xref: /linux/include/xen/interface/io/netif.h (revision f7af616c632ee2ac3af0876fe33bf9e0232e665a)
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  * "xdp-headroom" is used to request that extra space is added
165  * for XDP processing.  The value is measured in bytes and passed by
166  * the frontend to be consistent between both ends.
167  * If the value is greater than zero that means that
168  * an RX response is going to be passed to an XDP program for processing.
169  * XEN_NETIF_MAX_XDP_HEADROOM defines the maximum headroom offset in bytes
170  *
171  * "feature-xdp-headroom" is set to "1" by the netback side like other features
172  * so a guest can check if an XDP program can be processed.
173  */
174 #define XEN_NETIF_MAX_XDP_HEADROOM 0x7FFF
175 
176 /*
177  * Control ring
178  * ============
179  *
180  * Some features, such as hashing (detailed below), require a
181  * significant amount of out-of-band data to be passed from frontend to
182  * backend. Use of xenstore is not suitable for large quantities of data
183  * because of quota limitations and so a dedicated 'control ring' is used.
184  * The ability of the backend to use a control ring is advertised by
185  * setting:
186  *
187  * /local/domain/X/backend/<domid>/<vif>/feature-ctrl-ring = "1"
188  *
189  * The frontend provides a control ring to the backend by setting:
190  *
191  * /local/domain/<domid>/device/vif/<vif>/ctrl-ring-ref = <gref>
192  * /local/domain/<domid>/device/vif/<vif>/event-channel-ctrl = <port>
193  *
194  * where <gref> is the grant reference of the shared page used to
195  * implement the control ring and <port> is an event channel to be used
196  * as a mailbox interrupt. These keys must be set before the frontend
197  * moves into the connected state.
198  *
199  * The control ring uses a fixed request/response message size and is
200  * balanced (i.e. one request to one response), so operationally it is much
201  * the same as a transmit or receive ring.
202  * Note that there is no requirement that responses are issued in the same
203  * order as requests.
204  */
205 
206 /*
207  * Hash types
208  * ==========
209  *
210  * For the purposes of the definitions below, 'Packet[]' is an array of
211  * octets containing an IP packet without options, 'Array[X..Y]' means a
212  * sub-array of 'Array' containing bytes X thru Y inclusive, and '+' is
213  * used to indicate concatenation of arrays.
214  */
215 
216 /*
217  * A hash calculated over an IP version 4 header as follows:
218  *
219  * Buffer[0..8] = Packet[12..15] (source address) +
220  *                Packet[16..19] (destination address)
221  *
222  * Result = Hash(Buffer, 8)
223  */
224 #define _XEN_NETIF_CTRL_HASH_TYPE_IPV4 0
225 #define XEN_NETIF_CTRL_HASH_TYPE_IPV4 \
226 	(1 << _XEN_NETIF_CTRL_HASH_TYPE_IPV4)
227 
228 /*
229  * A hash calculated over an IP version 4 header and TCP header as
230  * follows:
231  *
232  * Buffer[0..12] = Packet[12..15] (source address) +
233  *                 Packet[16..19] (destination address) +
234  *                 Packet[20..21] (source port) +
235  *                 Packet[22..23] (destination port)
236  *
237  * Result = Hash(Buffer, 12)
238  */
239 #define _XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP 1
240 #define XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP \
241 	(1 << _XEN_NETIF_CTRL_HASH_TYPE_IPV4_TCP)
242 
243 /*
244  * A hash calculated over an IP version 6 header as follows:
245  *
246  * Buffer[0..32] = Packet[8..23]  (source address ) +
247  *                 Packet[24..39] (destination address)
248  *
249  * Result = Hash(Buffer, 32)
250  */
251 #define _XEN_NETIF_CTRL_HASH_TYPE_IPV6 2
252 #define XEN_NETIF_CTRL_HASH_TYPE_IPV6 \
253 	(1 << _XEN_NETIF_CTRL_HASH_TYPE_IPV6)
254 
255 /*
256  * A hash calculated over an IP version 6 header and TCP header as
257  * follows:
258  *
259  * Buffer[0..36] = Packet[8..23]  (source address) +
260  *                 Packet[24..39] (destination address) +
261  *                 Packet[40..41] (source port) +
262  *                 Packet[42..43] (destination port)
263  *
264  * Result = Hash(Buffer, 36)
265  */
266 #define _XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP 3
267 #define XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP \
268 	(1 << _XEN_NETIF_CTRL_HASH_TYPE_IPV6_TCP)
269 
270 /*
271  * Hash algorithms
272  * ===============
273  */
274 
275 #define XEN_NETIF_CTRL_HASH_ALGORITHM_NONE 0
276 
277 /*
278  * Toeplitz hash:
279  */
280 
281 #define XEN_NETIF_CTRL_HASH_ALGORITHM_TOEPLITZ 1
282 
283 /*
284  * This algorithm uses a 'key' as well as the data buffer itself.
