]> git.proxmox.com Git - mirror_ovs.git/blob - lib/netdev-provider.h
dpctl: Properly reflect a rule's offloaded to HW state
[mirror_ovs.git] / lib / netdev-provider.h
1 /*
2 * Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2016 Nicira, Inc.
3 *
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:
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
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.
15 */
16
17 #ifndef NETDEV_PROVIDER_H
18 #define NETDEV_PROVIDER_H 1
19
20 /* Generic interface to network devices. */
21
22 #include "connectivity.h"
23 #include "netdev.h"
24 #include "openvswitch/list.h"
25 #include "ovs-numa.h"
26 #include "packets.h"
27 #include "seq.h"
28 #include "openvswitch/shash.h"
29 #include "smap.h"
30
31 #ifdef __cplusplus
32 extern "C" {
33 #endif
34
35 struct netdev_tnl_build_header_params;
36 #define NETDEV_NUMA_UNSPEC OVS_NUMA_UNSPEC
37
38 /* A network device (e.g. an Ethernet device).
39 *
40 * Network device implementations may read these members but should not modify
41 * them. */
42 struct netdev {
43 /* The following do not change during the lifetime of a struct netdev. */
44 char *name; /* Name of network device. */
45 const struct netdev_class *netdev_class; /* Functions to control
46 this device. */
47
48 /* If this is 'true' the user did not specify a netdev_class when
49 * opening this device, and therefore got assigned to the "system" class */
50 bool auto_classified;
51
52 /* If this is 'true', the user explicitly specified an MTU for this
53 * netdev. Otherwise, Open vSwitch is allowed to override it. */
54 bool mtu_user_config;
55
56 int ref_cnt; /* Times this devices was opened. */
57
58 /* A sequence number which indicates changes in one of 'netdev''s
59 * properties. It must be nonzero so that users have a value which
60 * they may use as a reset when tracking 'netdev'.
61 *
62 * Minimally, the sequence number is required to change whenever
63 * 'netdev''s flags, features, ethernet address, or carrier changes. */
64 uint64_t change_seq;
65
66 /* A netdev provider might be unable to change some of the device's
67 * parameter (n_rxq, mtu) when the device is in use. In this case
68 * the provider can notify the upper layer by calling
69 * netdev_request_reconfigure(). The upper layer will react by stopping
70 * the operations on the device and calling netdev_reconfigure() to allow
71 * the configuration changes. 'last_reconfigure_seq' remembers the value
72 * of 'reconfigure_seq' when the last reconfiguration happened. */
73 struct seq *reconfigure_seq;
74 uint64_t last_reconfigure_seq;
75
76 /* The core netdev code initializes these at netdev construction and only
77 * provide read-only access to its client. Netdev implementations may
78 * modify them. */
79 int n_txq;
80 int n_rxq;
81 struct shash_node *node; /* Pointer to element in global map. */
82 struct ovs_list saved_flags_list; /* Contains "struct netdev_saved_flags". */
83 };
84
85 static inline void
86 netdev_change_seq_changed(const struct netdev *netdev_)
87 {
88 struct netdev *netdev = CONST_CAST(struct netdev *, netdev_);
89 seq_change(connectivity_seq_get());
90 netdev->change_seq++;
91 if (!netdev->change_seq) {
92 netdev->change_seq++;
93 }
94 }
95
96 static inline void
97 netdev_request_reconfigure(struct netdev *netdev)
98 {
99 seq_change(netdev->reconfigure_seq);
100 }
101
102 const char *netdev_get_type(const struct netdev *);
103 const struct netdev_class *netdev_get_class(const struct netdev *);
104 const char *netdev_get_name(const struct netdev *);
105 struct netdev *netdev_from_name(const char *name);
106 void netdev_get_devices(const struct netdev_class *,
107 struct shash *device_list);
108 struct netdev **netdev_get_vports(size_t *size);
109
110 /* A data structure for capturing packets received by a network device.
111 *
112 * Network device implementations may read these members but should not modify
113 * them.
114 *
115 * None of these members change during the lifetime of a struct netdev_rxq. */
116 struct netdev_rxq {
117 struct netdev *netdev; /* Owns a reference to the netdev. */
118 int queue_id;
119 };
120
121 struct netdev *netdev_rxq_get_netdev(const struct netdev_rxq *);
122
123
124 struct netdev_flow_dump {
125 struct netdev *netdev;
126 odp_port_t port;
127 bool terse;
128 struct nl_dump *nl_dump;
129 };
130
131 /* Network device class structure, to be defined by each implementation of a
132 * network device.
133 *
134 * These functions return 0 if successful or a positive errno value on failure,
135 * except where otherwise noted.
136 *
137 *
138 * Data Structures
139 * ===============
140 *
141 * These functions work primarily with two different kinds of data structures:
142 *
143 * - "struct netdev", which represents a network device.
