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. */
42 struct shash args; /* Argument list from last config. */
45 void netdev_dev_init(struct netdev_dev *, const char *name,
46 const struct shash *args,
47 const struct netdev_class *);
48 void netdev_dev_uninit(struct netdev_dev *, bool destroy);
49 const char *netdev_dev_get_type(const struct netdev_dev *);
50 const struct netdev_class *netdev_dev_get_class(const struct netdev_dev *);
51 const char *netdev_dev_get_name(const struct netdev_dev *);
52 struct netdev_dev *netdev_dev_from_name(const char *name);
53 void netdev_dev_get_devices(const struct netdev_class *,
54 struct shash *device_list);
56 static inline void netdev_dev_assert_class(const struct netdev_dev *netdev_dev,
57 const struct netdev_class *class_)
59 assert(netdev_dev->netdev_class == class_);
62 /* A instance of an open network device.
64 * This structure should be treated as opaque by network device
67 struct netdev_dev *netdev_dev; /* Parent netdev_dev. */
68 struct list node; /* Element in global list. */
70 enum netdev_flags save_flags; /* Initial device flags. */
71 enum netdev_flags changed_flags; /* Flags that we changed. */
74 void netdev_init(struct netdev *, struct netdev_dev *);
75 void netdev_uninit(struct netdev *, bool close);
76 struct netdev_dev *netdev_get_dev(const struct netdev *);
78 static inline void netdev_assert_class(const struct netdev *netdev,
79 const struct netdev_class *netdev_class)
81 netdev_dev_assert_class(netdev_get_dev(netdev), netdev_class);
84 /* A network device notifier.
86 * Network device implementations should use netdev_notifier_init() to
87 * initialize this structure, but they may freely read its members after
89 struct netdev_notifier {
90 struct netdev *netdev;
91 void (*cb)(struct netdev_notifier *);
94 void netdev_notifier_init(struct netdev_notifier *, struct netdev *,
95 void (*cb)(struct netdev_notifier *), void *aux);
97 /* Network device class structure, to be defined by each implementation of a
100 * These functions return 0 if successful or a positive errno value on failure,
101 * except where otherwise noted. */
102 struct netdev_class {
103 /* Type of netdevs in this class, e.g. "system", "tap", "gre", etc.
105 * One of the providers should supply a "system" type, since this is
106 * the type assumed if no type is specified when opening a netdev.
107 * The "system" type corresponds to an existing network device on
111 /* Called when the netdev provider is registered, typically at program
112 * startup. Returning an error from this function will prevent any network
113 * device in this class from being opened.
115 * This function may be set to null if a network device class needs no
116 * initialization at registration time. */
119 /* Performs periodic work needed by netdevs of this class. May be null if
120 * no periodic work is necessary. */
123 /* Arranges for poll_block() to wake up if the "run" member function needs
124 * to be called. May be null if nothing is needed here. */
127 /* Attempts to create a network device named 'name' with initial 'args' in
128 * 'netdev_class'. On success sets 'netdev_devp' to the newly created
130 int (*create)(const struct netdev_class *netdev_class, const char *name,
131 const struct shash *args, struct netdev_dev **netdev_devp);
133 /* Destroys 'netdev_dev'.
135 * Netdev devices maintain a reference count that is incremented on
136 * netdev_open() and decremented on netdev_close(). If 'netdev_dev'
137 * has a non-zero reference count, then this function will not be
139 void (*destroy)(struct netdev_dev *netdev_dev);
141 /* Changes the device 'netdev_dev''s configuration to 'args'.
143 * If this netdev class does not support reconfiguring a netdev
144 * device, this may be a null pointer.
146 int (*set_config)(struct netdev_dev *netdev_dev, const struct shash *args);
148 /* Attempts to open a network device. On success, sets 'netdevp'
149 * to the new network device.
