlate_xbridge_set(new_xbridge,
xbridge->dpif, xbridge->miss_rule,
xbridge->no_packet_in_rule, xbridge->ml, xbridge->stp,
- xbridge->mbridge, xbridge->sflow, xbridge->ipfix,
- xbridge->netflow, xbridge->frag, xbridge->forward_bpdu,
+ xbridge->rstp, xbridge->ms, xbridge->mbridge,
+ xbridge->sflow, xbridge->ipfix, xbridge->netflow,
+ xbridge->frag, xbridge->forward_bpdu,
xbridge->has_in_band, xbridge->enable_recirc,
xbridge->variable_length_userdata,
- xbridge->max_mpls_depth);
+ xbridge->max_mpls_depth, xbridge->masked_set_action);
LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) {
xlate_xbundle_copy(new_xbridge, xbundle);
}
xlate_xport_init(new_xcfg, new_xport);
xlate_xport_set(new_xport, xport->odp_port, xport->netdev, xport->cfm,
- xport->bfd, xport->stp_port_no, xport->config, xport->state,
- xport->is_tunnel, xport->may_enable);
+ xport->bfd, xport->stp_port_no, xport->rstp_port,
+ xport->config, xport->state, xport->is_tunnel,
+ xport->may_enable);
if (xport->peer) {
struct xport *peer = xport_lookup(new_xcfg, xport->peer->ofport);
struct dpif *dpif, struct rule_dpif *miss_rule,
struct rule_dpif *no_packet_in_rule,
const struct mac_learning *ml, struct stp *stp,
+ struct rstp *rstp, const struct mcast_snooping *ms,
const struct mbridge *mbridge,
const struct dpif_sflow *sflow,
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)
+ bool forward_bpdu, bool has_in_band, bool enable_recirc,
+ bool variable_length_userdata, size_t max_mpls_depth,
+ bool masked_set_action)
{
struct xbridge *xbridge;
xbridge->name = xstrdup(name);
xlate_xbridge_set(xbridge, dpif, miss_rule, no_packet_in_rule, ml, stp,
- mbridge, sflow, ipfix, netflow, frag, forward_bpdu,
- has_in_band, enable_recirc, variable_length_userdata,
- max_mpls_depth);
+ rstp, ms, mbridge, sflow, ipfix, netflow, frag,
+ forward_bpdu, has_in_band, enable_recirc,
+ variable_length_userdata, max_mpls_depth,
+ masked_set_action);
}
static void
hmap_remove(&xcfg->xbridges, &xbridge->hmap_node);
mac_learning_unref(xbridge->ml);
+ mcast_snooping_unref(xbridge->ms);
mbridge_unref(xbridge->mbridge);
dpif_sflow_unref(xbridge->sflow);
dpif_ipfix_unref(xbridge->ipfix);
stp_unref(xbridge->stp);
+ rstp_unref(xbridge->rstp);
hmap_destroy(&xbridge->xports);
free(xbridge->name);
free(xbridge);
odp_port_t odp_port, const struct netdev *netdev,
const struct cfm *cfm, const struct bfd *bfd,
struct ofport_dpif *peer, int stp_port_no,
+ const struct rstp_port *rstp_port,
const struct ofproto_port_queue *qdscp_list, size_t n_qdscp,
enum ofputil_port_config config,
enum ofputil_port_state state, bool is_tunnel,
ovs_assert(xport->ofp_port == ofp_port);
- xlate_xport_set(xport, odp_port, netdev, cfm, bfd, stp_port_no, config,
- state, is_tunnel, may_enable);
+ xlate_xport_set(xport, odp_port, netdev, cfm, bfd, stp_port_no,
+ rstp_port, config, state, is_tunnel, may_enable);
if (xport->peer) {
xport->peer->peer = NULL;
hmap_remove(&xport->xbridge->xports, &xport->ofp_node);
netdev_close(xport->netdev);
+ rstp_port_unref(xport->rstp_port);
cfm_unref(xport->cfm);
bfd_unref(xport->bfd);
free(xport);
xlate_xport_remove(new_xcfg, xport);
}
-/* Given a datpath, packet, and flow metadata ('backer', 'packet', and 'key'
- * respectively), populates 'flow' with the result of odp_flow_key_to_flow().
- * Optionally populates 'ofproto' with the ofproto_dpif, 'odp_in_port' with
- * the datapath in_port, that 'packet' ingressed, and 'ipfix', 'sflow', and
- * 'netflow' with the appropriate handles for those protocols if they're
- * enabled. Caller is responsible for unrefing them.
- *
- * If 'ofproto' is nonnull, requires 'flow''s in_port to exist. Otherwise sets
- * 'flow''s in_port to OFPP_NONE.
- *
- * This function does post-processing on data returned from
- * odp_flow_key_to_flow() to help make VLAN splinters transparent to the rest
- * of the upcall processing logic. In particular, if the extracted in_port is
- * a VLAN splinter port, it replaces flow->in_port by the "real" port, sets
- * flow->vlan_tci correctly for the VLAN of the VLAN splinter port, and pushes
- * a VLAN header onto 'packet' (if it is nonnull).
- *
- * Similarly, this function also includes some logic to help with tunnels. It
- * may modify 'flow' as necessary to make the tunneling implementation
- * transparent to the upcall processing logic.
- *
- * Returns 0 if successful, ENODEV if the parsed flow has no associated ofport,
- * or some other positive errno if there are other problems. */
-int
-xlate_receive(const struct dpif_backer *backer, struct ofpbuf *packet,
- const struct nlattr *key, size_t key_len, struct flow *flow,
- struct ofproto_dpif **ofproto, struct dpif_ipfix **ipfix,
- struct dpif_sflow **sflow, struct netflow **netflow,
- odp_port_t *odp_in_port)
+/* Given a datapath and flow metadata ('backer', and 'flow' respectively)
+ * returns the corresponding struct xport, or NULL if none is found. */
+static struct xport *
+xlate_lookup_xport(const struct dpif_backer *backer, const struct flow *flow)
{
struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp);
- int error = ENODEV;
+
+ return xport_lookup(xcfg, tnl_port_should_receive(flow)
+ ? tnl_port_receive(flow)
+ : odp_port_to_ofport(backer, flow->in_port.odp_port));
+}
+
+static struct ofproto_dpif *
+xlate_lookup_ofproto_(const struct dpif_backer *backer, const struct flow *flow,
+ ofp_port_t *ofp_in_port, const struct xport **xportp)
+{
const struct xport *xport;
- if (odp_flow_key_to_flow(key, key_len, flow) == ODP_FIT_ERROR) {
- error = EINVAL;
- return error;
- }
+ *xportp = xport = xlate_lookup_xport(backer, flow);
- if (odp_in_port) {
- *odp_in_port = flow->in_port.odp_port;
+ if (xport) {
+ if (ofp_in_port) {
+ *ofp_in_port = xport->ofp_port;
+ }
+ return xport->xbridge->ofproto;
}
- xport = xport_lookup(xcfg, tnl_port_should_receive(flow)
- ? tnl_port_receive(flow)
- : odp_port_to_ofport(backer, flow->in_port.odp_port));
+ return NULL;
+}
- flow->in_port.ofp_port = xport ? xport->ofp_port : OFPP_NONE;
- if (!xport) {
- return error;
+/* Given a datapath and flow metadata ('backer', and 'flow' respectively)
+ * returns the corresponding struct ofproto_dpif and OpenFlow port number. */
+struct ofproto_dpif *
+xlate_lookup_ofproto(const struct dpif_backer *backer, const struct flow *flow,
+ ofp_port_t *ofp_in_port)
+{
+ const struct xport *xport;
+
+ return xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport);
+}
+
+/* Given a datapath and flow metadata ('backer', and 'flow' respectively),
+ * optionally populates 'ofproto' with the ofproto_dpif, 'ofp_in_port' with the
+ * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate
+ * handles for those protocols if they're enabled. Caller may use the returned
+ * pointers until quiescing, for longer term use additional references must
+ * be taken.
