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 /* Enumerates the names of all network devices of this class.
150 * The caller has already initialized 'all_names' and might already have
151 * added some names to it. This function should not disturb any existing
152 * names in 'all_names'.
154 * If this netdev class does not support enumeration, this may be a null
156 int (*enumerate)(struct sset *all_names);
158 /* ## ----------------- ## */
159 /* ## Receiving Packets ## */
160 /* ## ----------------- ## */
162 /* The network provider interface is mostly used for inspecting and configuring
163 * device "metadata", not for sending and receiving packets directly. It may
164 * be impractical to implement these functions on some operating systems and
165 * hardware. These functions may all be NULL in such cases.
167 * (However, the "dpif-netdev" implementation, which is the easiest way to
168 * integrate Open vSwitch with a new operating system or hardware, does require
169 * the ability to receive packets.) */
171 /* Attempts to set up 'netdev' for receiving packets with ->recv().
172 * Returns 0 if successful, otherwise a positive errno value. Return
173 * EOPNOTSUPP to indicate that the network device does not implement packet
174 * reception through this interface. This function may be set to null if
175 * it would always return EOPNOTSUPP anyhow. (This will prevent the
176 * network device from being usefully used by the netdev-based "userspace
178 int (*listen)(struct netdev *netdev);
180 /* Attempts to receive a packet from 'netdev' into the 'size' bytes in
181 * 'buffer'. If successful, returns the number of bytes in the received
182 * packet, otherwise a negative errno value. Returns -EAGAIN immediately
183 * if no packet is ready to be received.
185 * This function can only be expected to return a packet if ->listen() has
186 * been called successfully.
188 * May be null if not needed, such as for a network device that does not
189 * implement packet reception through the 'recv' member function. */
190 int (*recv)(struct netdev *netdev, void *buffer, size_t size);
192 /* Registers with the poll loop to wake up from the next call to
193 * poll_block() when a packet is ready to be received with netdev_recv() on
196 * May be null if not needed, such as for a network device that does not
197 * implement packet reception through the 'recv' member function. */
198 void (*recv_wait)(struct netdev *netdev);
200 /* Discards all packets waiting to be received from 'netdev'.
202 * May be null if not needed, such as for a network device that does not
203 * implement packet reception through the 'recv' member function. */
204 int (*drain)(struct netdev *netdev);
206 /* Sends the 'size'-byte packet in 'buffer' on 'netdev'. Returns 0 if
207 * successful, otherwise a positive errno value. Returns EAGAIN without
208 * blocking if the packet cannot be queued immediately. Returns EMSGSIZE
209 * if a partial packet was transmitted or if the packet is too big or too
210 * small to transmit on the device.
212 * The caller retains ownership of 'buffer' in all cases.
214 * The network device is expected to maintain a packet transmission queue,
215 * so that the caller does not ordinarily have to do additional queuing of
218 * May return EOPNOTSUPP if a network device does not implement packet
219 * transmission through this interface. This function may be set to null
220 * if it would always return EOPNOTSUPP anyhow. (This will prevent the
221 * network device from being usefully used by the netdev-based "userspace
222 * datapath". It will also prevent the OVS implementation of bonding from
223 * working properly over 'netdev'.) */
224 int (*send)(struct netdev *netdev, const void *buffer, size_t size);
226 /* Registers with the poll loop to wake up from the next call to
227 * poll_block() when the packet transmission queue for 'netdev' has
228 * sufficient room to transmit a packet with netdev_send().
230 * The network device is expected to maintain a packet transmission queue,
231 * so that the caller does not ordinarily have to do additional queuing of
232 * packets. Thus, this function is unlikely to ever be useful.
234 * May be null if not needed, such as for a network device that does not
235 * implement packet transmission through the 'send' member function. */
236 void (*send_wait)(struct netdev *netdev);
238 /* Sets 'netdev''s Ethernet address to 'mac' */
239 int (*set_etheraddr)(struct netdev *netdev, const uint8_t mac[6]);
241 /* Retrieves 'netdev''s Ethernet address into 'mac'. */
242 int (*get_etheraddr)(const struct netdev *netdev, uint8_t mac[6]);
244 /* Retrieves 'netdev''s MTU into '*mtup'.
246 * The MTU is the maximum size of transmitted (and received) packets, in
247 * bytes, not including the hardware header; thus, this is typically 1500
248 * bytes for Ethernet devices.
250 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
251 * this function should set '*mtup' to INT_MAX. */
252 int (*get_mtu)(const struct netdev *netdev, int *mtup);
254 /* Returns the ifindex of 'netdev', if successful, as a positive number.