285  * (Buffer[] and Key[] are treated as shift-registers where the MSB of
286  * Buffer/Key[0] is considered 'left-most' and the LSB of Buffer/Key[N-1]
287  * is the 'right-most').
288  *
289  * Value = 0
290  * For number of bits in Buffer[]
291  *    If (left-most bit of Buffer[] is 1)
292  *        Value ^= left-most 32 bits of Key[]
293  *    Key[] << 1
294  *    Buffer[] << 1
295  *
296  * The code below is provided for convenience where an operating system
297  * does not already provide an implementation.
298  */
299 #ifdef XEN_NETIF_DEFINE_TOEPLITZ
300 static uint32_t xen_netif_toeplitz_hash(const uint8_t *key,
301 					unsigned int keylen,
302 					const uint8_t *buf, unsigned int buflen)
303 {
304 	unsigned int keyi, bufi;
305 	uint64_t prefix = 0;
306 	uint64_t hash = 0;
307 
308 	/* Pre-load prefix with the first 8 bytes of the key */
309 	for (keyi = 0; keyi < 8; keyi++) {
310 		prefix <<= 8;
311 		prefix |= (keyi < keylen) ? key[keyi] : 0;
312 	}
313 
314 	for (bufi = 0; bufi < buflen; bufi++) {
315 		uint8_t byte = buf[bufi];
316 		unsigned int bit;
317 
318 		for (bit = 0; bit < 8; bit++) {
319 			if (byte & 0x80)
320 				hash ^= prefix;
321 			prefix <<= 1;
322 			byte <<= 1;
323 		}
324 
325 		/*
326 		 * 'prefix' has now been left-shifted by 8, so
327 		 * OR in the next byte.
328 		 */
329 		prefix |= (keyi < keylen) ? key[keyi] : 0;
330 		keyi++;
331 	}
332 
333 	/* The valid part of the hash is in the upper 32 bits. */
334 	return hash >> 32;
335 }
336 #endif				/* XEN_NETIF_DEFINE_TOEPLITZ */
337 
338 /*
339  * Control requests (struct xen_netif_ctrl_request)
340  * ================================================
341  *
342  * All requests have the following format:
343  *
344  *    0     1     2     3     4     5     6     7  octet
345  * +-----+-----+-----+-----+-----+-----+-----+-----+
346  * |    id     |   type    |         data[0]       |
347  * +-----+-----+-----+-----+-----+-----+-----+-----+
348  * |         data[1]       |         data[2]       |
349  * +-----+-----+-----+-----+-----------------------+
350  *
351  * id: the request identifier, echoed in response.