151 * 'ethertype' may be a 16-bit Ethernet protocol value in host byte order
152 * to capture frames of that type received on the device. It may also be
153 * one of the 'enum netdev_pseudo_ethertype' values to receive frames in
154 * one of those categories. */
155 int (*open)(struct netdev_dev *netdev_dev, int ethertype,
156 struct netdev **netdevp);
158 /* Closes 'netdev'. */
159 void (*close)(struct netdev *netdev);
161 /* Enumerates the names of all network devices of this class.
163 * The caller has already initialized 'all_names' and might already have
164 * added some names to it. This function should not disturb any existing
165 * names in 'all_names'.
167 * If this netdev class does not support enumeration, this may be a null
169 int (*enumerate)(struct svec *all_names);
171 /* Attempts to receive a packet from 'netdev' into the 'size' bytes in
172 * 'buffer'. If successful, returns the number of bytes in the received
173 * packet, otherwise a negative errno value. Returns -EAGAIN immediately
174 * if no packet is ready to be received.
176 * May return -EOPNOTSUPP if a network device does not implement packet
177 * reception through this interface. This function may be set to null if
178 * it would always return -EOPNOTSUPP anyhow. (This will prevent the
179 * network device from being usefully used by the netdev-based "userspace
181 int (*recv)(struct netdev *netdev, void *buffer, size_t size);
183 /* Registers with the poll loop to wake up from the next call to
184 * poll_block() when a packet is ready to be received with netdev_recv() on
187 * May be null if not needed, such as for a network device that does not
188 * implement packet reception through the 'recv' member function. */
189 void (*recv_wait)(struct netdev *netdev);
191 /* Discards all packets waiting to be received from 'netdev'.
193 * May be null if not needed, such as for a network device that does not
194 * implement packet reception through the 'recv' member function. */
195 int (*drain)(struct netdev *netdev);
197 /* Sends the 'size'-byte packet in 'buffer' on 'netdev'. Returns 0 if
198 * successful, otherwise a positive errno value. Returns EAGAIN without
199 * blocking if the packet cannot be queued immediately. Returns EMSGSIZE
200 * if a partial packet was transmitted or if the packet is too big or too
201 * small to transmit on the device.
203 * The caller retains ownership of 'buffer' in all cases.
205 * The network device is expected to maintain a packet transmission queue,
206 * so that the caller does not ordinarily have to do additional queuing of
209 * May return EOPNOTSUPP if a network device does not implement packet
210 * transmission through this interface. This function may be set to null
211 * if it would always return EOPNOTSUPP anyhow. (This will prevent the
212 * network device from being usefully used by the netdev-based "userspace
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 set '*mtup' to INT_MAX. */
242 int (*get_mtu)(const struct netdev *netdev, int *mtup);
244 /* Returns the ifindex of 'netdev', if successful, as a positive number.
245 * On failure, returns a negative errno value.
247 * The desired semantics of the ifindex value are a combination of those
248 * specified by POSIX for if_nametoindex() and by SNMP for ifIndex. An
249 * ifindex value should be unique within a host and remain stable at least
250 * until reboot. SNMP says an ifindex "ranges between 1 and the value of
251 * ifNumber" but many systems do not follow this rule anyhow.
253 * This function may be set to null if it would always return -EOPNOTSUPP.
255 int (*get_ifindex)(const struct netdev *netdev);
257 /* Sets 'carrier' to true if carrier is active (link light is on) on
260 * May be null if device does not provide carrier status (will be always
261 * up as long as device is up).
263 int (*get_carrier)(const struct netdev *netdev, bool *carrier);
265 /* Sets 'miimon' to true if 'netdev' is up according to its MII. If
266 * 'netdev' does not support MII, may fall back to another method or return
269 * This function may be set to null if it would always return EOPNOTSUPP.
271 int (*get_miimon)(const struct netdev *netdev, bool *miimon);
273 /* Retrieves current device stats for 'netdev' into 'stats'.
275 * A network device that supports some statistics but not others, it should
276 * set the values of the unsupported statistics to all-1-bits
278 int (*get_stats)(const struct netdev *netdev, struct netdev_stats *);
280 /* Sets the device stats for 'netdev' to 'stats'.