144 *
145 * - "struct netdev_rxq", which represents a handle for capturing packets
146 * received on a network device
147 *
148 * Each of these data structures contains all of the implementation-independent
149 * generic state for the respective concept, called the "base" state. None of
150 * them contains any extra space for implementations to use. Instead, each
151 * implementation is expected to declare its own data structure that contains
152 * an instance of the generic data structure plus additional
153 * implementation-specific members, called the "derived" state. The
154 * implementation can use casts or (preferably) the CONTAINER_OF macro to
155 * obtain access to derived state given only a pointer to the embedded generic
156 * data structure.
157 *
158 *
159 * Life Cycle
160 * ==========
161 *
162 * Four stylized functions accompany each of these data structures:
163 *
164 * "alloc" "construct" "destruct" "dealloc"
165 * ------------ ---------------- --------------- --------------
166 * netdev ->alloc ->construct ->destruct ->dealloc
167 * netdev_rxq ->rxq_alloc ->rxq_construct ->rxq_destruct ->rxq_dealloc
168 *
169 * Any instance of a given data structure goes through the following life
170 * cycle:
171 *
172 * 1. The client calls the "alloc" function to obtain raw memory. If "alloc"
173 * fails, skip all the other steps.
174 *
175 * 2. The client initializes all of the data structure's base state. If this
176 * fails, skip to step 7.
177 *
178 * 3. The client calls the "construct" function. The implementation
179 * initializes derived state. It may refer to the already-initialized
180 * base state. If "construct" fails, skip to step 6.
181 *
182 * 4. The data structure is now initialized and in use.
183 *
184 * 5. When the data structure is no longer needed, the client calls the
185 * "destruct" function. The implementation uninitializes derived state.
186 * The base state has not been uninitialized yet, so the implementation
187 * may still refer to it.
188 *
189 * 6. The client uninitializes all of the data structure's base state.
190 *
191 * 7. The client calls the "dealloc" to free the raw memory. The
192 * implementation must not refer to base or derived state in the data
193 * structure, because it has already been uninitialized.
194 *
195 * If netdev support multi-queue IO then netdev->construct should set initialize
196 * netdev->n_rxq to number of queues.
197 *
198 * Each "alloc" function allocates and returns a new instance of the respective
199 * data structure. The "alloc" function is not given any information about the
200 * use of the new data structure, so it cannot perform much initialization.
201 * Its purpose is just to ensure that the new data structure has enough room
202 * for base and derived state. It may return a null pointer if memory is not
203 * available, in which case none of the other functions is called.
204 *
205 * Each "construct" function initializes derived state in its respective data
206 * structure. When "construct" is called, all of the base state has already
207 * been initialized, so the "construct" function may refer to it. The
208 * "construct" function is allowed to fail, in which case the client calls the
209 * "dealloc" function (but not the "destruct" function).
210 *
211 * Each "destruct" function uninitializes and frees derived state in its
212 * respective data structure. When "destruct" is called, the base state has
213 * not yet been uninitialized, so the "destruct" function may refer to it. The
214 * "destruct" function is not allowed to fail.
215 *
216 * Each "dealloc" function frees raw memory that was allocated by the
217 * "alloc" function. The memory's base and derived members might not have ever
218 * been initialized (but if "construct" returned successfully, then it has been
219 * "destruct"ed already). The "dealloc" function is not allowed to fail.
220 *
221 *
222 * Device Change Notification
223 * ==========================
224 *
225 * Minimally, implementations are required to report changes to netdev flags,
226 * features, ethernet address or carrier through connectivity_seq. Changes to
227 * other properties are allowed to cause notification through this interface,
228 * although implementations should try to avoid this. connectivity_seq_get()
229 * can be used to acquire a reference to the struct seq. The interface is
230 * described in detail in seq.h. */
231 struct netdev_class {
232 /* Type of netdevs in this class, e.g. "system", "tap", "gre", etc.
233 *
234 * One of the providers should supply a "system" type, since this is
235 * the type assumed if no type is specified when opening a netdev.
236 * The "system" type corresponds to an existing network device on
237 * the system. */
238 const char *type;
239
240 /* If 'true' then this netdev should be polled by PMD threads. */
241 bool is_pmd;
242
243 /* ## ------------------- ## */
244 /* ## Top-Level Functions ## */
245 /* ## ------------------- ## */
246
247 /* Called when the netdev provider is registered, typically at program
248 * startup. Returning an error from this function will prevent any network
249 * device in this class from being opened.
250 *
251 * This function may be set to null if a network device class needs no
252 * initialization at registration time. */
253 int (*init)(void);
254
255 /* Performs periodic work needed by netdevs of this class. May be null if
256 * no periodic work is necessary.