352  * type: the type of request (see below)
353  * data[]: any data associated with the request (determined by type)
354  */
355 
356 struct xen_netif_ctrl_request {
357 	uint16_t id;
358 	uint16_t type;
359 
360 #define XEN_NETIF_CTRL_TYPE_INVALID               0
361 #define XEN_NETIF_CTRL_TYPE_GET_HASH_FLAGS        1
362 #define XEN_NETIF_CTRL_TYPE_SET_HASH_FLAGS        2
363 #define XEN_NETIF_CTRL_TYPE_SET_HASH_KEY          3
364 #define XEN_NETIF_CTRL_TYPE_GET_HASH_MAPPING_SIZE 4
365 #define XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE 5
366 #define XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING      6
367 #define XEN_NETIF_CTRL_TYPE_SET_HASH_ALGORITHM    7
368 
369 	uint32_t data[3];
370 };
371 
372 /*
373  * Control responses (struct xen_netif_ctrl_response)
374  * ==================================================
375  *
376  * All responses have the following format:
377  *
378  *    0     1     2     3     4     5     6     7  octet
379  * +-----+-----+-----+-----+-----+-----+-----+-----+
380  * |    id     |   type    |         status        |
381  * +-----+-----+-----+-----+-----+-----+-----+-----+
382  * |         data          |
383  * +-----+-----+-----+-----+
384  *
385  * id: the corresponding request identifier
386  * type: the type of the corresponding request
387  * status: the status of request processing
388  * data: any data associated with the response (determined by type and
389  *       status)
390  */
391 
392 struct xen_netif_ctrl_response {
393 	uint16_t id;
394 	uint16_t type;
395 	uint32_t status;
396 
397 #define XEN_NETIF_CTRL_STATUS_SUCCESS           0
398 #define XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED     1
399 #define XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER 2
400 #define XEN_NETIF_CTRL_STATUS_BUFFER_OVERFLOW   3
401 
402 	uint32_t data;
403 };
404 
405 /*
406  * Control messages
407  * ================
408  *
409  * XEN_NETIF_CTRL_TYPE_SET_HASH_ALGORITHM
410  * --------------------------------------
411  *
412  * This is sent by the frontend to set the desired hash algorithm.
413  *
414  * Request:
415  *
416  *  type    = XEN_NETIF_CTRL_TYPE_SET_HASH_ALGORITHM
417  *  data[0] = a XEN_NETIF_CTRL_HASH_ALGORITHM_* value
418  *  data[1] = 0
419  *  data[2] = 0
420  *
421  * Response:
422  *
423  *  status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED     - Operation not
424  *                                                     supported
425  *           XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - The algorithm is not
426  *                                                     supported
427  *           XEN_NETIF_CTRL_STATUS_SUCCESS           - Operation successful
428  *
429  * NOTE: Setting data[0] to XEN_NETIF_CTRL_HASH_ALGORITHM_NONE disables
430  *       hashing and the backend is free to choose how it steers packets
431  *       to queues (which is the default behaviour).
432  *
433  * XEN_NETIF_CTRL_TYPE_GET_HASH_FLAGS
434  * ----------------------------------
435  *
436  * This is sent by the frontend to query the types of hash supported by
437  * the backend.
438  *
439  * Request:
440  *
441  *  type    = XEN_NETIF_CTRL_TYPE_GET_HASH_FLAGS
442  *  data[0] = 0
443  *  data[1] = 0
444  *  data[2] = 0
445  *
446  * Response:
447  *
448  *  status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED - Operation not supported
449  *           XEN_NETIF_CTRL_STATUS_SUCCESS       - Operation successful
450  *  data   = supported hash types (if operation was successful)
451  *
452  * NOTE: A valid hash algorithm must be selected before this operation can
453  *       succeed.
454  *
455  * XEN_NETIF_CTRL_TYPE_SET_HASH_FLAGS
456  * ----------------------------------
457  *
458  * This is sent by the frontend to set the types of hash that the backend
459  * should calculate. (See above for hash type definitions).
460  * Note that the 'maximal' type of hash should always be chosen. For
461  * example, if the frontend sets both IPV4 and IPV4_TCP hash types then
462  * the latter hash type should be calculated for any TCP packet and the
463  * former only calculated for non-TCP packets.
464  *
465  * Request:
466  *
467  *  type    = XEN_NETIF_CTRL_TYPE_SET_HASH_FLAGS
468  *  data[0] = bitwise OR of XEN_NETIF_CTRL_HASH_TYPE_* values
469  *  data[1] = 0
470  *  data[2] = 0
471  *
472  * Response:
473  *
474  *  status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED     - Operation not
475  *                                                     supported
476  *           XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - One or more flag
477  *                                                     value is invalid or
478  *                                                     unsupported
479  *           XEN_NETIF_CTRL_STATUS_SUCCESS           - Operation successful
480  *  data   = 0
481  *
482  * NOTE: A valid hash algorithm must be selected before this operation can
483  *       succeed.
