1/******************************************************************************
2 * xen_netif.h
3 *
4 * Unified network-device I/O interface for Xen guest OSes.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to
8 * deal in the Software without restriction, including without limitation the
9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10 * sell copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22 * DEALINGS IN THE SOFTWARE.
23 *
24 * Copyright (c) 2003-2004, Keir Fraser
25 */
26
27#ifndef __XEN_PUBLIC_IO_XEN_NETIF_H__
28#define __XEN_PUBLIC_IO_XEN_NETIF_H__
29
30#include "ring.h"
31#include "../grant_table.h"
32
33/*
34 * Older implementation of Xen network frontend / backend has an
35 * implicit dependency on the MAX_SKB_FRAGS as the maximum number of
36 * ring slots a skb can use. Netfront / netback may not work as
37 * expected when frontend and backend have different MAX_SKB_FRAGS.
38 *
39 * A better approach is to add mechanism for netfront / netback to
40 * negotiate this value. However we cannot fix all possible
41 * frontends, so we need to define a value which states the minimum
42 * slots backend must support.
43 *
44 * The minimum value derives from older Linux kernel's MAX_SKB_FRAGS
45 * (18), which is proved to work with most frontends. Any new backend
46 * which doesn't negotiate with frontend should expect frontend to
47 * send a valid packet using slots up to this value.
48 */
49#define XEN_NETIF_NR_SLOTS_MIN 18
50
51/*
52 * Notifications after enqueuing any type of message should be conditional on
53 * the appropriate req_event or rsp_event field in the shared ring.
54 * If the client sends notification for rx requests then it should specify
55 * feature 'feature-rx-notify' via xenbus. Otherwise the backend will assume
56 * that it cannot safely queue packets (as it may not be kicked to send them).
57 */
58
59/*
60 * "feature-split-event-channels" is introduced to separate guest TX
61 * and RX notification. Backend either doesn't support this feature or
62 * advertises it via xenstore as 0 (disabled) or 1 (enabled).
63 *
64 * To make use of this feature, frontend should allocate two event
65 * channels for TX and RX, advertise them to backend as
66 * "event-channel-tx" and "event-channel-rx" respectively. If frontend
67 * doesn't want to use this feature, it just writes "event-channel"
68 * node as before.
69 */
70
71/*
72 * Multiple transmit and receive queues:
73 * If supported, the backend will write the key "multi-queue-max-queues" to
74 * the directory for that vif, and set its value to the maximum supported
75 * number of queues.
76 * Frontends that are aware of this feature and wish to use it can write the
77 * key "multi-queue-num-queues", set to the number they wish to use, which
78 * must be greater than zero, and no more than the value reported by the backend
79 * in "multi-queue-max-queues".
80 *
81 * Queues replicate the shared rings and event channels.
82 * "feature-split-event-channels" may optionally be used when using
83 * multiple queues, but is not mandatory.
84 *
85 * Each queue consists of one shared ring pair, i.e. there must be the same
86 * number of tx and rx rings.
87 *
88 * For frontends requesting just one queue, the usual event-channel and
89 * ring-ref keys are written as before, simplifying the backend processing
90 * to avoid distinguishing between a frontend that doesn't understand the
91 * multi-queue feature, and one that does, but requested only one queue.
92 *
93 * Frontends requesting two or more queues must not write the toplevel
94 * event-channel (or event-channel-{tx,rx}) and {tx,rx}-ring-ref keys,
95 * instead writing those keys under sub-keys having the name "queue-N" where
96 * N is the integer ID of the queue for which those keys belong. Queues
97 * are indexed from zero. For example, a frontend with two queues and split
98 * event channels must write the following set of queue-related keys:
99 *
100 * /local/domain/1/device/vif/0/multi-queue-num-queues = "2"
101 * /local/domain/1/device/vif/0/queue-0 = ""
102 * /local/domain/1/device/vif/0/queue-0/tx-ring-ref = "<ring-ref-tx0>"
103 * /local/domain/1/device/vif/0/queue-0/rx-ring-ref = "<ring-ref-rx0>"
104 * /local/domain/1/device/vif/0/queue-0/event-channel-tx = "<evtchn-tx0>"
105 * /local/domain/1/device/vif/0/queue-0/event-channel-rx = "<evtchn-rx0>"
106 * /local/domain/1/device/vif/0/queue-1 = ""
107 * /local/domain/1/device/vif/0/queue-1/tx-ring-ref = "<ring-ref-tx1>"
108 * /local/domain/1/device/vif/0/queue-1/rx-ring-ref = "<ring-ref-rx1"
109 * /local/domain/1/device/vif/0/queue-1/event-channel-tx = "<evtchn-tx1>"
110 * /local/domain/1/device/vif/0/queue-1/event-channel-rx = "<evtchn-rx1>"
111 *
112 * If there is any inconsistency in the XenStore data, the backend may
113 * choose not to connect any queues, instead treating the request as an
114 * error. This includes scenarios where more (or fewer) queues were
115 * requested than the frontend provided details for.
