18ac98aedSDavid Woodhouse /* SPDX-License-Identifier: MIT */ 2a3434a2dSAnthony PERARD /****************************************************************************** 3a3434a2dSAnthony PERARD * blkif.h 4a3434a2dSAnthony PERARD * 5a3434a2dSAnthony PERARD * Unified block-device I/O interface for Xen guest OSes. 6a3434a2dSAnthony PERARD * 7a3434a2dSAnthony PERARD * Copyright (c) 2003-2004, Keir Fraser 8a3434a2dSAnthony PERARD * Copyright (c) 2012, Spectra Logic Corporation 9a3434a2dSAnthony PERARD */ 10a3434a2dSAnthony PERARD 11a3434a2dSAnthony PERARD #ifndef __XEN_PUBLIC_IO_BLKIF_H__ 12a3434a2dSAnthony PERARD #define __XEN_PUBLIC_IO_BLKIF_H__ 13a3434a2dSAnthony PERARD 14a3434a2dSAnthony PERARD #include "ring.h" 15a3434a2dSAnthony PERARD #include "../grant_table.h" 16a3434a2dSAnthony PERARD 17a3434a2dSAnthony PERARD /* 18a3434a2dSAnthony PERARD * Front->back notifications: When enqueuing a new request, sending a 19a3434a2dSAnthony PERARD * notification can be made conditional on req_event (i.e., the generic 20a3434a2dSAnthony PERARD * hold-off mechanism provided by the ring macros). Backends must set 21a3434a2dSAnthony PERARD * req_event appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()). 22a3434a2dSAnthony PERARD * 23a3434a2dSAnthony PERARD * Back->front notifications: When enqueuing a new response, sending a 24a3434a2dSAnthony PERARD * notification can be made conditional on rsp_event (i.e., the generic 25a3434a2dSAnthony PERARD * hold-off mechanism provided by the ring macros). Frontends must set 26a3434a2dSAnthony PERARD * rsp_event appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()). 27a3434a2dSAnthony PERARD */ 28a3434a2dSAnthony PERARD 29a3434a2dSAnthony PERARD #ifndef blkif_vdev_t 30a3434a2dSAnthony PERARD #define blkif_vdev_t uint16_t 31a3434a2dSAnthony PERARD #endif 32a3434a2dSAnthony PERARD #define blkif_sector_t uint64_t 33a3434a2dSAnthony PERARD 34a3434a2dSAnthony PERARD /* 35a3434a2dSAnthony PERARD * Feature and Parameter Negotiation 36a3434a2dSAnthony PERARD * ================================= 37a3434a2dSAnthony PERARD * The two halves of a Xen block driver utilize nodes within the XenStore to 38a3434a2dSAnthony PERARD * communicate capabilities and to negotiate operating parameters. This 39a3434a2dSAnthony PERARD * section enumerates these nodes which reside in the respective front and 40a3434a2dSAnthony PERARD * backend portions of the XenStore, following the XenBus convention. 41a3434a2dSAnthony PERARD * 42a3434a2dSAnthony PERARD * All data in the XenStore is stored as strings. Nodes specifying numeric 43a3434a2dSAnthony PERARD * values are encoded in decimal. Integer value ranges listed below are 44a3434a2dSAnthony PERARD * expressed as fixed sized integer types capable of storing the conversion 45a3434a2dSAnthony PERARD * of a properly formated node string, without loss of information. 46a3434a2dSAnthony PERARD * 47a3434a2dSAnthony PERARD * Any specified default value is in effect if the corresponding XenBus node 48a3434a2dSAnthony PERARD * is not present in the XenStore. 49a3434a2dSAnthony PERARD * 50a3434a2dSAnthony PERARD * XenStore nodes in sections marked "PRIVATE" are solely for use by the 51a3434a2dSAnthony PERARD * driver side whose XenBus tree contains them. 52a3434a2dSAnthony PERARD * 53a3434a2dSAnthony PERARD * XenStore nodes marked "DEPRECATED" in their notes section should only be 54a3434a2dSAnthony PERARD * used to provide interoperability with legacy implementations. 55a3434a2dSAnthony PERARD * 56a3434a2dSAnthony PERARD * See the XenBus state transition diagram below for details on when XenBus 57a3434a2dSAnthony PERARD * nodes must be published and when they can be queried. 58a3434a2dSAnthony PERARD * 59a3434a2dSAnthony PERARD ***************************************************************************** 60a3434a2dSAnthony PERARD * Backend XenBus Nodes 61a3434a2dSAnthony PERARD ***************************************************************************** 62a3434a2dSAnthony PERARD * 63a3434a2dSAnthony PERARD *------------------ Backend Device Identification (PRIVATE) ------------------ 64a3434a2dSAnthony PERARD * 65a3434a2dSAnthony PERARD * mode 66a3434a2dSAnthony PERARD * Values: "r" (read only), "w" (writable) 67a3434a2dSAnthony PERARD * 68a3434a2dSAnthony PERARD * The read or write access permissions to the backing store to be 69a3434a2dSAnthony PERARD * granted to the frontend. 70a3434a2dSAnthony PERARD * 71a3434a2dSAnthony PERARD * params 72a3434a2dSAnthony PERARD * Values: string 73a3434a2dSAnthony PERARD * 74a3434a2dSAnthony PERARD * A free formatted string providing sufficient information for the 75a3434a2dSAnthony PERARD * hotplug script to attach the device and provide a suitable 76a3434a2dSAnthony PERARD * handler (ie: a block device) for blkback to use. 77a3434a2dSAnthony PERARD * 78a3434a2dSAnthony PERARD * physical-device 79a3434a2dSAnthony PERARD * Values: "MAJOR:MINOR" 80a3434a2dSAnthony PERARD * Notes: 11 81a3434a2dSAnthony PERARD * 82a3434a2dSAnthony PERARD * MAJOR and MINOR are the major number and minor number of the 83a3434a2dSAnthony PERARD * backing device respectively. 