+ *
+ * '*ofp_in_port' is set to OFPP_NONE if 'flow''s in_port does not exist.
+ *
+ * Returns 0 if successful, ENODEV if the parsed flow has no associated ofport.
+ */
+int
+xlate_lookup(const struct dpif_backer *backer, const struct flow *flow,
+ struct ofproto_dpif **ofprotop, struct dpif_ipfix **ipfix,
+ struct dpif_sflow **sflow, struct netflow **netflow,
+ ofp_port_t *ofp_in_port)
+{
+ struct ofproto_dpif *ofproto;
+ const struct xport *xport;
+
+ ofproto = xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport);
+
+ if (ofp_in_port && !xport) {
+ *ofp_in_port = OFPP_NONE;
}
- if (vsp_adjust_flow(xport->xbridge->ofproto, flow)) {
- if (packet) {
- /* Make the packet resemble the flow, so that it gets sent to
- * an OpenFlow controller properly, so that it looks correct
- * for sFlow, and so that flow_extract() will get the correct
- * vlan_tci if it is called on 'packet'. */
- eth_push_vlan(packet, htons(ETH_TYPE_VLAN), flow->vlan_tci);
- }
+ if (!xport) {
+ return ENODEV;
}
- error = 0;
- if (ofproto) {
- *ofproto = xport->xbridge->ofproto;
+ if (ofprotop) {
+ *ofprotop = ofproto;
}
if (ipfix) {
- *ipfix = dpif_ipfix_ref(xport->xbridge->ipfix);
+ *ipfix = xport->xbridge->ipfix;
}
if (sflow) {
- *sflow = dpif_sflow_ref(xport->xbridge->sflow);
+ *sflow = xport->xbridge->sflow;
}
if (netflow) {
- *netflow = netflow_ref(xport->xbridge->netflow);
+ *netflow = xport->xbridge->netflow;
}
- return error;
+ return 0;
}
static struct xbridge *
}
static bool
-xport_stp_listen_state(const struct xport *xport)
+xport_stp_should_forward_bpdu(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);
+ return stp_should_forward_bpdu(sp ? stp_port_get_state(sp) : STP_DISABLED);
}
/* Returns true if STP should process 'flow'. Sets fields in 'wc' that
}
}
+static enum rstp_state
+xport_get_rstp_port_state(const struct xport *xport)
+{
+ return xport->rstp_port
+ ? rstp_port_get_state(xport->rstp_port)
+ : RSTP_DISABLED;
+}
+
+static bool
+xport_rstp_learn_state(const struct xport *xport)
+{
+ return rstp_learn_in_state(xport_get_rstp_port_state(xport));
+}
+
+static bool
+xport_rstp_forward_state(const struct xport *xport)
+{
+ return rstp_forward_in_state(xport_get_rstp_port_state(xport));
+}
+
+static bool
+xport_rstp_should_manage_bpdu(const struct xport *xport)
+{
+ return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport));
+}
+
+static void
+rstp_process_packet(const struct xport *xport, const struct ofpbuf *packet)
+{
+ struct ofpbuf payload = *packet;
+ struct eth_header *eth = ofpbuf_data(&payload);
+
+ /* Sink packets on ports that have no RSTP. */
+ if (!xport->rstp_port) {
+ return;
+ }
+
+ /* Trim off padding on payload. */
+ 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)) {
+ rstp_port_received_bpdu(xport->rstp_port, ofpbuf_data(&payload),
+ ofpbuf_size(&payload));
+ }
+}
+
static struct xport *
get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port)
{
"%s, which is reserved exclusively for mirroring",
ctx->xbridge->name, in_xbundle->name);
}
- ofpbuf_clear(&ctx->xout->odp_actions);
+ ofpbuf_clear(ctx->xout->odp_actions);
return;
}
}
}
+/* Updates multicast snooping table 'ms' given that a packet matching 'flow'
+ * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */
+static void
+update_mcast_snooping_table__(const struct xbridge *xbridge,
+ const struct flow *flow,
+ struct mcast_snooping *ms,
+ ovs_be32 ip4, int vlan,
+ struct xbundle *in_xbundle)
+ OVS_REQ_WRLOCK(ms->rwlock)
+{
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30);
+
+ switch (ntohs(flow->tp_src)) {
+ case IGMP_HOST_MEMBERSHIP_REPORT:
+ case IGMPV2_HOST_MEMBERSHIP_REPORT:
+ if (mcast_snooping_add_group(ms, ip4, vlan, in_xbundle->ofbundle)) {
+ VLOG_DBG_RL(&rl, "bridge %s: multicast snooping learned that "
+ IP_FMT" is on port %s in VLAN %d",
+ xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan);
+ }
+ break;
+ case IGMP_HOST_LEAVE_MESSAGE:
+ if (mcast_snooping_leave_group(ms, ip4, vlan, in_xbundle->ofbundle)) {
+ VLOG_DBG_RL(&rl, "bridge %s: multicast snooping leaving "
+ IP_FMT" is on port %s in VLAN %d",
+ xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan);
+ }
+ break;
+ case IGMP_HOST_MEMBERSHIP_QUERY:
+ if (flow->nw_src && mcast_snooping_add_mrouter(ms, vlan,
+ in_xbundle->ofbundle)) {
+ VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query from "
+ IP_FMT" is on port %s in VLAN %d",
+ xbridge->name, IP_ARGS(flow->nw_src),
+ in_xbundle->name, vlan);
+ }
+ break;
+ }
+}
+
+/* Updates multicast snooping table 'ms' given that a packet matching 'flow'
+ * was received on 'in_xbundle' in 'vlan'. */
+static void
+update_mcast_snooping_table(const struct xbridge *xbridge,
+ const struct flow *flow, int vlan,
+ struct xbundle *in_xbundle)
+{
+ struct mcast_snooping *ms = xbridge->ms;
+ struct xlate_cfg *xcfg;
+ struct xbundle *mcast_xbundle;
+ struct mcast_fport_bundle *fport;
+
