]> git.proxmox.com Git - mirror_ovs.git/blob - lib/netdev-provider.h
ovsdb-idl: Fix iteration over tracked rows with no actual data.
[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 "netdev-offload.h"
25 #include "openvswitch/list.h"
26 #include "ovs-numa.h"
27 #include "ovs-rcu.h"
28 #include "packets.h"
29 #include "seq.h"
30 #include "openvswitch/shash.h"
31 #include "smap.h"
32
33 #ifdef __cplusplus
34 extern "C" {
35 #endif
36
37 struct netdev_tnl_build_header_params;
38 #define NETDEV_NUMA_UNSPEC OVS_NUMA_UNSPEC
39
40 enum netdev_ol_flags {
41 NETDEV_TX_OFFLOAD_IPV4_CKSUM = 1 << 0,
42 NETDEV_TX_OFFLOAD_TCP_CKSUM = 1 << 1,
43 NETDEV_TX_OFFLOAD_UDP_CKSUM = 1 << 2,
44 NETDEV_TX_OFFLOAD_SCTP_CKSUM = 1 << 3,
45 NETDEV_TX_OFFLOAD_TCP_TSO = 1 << 4,
46 };
47
48 /* A network device (e.g. an Ethernet device).
49 *
50 * Network device implementations may read these members but should not modify
51 * them. */
52 struct netdev {
53 /* The following do not change during the lifetime of a struct netdev. */
54 char *name; /* Name of network device. */
55 const struct netdev_class *netdev_class; /* Functions to control
56 this device. */
57
58 /* If this is 'true' the user did not specify a netdev_class when
59 * opening this device, and therefore got assigned to the "system" class */
60 bool auto_classified;
61
62 /* This bitmask of the offloading features enabled by the netdev. */
63 uint64_t ol_flags;
64
65 /* If this is 'true', the user explicitly specified an MTU for this
66 * netdev. Otherwise, Open vSwitch is allowed to override it. */
67 bool mtu_user_config;
68
69 int ref_cnt; /* Times this devices was opened. */
70
71 /* A sequence number which indicates changes in one of 'netdev''s
72 * properties. It must be nonzero so that users have a value which
73 * they may use as a reset when tracking 'netdev'.
74 *
75 * Minimally, the sequence number is required to change whenever
76 * 'netdev''s flags, features, ethernet address, or carrier changes. */
77 atomic_uint64_t change_seq;
78
79 /* A netdev provider might be unable to change some of the device's
80 * parameter (n_rxq, mtu) when the device is in use. In this case
81 * the provider can notify the upper layer by calling
82 * netdev_request_reconfigure(). The upper layer will react by stopping
83 * the operations on the device and calling netdev_reconfigure() to allow
84 * the configuration changes. 'last_reconfigure_seq' remembers the value
85 * of 'reconfigure_seq' when the last reconfiguration happened. */
86 struct seq *reconfigure_seq;
87 uint64_t last_reconfigure_seq;
88
89 /* The core netdev code initializes these at netdev construction and only
90 * provide read-only access to its client. Netdev implementations may
91 * modify them. */
92 int n_txq;
93 int n_rxq;
94 struct shash_node *node; /* Pointer to element in global map. */
95 struct ovs_list saved_flags_list; /* Contains "struct netdev_saved_flags". */
96
97 /* Functions to control flow offloading. */
98 OVSRCU_TYPE(const struct netdev_flow_api *) flow_api;
99 const char *dpif_type; /* Type of dpif this netdev belongs to. */
100 struct netdev_hw_info hw_info; /* Offload-capable netdev info. */
101 };
102
103 static inline void
104 netdev_change_seq_changed(const struct netdev *netdev_)
105 {
106 uint64_t change_seq;
107 struct netdev *netdev = CONST_CAST(struct netdev *, netdev_);
108 seq_change(connectivity_seq_get());
109
110 atomic_read_relaxed(&netdev->change_seq, &change_seq);
111 change_seq++;
112 if (OVS_UNLIKELY(!change_seq)) {
113 change_seq++;
114 }
115 atomic_store_explicit(&netdev->change_seq, change_seq,
116 memory_order_release);
117 }
118
119 static inline void
120 netdev_request_reconfigure(struct netdev *netdev)
121 {
122 seq_change(netdev->reconfigure_seq);
123 }
124
125 const char *netdev_get_type(const struct netdev *);
126 const struct netdev_class *netdev_get_class(const struct netdev *);
127 const char *netdev_get_name(const struct netdev *);
128 struct netdev *netdev_from_name(const char *name);
129 void netdev_get_devices(const struct netdev_class *,
130 struct shash *device_list);
131 struct netdev **netdev_get_vports(size_t *size);
132
133 /* A data structure for capturing packets received by a network device.