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 * Some network devices might only allow setting their stats to 0. */
286 int (*set_stats)(struct netdev *netdev, const struct netdev_stats *);
288 /* Stores the features supported by 'netdev' into each of '*current',
289 * '*advertised', '*supported', and '*peer'. Each value is a bitmap of
290 * "enum ofp_port_features" bits, in host byte order.
292 * This function may be set to null if it would always return EOPNOTSUPP.
294 int (*get_features)(const struct netdev *netdev,
295 uint32_t *current, uint32_t *advertised,
296 uint32_t *supported, uint32_t *peer);
298 /* Set the features advertised by 'netdev' to 'advertise', which is a
299 * bitmap of "enum ofp_port_features" bits, in host byte order.
301 * This function may be set to null for a network device that does not
302 * support configuring advertisements. */
303 int (*set_advertisements)(struct netdev *netdev, uint32_t advertise);
305 /* If 'netdev' is a VLAN network device (e.g. one created with vconfig(8)),
306 * sets '*vlan_vid' to the VLAN VID associated with that device and returns
309 * Returns ENOENT if 'netdev' is a network device that is not a
312 * This function should be set to null if it doesn't make any sense for
313 * your network device (it probably doesn't). */
314 int (*get_vlan_vid)(const struct netdev *netdev, int *vlan_vid);
316 /* Attempts to set input rate limiting (policing) policy, such that up to
317 * 'kbits_rate' kbps of traffic is accepted, with a maximum accumulative
318 * burst size of 'kbits' kb.
320 * This function may be set to null if policing is not supported. */
321 int (*set_policing)(struct netdev *netdev, unsigned int kbits_rate,
322 unsigned int kbits_burst);
324 /* Adds to 'types' all of the forms of QoS supported by 'netdev', or leaves
325 * it empty if 'netdev' does not support QoS. Any names added to 'types'
326 * should be documented as valid for the "type" column in the "QoS" table
327 * in vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
329 * Every network device must support disabling QoS with a type of "", but
330 * this function must not add "" to 'types'.
332 * The caller is responsible for initializing 'types' (e.g. with
333 * svec_init()) before calling this function. The caller takes ownership
334 * of the strings added to 'types'.
336 * May be NULL if 'netdev' does not support QoS at all. */
337 int (*get_qos_types)(const struct netdev *netdev, struct svec *types);
339 /* Queries 'netdev' for its capabilities regarding the specified 'type' of
340 * QoS. On success, initializes 'caps' with the QoS capabilities.
342 * Should return EOPNOTSUPP if 'netdev' does not support 'type'. May be
343 * NULL if 'netdev' does not support QoS at all. */
344 int (*get_qos_capabilities)(const struct netdev *netdev,
346 struct netdev_qos_capabilities *caps);
348 /* Queries 'netdev' about its currently configured form of QoS. If
349 * successful, stores the name of the current form of QoS into '*typep'
350 * and any details of configuration as string key-value pairs in
353 * A '*typep' of "" indicates that QoS is currently disabled on 'netdev'.
355 * The caller initializes 'details' before calling this function. The
356 * caller takes ownership of the string key-values pairs added to
359 * The netdev retains ownership of '*typep'.
361 * '*typep' will be one of the types returned by netdev_get_qos_types() for
362 * 'netdev'. The contents of 'details' should be documented as valid for
363 * '*typep' in the "other_config" column in the "QoS" table in
364 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
366 * May be NULL if 'netdev' does not support QoS at all. */
367 int (*get_qos)(const struct netdev *netdev,
368 const char **typep, struct shash *details);
370 /* Attempts to reconfigure QoS on 'netdev', changing the form of QoS to
371 * 'type' with details of configuration from 'details'.
373 * On error, the previous QoS configuration is retained.
375 * When this function changes the type of QoS (not just 'details'), this
376 * also resets all queue configuration for 'netdev' to their defaults
377 * (which depend on the specific type of QoS). Otherwise, the queue
378 * configuration for 'netdev' is unchanged.