116 *
117 * Mapping of packets to queues is considered to be a function of the
118 * transmitting system (backend or frontend) and is not negotiated
119 * between the two. Guests are free to transmit packets on any queue
120 * they choose, provided it has been set up correctly. Guests must be
121 * prepared to receive packets on any queue they have requested be set up.
122 */
123
124/*
125 * "feature-no-csum-offload" should be used to turn IPv4 TCP/UDP checksum
126 * offload off or on. If it is missing then the feature is assumed to be on.
127 * "feature-ipv6-csum-offload" should be used to turn IPv6 TCP/UDP checksum
128 * offload on or off. If it is missing then the feature is assumed to be off.
129 */
130
131/*
132 * "feature-gso-tcpv4" and "feature-gso-tcpv6" advertise the capability to
133 * handle large TCP packets (in IPv4 or IPv6 form respectively). Neither
134 * frontends nor backends are assumed to be capable unless the flags are
135 * present.
136 */
137
138/*
139 * "feature-multicast-control" and "feature-dynamic-multicast-control"
140 * advertise the capability to filter ethernet multicast packets in the
141 * backend. If the frontend wishes to take advantage of this feature then
142 * it may set "request-multicast-control". If the backend only advertises
143 * "feature-multicast-control" then "request-multicast-control" must be set
144 * before the frontend moves into the connected state. The backend will
145 * sample the value on this state transition and any subsequent change in
146 * value will have no effect. However, if the backend also advertises
147 * "feature-dynamic-multicast-control" then "request-multicast-control"
148 * may be set by the frontend at any time. In this case, the backend will
149 * watch the value and re-sample on watch events.
150 *
151 * If the sampled value of "request-multicast-control" is set then the
152 * backend transmit side should no longer flood multicast packets to the
153 * frontend, it should instead drop any multicast packet that does not
154 * match in a filter list.
155 * The list is amended by the frontend by sending dummy transmit requests
156 * containing XEN_NETIF_EXTRA_TYPE_MCAST_{ADD,DEL} extra-info fragments as
157 * specified below.
158 * Note that the filter list may be amended even if the sampled value of
159 * "request-multicast-control" is not set, however the filter should only
160 * be applied if it is set.
161 */
162
163/*
164 * Control ring
165 * ============
166 *
167 * Some features, such as hashing (detailed below), require a
168 * significant amount of out-of-band data to be passed from frontend to
169 * backend. Use of xenstore is not suitable for large quantities of data
170 * because of quota limitations and so a dedicated 'control ring' is used.
171 * The ability of the backend to use a control ring is advertised by
172 * setting:
173 *
174 * /local/domain/X/backend/<domid>/<vif>/feature-ctrl-ring = "1"
175 *
176 * The frontend provides a control ring to the backend by setting:
177 *
178 * /local/domain/<domid>/device/vif/<vif>/ctrl-ring-ref = <gref>
179 * /local/domain/<domid>/device/vif/<vif>/event-channel-ctrl = <port>
180 *
181 * where <gref> is the grant reference of the shared page used to
182 * implement the control ring and <port> is an event channel to be used
183 * as a mailbox interrupt. These keys must be set before the frontend
184 * moves into the connected state.
185 *
186 * The control ring uses a fixed request/response message size and is
187 * balanced (i.e. one request to one response), so operationally it is much
188 * the same as a transmit or receive ring.
189 * Note that there is no requirement that responses are issued in the same
190 * order as requests.
191 */
192
193/*
194 * Hash types
195 * ==========
196 *
197 * For the purposes of the definitions below, 'Packet[]' is an array of
198 * octets containing an IP packet without options, 'Array[X..Y]' means a
199 * sub-array of 'Array' containing bytes X thru Y inclusive, and '+' is
200 * used to indicate concatenation of arrays.