84a3434a2dSAnthony PERARD * 85a3434a2dSAnthony PERARD * physical-device-path 86a3434a2dSAnthony PERARD * Values: path string 87a3434a2dSAnthony PERARD * 88a3434a2dSAnthony PERARD * A string that contains the absolute path to the disk image. On 89a3434a2dSAnthony PERARD * NetBSD and Linux this is always a block device, while on FreeBSD 90a3434a2dSAnthony PERARD * it can be either a block device or a regular file. 91a3434a2dSAnthony PERARD * 92a3434a2dSAnthony PERARD * type 93a3434a2dSAnthony PERARD * Values: "file", "phy", "tap" 94a3434a2dSAnthony PERARD * 95a3434a2dSAnthony PERARD * The type of the backing device/object. 96a3434a2dSAnthony PERARD * 97a3434a2dSAnthony PERARD * 98a3434a2dSAnthony PERARD * direct-io-safe 99a3434a2dSAnthony PERARD * Values: 0/1 (boolean) 100a3434a2dSAnthony PERARD * Default Value: 0 101a3434a2dSAnthony PERARD * 102a3434a2dSAnthony PERARD * The underlying storage is not affected by the direct IO memory 103a3434a2dSAnthony PERARD * lifetime bug. See: 10450c88402SJoao Martins * https://lists.xen.org/archives/html/xen-devel/2012-12/msg01154.html 105a3434a2dSAnthony PERARD * 106a3434a2dSAnthony PERARD * Therefore this option gives the backend permission to use 107a3434a2dSAnthony PERARD * O_DIRECT, notwithstanding that bug. 108a3434a2dSAnthony PERARD * 109a3434a2dSAnthony PERARD * That is, if this option is enabled, use of O_DIRECT is safe, 110a3434a2dSAnthony PERARD * in circumstances where we would normally have avoided it as a 111a3434a2dSAnthony PERARD * workaround for that bug. This option is not relevant for all 112a3434a2dSAnthony PERARD * backends, and even not necessarily supported for those for 113a3434a2dSAnthony PERARD * which it is relevant. A backend which knows that it is not 114a3434a2dSAnthony PERARD * affected by the bug can ignore this option. 115a3434a2dSAnthony PERARD * 116a3434a2dSAnthony PERARD * This option doesn't require a backend to use O_DIRECT, so it 117a3434a2dSAnthony PERARD * should not be used to try to control the caching behaviour. 118a3434a2dSAnthony PERARD * 119a3434a2dSAnthony PERARD *--------------------------------- Features --------------------------------- 120a3434a2dSAnthony PERARD * 121a3434a2dSAnthony PERARD * feature-barrier 122a3434a2dSAnthony PERARD * Values: 0/1 (boolean) 123a3434a2dSAnthony PERARD * Default Value: 0 124a3434a2dSAnthony PERARD * 125a3434a2dSAnthony PERARD * A value of "1" indicates that the backend can process requests 126a3434a2dSAnthony PERARD * containing the BLKIF_OP_WRITE_BARRIER request opcode. Requests 127a3434a2dSAnthony PERARD * of this type may still be returned at any time with the 128a3434a2dSAnthony PERARD * BLKIF_RSP_EOPNOTSUPP result code. 129a3434a2dSAnthony PERARD * 130a3434a2dSAnthony PERARD * feature-flush-cache 131a3434a2dSAnthony PERARD * Values: 0/1 (boolean) 132a3434a2dSAnthony PERARD * Default Value: 0 133a3434a2dSAnthony PERARD * 134a3434a2dSAnthony PERARD * A value of "1" indicates that the backend can process requests 135a3434a2dSAnthony PERARD * containing the BLKIF_OP_FLUSH_DISKCACHE request opcode. Requests 136a3434a2dSAnthony PERARD * of this type may still be returned at any time with the 137a3434a2dSAnthony PERARD * BLKIF_RSP_EOPNOTSUPP result code. 138a3434a2dSAnthony PERARD * 139a3434a2dSAnthony PERARD * feature-discard 140a3434a2dSAnthony PERARD * Values: 0/1 (boolean) 141a3434a2dSAnthony PERARD * Default Value: 0 142a3434a2dSAnthony PERARD * 143a3434a2dSAnthony PERARD * A value of "1" indicates that the backend can process requests 144a3434a2dSAnthony PERARD * containing the BLKIF_OP_DISCARD request opcode. Requests 145a3434a2dSAnthony PERARD * of this type may still be returned at any time with the 146a3434a2dSAnthony PERARD * BLKIF_RSP_EOPNOTSUPP result code. 147a3434a2dSAnthony PERARD * 148a3434a2dSAnthony PERARD * feature-persistent 149a3434a2dSAnthony PERARD * Values: 0/1 (boolean) 150a3434a2dSAnthony PERARD * Default Value: 0 151a3434a2dSAnthony PERARD * Notes: 7 152a3434a2dSAnthony PERARD * 153a3434a2dSAnthony PERARD * A value of "1" indicates that the backend can keep the grants used 154a3434a2dSAnthony PERARD * by the frontend driver mapped, so the same set of grants should be 155a3434a2dSAnthony PERARD * used in all transactions. The maximum number of grants the backend 156a3434a2dSAnthony PERARD * can map persistently depends on the implementation, but ideally it 157a3434a2dSAnthony PERARD * should be RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST. Using this 158a3434a2dSAnthony PERARD * feature the backend doesn't need to unmap each grant, preventing 159a3434a2dSAnthony PERARD * costly TLB flushes. The backend driver should only map grants 160a3434a2dSAnthony PERARD * persistently if the frontend supports it. If a backend driver chooses 161a3434a2dSAnthony PERARD * to use the persistent protocol when the frontend doesn't support it, 162a3434a2dSAnthony PERARD * it will probably hit the maximum number of persistently mapped grants 163a3434a2dSAnthony PERARD * (due to the fact that the frontend won't be reusing the same grants), 164a3434a2dSAnthony PERARD * and fall back to non-persistent mode. Backend implementations may 165a3434a2dSAnthony PERARD * shrink or expand the number of persistently mapped grants without 166a3434a2dSAnthony PERARD * notifying the frontend depending on memory constraints (this might 167a3434a2dSAnthony PERARD * cause a performance degradation). 