255 * On failure, returns a negative errno value.
257 * The desired semantics of the ifindex value are a combination of those
258 * specified by POSIX for if_nametoindex() and by SNMP for ifIndex. An
259 * ifindex value should be unique within a host and remain stable at least
260 * until reboot. SNMP says an ifindex "ranges between 1 and the value of
261 * ifNumber" but many systems do not follow this rule anyhow.
263 * This function may be set to null if it would always return -EOPNOTSUPP.
265 int (*get_ifindex)(const struct netdev *netdev);
267 /* Sets 'carrier' to true if carrier is active (link light is on) on
270 * May be null if device does not provide carrier status (will be always
271 * up as long as device is up).
273 int (*get_carrier)(const struct netdev *netdev, bool *carrier);
275 /* Forces ->get_carrier() to poll 'netdev''s MII registers for link status
276 * instead of checking 'netdev''s carrier. 'netdev''s MII registers will
277 * be polled once ever 'interval' milliseconds. If 'netdev' does not
278 * support MII, another method may be used as a fallback. If 'interval' is
279 * less than or equal to zero, reverts ->get_carrier() to its normal
282 * Most network devices won't support this feature and will set this
283 * function pointer to NULL, which is equivalent to returning EOPNOTSUPP.
285 int (*set_miimon_interval)(struct netdev *netdev, long long int interval);
287 /* Retrieves current device stats for 'netdev' into 'stats'.
289 * A network device that supports some statistics but not others, it should
290 * set the values of the unsupported statistics to all-1-bits
292 int (*get_stats)(const struct netdev *netdev, struct netdev_stats *);
294 /* Sets the device stats for 'netdev' to 'stats'.
296 * Most network devices won't support this feature and will set this
297 * function pointer to NULL, which is equivalent to returning EOPNOTSUPP.
299 * Some network devices might only allow setting their stats to 0. */
300 int (*set_stats)(struct netdev *netdev, const struct netdev_stats *);
302 /* Stores the features supported by 'netdev' into each of '*current',
303 * '*advertised', '*supported', and '*peer'. Each value is a bitmap of
304 * "enum ofp_port_features" bits, in host byte order.
306 * This function may be set to null if it would always return EOPNOTSUPP.
308 int (*get_features)(const struct netdev *netdev,
309 uint32_t *current, uint32_t *advertised,
310 uint32_t *supported, uint32_t *peer);
312 /* Set the features advertised by 'netdev' to 'advertise', which is a
313 * bitmap of "enum ofp_port_features" bits, in host byte order.
315 * This function may be set to null for a network device that does not
316 * support configuring advertisements. */
317 int (*set_advertisements)(struct netdev *netdev, uint32_t advertise);
319 /* If 'netdev' is a VLAN network device (e.g. one created with vconfig(8)),
320 * sets '*vlan_vid' to the VLAN VID associated with that device and returns
323 * Returns ENOENT if 'netdev' is a network device that is not a
326 * This function should be set to null if it doesn't make any sense for
327 * your network device (it probably doesn't). */
328 int (*get_vlan_vid)(const struct netdev *netdev, int *vlan_vid);
330 /* Attempts to set input rate limiting (policing) policy, such that up to
331 * 'kbits_rate' kbps of traffic is accepted, with a maximum accumulative
332 * burst size of 'kbits' kb.
334 * This function may be set to null if policing is not supported. */
335 int (*set_policing)(struct netdev *netdev, unsigned int kbits_rate,
336 unsigned int kbits_burst);
338 /* Adds to 'types' all of the forms of QoS supported by 'netdev', or leaves
339 * it empty if 'netdev' does not support QoS. Any names added to 'types'
340 * should be documented as valid for the "type" column in the "QoS" table
341 * in vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
343 * Every network device must support disabling QoS with a type of "", but
344 * this function must not add "" to 'types'.
346 * The caller is responsible for initializing 'types' (e.g. with
347 * sset_init()) before calling this function. The caller retains ownership
350 * May be NULL if 'netdev' does not support QoS at all. */
351 int (*get_qos_types)(const struct netdev *netdev, struct sset *types);
353 /* Queries 'netdev' for its capabilities regarding the specified 'type' of
354 * QoS. On success, initializes 'caps' with the QoS capabilities.
356 * Should return EOPNOTSUPP if 'netdev' does not support 'type'. May be
357 * NULL if 'netdev' does not support QoS at all. */
358 int (*get_qos_capabilities)(const struct netdev *netdev,
360 struct netdev_qos_capabilities *caps);
362 /* Queries 'netdev' about its currently configured form of QoS. If
363 * successful, stores the name of the current form of QoS into '*typep'
364 * and any details of configuration as string key-value pairs in
367 * A '*typep' of "" indicates that QoS is currently disabled on 'netdev'.