257 *
258 * 'netdev_class' points to the class. It is useful in case the same
259 * function is used to implement different classes. */
260 void (*run)(const struct netdev_class *netdev_class);
261
262 /* Arranges for poll_block() to wake up if the "run" member function needs
263 * to be called. Implementations are additionally required to wake
264 * whenever something changes in any of its netdevs which would cause their
265 * ->change_seq() function to change its result. May be null if nothing is
266 * needed here.
267 *
268 * 'netdev_class' points to the class. It is useful in case the same
269 * function is used to implement different classes. */
270 void (*wait)(const struct netdev_class *netdev_class);
271
272 /* ## ---------------- ## */
273 /* ## netdev Functions ## */
274 /* ## ---------------- ## */
275
276 /* Life-cycle functions for a netdev. See the large comment above on
277 * struct netdev_class. */
278 struct netdev *(*alloc)(void);
279 int (*construct)(struct netdev *);
280 void (*destruct)(struct netdev *);
281 void (*dealloc)(struct netdev *);
282
283 /* Fetches the device 'netdev''s configuration, storing it in 'args'.
284 * The caller owns 'args' and pre-initializes it to an empty smap.
285 *
286 * If this netdev class does not have any configuration options, this may
287 * be a null pointer. */
288 int (*get_config)(const struct netdev *netdev, struct smap *args);
289
290 /* Changes the device 'netdev''s configuration to 'args'.
291 *
292 * If this netdev class does not support configuration, this may be a null
293 * pointer.
294 *
295 * If the return value is not zero (meaning that an error occurred),
296 * the provider can allocate a string with an error message in '*errp'.
297 * The caller has to call free on it. */
298 int (*set_config)(struct netdev *netdev, const struct smap *args,
299 char **errp);
300
301 /* Returns the tunnel configuration of 'netdev'. If 'netdev' is
302 * not a tunnel, returns null.
303 *
304 * If this function would always return null, it may be null instead. */
305 const struct netdev_tunnel_config *
306 (*get_tunnel_config)(const struct netdev *netdev);
307
308 /* Build Tunnel header. Ethernet and ip header parameters are passed to
309 * tunnel implementation to build entire outer header for given flow. */
310 int (*build_header)(const struct netdev *, struct ovs_action_push_tnl *data,
311 const struct netdev_tnl_build_header_params *params);
312
313 /* build_header() can not build entire header for all packets for given
314 * flow. Push header is called for packet to build header specific to
315 * a packet on actual transmit. It uses partial header build by
316 * build_header() which is passed as data. */
317 void (*push_header)(const struct netdev *,
318 struct dp_packet *packet,
319 const struct ovs_action_push_tnl *data);
320
321 /* Pop tunnel header from packet, build tunnel metadata and resize packet
322 * for further processing.
323 * Returns NULL in case of error or tunnel implementation queued packet for further
324 * processing. */
325 struct dp_packet * (*pop_header)(struct dp_packet *packet);
326
327 /* Returns the id of the numa node the 'netdev' is on. If there is no
328 * such info, returns NETDEV_NUMA_UNSPEC. */
329 int (*get_numa_id)(const struct netdev *netdev);
330
331 /* Configures the number of tx queues of 'netdev'. Returns 0 if successful,
332 * otherwise a positive errno value.
333 *
334 * 'n_txq' specifies the exact number of transmission queues to create.
335 *
336 * The caller will call netdev_reconfigure() (if necessary) before using
337 * netdev_send() on any of the newly configured queues, giving the provider
338 * a chance to adjust its settings.
339 *
340 * On error, the tx queue configuration is unchanged. */
341 int (*set_tx_multiq)(struct netdev *netdev, unsigned int n_txq);
342
343 /* Sends buffers on 'netdev'.
344 * Returns 0 if successful (for every buffer), otherwise a positive errno
345 * value. Returns EAGAIN without blocking if one or more packets cannot be
346 * queued immediately. Returns EMSGSIZE if a partial packet was transmitted
347 * or if a packet is too big or too small to transmit on the device.
348 *
349 * If the function returns a non-zero value, some of the packets might have
350 * been sent anyway.
351 *
352 * The caller transfers ownership of all the packets to the network
353 * device, regardless of success.
354 *
355 * If 'concurrent_txq' is true, the caller may perform concurrent calls
356 * to netdev_send() with the same 'qid'. The netdev provider is responsible
357 * for making sure that these concurrent calls do not create a race
358 * condition by using locking or other synchronization if required.
359 *
360 * The network device is expected to maintain one or more packet
361 * transmission queues, so that the caller does not ordinarily have to
362 * do additional queuing of packets. 'qid' specifies the queue to use
363 * and can be ignored if the implementation does not support multiple
364 * queues.