134 *
135 * Network device implementations may read these members but should not modify
136 * them.
137 *
138 * None of these members change during the lifetime of a struct netdev_rxq. */
139 struct netdev_rxq {
140 struct netdev *netdev; /* Owns a reference to the netdev. */
141 int queue_id;
142 };
143
144 struct netdev *netdev_rxq_get_netdev(const struct netdev_rxq *);
145
146
147 /* Network device class structure, to be defined by each implementation of a
148 * network device.
149 *
150 * These functions return 0 if successful or a positive errno value on failure,
151 * except where otherwise noted.
152 *
153 *
154 * Data Structures
155 * ===============
156 *
157 * These functions work primarily with two different kinds of data structures:
158 *
159 * - "struct netdev", which represents a network device.
160 *
161 * - "struct netdev_rxq", which represents a handle for capturing packets
162 * received on a network device
163 *
164 * Each of these data structures contains all of the implementation-independent
165 * generic state for the respective concept, called the "base" state. None of
166 * them contains any extra space for implementations to use. Instead, each
167 * implementation is expected to declare its own data structure that contains
168 * an instance of the generic data structure plus additional
169 * implementation-specific members, called the "derived" state. The
170 * implementation can use casts or (preferably) the CONTAINER_OF macro to
171 * obtain access to derived state given only a pointer to the embedded generic
172 * data structure.
173 *
174 *
175 * Life Cycle
176 * ==========
177 *
178 * Four stylized functions accompany each of these data structures:
179 *
180 * "alloc" "construct" "destruct" "dealloc"
181 * ------------ ---------------- --------------- --------------
182 * netdev ->alloc ->construct ->destruct ->dealloc
183 * netdev_rxq ->rxq_alloc ->rxq_construct ->rxq_destruct ->rxq_dealloc
184 *
185 * Any instance of a given data structure goes through the following life
186 * cycle:
187 *
188 * 1. The client calls the "alloc" function to obtain raw memory. If "alloc"
189 * fails, skip all the other steps.
190 *
191 * 2. The client initializes all of the data structure's base state. If this
192 * fails, skip to step 7.
193 *
194 * 3. The client calls the "construct" function. The implementation
195 * initializes derived state. It may refer to the already-initialized
196 * base state. If "construct" fails, skip to step 6.
197 *
198 * 4. The data structure is now initialized and in use.
199 *
200 * 5. When the data structure is no longer needed, the client calls the
201 * "destruct" function. The implementation uninitializes derived state.
202 * The base state has not been uninitialized yet, so the implementation
203 * may still refer to it.
204 *
205 * 6. The client uninitializes all of the data structure's base state.
206 *
207 * 7. The client calls the "dealloc" to free the raw memory. The
208 * implementation must not refer to base or derived state in the data
209 * structure, because it has already been uninitialized.
210 *
211 * If netdev support multi-queue IO then netdev->construct should set initialize
212 * netdev->n_rxq to number of queues.
213 *
214 * Each "alloc" function allocates and returns a new instance of the respective
215 * data structure. The "alloc" function is not given any information about the
216 * use of the new data structure, so it cannot perform much initialization.
217 * Its purpose is just to ensure that the new data structure has enough room
218 * for base and derived state. It may return a null pointer if memory is not
219 * available, in which case none of the other functions is called.
220 *
221 * Each "construct" function initializes derived state in its respective data
222 * structure. When "construct" is called, all of the base state has already
223 * been initialized, so the "construct" function may refer to it. The
224 * "construct" function is allowed to fail, in which case the client calls the
225 * "dealloc" function (but not the "destruct" function).
226 *
227 * Each "destruct" function uninitializes and frees derived state in its
228 * respective data structure. When "destruct" is called, the base state has
229 * not yet been uninitialized, so the "destruct" function may refer to it. The
230 * "destruct" function is not allowed to fail.
231 *
232 * Each "dealloc" function frees raw memory that was allocated by the
233 * "alloc" function. The memory's base and derived members might not have ever
234 * been initialized (but if "construct" returned successfully, then it has been
235 * "destruct"ed already). The "dealloc" function is not allowed to fail.
236 *
237 *
238 * Device Change Notification
239 * ==========================
240 *
241 * Minimally, implementations are required to report changes to netdev flags,
242 * features, ethernet address or carrier through connectivity_seq. Changes to
243 * other properties are allowed to cause notification through this interface,
244 * although implementations should try to avoid this. connectivity_seq_get()
245 * can be used to acquire a reference to the struct seq. The interface is
246 * described in detail in seq.h. */
247 struct netdev_class {
248 /* Type of netdevs in this class, e.g. "system", "tap", "gre", etc.
