*
* In Open vSwitch userspace, "struct flow" is the typical way to describe
* a flow, but the datapath interface uses a different data format to
- * allow ABI forward- and backward-compatibility. datapath/README
+ * allow ABI forward- and backward-compatibility. datapath/README.md
* describes the rationale and design. Refer to OVS_KEY_ATTR_* and
- * "struct ovs_key_*" in include/linux/openvswitch.h for details.
+ * "struct ovs_key_*" in include/odp-netlink.h for details.
* lib/odp-util.h defines several functions for working with these flows.
*
* - A "mask" that, for each bit in the flow, specifies whether the datapath
* within a flow. Some examples of actions are OVS_ACTION_ATTR_OUTPUT,
* which transmits the packet out a port, and OVS_ACTION_ATTR_SET, which
* modifies packet headers. Refer to OVS_ACTION_ATTR_* and "struct
- * ovs_action_*" in include/linux/openvswitch.h for details.
- * lib/odp-util.h defines several functions for working with datapath
- * actions.
+ * ovs_action_*" in include/odp-netlink.h for details. lib/odp-util.h
+ * defines several functions for working with datapath actions.
*
* The actions list may be empty. This indicates that nothing should be
* done to matching packets, that is, they should be dropped.
#include <stddef.h>
#include <stdint.h>
#include "netdev.h"
-#include "ofpbuf.h"
+#include "dp-packet.h"
#include "openflow/openflow.h"
+#include "ovs-numa.h"
#include "packets.h"
#include "util.h"
#endif
struct dpif;
+struct dpif_class;
+struct dpif_flow;
struct ds;
struct flow;
+struct flow_wildcards;
struct nlattr;
struct sset;
-struct dpif_class;
int dp_register_provider(const struct dpif_class *);
int dp_unregister_provider(const char *type);
int dpif_create_and_open(const char *name, const char *type, struct dpif **);
void dpif_close(struct dpif *);
-void dpif_run(struct dpif *);
+bool dpif_run(struct dpif *);
void dpif_wait(struct dpif *);
const char *dpif_name(const struct dpif *);
uint16_t tcp_flags;
};
-void dpif_flow_stats_extract(const struct flow *, const struct ofpbuf *packet,
+void dpif_flow_stats_extract(const struct flow *, const struct dp_packet *packet,
long long int used, struct dpif_flow_stats *);
void dpif_flow_stats_format(const struct dpif_flow_stats *, struct ds *);
enum dpif_flow_put_flags {
DPIF_FP_CREATE = 1 << 0, /* Allow creating a new flow. */
DPIF_FP_MODIFY = 1 << 1, /* Allow modifying an existing flow. */
- DPIF_FP_ZERO_STATS = 1 << 2 /* Zero the stats of an existing flow. */
+ DPIF_FP_ZERO_STATS = 1 << 2, /* Zero the stats of an existing flow. */
+ DPIF_FP_PROBE = 1 << 3 /* Suppress error messages, if any. */
};
+bool dpif_probe_feature(struct dpif *, const char *name,
+ const struct ofpbuf *key, const ovs_u128 *ufid);
+void dpif_flow_hash(const struct dpif *, const void *key, size_t key_len,
+ ovs_u128 *hash);
int dpif_flow_flush(struct dpif *);
int dpif_flow_put(struct dpif *, enum dpif_flow_put_flags,
const struct nlattr *key, size_t key_len,
const struct nlattr *mask, size_t mask_len,
const struct nlattr *actions, size_t actions_len,
+ const ovs_u128 *ufid, const unsigned pmd_id,
struct dpif_flow_stats *);
int dpif_flow_del(struct dpif *,
const struct nlattr *key, size_t key_len,
+ const ovs_u128 *ufid, const unsigned pmd_id,
struct dpif_flow_stats *);
-int dpif_flow_get(const struct dpif *,
+int dpif_flow_get(struct dpif *,
const struct nlattr *key, size_t key_len,
- struct ofpbuf **actionsp, struct dpif_flow_stats *);
+ const ovs_u128 *ufid, const unsigned pmd_id,
+ struct ofpbuf *, struct dpif_flow *);
\f
/* Flow dumping interface
* ======================
*
* All error reporting is deferred to the call to dpif_flow_dump_destroy().
