2 * Copyright (c) 2009, 2010, 2011, 2012 Nicira, Inc.
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 * Returns -EMSGSIZE, and discards the packet, if the received packet is
176 * longer than 'size' bytes.
178 * This function can only be expected to return a packet if ->listen() has
179 * been called successfully.
181 * May be null if not needed, such as for a network device that does not
182 * implement packet reception through the 'recv' member function. */
183 int (*recv)(struct netdev *netdev, void *buffer, size_t size);
185 /* Registers with the poll loop to wake up from the next call to
186 * poll_block() when a packet is ready to be received with netdev_recv() on
189 * May be null if not needed, such as for a network device that does not
190 * implement packet reception through the 'recv' member function. */
191 void (*recv_wait)(struct netdev *netdev);
193 /* Discards all packets waiting to be received from 'netdev'.
195 * May be null if not needed, such as for a network device that does not
196 * implement packet reception through the 'recv' member function. */
197 int (*drain)(struct netdev *netdev);
199 /* Sends the 'size'-byte packet in 'buffer' on 'netdev'. Returns 0 if
200 * successful, otherwise a positive errno value. Returns EAGAIN without
201 * blocking if the packet cannot be queued immediately. Returns EMSGSIZE
202 * if a partial packet was transmitted or if the packet is too big or too
203 * small to transmit on the device.
205 * The caller retains ownership of 'buffer' in all cases.
207 * The network device is expected to maintain a packet transmission queue,
208 * so that the caller does not ordinarily have to do additional queuing of
211 * May return EOPNOTSUPP if a network device does not implement packet
212 * transmission through this interface. This function may be set to null
213 * if it would always return EOPNOTSUPP anyhow. (This will prevent the
214 * network device from being usefully used by the netdev-based "userspace
215 * datapath". It will also prevent the OVS implementation of bonding from
216 * working properly over 'netdev'.) */
217 int (*send)(struct netdev *netdev, const void *buffer, size_t size);
219 /* Registers with the poll loop to wake up from the next call to
220 * poll_block() when the packet transmission queue for 'netdev' has
221 * sufficient room to transmit a packet with netdev_send().
223 * The network device is expected to maintain a packet transmission queue,
224 * so that the caller does not ordinarily have to do additional queuing of
225 * packets. Thus, this function is unlikely to ever be useful.
227 * May be null if not needed, such as for a network device that does not
228 * implement packet transmission through the 'send' member function. */
229 void (*send_wait)(struct netdev *netdev);
231 /* Sets 'netdev''s Ethernet address to 'mac' */
232 int (*set_etheraddr)(struct netdev *netdev, const uint8_t mac[6]);
234 /* Retrieves 'netdev''s Ethernet address into 'mac'.
236 * This address will be advertised as 'netdev''s MAC address through the
237 * OpenFlow protocol, among other uses. */
238 int (*get_etheraddr)(const struct netdev *netdev, uint8_t mac[6]);
240 /* Retrieves 'netdev''s MTU into '*mtup'.
242 * The MTU is the maximum size of transmitted (and received) packets, in
243 * bytes, not including the hardware header; thus, this is typically 1500
244 * bytes for Ethernet devices.
246 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
247 * this function should return EOPNOTSUPP. This function may be set to
248 * null if it would always return EOPNOTSUPP. */
249 int (*get_mtu)(const struct netdev *netdev, int *mtup);
251 /* Sets 'netdev''s MTU to 'mtu'.
253 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
254 * this function should return EOPNOTSUPP. This function may be set to
255 * null if it would always return EOPNOTSUPP. */
256 int (*set_mtu)(const struct netdev *netdev, int mtu);
258 /* Returns the ifindex of 'netdev', if successful, as a positive number.
259 * On failure, returns a negative errno value.
261 * The desired semantics of the ifindex value are a combination of those
262 * specified by POSIX for if_nametoindex() and by SNMP for ifIndex. An
263 * ifindex value should be unique within a host and remain stable at least
264 * until reboot. SNMP says an ifindex "ranges between 1 and the value of
265 * ifNumber" but many systems do not follow this rule anyhow.
267 * This function may be set to null if it would always return -EOPNOTSUPP.
269 int (*get_ifindex)(const struct netdev *netdev);
271 /* Sets 'carrier' to true if carrier is active (link light is on) on
274 * May be null if device does not provide carrier status (will be always
275 * up as long as device is up).