249 *
250 * One of the providers should supply a "system" type, since this is
251 * the type assumed if no type is specified when opening a netdev.
252 * The "system" type corresponds to an existing network device on
253 * the system. */
254 const char *type;
255
256 /* If 'true' then this netdev should be polled by PMD threads. */
257 bool is_pmd;
258
259 /* ## ------------------- ## */
260 /* ## Top-Level Functions ## */
261 /* ## ------------------- ## */
262
263 /* Called when the netdev provider is registered, typically at program
264 * startup. Returning an error from this function will prevent any network
265 * device in this class from being opened.
266 *
267 * This function may be set to null if a network device class needs no
268 * initialization at registration time. */
269 int (*init)(void);
270
271 /* Performs periodic work needed by netdevs of this class. May be null if
272 * no periodic work is necessary.
273 *
274 * 'netdev_class' points to the class. It is useful in case the same
275 * function is used to implement different classes. */
276 void (*run)(const struct netdev_class *netdev_class);
277
278 /* Arranges for poll_block() to wake up if the "run" member function needs
279 * to be called. Implementations are additionally required to wake
280 * whenever something changes in any of its netdevs which would cause their
281 * ->change_seq() function to change its result. May be null if nothing is
282 * needed here.
283 *
284 * 'netdev_class' points to the class. It is useful in case the same
285 * function is used to implement different classes. */
286 void (*wait)(const struct netdev_class *netdev_class);
287
288 /* ## ---------------- ## */
289 /* ## netdev Functions ## */
290 /* ## ---------------- ## */
291
292 /* Life-cycle functions for a netdev. See the large comment above on
293 * struct netdev_class. */
294 struct netdev *(*alloc)(void);
295 int (*construct)(struct netdev *);
296 void (*destruct)(struct netdev *);
297 void (*dealloc)(struct netdev *);
298
299 /* Fetches the device 'netdev''s configuration, storing it in 'args'.
300 * The caller owns 'args' and pre-initializes it to an empty smap.
301 *
302 * If this netdev class does not have any configuration options, this may
303 * be a null pointer. */
304 int (*get_config)(const struct netdev *netdev, struct smap *args);
305
306 /* Changes the device 'netdev''s configuration to 'args'.
307 *
308 * If this netdev class does not support configuration, this may be a null
309 * pointer.
310 *
311 * If the return value is not zero (meaning that an error occurred),
312 * the provider can allocate a string with an error message in '*errp'.
313 * The caller has to call free on it. */
314 int (*set_config)(struct netdev *netdev, const struct smap *args,
315 char **errp);
316
317 /* Returns the tunnel configuration of 'netdev'. If 'netdev' is
318 * not a tunnel, returns null.
319 *
320 * If this function would always return null, it may be null instead. */
321 const struct netdev_tunnel_config *
322 (*get_tunnel_config)(const struct netdev *netdev);
323
324 /* Build Tunnel header. Ethernet and ip header parameters are passed to
325 * tunnel implementation to build entire outer header for given flow. */
326 int (*build_header)(const struct netdev *, struct ovs_action_push_tnl *data,
327 const struct netdev_tnl_build_header_params *params);
328
329 /* build_header() can not build entire header for all packets for given
330 * flow. Push header is called for packet to build header specific to
331 * a packet on actual transmit. It uses partial header build by
332 * build_header() which is passed as data. */
333 void (*push_header)(const struct netdev *,
334 struct dp_packet *packet,
335 const struct ovs_action_push_tnl *data);
336
337 /* Pop tunnel header from packet, build tunnel metadata and resize packet
338 * for further processing.
339 * Returns NULL in case of error or tunnel implementation queued packet for further
340 * processing. */
341 struct dp_packet * (*pop_header)(struct dp_packet *packet);
342
343 /* Returns the id of the numa node the 'netdev' is on. If there is no
344 * such info, returns NETDEV_NUMA_UNSPEC. */
345 int (*get_numa_id)(const struct netdev *netdev);
346
347 /* Configures the number of tx queues of 'netdev'. Returns 0 if successful,
348 * otherwise a positive errno value.
349 *
350 * 'n_txq' specifies the exact number of transmission queues to create.
351 *
352 * The caller will call netdev_reconfigure() (if necessary) before using
353 * netdev_send() on any of the newly configured queues, giving the provider
354 * a chance to adjust its settings.
