2 * Copyright (c) 2009, 2010, 2011, 2012, 2013 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. */
31 /* A network device (e.g. an Ethernet device).
33 * This structure should be treated as opaque by network device
36 char *name; /* Name of network device. */
37 const struct netdev_class *netdev_class; /* Functions to control
39 int ref_cnt; /* Times this devices was opened. */
40 struct shash_node *node; /* Pointer to element in global map. */
41 struct list saved_flags_list; /* Contains "struct netdev_saved_flags". */
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 ovs_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. */
69 void netdev_init(struct netdev *, struct netdev_dev *);
70 void netdev_uninit(struct netdev *, bool close);
71 struct netdev_dev *netdev_get_dev(const struct netdev *);
73 static inline void netdev_assert_class(const struct netdev *netdev,
74 const struct netdev_class *netdev_class)
76 netdev_dev_assert_class(netdev_get_dev(netdev), netdev_class);
79 /* Network device class structure, to be defined by each implementation of a
82 * These functions return 0 if successful or a positive errno value on failure,
83 * except where otherwise noted. */
85 /* Type of netdevs in this class, e.g. "system", "tap", "gre", etc.
87 * One of the providers should supply a "system" type, since this is
88 * the type assumed if no type is specified when opening a netdev.
89 * The "system" type corresponds to an existing network device on
93 /* Called when the netdev provider is registered, typically at program
94 * startup. Returning an error from this function will prevent any network
95 * device in this class from being opened.
97 * This function may be set to null if a network device class needs no
98 * initialization at registration time. */
101 /* Performs periodic work needed by netdevs of this class. May be null if
102 * no periodic work is necessary. */
105 /* Arranges for poll_block() to wake up if the "run" member function needs
106 * to be called. Implementations are additionally required to wake
107 * whenever something changes in any of its netdevs which would cause their
108 * ->change_seq() function to change its result. May be null if nothing is
112 /* Attempts to create a network device named 'name' in 'netdev_class'. On
113 * success sets 'netdev_devp' to the newly created device. */
114 int (*create)(const struct netdev_class *netdev_class, const char *name,
115 struct netdev_dev **netdev_devp);
117 /* Destroys 'netdev_dev'.
119 * Netdev devices maintain a reference count that is incremented on
120 * netdev_open() and decremented on netdev_close(). If 'netdev_dev'
121 * has a non-zero reference count, then this function will not be
123 void (*destroy)(struct netdev_dev *netdev_dev);
125 /* Fetches the device 'netdev_dev''s configuration, storing it in 'args'.
126 * The caller owns 'args' and pre-initializes it to an empty smap.
128 * If this netdev class does not have any configuration options, this may
129 * be a null pointer. */
130 int (*get_config)(struct netdev_dev *netdev_dev, struct smap *args);
132 /* Changes the device 'netdev_dev''s configuration to 'args'.
134 * If this netdev class does not support configuration, this may be a null
136 int (*set_config)(struct netdev_dev *netdev_dev, const struct smap *args);
138 /* Returns the tunnel configuration of 'netdev_dev'. If 'netdev_dev' is
139 * not a tunnel, returns null.
141 * If this function would always return null, it may be null instead. */
142 const struct netdev_tunnel_config *
143 (*get_tunnel_config)(const struct netdev_dev *netdev_dev);
145 /* Attempts to open a network device. On success, sets 'netdevp'
146 * to the new network device. */
147 int (*open)(struct netdev_dev *netdev_dev, struct netdev **netdevp);
149 /* Closes 'netdev'. */
150 void (*close)(struct netdev *netdev);
152 /* ## ----------------- ## */
153 /* ## Receiving Packets ## */
154 /* ## ----------------- ## */
156 /* The network provider interface is mostly used for inspecting and configuring
157 * device "metadata", not for sending and receiving packets directly. It may
158 * be impractical to implement these functions on some operating systems and
159 * hardware. These functions may all be NULL in such cases.
161 * (However, the "dpif-netdev" implementation, which is the easiest way to
162 * integrate Open vSwitch with a new operating system or hardware, does require
163 * the ability to receive packets.) */
165 /* Attempts to set up 'netdev' for receiving packets with ->recv().