365 *
366 * May return EOPNOTSUPP if a network device does not implement packet
367 * transmission through this interface. This function may be set to null
368 * if it would always return EOPNOTSUPP anyhow. (This will prevent the
369 * network device from being usefully used by the netdev-based "userspace
370 * datapath". It will also prevent the OVS implementation of bonding from
371 * working properly over 'netdev'.) */
372 int (*send)(struct netdev *netdev, int qid, struct dp_packet_batch *batch,
373 bool concurrent_txq);
374
375 /* Registers with the poll loop to wake up from the next call to
376 * poll_block() when the packet transmission queue for 'netdev' has
377 * sufficient room to transmit a packet with netdev_send().
378 *
379 * The network device is expected to maintain one or more packet
380 * transmission queues, so that the caller does not ordinarily have to
381 * do additional queuing of packets. 'qid' specifies the queue to use
382 * and can be ignored if the implementation does not support multiple
383 * queues.
384 *
385 * May be null if not needed, such as for a network device that does not
386 * implement packet transmission through the 'send' member function. */
387 void (*send_wait)(struct netdev *netdev, int qid);
388
389 /* Sets 'netdev''s Ethernet address to 'mac' */
390 int (*set_etheraddr)(struct netdev *netdev, const struct eth_addr mac);
391
392 /* Retrieves 'netdev''s Ethernet address into 'mac'.
393 *
394 * This address will be advertised as 'netdev''s MAC address through the
395 * OpenFlow protocol, among other uses. */
396 int (*get_etheraddr)(const struct netdev *netdev, struct eth_addr *mac);
397
398 /* Retrieves 'netdev''s MTU into '*mtup'.
399 *
400 * The MTU is the maximum size of transmitted (and received) packets, in
401 * bytes, not including the hardware header; thus, this is typically 1500
402 * bytes for Ethernet devices.
403 *
404 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
405 * this function should return EOPNOTSUPP. This function may be set to
406 * null if it would always return EOPNOTSUPP. */
407 int (*get_mtu)(const struct netdev *netdev, int *mtup);
408
409 /* Sets 'netdev''s MTU to 'mtu'.
410 *
411 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
412 * this function should return EOPNOTSUPP. This function may be set to
413 * null if it would always return EOPNOTSUPP. */
414 int (*set_mtu)(struct netdev *netdev, int mtu);
415
416 /* Returns the ifindex of 'netdev', if successful, as a positive number.
417 * On failure, returns a negative errno value.
418 *
419 * The desired semantics of the ifindex value are a combination of those
420 * specified by POSIX for if_nametoindex() and by SNMP for ifIndex. An
421 * ifindex value should be unique within a host and remain stable at least
422 * until reboot. SNMP says an ifindex "ranges between 1 and the value of
423 * ifNumber" but many systems do not follow this rule anyhow.
424 *
425 * This function may be set to null if it would always return -EOPNOTSUPP.
426 */
427 int (*get_ifindex)(const struct netdev *netdev);
428
429 /* Sets 'carrier' to true if carrier is active (link light is on) on
430 * 'netdev'.
431 *
432 * May be null if device does not provide carrier status (will be always
433 * up as long as device is up).
434 */
435 int (*get_carrier)(const struct netdev *netdev, bool *carrier);
436
437 /* Returns the number of times 'netdev''s carrier has changed since being
438 * initialized.
439 *
440 * If null, callers will assume the number of carrier resets is zero. */
441 long long int (*get_carrier_resets)(const struct netdev *netdev);
442
443 /* Forces ->get_carrier() to poll 'netdev''s MII registers for link status
444 * instead of checking 'netdev''s carrier. 'netdev''s MII registers will
445 * be polled once every 'interval' milliseconds. If 'netdev' does not
446 * support MII, another method may be used as a fallback. If 'interval' is
447 * less than or equal to zero, reverts ->get_carrier() to its normal
448 * behavior.
449 *
450 * Most network devices won't support this feature and will set this
451 * function pointer to NULL, which is equivalent to returning EOPNOTSUPP.
452 */
453 int (*set_miimon_interval)(struct netdev *netdev, long long int interval);
454
455 /* Retrieves current device stats for 'netdev' into 'stats'.
456 *
457 * A network device that supports some statistics but not others, it should
458 * set the values of the unsupported statistics to all-1-bits
459 * (UINT64_MAX). */
460 int (*get_stats)(const struct netdev *netdev, struct netdev_stats *);
461
462 /* Retrieves current device custom stats for 'netdev' into 'custom_stats'.
463 *
464 * A network device should return only available statistics (if any).
465 * If there are not statistics available, empty array should be
466 * returned.
467 *
468 * The caller initializes 'custom_stats' before calling this function.
469 * The caller takes ownership over allocated array of counters inside
470 * structure netdev_custom_stats.
471 * */
472 int (*get_custom_stats)(const struct netdev *netdev,
473 struct netdev_custom_stats *custom_stats);
474
475 /* Stores the features supported by 'netdev' into each of '*current',
476 * '*advertised', '*supported', and '*peer'. Each value is a bitmap of
477 * NETDEV_F_* bits.