484  *       Also, setting data[0] to zero disables hashing and the backend
485  *       is free to choose how it steers packets to queues.
486  *
487  * XEN_NETIF_CTRL_TYPE_SET_HASH_KEY
488  * --------------------------------
489  *
490  * This is sent by the frontend to set the key of the hash if the algorithm
491  * requires it. (See hash algorithms above).
492  *
493  * Request:
494  *
495  *  type    = XEN_NETIF_CTRL_TYPE_SET_HASH_KEY
496  *  data[0] = grant reference of page containing the key (assumed to
497  *            start at beginning of grant)
498  *  data[1] = size of key in octets
499  *  data[2] = 0
500  *
501  * Response:
502  *
503  *  status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED     - Operation not
504  *                                                     supported
505  *           XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - Key size is invalid
506  *           XEN_NETIF_CTRL_STATUS_BUFFER_OVERFLOW   - Key size is larger
507  *                                                     than the backend
508  *                                                     supports
509  *           XEN_NETIF_CTRL_STATUS_SUCCESS           - Operation successful
510  *  data   = 0
511  *
512  * NOTE: Any key octets not specified are assumed to be zero (the key
513  *       is assumed to be empty by default) and specifying a new key
514  *       invalidates any previous key, hence specifying a key size of
515  *       zero will clear the key (which ensures that the calculated hash
516  *       will always be zero).
517  *       The maximum size of key is algorithm and backend specific, but
518  *       is also limited by the single grant reference.
519  *       The grant reference may be read-only and must remain valid until
520  *       the response has been processed.
521  *
522  * XEN_NETIF_CTRL_TYPE_GET_HASH_MAPPING_SIZE
523  * -----------------------------------------
524  *
525  * This is sent by the frontend to query the maximum size of mapping
526  * table supported by the backend. The size is specified in terms of
527  * table entries.
528  *
529  * Request:
530  *
531  *  type    = XEN_NETIF_CTRL_TYPE_GET_HASH_MAPPING_SIZE
532  *  data[0] = 0
533  *  data[1] = 0
534  *  data[2] = 0
535  *
536  * Response:
537  *
538  *  status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED - Operation not supported
539  *           XEN_NETIF_CTRL_STATUS_SUCCESS       - Operation successful
540  *  data   = maximum number of entries allowed in the mapping table
541  *           (if operation was successful) or zero if a mapping table is
542  *           not supported (i.e. hash mapping is done only by modular
543  *           arithmetic).
544  *
545  * XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE
546  * -------------------------------------
547  *
548  * This is sent by the frontend to set the actual size of the mapping
549  * table to be used by the backend. The size is specified in terms of
550  * table entries.
551  * Any previous table is invalidated by this message and any new table
552  * is assumed to be zero filled.
553  *
554  * Request:
555  *
556  *  type    = XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE
557  *  data[0] = number of entries in mapping table
558  *  data[1] = 0
559  *  data[2] = 0
560  *
561  * Response:
562  *
563  *  status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED     - Operation not
564  *                                                     supported
565  *           XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - Table size is invalid
566  *           XEN_NETIF_CTRL_STATUS_SUCCESS           - Operation successful
567  *  data   = 0
568  *
569  * NOTE: Setting data[0] to 0 means that hash mapping should be done
570  *       using modular arithmetic.
571  *
572  * XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING
573  * ------------------------------------
574  *
575  * This is sent by the frontend to set the content of the table mapping
576  * hash value to queue number. The backend should calculate the hash from
577  * the packet header, use it as an index into the table (modulo the size
578  * of the table) and then steer the packet to the queue number found at
579  * that index.