166 * Returns 0 if successful, otherwise a positive errno value. Return
167 * EOPNOTSUPP to indicate that the network device does not implement packet
168 * reception through this interface. This function may be set to null if
169 * it would always return EOPNOTSUPP anyhow. (This will prevent the
170 * network device from being usefully used by the netdev-based "userspace
172 int (*listen)(struct netdev *netdev);
174 /* Attempts to receive a packet from 'netdev' into the 'size' bytes in
175 * 'buffer'. If successful, returns the number of bytes in the received
176 * packet, otherwise a negative errno value. Returns -EAGAIN immediately
177 * if no packet is ready to be received.
179 * Returns -EMSGSIZE, and discards the packet, if the received packet is
180 * longer than 'size' bytes.
182 * This function can only be expected to return a packet if ->listen() has
183 * been called successfully.
185 * May be null if not needed, such as for a network device that does not
186 * implement packet reception through the 'recv' member function. */
187 int (*recv)(struct netdev *netdev, void *buffer, size_t size);
189 /* Registers with the poll loop to wake up from the next call to
190 * poll_block() when a packet is ready to be received with netdev_recv() on
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 void (*recv_wait)(struct netdev *netdev);
197 /* Discards all packets waiting to be received from 'netdev'.
199 * May be null if not needed, such as for a network device that does not
200 * implement packet reception through the 'recv' member function. */
201 int (*drain)(struct netdev *netdev);
203 /* Sends the 'size'-byte packet in 'buffer' on 'netdev'. Returns 0 if
204 * successful, otherwise a positive errno value. Returns EAGAIN without
205 * blocking if the packet cannot be queued immediately. Returns EMSGSIZE
206 * if a partial packet was transmitted or if the packet is too big or too
207 * small to transmit on the device.
209 * The caller retains ownership of 'buffer' in all cases.
211 * The network device is expected to maintain a packet transmission queue,
212 * so that the caller does not ordinarily have to do additional queuing of
215 * May return EOPNOTSUPP if a network device does not implement packet
216 * transmission through this interface. This function may be set to null
217 * if it would always return EOPNOTSUPP anyhow. (This will prevent the
218 * network device from being usefully used by the netdev-based "userspace
219 * datapath". It will also prevent the OVS implementation of bonding from
220 * working properly over 'netdev'.) */
221 int (*send)(struct netdev *netdev, const void *buffer, size_t size);
223 /* Registers with the poll loop to wake up from the next call to
224 * poll_block() when the packet transmission queue for 'netdev' has
225 * sufficient room to transmit a packet with netdev_send().
227 * The network device is expected to maintain a packet transmission queue,
228 * so that the caller does not ordinarily have to do additional queuing of
229 * packets. Thus, this function is unlikely to ever be useful.
231 * May be null if not needed, such as for a network device that does not
232 * implement packet transmission through the 'send' member function. */
233 void (*send_wait)(struct netdev *netdev);
235 /* Sets 'netdev''s Ethernet address to 'mac' */
236 int (*set_etheraddr)(struct netdev *netdev, const uint8_t mac[6]);
238 /* Retrieves 'netdev''s Ethernet address into 'mac'.
240 * This address will be advertised as 'netdev''s MAC address through the
241 * OpenFlow protocol, among other uses. */
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 return EOPNOTSUPP. This function may be set to
252 * null if it would always return EOPNOTSUPP. */
253 int (*get_mtu)(const struct netdev *netdev, int *mtup);
255 /* Sets 'netdev''s MTU to 'mtu'.
257 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
258 * this function should return EOPNOTSUPP. This function may be set to
259 * null if it would always return EOPNOTSUPP. */
260 int (*set_mtu)(const struct netdev *netdev, int mtu);
262 /* Returns the ifindex of 'netdev', if successful, as a positive number.
263 * On failure, returns a negative errno value.
265 * The desired semantics of the ifindex value are a combination of those
266 * specified by POSIX for if_nametoindex() and by SNMP for ifIndex. An
267 * ifindex value should be unique within a host and remain stable at least
268 * until reboot. SNMP says an ifindex "ranges between 1 and the value of
269 * ifNumber" but many systems do not follow this rule anyhow.
271 * This function may be set to null if it would always return -EOPNOTSUPP.
273 int (*get_ifindex)(const struct netdev *netdev);
275 /* Sets 'carrier' to true if carrier is active (link light is on) on
278 * May be null if device does not provide carrier status (will be always
279 * up as long as device is up).
