2 * Copyright (c) 2009, 2010, 2011 Nicira Networks.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
17 #ifndef NETDEV_PROVIDER_H
18 #define NETDEV_PROVIDER_H 1
20 /* Generic interface to network devices. */
32 /* A network device (e.g. an Ethernet device).
34 * This structure should be treated as opaque by network device
37 char *name; /* Name of network device. */
38 const struct netdev_class *netdev_class; /* Functions to control
40 int ref_cnt; /* Times this devices was opened. */
41 struct shash_node *node; /* Pointer to element in global map. */
44 void netdev_dev_init(struct netdev_dev *, const char *name,
45 const struct netdev_class *);
46 void netdev_dev_uninit(struct netdev_dev *, bool destroy);
47 const char *netdev_dev_get_type(const struct netdev_dev *);
48 const struct netdev_class *netdev_dev_get_class(const struct netdev_dev *);
49 const char *netdev_dev_get_name(const struct netdev_dev *);
50 struct netdev_dev *netdev_dev_from_name(const char *name);
51 void netdev_dev_get_devices(const struct netdev_class *,
52 struct shash *device_list);
54 static inline void netdev_dev_assert_class(const struct netdev_dev *netdev_dev,
55 const struct netdev_class *class_)
57 assert(netdev_dev->netdev_class == class_);
60 /* A instance of an open network device.
62 * This structure should be treated as opaque by network device
65 struct netdev_dev *netdev_dev; /* Parent netdev_dev. */
66 struct list node; /* Element in global list. */
68 enum netdev_flags save_flags; /* Initial device flags. */
69 enum netdev_flags changed_flags; /* Flags that we changed. */
72 void netdev_init(struct netdev *, struct netdev_dev *);
73 void netdev_uninit(struct netdev *, bool close);
74 struct netdev_dev *netdev_get_dev(const struct netdev *);
76 static inline void netdev_assert_class(const struct netdev *netdev,
77 const struct netdev_class *netdev_class)
79 netdev_dev_assert_class(netdev_get_dev(netdev), netdev_class);
82 /* Network device class structure, to be defined by each implementation of a
85 * These functions return 0 if successful or a positive errno value on failure,
86 * except where otherwise noted. */
88 /* Type of netdevs in this class, e.g. "system", "tap", "gre", etc.
90 * One of the providers should supply a "system" type, since this is
91 * the type assumed if no type is specified when opening a netdev.
92 * The "system" type corresponds to an existing network device on
96 /* Called when the netdev provider is registered, typically at program
97 * startup. Returning an error from this function will prevent any network
98 * device in this class from being opened.
100 * This function may be set to null if a network device class needs no
101 * initialization at registration time. */
104 /* Performs periodic work needed by netdevs of this class. May be null if
105 * no periodic work is necessary. */
108 /* Arranges for poll_block() to wake up if the "run" member function needs
109 * to be called. Implementations are additionally required to wake
110 * whenever something changes in any of its netdevs which would cause their
111 * ->change_seq() function to change its result. May be null if nothing is
115 /* Attempts to create a network device named 'name' in 'netdev_class'. On
116 * success sets 'netdev_devp' to the newly created device. */
117 int (*create)(const struct netdev_class *netdev_class, const char *name,
118 struct netdev_dev **netdev_devp);
120 /* Destroys 'netdev_dev'.
122 * Netdev devices maintain a reference count that is incremented on
123 * netdev_open() and decremented on netdev_close(). If 'netdev_dev'
124 * has a non-zero reference count, then this function will not be
126 void (*destroy)(struct netdev_dev *netdev_dev);
128 /* Fetches the device 'netdev_dev''s configuration, storing it in 'args'.
129 * The caller owns 'args' and pre-initializes it to an empty shash.
131 * If this netdev class does not have any configuration options, this may
132 * be a null pointer. */
133 int (*get_config)(struct netdev_dev *netdev_dev, struct shash *args);
135 /* Changes the device 'netdev_dev''s configuration to 'args'.
137 * If this netdev class does not support configuration, this may be a null
139 int (*set_config)(struct netdev_dev *netdev_dev, const struct shash *args);
141 /* Attempts to open a network device. On success, sets 'netdevp'
142 * to the new network device. */
143 int (*open)(struct netdev_dev *netdev_dev, struct netdev **netdevp);
145 /* Closes 'netdev'. */
146 void (*close)(struct netdev *netdev);
148 /* ## ----------------- ## */
149 /* ## Receiving Packets ## */
150 /* ## ----------------- ## */
152 /* The network provider interface is mostly used for inspecting and configuring
153 * device "metadata", not for sending and receiving packets directly. It may
154 * be impractical to implement these functions on some operating systems and
155 * hardware. These functions may all be NULL in such cases.