355 *
356 * On error, the tx queue configuration is unchanged. */
357 int (*set_tx_multiq)(struct netdev *netdev, unsigned int n_txq);
358
359 /* Sends buffers on 'netdev'.
360 * Returns 0 if successful (for every buffer), otherwise a positive errno
361 * value. Returns EAGAIN without blocking if one or more packets cannot be
362 * queued immediately. Returns EMSGSIZE if a partial packet was transmitted
363 * or if a packet is too big or too small to transmit on the device.
364 *
365 * If the function returns a non-zero value, some of the packets might have
366 * been sent anyway.
367 *
368 * The caller transfers ownership of all the packets to the network
369 * device, regardless of success.
370 *
371 * If 'concurrent_txq' is true, the caller may perform concurrent calls
372 * to netdev_send() with the same 'qid'. The netdev provider is responsible
373 * for making sure that these concurrent calls do not create a race
374 * condition by using locking or other synchronization if required.
375 *
376 * The network device is expected to maintain one or more packet
377 * transmission queues, so that the caller does not ordinarily have to
378 * do additional queuing of packets. 'qid' specifies the queue to use
379 * and can be ignored if the implementation does not support multiple
380 * queues.
381 *
382 * May return EOPNOTSUPP if a network device does not implement packet
383 * transmission through this interface. This function may be set to null
384 * if it would always return EOPNOTSUPP anyhow. (This will prevent the
385 * network device from being usefully used by the netdev-based "userspace
386 * datapath". It will also prevent the OVS implementation of bonding from
387 * working properly over 'netdev'.) */
388 int (*send)(struct netdev *netdev, int qid, struct dp_packet_batch *batch,
389 bool concurrent_txq);
390
391 /* Registers with the poll loop to wake up from the next call to
392 * poll_block() when the packet transmission queue for 'netdev' has
393 * sufficient room to transmit a packet with netdev_send().
394 *
395 * The network device is expected to maintain one or more packet
396 * transmission queues, so that the caller does not ordinarily have to
397 * do additional queuing of packets. 'qid' specifies the queue to use
398 * and can be ignored if the implementation does not support multiple
399 * queues.
400 *
401 * May be null if not needed, such as for a network device that does not
402 * implement packet transmission through the 'send' member function. */
403 void (*send_wait)(struct netdev *netdev, int qid);
404
405 /* Sets 'netdev''s Ethernet address to 'mac' */
406 int (*set_etheraddr)(struct netdev *netdev, const struct eth_addr mac);
407
408 /* Retrieves 'netdev''s Ethernet address into 'mac'.
409 *
410 * This address will be advertised as 'netdev''s MAC address through the
411 * OpenFlow protocol, among other uses. */
412 int (*get_etheraddr)(const struct netdev *netdev, struct eth_addr *mac);
413
414 /* Retrieves 'netdev''s MTU into '*mtup'.
415 *
416 * The MTU is the maximum size of transmitted (and received) packets, in
417 * bytes, not including the hardware header; thus, this is typically 1500
418 * bytes for Ethernet devices.
419 *
420 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
421 * this function should return EOPNOTSUPP. This function may be set to
422 * null if it would always return EOPNOTSUPP. */
423 int (*get_mtu)(const struct netdev *netdev, int *mtup);
424
425 /* Sets 'netdev''s MTU to 'mtu'.
426 *
427 * If 'netdev' does not have an MTU (e.g. as some tunnels do not), then
428 * this function should return EOPNOTSUPP. This function may be set to
429 * null if it would always return EOPNOTSUPP. */
430 int (*set_mtu)(struct netdev *netdev, int mtu);
431
432 /* Returns the ifindex of 'netdev', if successful, as a positive number.
433 * On failure, returns a negative errno value.
434 *
435 * The desired semantics of the ifindex value are a combination of those
436 * specified by POSIX for if_nametoindex() and by SNMP for ifIndex. An
437 * ifindex value should be unique within a host and remain stable at least
438 * until reboot. SNMP says an ifindex "ranges between 1 and the value of
439 * ifNumber" but many systems do not follow this rule anyhow.
440 *
441 * This function may be set to null if it would always return -EOPNOTSUPP.
442 */
443 int (*get_ifindex)(const struct netdev *netdev);
444
445 /* Sets 'carrier' to true if carrier is active (link light is on) on
446 * 'netdev'.
447 *
448 * May be null if device does not provide carrier status (will be always
449 * up as long as device is up).
450 */
451 int (*get_carrier)(const struct netdev *netdev, bool *carrier);
452
453 /* Returns the number of times 'netdev''s carrier has changed since being
454 * initialized.