580  *
581  * Request:
582  *
583  *  type    = XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING
584  *  data[0] = grant reference of page containing the mapping (sub-)table
585  *            (assumed to start at beginning of grant)
586  *  data[1] = size of (sub-)table in entries
587  *  data[2] = offset, in entries, of sub-table within overall table
588  *
589  * Response:
590  *
591  *  status = XEN_NETIF_CTRL_STATUS_NOT_SUPPORTED     - Operation not
592  *                                                     supported
593  *           XEN_NETIF_CTRL_STATUS_INVALID_PARAMETER - Table size or content
594  *                                                     is invalid
595  *           XEN_NETIF_CTRL_STATUS_BUFFER_OVERFLOW   - Table size is larger
596  *                                                     than the backend
597  *                                                     supports
598  *           XEN_NETIF_CTRL_STATUS_SUCCESS           - Operation successful
599  *  data   = 0
600  *
601  * NOTE: The overall table has the following format:
602  *
603  *          0     1     2     3     4     5     6     7  octet
604  *       +-----+-----+-----+-----+-----+-----+-----+-----+
605  *       |       mapping[0]      |       mapping[1]      |
606  *       +-----+-----+-----+-----+-----+-----+-----+-----+
607  *       |                       .                       |
608  *       |                       .                       |
609  *       |                       .                       |
610  *       +-----+-----+-----+-----+-----+-----+-----+-----+
611  *       |      mapping[N-2]     |      mapping[N-1]     |
612  *       +-----+-----+-----+-----+-----+-----+-----+-----+
613  *
614  *       where N is specified by a XEN_NETIF_CTRL_TYPE_SET_HASH_MAPPING_SIZE
615  *       message and each  mapping must specifies a queue between 0 and
616  *       "multi-queue-num-queues" (see above).
617  *       The backend may support a mapping table larger than can be
618  *       mapped by a single grant reference. Thus sub-tables within a
619  *       larger table can be individually set by sending multiple messages
620  *       with differing offset values. Specifying a new sub-table does not
621  *       invalidate any table data outside that range.
622  *       The grant reference may be read-only and must remain valid until
623  *       the response has been processed.
624  */
625 
626 DEFINE_RING_TYPES(xen_netif_ctrl,
627 		  struct xen_netif_ctrl_request,
628 		  struct xen_netif_ctrl_response);
629 
630 /*
631  * Guest transmit
632  * ==============
633  *
634  * This is the 'wire' format for transmit (frontend -> backend) packets:
635  *
636  *  Fragment 1: xen_netif_tx_request_t  - flags = XEN_NETTXF_*
637  *                                    size = total packet size
638  * [Extra 1: xen_netif_extra_info_t]    - (only if fragment 1 flags include
639  *                                     XEN_NETTXF_extra_info)
640  *  ...
641  * [Extra N: xen_netif_extra_info_t]    - (only if extra N-1 flags include
642  *                                     XEN_NETIF_EXTRA_MORE)
643  *  ...
644  *  Fragment N: xen_netif_tx_request_t  - (only if fragment N-1 flags include
645  *                                     XEN_NETTXF_more_data - flags on preceding
646  *                                     extras are not relevant here)
647  *                                    flags = 0
648  *                                    size = fragment size
649  *
650  * NOTE:
651  *
652  * This format slightly is different from that used for receive
653  * (backend -> frontend) packets. Specifically, in a multi-fragment
654  * packet the actual size of fragment 1 can only be determined by
655  * subtracting the sizes of fragments 2..N from the total packet size.
656  *
657  * Ring slot size is 12 octets, however not all request/response
658  * structs use the full size.
659  *
660  * tx request data (xen_netif_tx_request_t)
661  * ------------------------------------
662  *
663  *    0     1     2     3     4     5     6     7  octet
664  * +-----+-----+-----+-----+-----+-----+-----+-----+
665  * | grant ref             | offset    | flags     |
666  * +-----+-----+-----+-----+-----+-----+-----+-----+
667  * | id        | size      |
668  * +-----+-----+-----+-----+
669  *
670  * grant ref: Reference to buffer page.
671  * offset: Offset within buffer page.
672  * flags: XEN_NETTXF_*.
673  * id: request identifier, echoed in response.
674  * size: packet size in bytes.