+ /* Don't learn the OFPP_NONE port. */
+ if (in_xbundle == &ofpp_none_bundle) {
+ return;
+ }
+
+ /* Don't learn from flood ports */
+ mcast_xbundle = NULL;
+ ovs_rwlock_wrlock(&ms->rwlock);
+ xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp);
+ LIST_FOR_EACH(fport, fport_node, &ms->fport_list) {
+ mcast_xbundle = xbundle_lookup(xcfg, fport->port);
+ if (mcast_xbundle == in_xbundle) {
+ break;
+ }
+ }
+
+ if (!mcast_xbundle || mcast_xbundle != in_xbundle) {
+ update_mcast_snooping_table__(xbridge, flow, ms, flow->igmp_group_ip4,
+ vlan, in_xbundle);
+ }
+ ovs_rwlock_unlock(&ms->rwlock);
+}
+
+/* send the packet to ports having the multicast group learned */
+static void
+xlate_normal_mcast_send_group(struct xlate_ctx *ctx,
+ struct mcast_snooping *ms OVS_UNUSED,
+ struct mcast_group *grp,
+ struct xbundle *in_xbundle, uint16_t vlan)
+ OVS_REQ_RDLOCK(ms->rwlock)
+{
+ struct xlate_cfg *xcfg;
+ struct mcast_group_bundle *b;
+ struct xbundle *mcast_xbundle;
+
+ xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp);
+ LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) {
+ mcast_xbundle = xbundle_lookup(xcfg, b->port);
+ if (mcast_xbundle && mcast_xbundle != in_xbundle) {
+ xlate_report(ctx, "forwarding to mcast group port");
+ output_normal(ctx, mcast_xbundle, vlan);
+ } else if (!mcast_xbundle) {
+ xlate_report(ctx, "mcast group port is unknown, dropping");
+ } else {
+ xlate_report(ctx, "mcast group port is input port, dropping");
+ }
+ }
+}
+
+/* send the packet to ports connected to multicast routers */
+static void
+xlate_normal_mcast_send_mrouters(struct xlate_ctx *ctx,
+ struct mcast_snooping *ms,
+ struct xbundle *in_xbundle, uint16_t vlan)
+ OVS_REQ_RDLOCK(ms->rwlock)
+{
+ struct xlate_cfg *xcfg;
+ struct mcast_mrouter_bundle *mrouter;
+ struct xbundle *mcast_xbundle;
+
+ xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp);
+ LIST_FOR_EACH(mrouter, mrouter_node, &ms->mrouter_lru) {
+ mcast_xbundle = xbundle_lookup(xcfg, mrouter->port);
+ if (mcast_xbundle && mcast_xbundle != in_xbundle) {
+ xlate_report(ctx, "forwarding to mcast router port");
+ output_normal(ctx, mcast_xbundle, vlan);
+ } else if (!mcast_xbundle) {
+ xlate_report(ctx, "mcast router port is unknown, dropping");
+ } else {
+ xlate_report(ctx, "mcast router port is input port, dropping");
+ }
+ }
+}
+
+/* send the packet to ports flagged to be flooded */
+static void
+xlate_normal_mcast_send_fports(struct xlate_ctx *ctx,
+ struct mcast_snooping *ms,
+ struct xbundle *in_xbundle, uint16_t vlan)
+ OVS_REQ_RDLOCK(ms->rwlock)
+{
+ struct xlate_cfg *xcfg;
+ struct mcast_fport_bundle *fport;
+ struct xbundle *mcast_xbundle;
+
+ xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp);
+ LIST_FOR_EACH(fport, fport_node, &ms->fport_list) {
+ mcast_xbundle = xbundle_lookup(xcfg, fport->port);
+ if (mcast_xbundle && mcast_xbundle != in_xbundle) {
+ xlate_report(ctx, "forwarding to mcast flood port");
+ output_normal(ctx, mcast_xbundle, vlan);
+ } else if (!mcast_xbundle) {
+ xlate_report(ctx, "mcast flood port is unknown, dropping");
+ } else {
+ xlate_report(ctx, "mcast flood port is input port, dropping");
+ }
+ }
+}
+
static void
xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle,
uint16_t vlan)
}
/* Determine output bundle. */
- ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock);
- mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan);
- mac_port = mac ? mac->port.p : NULL;
- ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock);
+ if (mcast_snooping_enabled(ctx->xbridge->ms)
+ && !eth_addr_is_broadcast(flow->dl_dst)
+ && eth_addr_is_multicast(flow->dl_dst)
+ && flow->dl_type == htons(ETH_TYPE_IP)) {
+ struct mcast_snooping *ms = ctx->xbridge->ms;
+ struct mcast_group *grp;
+
+ if (flow->nw_proto == IPPROTO_IGMP) {
+ if (ctx->xin->may_learn) {
+ if (mcast_snooping_is_membership(flow->tp_src) ||
+ mcast_snooping_is_query(flow->tp_src)) {
+ update_mcast_snooping_table(ctx->xbridge, flow, vlan,
+ in_xbundle);
+ }
+ }
- if (mac_port) {
- struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp);
- struct xbundle *mac_xbundle = xbundle_lookup(xcfg, mac_port);
- if (mac_xbundle && mac_xbundle != in_xbundle) {
- xlate_report(ctx, "forwarding to learned port");
- output_normal(ctx, mac_xbundle, vlan);
- } else if (!mac_xbundle) {
- xlate_report(ctx, "learned port is unknown, dropping");
+ if (mcast_snooping_is_membership(flow->tp_src)) {
+ ovs_rwlock_rdlock(&ms->rwlock);
+ xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan);
+ ovs_rwlock_unlock(&ms->rwlock);
+ } else {
+ xlate_report(ctx, "multicast traffic, flooding");
+ xlate_normal_flood(ctx, in_xbundle, vlan);
+ }
+ return;
+ } else {
+ if (ip_is_local_multicast(flow->nw_dst)) {
+ /* RFC4541: section 2.1.2, item 2: Packets with a dst IP
+ * address in the 224.0.0.x range which are not IGMP must
+ * be forwarded on all ports */
+ xlate_report(ctx, "RFC4541: section 2.1.2, item 2, flooding");