455 *
456 * If null, callers will assume the number of carrier resets is zero. */
457 long long int (*get_carrier_resets)(const struct netdev *netdev);
458
459 /* Forces ->get_carrier() to poll 'netdev''s MII registers for link status
460 * instead of checking 'netdev''s carrier. 'netdev''s MII registers will
461 * be polled once every 'interval' milliseconds. If 'netdev' does not
462 * support MII, another method may be used as a fallback. If 'interval' is
463 * less than or equal to zero, reverts ->get_carrier() to its normal
464 * behavior.
465 *
466 * Most network devices won't support this feature and will set this
467 * function pointer to NULL, which is equivalent to returning EOPNOTSUPP.
468 */
469 int (*set_miimon_interval)(struct netdev *netdev, long long int interval);
470
471 /* Retrieves current device stats for 'netdev' into 'stats'.
472 *
473 * A network device that supports some statistics but not others, it should
474 * set the values of the unsupported statistics to all-1-bits
475 * (UINT64_MAX). */
476 int (*get_stats)(const struct netdev *netdev, struct netdev_stats *);
477
478 /* Retrieves current device custom stats for 'netdev' into 'custom_stats'.
479 *
480 * A network device should return only available statistics (if any).
481 * If there are not statistics available, empty array should be
482 * returned.
483 *
484 * The caller initializes 'custom_stats' before calling this function.
485 * The caller takes ownership over allocated array of counters inside
486 * structure netdev_custom_stats.
487 * */
488 int (*get_custom_stats)(const struct netdev *netdev,
489 struct netdev_custom_stats *custom_stats);
490
491 /* Stores the features supported by 'netdev' into each of '*current',
492 * '*advertised', '*supported', and '*peer'. Each value is a bitmap of
493 * NETDEV_F_* bits.
494 *
495 * This function may be set to null if it would always return EOPNOTSUPP.
496 */
497 int (*get_features)(const struct netdev *netdev,
498 enum netdev_features *current,
499 enum netdev_features *advertised,
500 enum netdev_features *supported,
501 enum netdev_features *peer);
502
503 /* Set the features advertised by 'netdev' to 'advertise', which is a
504 * set of NETDEV_F_* bits.
505 *
506 * This function may be set to null for a network device that does not
507 * support configuring advertisements. */
508 int (*set_advertisements)(struct netdev *netdev,
509 enum netdev_features advertise);
510
511 /* Returns 'netdev''s configured packet_type mode.
512 *
513 * This function may be set to null if it would always return
514 * NETDEV_PT_LEGACY_L2. */
515 enum netdev_pt_mode (*get_pt_mode)(const struct netdev *netdev);
516
517 /* Attempts to set input rate limiting (policing) policy, such that up to
518 * 'kbits_rate' kbps of traffic is accepted, with a maximum accumulative
519 * burst size of 'kbits' kb.
520 *
521 * This function may be set to null if policing is not supported. */
522 int (*set_policing)(struct netdev *netdev, unsigned int kbits_rate,
523 unsigned int kbits_burst);
524
525 /* Adds to 'types' all of the forms of QoS supported by 'netdev', or leaves
526 * it empty if 'netdev' does not support QoS. Any names added to 'types'
527 * should be documented as valid for the "type" column in the "QoS" table
528 * in vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
529 *
530 * Every network device must support disabling QoS with a type of "", but
531 * this function must not add "" to 'types'.
532 *
533 * The caller is responsible for initializing 'types' (e.g. with
534 * sset_init()) before calling this function. The caller retains ownership
535 * of 'types'.
536 *
537 * May be NULL if 'netdev' does not support QoS at all. */
538 int (*get_qos_types)(const struct netdev *netdev, struct sset *types);
539
540 /* Queries 'netdev' for its capabilities regarding the specified 'type' of
541 * QoS. On success, initializes 'caps' with the QoS capabilities.
542 *
543 * Should return EOPNOTSUPP if 'netdev' does not support 'type'. May be
544 * NULL if 'netdev' does not support QoS at all. */
545 int (*get_qos_capabilities)(const struct netdev *netdev,
546 const char *type,
547 struct netdev_qos_capabilities *caps);
548
549 /* Queries 'netdev' about its currently configured form of QoS. If
550 * successful, stores the name of the current form of QoS into '*typep'
551 * and any details of configuration as string key-value pairs in
552 * 'details'.
553 *
554 * A '*typep' of "" indicates that QoS is currently disabled on 'netdev'.
555 *
556 * The caller initializes 'details' before calling this function. The
557 * caller takes ownership of the string key-values pairs added to
558 * 'details'.
559 *
560 * The netdev retains ownership of '*typep'.