675  *
676  * tx response (xen_netif_tx_response_t)
677  * ---------------------------------
678  *
679  *    0     1     2     3     4     5     6     7  octet
680  * +-----+-----+-----+-----+-----+-----+-----+-----+
681  * | id        | status    | unused                |
682  * +-----+-----+-----+-----+-----+-----+-----+-----+
683  * | unused                |
684  * +-----+-----+-----+-----+
685  *
686  * id: reflects id in transmit request
687  * status: XEN_NETIF_RSP_*
688  *
689  * Guest receive
690  * =============
691  *
692  * This is the 'wire' format for receive (backend -> frontend) packets:
693  *
694  *  Fragment 1: xen_netif_rx_request_t  - flags = XEN_NETRXF_*
695  *                                    size = fragment size
696  * [Extra 1: xen_netif_extra_info_t]    - (only if fragment 1 flags include
697  *                                     XEN_NETRXF_extra_info)
698  *  ...
699  * [Extra N: xen_netif_extra_info_t]    - (only if extra N-1 flags include
700  *                                     XEN_NETIF_EXTRA_MORE)
701  *  ...
702  *  Fragment N: xen_netif_rx_request_t  - (only if fragment N-1 flags include
703  *                                     XEN_NETRXF_more_data - flags on preceding
704  *                                     extras are not relevant here)
705  *                                    flags = 0
706  *                                    size = fragment size
707  *
708  * NOTE:
709  *
710  * This format slightly is different from that used for transmit
711  * (frontend -> backend) packets. Specifically, in a multi-fragment
712  * packet the size of the packet can only be determined by summing the
713  * sizes of fragments 1..N.
714  *
715  * Ring slot size is 8 octets.
716  *
717  * rx request (xen_netif_rx_request_t)
718  * -------------------------------
719  *
720  *    0     1     2     3     4     5     6     7  octet
721  * +-----+-----+-----+-----+-----+-----+-----+-----+
722  * | id        | pad       | gref                  |
723  * +-----+-----+-----+-----+-----+-----+-----+-----+
724  *
725  * id: request identifier, echoed in response.
726  * gref: reference to incoming granted frame.
727  *
728  * rx response (xen_netif_rx_response_t)
729  * ---------------------------------
730  *
731  *    0     1     2     3     4     5     6     7  octet
732  * +-----+-----+-----+-----+-----+-----+-----+-----+
733  * | id        | offset    | flags     | status    |
734  * +-----+-----+-----+-----+-----+-----+-----+-----+
735  *
736  * id: reflects id in receive request
737  * offset: offset in page of start of received packet
738  * flags: XEN_NETRXF_*
739  * status: -ve: XEN_NETIF_RSP_*; +ve: Rx'ed pkt size.
740  *
741  * NOTE: Historically, to support GSO on the frontend receive side, Linux
742  *       netfront does not make use of the rx response id (because, as
743  *       described below, extra info structures overlay the id field).
744  *       Instead it assumes that responses always appear in the same ring
745  *       slot as their corresponding request. Thus, to maintain
746  *       compatibility, backends must make sure this is the case.
747  *
748  * Extra Info
749  * ==========
750  *
751  * Can be present if initial request or response has NET{T,R}XF_extra_info,
752  * or previous extra request has XEN_NETIF_EXTRA_MORE.
753  *
754  * The struct therefore needs to fit into either a tx or rx slot and
755  * is therefore limited to 8 octets.
756  *
757  * NOTE: Because extra info data overlays the usual request/response
758  *       structures, there is no id information in the opposite direction.
759  *       So, if an extra info overlays an rx response the frontend can
760  *       assume that it is in the same ring slot as the request that was
761  *       consumed to make the slot available, and the backend must ensure
762  *       this assumption is true.
763  *
764  * extra info (xen_netif_extra_info_t)
765  * -------------------------------
766  *
767  * General format:
768  *
769  *    0     1     2     3     4     5     6     7  octet
770  * +-----+-----+-----+-----+-----+-----+-----+-----+
771  * |type |flags| type specific data                |
772  * +-----+-----+-----+-----+-----+-----+-----+-----+
773  * | padding for tx        |
774  * +-----+-----+-----+-----+
775  *
776  * type: XEN_NETIF_EXTRA_TYPE_*
777  * flags: XEN_NETIF_EXTRA_FLAG_*
778  * padding for tx: present only in the tx case due to 8 octet limit
779  *                 from rx case. Not shown in type specific entries
780  *                 below.