380 * 'type' should be "" (to disable QoS) or one of the types returned by
381 * netdev_get_qos_types() for 'netdev'. The contents of 'details' should
382 * be documented as valid for the given 'type' in the "other_config" column
383 * in the "QoS" table in vswitchd/vswitch.xml (which is built as
384 * ovs-vswitchd.conf.db(8)).
386 * May be NULL if 'netdev' does not support QoS at all. */
387 int (*set_qos)(struct netdev *netdev,
388 const char *type, const struct shash *details);
390 /* Queries 'netdev' for information about the queue numbered 'queue_id'.
391 * If successful, adds that information as string key-value pairs to
392 * 'details'. Returns 0 if successful, otherwise a positive errno value.
394 * Should return EINVAL if 'queue_id' is greater than or equal to the
395 * number of supported queues (as reported in the 'n_queues' member of
396 * struct netdev_qos_capabilities by 'get_qos_capabilities').
398 * The caller initializes 'details' before calling this function. The
399 * caller takes ownership of the string key-values pairs added to
402 * The returned contents of 'details' should be documented as valid for the
403 * given 'type' in the "other_config" column in the "Queue" table in
404 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
406 int (*get_queue)(const struct netdev *netdev,
407 unsigned int queue_id, struct shash *details);
409 /* Configures the queue numbered 'queue_id' on 'netdev' with the key-value
410 * string pairs in 'details'. The contents of 'details' should be
411 * documented as valid for the given 'type' in the "other_config" column in
412 * the "Queue" table in vswitchd/vswitch.xml (which is built as
413 * ovs-vswitchd.conf.db(8)). Returns 0 if successful, otherwise a positive
414 * errno value. On failure, the given queue's configuration should be
417 * Should return EINVAL if 'queue_id' is greater than or equal to the
418 * number of supported queues (as reported in the 'n_queues' member of
419 * struct netdev_qos_capabilities by 'get_qos_capabilities'), or if
420 * 'details' is invalid for the type of queue.
422 * This function does not modify 'details', and the caller retains
425 * May be NULL if 'netdev' does not support QoS at all. */
426 int (*set_queue)(struct netdev *netdev,
427 unsigned int queue_id, const struct shash *details);
429 /* Attempts to delete the queue numbered 'queue_id' from 'netdev'.
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'). Should
434 * return EOPNOTSUPP if 'queue_id' is valid but may not be deleted (e.g. if
435 * 'netdev' has a fixed set of queues with the current QoS mode).
437 * May be NULL if 'netdev' does not support QoS at all, or if all of its
438 * QoS modes have fixed sets of queues. */
439 int (*delete_queue)(struct netdev *netdev, unsigned int queue_id);
441 /* Obtains statistics about 'queue_id' on 'netdev'. Fills 'stats' with the
442 * queue's statistics. May set individual members of 'stats' to all-1-bits
443 * if the statistic is unavailable.
445 * May be NULL if 'netdev' does not support QoS at all. */
446 int (*get_queue_stats)(const struct netdev *netdev, unsigned int queue_id,
447 struct netdev_queue_stats *stats);
449 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
450 * ID, its configuration, and the 'aux' specified by the caller. The order
451 * of iteration is unspecified, but (when successful) each queue is visited
454 * 'cb' will not modify or free the 'details' argument passed in. */
455 int (*dump_queues)(const struct netdev *netdev,
456 void (*cb)(unsigned int queue_id,
457 const struct shash *details,
461 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
462 * ID, its statistics, and the 'aux' specified by the caller. The order of
463 * iteration is unspecified, but (when successful) each queue must be
464 * visited exactly once.
466 * 'cb' will not modify or free the statistics passed in. */
467 int (*dump_queue_stats)(const struct netdev *netdev,
468 void (*cb)(unsigned int queue_id,
469 struct netdev_queue_stats *,
473 /* If 'netdev' has an assigned IPv4 address, sets '*address' to that
474 * address and '*netmask' to the associated netmask.
476 * The following error values have well-defined meanings:
478 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv4 address.