+ xlate_normal_flood(ctx, in_xbundle, vlan);
+ return;
+ }
+ }
+
+ /* forwarding to group base ports */
+ ovs_rwlock_rdlock(&ms->rwlock);
+ grp = mcast_snooping_lookup(ms, flow->nw_dst, vlan);
+ if (grp) {
+ xlate_normal_mcast_send_group(ctx, ms, grp, in_xbundle, vlan);
+ xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan);
+ xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan);
} else {
- xlate_report(ctx, "learned port is input port, dropping");
+ if (mcast_snooping_flood_unreg(ms)) {
+ xlate_report(ctx, "unregistered multicast, flooding");
+ xlate_normal_flood(ctx, in_xbundle, vlan);
+ } else {
+ xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan);
+ xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan);
+ }
}
+ ovs_rwlock_unlock(&ms->rwlock);
} else {
- xlate_report(ctx, "no learned MAC for destination, flooding");
- xlate_normal_flood(ctx, in_xbundle, vlan);
+ ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock);
+ mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan);
+ mac_port = mac ? mac->port.p : NULL;
+ ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock);
+
+ if (mac_port) {
+ struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp);
+ struct xbundle *mac_xbundle = xbundle_lookup(xcfg, mac_port);
+ if (mac_xbundle && mac_xbundle != in_xbundle) {
+ xlate_report(ctx, "forwarding to learned port");
+ output_normal(ctx, mac_xbundle, vlan);
+ } else if (!mac_xbundle) {
+ xlate_report(ctx, "learned port is unknown, dropping");
+ } else {
+ xlate_report(ctx, "learned port is input port, dropping");
+ }
+ } else {
+ xlate_report(ctx, "no learned MAC for destination, flooding");
+ xlate_normal_flood(ctx, in_xbundle, vlan);
+ }
}
}
const struct flow *flow,
const uint32_t probability,
const union user_action_cookie *cookie,
- const size_t cookie_size)
+ const size_t cookie_size,
+ const odp_port_t tunnel_out_port)
{
size_t sample_offset, actions_offset;
odp_port_t odp_port;
pid = dpif_port_get_pid(xbridge->dpif, odp_port,
flow_hash_5tuple(flow, 0));
cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size,
- odp_actions);
+ tunnel_out_port, odp_actions);
nl_msg_end_nested(odp_actions, actions_offset);
nl_msg_end_nested(odp_actions, sample_offset);
odp_port == ODPP_NONE ? 0 : 1, &cookie);
return compose_sample_action(xbridge, odp_actions, flow, probability,
- &cookie, sizeof cookie.sflow);
+ &cookie, sizeof cookie.sflow, ODPP_NONE);
}
static void
}
static void
-compose_ipfix_cookie(union user_action_cookie *cookie)
+compose_ipfix_cookie(union user_action_cookie *cookie,
+ odp_port_t output_odp_port)
{
cookie->type = USER_ACTION_COOKIE_IPFIX;
+ cookie->ipfix.output_odp_port = output_odp_port;
}
/* Compose SAMPLE action for IPFIX bridge sampling. */
static void
compose_ipfix_action(const struct xbridge *xbridge,
struct ofpbuf *odp_actions,
- const struct flow *flow)
+ const struct flow *flow,
+ odp_port_t output_odp_port)
{
uint32_t probability;
union user_action_cookie cookie;
+ odp_port_t tunnel_out_port = ODPP_NONE;
if (!xbridge->ipfix || flow->in_port.ofp_port == OFPP_NONE) {
return;
}
+ /* For input case, output_odp_port is ODPP_NONE, which is an invalid port
+ * number. */
+ if (output_odp_port == ODPP_NONE &&
+ !dpif_ipfix_get_bridge_exporter_input_sampling(xbridge->ipfix)) {
+ return;
+ }
+
+ /* For output case, output_odp_port is valid*/
+ if (output_odp_port != ODPP_NONE) {
+ if (!dpif_ipfix_get_bridge_exporter_output_sampling(xbridge->ipfix)) {
+ return;
+ }
+ /* If tunnel sampling is enabled, put an additional option attribute:
+ * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT
+ */
+ if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(xbridge->ipfix) &&
+ dpif_ipfix_get_tunnel_port(xbridge->ipfix, output_odp_port) ) {
+ tunnel_out_port = output_odp_port;
+ }
+ }
+
probability = dpif_ipfix_get_bridge_exporter_probability(xbridge->ipfix);
- compose_ipfix_cookie(&cookie);
+ compose_ipfix_cookie(&cookie, output_odp_port);
compose_sample_action(xbridge, odp_actions, flow, probability,
- &cookie, sizeof cookie.ipfix);
+ &cookie, sizeof cookie.ipfix, tunnel_out_port);
}
/* SAMPLE action for sFlow must be first action in any given list of
add_sflow_action(struct xlate_ctx *ctx)
{
ctx->user_cookie_offset = compose_sflow_action(ctx->xbridge,
- &ctx->xout->odp_actions,
+ ctx->xout->odp_actions,
&ctx->xin->flow, ODPP_NONE);
ctx->sflow_odp_port = 0;
ctx->sflow_n_outputs = 0;
static void
add_ipfix_action(struct xlate_ctx *ctx)
{
- compose_ipfix_action(ctx->xbridge, &ctx->xout->odp_actions,
- &ctx->xin->flow);
+ compose_ipfix_action(ctx->xbridge, ctx->xout->odp_actions,
+ &ctx->xin->flow, ODPP_NONE);
+}
+
+static void
+add_ipfix_output_action(struct xlate_ctx *ctx, odp_port_t port)
+{
+ compose_ipfix_action(ctx->xbridge, ctx->xout->odp_actions,
+ &ctx->xin->flow, port);
}
/* Fix SAMPLE action according to data collected while composing ODP actions.