157 * (However, the "dpif-netdev" implementation, which is the easiest way to
158 * integrate Open vSwitch with a new operating system or hardware, does require
159 * the ability to receive packets.) */
161 /* Attempts to set up 'netdev' for receiving packets with ->recv().
162 * Returns 0 if successful, otherwise a positive errno value. Return
163 * EOPNOTSUPP to indicate that the network device does not implement packet
164 * reception through this interface. This function may be set to null if
165 * it would always return EOPNOTSUPP anyhow. (This will prevent the
166 * network device from being usefully used by the netdev-based "userspace
168 int (*listen)(struct netdev *netdev);
170 /* Attempts to receive a packet from 'netdev' into the 'size' bytes in
171 * 'buffer'. If successful, returns the number of bytes in the received
172 * packet, otherwise a negative errno value. Returns -EAGAIN immediately
173 * if no packet is ready to be received.
175 * This function can only be expected to return a packet if ->listen() has
176 * been called successfully.
178 * May be null if not needed, such as for a network device that does not
179 * implement packet reception through the 'recv' member function. */
180 int (*recv)(struct netdev *netdev, void *buffer, size_t size);
182 /* Registers with the poll loop to wake up from the next call to
183 * poll_block() when a packet is ready to be received with netdev_recv() on
186 * May be null if not needed, such as for a network device that does not
187 * implement packet reception through the 'recv' member function. */
188 void (*recv_wait)(struct netdev *netdev);
190 /* Discards all packets waiting to be received from 'netdev'.
192 * May be null if not needed, such as for a network device that does not
193 * implement packet reception through the 'recv' member function. */
194 int (*drain)(struct netdev *netdev);
196 /* Sends the 'size'-byte packet in 'buffer' on 'netdev'. Returns 0 if
197 * successful, otherwise a positive errno value. Returns EAGAIN without
198 * blocking if the packet cannot be queued immediately. Returns EMSGSIZE
199 * if a partial packet was transmitted or if the packet is too big or too
200 * small to transmit on the device.
202 * The caller retains ownership of 'buffer' in all cases.
204 * The network device is expected to maintain a packet transmission queue,
205 * so that the caller does not ordinarily have to do additional queuing of
208 * May return EOPNOTSUPP if a network device does not implement packet
209 * transmission through this interface. This function may be set to null
210 * if it would always return EOPNOTSUPP anyhow. (This will prevent the
211 * network device from being usefully used by the netdev-based "userspace
212 * datapath". It will also prevent the OVS implementation of bonding from
213 * working properly over 'netdev'.) */
214 int (*send)(struct netdev *netdev, const void *buffer, size_t size);
216 /* Registers with the poll loop to wake up from the next call to
217 * poll_block() when the packet transmission queue for 'netdev' has
218 * sufficient room to transmit a packet with netdev_send().
220 * The network device is expected to maintain a packet transmission queue,
221 * so that the caller does not ordinarily have to do additional queuing of
222 * packets. Thus, this function is unlikely to ever be useful.
224 * May be null if not needed, such as for a network device that does not
225 * implement packet transmission through the 'send' member function. */
226 void (*send_wait)(struct netdev *netdev);
228 /* Sets 'netdev''s Ethernet address to 'mac' */
229 int (*set_etheraddr)(struct netdev *netdev, const uint8_t mac[6]);
231 /* Retrieves 'netdev''s Ethernet address into 'mac'. */
232 int (*get_etheraddr)(const struct netdev *netdev, uint8_t mac[6]);
234 /* Retrieves 'netdev''s MTU into '*mtup'.
236 * The MTU is the maximum size of transmitted (and received) packets, in
237 * bytes, not including the hardware header; thus, this is typically 1500
238 * bytes for Ethernet devices.
240 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
241 * this function should return EOPNOTSUPP. This function may be set to
242 * null if it would always return EOPNOTSUPP. */
243 int (*get_mtu)(const struct netdev *netdev, int *mtup);
245 /* Sets 'netdev''s MTU to 'mtu'.
247 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
248 * this function should return EOPNOTSUPP. This function may be set to
249 * null if it would always return EOPNOTSUPP. */
250 int (*set_mtu)(const struct netdev *netdev, int mtu);
252 /* Returns the ifindex of 'netdev', if successful, as a positive number.
253 * On failure, returns a negative errno value.