*/
-struct dpif_flow_dump *dpif_flow_dump_create(const struct dpif *);
+struct dpif_flow_dump *dpif_flow_dump_create(const struct dpif *, bool terse);
int dpif_flow_dump_destroy(struct dpif_flow_dump *);
struct dpif_flow_dump_thread *dpif_flow_dump_thread_create(
struct dpif_flow_dump *);
void dpif_flow_dump_thread_destroy(struct dpif_flow_dump_thread *);
+#define PMD_ID_NULL OVS_CORE_UNSPEC
+
/* A datapath flow as dumped by dpif_flow_dump_next(). */
struct dpif_flow {
const struct nlattr *key; /* Flow key, as OVS_KEY_ATTR_* attrs. */
size_t mask_len; /* 'mask' length in bytes. */
const struct nlattr *actions; /* Actions, as OVS_ACTION_ATTR_ */
size_t actions_len; /* 'actions' length in bytes. */
+ ovs_u128 ufid; /* Unique flow identifier. */
+ bool ufid_present; /* True if 'ufid' was provided by datapath.*/
+ unsigned pmd_id; /* Datapath poll mode driver id. */
struct dpif_flow_stats stats; /* Flow statistics. */
};
int dpif_flow_dump_next(struct dpif_flow_dump_thread *,
struct dpif_flow *flows, int max_flows);
+
+#define DPIF_FLOW_BUFSIZE 2048
\f
/* Operation batching interface.
*
DPIF_OP_FLOW_PUT = 1,
DPIF_OP_FLOW_DEL,
DPIF_OP_EXECUTE,
+ DPIF_OP_FLOW_GET,
};
+/* Add or modify a flow.
+ *
+ * The flow is specified by the Netlink attributes with types OVS_KEY_ATTR_* in
+ * the 'key_len' bytes starting at 'key'. The associated actions are specified
+ * by the Netlink attributes with types OVS_ACTION_ATTR_* in the 'actions_len'
+ * bytes starting at 'actions'.
+ *
+ * - If the flow's key does not exist in the dpif, then the flow will be
+ * added if 'flags' includes DPIF_FP_CREATE. Otherwise the operation will
+ * fail with ENOENT.
+ *
+ * If the operation succeeds, then 'stats', if nonnull, will be zeroed.
+ *
+ * - If the flow's key does exist in the dpif, then the flow's actions will
+ * be updated if 'flags' includes DPIF_FP_MODIFY. Otherwise the operation
+ * will fail with EEXIST. If the flow's actions are updated, then its
+ * statistics will be zeroed if 'flags' includes DPIF_FP_ZERO_STATS, and
+ * left as-is otherwise.
+ *
+ * If the operation succeeds, then 'stats', if nonnull, will be set to the
+ * flow's statistics before the update.
+ *
+ * - If the datapath implements multiple pmd thread with its own flow
+ * table, 'pmd_id' should be used to specify the particular polling
+ * thread for the operation.
+ */
struct dpif_flow_put {
/* Input. */
enum dpif_flow_put_flags flags; /* DPIF_FP_*. */
size_t mask_len; /* Length of 'mask' in bytes. */
const struct nlattr *actions; /* Actions to perform on flow. */
size_t actions_len; /* Length of 'actions' in bytes. */
+ const ovs_u128 *ufid; /* Optional unique flow identifier. */
+ unsigned pmd_id; /* Datapath poll mode driver id. */
/* Output. */
struct dpif_flow_stats *stats; /* Optional flow statistics. */
};
+/* Delete a flow.
+ *
+ * The flow is specified by the Netlink attributes with types OVS_KEY_ATTR_* in
+ * the 'key_len' bytes starting at 'key', or the unique identifier 'ufid'. If
+ * the flow was created using 'ufid', then 'ufid' must be specified to delete
+ * the flow. If both are specified, 'key' will be ignored for flow deletion.
+ * Succeeds with status 0 if the flow is deleted, or fails with ENOENT if the
+ * dpif does not contain such a flow.
+ *
+ * Callers should always provide the 'key' to improve dpif logging in the event
+ * of errors or unexpected behaviour.