return;
}
- cookie = ofpbuf_at(&ctx->xout->odp_actions, ctx->user_cookie_offset,
+ cookie = ofpbuf_at(ctx->xout->odp_actions, ctx->user_cookie_offset,
sizeof cookie->sflow);
ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW);
lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet);
}
return SLOW_LACP;
- } else if (xbridge->stp && stp_should_process_flow(flow, wc)) {
+ } else if ((xbridge->stp || xbridge->rstp) &&
+ stp_should_process_flow(flow, wc)) {
if (packet) {
- stp_process_packet(xport, packet);
+ xbridge->stp
+ ? stp_process_packet(xport, packet)
+ : rstp_process_packet(xport, packet);
}
return SLOW_STP;
} else {
/* If 'struct flow' gets additional metadata, we'll need to zero it out
* before traversing a patch port. */
- BUILD_ASSERT_DECL(FLOW_WC_SEQ == 26);
+ BUILD_ASSERT_DECL(FLOW_WC_SEQ == 27);
if (!xport) {
xlate_report(ctx, "Nonexistent output port");
return;
} else if (check_stp) {
if (is_stp(&ctx->base_flow)) {
- if (!xport_stp_listen_state(xport)) {
- xlate_report(ctx, "STP not in listening state, "
- "skipping bpdu output");
+ if (!xport_stp_should_forward_bpdu(xport) &&
+ !xport_rstp_should_manage_bpdu(xport)) {
+ if (ctx->xbridge->stp != NULL) {
+ xlate_report(ctx, "STP not in listening state, "
+ "skipping bpdu output");
+ } else if (ctx->xbridge->rstp != NULL) {
+ xlate_report(ctx, "RSTP not managing BPDU in this state, "
+ "skipping bpdu output");
+ }
return;
}
- } else if (!xport_stp_forward_state(xport)) {
- xlate_report(ctx, "STP not in forwarding state, "
- "skipping output");
+ } else if (!xport_stp_forward_state(xport) ||
+ !xport_rstp_forward_state(xport)) {
+ if (ctx->xbridge->stp != NULL) {
+ xlate_report(ctx, "STP not in forwarding state, "
+ "skipping output");
+ } else if (ctx->xbridge->rstp != NULL) {
+ xlate_report(ctx, "RSTP not in forwarding state, "
+ "skipping output");
+ }
return;
}
}
if (special) {
ctx->xout->slow |= special;
} else if (may_receive(peer, ctx)) {
- if (xport_stp_forward_state(peer)) {
+ if (xport_stp_forward_state(peer) && xport_rstp_forward_state(peer)) {
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. */
+ /* Forwarding is disabled by STP and RSTP. 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 = ofpbuf_size(&ctx->xout->odp_actions);
+ 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, true);
ctx->xout->mirrors = old_mirrors;
ctx->base_flow = old_base_flow;
- ofpbuf_set_size(&ctx->xout->odp_actions, old_size);
+ ofpbuf_set_size(ctx->xout->odp_actions, old_size);
}
}
flow_pkt_mark = flow->pkt_mark;
flow_nw_tos = flow->nw_tos;
- if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) {
- wc->masks.nw_tos |= IP_DSCP_MASK;
- flow->nw_tos &= ~IP_DSCP_MASK;
- flow->nw_tos |= dscp;
+ if (count_skb_priorities(xport)) {
+ memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority);
+ if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) {
+ wc->masks.nw_tos |= IP_DSCP_MASK;
+ flow->nw_tos &= ~IP_DSCP_MASK;
+ flow->nw_tos |= dscp;
+ }
}
if (xport->is_tunnel) {
}
out_port = odp_port;
commit_odp_tunnel_action(flow, &ctx->base_flow,
- &ctx->xout->odp_actions);
+ ctx->xout->odp_actions);
flow->tunnel = flow_tnl; /* Restore tunnel metadata */
} else {
odp_port = xport->odp_port;
if (out_port != ODPP_NONE) {
ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow,
- &ctx->xout->odp_actions,
- &ctx->xout->wc);
+ ctx->xout->odp_actions,
+ &ctx->xout->wc,
+ ctx->xbridge->masked_set_action);
if (ctx->use_recirc) {
struct ovs_action_hash *act_hash;
struct xlate_recirc *xr = &ctx->recirc;
/* Hash action. */
- act_hash = nl_msg_put_unspec_uninit(&ctx->xout->odp_actions,
+ act_hash = nl_msg_put_unspec_uninit(ctx->xout->odp_actions,
OVS_ACTION_ATTR_HASH,
sizeof *act_hash);
act_hash->hash_alg = xr->hash_alg;
act_hash->hash_basis = xr->hash_basis;
/* Recirc action. */
- nl_msg_put_u32(&ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC,
+ nl_msg_put_u32(ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC,
xr->recirc_id);
} else {
- nl_msg_put_odp_port(&ctx->xout->odp_actions, OVS_ACTION_ATTR_OUTPUT,
+ add_ipfix_output_action(ctx, out_port);
+ nl_msg_put_odp_port(ctx->xout->odp_actions, OVS_ACTION_ATTR_OUTPUT,
out_port);
}
MAX_RESUBMIT_RECURSION);
} else if (ctx->resubmits >= MAX_RESUBMITS + MAX_INTERNAL_RESUBMITS) {
VLOG_ERR_RL(&rl, "over %d resubmit actions", MAX_RESUBMITS);
- } else if (ofpbuf_size(&ctx->xout->odp_actions) > 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 (ofpbuf_size(&ctx->stack) >= 65536) {
VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of stack");
struct ofputil_bucket *bucket;
uint32_t basis;
- basis = hash_mac(ctx->xin->flow.dl_dst, 0, 0);
+ basis = flow_hash_symmetric_l4(&ctx->xin->flow, 0);
bucket = group_best_live_bucket(ctx, group, basis);
if (bucket) {
memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
+ memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src);
+ memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type);
+ memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst);
+ memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
+ memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src);
+ memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
+ memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src);
+ memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci);
+
xlate_group_bucket(ctx, bucket);
xlate_group_stats(ctx, group, bucket);
}
uint16_t controller_id)
{
struct ofproto_packet_in *pin;
- struct ofpbuf *packet;
- struct pkt_metadata md = PKT_METADATA_INITIALIZER(0);
+ struct dpif_packet *packet;
ctx->xout->slow |= SLOW_CONTROLLER;
if (!ctx->xin->packet) {
return;
}
- packet = ofpbuf_clone(ctx->xin->packet);
+ packet = dpif_packet_clone_from_ofpbuf(ctx->xin->packet);
ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
- &ctx->xout->odp_actions,
- &ctx->xout->wc);
+ ctx->xout->odp_actions,
+ &ctx->xout->wc,
+ ctx->xbridge->masked_set_action);
- odp_execute_actions(NULL, packet, false, &md,
- ofpbuf_data(&ctx->xout->odp_actions),
- ofpbuf_size(&ctx->xout->odp_actions), NULL);
+ odp_execute_actions(NULL, &packet, 1, false,
+ ofpbuf_data(ctx->xout->odp_actions),
+ ofpbuf_size(ctx->xout->odp_actions), NULL);
pin = xmalloc(sizeof *pin);
- pin->up.packet_len = ofpbuf_size(packet);
- pin->up.packet = ofpbuf_steal_data(packet);
+ pin->up.packet_len = ofpbuf_size(&packet->ofpbuf);
+ pin->up.packet = ofpbuf_steal_data(&packet->ofpbuf);
pin->up.reason = reason;
pin->up.table_id = ctx->table_id;
pin->up.cookie = (ctx->rule
}
}
ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, pin);
- ofpbuf_delete(packet);
+ dpif_packet_delete(packet);
+}
+
+static void
+compose_recirculate_action(struct xlate_ctx *ctx,
+ const struct ofpact *ofpacts_base,
+ const struct ofpact *ofpact_current,
+ size_t ofpacts_base_len)
+{
+ uint32_t id;
+ int error;
+ unsigned ofpacts_len;
+ struct match match;
+ struct rule *rule;
+ struct ofpbuf ofpacts;
+
+ ctx->exit = true;
+
+ ofpacts_len = ofpacts_base_len -
+ ((uint8_t *)ofpact_current - (uint8_t *)ofpacts_base);
+
+ if (ctx->rule) {
+ id = rule_dpif_get_recirc_id(ctx->rule);
+ } else {
+ /* In the case where ctx has no rule then allocate a recirc id.