201 */
202
203/*
204 * A hash calculated over an IP version 4 header as follows:
205 *
206 * Buffer[0..8] = Packet[12..15] (source address) +
207 * Packet[16..19] (destination address)
208 *
209 * Result = Hash(Buffer, 8)
210 */
211#define _XEN_NETIF_CTRL_HASH_TYPE_IPV4 0
212#define XEN_NETIF_CTRL_HASH_TYPE_IPV4 \
213 (1 << _XEN_NETIF_CTRL_HASH_TYPE_IPV4)
214
215/*
216 * A hash calculated over an IP version 4 header and TCP header as
217 * follows:
218 *
219 * Buffer[0..12] = Packet[12..15] (source address) +
220 * Packet[16..19] (destination address) +
221 * Packet[20..21] (source port) +
222 * Packet[22..23] (destination port)
223 *
224 * Result = Hash(Buffer, 12)
225 */
226#define _XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP 1
227#define XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP \
228 (1 << _XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP)
229
230/*
231 * A hash calculated over an IP version 6 header as follows:
232 *
233 * Buffer[0..32] = Packet[8..23] (source address ) +
234 * Packet[24..39] (destination address)
235 *
236 * Result = Hash(Buffer, 32)
237 */
238#define _XEN_NETIF_CTRL_HASH_TYPE_IPV6 2
239#define XEN_NETIF_CTRL_HASH_TYPE_IPV6 \
240 (1 << _XEN_NETIF_CTRL_HASH_TYPE_IPV6)
241
242/*
243 * A hash calculated over an IP version 6 header and TCP header as
244 * follows:
245 *
246 * Buffer[0..36] = Packet[8..23] (source address) +
247 * Packet[24..39] (destination address) +
248 * Packet[40..41] (source port) +
249 * Packet[42..43] (destination port)
250 *
251 * Result = Hash(Buffer, 36)
252 */
253#define _XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP 3
254#define XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP \
255 (1 << _XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP)
256
257/*
258 * Hash algorithms
259 * ===============
260 */
261
262#define XEN_NETIF_CTRL_HASH_ALGORITHM_NONE 0
263
264/*
265 * Toeplitz hash:
266 */
267
268#define XEN_NETIF_CTRL_HASH_ALGORITHM_TOEPLITZ 1
269
270/*
271 * This algorithm uses a 'key' as well as the data buffer itself.
272 * (Buffer[] and Key[] are treated as shift-registers where the MSB of
273 * Buffer/Key[0] is considered 'left-most' and the LSB of Buffer/Key[N-1]
274 * is the 'right-most').
275 *
276 * Value = 0
277 * For number of bits in Buffer[]
278 * If (left-most bit of Buffer[] is 1)
279 * Value ^= left-most 32 bits of Key[]
280 * Key[] << 1
281 * Buffer[] << 1
282 *
283 * The code below is provided for convenience where an operating system
284 * does not already provide an implementation.
285 */
286#ifdef XEN_NETIF_DEFINE_TOEPLITZ
287static uint32_t xen_netif_toeplitz_hash(const uint8_t *key,
288 unsigned int keylen,
289 const uint8_t *buf, unsigned int buflen)
290{
291 unsigned int keyi, bufi;
292 uint64_t prefix = 0;
293 uint64_t hash = 0;
294
295 /* Pre-load prefix with the first 8 bytes of the key */
296 for (keyi = 0; keyi < 8; keyi++) {
297 prefix <<= 8;
298 prefix |= (keyi < keylen) ? key[keyi] : 0;
299 }
300
301 for (bufi = 0; bufi < buflen; bufi++) {
302 uint8_t byte = buf[bufi];
303 unsigned int bit;
304
305 for (bit = 0; bit < 8; bit++) {
306 if (byte & 0x80)
307 hash ^= prefix;
308 prefix <<= 1;
309 byte <<= 1;
310 }
311
312 /*
313 * 'prefix' has now been left-shifted by 8, so
314 * OR in the next byte.
315 */
316 prefix |= (keyi < keylen) ? key[keyi] : 0;
317 keyi++;
318 }
319
320 /* The valid part of the hash is in the upper 32 bits. */
321 return hash >> 32;
322}
323#endif /* XEN_NETIF_DEFINE_TOEPLITZ */
324
325/*
326 * Control requests (struct xen_netif_ctrl_request)
327 * ================================================
328 *
329 * All requests have the following format:
330 *
331 * 0 1 2 3 4 5 6 7 octet
332 * +-----+-----+-----+-----+-----+-----+-----+-----+
333 * | id | type | data[0] |
334 * +-----+-----+-----+-----+-----+-----+-----+-----+
335 * | data[1] | data[2] |
336 * +-----+-----+-----+-----+-----------------------+
337 *
338 * id: the request identifier, echoed in response.