168a3434a2dSAnthony PERARD * 169a3434a2dSAnthony PERARD * If a backend driver wants to limit the maximum number of persistently 170a3434a2dSAnthony PERARD * mapped grants to a value less than RING_SIZE * 171a3434a2dSAnthony PERARD * BLKIF_MAX_SEGMENTS_PER_REQUEST a LRU strategy should be used to 172a3434a2dSAnthony PERARD * discard the grants that are less commonly used. Using a LRU in the 173a3434a2dSAnthony PERARD * backend driver paired with a LIFO queue in the frontend will 174a3434a2dSAnthony PERARD * allow us to have better performance in this scenario. 175a3434a2dSAnthony PERARD * 176a3434a2dSAnthony PERARD *----------------------- Request Transport Parameters ------------------------ 177a3434a2dSAnthony PERARD * 178a3434a2dSAnthony PERARD * max-ring-page-order 179a3434a2dSAnthony PERARD * Values: <uint32_t> 180a3434a2dSAnthony PERARD * Default Value: 0 181a3434a2dSAnthony PERARD * Notes: 1, 3 182a3434a2dSAnthony PERARD * 183a3434a2dSAnthony PERARD * The maximum supported size of the request ring buffer in units of 184a3434a2dSAnthony PERARD * lb(machine pages). (e.g. 0 == 1 page, 1 = 2 pages, 2 == 4 pages, 185a3434a2dSAnthony PERARD * etc.). 186a3434a2dSAnthony PERARD * 187a3434a2dSAnthony PERARD * max-ring-pages 188a3434a2dSAnthony PERARD * Values: <uint32_t> 189a3434a2dSAnthony PERARD * Default Value: 1 190a3434a2dSAnthony PERARD * Notes: DEPRECATED, 2, 3 191a3434a2dSAnthony PERARD * 192a3434a2dSAnthony PERARD * The maximum supported size of the request ring buffer in units of 193a3434a2dSAnthony PERARD * machine pages. The value must be a power of 2. 194a3434a2dSAnthony PERARD * 195a3434a2dSAnthony PERARD *------------------------- Backend Device Properties ------------------------- 196a3434a2dSAnthony PERARD * 197a3434a2dSAnthony PERARD * discard-enable 198a3434a2dSAnthony PERARD * Values: 0/1 (boolean) 199a3434a2dSAnthony PERARD * Default Value: 1 200a3434a2dSAnthony PERARD * 201a3434a2dSAnthony PERARD * This optional property, set by the toolstack, instructs the backend 202a3434a2dSAnthony PERARD * to offer (or not to offer) discard to the frontend. If the property 203a3434a2dSAnthony PERARD * is missing the backend should offer discard if the backing storage 204a3434a2dSAnthony PERARD * actually supports it. 205a3434a2dSAnthony PERARD * 206a3434a2dSAnthony PERARD * discard-alignment 207a3434a2dSAnthony PERARD * Values: <uint32_t> 208a3434a2dSAnthony PERARD * Default Value: 0 209a3434a2dSAnthony PERARD * Notes: 4, 5 210a3434a2dSAnthony PERARD * 211a3434a2dSAnthony PERARD * The offset, in bytes from the beginning of the virtual block device, 212a3434a2dSAnthony PERARD * to the first, addressable, discard extent on the underlying device. 213a3434a2dSAnthony PERARD * 214a3434a2dSAnthony PERARD * discard-granularity 215a3434a2dSAnthony PERARD * Values: <uint32_t> 216a3434a2dSAnthony PERARD * Default Value: <"sector-size"> 217a3434a2dSAnthony PERARD * Notes: 4 218a3434a2dSAnthony PERARD * 219a3434a2dSAnthony PERARD * The size, in bytes, of the individually addressable discard extents 220a3434a2dSAnthony PERARD * of the underlying device. 221a3434a2dSAnthony PERARD * 222a3434a2dSAnthony PERARD * discard-secure 223a3434a2dSAnthony PERARD * Values: 0/1 (boolean) 224a3434a2dSAnthony PERARD * Default Value: 0 225a3434a2dSAnthony PERARD * Notes: 10 226a3434a2dSAnthony PERARD * 227a3434a2dSAnthony PERARD * A value of "1" indicates that the backend can process BLKIF_OP_DISCARD 228a3434a2dSAnthony PERARD * requests with the BLKIF_DISCARD_SECURE flag set. 229a3434a2dSAnthony PERARD * 230a3434a2dSAnthony PERARD * info 231a3434a2dSAnthony PERARD * Values: <uint32_t> (bitmap) 232a3434a2dSAnthony PERARD * 233a3434a2dSAnthony PERARD * A collection of bit flags describing attributes of the backing 234a3434a2dSAnthony PERARD * device. The VDISK_* macros define the meaning of each bit 235a3434a2dSAnthony PERARD * location. 236a3434a2dSAnthony PERARD * 237a3434a2dSAnthony PERARD * sector-size 238a3434a2dSAnthony PERARD * Values: <uint32_t> 239a3434a2dSAnthony PERARD * 240a3434a2dSAnthony PERARD * The logical block size, in bytes, of the underlying storage. This 241a3434a2dSAnthony PERARD * must be a power of two with a minimum value of 512. 242a3434a2dSAnthony PERARD * 243a3434a2dSAnthony PERARD * NOTE: Because of implementation bugs in some frontends this must be 244a3434a2dSAnthony PERARD * set to 512, unless the frontend advertizes a non-zero value 245a3434a2dSAnthony PERARD * in its "feature-large-sector-size" xenbus node. (See below). 246a3434a2dSAnthony PERARD * 247a3434a2dSAnthony PERARD * physical-sector-size 248a3434a2dSAnthony PERARD * Values: <uint32_t> 249a3434a2dSAnthony PERARD * Default Value: <"sector-size"> 250a3434a2dSAnthony PERARD * 251a3434a2dSAnthony PERARD * The physical block size, in bytes, of the backend storage. This 252a3434a2dSAnthony PERARD * must be an integer multiple of "sector-size". 