369 * The caller initializes 'details' before calling this function. The
370 * caller takes ownership of the string key-values pairs added to
373 * The netdev retains ownership of '*typep'.
375 * '*typep' will be one of the types returned by netdev_get_qos_types() for
376 * 'netdev'. The contents of 'details' should be documented as valid for
377 * '*typep' in the "other_config" column in the "QoS" table in
378 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
380 * May be NULL if 'netdev' does not support QoS at all. */
381 int (*get_qos)(const struct netdev *netdev,
382 const char **typep, struct shash *details);
384 /* Attempts to reconfigure QoS on 'netdev', changing the form of QoS to
385 * 'type' with details of configuration from 'details'.
387 * On error, the previous QoS configuration is retained.
389 * When this function changes the type of QoS (not just 'details'), this
390 * also resets all queue configuration for 'netdev' to their defaults
391 * (which depend on the specific type of QoS). Otherwise, the queue
392 * configuration for 'netdev' is unchanged.
394 * 'type' should be "" (to disable QoS) or one of the types returned by
395 * netdev_get_qos_types() for 'netdev'. The contents of 'details' should
396 * be documented as valid for the given 'type' in the "other_config" column
397 * in the "QoS" table in vswitchd/vswitch.xml (which is built as
398 * ovs-vswitchd.conf.db(8)).
400 * May be NULL if 'netdev' does not support QoS at all. */
401 int (*set_qos)(struct netdev *netdev,
402 const char *type, const struct shash *details);
404 /* Queries 'netdev' for information about the queue numbered 'queue_id'.
405 * If successful, adds that information as string key-value pairs to
406 * 'details'. Returns 0 if successful, otherwise a positive errno value.
408 * Should return EINVAL if 'queue_id' is greater than or equal to the
409 * number of supported queues (as reported in the 'n_queues' member of
410 * struct netdev_qos_capabilities by 'get_qos_capabilities').
412 * The caller initializes 'details' before calling this function. The
413 * caller takes ownership of the string key-values pairs added to
416 * The returned contents of 'details' should be documented as valid for the
417 * given 'type' in the "other_config" column in the "Queue" table in
418 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
420 int (*get_queue)(const struct netdev *netdev,
421 unsigned int queue_id, struct shash *details);
423 /* Configures the queue numbered 'queue_id' on 'netdev' with the key-value
424 * string pairs in 'details'. The contents of 'details' should be
425 * documented as valid for the given 'type' in the "other_config" column in
426 * the "Queue" table in vswitchd/vswitch.xml (which is built as
427 * ovs-vswitchd.conf.db(8)). Returns 0 if successful, otherwise a positive
428 * errno value. On failure, the given queue's configuration should be
431 * Should return EINVAL if 'queue_id' is greater than or equal to the
432 * number of supported queues (as reported in the 'n_queues' member of
433 * struct netdev_qos_capabilities by 'get_qos_capabilities'), or if
434 * 'details' is invalid for the type of queue.
436 * This function does not modify 'details', and the caller retains
439 * May be NULL if 'netdev' does not support QoS at all. */
440 int (*set_queue)(struct netdev *netdev,
441 unsigned int queue_id, const struct shash *details);
443 /* Attempts to delete the queue numbered 'queue_id' from 'netdev'.
445 * Should return EINVAL if 'queue_id' is greater than or equal to the
446 * number of supported queues (as reported in the 'n_queues' member of
447 * struct netdev_qos_capabilities by 'get_qos_capabilities'). Should
448 * return EOPNOTSUPP if 'queue_id' is valid but may not be deleted (e.g. if
449 * 'netdev' has a fixed set of queues with the current QoS mode).
451 * May be NULL if 'netdev' does not support QoS at all, or if all of its
452 * QoS modes have fixed sets of queues. */
453 int (*delete_queue)(struct netdev *netdev, unsigned int queue_id);
455 /* Obtains statistics about 'queue_id' on 'netdev'. Fills 'stats' with the
456 * queue's statistics. May set individual members of 'stats' to all-1-bits
457 * if the statistic is unavailable.
459 * May be NULL if 'netdev' does not support QoS at all. */
460 int (*get_queue_stats)(const struct netdev *netdev, unsigned int queue_id,
461 struct netdev_queue_stats *stats);
463 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
464 * ID, its configuration, and the 'aux' specified by the caller. The order
465 * of iteration is unspecified, but (when successful) each queue is visited
468 * 'cb' will not modify or free the 'details' argument passed in. */
469 int (*dump_queues)(const struct netdev *netdev,
470 void (*cb)(unsigned int queue_id,
471 const struct shash *details,
475 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
476 * ID, its statistics, and the 'aux' specified by the caller. The order of
477 * iteration is unspecified, but (when successful) each queue must be
478 * visited exactly once.