281 int (*get_carrier)(const struct netdev *netdev, bool *carrier);
283 /* Returns the number of times 'netdev''s carrier has changed since being
286 * If null, callers will assume the number of carrier resets is zero. */
287 long long int (*get_carrier_resets)(const struct netdev *netdev);
289 /* Forces ->get_carrier() to poll 'netdev''s MII registers for link status
290 * instead of checking 'netdev''s carrier. 'netdev''s MII registers will
291 * be polled once ever 'interval' milliseconds. If 'netdev' does not
292 * support MII, another method may be used as a fallback. If 'interval' is
293 * less than or equal to zero, reverts ->get_carrier() to its normal
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 int (*set_miimon_interval)(struct netdev *netdev, long long int interval);
301 /* Retrieves current device stats for 'netdev' into 'stats'.
303 * A network device that supports some statistics but not others, it should
304 * set the values of the unsupported statistics to all-1-bits
306 int (*get_stats)(const struct netdev *netdev, struct netdev_stats *);
308 /* Sets the device stats for 'netdev' to 'stats'.
310 * Most network devices won't support this feature and will set this
311 * function pointer to NULL, which is equivalent to returning EOPNOTSUPP.
313 * Some network devices might only allow setting their stats to 0. */
314 int (*set_stats)(struct netdev *netdev, const struct netdev_stats *);
316 /* Stores the features supported by 'netdev' into each of '*current',
317 * '*advertised', '*supported', and '*peer'. Each value is a bitmap of
320 * This function may be set to null if it would always return EOPNOTSUPP.
322 int (*get_features)(const struct netdev *netdev,
323 enum netdev_features *current,
324 enum netdev_features *advertised,
325 enum netdev_features *supported,
326 enum netdev_features *peer);
328 /* Set the features advertised by 'netdev' to 'advertise', which is a
329 * set of NETDEV_F_* bits.
331 * This function may be set to null for a network device that does not
332 * support configuring advertisements. */
333 int (*set_advertisements)(struct netdev *netdev,
334 enum netdev_features advertise);
336 /* Attempts to set input rate limiting (policing) policy, such that up to
337 * 'kbits_rate' kbps of traffic is accepted, with a maximum accumulative
338 * burst size of 'kbits' kb.
340 * This function may be set to null if policing is not supported. */
341 int (*set_policing)(struct netdev *netdev, unsigned int kbits_rate,
342 unsigned int kbits_burst);
344 /* Adds to 'types' all of the forms of QoS supported by 'netdev', or leaves
345 * it empty if 'netdev' does not support QoS. Any names added to 'types'
346 * should be documented as valid for the "type" column in the "QoS" table
347 * in vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
349 * Every network device must support disabling QoS with a type of "", but
350 * this function must not add "" to 'types'.
352 * The caller is responsible for initializing 'types' (e.g. with
353 * sset_init()) before calling this function. The caller retains ownership
356 * May be NULL if 'netdev' does not support QoS at all. */
357 int (*get_qos_types)(const struct netdev *netdev, struct sset *types);
359 /* Queries 'netdev' for its capabilities regarding the specified 'type' of
360 * QoS. On success, initializes 'caps' with the QoS capabilities.
362 * Should return EOPNOTSUPP if 'netdev' does not support 'type'. May be
363 * NULL if 'netdev' does not support QoS at all. */
364 int (*get_qos_capabilities)(const struct netdev *netdev,
366 struct netdev_qos_capabilities *caps);
368 /* Queries 'netdev' about its currently configured form of QoS. If
369 * successful, stores the name of the current form of QoS into '*typep'
370 * and any details of configuration as string key-value pairs in
373 * A '*typep' of "" indicates that QoS is currently disabled on 'netdev'.
375 * The caller initializes 'details' before calling this function. The
376 * caller takes ownership of the string key-values pairs added to
379 * The netdev retains ownership of '*typep'.
381 * '*typep' will be one of the types returned by netdev_get_qos_types() for
382 * 'netdev'. The contents of 'details' should be documented as valid for
383 * '*typep' in the "other_config" column in the "QoS" table in
384 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
386 * May be NULL if 'netdev' does not support QoS at all. */
387 int (*get_qos)(const struct netdev *netdev,
388 const char **typep, struct smap *details);
390 /* Attempts to reconfigure QoS on 'netdev', changing the form of QoS to
391 * 'type' with details of configuration from 'details'.
393 * On error, the previous QoS configuration is retained.