253a3434a2dSAnthony PERARD * 254a3434a2dSAnthony PERARD * sectors 255a3434a2dSAnthony PERARD * Values: <uint64_t> 256a3434a2dSAnthony PERARD * 257a3434a2dSAnthony PERARD * The size of the backend device, expressed in units of "sector-size". 258a3434a2dSAnthony PERARD * The product of "sector-size" and "sectors" must also be an integer 259a3434a2dSAnthony PERARD * multiple of "physical-sector-size", if that node is present. 260a3434a2dSAnthony PERARD * 261a3434a2dSAnthony PERARD ***************************************************************************** 262a3434a2dSAnthony PERARD * Frontend XenBus Nodes 263a3434a2dSAnthony PERARD ***************************************************************************** 264a3434a2dSAnthony PERARD * 265a3434a2dSAnthony PERARD *----------------------- Request Transport Parameters ----------------------- 266a3434a2dSAnthony PERARD * 267a3434a2dSAnthony PERARD * event-channel 268a3434a2dSAnthony PERARD * Values: <uint32_t> 269a3434a2dSAnthony PERARD * 270a3434a2dSAnthony PERARD * The identifier of the Xen event channel used to signal activity 271a3434a2dSAnthony PERARD * in the ring buffer. 272a3434a2dSAnthony PERARD * 273a3434a2dSAnthony PERARD * ring-ref 274a3434a2dSAnthony PERARD * Values: <uint32_t> 275a3434a2dSAnthony PERARD * Notes: 6 276a3434a2dSAnthony PERARD * 277a3434a2dSAnthony PERARD * The Xen grant reference granting permission for the backend to map 278a3434a2dSAnthony PERARD * the sole page in a single page sized ring buffer. 279a3434a2dSAnthony PERARD * 280a3434a2dSAnthony PERARD * ring-ref%u 281a3434a2dSAnthony PERARD * Values: <uint32_t> 282a3434a2dSAnthony PERARD * Notes: 6 283a3434a2dSAnthony PERARD * 284a3434a2dSAnthony PERARD * For a frontend providing a multi-page ring, a "number of ring pages" 285a3434a2dSAnthony PERARD * sized list of nodes, each containing a Xen grant reference granting 286a3434a2dSAnthony PERARD * permission for the backend to map the page of the ring located 287a3434a2dSAnthony PERARD * at page index "%u". Page indexes are zero based. 288a3434a2dSAnthony PERARD * 289a3434a2dSAnthony PERARD * protocol 290a3434a2dSAnthony PERARD * Values: string (XEN_IO_PROTO_ABI_*) 291a3434a2dSAnthony PERARD * Default Value: XEN_IO_PROTO_ABI_NATIVE 292a3434a2dSAnthony PERARD * 293a3434a2dSAnthony PERARD * The machine ABI rules governing the format of all ring request and 294a3434a2dSAnthony PERARD * response structures. 295a3434a2dSAnthony PERARD * 296a3434a2dSAnthony PERARD * ring-page-order 297a3434a2dSAnthony PERARD * Values: <uint32_t> 298a3434a2dSAnthony PERARD * Default Value: 0 299a3434a2dSAnthony PERARD * Maximum Value: MAX(ffs(max-ring-pages) - 1, max-ring-page-order) 300a3434a2dSAnthony PERARD * Notes: 1, 3 301a3434a2dSAnthony PERARD * 302a3434a2dSAnthony PERARD * The size of the frontend allocated request ring buffer in units 303a3434a2dSAnthony PERARD * of lb(machine pages). (e.g. 0 == 1 page, 1 = 2 pages, 2 == 4 pages, 304a3434a2dSAnthony PERARD * etc.). 305a3434a2dSAnthony PERARD * 306a3434a2dSAnthony PERARD * num-ring-pages 307a3434a2dSAnthony PERARD * Values: <uint32_t> 308a3434a2dSAnthony PERARD * Default Value: 1 309a3434a2dSAnthony PERARD * Maximum Value: MAX(max-ring-pages,(0x1 << max-ring-page-order)) 310a3434a2dSAnthony PERARD * Notes: DEPRECATED, 2, 3 311a3434a2dSAnthony PERARD * 312a3434a2dSAnthony PERARD * The size of the frontend allocated request ring buffer in units of 313a3434a2dSAnthony PERARD * machine pages. The value must be a power of 2. 314a3434a2dSAnthony PERARD * 315a3434a2dSAnthony PERARD *--------------------------------- Features --------------------------------- 316a3434a2dSAnthony PERARD * 317a3434a2dSAnthony PERARD * feature-persistent 318a3434a2dSAnthony PERARD * Values: 0/1 (boolean) 319a3434a2dSAnthony PERARD * Default Value: 0 320a3434a2dSAnthony PERARD * Notes: 7, 8, 9 321a3434a2dSAnthony PERARD * 322a3434a2dSAnthony PERARD * A value of "1" indicates that the frontend will reuse the same grants 323a3434a2dSAnthony PERARD * for all transactions, allowing the backend to map them with write 324a3434a2dSAnthony PERARD * access (even when it should be read-only). If the frontend hits the 325a3434a2dSAnthony PERARD * maximum number of allowed persistently mapped grants, it can fallback 326a3434a2dSAnthony PERARD * to non persistent mode. This will cause a performance degradation, 32750c88402SJoao Martins * since the the backend driver will still try to map those grants 328a3434a2dSAnthony PERARD * persistently. Since the persistent grants protocol is compatible with 329a3434a2dSAnthony PERARD * the previous protocol, a frontend driver can choose to work in 330a3434a2dSAnthony PERARD * persistent mode even when the backend doesn't support it. 331a3434a2dSAnthony PERARD * 332a3434a2dSAnthony PERARD * It is recommended that the frontend driver stores the persistently 333a3434a2dSAnthony PERARD * mapped grants in a LIFO