339 * type: the type of request (see below)
340 * data[]: any data associated with the request (determined by type)
341 */
342
343struct xen_netif_ctrl_request {
344 uint16_t id;
345 uint16_t type;
346
347#define XEN_NETIF_CTRL_TYPE_INVALID 0
348#define XEN_NETIF_CTRL_TYPE_GET_HASH_FLAGS 1
349#define XEN_NETIF_CTRL_TYPE_SET_HASH_FLAGS 2
350#define XEN_NETIF_CTRL_TYPE_SET_HASH_KEY 3
351#define XEN_NETIF_CTRL_TYPE_GET_HASH_MAPPING_SIZE 4
352#define XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE 5
353#define XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING 6
354#define XEN_NETIF_CTRL_TYPE_SET_HASH_ALGORITHM 7
355
356 uint32_t data[3];
357};
358
359/*
360 * Control responses (struct xen_netif_ctrl_response)
361 * ==================================================
362 *
363 * All responses have the following format:
364 *
365 * 0 1 2 3 4 5 6 7 octet
366 * +-----+-----+-----+-----+-----+-----+-----+-----+
367 * | id | type | status |
368 * +-----+-----+-----+-----+-----+-----+-----+-----+
369 * | data |
370 * +-----+-----+-----+-----+
371 *
372 * id: the corresponding request identifier
373 * type: the type of the corresponding request
374 * status: the status of request processing
375 * data: any data associated with the response (determined by type and
376 * status)
377 */
378
379struct xen_netif_ctrl_response {
380 uint16_t id;
381 uint16_t type;
382 uint32_t status;
383
384#define XEN_NETIF_CTRL_STATUS_SUCCESS 0
385#define XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED 1
386#define XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER 2
387#define XEN_NETIF_CTRL_STATUS_BUFFER_OVERFLOW 3
388
389 uint32_t data;
390};
391
392/*
393 * Control messages
394 * ================
395 *
396 * XEN_NETIF_CTRL_TYPE_SET_HASH_ALGORITHM
397 * --------------------------------------
398 *
399 * This is sent by the frontend to set the desired hash algorithm.
400 *
401 * Request:
402 *
403 * type = XEN_NETIF_CTRL_TYPE_SET_HASH_ALGORITHM
404 * data[0] = a XEN_NETIF_CTRL_HASH_ALGORITHM_* value
405 * data[1] = 0
406 * data[2] = 0
407 *
408 * Response:
409 *
410 * status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED - Operation not
411 * supported
412 * XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - The algorithm is not
413 * supported
414 * XEN_NETIF_CTRL_STATUS_SUCCESS - Operation successful
415 *
416 * NOTE: Setting data[0] to XEN_NETIF_CTRL_HASH_ALGORITHM_NONE disables
417 * hashing and the backend is free to choose how it steers packets
418 * to queues (which is the default behaviour).
419 *
420 * XEN_NETIF_CTRL_TYPE_GET_HASH_FLAGS
421 * ----------------------------------
422 *
423 * This is sent by the frontend to query the types of hash supported by
424 * the backend.
425 *
426 * Request:
427 *
428 * type = XEN_NETIF_CTRL_TYPE_GET_HASH_FLAGS
429 * data[0] = 0
430 * data[1] = 0
431 * data[2] = 0
432 *
433 * Response:
434 *
435 * status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED - Operation not supported
436 * XEN_NETIF_CTRL_STATUS_SUCCESS - Operation successful
437 * data = supported hash types (if operation was successful)
438 *
439 * NOTE: A valid hash algorithm must be selected before this operation can
440 * succeed.
441 *
442 * XEN_NETIF_CTRL_TYPE_SET_HASH_FLAGS
443 * ----------------------------------
444 *
445 * This is sent by the frontend to set the types of hash that the backend
446 * should calculate. (See above for hash type definitions).
447 * Note that the 'maximal' type of hash should always be chosen. For
448 * example, if the frontend sets both IPV4 and IPV4_TCP hash types then
449 * the latter hash type should be calculated for any TCP packet and the
450 * former only calculated for non-TCP packets.
451 *
452 * Request:
453 *
454 * type = XEN_NETIF_CTRL_TYPE_SET_HASH_FLAGS
455 * data[0] = bitwise OR of XEN_NETIF_CTRL_HASH_TYPE_* values
456 * data[1] = 0
457 * data[2] = 0
458 *
459 * Response:
460 *
461 * status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED - Operation not
462 * supported
463 * XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - One or more flag
464 * value is invalid or
465 * unsupported
466 * XEN_NETIF_CTRL_STATUS_SUCCESS - Operation successful
467 * data = 0
468 *
469 * NOTE: A valid hash algorithm must be selected before this operation can
470 * succeed.
471 * Also, setting data[0] to zero disables hashing and the backend
472 * is free to choose how it steers packets to queues.
473 *
474 * XEN_NETIF_CTRL_TYPE_SET_HASH_KEY
475 * --------------------------------
476 *
477 * This is sent by the frontend to set the key of the hash if the algorithm
478 * requires it. (See hash algorithms above).