+ *
+ * If the datapath implements multiple polling thread with its own flow table,
+ * 'pmd_id' should be used to specify the particular polling thread for the
+ * operation.
+ *
+ * If the operation succeeds, then 'stats', if nonnull, will be set to the
+ * flow's statistics before its deletion. */
struct dpif_flow_del {
/* Input. */
const struct nlattr *key; /* Flow to delete. */
size_t key_len; /* Length of 'key' in bytes. */
+ const ovs_u128 *ufid; /* Unique identifier of flow to delete. */
+ bool terse; /* OK to skip sending/receiving full flow
+ * info? */
+ unsigned pmd_id; /* Datapath poll mode driver id. */
/* Output. */
struct dpif_flow_stats *stats; /* Optional flow statistics. */
};
+/* Executes actions on a specified packet.
+ *
+ * Performs the 'actions_len' bytes of actions in 'actions' on the Ethernet
+ * frame in 'packet' and on the packet metadata in 'md'. May modify both
+ * 'packet' and 'md'.
+ *
+ * Some dpif providers do not implement every action. The Linux kernel
+ * datapath, in particular, does not implement ARP field modification. If
+ * 'needs_help' is true, the dpif layer executes in userspace all of the
+ * actions that it can, and for OVS_ACTION_ATTR_OUTPUT and
+ * OVS_ACTION_ATTR_USERSPACE actions it passes the packet through to the dpif
+ * implementation.
+ *
+ * This works even if 'actions_len' is too long for a Netlink attribute. */
struct dpif_execute {
- /* Raw support for execute passed along to the provider. */
+ /* Input. */
const struct nlattr *actions; /* Actions to execute on packet. */
size_t actions_len; /* Length of 'actions' in bytes. */
- struct ofpbuf *packet; /* Packet to execute. */
- struct pkt_metadata md; /* Packet metadata. */
-
- /* Some dpif providers do not implement every action. The Linux kernel
- * datapath, in particular, does not implement ARP field modification.
- *
- * If this member is set to true, the dpif layer executes in userspace all
- * of the actions that it can, and for OVS_ACTION_ATTR_OUTPUT and
- * OVS_ACTION_ATTR_USERSPACE actions it passes the packet through to the
- * dpif implementation. */
bool needs_help;
+ bool probe; /* Suppress error messages. */
+ unsigned int mtu; /* Maximum transmission unit to fragment.
+ 0 if not a fragmented packet */
+
+ /* Input, but possibly modified as a side effect of execution. */
+ struct dp_packet *packet; /* Packet to execute. */
+};
+
+/* Queries the dpif for a flow entry.
+ *
+ * The flow is specified by the Netlink attributes with types OVS_KEY_ATTR_* in
+ * the 'key_len' bytes starting at 'key', or the unique identifier 'ufid'. If
+ * the flow was created using 'ufid', then 'ufid' must be specified to fetch
+ * the flow. If both are specified, 'key' will be ignored for the flow query.
+ * 'buffer' must point to an initialized buffer, with a recommended size of
+ * DPIF_FLOW_BUFSIZE bytes.
+ *
+ * On success, 'flow' will be populated with the mask, actions and stats for
+ * the datapath flow corresponding to 'key'. The mask and actions may point
+ * within '*buffer', or may point at RCU-protected data. Therefore, callers
+ * that wish to hold these over quiescent periods must make a copy of these
+ * fields before quiescing.
+ *
+ * Callers should always provide 'key' to improve dpif logging in the event of
+ * errors or unexpected behaviour.
+ *
+ * If the datapath implements multiple polling thread with its own flow table,
+ * 'pmd_id' should be used to specify the particular polling thread for the
+ * operation.
+ *
+ * Succeeds with status 0 if the flow is fetched, or fails with ENOENT if no
+ * such flow exists. Other failures are indicated with a positive errno value.