queue, so a subset of all persistently mapped 334a3434a2dSAnthony PERARD * grants gets used commonly. This is done in case the backend driver 335a3434a2dSAnthony PERARD * decides to limit the maximum number of persistently mapped grants 336a3434a2dSAnthony PERARD * to a value less than RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST. 337a3434a2dSAnthony PERARD * 338a3434a2dSAnthony PERARD * feature-large-sector-size 339a3434a2dSAnthony PERARD * Values: 0/1 (boolean) 340a3434a2dSAnthony PERARD * Default Value: 0 341a3434a2dSAnthony PERARD * 342a3434a2dSAnthony PERARD * A value of "1" indicates that the frontend will correctly supply and 343a3434a2dSAnthony PERARD * interpret all sector-based quantities in terms of the "sector-size" 344a3434a2dSAnthony PERARD * value supplied in the backend info, whatever that may be set to. 345a3434a2dSAnthony PERARD * If this node is not present or its value is "0" then it is assumed 346a3434a2dSAnthony PERARD * that the frontend requires that the logical block size is 512 as it 347a3434a2dSAnthony PERARD * is hardcoded (which is the case in some frontend implementations). 348a3434a2dSAnthony PERARD * 3498ac98aedSDavid Woodhouse * trusted 3508ac98aedSDavid Woodhouse * Values: 0/1 (boolean) 3518ac98aedSDavid Woodhouse * Default value: 1 3528ac98aedSDavid Woodhouse * 3538ac98aedSDavid Woodhouse * A value of "0" indicates that the frontend should not trust the 3548ac98aedSDavid Woodhouse * backend, and should deploy whatever measures available to protect from 3558ac98aedSDavid Woodhouse * a malicious backend on the other end. 3568ac98aedSDavid Woodhouse * 357a3434a2dSAnthony PERARD *------------------------- Virtual Device Properties ------------------------- 358a3434a2dSAnthony PERARD * 359a3434a2dSAnthony PERARD * device-type 360a3434a2dSAnthony PERARD * Values: "disk", "cdrom", "floppy", etc. 361a3434a2dSAnthony PERARD * 362a3434a2dSAnthony PERARD * virtual-device 363a3434a2dSAnthony PERARD * Values: <uint32_t> 364a3434a2dSAnthony PERARD * 365a3434a2dSAnthony PERARD * A value indicating the physical device to virtualize within the 366a3434a2dSAnthony PERARD * frontend's domain. (e.g. "The first ATA disk", "The third SCSI 367a3434a2dSAnthony PERARD * disk", etc.) 368a3434a2dSAnthony PERARD * 369a3434a2dSAnthony PERARD * See docs/misc/vbd-interface.txt for details on the format of this 370a3434a2dSAnthony PERARD * value. 371a3434a2dSAnthony PERARD * 372a3434a2dSAnthony PERARD * Notes 373a3434a2dSAnthony PERARD * ----- 374a3434a2dSAnthony PERARD * (1) Multi-page ring buffer scheme first developed in the Citrix XenServer 375a3434a2dSAnthony PERARD * PV drivers. 376a3434a2dSAnthony PERARD * (2) Multi-page ring buffer scheme first used in some RedHat distributions 377a3434a2dSAnthony PERARD * including a distribution deployed on certain nodes of the Amazon 378a3434a2dSAnthony PERARD * EC2 cluster. 379a3434a2dSAnthony PERARD * (3) Support for multi-page ring buffers was implemented independently, 380a3434a2dSAnthony PERARD * in slightly different forms, by both Citrix and RedHat/Amazon. 381a3434a2dSAnthony PERARD * For full interoperability, block front and backends should publish 382a3434a2dSAnthony PERARD * identical ring parameters, adjusted for unit differences, to the 383a3434a2dSAnthony PERARD * XenStore nodes used in both schemes. 384a3434a2dSAnthony PERARD * (4) Devices that support discard functionality may internally allocate space 385a3434a2dSAnthony PERARD * (discardable extents) in units that are larger than the exported logical 386a3434a2dSAnthony PERARD * block size. If the backing device has such discardable extents the 387a3434a2dSAnthony PERARD * backend should provide both discard-granularity and discard-alignment. 388a3434a2dSAnthony PERARD * Providing just one of the two may be considered an error by the frontend. 389a3434a2dSAnthony PERARD * Backends supporting discard should include discard-granularity and 390a3434a2dSAnthony PERARD * discard-alignment even if it supports discarding individual sectors. 391a3434a2dSAnthony PERARD * Frontends should assume discard-alignment == 0 and discard-granularity 392a3434a2dSAnthony PERARD * == sector size if these keys are missing. 393a3434a2dSAnthony PERARD * (5) The discard-alignment parameter allows a physical device to be 394a3434a2dSAnthony PERARD * partitioned into virtual devices that do not necessarily begin or 395a3434a2dSAnthony PERARD * end on a discardable extent boundary. 396a3434a2dSAnthony PERARD * (6) When there is only a single page allocated to the request ring, 397a3434a2dSAnthony PERARD * 'ring-ref' is used to communicate the grant reference for this 398a3434a2dSAnthony PERARD * page to the backend. When using a multi-page ring, the 'ring-ref' 399a3434a2dSAnthony PERARD * node is not created. Instead 'ring-ref0' - 'ring-refN' are used. 400a3434a2dSAnthony PERARD * (7) When using persistent grants data has to be copied from/to the page 401a3434a2dSAnthony PERARD * where the grant is currently mapped. The overhead of doing this copy 402a3434a2dSAnthony PERARD * however doesn't suppress the speed improvement of not having to unmap 403a3434a2dSAnthony PERARD * the grants. 