561 *
562 * '*typep' will be one of the types returned by netdev_get_qos_types() for
563 * 'netdev'. The contents of 'details' should be documented as valid for
564 * '*typep' in the "other_config" column in the "QoS" table in
565 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
566 *
567 * May be NULL if 'netdev' does not support QoS at all. */
568 int (*get_qos)(const struct netdev *netdev,
569 const char **typep, struct smap *details);
570
571 /* Attempts to reconfigure QoS on 'netdev', changing the form of QoS to
572 * 'type' with details of configuration from 'details'.
573 *
574 * On error, the previous QoS configuration is retained.
575 *
576 * When this function changes the type of QoS (not just 'details'), this
577 * also resets all queue configuration for 'netdev' to their defaults
578 * (which depend on the specific type of QoS). Otherwise, the queue
579 * configuration for 'netdev' is unchanged.
580 *
581 * 'type' should be "" (to disable QoS) or one of the types returned by
582 * netdev_get_qos_types() for 'netdev'. The contents of 'details' should
583 * be documented as valid for the given 'type' in the "other_config" column
584 * in the "QoS" table in vswitchd/vswitch.xml (which is built as
585 * ovs-vswitchd.conf.db(8)).
586 *
587 * May be NULL if 'netdev' does not support QoS at all. */
588 int (*set_qos)(struct netdev *netdev,
589 const char *type, const struct smap *details);
590
591 /* Queries 'netdev' for information about the queue numbered 'queue_id'.
592 * If successful, adds that information as string key-value pairs to
593 * 'details'. Returns 0 if successful, otherwise a positive errno value.
594 *
595 * Should return EINVAL if 'queue_id' is greater than or equal to the
596 * number of supported queues (as reported in the 'n_queues' member of
597 * struct netdev_qos_capabilities by 'get_qos_capabilities').
598 *
599 * The caller initializes 'details' before calling this function. The
600 * caller takes ownership of the string key-values pairs added to
601 * 'details'.
602 *
603 * The returned contents of 'details' should be documented as valid for the
604 * given 'type' in the "other_config" column in the "Queue" table in
605 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
606 */
607 int (*get_queue)(const struct netdev *netdev,
608 unsigned int queue_id, struct smap *details);
609
610 /* Configures the queue numbered 'queue_id' on 'netdev' with the key-value
611 * string pairs in 'details'. The contents of 'details' should be
612 * documented as valid for the given 'type' in the "other_config" column in
613 * the "Queue" table in vswitchd/vswitch.xml (which is built as
614 * ovs-vswitchd.conf.db(8)). Returns 0 if successful, otherwise a positive
615 * errno value. On failure, the given queue's configuration should be
616 * unmodified.
617 *
618 * Should return EINVAL if 'queue_id' is greater than or equal to the
619 * number of supported queues (as reported in the 'n_queues' member of
620 * struct netdev_qos_capabilities by 'get_qos_capabilities'), or if
621 * 'details' is invalid for the type of queue.
622 *
623 * This function does not modify 'details', and the caller retains
624 * ownership of it.
625 *
626 * May be NULL if 'netdev' does not support QoS at all. */
627 int (*set_queue)(struct netdev *netdev,
628 unsigned int queue_id, const struct smap *details);
629
630 /* Attempts to delete the queue numbered 'queue_id' from 'netdev'.
631 *
632 * Should return EINVAL if 'queue_id' is greater than or equal to the
633 * number of supported queues (as reported in the 'n_queues' member of
634 * struct netdev_qos_capabilities by 'get_qos_capabilities'). Should
635 * return EOPNOTSUPP if 'queue_id' is valid but may not be deleted (e.g. if
636 * 'netdev' has a fixed set of queues with the current QoS mode).
637 *
638 * May be NULL if 'netdev' does not support QoS at all, or if all of its
639 * QoS modes have fixed sets of queues. */
640 int (*delete_queue)(struct netdev *netdev, unsigned int queue_id);
641
642 /* Obtains statistics about 'queue_id' on 'netdev'. Fills 'stats' with the
643 * queue's statistics. May set individual members of 'stats' to all-1-bits
644 * if the statistic is unavailable.
645 *
646 * May be NULL if 'netdev' does not support QoS at all. */
647 int (*get_queue_stats)(const struct netdev *netdev, unsigned int queue_id,
648 struct netdev_queue_stats *stats);
649
650 /* Attempts to begin dumping the queues in 'netdev'. On success, returns 0
651 * and initializes '*statep' with any data needed for iteration. On
652 * failure, returns a positive errno value.
