1 /******************************************************************************
4 * Unified block-device I/O interface for Xen guest OSes.
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
24 * Copyright (c) 2003-2004, Keir Fraser
25 * Copyright (c) 2012, Spectra Logic Corporation
28 #ifndef __XEN_PUBLIC_IO_BLKIF_H__
29 #define __XEN_PUBLIC_IO_BLKIF_H__
32 #include "../grant_table.h"
35 * Front->back notifications: When enqueuing a new request, sending a
36 * notification can be made conditional on req_event (i.e., the generic
37 * hold-off mechanism provided by the ring macros). Backends must set
38 * req_event appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
40 * Back->front notifications: When enqueuing a new response, sending a
41 * notification can be made conditional on rsp_event (i.e., the generic
42 * hold-off mechanism provided by the ring macros). Frontends must set
43 * rsp_event appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
47 #define blkif_vdev_t uint16_t
49 #define blkif_sector_t uint64_t
52 * Feature and Parameter Negotiation
53 * =================================
54 * The two halves of a Xen block driver utilize nodes within the XenStore to
55 * communicate capabilities and to negotiate operating parameters. This
56 * section enumerates these nodes which reside in the respective front and
57 * backend portions of the XenStore, following the XenBus convention.
59 * All data in the XenStore is stored as strings. Nodes specifying numeric
60 * values are encoded in decimal. Integer value ranges listed below are
61 * expressed as fixed sized integer types capable of storing the conversion
62 * of a properly formatted node string, without loss of information.
64 * Any specified default value is in effect if the corresponding XenBus node
65 * is not present in the XenStore.
67 * XenStore nodes in sections marked "PRIVATE" are solely for use by the
68 * driver side whose XenBus tree contains them.
70 * XenStore nodes marked "DEPRECATED" in their notes section should only be
71 * used to provide interoperability with legacy implementations.
73 * See the XenBus state transition diagram below for details on when XenBus
74 * nodes must be published and when they can be queried.
76 *****************************************************************************
77 * Backend XenBus Nodes
78 *****************************************************************************
80 *------------------ Backend Device Identification (PRIVATE) ------------------
83 * Values: "r" (read only), "w" (writable)
85 * The read or write access permissions to the backing store to be
86 * granted to the frontend.
91 * Data used by the backend driver to locate and configure the backing
92 * device. The format and semantics of this data vary according to the
93 * backing device in use and are outside the scope of this specification.
96 * Values: "file", "phy", "tap"
98 * The type of the backing device/object.
100 *--------------------------------- Features ---------------------------------
103 * Values: 0/1 (boolean)
106 * A value of "1" indicates that the backend can process requests
107 * containing the BLKIF_OP_WRITE_BARRIER request opcode. Requests
108 * of this type may still be returned at any time with the
109 * BLKIF_RSP_EOPNOTSUPP result code.
111 * feature-flush-cache
112 * Values: 0/1 (boolean)
115 * A value of "1" indicates that the backend can process requests
116 * containing the BLKIF_OP_FLUSH_DISKCACHE request opcode. Requests
117 * of this type may still be returned at any time with the
118 * BLKIF_RSP_EOPNOTSUPP result code.
121 * Values: 0/1 (boolean)
124 * A value of "1" indicates that the backend can process requests
125 * containing the BLKIF_OP_DISCARD request opcode. Requests
126 * of this type may still be returned at any time with the
127 * BLKIF_RSP_EOPNOTSUPP result code.
129 *----------------------- Request Transport Parameters ------------------------
131 * max-ring-page-order
136 * The maximum supported size of the request ring buffer in units of
137 * lb(machine pages). (e.g. 0 == 1 page, 1 = 2 pages, 2 == 4 pages,
143 * Notes: DEPRECATED, 2, 3
145 * The maximum supported size of the request ring buffer in units of
146 * machine pages. The value must be a power of 2.
148 *------------------------- Backend Device Properties -------------------------
155 * The offset, in bytes from the beginning of the virtual block device,
156 * to the first, addressable, discard extent on the underlying device.
158 * discard-granularity
160 * Default Value: <"sector-size">
163 * The size, in bytes, of the individually addressable discard extents
164 * of the underlying device.
167 * Values: 0/1 (boolean)
170 * A value of "1" indicates that the backend can process BLKIF_OP_DISCARD
171 * requests with the BLKIF_DISCARD_SECURE flag set.
174 * Values: <uint32_t> (bitmap)
176 * A collection of bit flags describing attributes of the backing
177 * device. The VDISK_* macros define the meaning of each bit
183 * The size, in bytes, of the individually addressible data blocks
184 * on the backend device.