+ * The life-cycle of this recirc id is managed by associating it
+ * with the internal rule that is created to to handle
+ * recirculation below.
+ *
+ * The known use-case of this is packet_out which
+ * translates actions without a rule */
+ id = ofproto_dpif_alloc_recirc_id(ctx->xbridge->ofproto);
+ }
+ if (!id) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_ERR_RL(&rl, "Failed to allocate recirculation id");
+ return;
+ }
+
+ match_init_catchall(&match);
+ match_set_recirc_id(&match, id);
+ ofpbuf_use_const(&ofpacts, ofpact_current, ofpacts_len);
+ error = ofproto_dpif_add_internal_flow(ctx->xbridge->ofproto, &match,
+ RECIRC_RULE_PRIORITY,
+ RECIRC_TIMEOUT, &ofpacts, &rule);
+ if (error) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_ERR_RL(&rl, "Failed to add post recirculation flow %s",
+ match_to_string(&match, 0));
+ return;
+ }
+ /* If ctx has no rule then associate the recirc id, which
+ * was allocated above, with the internal rule. This allows
+ * the recirc id to be released when the internal rule times out. */
+ if (!ctx->rule) {
+ rule_set_recirc_id(rule, id);
+ }
+
+ ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
+ ctx->xout->odp_actions,
+ &ctx->xout->wc,
+ ctx->xbridge->masked_set_action);
+ nl_msg_put_u32(ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC, id);
}
static void
n = flow_count_mpls_labels(flow, wc);
if (!n) {
ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow,
- &ctx->xout->odp_actions,
- &ctx->xout->wc);
+ ctx->xout->odp_actions,
+ &ctx->xout->wc,
+ ctx->xbridge->masked_set_action);
} else if (n >= FLOW_MAX_MPLS_LABELS) {
if (ctx->xin->packet != NULL) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
struct flow *flow = &ctx->xin->flow;
int n = flow_count_mpls_labels(flow, wc);
- if (!flow_pop_mpls(flow, n, eth_type, wc) && n >= FLOW_MAX_MPLS_LABELS) {
+ if (flow_pop_mpls(flow, n, eth_type, wc)) {
+ if (ctx->xbridge->enable_recirc && !eth_type_mpls(eth_type)) {
+ ctx->was_mpls = true;
+ }
+ } else if (n >= FLOW_MAX_MPLS_LABELS) {
if (ctx->xin->packet != NULL) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an "
ctx->xbridge->name, FLOW_MAX_MPLS_LABELS);
}
ctx->exit = true;
- ofpbuf_clear(&ctx->xout->odp_actions);
+ ofpbuf_clear(ctx->xout->odp_actions);
}
}
compose_dec_mpls_ttl_action(struct xlate_ctx *ctx)
{
struct flow *flow = &ctx->xin->flow;
- uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]);
struct flow_wildcards *wc = &ctx->xout->wc;
- memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
if (eth_type_mpls(flow->dl_type)) {
+ uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]);
+
+ wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK);
if (ttl > 1) {
ttl--;
set_mpls_lse_ttl(&flow->mpls_lse[0], ttl);
return false;
} else {
execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0);
-
- /* Stop processing for current table. */
- return true;
}
- } else {
- return true;
}
+
+ /* Stop processing for current table. */
+ return true;
}
static void
}
static void
-xlate_learn_action(struct xlate_ctx *ctx,
- const struct ofpact_learn *learn)
+xlate_learn_action__(struct xlate_ctx *ctx, const struct ofpact_learn *learn,
+ struct ofputil_flow_mod *fm, struct ofpbuf *ofpacts)
{
- uint64_t ofpacts_stub[1024 / 8];
- struct ofputil_flow_mod fm;
- struct ofpbuf ofpacts;
+ learn_execute(learn, &ctx->xin->flow, fm, ofpacts);
+ if (ctx->xin->may_learn) {
+ ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm);
+ }
+}
+static void
+xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn)
+{
ctx->xout->has_learn = true;
-
learn_mask(learn, &ctx->xout->wc);
- if (!ctx->xin->may_learn) {
- return;
- }
-
- ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub);
- learn_execute(learn, &ctx->xin->flow, &fm, &ofpacts);
- ofproto_dpif_flow_mod(ctx->xbridge->ofproto, &fm);
- ofpbuf_uninit(&ofpacts);
-
if (ctx->xin->xcache) {
struct xc_entry *entry;
entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN);
- /* Lookup the learned rule, taking a reference on it. The reference
- * is released when this cache entry is deleted. */
- rule_dpif_lookup(ctx->xbridge->ofproto, &ctx->xin->flow, NULL,
- &entry->u.learn.rule, true, NULL);
+ entry->u.learn.ofproto = ctx->xbridge->ofproto;
+ entry->u.learn.fm = xmalloc(sizeof *entry->u.learn.fm);
+ entry->u.learn.ofpacts = ofpbuf_new(64);
+ xlate_learn_action__(ctx, learn, entry->u.learn.fm,
+ entry->u.learn.ofpacts);
+ } else if (ctx->xin->may_learn) {
+ uint64_t ofpacts_stub[1024 / 8];
+ struct ofputil_flow_mod fm;
+ struct ofpbuf ofpacts;
+
+ ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub);
+ xlate_learn_action__(ctx, learn, &fm, &ofpacts);
+ ofpbuf_uninit(&ofpacts);
}
}
}
ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
- &ctx->xout->odp_actions,
- &ctx->xout->wc);
+ ctx->xout->odp_actions,
+ &ctx->xout->wc,
+ ctx->xbridge->masked_set_action);
compose_flow_sample_cookie(os->probability, os->collector_set_id,
os->obs_domain_id, os->obs_point_id, &cookie);
- compose_sample_action(ctx->xbridge, &ctx->xout->odp_actions, &ctx->xin->flow,
- probability, &cookie, sizeof cookie.flow_sample);
+ compose_sample_action(ctx->xbridge, ctx->xout->odp_actions, &ctx->xin->flow,
+ probability, &cookie, sizeof cookie.flow_sample,
+ ODPP_NONE);
}
static bool
* disabled. If just learning is enabled, we need to have
* OFPP_NORMAL and the learning action have a look at the packet
* before we can drop it. */
- if (!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) {
+ if ((!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) ||