479 *
480 * Request:
481 *
482 * type = XEN_NETIF_CTRL_TYPE_SET_HASH_KEY
483 * data[0] = grant reference of page containing the key (assumed to
484 * start at beginning of grant)
485 * data[1] = size of key in octets
486 * data[2] = 0
487 *
488 * Response:
489 *
490 * status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED - Operation not
491 * supported
492 * XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - Key size is invalid
493 * XEN_NETIF_CTRL_STATUS_BUFFER_OVERFLOW - Key size is larger
494 * than the backend
495 * supports
496 * XEN_NETIF_CTRL_STATUS_SUCCESS - Operation successful
497 * data = 0
498 *
499 * NOTE: Any key octets not specified are assumed to be zero (the key
500 * is assumed to be empty by default) and specifying a new key
501 * invalidates any previous key, hence specifying a key size of
502 * zero will clear the key (which ensures that the calculated hash
503 * will always be zero).
504 * The maximum size of key is algorithm and backend specific, but
505 * is also limited by the single grant reference.
506 * The grant reference may be read-only and must remain valid until
507 * the response has been processed.
508 *
509 * XEN_NETIF_CTRL_TYPE_GET_HASH_MAPPING_SIZE
510 * -----------------------------------------
511 *
512 * This is sent by the frontend to query the maximum size of mapping
513 * table supported by the backend. The size is specified in terms of
514 * table entries.
515 *
516 * Request:
517 *
518 * type = XEN_NETIF_CTRL_TYPE_GET_HASH_MAPPING_SIZE
519 * data[0] = 0
520 * data[1] = 0
521 * data[2] = 0
522 *
523 * Response:
524 *
525 * status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED - Operation not supported
526 * XEN_NETIF_CTRL_STATUS_SUCCESS - Operation successful
527 * data = maximum number of entries allowed in the mapping table
528 * (if operation was successful) or zero if a mapping table is
529 * not supported (i.e. hash mapping is done only by modular
530 * arithmetic).
531 *
532 * XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE
533 * -------------------------------------
534 *
535 * This is sent by the frontend to set the actual size of the mapping
536 * table to be used by the backend. The size is specified in terms of
537 * table entries.
538 * Any previous table is invalidated by this message and any new table
539 * is assumed to be zero filled.
540 *
541 * Request:
542 *
543 * type = XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE
544 * data[0] = number of entries in mapping table
545 * data[1] = 0
546 * data[2] = 0
547 *
548 * Response:
549 *
550 * status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED - Operation not
551 * supported
552 * XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - Table size is invalid
553 * XEN_NETIF_CTRL_STATUS_SUCCESS - Operation successful
554 * data = 0
555 *
556 * NOTE: Setting data[0] to 0 means that hash mapping should be done
557 * using modular arithmetic.
558 *
559 * XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING
560 * ------------------------------------
561 *
562 * This is sent by the frontend to set the content of the table mapping
563 * hash value to queue number. The backend should calculate the hash from
564 * the packet header, use it as an index into the table (modulo the size
565 * of the table) and then steer the packet to the queue number found at
566 * that index.
567 *
568 * Request:
569 *
570 * type = XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING
571 * data[0] = grant reference of page containing the mapping (sub-)table
572 * (assumed to start at beginning of grant)
573 * data[1] = size of (sub-)table in entries
574 * data[2] = offset, in entries, of sub-table within overall table
575 *
576 * Response:
577 *
578 * status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED - Operation not
579 * supported
580 * XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - Table size or content
581 * is invalid
582 * XEN_NETIF_CTRL_STATUS_BUFFER_OVERFLOW - Table size is larger
583 * than the backend
584 * supports
585 * XEN_NETIF_CTRL_STATUS_SUCCESS - Operation successful
586 * data = 0
587 *
588 * NOTE: The overall table has the following format:
589 *
590 * 0 1 2 3 4 5 6 7 octet
591 * +-----+-----+-----+-----+-----+-----+-----+-----+
592 * | mapping[0] | mapping[1] |
593 * +-----+-----+-----+-----+-----+-----+-----+-----+
594 * | . |
595 * | . |
596 * | . |
597 * +-----+-----+-----+-----+-----+-----+-----+-----+
598 * | mapping[N-2] | mapping[N-1] |
599 * +-----+-----+-----+-----+-----+-----+-----+-----+
600 *
601 * where N is specified by a XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE
602 * message and each mapping must specifies a queue between 0 and
603 * "multi-queue-num-queues" (see above).
604 * The backend may support a mapping table larger than can be
605 * mapped by a single grant reference. Thus sub-tables within a
606 * larger table can be individually set by sending multiple messages
607 * with differing offset values. Specifying a new sub-table does not
608 * invalidate any table data outside that range.