653 *
654 * May be NULL if 'netdev' does not support QoS at all. */
655 int (*queue_dump_start)(const struct netdev *netdev, void **statep);
656
657 /* Attempts to retrieve another queue from 'netdev' for 'state', which was
658 * initialized by a successful call to the 'queue_dump_start' function for
659 * 'netdev'. On success, stores a queue ID into '*queue_id' and fills
660 * 'details' with the configuration of the queue with that ID. Returns EOF
661 * if the last queue has been dumped, or a positive errno value on error.
662 * This function will not be called again once it returns nonzero once for
663 * a given iteration (but the 'queue_dump_done' function will be called
664 * afterward).
665 *
666 * The caller initializes and clears 'details' before calling this
667 * function. The caller takes ownership of the string key-values pairs
668 * added to 'details'.
669 *
670 * The returned contents of 'details' should be documented as valid for the
671 * given 'type' in the "other_config" column in the "Queue" table in
672 * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
673 *
674 * May be NULL if 'netdev' does not support QoS at all. */
675 int (*queue_dump_next)(const struct netdev *netdev, void *state,
676 unsigned int *queue_id, struct smap *details);
677
678 /* Releases resources from 'netdev' for 'state', which was initialized by a
679 * successful call to the 'queue_dump_start' function for 'netdev'.
680 *
681 * May be NULL if 'netdev' does not support QoS at all. */
682 int (*queue_dump_done)(const struct netdev *netdev, void *state);
683
684 /* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's
685 * ID, its statistics, and the 'aux' specified by the caller. The order of
686 * iteration is unspecified, but (when successful) each queue must be
687 * visited exactly once.
688 *
689 * 'cb' will not modify or free the statistics passed in. */
690 int (*dump_queue_stats)(const struct netdev *netdev,
691 void (*cb)(unsigned int queue_id,
692 struct netdev_queue_stats *,
693 void *aux),
694 void *aux);
695
696 /* Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
697 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared.
698 *
699 * This function may be set to null if it would always return EOPNOTSUPP
700 * anyhow. */
701 int (*set_in4)(struct netdev *netdev, struct in_addr addr,
702 struct in_addr mask);
703
704 /* Returns all assigned IP address to 'netdev' and returns 0.
705 * API allocates array of address and masks and set it to
706 * '*addr' and '*mask'.
707 * Otherwise, returns a positive errno value and sets '*addr', '*mask
708 * and '*n_addr' to NULL.
709 *
710 * The following error values have well-defined meanings:
711 *
712 * - EADDRNOTAVAIL: 'netdev' has no assigned IPv6 address.
713 *
714 * - EOPNOTSUPP: No IPv6 network stack attached to 'netdev'.
715 *
716 * 'addr' may be null, in which case the address itself is not reported. */
717 int (*get_addr_list)(const struct netdev *netdev, struct in6_addr **in,
718 struct in6_addr **mask, int *n_in6);
719
720 /* Adds 'router' as a default IP gateway for the TCP/IP stack that
721 * corresponds to 'netdev'.
722 *
723 * This function may be set to null if it would always return EOPNOTSUPP
724 * anyhow. */
725 int (*add_router)(struct netdev *netdev, struct in_addr router);
726
727 /* Looks up the next hop for 'host' in the host's routing table. If
728 * successful, stores the next hop gateway's address (0 if 'host' is on a
729 * directly connected network) in '*next_hop' and a copy of the name of the
730 * device to reach 'host' in '*netdev_name', and returns 0. The caller is
731 * responsible for freeing '*netdev_name' (by calling free()).
732 *
733 * This function may be set to null if it would always return EOPNOTSUPP
734 * anyhow. */
735 int (*get_next_hop)(const struct in_addr *host, struct in_addr *next_hop,
736 char **netdev_name);
737
738 /* Retrieves driver information of the device.
739 *
740 * Populates 'smap' with key-value pairs representing the status of the
741 * device. 'smap' is a set of key-value string pairs representing netdev
742 * type specific information. For more information see
743 * ovs-vswitchd.conf.db(5).
744 *
745 * The caller is responsible for destroying 'smap' and its data.
746 *
747 * This function may be set to null if it would always return EOPNOTSUPP
748 * anyhow. */
749 int (*get_status)(const struct netdev *netdev, struct smap *smap);
750
751 /* Looks up the ARP table entry for 'ip' on 'netdev' and stores the
752 * corresponding MAC address in 'mac'. A return value of ENXIO, in
753 * particular, indicates that there is no ARP table entry for 'ip' on
754 * 'netdev'.