404a3434a2dSAnthony PERARD * (8) The frontend driver has to allow the backend driver to map all grants 405a3434a2dSAnthony PERARD * with write access, even when they should be mapped read-only, since 406a3434a2dSAnthony PERARD * further requests may reuse these grants and require write permissions. 407a3434a2dSAnthony PERARD * (9) Linux implementation doesn't have a limit on the maximum number of 408a3434a2dSAnthony PERARD * grants that can be persistently mapped in the frontend driver, but 409a3434a2dSAnthony PERARD * due to the frontent driver implementation it should never be bigger 410a3434a2dSAnthony PERARD * than RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST. 411a3434a2dSAnthony PERARD *(10) The discard-secure property may be present and will be set to 1 if the 412a3434a2dSAnthony PERARD * backing device supports secure discard. 413a3434a2dSAnthony PERARD *(11) Only used by Linux and NetBSD. 414a3434a2dSAnthony PERARD */ 415a3434a2dSAnthony PERARD 416a3434a2dSAnthony PERARD /* 417a3434a2dSAnthony PERARD * Multiple hardware queues/rings: 418a3434a2dSAnthony PERARD * If supported, the backend will write the key "multi-queue-max-queues" to 419a3434a2dSAnthony PERARD * the directory for that vbd, and set its value to the maximum supported 420a3434a2dSAnthony PERARD * number of queues. 421a3434a2dSAnthony PERARD * Frontends that are aware of this feature and wish to use it can write the 422a3434a2dSAnthony PERARD * key "multi-queue-num-queues" with the number they wish to use, which must be 423a3434a2dSAnthony PERARD * greater than zero, and no more than the value reported by the backend in 424a3434a2dSAnthony PERARD * "multi-queue-max-queues". 425a3434a2dSAnthony PERARD * 426a3434a2dSAnthony PERARD * For frontends requesting just one queue, the usual event-channel and 427a3434a2dSAnthony PERARD * ring-ref keys are written as before, simplifying the backend processing 428a3434a2dSAnthony PERARD * to avoid distinguishing between a frontend that doesn't understand the 429a3434a2dSAnthony PERARD * multi-queue feature, and one that does, but requested only one queue. 430a3434a2dSAnthony PERARD * 431a3434a2dSAnthony PERARD * Frontends requesting two or more queues must not write the toplevel 432a3434a2dSAnthony PERARD * event-channel and ring-ref keys, instead writing those keys under sub-keys 433a3434a2dSAnthony PERARD * having the name "queue-N" where N is the integer ID of the queue/ring for 434a3434a2dSAnthony PERARD * which those keys belong. Queues are indexed from zero. 435a3434a2dSAnthony PERARD * For example, a frontend with two queues must write the following set of 436a3434a2dSAnthony PERARD * queue-related keys: 437a3434a2dSAnthony PERARD * 438a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/multi-queue-num-queues = "2" 439a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-0 = "" 440a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-0/ring-ref = "<ring-ref#0>" 441a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-0/event-channel = "<evtchn#0>" 442a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-1 = "" 443a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-1/ring-ref = "<ring-ref#1>" 444a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-1/event-channel = "<evtchn#1>" 445a3434a2dSAnthony PERARD * 446a3434a2dSAnthony PERARD * It is also possible to use multiple queues/rings together with 447a3434a2dSAnthony PERARD * feature multi-page ring buffer. 448a3434a2dSAnthony PERARD * For example, a frontend requests two queues/rings and the size of each ring 449a3434a2dSAnthony PERARD * buffer is two pages must write the following set of related keys: 450a3434a2dSAnthony PERARD * 451a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/multi-queue-num-queues = "2" 452a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/ring-page-order = "1" 453a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-0 = "" 454a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-0/ring-ref0 = "<ring-ref#0>" 455a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-0/ring-ref1 = "<ring-ref#1>" 456a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-0/event-channel = "<evtchn#0>" 457a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-1 = "" 458a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-1/ring-ref0 = "<ring-ref#2>" 459a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-1/ring-ref1 = "<ring-ref#3>" 460a3434a2dSAnthony PERARD * /local/domain/1/device/vbd/0/queue-1/event-channel = "<evtchn#1>" 461a3434a2dSAnthony PERARD * 462a3434a2dSAnthony PERARD */ 463a3434a2dSAnthony PERARD 464a3434a2dSAnthony PERARD /* 465a3434a2dSAnthony PERARD * STATE DIAGRAMS 466a3434a2dSAnthony PERARD * 467a3434a2dSAnthony PERARD ***************************************************************************** 468a3434a2dSAnthony PERARD * Startup * 469a3434a2dSAnthony PERARD ***************************************************************************** 470a3434a2dSAnthony PERARD * 471a3434a2dSAnthony PERARD * Tool stack creates front and back nodes with state XenbusStateInitialising. 