609 * The grant reference may be read-only and must remain valid until
610 * the response has been processed.
611 */
612
613DEFINE_RING_TYPES(xen_netif_ctrl,
614 struct xen_netif_ctrl_request,
615 struct xen_netif_ctrl_response);
616
617/*
618 * Guest transmit
619 * ==============
620 *
621 * This is the 'wire' format for transmit (frontend -> backend) packets:
622 *
623 * Fragment 1: xen_netif_tx_request_t - flags = XEN_NETTXF_*
624 * size = total packet size
625 * [Extra 1: xen_netif_extra_info_t] - (only if fragment 1 flags include
626 * XEN_NETTXF_extra_info)
627 * ...
628 * [Extra N: xen_netif_extra_info_t] - (only if extra N-1 flags include
629 * XEN_NETIF_EXTRA_MORE)
630 * ...
631 * Fragment N: xen_netif_tx_request_t - (only if fragment N-1 flags include
632 * XEN_NETTXF_more_data - flags on preceding
633 * extras are not relevant here)
634 * flags = 0
635 * size = fragment size
636 *
637 * NOTE:
638 *
639 * This format slightly is different from that used for receive
640 * (backend -> frontend) packets. Specifically, in a multi-fragment
641 * packet the actual size of fragment 1 can only be determined by
642 * subtracting the sizes of fragments 2..N from the total packet size.
643 *
644 * Ring slot size is 12 octets, however not all request/response
645 * structs use the full size.
646 *
647 * tx request data (xen_netif_tx_request_t)
648 * ------------------------------------
649 *
650 * 0 1 2 3 4 5 6 7 octet
651 * +-----+-----+-----+-----+-----+-----+-----+-----+
652 * | grant ref | offset | flags |
653 * +-----+-----+-----+-----+-----+-----+-----+-----+
654 * | id | size |
655 * +-----+-----+-----+-----+
656 *
657 * grant ref: Reference to buffer page.
658 * offset: Offset within buffer page.
659 * flags: XEN_NETTXF_*.
660 * id: request identifier, echoed in response.
661 * size: packet size in bytes.
662 *
663 * tx response (xen_netif_tx_response_t)
664 * ---------------------------------
665 *
666 * 0 1 2 3 4 5 6 7 octet
667 * +-----+-----+-----+-----+-----+-----+-----+-----+
668 * | id | status | unused |
669 * +-----+-----+-----+-----+-----+-----+-----+-----+
670 * | unused |
671 * +-----+-----+-----+-----+
672 *
673 * id: reflects id in transmit request
674 * status: XEN_NETIF_RSP_*
675 *
676 * Guest receive
677 * =============
678 *
679 * This is the 'wire' format for receive (backend -> frontend) packets:
680 *
681 * Fragment 1: xen_netif_rx_request_t - flags = XEN_NETRXF_*
682 * size = fragment size
683 * [Extra 1: xen_netif_extra_info_t] - (only if fragment 1 flags include
684 * XEN_NETRXF_extra_info)
685 * ...
686 * [Extra N: xen_netif_extra_info_t] - (only if extra N-1 flags include
687 * XEN_NETIF_EXTRA_MORE)
688 * ...
689 * Fragment N: xen_netif_rx_request_t - (only if fragment N-1 flags include
690 * XEN_NETRXF_more_data - flags on preceding
691 * extras are not relevant here)
692 * flags = 0
693 * size = fragment size
694 *
695 * NOTE:
696 *
697 * This format slightly is different from that used for transmit
698 * (frontend -> backend) packets. Specifically, in a multi-fragment
699 * packet the size of the packet can only be determined by summing the
700 * sizes of fragments 1..N.
701 *
702 * Ring slot size is 8 octets.
703 *
704 * rx request (xen_netif_rx_request_t)
705 * -------------------------------
706 *
707 * 0 1 2 3 4 5 6 7 octet
708 * +-----+-----+-----+-----+-----+-----+-----+-----+
709 * | id | pad | gref |
710 * +-----+-----+-----+-----+-----+-----+-----+-----+
711 *
712 * id: request identifier, echoed in response.
713 * gref: reference to incoming granted frame.
714 *
715 * rx response (xen_netif_rx_response_t)
716 * ---------------------------------
717 *
718 * 0 1 2 3 4 5 6 7 octet
719 * +-----+-----+-----+-----+-----+-----+-----+-----+
720 * | id | offset | flags | status |
721 * +-----+-----+-----+-----+-----+-----+-----+-----+
722 *
723 * id: reflects id in receive request
724 * offset: offset in page of start of received packet
725 * flags: XEN_NETRXF_*
726 * status: -ve: XEN_NETIF_RSP_*; +ve: Rx'ed pkt size.