480 * - EOPNOTSUPP: No IPv4 network stack attached to 'netdev'.
482 * This function may be set to null if it would always return EOPNOTSUPP
484 int (*get_in4)(const struct netdev *netdev, struct in_addr *address,
485 struct in_addr *netmask);
487 /* Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
488 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared.
490 * This function may be set to null if it would always return EOPNOTSUPP
492 int (*set_in4)(struct netdev *netdev, struct in_addr addr,
493 struct in_addr mask);
495 /* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address.
497 * The following error values have well-defined meanings:
499 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv6 address.
501 * - EOPNOTSUPP: No IPv6 network stack attached to 'netdev'.
503 * This function may be set to null if it would always return EOPNOTSUPP
505 int (*get_in6)(const struct netdev *netdev, struct in6_addr *in6);
507 /* Adds 'router' as a default IP gateway for the TCP/IP stack that
508 * corresponds to 'netdev'.
510 * This function may be set to null if it would always return EOPNOTSUPP
512 int (*add_router)(struct netdev *netdev, struct in_addr router);
514 /* Looks up the next hop for 'host'. If succesful, stores the next hop
515 * gateway's address (0 if 'host' is on a directly connected network) in
516 * '*next_hop' and a copy of the name of the device to reach 'host' in
517 * '*netdev_name', and returns 0. The caller is responsible for freeing
518 * '*netdev_name' (by calling free()).
520 * This function may be set to null if it would always return EOPNOTSUPP
522 int (*get_next_hop)(const struct in_addr *host, struct in_addr *next_hop,
525 /* Retrieves the status of the device.
527 * Populates 'sh' with key-value pairs representing the status of the
528 * device. A device's status is a set of key-value string pairs
529 * representing netdev type specific information. For more information see
530 * ovs-vswitchd.conf.db(5).
532 * The data of 'sh' are heap allocated strings which the caller is
533 * responsible for deallocating.
535 * This function may be set to null if it would always return EOPNOTSUPP
537 int (*get_status)(const struct netdev *netdev, struct shash *sh);
539 /* Looks up the ARP table entry for 'ip' on 'netdev' and stores the
540 * corresponding MAC address in 'mac'. A return value of ENXIO, in
541 * particular, indicates that there is no ARP table entry for 'ip' on
544 * This function may be set to null if it would always return EOPNOTSUPP
546 int (*arp_lookup)(const struct netdev *netdev, uint32_t ip, uint8_t mac[6]);
548 /* Retrieves the current set of flags on 'netdev' into '*old_flags'.
549 * Then, turns off the flags that are set to 1 in 'off' and turns on the
550 * flags that are set to 1 in 'on'. (No bit will be set to 1 in both 'off'
551 * and 'on'; that is, off & on == 0.)
553 * This function may be invoked from a signal handler. Therefore, it
554 * should not do anything that is not signal-safe (such as logging). */
555 int (*update_flags)(struct netdev *netdev, enum netdev_flags off,
556 enum netdev_flags on, enum netdev_flags *old_flags);
558 /* Arranges for 'cb' to be called whenever one of the attributes of
559 * 'netdev' changes and sets '*notifierp' to a newly created
560 * netdev_notifier that represents this arrangement. The created notifier
561 * will have its 'netdev', 'cb', and 'aux' members set to the values of the
562 * corresponding parameters. */
563 int (*poll_add)(struct netdev *netdev,
564 void (*cb)(struct netdev_notifier *notifier), void *aux,
565 struct netdev_notifier **notifierp);
567 /* Cancels poll notification for 'notifier'. */
568 void (*poll_remove)(struct netdev_notifier *notifier);
571 int netdev_register_provider(const struct netdev_class *);
572 int netdev_unregister_provider(const char *type);
573 const struct netdev_class *netdev_lookup_provider(const char *type);
575 extern const struct netdev_class netdev_linux_class;
576 extern const struct netdev_class netdev_internal_class;
577 extern const struct netdev_class netdev_tap_class;
583 #endif /* netdev.h */