255 * The desired semantics of the ifindex value are a combination of those
256 * specified by POSIX for if_nametoindex() and by SNMP for ifIndex. An
257 * ifindex value should be unique within a host and remain stable at least
258 * until reboot. SNMP says an ifindex "ranges between 1 and the value of
259 * ifNumber" but many systems do not follow this rule anyhow.
261 * This function may be set to null if it would always return -EOPNOTSUPP.
263 int (*get_ifindex)(const struct netdev *netdev);
265 /* Sets 'carrier' to true if carrier is active (link light is on) on
268 * May be null if device does not provide carrier status (will be always
269 * up as long as device is up).
271 int (*get_carrier)(const struct netdev *netdev, bool *carrier);
273 /* Forces ->get_carrier() to poll 'netdev''s MII registers for link status
274 * instead of checking 'netdev''s carrier. 'netdev''s MII registers will
275 * be polled once ever 'interval' milliseconds. If 'netdev' does not
276 * support MII, another method may be used as a fallback. If 'interval' is
277 * less than or equal to zero, reverts ->get_carrier() to its normal
280 * Most network devices won't support this feature and will set this
281 * function pointer to NULL, which is equivalent to returning EOPNOTSUPP.
283 int (*set_miimon_interval)(struct netdev *netdev, long long int interval);
285 /* Retrieves current device stats for 'netdev' into 'stats'.
287 * A network device that supports some statistics but not others, it should
288 * set the values of the unsupported statistics to all-1-bits
290 int (*get_stats)(const struct netdev *netdev, struct netdev_stats *);
292 /* Sets the device stats for 'netdev' to 'stats'.
294 * Most network devices won't support this feature and will set this
295 * function pointer to NULL, which is equivalent to returning EOPNOTSUPP.
297 * Some network devices might only allow setting their stats to 0. */
298 int (*set_stats)(struct netdev *netdev, const struct netdev_stats *);
300 /* Stores the features supported by 'netdev' into each of '*current',
301 * '*advertised', '*supported', and '*peer'. Each value is a bitmap of
302 * "enum ofp_port_features" bits, in host byte order.
304 * This function may be set to null if it would always return EOPNOTSUPP.
306 int (*get_features)(const struct netdev *netdev,
307 uint32_t *current, uint32_t *advertised,
308 uint32_t *supported, uint32_t *peer);
310 /* Set the features advertised by 'netdev' to 'advertise', which is a
311 * bitmap of "enum ofp_port_features" bits, in host byte order.
313 * This function may be set to null for a network device that does not
314 * support configuring advertisements. */
315 int (*set_advertisements)(struct netdev *netdev, uint32_t advertise);
317 /* If 'netdev' is a VLAN network device (e.g. one created with vconfig(8)),
318 * sets '*vlan_vid' to the VLAN VID associated with that device and returns
321 * Returns ENOENT if 'netdev' is a network device that is not a
324 * This function should be set to null if it doesn't make any sense for
325 * your network device (it probably doesn't). */
326 int (*get_vlan_vid)(const struct netdev *netdev, int *vlan_vid);
328 /* Attempts to set input rate limiting (policing) policy, such that up to
329 * 'kbits_rate' kbps of traffic is accepted, with a maximum accumulative
330 * burst size of 'kbits' kb.
332 * This function may be set to null if policing is not supported. */
333 int (*set_policing)(struct netdev *netdev, unsigned int kbits_rate,
334 unsigned int kbits_burst);
336 /* Adds to 'types' all of the forms of QoS supported by 'netdev', or leaves
337 * it empty if 'netdev' does not support QoS. Any names added to 'types'
338 * should be documented as valid for the "type" column in the "QoS" table
339 * in vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
341 * Every network device must support disabling QoS with a type of "", but
342 * this function must not add "" to 'types'.
344 * The caller is responsible for initializing 'types' (e.g. with
345 * sset_init()) before calling this function. The caller retains ownership
348 * May be NULL if 'netdev' does not support QoS at all. */
349 int (*get_qos_types)(const struct netdev *netdev, struct sset *types);
351 /* Queries 'netdev' for its capabilities regarding the specified 'type' of
352 * QoS. On success, initializes 'caps' with the QoS capabilities.
354 * Should return EOPNOTSUPP if 'netdev' does not support 'type'. May be
355 * NULL if 'netdev' does not support QoS at all. */
356 int (*get_qos_capabilities)(const struct netdev *netdev,
358 struct netdev_qos_capabilities *caps);
360 /* Queries 'netdev' about its currently configured form of QoS. If
361 * successful, stores the name of the current form of QoS into '*typep'
362 * and any details of configuration as string key-value pairs in
365 * A '*typep' of "" indicates that QoS is currently disabled on 'netdev'.