727 *
728 * NOTE: Historically, to support GSO on the frontend receive side, Linux
729 * netfront does not make use of the rx response id (because, as
730 * described below, extra info structures overlay the id field).
731 * Instead it assumes that responses always appear in the same ring
732 * slot as their corresponding request. Thus, to maintain
733 * compatibility, backends must make sure this is the case.
734 *
735 * Extra Info
736 * ==========
737 *
738 * Can be present if initial request or response has NET{T,R}XF_extra_info,
739 * or previous extra request has XEN_NETIF_EXTRA_MORE.
740 *
741 * The struct therefore needs to fit into either a tx or rx slot and
742 * is therefore limited to 8 octets.
743 *
744 * NOTE: Because extra info data overlays the usual request/response
745 * structures, there is no id information in the opposite direction.
746 * So, if an extra info overlays an rx response the frontend can
747 * assume that it is in the same ring slot as the request that was
748 * consumed to make the slot available, and the backend must ensure
749 * this assumption is true.
750 *
751 * extra info (xen_netif_extra_info_t)
752 * -------------------------------
753 *
754 * General format:
755 *
756 * 0 1 2 3 4 5 6 7 octet
757 * +-----+-----+-----+-----+-----+-----+-----+-----+
758 * |type |flags| type specific data |
759 * +-----+-----+-----+-----+-----+-----+-----+-----+
760 * | padding for tx |
761 * +-----+-----+-----+-----+
762 *
763 * type: XEN_NETIF_EXTRA_TYPE_*
764 * flags: XEN_NETIF_EXTRA_FLAG_*
765 * padding for tx: present only in the tx case due to 8 octet limit
766 * from rx case. Not shown in type specific entries
767 * below.
768 *
769 * XEN_NETIF_EXTRA_TYPE_GSO:
770 *
771 * 0 1 2 3 4 5 6 7 octet
772 * +-----+-----+-----+-----+-----+-----+-----+-----+
773 * |type |flags| size |type | pad | features |
774 * +-----+-----+-----+-----+-----+-----+-----+-----+
775 *
776 * type: Must be XEN_NETIF_EXTRA_TYPE_GSO
777 * flags: XEN_NETIF_EXTRA_FLAG_*
778 * size: Maximum payload size of each segment. For example,
779 * for TCP this is just the path MSS.
780 * type: XEN_NETIF_GSO_TYPE_*: This determines the protocol of
781 * the packet and any extra features required to segment the
782 * packet properly.
783 * features: EN_XEN_NETIF_GSO_FEAT_*: This specifies any extra GSO
784 * features required to process this packet, such as ECN
785 * support for TCPv4.
786 *
787 * XEN_NETIF_EXTRA_TYPE_MCAST_{ADD,DEL}:
788 *
789 * 0 1 2 3 4 5 6 7 octet
790 * +-----+-----+-----+-----+-----+-----+-----+-----+
791 * |type |flags| addr |
792 * +-----+-----+-----+-----+-----+-----+-----+-----+
793 *
794 * type: Must be XEN_NETIF_EXTRA_TYPE_MCAST_{ADD,DEL}
795 * flags: XEN_NETIF_EXTRA_FLAG_*
796 * addr: address to add/remove
797 *
798 * XEN_NETIF_EXTRA_TYPE_HASH:
799 *
800 * A backend that supports teoplitz hashing is assumed to accept
801 * this type of extra info in transmit packets.
802 * A frontend that enables hashing is assumed to accept
803 * this type of extra info in receive packets.