189 * The size of the backend device, expressed in units of its native
190 * sector size ("sector-size").
192 *****************************************************************************
193 * Frontend XenBus Nodes
194 *****************************************************************************
196 *----------------------- Request Transport Parameters -----------------------
201 * The identifier of the Xen event channel used to signal activity
202 * in the ring buffer.
208 * The Xen grant reference granting permission for the backend to map
209 * the sole page in a single page sized ring buffer.
215 * For a frontend providing a multi-page ring, a "number of ring pages"
216 * sized list of nodes, each containing a Xen grant reference granting
217 * permission for the backend to map the page of the ring located
218 * at page index "%u". Page indexes are zero based.
221 * Values: string (XEN_IO_PROTO_ABI_*)
222 * Default Value: XEN_IO_PROTO_ABI_NATIVE
224 * The machine ABI rules governing the format of all ring request and
225 * response structures.
230 * Maximum Value: MAX(ffs(max-ring-pages) - 1, max-ring-page-order)
233 * The size of the frontend allocated request ring buffer in units
234 * of lb(machine pages). (e.g. 0 == 1 page, 1 = 2 pages, 2 == 4 pages,
240 * Maximum Value: MAX(max-ring-pages,(0x1 << max-ring-page-order))
241 * Notes: DEPRECATED, 2, 3
243 * The size of the frontend allocated request ring buffer in units of
244 * machine pages. The value must be a power of 2.
246 *------------------------- Virtual Device Properties -------------------------
249 * Values: "disk", "cdrom", "floppy", etc.
254 * A value indicating the physical device to virtualize within the
255 * frontend's domain. (e.g. "The first ATA disk", "The third SCSI
258 * See docs/misc/vbd-interface.txt for details on the format of this
263 * (1) Multi-page ring buffer scheme first developed in the Citrix XenServer
265 * (2) Multi-page ring buffer scheme first used in some Red Hat distributions
266 * including a distribution deployed on certain nodes of the Amazon
268 * (3) Support for multi-page ring buffers was implemented independently,
269 * in slightly different forms, by both Citrix and Red Hat/Amazon.
270 * For full interoperability, block front and backends should publish
271 * identical ring parameters, adjusted for unit differences, to the
272 * XenStore nodes used in both schemes.
273 * (4) Devices that support discard functionality may internally allocate
274 * space (discardable extents) in units that are larger than the
275 * exported logical block size.
276 * (5) The discard-alignment parameter allows a physical device to be
277 * partitioned into virtual devices that do not necessarily begin or
278 * end on a discardable extent boundary.
279 * (6) When there is only a single page allocated to the request ring,
280 * 'ring-ref' is used to communicate the grant reference for this
281 * page to the backend. When using a multi-page ring, the 'ring-ref'
282 * node is not created. Instead 'ring-ref0' - 'ring-refN' are used.
288 *****************************************************************************
290 *****************************************************************************
292 * Tool stack creates front and back nodes with state XenbusStateInitialising.
295 * ================================= =====================================
296 * XenbusStateInitialising XenbusStateInitialising
297 * o Query virtual device o Query backend device identification
299 * o Setup OS device instance. o Open and validate backend device.
300 * o Publish backend features and
301 * transport parameters.
305 * XenbusStateInitWait
307 * o Query backend features and
308 * transport parameters.
309 * o Allocate and initialize the
311 * o Publish transport parameters
312 * that will be in effect during
317 * XenbusStateInitialised
319 * o Query frontend transport parameters.
320 * o Connect to the request ring and
322 * o Publish backend device properties.
326 * XenbusStateConnected
328 * o Query backend device properties.
329 * o Finalize OS virtual device
334 * XenbusStateConnected
336 * Note: Drivers that do not support any optional features, or the negotiation
337 * of transport parameters, can skip certain states in the state machine:
339 * o A frontend may transition to XenbusStateInitialised without
340 * waiting for the backend to enter XenbusStateInitWait. In this
341 * case, default transport parameters are in effect and any
342 * transport parameters published by the frontend must contain
343 * their default values.
345 * o A backend may transition to XenbusStateInitialised, bypassing
346 * XenbusStateInitWait, without waiting for the frontend to first
347 * enter the XenbusStateInitialised state. In this case, default
348 * transport parameters are in effect and any transport parameters
349 * published by the backend must contain their default values.
351 * Drivers that support optional features and/or transport parameter
352 * negotiation must tolerate these additional state transition paths.
353 * In general this means performing the work of any skipped state
354 * transition, if it has not already been performed, in addition to the
355 * work associated with entry into the current state.