781  *
782  * XEN_NETIF_EXTRA_TYPE_GSO:
783  *
784  *    0     1     2     3     4     5     6     7  octet
785  * +-----+-----+-----+-----+-----+-----+-----+-----+
786  * |type |flags| size      |type | pad | features  |
787  * +-----+-----+-----+-----+-----+-----+-----+-----+
788  *
789  * type: Must be XEN_NETIF_EXTRA_TYPE_GSO
790  * flags: XEN_NETIF_EXTRA_FLAG_*
791  * size: Maximum payload size of each segment. For example,
792  *       for TCP this is just the path MSS.
793  * type: XEN_NETIF_GSO_TYPE_*: This determines the protocol of
794  *       the packet and any extra features required to segment the
795  *       packet properly.
796  * features: EN_XEN_NETIF_GSO_FEAT_*: This specifies any extra GSO
797  *           features required to process this packet, such as ECN
798  *           support for TCPv4.
799  *
800  * XEN_NETIF_EXTRA_TYPE_MCAST_{ADD,DEL}:
801  *
802  *    0     1     2     3     4     5     6     7  octet
803  * +-----+-----+-----+-----+-----+-----+-----+-----+
804  * |type |flags| addr                              |
805  * +-----+-----+-----+-----+-----+-----+-----+-----+
806  *
807  * type: Must be XEN_NETIF_EXTRA_TYPE_MCAST_{ADD,DEL}
808  * flags: XEN_NETIF_EXTRA_FLAG_*
809  * addr: address to add/remove
810  *
811  * XEN_NETIF_EXTRA_TYPE_HASH:
812  *
813  * A backend that supports teoplitz hashing is assumed to accept
814  * this type of extra info in transmit packets.
815  * A frontend that enables hashing is assumed to accept
816  * this type of extra info in receive packets.
817  *
818  *    0     1     2     3     4     5     6     7  octet
819  * +-----+-----+-----+-----+-----+-----+-----+-----+
820  * |type |flags|htype| alg |LSB ---- value ---- MSB|
821  * +-----+-----+-----+-----+-----+-----+-----+-----+
822  *
823  * type: Must be XEN_NETIF_EXTRA_TYPE_HASH
824  * flags: XEN_NETIF_EXTRA_FLAG_*
825  * htype: Hash type (one of _XEN_NETIF_CTRL_HASH_TYPE_* - see above)
826  * alg: The algorithm used to calculate the hash (one of
827  *      XEN_NETIF_CTRL_HASH_TYPE_ALGORITHM_* - see above)
828  * value: Hash value
829  */
830 
831 /* Protocol checksum field is blank in the packet (hardware offload)? */
832 #define _XEN_NETTXF_csum_blank     (0)
833 #define  XEN_NETTXF_csum_blank     (1U<<_XEN_NETTXF_csum_blank)
834 
835 /* Packet data has been validated against protocol checksum. */
836 #define _XEN_NETTXF_data_validated (1)
837 #define  XEN_NETTXF_data_validated (1U<<_XEN_NETTXF_data_validated)
838 
839 /* Packet continues in the next request descriptor. */
840 #define _XEN_NETTXF_more_data      (2)
841 #define  XEN_NETTXF_more_data      (1U<<_XEN_NETTXF_more_data)
842 
843 /* Packet to be followed by extra descriptor(s). */
844 #define _XEN_NETTXF_extra_info     (3)
845 #define  XEN_NETTXF_extra_info     (1U<<_XEN_NETTXF_extra_info)
846 
847 #define XEN_NETIF_MAX_TX_SIZE 0xFFFF
848 struct xen_netif_tx_request {
849 	grant_ref_t gref;
850 	uint16_t offset;
851 	uint16_t flags;
852 	uint16_t id;
853 	uint16_t size;
854 };
855 
856 /* Types of xen_netif_extra_info descriptors. */
857 #define XEN_NETIF_EXTRA_TYPE_NONE      (0)	/* Never used - invalid */
858 #define XEN_NETIF_EXTRA_TYPE_GSO       (1)	/* u.gso */
859 #define XEN_NETIF_EXTRA_TYPE_MCAST_ADD (2)	/* u.mcast */
860 #define XEN_NETIF_EXTRA_TYPE_MCAST_DEL (3)	/* u.mcast */
861 #define XEN_NETIF_EXTRA_TYPE_HASH      (4)	/* u.hash */
862 #define XEN_NETIF_EXTRA_TYPE_XDP       (5)	/* u.xdp */
863 #define XEN_NETIF_EXTRA_TYPE_MAX       (6)
864 
865 /* xen_netif_extra_info_t flags. */
866 #define _XEN_NETIF_EXTRA_FLAG_MORE (0)
867 #define XEN_NETIF_EXTRA_FLAG_MORE  (1U<<_XEN_NETIF_EXTRA_FLAG_MORE)
868 
869 /* GSO types */
870 #define XEN_NETIF_GSO_TYPE_NONE         (0)
871 #define XEN_NETIF_GSO_TYPE_TCPV4        (1)
872 #define XEN_NETIF_GSO_TYPE_TCPV6        (2)
873 
874 /*
875  * This structure needs to fit within both xen_netif_tx_request_t and
876  * xen_netif_rx_response_t for compatibility.