755 *
756 * This function may be set to null if it would always return EOPNOTSUPP
757 * anyhow. */
758 int (*arp_lookup)(const struct netdev *netdev, ovs_be32 ip,
759 struct eth_addr *mac);
760
761 /* Retrieves the current set of flags on 'netdev' into '*old_flags'. Then,
762 * turns off the flags that are set to 1 in 'off' and turns on the flags
763 * that are set to 1 in 'on'. (No bit will be set to 1 in both 'off' and
764 * 'on'; that is, off & on == 0.)
765 *
766 * This function may be invoked from a signal handler. Therefore, it
767 * should not do anything that is not signal-safe (such as logging). */
768 int (*update_flags)(struct netdev *netdev, enum netdev_flags off,
769 enum netdev_flags on, enum netdev_flags *old_flags);
770
771 /* If the provider called netdev_request_reconfigure(), the upper layer
772 * will eventually call this. The provider can update the device
773 * configuration knowing that the upper layer will not call rxq_recv() or
774 * send() until this function returns.
775 *
776 * On error, the configuration is indeterminant and the device cannot be
777 * used to send and receive packets until a successful configuration is
778 * applied. */
779 int (*reconfigure)(struct netdev *netdev);
780 /* ## -------------------- ## */
781 /* ## netdev_rxq Functions ## */
782 /* ## -------------------- ## */
783
784 /* If a particular netdev class does not support receiving packets, all these
785 * function pointers must be NULL. */
786
787 /* Life-cycle functions for a netdev_rxq. See the large comment above on
788 * struct netdev_class. */
789 struct netdev_rxq *(*rxq_alloc)(void);
790 int (*rxq_construct)(struct netdev_rxq *);
791 void (*rxq_destruct)(struct netdev_rxq *);
792 void (*rxq_dealloc)(struct netdev_rxq *);
793
794 /* Retrieves the current state of rx queue. 'false' means that queue won't
795 * get traffic in a short term and could be not polled.
796 *
797 * This function may be set to null if it would always return 'true'
798 * anyhow. */
799 bool (*rxq_enabled)(struct netdev_rxq *);
800
801 /* Attempts to receive a batch of packets from 'rx'. In 'batch', the
802 * caller supplies 'packets' as the pointer to the beginning of an array
803 * of NETDEV_MAX_BURST pointers to dp_packet. If successful, the
804 * implementation stores pointers to up to NETDEV_MAX_BURST dp_packets into
805 * the array, transferring ownership of the packets to the caller, stores
806 * the number of received packets into 'count', and returns 0.
807 *
808 * The implementation does not necessarily initialize any non-data members
809 * of 'packets' in 'batch'. That is, the caller must initialize layer
810 * pointers and metadata itself, if desired, e.g. with pkt_metadata_init()
811 * and miniflow_extract().
812 *
813 * Implementations should allocate buffers with DP_NETDEV_HEADROOM bytes of
814 * headroom.
815 *
816 * If the caller provides a non-NULL qfill pointer, the implementation
817 * should return the number (zero or more) of remaining packets in the
818 * queue after the reception the current batch, if it supports that,
819 * or -ENOTSUP otherwise.
820 *
821 * Returns EAGAIN immediately if no packet is ready to be received or
822 * another positive errno value if an error was encountered. */
823 int (*rxq_recv)(struct netdev_rxq *rx, struct dp_packet_batch *batch,
824 int *qfill);
825
826 /* Registers with the poll loop to wake up from the next call to
827 * poll_block() when a packet is ready to be received with
828 * netdev_rxq_recv() on 'rx'. */
829 void (*rxq_wait)(struct netdev_rxq *rx);
830
831 /* Discards all packets waiting to be received from 'rx'. */
832 int (*rxq_drain)(struct netdev_rxq *rx);
833
834 /* Get a block_id from the netdev.
835 * Returns the block_id or 0 if none exists for netdev. */
836 uint32_t (*get_block_id)(struct netdev *);
837 };
838
839 int netdev_register_provider(const struct netdev_class *);
840 int netdev_unregister_provider(const char *type);
841
842 #if defined(__FreeBSD__) || defined(__NetBSD__)
843 extern const struct netdev_class netdev_bsd_class;
844 #elif defined(_WIN32)
845 extern const struct netdev_class netdev_windows_class;
846 #else
847 extern const struct netdev_class netdev_linux_class;
848 #endif
849 extern const struct netdev_class netdev_internal_class;
850 extern const struct netdev_class netdev_tap_class;
851
852 #ifdef HAVE_AF_XDP
853 extern const struct netdev_class netdev_afxdp_class;
854 extern const struct netdev_class netdev_afxdp_nonpmd_class;
855 #endif
856 #ifdef __cplusplus
857 }
858 #endif
859
860 #endif /* netdev.h */