+ (!xport_rstp_forward_state(xport) && !xport_rstp_learn_state(xport))) {
return false;
}
ofpbuf_uninit(&action_list);
}
+static bool
+ofpact_needs_recirculation_after_mpls(const struct xlate_ctx *ctx,
+ const struct ofpact *a)
+{
+ struct flow_wildcards *wc = &ctx->xout->wc;
+ struct flow *flow = &ctx->xin->flow;
+
+ switch (a->type) {
+ case OFPACT_OUTPUT:
+ case OFPACT_GROUP:
+ case OFPACT_CONTROLLER:
+ case OFPACT_STRIP_VLAN:
+ case OFPACT_SET_VLAN_PCP:
+ case OFPACT_SET_VLAN_VID:
+ case OFPACT_ENQUEUE:
+ case OFPACT_PUSH_VLAN:
+ case OFPACT_SET_ETH_SRC:
+ case OFPACT_SET_ETH_DST:
+ case OFPACT_SET_TUNNEL:
+ case OFPACT_SET_QUEUE:
+ case OFPACT_POP_QUEUE:
+ case OFPACT_POP_MPLS:
+ case OFPACT_DEC_MPLS_TTL:
+ case OFPACT_SET_MPLS_TTL:
+ case OFPACT_SET_MPLS_TC:
+ case OFPACT_SET_MPLS_LABEL:
+ case OFPACT_NOTE:
+ case OFPACT_OUTPUT_REG:
+ case OFPACT_EXIT:
+ case OFPACT_METER:
+ case OFPACT_WRITE_METADATA:
+ case OFPACT_WRITE_ACTIONS:
+ case OFPACT_CLEAR_ACTIONS:
+ case OFPACT_SAMPLE:
+ return false;
+
+ case OFPACT_SET_IPV4_SRC:
+ case OFPACT_SET_IPV4_DST:
+ case OFPACT_SET_IP_DSCP:
+ case OFPACT_SET_IP_ECN:
+ case OFPACT_SET_IP_TTL:
+ case OFPACT_SET_L4_SRC_PORT:
+ case OFPACT_SET_L4_DST_PORT:
+ case OFPACT_RESUBMIT:
+ case OFPACT_STACK_PUSH:
+ case OFPACT_STACK_POP:
+ case OFPACT_DEC_TTL:
+ case OFPACT_MULTIPATH:
+ case OFPACT_BUNDLE:
+ case OFPACT_LEARN:
+ case OFPACT_FIN_TIMEOUT:
+ case OFPACT_GOTO_TABLE:
+ return true;
+
+ case OFPACT_REG_MOVE:
+ return (mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->dst.field) ||
+ mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->src.field));
+
+ case OFPACT_SET_FIELD:
+ return mf_is_l3_or_higher(ofpact_get_SET_FIELD(a)->field);
+
+ case OFPACT_PUSH_MPLS:
+ /* Recirculate if it is an IP packet with a zero ttl. This may
+ * indicate that the packet was previously MPLS and an MPLS pop action
+ * converted it to IP. In this case recirculating should reveal the IP
+ * TTL which is used as the basis for a new MPLS LSE. */
+ return (!flow_count_mpls_labels(flow, wc)
+ && flow->nw_ttl == 0
+ && is_ip_any(flow));
+ }
+
+ OVS_NOT_REACHED();
+}
+
static void
do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len,
struct xlate_ctx *ctx)
break;
}
+ if (ctx->was_mpls && ofpact_needs_recirculation_after_mpls(ctx, a)) {
+ compose_recirculate_action(ctx, ofpacts, a, ofpacts_len);
+ return;
+ }
+
switch (a->type) {
case OFPACT_OUTPUT:
xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port,
break;
case OFPACT_ENQUEUE:
+ memset(&wc->masks.skb_priority, 0xff,
+ sizeof wc->masks.skb_priority);
xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a));
break;
break;
case OFPACT_SET_QUEUE:
+ memset(&wc->masks.skb_priority, 0xff,
+ sizeof wc->masks.skb_priority);
xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id);
break;
case OFPACT_POP_QUEUE:
+ memset(&wc->masks.skb_priority, 0xff,
+ sizeof wc->masks.skb_priority);
flow->skb_priority = ctx->orig_skb_priority;
break;
nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc);
break;
- case OFPACT_REG_LOAD:
- nxm_execute_reg_load(ofpact_get_REG_LOAD(a), flow, wc);
- break;
-
case OFPACT_SET_FIELD:
set_field = ofpact_get_SET_FIELD(a);
mf = set_field->field;
}
mf_mask_field_and_prereqs(mf, &wc->masks);
- mf_set_flow_value(mf, &set_field->value, flow);
+ mf_set_flow_value_masked(mf, &set_field->value, &set_field->mask,
+ flow);
break;
case OFPACT_STACK_PUSH:
case OFPACT_GOTO_TABLE: {
struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a);
- ovs_assert(ctx->table_id < ogt->table_id);
+ /* 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);
xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
ogt->table_id, true, true);
break;
void
xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto,
- const struct flow *flow, struct rule_dpif *rule,
- uint16_t tcp_flags, const struct ofpbuf *packet)
+ const struct flow *flow, ofp_port_t in_port,
+ struct rule_dpif *rule, uint16_t tcp_flags,
+ const struct ofpbuf *packet)
{
xin->ofproto = ofproto;
xin->flow = *flow;
+ xin->flow.in_port.ofp_port = in_port;
xin->packet = packet;
xin->may_learn = packet != NULL;
xin->rule = rule;
xin->report_hook = NULL;
xin->resubmit_stats = NULL;
xin->skip_wildcards = false;
+ xin->odp_actions = NULL;
}
void
xlate_out_uninit(struct xlate_out *xout)
{
- if (xout) {
- ofpbuf_uninit(&xout->odp_actions);
+ if (xout && xout->odp_actions == &xout->odp_actions_buf) {
+ ofpbuf_uninit(xout->odp_actions);
}
}
dst->nf_output_iface = src->nf_output_iface;
dst->mirrors = src->mirrors;
- ofpbuf_use_stub(&dst->odp_actions, dst->odp_actions_stub,
+ dst->odp_actions = &dst->odp_actions_buf;
+ ofpbuf_use_stub(dst->odp_actions, dst->odp_actions_stub,
sizeof dst->odp_actions_stub);
- ofpbuf_put(&dst->odp_actions, ofpbuf_data(&src->odp_actions),
- ofpbuf_size(&src->odp_actions));
+ ofpbuf_put(dst->odp_actions, ofpbuf_data(src->odp_actions),
+ ofpbuf_size(src->odp_actions));
}
\f
static struct skb_priority_to_dscp *
return pdscp != NULL;
}
+static size_t
+count_skb_priorities(const struct xport *xport)
+{
+ return hmap_count(&xport->skb_priorities);
+}
+
static void
clear_skb_priorities(struct xport *xport)
{
const struct nlattr *a;
unsigned int left;
- NL_ATTR_FOR_EACH_UNSAFE (a, left, ofpbuf_data(&ctx->xout->odp_actions),
- ofpbuf_size(&ctx->xout->odp_actions)) {
+ 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;
return false;
}
+#if defined(__linux__)
+/* Returns the maximum number of packets that the Linux kernel is willing to