277 int (*get_carrier)(const struct netdev *netdev, bool *carrier);
279 /* Returns the number of times 'netdev''s carrier has changed since being
282 * If null, callers will assume the number of carrier resets is zero. */
283 long long int (*get_carrier_resets)(const struct netdev *netdev);
285 /* Forces ->get_carrier() to poll 'netdev''s MII registers for link status
286 * instead of checking 'netdev''s carrier. 'netdev''s MII registers will
287 * be polled once ever 'interval' milliseconds. If 'netdev' does not
288 * support MII, another method may be used as a fallback. If 'interval' is
289 * less than or equal to zero, reverts ->get_carrier() to its normal
292 * Most network devices won't support this feature and will set this
293 * function pointer to NULL, which is equivalent to returning EOPNOTSUPP.
295 int (*set_miimon_interval)(struct netdev *netdev, long long int interval);
297 /* Retrieves current device stats for 'netdev' into 'stats'.
299 * A network device that supports some statistics but not others, it should
300 * set the values of the unsupported statistics to all-1-bits
302 int (*get_stats)(const struct netdev *netdev, struct netdev_stats *);
304 /* Sets the device stats for 'netdev' to 'stats'.
306 * Most network devices won't support this feature and will set this
307 * function pointer to NULL, which is equivalent to returning EOPNOTSUPP.
309 * Some network devices might only allow setting their stats to 0. */
310 int (*set_stats)(struct netdev *netdev, const struct netdev_stats *);
312 /* Stores the features supported by 'netdev' into each of '*current',
313 * '*advertised', '*supported', and '*peer'. Each value is a bitmap of
316 * This function may be set to null if it would always return EOPNOTSUPP.
318 int (*get_features)(const struct netdev *netdev,
319 enum netdev_features *current,
320 enum netdev_features *advertised,
321 enum netdev_features *supported,
322 enum netdev_features *peer);
324 /* Set the features advertised by 'netdev' to 'advertise', which is a
325 * set of NETDEV_F_* bits.
327 * This function may be set to null for a network device that does not
328 * support configuring advertisements. */
329 int (*set_advertisements)(struct netdev *netdev,
330 enum netdev_features advertise);
332 /* Attempts to set input rate limiting (policing) policy, such that up to
333 * 'kbits_rate' kbps of traffic is accepted, with a maximum accumulative
334 * burst size of 'kbits' kb.
336 * This function may be set to null if policing is not supported. */
337 int (*set_policing)(struct netdev *netdev, unsigned int kbits_rate,
338 unsigned int kbits_burst);
340 /* Adds to 'types' all of the forms of QoS supported by 'netdev', or leaves
341 * it empty if 'netdev' does not support QoS. Any names added to 'types'
342 * should be documented as valid for the "type" column in the "QoS" table
343 * in vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
345 * Every network device must support disabling QoS with a type of "", but
346 * this function must not add "" to 'types'.
348 * The caller is responsible for initializing 'types' (e.g. with
349 * sset_init()) before calling this function. The caller retains ownership
352 * May be NULL if 'netdev' does not support QoS at all. */
353 int (*get_qos_types)(const struct netdev *netdev, struct sset *types);
355 /* Queries 'netdev' for its capabilities regarding the specified 'type' of
356 * QoS. On success, initializes 'caps' with the QoS capabilities.
358 * Should return EOPNOTSUPP if 'netdev' does not support 'type'. May be
359 * NULL if 'netdev' does not support QoS at all. */
360 int (*get_qos_capabilities)(const struct netdev *netdev,
362 struct netdev_qos_capabilities *caps);
364 /* Queries 'netdev' about its currently configured form of QoS. If
365 * successful, stores the name of the current form of QoS into '*typep'
366 * and any details of configuration as string key-value pairs in
369 * A '*typep' of "" indicates that QoS is currently disabled on 'netdev'.
371 * The caller initializes 'details' before calling this function. The
372 * caller takes ownership of the string key-values pairs added to
375 * The netdev retains ownership of '*typep'.
377 * '*typep' will be one of the types returned by netdev_get_qos_types() for
378 * 'netdev'. The contents of 'details' should be documented as valid for
379 * '*typep' in the "other_config" column in the "QoS" table in
380 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
382 * May be NULL if 'netdev' does not support QoS at all. */
383 int (*get_qos)(const struct netdev *netdev,
384 const char **typep, struct shash *details);
386 /* Attempts to reconfigure QoS on 'netdev', changing the form of QoS to
387 * 'type' with details of configuration from 'details'.