472a3434a2dSAnthony PERARD * 473a3434a2dSAnthony PERARD * Front Back 474a3434a2dSAnthony PERARD * ================================= ===================================== 475a3434a2dSAnthony PERARD * XenbusStateInitialising XenbusStateInitialising 476a3434a2dSAnthony PERARD * o Query virtual device o Query backend device identification 477a3434a2dSAnthony PERARD * properties. data. 478a3434a2dSAnthony PERARD * o Setup OS device instance. o Open and validate backend device. 479a3434a2dSAnthony PERARD * o Publish backend features and 480a3434a2dSAnthony PERARD * transport parameters. 481a3434a2dSAnthony PERARD * | 482a3434a2dSAnthony PERARD * | 483a3434a2dSAnthony PERARD * V 484a3434a2dSAnthony PERARD * XenbusStateInitWait 485a3434a2dSAnthony PERARD * 486a3434a2dSAnthony PERARD * o Query backend features and 487a3434a2dSAnthony PERARD * transport parameters. 488a3434a2dSAnthony PERARD * o Allocate and initialize the 489a3434a2dSAnthony PERARD * request ring. 490a3434a2dSAnthony PERARD * o Publish transport parameters 491a3434a2dSAnthony PERARD * that will be in effect during 492a3434a2dSAnthony PERARD * this connection. 493a3434a2dSAnthony PERARD * | 494a3434a2dSAnthony PERARD * | 495a3434a2dSAnthony PERARD * V 496a3434a2dSAnthony PERARD * XenbusStateInitialised 497a3434a2dSAnthony PERARD * 498a3434a2dSAnthony PERARD * o Query frontend transport parameters. 499a3434a2dSAnthony PERARD * o Connect to the request ring and 500a3434a2dSAnthony PERARD * event channel. 501a3434a2dSAnthony PERARD * o Publish backend device properties. 502a3434a2dSAnthony PERARD * | 503a3434a2dSAnthony PERARD * | 504a3434a2dSAnthony PERARD * V 505a3434a2dSAnthony PERARD * XenbusStateConnected 506a3434a2dSAnthony PERARD * 507a3434a2dSAnthony PERARD * o Query backend device properties. 508a3434a2dSAnthony PERARD * o Finalize OS virtual device 509a3434a2dSAnthony PERARD * instance. 510a3434a2dSAnthony PERARD * | 511a3434a2dSAnthony PERARD * | 512a3434a2dSAnthony PERARD * V 513a3434a2dSAnthony PERARD * XenbusStateConnected 514a3434a2dSAnthony PERARD * 515a3434a2dSAnthony PERARD * Note: Drivers that do not support any optional features, or the negotiation 516a3434a2dSAnthony PERARD * of transport parameters, can skip certain states in the state machine: 517a3434a2dSAnthony PERARD * 518a3434a2dSAnthony PERARD * o A frontend may transition to XenbusStateInitialised without 519a3434a2dSAnthony PERARD * waiting for the backend to enter XenbusStateInitWait. In this 520a3434a2dSAnthony PERARD * case, default transport parameters are in effect and any 521a3434a2dSAnthony PERARD * transport parameters published by the frontend must contain 522a3434a2dSAnthony PERARD * their default values. 523a3434a2dSAnthony PERARD * 524a3434a2dSAnthony PERARD * o A backend may transition to XenbusStateInitialised, bypassing 525a3434a2dSAnthony PERARD * XenbusStateInitWait, without waiting for the frontend to first 526a3434a2dSAnthony PERARD * enter the XenbusStateInitialised state. In this case, default 527a3434a2dSAnthony PERARD * transport parameters are in effect and any transport parameters 528a3434a2dSAnthony PERARD * published by the backend must contain their default values. 529a3434a2dSAnthony PERARD * 530a3434a2dSAnthony PERARD * Drivers that support optional features and/or transport parameter 531a3434a2dSAnthony PERARD * negotiation must tolerate these additional state transition paths. 532a3434a2dSAnthony PERARD * In general this means performing the work of any skipped state 533a3434a2dSAnthony PERARD * transition, if it has not already been performed, in addition to the 534a3434a2dSAnthony PERARD * work associated with entry into the current state. 535a3434a2dSAnthony PERARD */ 536a3434a2dSAnthony PERARD 537a3434a2dSAnthony PERARD /* 538a3434a2dSAnthony PERARD * REQUEST CODES. 539a3434a2dSAnthony PERARD */ 540a3434a2dSAnthony PERARD #define BLKIF_OP_READ 0 541a3434a2dSAnthony PERARD #define BLKIF_OP_WRITE 1 542a3434a2dSAnthony PERARD /* 543a3434a2dSAnthony PERARD * All writes issued prior to a request with the BLKIF_OP_WRITE_BARRIER 544a3434a2dSAnthony PERARD * operation code ("barrier request") must be completed prior to the 545a3434a2dSAnthony PERARD * execution of the barrier request. All writes issued after the barrier 546a3434a2dSAnthony PERARD * request must not execute until after the completion of the barrier request. 547a3434a2dSAnthony PERARD * 548a3434a2dSAnthony PERARD * Optional. See "feature-barrier" XenBus node documentation above. 549a3434a2dSAnthony PERARD */ 550a3434a2dSAnthony PERARD #define BLKIF_OP_WRITE_BARRIER 2 551a3434a2dSAnthony PERARD /* 552a3434a2dSAnthony PERARD * Commit any uncommitted contents of the backing device's volatile cache 553a3434a2dSAnthony PERARD * to stable storage. 554a3434a2dSAnthony PERARD * 555a3434a2dSAnthony PERARD * Optional. See "feature-flush-cache" XenBus node documentation above. 