+ */
+struct dpif_flow_get {
+ /* Input. */
+ const struct nlattr *key; /* Flow to get. */
+ size_t key_len; /* Length of 'key' in bytes. */
+ const ovs_u128 *ufid; /* Unique identifier of flow to get. */
+ unsigned pmd_id; /* Datapath poll mode driver id. */
+ struct ofpbuf *buffer; /* Storage for output parameters. */
+
+ /* Output. */
+ struct dpif_flow *flow; /* Resulting flow from datapath. */
};
int dpif_execute(struct dpif *, struct dpif_execute *);
struct dpif_flow_put flow_put;
struct dpif_flow_del flow_del;
struct dpif_execute execute;
+ struct dpif_flow_get flow_get;
} u;
};
struct dpif_upcall {
/* All types. */
enum dpif_upcall_type type;
- struct ofpbuf packet; /* Packet data. */
+ struct dp_packet packet; /* Packet data. */
struct nlattr *key; /* Flow key. */
size_t key_len; /* Length of 'key' in bytes. */
+ ovs_u128 ufid; /* Unique flow identifier for 'key'. */
+ struct nlattr *mru; /* Maximum receive unit. */
/* DPIF_UC_ACTION only. */
struct nlattr *userdata; /* Argument to OVS_ACTION_ATTR_USERSPACE. */
+ struct nlattr *out_tun_key; /* Output tunnel key. */
+ struct nlattr *actions; /* Argument to OVS_ACTION_ATTR_USERSPACE. */
};
+/* A callback to notify higher layer of dpif about to be purged, so that
+ * higher layer could try reacting to this (e.g. grabbing all flow stats
+ * before they are gone). This function is currently implemented only by
+ * dpif-netdev.
+ *
+ * The caller needs to provide the 'aux' pointer passed down by higher
+ * layer from the dpif_register_notify_cb() function and the 'pmd_id' of
+ * the polling thread.
+ */
+ typedef void dp_purge_callback(void *aux, unsigned pmd_id);
+
+void dpif_register_dp_purge_cb(struct dpif *, dp_purge_callback *, void *aux);
+
+/* A callback to process an upcall, currently implemented only by dpif-netdev.
+ *
+ * The caller provides the 'packet' and 'flow' to process, the corresponding
+ * 'ufid' as generated by dpif_flow_hash(), the polling thread id 'pmd_id',
+ * the 'type' of the upcall, and if 'type' is DPIF_UC_ACTION then the
+ * 'userdata' attached to the action.
+ *
+ * The callback must fill in 'actions' with the datapath actions to apply to
+ * 'packet'. 'wc' and 'put_actions' will either be both null or both nonnull.
+ * If they are nonnull, then the caller will install a flow entry to process
+ * all future packets that match 'flow' and 'wc'; the callback must store a
+ * wildcard mask suitable for that purpose into 'wc'. If the actions to store
+ * into the flow entry are the same as 'actions', then the callback may leave
+ * 'put_actions' empty; otherwise it must store the desired actions into
+ * 'put_actions'.
+ *
+ * Returns 0 if successful, ENOSPC if the flow limit has been reached and no
+ * flow should be installed, or some otherwise a positive errno value. */
+typedef int upcall_callback(const struct dp_packet *packet,
+ const struct flow *flow,
+ ovs_u128 *ufid,
+ unsigned pmd_id,
+ enum dpif_upcall_type type,
+ const struct nlattr *userdata,
+ struct ofpbuf *actions,
+ struct flow_wildcards *wc,
+ struct ofpbuf *put_actions,
+ void *aux);
+
+void dpif_register_upcall_cb(struct dpif *, upcall_callback *, void *aux);
+
int dpif_recv_set(struct dpif *, bool enable);
int dpif_handlers_set(struct dpif *, uint32_t n_handlers);
+int dpif_poll_threads_set(struct dpif *, unsigned int n_rxqs,
+ const char *cmask);
int dpif_recv(struct dpif *, uint32_t handler_id, struct dpif_upcall *,
struct ofpbuf *);
void dpif_recv_purge(struct dpif *);
void dpif_recv_wait(struct dpif *, uint32_t handler_id);
+void dpif_enable_upcall(struct dpif *);
+void dpif_disable_upcall(struct dpif *);
+
+void dpif_print_packet(struct dpif *, struct dpif_upcall *);
\f
/* Miscellaneous. */
int dpif_queue_to_priority(const struct dpif *, uint32_t queue_id,
uint32_t *priority);
+char *dpif_get_dp_version(const struct dpif *);
+bool dpif_supports_tnl_push_pop(const struct dpif *);
#ifdef __cplusplus
}
#endif