395 * When this function changes the type of QoS (not just 'details'), this
396 * also resets all queue configuration for 'netdev' to their defaults
397 * (which depend on the specific type of QoS). Otherwise, the queue
398 * configuration for 'netdev' is unchanged.
400 * 'type' should be "" (to disable QoS) or one of the types returned by
401 * netdev_get_qos_types() for 'netdev'. The contents of 'details' should
402 * be documented as valid for the given 'type' in the "other_config" column
403 * in the "QoS" table in vswitchd/vswitch.xml (which is built as
404 * ovs-vswitchd.conf.db(8)).
406 * May be NULL if 'netdev' does not support QoS at all. */
407 int (*set_qos)(struct netdev *netdev,
408 const char *type, const struct smap *details);
410 /* Queries 'netdev' for information about the queue numbered 'queue_id'.
411 * If successful, adds that information as string key-value pairs to
412 * 'details'. Returns 0 if successful, otherwise a positive errno value.
414 * Should return EINVAL if 'queue_id' is greater than or equal to the
415 * number of supported queues (as reported in the 'n_queues' member of
416 * struct netdev_qos_capabilities by 'get_qos_capabilities').
418 * The caller initializes 'details' before calling this function. The
419 * caller takes ownership of the string key-values pairs added to
422 * The returned contents of 'details' should be documented as valid for the
423 * given 'type' in the "other_config" column in the "Queue" table in
424 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
426 int (*get_queue)(const struct netdev *netdev,
427 unsigned int queue_id, struct smap *details);
429 /* Configures the queue numbered 'queue_id' on 'netdev' with the key-value
430 * string pairs in 'details'. The contents of 'details' should be
431 * documented as valid for the given 'type' in the "other_config" column in
432 * the "Queue" table in vswitchd/vswitch.xml (which is built as
433 * ovs-vswitchd.conf.db(8)). Returns 0 if successful, otherwise a positive
434 * errno value. On failure, the given queue's configuration should be
437 * Should return EINVAL if 'queue_id' is greater than or equal to the
438 * number of supported queues (as reported in the 'n_queues' member of
439 * struct netdev_qos_capabilities by 'get_qos_capabilities'), or if
440 * 'details' is invalid for the type of queue.
442 * This function does not modify 'details', and the caller retains
445 * May be NULL if 'netdev' does not support QoS at all. */
446 int (*set_queue)(struct netdev *netdev,
447 unsigned int queue_id, const struct smap *details);
449 /* Attempts to delete the queue numbered 'queue_id' from 'netdev'.
451 * Should return EINVAL if 'queue_id' is greater than or equal to the
452 * number of supported queues (as reported in the 'n_queues' member of
453 * struct netdev_qos_capabilities by 'get_qos_capabilities'). Should
454 * return EOPNOTSUPP if 'queue_id' is valid but may not be deleted (e.g. if
455 * 'netdev' has a fixed set of queues with the current QoS mode).
457 * May be NULL if 'netdev' does not support QoS at all, or if all of its
458 * QoS modes have fixed sets of queues. */
459 int (*delete_queue)(struct netdev *netdev, unsigned int queue_id);
461 /* Obtains statistics about 'queue_id' on 'netdev'. Fills 'stats' with the
462 * queue's statistics. May set individual members of 'stats' to all-1-bits
463 * if the statistic is unavailable.
465 * May be NULL if 'netdev' does not support QoS at all. */
466 int (*get_queue_stats)(const struct netdev *netdev, unsigned int queue_id,
467 struct netdev_queue_stats *stats);
469 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
470 * ID, its configuration, and the 'aux' specified by the caller. The order
471 * of iteration is unspecified, but (when successful) each queue is visited
474 * 'cb' will not modify or free the 'details' argument passed in. It may
475 * delete or modify the queue passed in as its 'queue_id' argument. It may
476 * modify but will not delete any other queue within 'netdev'. If 'cb'
477 * adds new queues, then ->dump_queues is allowed to visit some queues
478 * twice or not at all.
480 int (*dump_queues)(const struct netdev *netdev,
481 void (*cb)(unsigned int queue_id,
482 const struct smap *details,
486 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
487 * ID, its statistics, and the 'aux' specified by the caller. The order of
488 * iteration is unspecified, but (when successful) each queue must be
489 * visited exactly once.