361 #define BLKIF_OP_READ 0
362 #define BLKIF_OP_WRITE 1
364 * All writes issued prior to a request with the BLKIF_OP_WRITE_BARRIER
365 * operation code ("barrier request") must be completed prior to the
366 * execution of the barrier request. All writes issued after the barrier
367 * request must not execute until after the completion of the barrier request.
369 * Optional. See "feature-barrier" XenBus node documentation above.
371 #define BLKIF_OP_WRITE_BARRIER 2
373 * Commit any uncommitted contents of the backing device's volatile cache
376 * Optional. See "feature-flush-cache" XenBus node documentation above.
378 #define BLKIF_OP_FLUSH_DISKCACHE 3
380 * Used in SLES sources for device specific command packet
381 * contained within the request. Reserved for that purpose.
383 #define BLKIF_OP_RESERVED_1 4
385 * Indicate to the backend device that a region of storage is no longer in
386 * use, and may be discarded at any time without impact to the client. If
387 * the BLKIF_DISCARD_SECURE flag is set on the request, all copies of the
388 * discarded region on the device must be rendered unrecoverable before the
391 * This operation is analogous to performing a trim (ATA) or unmap (SCSI),
392 * command on a native device.
394 * More information about trim/unmap operations can be found at:
395 * http://t13.org/Documents/UploadedDocuments/docs2008/
396 * e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc
397 * http://www.seagate.com/staticfiles/support/disc/manuals/
398 * Interface%20manuals/100293068c.pdf
400 * Optional. See "feature-discard", "discard-alignment",
401 * "discard-granularity", and "discard-secure" in the XenBus node
402 * documentation above.
404 #define BLKIF_OP_DISCARD 5
407 * Maximum scatter/gather segments per request.
408 * This is carefully chosen so that sizeof(blkif_ring_t) <= PAGE_SIZE.
409 * NB. This could be 12 if the ring indexes weren't stored in the same page.
411 #define BLKIF_MAX_SEGMENTS_PER_REQUEST 11
414 * NB. first_sect and last_sect in blkif_request_segment, as well as
415 * sector_number in blkif_request, are always expressed in 512-byte units.
416 * However they must be properly aligned to the real sector size of the
417 * physical disk, which is reported in the "sector-size" node in the backend
418 * xenbus info. Also the xenbus "sectors" node is expressed in 512-byte units.
420 struct blkif_request_segment {
421 grant_ref_t gref; /* reference to I/O buffer frame */
422 /* @first_sect: first sector in frame to transfer (inclusive). */
423 /* @last_sect: last sector in frame to transfer (inclusive). */
424 uint8_t first_sect, last_sect;
426 typedef struct blkif_request_segment blkif_request_segment_t;
429 * Starting ring element for any I/O request.
431 struct blkif_request {
432 uint8_t operation; /* BLKIF_OP_??? */
433 uint8_t nr_segments; /* number of segments */
434 blkif_vdev_t handle; /* only for read/write requests */
435 uint64_t id; /* private guest value, echoed in resp */
436 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
437 blkif_request_segment_t seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
439 typedef struct blkif_request blkif_request_t;
442 * Cast to this structure when blkif_request.operation == BLKIF_OP_DISCARD
443 * sizeof(struct blkif_request_discard) <= sizeof(struct blkif_request)
445 struct blkif_request_discard {
446 uint8_t operation; /* BLKIF_OP_DISCARD */
447 uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */
448 #define BLKIF_DISCARD_SECURE (1<<0) /* ignored if discard-secure=0 */
449 blkif_vdev_t handle; /* same as for read/write requests */
450 uint64_t id; /* private guest value, echoed in resp */
451 blkif_sector_t sector_number;/* start sector idx on disk */
452 uint64_t nr_sectors; /* number of contiguous sectors to discard*/
454 typedef struct blkif_request_discard blkif_request_discard_t;
456 struct blkif_response {
457 uint64_t id; /* copied from request */
458 uint8_t operation; /* copied from request */
459 int16_t status; /* BLKIF_RSP_??? */
461 typedef struct blkif_response blkif_response_t;
464 * STATUS RETURN CODES.
466 /* Operation not supported (only happens on barrier writes). */
467 #define BLKIF_RSP_EOPNOTSUPP -2
468 /* Operation failed for some unspecified reason (-EIO). */
469 #define BLKIF_RSP_ERROR -1
470 /* Operation completed successfully. */
471 #define BLKIF_RSP_OKAY 0
474 * Generate blkif ring structures and types.
476 DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response);
478 #define VDISK_CDROM 0x1
479 #define VDISK_REMOVABLE 0x2
480 #define VDISK_READONLY 0x4
482 #endif /* __XEN_PUBLIC_IO_BLKIF_H__ */
490 * indent-tabs-mode: nil