804 *
805 * 0 1 2 3 4 5 6 7 octet
806 * +-----+-----+-----+-----+-----+-----+-----+-----+
807 * |type |flags|htype| alg |LSB ---- value ---- MSB|
808 * +-----+-----+-----+-----+-----+-----+-----+-----+
809 *
810 * type: Must be XEN_NETIF_EXTRA_TYPE_HASH
811 * flags: XEN_NETIF_EXTRA_FLAG_*
812 * htype: Hash type (one of _XEN_NETIF_CTRL_HASH_TYPE_* - see above)
813 * alg: The algorithm used to calculate the hash (one of
814 * XEN_NETIF_CTRL_HASH_TYPE_ALGORITHM_* - see above)
815 * value: Hash value
816 */
817
818/* Protocol checksum field is blank in the packet (hardware offload)? */
819#define _XEN_NETTXF_csum_blank (0)
820#define XEN_NETTXF_csum_blank (1U<<_XEN_NETTXF_csum_blank)
821
822/* Packet data has been validated against protocol checksum. */
823#define _XEN_NETTXF_data_validated (1)
824#define XEN_NETTXF_data_validated (1U<<_XEN_NETTXF_data_validated)
825
826/* Packet continues in the next request descriptor. */
827#define _XEN_NETTXF_more_data (2)
828#define XEN_NETTXF_more_data (1U<<_XEN_NETTXF_more_data)
829
830/* Packet to be followed by extra descriptor(s). */
831#define _XEN_NETTXF_extra_info (3)
832#define XEN_NETTXF_extra_info (1U<<_XEN_NETTXF_extra_info)
833
834#define XEN_NETIF_MAX_TX_SIZE 0xFFFF
835struct xen_netif_tx_request {
836 grant_ref_t gref;
837 uint16_t offset;
838 uint16_t flags;
839 uint16_t id;
840 uint16_t size;
841};
842
843/* Types of xen_netif_extra_info descriptors. */
844#define XEN_NETIF_EXTRA_TYPE_NONE (0) /* Never used - invalid */
845#define XEN_NETIF_EXTRA_TYPE_GSO (1) /* u.gso */
846#define XEN_NETIF_EXTRA_TYPE_MCAST_ADD (2) /* u.mcast */
847#define XEN_NETIF_EXTRA_TYPE_MCAST_DEL (3) /* u.mcast */
848#define XEN_NETIF_EXTRA_TYPE_HASH (4) /* u.hash */
849#define XEN_NETIF_EXTRA_TYPE_MAX (5)
850
851/* xen_netif_extra_info_t flags. */
852#define _XEN_NETIF_EXTRA_FLAG_MORE (0)
853#define XEN_NETIF_EXTRA_FLAG_MORE (1U<<_XEN_NETIF_EXTRA_FLAG_MORE)
854
855/* GSO types */
856#define XEN_NETIF_GSO_TYPE_NONE (0)
857#define XEN_NETIF_GSO_TYPE_TCPV4 (1)
858#define XEN_NETIF_GSO_TYPE_TCPV6 (2)
859
860/*
861 * This structure needs to fit within both xen_netif_tx_request_t and
862 * xen_netif_rx_response_t for compatibility.
863 */
864struct xen_netif_extra_info {
865 uint8_t type;
866 uint8_t flags;
867 union {
868 struct {
869 uint16_t size;
870 uint8_t type;
871 uint8_t pad;
872 uint16_t features;
873 } gso;
874 struct {
875 uint8_t addr[6];
876 } mcast;
877 struct {
878 uint8_t type;
879 uint8_t algorithm;
880 uint8_t value[4];
881 } hash;
882 uint16_t pad[3];
883 } u;
884};
885
886struct xen_netif_tx_response {
887 uint16_t id;
888 int16_t status;
889};
890
891struct xen_netif_rx_request {
892 uint16_t id; /* Echoed in response message. */
893 uint16_t pad;
894 grant_ref_t gref;
895};
896
897/* Packet data has been validated against protocol checksum. */
898#define _XEN_NETRXF_data_validated (0)
899#define XEN_NETRXF_data_validated (1U<<_XEN_NETRXF_data_validated)
900
901/* Protocol checksum field is blank in the packet (hardware offload)? */
902#define _XEN_NETRXF_csum_blank (1)
903#define XEN_NETRXF_csum_blank (1U<<_XEN_NETRXF_csum_blank)
904
905/* Packet continues in the next request descriptor. */
906#define _XEN_NETRXF_more_data (2)
907#define XEN_NETRXF_more_data (1U<<_XEN_NETRXF_more_data)
908
909/* Packet to be followed by extra descriptor(s). */
910#define _XEN_NETRXF_extra_info (3)
911#define XEN_NETRXF_extra_info (1U<<_XEN_NETRXF_extra_info)
912
913/* Packet has GSO prefix. Deprecated but included for compatibility */
914#define _XEN_NETRXF_gso_prefix (4)
915#define XEN_NETRXF_gso_prefix (1U<<_XEN_NETRXF_gso_prefix)
916
917struct xen_netif_rx_response {
918 uint16_t id;
919 uint16_t offset;
920 uint16_t flags;
921 int16_t status;
922};
923
924/*
925 * Generate xen_netif ring structures and types.
926 */
927
928DEFINE_RING_TYPES(xen_netif_tx, struct xen_netif_tx_request,
929 struct xen_netif_tx_response);
930DEFINE_RING_TYPES(xen_netif_rx, struct xen_netif_rx_request,
931 struct xen_netif_rx_response);
932
933#define XEN_NETIF_RSP_DROPPED -2
934#define XEN_NETIF_RSP_ERROR -1
935#define XEN_NETIF_RSP_OKAY 0
936/* No response: used for auxiliary requests (e.g., xen_netif_extra_info_t). */
937#define XEN_NETIF_RSP_NULL 1
938
939#endif
940