478 *
479 * This function may be set to null if it would always return EOPNOTSUPP.
480 */
481 int (*get_features)(const struct netdev *netdev,
482 enum netdev_features *current,
483 enum netdev_features *advertised,
484 enum netdev_features *supported,
485 enum netdev_features *peer);
486
487 /* Set the features advertised by 'netdev' to 'advertise', which is a
488 * set of NETDEV_F_* bits.
489 *
490 * This function may be set to null for a network device that does not
491 * support configuring advertisements. */
492 int (*set_advertisements)(struct netdev *netdev,
493 enum netdev_features advertise);
494
495 /* Returns 'netdev''s configured packet_type mode.
496 *
497 * This function may be set to null if it would always return
498 * NETDEV_PT_LEGACY_L2. */
499 enum netdev_pt_mode (*get_pt_mode)(const struct netdev *netdev);
500
501 /* Attempts to set input rate limiting (policing) policy, such that up to
502 * 'kbits_rate' kbps of traffic is accepted, with a maximum accumulative
503 * burst size of 'kbits' kb.
504 *
505 * This function may be set to null if policing is not supported. */
506 int (*set_policing)(struct netdev *netdev, unsigned int kbits_rate,
507 unsigned int kbits_burst);
508
509 /* Adds to 'types' all of the forms of QoS supported by 'netdev', or leaves
510 * it empty if 'netdev' does not support QoS. Any names added to 'types'
511 * should be documented as valid for the "type" column in the "QoS" table
512 * in vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
513 *
514 * Every network device must support disabling QoS with a type of "", but
515 * this function must not add "" to 'types'.
516 *
517 * The caller is responsible for initializing 'types' (e.g. with
518 * sset_init()) before calling this function. The caller retains ownership
519 * of 'types'.
520 *
521 * May be NULL if 'netdev' does not support QoS at all. */
522 int (*get_qos_types)(const struct netdev *netdev, struct sset *types);
523
524 /* Queries 'netdev' for its capabilities regarding the specified 'type' of
525 * QoS. On success, initializes 'caps' with the QoS capabilities.
526 *
527 * Should return EOPNOTSUPP if 'netdev' does not support 'type'. May be
528 * NULL if 'netdev' does not support QoS at all. */
529 int (*get_qos_capabilities)(const struct netdev *netdev,
530 const char *type,
531 struct netdev_qos_capabilities *caps);
532
533 /* Queries 'netdev' about its currently configured form of QoS. If
534 * successful, stores the name of the current form of QoS into '*typep'
535 * and any details of configuration as string key-value pairs in
536 * 'details'.
537 *
538 * A '*typep' of "" indicates that QoS is currently disabled on 'netdev'.
539 *
540 * The caller initializes 'details' before calling this function. The
541 * caller takes ownership of the string key-values pairs added to
542 * 'details'.
543 *
544 * The netdev retains ownership of '*typep'.
545 *
546 * '*typep' will be one of the types returned by netdev_get_qos_types() for
547 * 'netdev'. The contents of 'details' should be documented as valid for
548 * '*typep' in the "other_config" column in the "QoS" table in
549 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
550 *
551 * May be NULL if 'netdev' does not support QoS at all. */
552 int (*get_qos)(const struct netdev *netdev,
553 const char **typep, struct smap *details);
554
555 /* Attempts to reconfigure QoS on 'netdev', changing the form of QoS to
556 * 'type' with details of configuration from 'details'.
557 *
558 * On error, the previous QoS configuration is retained.
559 *
560 * When this function changes the type of QoS (not just 'details'), this
561 * also resets all queue configuration for 'netdev' to their defaults
562 * (which depend on the specific type of QoS). Otherwise, the queue
563 * configuration for 'netdev' is unchanged.
564 *
565 * 'type' should be "" (to disable QoS) or one of the types returned by
566 * netdev_get_qos_types() for 'netdev'. The contents of 'details' should
567 * be documented as valid for the given 'type' in the "other_config" column
568 * in the "QoS" table in vswitchd/vswitch.xml (which is built as
569 * ovs-vswitchd.conf.db(8)).
570 *
571 * May be NULL if 'netdev' does not support QoS at all. */
572 int (*set_qos)(struct netdev *netdev,
573 const char *type, const struct smap *details);
574
575 /* Queries 'netdev' for information about the queue numbered 'queue_id'.
576 * If successful, adds that information as string key-value pairs to
577 * 'details'. Returns 0 if successful, otherwise a positive errno value.
578 *
579 * Should return EINVAL if 'queue_id' is greater than or equal to the
580 * number of supported queues (as reported in the 'n_queues' member of
581 * struct netdev_qos_capabilities by 'get_qos_capabilities').