367 * The caller initializes 'details' before calling this function. The
368 * caller takes ownership of the string key-values pairs added to
371 * The netdev retains ownership of '*typep'.
373 * '*typep' will be one of the types returned by netdev_get_qos_types() for
374 * 'netdev'. The contents of 'details' should be documented as valid for
375 * '*typep' in the "other_config" column in the "QoS" table in
376 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
378 * May be NULL if 'netdev' does not support QoS at all. */
379 int (*get_qos)(const struct netdev *netdev,
380 const char **typep, struct shash *details);
382 /* Attempts to reconfigure QoS on 'netdev', changing the form of QoS to
383 * 'type' with details of configuration from 'details'.
385 * On error, the previous QoS configuration is retained.
387 * When this function changes the type of QoS (not just 'details'), this
388 * also resets all queue configuration for 'netdev' to their defaults
389 * (which depend on the specific type of QoS). Otherwise, the queue
390 * configuration for 'netdev' is unchanged.
392 * 'type' should be "" (to disable QoS) or one of the types returned by
393 * netdev_get_qos_types() for 'netdev'. The contents of 'details' should
394 * be documented as valid for the given 'type' in the "other_config" column
395 * in the "QoS" table in vswitchd/vswitch.xml (which is built as
396 * ovs-vswitchd.conf.db(8)).
398 * May be NULL if 'netdev' does not support QoS at all. */
399 int (*set_qos)(struct netdev *netdev,
400 const char *type, const struct shash *details);
402 /* Queries 'netdev' for information about the queue numbered 'queue_id'.
403 * If successful, adds that information as string key-value pairs to
404 * 'details'. Returns 0 if successful, otherwise a positive errno value.
406 * Should return EINVAL if 'queue_id' is greater than or equal to the
407 * number of supported queues (as reported in the 'n_queues' member of
408 * struct netdev_qos_capabilities by 'get_qos_capabilities').
410 * The caller initializes 'details' before calling this function. The
411 * caller takes ownership of the string key-values pairs added to
414 * The returned contents of 'details' should be documented as valid for the
415 * given 'type' in the "other_config" column in the "Queue" table in
416 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
418 int (*get_queue)(const struct netdev *netdev,
419 unsigned int queue_id, struct shash *details);
421 /* Configures the queue numbered 'queue_id' on 'netdev' with the key-value
422 * string pairs in 'details'. The contents of 'details' should be
423 * documented as valid for the given 'type' in the "other_config" column in
424 * the "Queue" table in vswitchd/vswitch.xml (which is built as
425 * ovs-vswitchd.conf.db(8)). Returns 0 if successful, otherwise a positive
426 * errno value. On failure, the given queue's configuration should be
429 * Should return EINVAL if 'queue_id' is greater than or equal to the
430 * number of supported queues (as reported in the 'n_queues' member of
431 * struct netdev_qos_capabilities by 'get_qos_capabilities'), or if
432 * 'details' is invalid for the type of queue.
434 * This function does not modify 'details', and the caller retains
437 * May be NULL if 'netdev' does not support QoS at all. */
438 int (*set_queue)(struct netdev *netdev,
439 unsigned int queue_id, const struct shash *details);
441 /* Attempts to delete the queue numbered 'queue_id' from 'netdev'.
443 * Should return EINVAL if 'queue_id' is greater than or equal to the
444 * number of supported queues (as reported in the 'n_queues' member of
445 * struct netdev_qos_capabilities by 'get_qos_capabilities'). Should
446 * return EOPNOTSUPP if 'queue_id' is valid but may not be deleted (e.g. if
447 * 'netdev' has a fixed set of queues with the current QoS mode).
449 * May be NULL if 'netdev' does not support QoS at all, or if all of its
450 * QoS modes have fixed sets of queues. */
451 int (*delete_queue)(struct netdev *netdev, unsigned int queue_id);
453 /* Obtains statistics about 'queue_id' on 'netdev'. Fills 'stats' with the
454 * queue's statistics. May set individual members of 'stats' to all-1-bits
455 * if the statistic is unavailable.
457 * May be NULL if 'netdev' does not support QoS at all. */
458 int (*get_queue_stats)(const struct netdev *netdev, unsigned int queue_id,
459 struct netdev_queue_stats *stats);
461 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
462 * ID, its configuration, and the 'aux' specified by the caller. The order
463 * of iteration is unspecified, but (when successful) each queue is visited
466 * 'cb' will not modify or free the 'details' argument passed in. */
467 int (*dump_queues)(const struct netdev *netdev,
468 void (*cb)(unsigned int queue_id,
469 const struct shash *details,
473 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
474 * ID, its statistics, and the 'aux' specified by the caller. The order of
475 * iteration is unspecified, but (when successful) each queue must be
476 * visited exactly once.