389 * On error, the previous QoS configuration is retained.
391 * When this function changes the type of QoS (not just 'details'), this
392 * also resets all queue configuration for 'netdev' to their defaults
393 * (which depend on the specific type of QoS). Otherwise, the queue
394 * configuration for 'netdev' is unchanged.
396 * 'type' should be "" (to disable QoS) or one of the types returned by
397 * netdev_get_qos_types() for 'netdev'. The contents of 'details' should
398 * be documented as valid for the given 'type' in the "other_config" column
399 * in the "QoS" table in vswitchd/vswitch.xml (which is built as
400 * ovs-vswitchd.conf.db(8)).
402 * May be NULL if 'netdev' does not support QoS at all. */
403 int (*set_qos)(struct netdev *netdev,
404 const char *type, const struct shash *details);
406 /* Queries 'netdev' for information about the queue numbered 'queue_id'.
407 * If successful, adds that information as string key-value pairs to
408 * 'details'. Returns 0 if successful, otherwise a positive errno value.
410 * Should return EINVAL if 'queue_id' is greater than or equal to the
411 * number of supported queues (as reported in the 'n_queues' member of
412 * struct netdev_qos_capabilities by 'get_qos_capabilities').
414 * The caller initializes 'details' before calling this function. The
415 * caller takes ownership of the string key-values pairs added to
418 * The returned contents of 'details' should be documented as valid for the
419 * given 'type' in the "other_config" column in the "Queue" table in
420 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
422 int (*get_queue)(const struct netdev *netdev,
423 unsigned int queue_id, struct shash *details);
425 /* Configures the queue numbered 'queue_id' on 'netdev' with the key-value
426 * string pairs in 'details'. The contents of 'details' should be
427 * documented as valid for the given 'type' in the "other_config" column in
428 * the "Queue" table in vswitchd/vswitch.xml (which is built as
429 * ovs-vswitchd.conf.db(8)). Returns 0 if successful, otherwise a positive
430 * errno value. On failure, the given queue's configuration should be
433 * Should return EINVAL if 'queue_id' is greater than or equal to the
434 * number of supported queues (as reported in the 'n_queues' member of
435 * struct netdev_qos_capabilities by 'get_qos_capabilities'), or if
436 * 'details' is invalid for the type of queue.
438 * This function does not modify 'details', and the caller retains
441 * May be NULL if 'netdev' does not support QoS at all. */
442 int (*set_queue)(struct netdev *netdev,
443 unsigned int queue_id, const struct shash *details);
445 /* Attempts to delete the queue numbered 'queue_id' from 'netdev'.
447 * Should return EINVAL if 'queue_id' is greater than or equal to the
448 * number of supported queues (as reported in the 'n_queues' member of
449 * struct netdev_qos_capabilities by 'get_qos_capabilities'). Should
450 * return EOPNOTSUPP if 'queue_id' is valid but may not be deleted (e.g. if
451 * 'netdev' has a fixed set of queues with the current QoS mode).
453 * May be NULL if 'netdev' does not support QoS at all, or if all of its
454 * QoS modes have fixed sets of queues. */
455 int (*delete_queue)(struct netdev *netdev, unsigned int queue_id);
457 /* Obtains statistics about 'queue_id' on 'netdev'. Fills 'stats' with the
458 * queue's statistics. May set individual members of 'stats' to all-1-bits
459 * if the statistic is unavailable.
461 * May be NULL if 'netdev' does not support QoS at all. */
462 int (*get_queue_stats)(const struct netdev *netdev, unsigned int queue_id,
463 struct netdev_queue_stats *stats);
465 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
466 * ID, its configuration, and the 'aux' specified by the caller. The order
467 * of iteration is unspecified, but (when successful) each queue is visited
470 * 'cb' will not modify or free the 'details' argument passed in. It may
471 * delete or modify the queue passed in as its 'queue_id' argument. It may
472 * modify but will not delete any other queue within 'netdev'. If 'cb'
473 * adds new queues, then ->dump_queues is allowed to visit some queues
474 * twice or not at all.
476 int (*dump_queues)(const struct netdev *netdev,
477 void (*cb)(unsigned int queue_id,
478 const struct shash *details,
482 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
483 * ID, its statistics, and the 'aux' specified by the caller. The order of
484 * iteration is unspecified, but (when successful) each queue must be
485 * visited exactly once.