480 * 'cb' will not modify or free the statistics passed in. */
481 int (*dump_queue_stats)(const struct netdev *netdev,
482 void (*cb)(unsigned int queue_id,
483 struct netdev_queue_stats *,
487 /* If 'netdev' has an assigned IPv4 address, sets '*address' to that
488 * address and '*netmask' to the associated netmask.
490 * The following error values have well-defined meanings:
492 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv4 address.
494 * - EOPNOTSUPP: No IPv4 network stack attached to 'netdev'.
496 * This function may be set to null if it would always return EOPNOTSUPP
498 int (*get_in4)(const struct netdev *netdev, struct in_addr *address,
499 struct in_addr *netmask);
501 /* Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
502 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared.
504 * This function may be set to null if it would always return EOPNOTSUPP
506 int (*set_in4)(struct netdev *netdev, struct in_addr addr,
507 struct in_addr mask);
509 /* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address.
511 * The following error values have well-defined meanings:
513 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv6 address.
515 * - EOPNOTSUPP: No IPv6 network stack attached to 'netdev'.
517 * This function may be set to null if it would always return EOPNOTSUPP
519 int (*get_in6)(const struct netdev *netdev, struct in6_addr *in6);
521 /* Adds 'router' as a default IP gateway for the TCP/IP stack that
522 * corresponds to 'netdev'.
524 * This function may be set to null if it would always return EOPNOTSUPP
526 int (*add_router)(struct netdev *netdev, struct in_addr router);
528 /* Looks up the next hop for 'host'. If succesful, stores the next hop
529 * gateway's address (0 if 'host' is on a directly connected network) in
530 * '*next_hop' and a copy of the name of the device to reach 'host' in
531 * '*netdev_name', and returns 0. The caller is responsible for freeing
532 * '*netdev_name' (by calling free()).
534 * This function may be set to null if it would always return EOPNOTSUPP
536 int (*get_next_hop)(const struct in_addr *host, struct in_addr *next_hop,
539 /* Retrieves the status of the device.
541 * Populates 'sh' with key-value pairs representing the status of the
542 * device. A device's status is a set of key-value string pairs
543 * representing netdev type specific information. For more information see
544 * ovs-vswitchd.conf.db(5).
546 * The data of 'sh' are heap allocated strings which the caller is
547 * responsible for deallocating.
549 * This function may be set to null if it would always return EOPNOTSUPP
551 int (*get_status)(const struct netdev *netdev, struct shash *sh);
553 /* Looks up the ARP table entry for 'ip' on 'netdev' and stores the
554 * corresponding MAC address in 'mac'. A return value of ENXIO, in
555 * particular, indicates that there is no ARP table entry for 'ip' on
558 * This function may be set to null if it would always return EOPNOTSUPP
560 int (*arp_lookup)(const struct netdev *netdev, ovs_be32 ip,
563 /* Retrieves the current set of flags on 'netdev' into '*old_flags'.
564 * Then, turns off the flags that are set to 1 in 'off' and turns on the
565 * flags that are set to 1 in 'on'. (No bit will be set to 1 in both 'off'
566 * and 'on'; that is, off & on == 0.)
568 * This function may be invoked from a signal handler. Therefore, it
569 * should not do anything that is not signal-safe (such as logging). */
570 int (*update_flags)(struct netdev *netdev, enum netdev_flags off,
571 enum netdev_flags on, enum netdev_flags *old_flags);
573 /* Returns a sequence number which indicates changes in one of 'netdev''s
574 * properties. The returned sequence number must be nonzero so that
575 * callers have a value which they may use as a reset when tracking
578 * Minimally, the returned sequence number is required to change whenever
579 * 'netdev''s flags, features, ethernet address, or carrier changes. The
580 * returned sequence number is allowed to change even when 'netdev' doesn't
581 * change, although implementations should try to avoid this. */
582 unsigned int (*change_seq)(const struct netdev *netdev);
585 int netdev_register_provider(const struct netdev_class *);
586 int netdev_unregister_provider(const char *type);
587 const struct netdev_class *netdev_lookup_provider(const char *type);
589 extern const struct netdev_class netdev_linux_class;
590 extern const struct netdev_class netdev_internal_class;
591 extern const struct netdev_class netdev_tap_class;
597 #endif /* netdev.h */