582 *
583 * The caller initializes 'details' before calling this function. The
584 * caller takes ownership of the string key-values pairs added to
585 * 'details'.
586 *
587 * The returned contents of 'details' should be documented as valid for the
588 * given 'type' in the "other_config" column in the "Queue" table in
589 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
590 */
591 int (*get_queue)(const struct netdev *netdev,
592 unsigned int queue_id, struct smap *details);
593
594 /* Configures the queue numbered 'queue_id' on 'netdev' with the key-value
595 * string pairs in 'details'. The contents of 'details' should be
596 * documented as valid for the given 'type' in the "other_config" column in
597 * the "Queue" table in vswitchd/vswitch.xml (which is built as
598 * ovs-vswitchd.conf.db(8)). Returns 0 if successful, otherwise a positive
599 * errno value. On failure, the given queue's configuration should be
600 * unmodified.
601 *
602 * Should return EINVAL if 'queue_id' is greater than or equal to the
603 * number of supported queues (as reported in the 'n_queues' member of
604 * struct netdev_qos_capabilities by 'get_qos_capabilities'), or if
605 * 'details' is invalid for the type of queue.
606 *
607 * This function does not modify 'details', and the caller retains
608 * ownership of it.
609 *
610 * May be NULL if 'netdev' does not support QoS at all. */
611 int (*set_queue)(struct netdev *netdev,
612 unsigned int queue_id, const struct smap *details);
613
614 /* Attempts to delete the queue numbered 'queue_id' from 'netdev'.
615 *
616 * Should return EINVAL if 'queue_id' is greater than or equal to the
617 * number of supported queues (as reported in the 'n_queues' member of
618 * struct netdev_qos_capabilities by 'get_qos_capabilities'). Should
619 * return EOPNOTSUPP if 'queue_id' is valid but may not be deleted (e.g. if
620 * 'netdev' has a fixed set of queues with the current QoS mode).
621 *
622 * May be NULL if 'netdev' does not support QoS at all, or if all of its
623 * QoS modes have fixed sets of queues. */
624 int (*delete_queue)(struct netdev *netdev, unsigned int queue_id);
625
626 /* Obtains statistics about 'queue_id' on 'netdev'. Fills 'stats' with the
627 * queue's statistics. May set individual members of 'stats' to all-1-bits
628 * if the statistic is unavailable.
629 *
630 * May be NULL if 'netdev' does not support QoS at all. */
631 int (*get_queue_stats)(const struct netdev *netdev, unsigned int queue_id,
632 struct netdev_queue_stats *stats);
633
634 /* Attempts to begin dumping the queues in 'netdev'. On success, returns 0
635 * and initializes '*statep' with any data needed for iteration. On
636 * failure, returns a positive errno value.
637 *
638 * May be NULL if 'netdev' does not support QoS at all. */
639 int (*queue_dump_start)(const struct netdev *netdev, void **statep);
640
641 /* Attempts to retrieve another queue from 'netdev' for 'state', which was
642 * initialized by a successful call to the 'queue_dump_start' function for
643 * 'netdev'. On success, stores a queue ID into '*queue_id' and fills
644 * 'details' with the configuration of the queue with that ID. Returns EOF
645 * if the last queue has been dumped, or a positive errno value on error.
646 * This function will not be called again once it returns nonzero once for
647 * a given iteration (but the 'queue_dump_done' function will be called
648 * afterward).
649 *
650 * The caller initializes and clears 'details' before calling this
651 * function. The caller takes ownership of the string key-values pairs
652 * added to 'details'.
653 *
654 * The returned contents of 'details' should be documented as valid for the
655 * given 'type' in the "other_config" column in the "Queue" table in
656 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
657 *
658 * May be NULL if 'netdev' does not support QoS at all. */
659 int (*queue_dump_next)(const struct netdev *netdev, void *state,
660 unsigned int *queue_id, struct smap *details);
661
662 /* Releases resources from 'netdev' for 'state', which was initialized by a
663 * successful call to the 'queue_dump_start' function for 'netdev'.
664 *
665 * May be NULL if 'netdev' does not support QoS at all. */
666 int (*queue_dump_done)(const struct netdev *netdev, void *state);
667
668 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
669 * ID, its statistics, and the 'aux' specified by the caller. The order of
670 * iteration is unspecified, but (when successful) each queue must be
671 * visited exactly once.
672 *
673 * 'cb' will not modify or free the statistics passed in. */
674 int (*dump_queue_stats)(const struct netdev *netdev,
675 void (*cb)(unsigned int queue_id,
676 struct netdev_queue_stats *,
677 void *aux),
678 void *aux);
679
680 /* Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
681 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared.
682 *
683 * This function may be set to null if it would always return EOPNOTSUPP
684 * anyhow. */
685 int (*set_in4)(struct netdev *netdev, struct in_addr addr,
686 struct in_addr mask);
687
688 /* Returns all assigned IP address to 'netdev' and returns 0.