487 * 'cb' will not modify or free the statistics passed in. */
488 int (*dump_queue_stats)(const struct netdev *netdev,
489 void (*cb)(unsigned int queue_id,
490 struct netdev_queue_stats *,
494 /* If 'netdev' has an assigned IPv4 address, sets '*address' to that
495 * address and '*netmask' to the associated netmask.
497 * The following error values have well-defined meanings:
499 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv4 address.
501 * - EOPNOTSUPP: No IPv4 network stack attached to 'netdev'.
503 * This function may be set to null if it would always return EOPNOTSUPP
505 int (*get_in4)(const struct netdev *netdev, struct in_addr *address,
506 struct in_addr *netmask);
508 /* Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
509 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared.
511 * This function may be set to null if it would always return EOPNOTSUPP
513 int (*set_in4)(struct netdev *netdev, struct in_addr addr,
514 struct in_addr mask);
516 /* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address.
518 * The following error values have well-defined meanings:
520 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv6 address.
522 * - EOPNOTSUPP: No IPv6 network stack attached to 'netdev'.
524 * This function may be set to null if it would always return EOPNOTSUPP
526 int (*get_in6)(const struct netdev *netdev, struct in6_addr *in6);
528 /* Adds 'router' as a default IP gateway for the TCP/IP stack that
529 * corresponds to 'netdev'.
531 * This function may be set to null if it would always return EOPNOTSUPP
533 int (*add_router)(struct netdev *netdev, struct in_addr router);
535 /* Looks up the next hop for 'host'. If succesful, stores the next hop
536 * gateway's address (0 if 'host' is on a directly connected network) in
537 * '*next_hop' and a copy of the name of the device to reach 'host' in
538 * '*netdev_name', and returns 0. The caller is responsible for freeing
539 * '*netdev_name' (by calling free()).
541 * This function may be set to null if it would always return EOPNOTSUPP
543 int (*get_next_hop)(const struct in_addr *host, struct in_addr *next_hop,
546 /* Retrieves driver information of the device.
548 * Populates 'sh' with key-value pairs representing the status of the
549 * device. Driver info is a set of key-value string pairs
550 * representing netdev type specific information. For more information see
551 * ovs-vswitchd.conf.db(5).
553 * The data of 'sh' are heap allocated strings which the caller is
554 * responsible for deallocating.
556 * This function may be set to null if it would always return EOPNOTSUPP
558 int (*get_drv_info)(const struct netdev *netdev, struct shash *sh);
560 /* Looks up the ARP table entry for 'ip' on 'netdev' and stores the
561 * corresponding MAC address in 'mac'. A return value of ENXIO, in
562 * particular, indicates that there is no ARP table entry for 'ip' on
565 * This function may be set to null if it would always return EOPNOTSUPP
567 int (*arp_lookup)(const struct netdev *netdev, ovs_be32 ip,
570 /* Retrieves the current set of flags on 'netdev' into '*old_flags'.
571 * Then, turns off the flags that are set to 1 in 'off' and turns on the
572 * flags that are set to 1 in 'on'. (No bit will be set to 1 in both 'off'
573 * and 'on'; that is, off & on == 0.)
575 * This function may be invoked from a signal handler. Therefore, it
576 * should not do anything that is not signal-safe (such as logging). */
577 int (*update_flags)(struct netdev *netdev, enum netdev_flags off,
578 enum netdev_flags on, enum netdev_flags *old_flags);
580 /* Returns a sequence number which indicates changes in one of 'netdev''s
581 * properties. The returned sequence number must be nonzero so that
582 * callers have a value which they may use as a reset when tracking
585 * Minimally, the returned sequence number is required to change whenever
586 * 'netdev''s flags, features, ethernet address, or carrier changes. The
587 * returned sequence number is allowed to change even when 'netdev' doesn't
588 * change, although implementations should try to avoid this. */
589 unsigned int (*change_seq)(const struct netdev *netdev);
592 int netdev_register_provider(const struct netdev_class *);
593 int netdev_unregister_provider(const char *type);
594 const struct netdev_class *netdev_lookup_provider(const char *type);
596 extern const struct netdev_class netdev_linux_class;
597 extern const struct netdev_class netdev_internal_class;
598 extern const struct netdev_class netdev_tap_class;
604 #endif /* netdev.h */