877  */
878 struct xen_netif_extra_info {
879 	uint8_t type;
880 	uint8_t flags;
881 	union {
882 		struct {
883 			uint16_t size;
884 			uint8_t type;
885 			uint8_t pad;
886 			uint16_t features;
887 		} gso;
888 		struct {
889 			uint8_t addr[6];
890 		} mcast;
891 		struct {
892 			uint8_t type;
893 			uint8_t algorithm;
894 			uint8_t value[4];
895 		} hash;
896 		struct {
897 			uint16_t headroom;
898 			uint16_t pad[2];
899 		} xdp;
900 		uint16_t pad[3];
901 	} u;
902 };
903 
904 struct xen_netif_tx_response {
905 	uint16_t id;
906 	int16_t status;
907 };
908 
909 struct xen_netif_rx_request {
910 	uint16_t id;		/* Echoed in response message.        */
911 	uint16_t pad;
912 	grant_ref_t gref;
913 };
914 
915 /* Packet data has been validated against protocol checksum. */
916 #define _XEN_NETRXF_data_validated (0)
917 #define  XEN_NETRXF_data_validated (1U<<_XEN_NETRXF_data_validated)
918 
919 /* Protocol checksum field is blank in the packet (hardware offload)? */
920 #define _XEN_NETRXF_csum_blank     (1)
921 #define  XEN_NETRXF_csum_blank     (1U<<_XEN_NETRXF_csum_blank)
922 
923 /* Packet continues in the next request descriptor. */
924 #define _XEN_NETRXF_more_data      (2)
925 #define  XEN_NETRXF_more_data      (1U<<_XEN_NETRXF_more_data)
926 
927 /* Packet to be followed by extra descriptor(s). */
928 #define _XEN_NETRXF_extra_info     (3)
929 #define  XEN_NETRXF_extra_info     (1U<<_XEN_NETRXF_extra_info)
930 
931 /* Packet has GSO prefix. Deprecated but included for compatibility */
932 #define _XEN_NETRXF_gso_prefix     (4)
933 #define  XEN_NETRXF_gso_prefix     (1U<<_XEN_NETRXF_gso_prefix)
934 
935 struct xen_netif_rx_response {
936 	uint16_t id;
937 	uint16_t offset;
938 	uint16_t flags;
939 	int16_t status;
940 };
941 
942 /*
943  * Generate xen_netif ring structures and types.
944  */
945 
946 DEFINE_RING_TYPES(xen_netif_tx, struct xen_netif_tx_request,
947 		  struct xen_netif_tx_response);
948 DEFINE_RING_TYPES(xen_netif_rx, struct xen_netif_rx_request,
949 		  struct xen_netif_rx_response);
950 
951 #define XEN_NETIF_RSP_DROPPED         -2
952 #define XEN_NETIF_RSP_ERROR           -1
953 #define XEN_NETIF_RSP_OKAY             0
954 /* No response: used for auxiliary requests (e.g., xen_netif_extra_info_t). */
955 #define XEN_NETIF_RSP_NULL             1
956 
957 #endif
958