478 * 'cb' will not modify or free the statistics passed in. */
479 int (*dump_queue_stats)(const struct netdev *netdev,
480 void (*cb)(unsigned int queue_id,
481 struct netdev_queue_stats *,
485 /* If 'netdev' has an assigned IPv4 address, sets '*address' to that
486 * address and '*netmask' to the associated netmask.
488 * The following error values have well-defined meanings:
490 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv4 address.
492 * - EOPNOTSUPP: No IPv4 network stack attached to 'netdev'.
494 * This function may be set to null if it would always return EOPNOTSUPP
496 int (*get_in4)(const struct netdev *netdev, struct in_addr *address,
497 struct in_addr *netmask);
499 /* Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
500 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared.
502 * This function may be set to null if it would always return EOPNOTSUPP
504 int (*set_in4)(struct netdev *netdev, struct in_addr addr,
505 struct in_addr mask);
507 /* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address.
509 * The following error values have well-defined meanings:
511 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv6 address.
513 * - EOPNOTSUPP: No IPv6 network stack attached to 'netdev'.
515 * This function may be set to null if it would always return EOPNOTSUPP
517 int (*get_in6)(const struct netdev *netdev, struct in6_addr *in6);
519 /* Adds 'router' as a default IP gateway for the TCP/IP stack that
520 * corresponds to 'netdev'.
522 * This function may be set to null if it would always return EOPNOTSUPP
524 int (*add_router)(struct netdev *netdev, struct in_addr router);
526 /* Looks up the next hop for 'host'. If succesful, stores the next hop
527 * gateway's address (0 if 'host' is on a directly connected network) in
528 * '*next_hop' and a copy of the name of the device to reach 'host' in
529 * '*netdev_name', and returns 0. The caller is responsible for freeing
530 * '*netdev_name' (by calling free()).
532 * This function may be set to null if it would always return EOPNOTSUPP
534 int (*get_next_hop)(const struct in_addr *host, struct in_addr *next_hop,
537 /* Retrieves the status of the device.
539 * Populates 'sh' with key-value pairs representing the status of the
540 * device. A device's status is a set of key-value string pairs
541 * representing netdev type specific information. For more information see
542 * ovs-vswitchd.conf.db(5).
544 * The data of 'sh' are heap allocated strings which the caller is
545 * responsible for deallocating.
547 * This function may be set to null if it would always return EOPNOTSUPP
549 int (*get_status)(const struct netdev *netdev, struct shash *sh);
551 /* Looks up the ARP table entry for 'ip' on 'netdev' and stores the
552 * corresponding MAC address in 'mac'. A return value of ENXIO, in
553 * particular, indicates that there is no ARP table entry for 'ip' on
556 * This function may be set to null if it would always return EOPNOTSUPP
558 int (*arp_lookup)(const struct netdev *netdev, ovs_be32 ip,
561 /* Retrieves the current set of flags on 'netdev' into '*old_flags'.
562 * Then, turns off the flags that are set to 1 in 'off' and turns on the
563 * flags that are set to 1 in 'on'. (No bit will be set to 1 in both 'off'
564 * and 'on'; that is, off & on == 0.)
566 * This function may be invoked from a signal handler. Therefore, it
567 * should not do anything that is not signal-safe (such as logging). */
568 int (*update_flags)(struct netdev *netdev, enum netdev_flags off,
569 enum netdev_flags on, enum netdev_flags *old_flags);
571 /* Returns a sequence number which indicates changes in one of 'netdev''s
572 * properties. The returned sequence number must be nonzero so that
573 * callers have a value which they may use as a reset when tracking
576 * Minimally, the returned sequence number is required to change whenever
577 * 'netdev''s flags, features, ethernet address, or carrier changes. The
578 * returned sequence number is allowed to change even when 'netdev' doesn't
579 * change, although implementations should try to avoid this. */
580 unsigned int (*change_seq)(const struct netdev *netdev);
583 int netdev_register_provider(const struct netdev_class *);
584 int netdev_unregister_provider(const char *type);
585 const struct netdev_class *netdev_lookup_provider(const char *type);
587 extern const struct netdev_class netdev_linux_class;
588 extern const struct netdev_class netdev_internal_class;
589 extern const struct netdev_class netdev_tap_class;
595 #endif /* netdev.h */