491 * 'cb' will not modify or free the statistics passed in. */
492 int (*dump_queue_stats)(const struct netdev *netdev,
493 void (*cb)(unsigned int queue_id,
494 struct netdev_queue_stats *,
498 /* If 'netdev' has an assigned IPv4 address, sets '*address' to that
499 * address and '*netmask' to the associated netmask.
501 * The following error values have well-defined meanings:
503 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv4 address.
505 * - EOPNOTSUPP: No IPv4 network stack attached to 'netdev'.
507 * This function may be set to null if it would always return EOPNOTSUPP
509 int (*get_in4)(const struct netdev *netdev, struct in_addr *address,
510 struct in_addr *netmask);
512 /* Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
513 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared.
515 * This function may be set to null if it would always return EOPNOTSUPP
517 int (*set_in4)(struct netdev *netdev, struct in_addr addr,
518 struct in_addr mask);
520 /* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address.
522 * The following error values have well-defined meanings:
524 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv6 address.
526 * - EOPNOTSUPP: No IPv6 network stack attached to 'netdev'.
528 * This function may be set to null if it would always return EOPNOTSUPP
530 int (*get_in6)(const struct netdev *netdev, struct in6_addr *in6);
532 /* Adds 'router' as a default IP gateway for the TCP/IP stack that
533 * corresponds to 'netdev'.
535 * This function may be set to null if it would always return EOPNOTSUPP
537 int (*add_router)(struct netdev *netdev, struct in_addr router);
539 /* Looks up the next hop for 'host'. If succesful, stores the next hop
540 * gateway's address (0 if 'host' is on a directly connected network) in
541 * '*next_hop' and a copy of the name of the device to reach 'host' in
542 * '*netdev_name', and returns 0. The caller is responsible for freeing
543 * '*netdev_name' (by calling free()).
545 * This function may be set to null if it would always return EOPNOTSUPP
547 int (*get_next_hop)(const struct in_addr *host, struct in_addr *next_hop,
550 /* Retrieves driver information of the device.
552 * Populates 'smap' with key-value pairs representing the status of the
553 * device. 'smap' is a set of key-value string pairs representing netdev
554 * type specific information. For more information see
555 * ovs-vswitchd.conf.db(5).
557 * The caller is responsible for destroying 'smap' and its data.
559 * This function may be set to null if it would always return EOPNOTSUPP
561 int (*get_status)(const struct netdev *netdev, struct smap *smap);
563 /* Looks up the ARP table entry for 'ip' on 'netdev' and stores the
564 * corresponding MAC address in 'mac'. A return value of ENXIO, in
565 * particular, indicates that there is no ARP table entry for 'ip' on
568 * This function may be set to null if it would always return EOPNOTSUPP
570 int (*arp_lookup)(const struct netdev *netdev, ovs_be32 ip,
573 /* Retrieves the current set of flags on 'dev' into '*old_flags'. Then,
574 * turns off the flags that are set to 1 in 'off' and turns on the flags
575 * that are set to 1 in 'on'. (No bit will be set to 1 in both 'off' and
576 * 'on'; that is, off & on == 0.)
578 * This function may be invoked from a signal handler. Therefore, it
579 * should not do anything that is not signal-safe (such as logging). */
580 int (*update_flags)(struct netdev_dev *dev, enum netdev_flags off,
581 enum netdev_flags on, enum netdev_flags *old_flags);
583 /* Returns a sequence number which indicates changes in one of 'netdev''s
584 * properties. The returned sequence number must be nonzero so that
585 * callers have a value which they may use as a reset when tracking
588 * Minimally, the returned sequence number is required to change whenever
589 * 'netdev''s flags, features, ethernet address, or carrier changes. The
590 * returned sequence number is allowed to change even when 'netdev' doesn't
591 * change, although implementations should try to avoid this. */
592 unsigned int (*change_seq)(const struct netdev *netdev);
595 int netdev_register_provider(const struct netdev_class *);
596 int netdev_unregister_provider(const char *type);
597 const struct netdev_class *netdev_lookup_provider(const char *type);
599 extern const struct netdev_class netdev_linux_class;
600 extern const struct netdev_class netdev_internal_class;
601 extern const struct netdev_class netdev_tap_class;
603 extern const struct netdev_class netdev_bsd_class;
606 extern const struct netdev_class netdev_tunnel_class;
607 extern const struct netdev_class netdev_pltap_class;
613 #endif /* netdev.h */