689 * API allocates array of address and masks and set it to
690 * '*addr' and '*mask'.
691 * Otherwise, returns a positive errno value and sets '*addr', '*mask
692 * and '*n_addr' to NULL.
693 *
694 * The following error values have well-defined meanings:
695 *
696 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv6 address.
697 *
698 * - EOPNOTSUPP: No IPv6 network stack attached to 'netdev'.
699 *
700 * 'addr' may be null, in which case the address itself is not reported. */
701 int (*get_addr_list)(const struct netdev *netdev, struct in6_addr **in,
702 struct in6_addr **mask, int *n_in6);
703
704 /* Adds 'router' as a default IP gateway for the TCP/IP stack that
705 * corresponds to 'netdev'.
706 *
707 * This function may be set to null if it would always return EOPNOTSUPP
708 * anyhow. */
709 int (*add_router)(struct netdev *netdev, struct in_addr router);
710
711 /* Looks up the next hop for 'host' in the host's routing table. If
712 * successful, stores the next hop gateway's address (0 if 'host' is on a
713 * directly connected network) in '*next_hop' and a copy of the name of the
714 * device to reach 'host' in '*netdev_name', and returns 0. The caller is
715 * responsible for freeing '*netdev_name' (by calling free()).
716 *
717 * This function may be set to null if it would always return EOPNOTSUPP
718 * anyhow. */
719 int (*get_next_hop)(const struct in_addr *host, struct in_addr *next_hop,
720 char **netdev_name);
721
722 /* Retrieves driver information of the device.
723 *
724 * Populates 'smap' with key-value pairs representing the status of the
725 * device. 'smap' is a set of key-value string pairs representing netdev
726 * type specific information. For more information see
727 * ovs-vswitchd.conf.db(5).
728 *
729 * The caller is responsible for destroying 'smap' and its data.
730 *
731 * This function may be set to null if it would always return EOPNOTSUPP
732 * anyhow. */
733 int (*get_status)(const struct netdev *netdev, struct smap *smap);
734
735 /* Looks up the ARP table entry for 'ip' on 'netdev' and stores the
736 * corresponding MAC address in 'mac'. A return value of ENXIO, in
737 * particular, indicates that there is no ARP table entry for 'ip' on
738 * 'netdev'.
739 *
740 * This function may be set to null if it would always return EOPNOTSUPP
741 * anyhow. */
742 int (*arp_lookup)(const struct netdev *netdev, ovs_be32 ip,
743 struct eth_addr *mac);
744
745 /* Retrieves the current set of flags on 'netdev' into '*old_flags'. Then,
746 * turns off the flags that are set to 1 in 'off' and turns on the flags
747 * that are set to 1 in 'on'. (No bit will be set to 1 in both 'off' and
748 * 'on'; that is, off & on == 0.)
749 *
750 * This function may be invoked from a signal handler. Therefore, it
751 * should not do anything that is not signal-safe (such as logging). */
752 int (*update_flags)(struct netdev *netdev, enum netdev_flags off,
753 enum netdev_flags on, enum netdev_flags *old_flags);
754
755 /* If the provider called netdev_request_reconfigure(), the upper layer
756 * will eventually call this. The provider can update the device
757 * configuration knowing that the upper layer will not call rxq_recv() or
758 * send() until this function returns.
759 *
760 * On error, the configuration is indeterminant and the device cannot be
761 * used to send and receive packets until a successful configuration is
762 * applied. */
763 int (*reconfigure)(struct netdev *netdev);
764 /* ## -------------------- ## */
765 /* ## netdev_rxq Functions ## */
766 /* ## -------------------- ## */
767
768 /* If a particular netdev class does not support receiving packets, all these
769 * function pointers must be NULL. */
770
771 /* Life-cycle functions for a netdev_rxq. See the large comment above on
772 * struct netdev_class. */
773 struct netdev_rxq *(*rxq_alloc)(void);
774 int (*rxq_construct)(struct netdev_rxq *);
775 void (*rxq_destruct)(struct netdev_rxq *);
776 void (*rxq_dealloc)(struct netdev_rxq *);
777
778 /* Attempts to receive a batch of packets from 'rx'. In 'batch', the
779 * caller supplies 'packets' as the pointer to the beginning of an array
780 * of NETDEV_MAX_BURST pointers to dp_packet. If successful, the
781 * implementation stores pointers to up to NETDEV_MAX_BURST dp_packets into
782 * the array, transferring ownership of the packets to the caller, stores
783 * the number of received packets into 'count', and returns 0.
784 *
785 * The implementation does not necessarily initialize any non-data members
786 * of 'packets' in 'batch'. That is, the caller must initialize layer
787 * pointers and metadata itself, if desired, e.g. with pkt_metadata_init()
788 * and miniflow_extract().