+ * queue up internally to certain kinds of software-implemented ports, or the
+ * default (and rarely modified) value if it cannot be determined. */
+static int
+netdev_max_backlog(void)
+{
+ static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
+ static int max_backlog = 1000; /* The normal default value. */
+
+ if (ovsthread_once_start(&once)) {
+ static const char filename[] = "/proc/sys/net/core/netdev_max_backlog";
+ FILE *stream;
+ int n;
+
+ stream = fopen(filename, "r");
+ if (!stream) {
+ VLOG_WARN("%s: open failed (%s)", filename, ovs_strerror(errno));
+ } else {
+ if (fscanf(stream, "%d", &n) != 1) {
+ VLOG_WARN("%s: read error", filename);
+ } else if (n <= 100) {
+ VLOG_WARN("%s: unexpectedly small value %d", filename, n);
+ } else {
+ max_backlog = n;
+ }
+ fclose(stream);
+ }
+ ovsthread_once_done(&once);
+
+ VLOG_DBG("%s: using %d max_backlog", filename, max_backlog);
+ }
+
+ return max_backlog;
+}
+
+/* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in
+ * 'odp_actions'. */
+static int
+count_output_actions(const struct ofpbuf *odp_actions)
+{
+ const struct nlattr *a;
+ size_t left;
+ int n = 0;
+
+ NL_ATTR_FOR_EACH_UNSAFE (a, left, ofpbuf_data(odp_actions),
+ ofpbuf_size(odp_actions)) {
+ if (a->nla_type == OVS_ACTION_ATTR_OUTPUT) {
+ n++;
+ }
+ }
+ return n;
+}
+#endif /* defined(__linux__) */
+
+/* Returns true if 'odp_actions' contains more output actions than the datapath
+ * can reliably handle in one go. On Linux, this is the value of the
+ * net.core.netdev_max_backlog sysctl, which limits the maximum number of
+ * packets that the kernel is willing to queue up for processing while the
+ * datapath is processing a set of actions. */
+static bool
+too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED)
+{
+#ifdef __linux__
+ return (ofpbuf_size(odp_actions) / NL_A_U32_SIZE > netdev_max_backlog()
+ && count_output_actions(odp_actions) > netdev_max_backlog());
+#else
+ /* OSes other than Linux might have similar limits, but we don't know how
+ * to determine them.*/
+ return false;
+#endif
+}
+
/* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts'
* into datapath actions in 'odp_actions', using 'ctx'.
*
ctx.xout->has_fin_timeout = false;
ctx.xout->nf_output_iface = NF_OUT_DROP;
ctx.xout->mirrors = 0;
- ofpbuf_use_stub(&ctx.xout->odp_actions, ctx.xout->odp_actions_stub,
- sizeof ctx.xout->odp_actions_stub);
- ofpbuf_reserve(&ctx.xout->odp_actions, NL_A_U32_SIZE);
+
+ xout->odp_actions = xin->odp_actions;
+ if (!xout->odp_actions) {
+ xout->odp_actions = &xout->odp_actions_buf;
+ ofpbuf_use_stub(xout->odp_actions, xout->odp_actions_stub,
+ sizeof xout->odp_actions_stub);
+ }
+ ofpbuf_reserve(xout->odp_actions, NL_A_U32_SIZE);
ctx.xbridge = xbridge_lookup(xcfg, xin->ofproto);
if (!ctx.xbridge) {
flow_wildcards_init_catchall(wc);
memset(&wc->masks.in_port, 0xff, sizeof wc->masks.in_port);
- memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority);
memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type);
if (is_ip_any(flow)) {
wc->masks.nw_frag |= FLOW_NW_FRAG_MASK;
ctx.table_id = 0;
ctx.exit = false;
ctx.use_recirc = false;
+ ctx.was_mpls = false;
if (!xin->ofpacts && !ctx.rule) {
ctx.table_id = rule_dpif_lookup(ctx.xbridge->ofproto, flow,
add_sflow_action(&ctx);
add_ipfix_action(&ctx);
- sample_actions_len = ofpbuf_size(&ctx.xout->odp_actions);
+ 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)) {
- ofpbuf_set_size(&ctx.xout->odp_actions, sample_actions_len);
+ if (in_port && (!xport_stp_forward_state(in_port) ||
+ !xport_rstp_forward_state(in_port))) {
+ ofpbuf_set_size(ctx.xout->odp_actions, sample_actions_len);
}
}
}
}
- if (nl_attr_oversized(ofpbuf_size(&ctx.xout->odp_actions))) {
+ 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
* prevent the flow from being installed. */
COVERAGE_INC(xlate_actions_oversize);
ctx.xout->slow |= SLOW_ACTION;
+ } else if (too_many_output_actions(ctx.xout->odp_actions)) {
+ COVERAGE_INC(xlate_actions_too_many_output);
+ ctx.xout->slow |= SLOW_ACTION;
}
if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
/* Push stats and perform side effects of flow translation. */
void
-xlate_push_stats(struct xlate_cache *xcache, bool may_learn,
+xlate_push_stats(struct xlate_cache *xcache,
const struct dpif_flow_stats *stats)
{
struct xc_entry *entry;
struct ofpbuf entries = xcache->entries;
+ if (!stats->n_packets) {
+ return;
+ }
+
XC_ENTRY_FOR_EACH (entry, entries, xcache) {
switch (entry->type) {
case XC_RULE:
stats->n_packets, stats->n_bytes);
break;
case XC_LEARN:
- if (may_learn) {
- struct rule_dpif *rule = entry->u.learn.rule;
-
- /* Reset the modified time for a rule that is equivalent to
- * the currently cached rule. If the rule is not the exact
- * rule we have cached, update the reference that we have. */
- entry->u.learn.rule = ofproto_dpif_refresh_rule(rule);
- }
+ ofproto_dpif_flow_mod(entry->u.learn.ofproto, entry->u.learn.fm);
break;
case XC_NORMAL:
xlate_cache_normal(entry->u.normal.ofproto, entry->u.normal.flow,
static void
xlate_cache_clear_netflow(struct netflow *netflow, struct flow *flow)
{
- netflow_expire(netflow, flow);
netflow_flow_clear(netflow, flow);
netflow_unref(netflow);
free(flow);
mbridge_unref(entry->u.mirror.mbridge);
break;
case XC_LEARN:
- /* 'u.learn.rule' is the learned rule. */
- rule_dpif_unref(entry->u.learn.rule);
+ free(entry->u.learn.fm);
+ ofpbuf_delete(entry->u.learn.ofpacts);
break;
case XC_NORMAL:
free(entry->u.normal.flow);
projects / cascardo / ovs.git / blobdiff