789 *
790 * Implementations should allocate buffers with DP_NETDEV_HEADROOM bytes of
791 * headroom.
792 *
793 * If the caller provides a non-NULL qfill pointer, the implementation
794 * should return the number (zero or more) of remaining packets in the
795 * queue after the reception the current batch, if it supports that,
796 * or -ENOTSUP otherwise.
797 *
798 * Returns EAGAIN immediately if no packet is ready to be received or
799 * another positive errno value if an error was encountered. */
800 int (*rxq_recv)(struct netdev_rxq *rx, struct dp_packet_batch *batch,
801 int *qfill);
802
803 /* Registers with the poll loop to wake up from the next call to
804 * poll_block() when a packet is ready to be received with
805 * netdev_rxq_recv() on 'rx'. */
806 void (*rxq_wait)(struct netdev_rxq *rx);
807
808 /* Discards all packets waiting to be received from 'rx'. */
809 int (*rxq_drain)(struct netdev_rxq *rx);
810
811 /* ## -------------------------------- ## */
812 /* ## netdev flow offloading functions ## */
813 /* ## -------------------------------- ## */
814
815 /* If a particular netdev class does not support offloading flows,
816 * all these function pointers must be NULL. */
817
818 /* Flush all offloaded flows from a netdev.
819 * Return 0 if successful, otherwise returns a positive errno value. */
820 int (*flow_flush)(struct netdev *);
821
822 /* Flow dumping interface.
823 *
824 * This is the back-end for the flow dumping interface described in
825 * dpif.h. Please read the comments there first, because this code
826 * closely follows it.
827 *
828 * On success returns 0 and allocates data, on failure returns
829 * positive errno. */
830 int (*flow_dump_create)(struct netdev *, struct netdev_flow_dump **dump);
831 int (*flow_dump_destroy)(struct netdev_flow_dump *);
832
833 /* Returns true if there are more flows to dump.
834 * 'rbuffer' is used as a temporary buffer and needs to be pre allocated
835 * by the caller. While there are more flows the same 'rbuffer'
836 * should be provided. 'wbuffer' is used to store dumped actions and needs
837 * to be pre allocated by the caller. */
838 bool (*flow_dump_next)(struct netdev_flow_dump *, struct match *,
839 struct nlattr **actions,
840 struct dpif_flow_stats *stats,
841 struct dpif_flow_attrs *attrs, ovs_u128 *ufid,
842 struct ofpbuf *rbuffer, struct ofpbuf *wbuffer);
843
844 /* Offload the given flow on netdev.
845 * To modify a flow, use the same ufid.
846 * 'actions' are in netlink format, as with struct dpif_flow_put.
847 * 'info' is extra info needed to offload the flow.
848 * 'stats' is populated according to the rules set out in the description
849 * above 'struct dpif_flow_put'.
850 * Return 0 if successful, otherwise returns a positive errno value. */
851 int (*flow_put)(struct netdev *, struct match *, struct nlattr *actions,
852 size_t actions_len, const ovs_u128 *ufid,
853 struct offload_info *info, struct dpif_flow_stats *);
854
855 /* Queries a flow specified by ufid on netdev.
856 * Fills output buffer as 'wbuffer' in flow_dump_next, which
857 * needs to be be pre allocated.
858 * Return 0 if successful, otherwise returns a positive errno value. */
859 int (*flow_get)(struct netdev *, struct match *, struct nlattr **actions,
860 const ovs_u128 *ufid, struct dpif_flow_stats *,
861 struct dpif_flow_attrs *, struct ofpbuf *wbuffer);
862
863 /* Delete a flow specified by ufid from netdev.
864 * 'stats' is populated according to the rules set out in the description
865 * above 'struct dpif_flow_del'.
866 * Return 0 if successful, otherwise returns a positive errno value. */
867 int (*flow_del)(struct netdev *, const ovs_u128 *ufid,
868 struct dpif_flow_stats *);
869
870 /* Initializies the netdev flow api.
871 * Return 0 if successful, otherwise returns a positive errno value. */
872 int (*init_flow_api)(struct netdev *);
873 };
874
875 int netdev_register_provider(const struct netdev_class *);
876 int netdev_unregister_provider(const char *type);
877
878 #if defined(__FreeBSD__) || defined(__NetBSD__)
879 extern const struct netdev_class netdev_bsd_class;
880 #elif defined(_WIN32)
881 extern const struct netdev_class netdev_windows_class;
882 #else
883 extern const struct netdev_class netdev_linux_class;
884 #endif
885 extern const struct netdev_class netdev_internal_class;
886 extern const struct netdev_class netdev_tap_class;
887
888 #ifdef __cplusplus
889 }
890 #endif
891
892 #define NO_OFFLOAD_API NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL
893
894 #endif /* netdev.h */