556a3434a2dSAnthony PERARD */ 557a3434a2dSAnthony PERARD #define BLKIF_OP_FLUSH_DISKCACHE 3 558a3434a2dSAnthony PERARD /* 559a3434a2dSAnthony PERARD * Used in SLES sources for device specific command packet 560a3434a2dSAnthony PERARD * contained within the request. Reserved for that purpose. 561a3434a2dSAnthony PERARD */ 562a3434a2dSAnthony PERARD #define BLKIF_OP_RESERVED_1 4 563a3434a2dSAnthony PERARD /* 564a3434a2dSAnthony PERARD * Indicate to the backend device that a region of storage is no longer in 565a3434a2dSAnthony PERARD * use, and may be discarded at any time without impact to the client. If 566a3434a2dSAnthony PERARD * the BLKIF_DISCARD_SECURE flag is set on the request, all copies of the 567a3434a2dSAnthony PERARD * discarded region on the device must be rendered unrecoverable before the 568a3434a2dSAnthony PERARD * command returns. 569a3434a2dSAnthony PERARD * 570a3434a2dSAnthony PERARD * This operation is analogous to performing a trim (ATA) or unamp (SCSI), 571a3434a2dSAnthony PERARD * command on a native device. 572a3434a2dSAnthony PERARD * 573a3434a2dSAnthony PERARD * More information about trim/unmap operations can be found at: 574a3434a2dSAnthony PERARD * http://t13.org/Documents/UploadedDocuments/docs2008/ 575a3434a2dSAnthony PERARD * e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc 576a3434a2dSAnthony PERARD * http://www.seagate.com/staticfiles/support/disc/manuals/ 577a3434a2dSAnthony PERARD * Interface%20manuals/100293068c.pdf 578a3434a2dSAnthony PERARD * 579a3434a2dSAnthony PERARD * Optional. See "feature-discard", "discard-alignment", 580a3434a2dSAnthony PERARD * "discard-granularity", and "discard-secure" in the XenBus node 581a3434a2dSAnthony PERARD * documentation above. 582a3434a2dSAnthony PERARD */ 583a3434a2dSAnthony PERARD #define BLKIF_OP_DISCARD 5 584a3434a2dSAnthony PERARD 585a3434a2dSAnthony PERARD /* 586a3434a2dSAnthony PERARD * Recognized if "feature-max-indirect-segments" in present in the backend 587a3434a2dSAnthony PERARD * xenbus info. The "feature-max-indirect-segments" node contains the maximum 588a3434a2dSAnthony PERARD * number of segments allowed by the backend per request. If the node is 589a3434a2dSAnthony PERARD * present, the frontend might use blkif_request_indirect structs in order to 590a3434a2dSAnthony PERARD * issue requests with more than BLKIF_MAX_SEGMENTS_PER_REQUEST (11). The 591a3434a2dSAnthony PERARD * maximum number of indirect segments is fixed by the backend, but the 592a3434a2dSAnthony PERARD * frontend can issue requests with any number of indirect segments as long as 593a3434a2dSAnthony PERARD * it's less than the number provided by the backend. The indirect_grefs field 594a3434a2dSAnthony PERARD * in blkif_request_indirect should be filled by the frontend with the 595a3434a2dSAnthony PERARD * grant references of the pages that are holding the indirect segments. 596a3434a2dSAnthony PERARD * These pages are filled with an array of blkif_request_segment that hold the 597a3434a2dSAnthony PERARD * information about the segments. The number of indirect pages to use is 598a3434a2dSAnthony PERARD * determined by the number of segments an indirect request contains. Every 599a3434a2dSAnthony PERARD * indirect page can contain a maximum of 600a3434a2dSAnthony PERARD * (PAGE_SIZE / sizeof(struct blkif_request_segment)) segments, so to 601a3434a2dSAnthony PERARD * calculate the number of indirect pages to use we have to do 602a3434a2dSAnthony PERARD * ceil(indirect_segments / (PAGE_SIZE / sizeof(struct blkif_request_segment))). 603a3434a2dSAnthony PERARD * 604a3434a2dSAnthony PERARD * If a backend does not recognize BLKIF_OP_INDIRECT, it should *not* 605a3434a2dSAnthony PERARD * create the "feature-max-indirect-segments" node! 606a3434a2dSAnthony PERARD */ 607a3434a2dSAnthony PERARD #define BLKIF_OP_INDIRECT 6 608a3434a2dSAnthony PERARD 609a3434a2dSAnthony PERARD /* 610a3434a2dSAnthony PERARD * Maximum scatter/gather segments per request. 611a3434a2dSAnthony PERARD * This is carefully chosen so that sizeof(blkif_ring_t) <= PAGE_SIZE. 612a3434a2dSAnthony PERARD * NB. This could be 12 if the ring indexes weren't stored in the same page. 613a3434a2dSAnthony PERARD */ 614a3434a2dSAnthony PERARD #define BLKIF_MAX_SEGMENTS_PER_REQUEST 11 615a3434a2dSAnthony PERARD 616a3434a2dSAnthony PERARD /* 617a3434a2dSAnthony PERARD * Maximum number of indirect pages to use per request. 618a3434a2dSAnthony PERARD */ 619a3434a2dSAnthony PERARD #define BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST 8 620a3434a2dSAnthony PERARD 621a3434a2dSAnthony PERARD /* 622a3434a2dSAnthony PERARD * NB. 'first_sect' and 'last_sect' in blkif_request_segment, as well as 623a3434a2dSAnthony PERARD * 'sector_number' in blkif_request, blkif_request_discard and 624a3434a2dSAnthony PERARD * blkif_request_indirect are sector-based quantities. See the description 625a3434a2dSAnthony PERARD * of the "feature-large-sector-size" frontend xenbus node above for 626a3434a2dSAnthony PERARD * more information. 627a3434a2dSAnthony PERARD */ 628a3434a2dSAnthony PERARD struct blkif_request_segment { 629a3434a2dSAnthony PERARD grant_ref_t gref; /* reference to I/O buffer frame */ 630a3434a2dSAnthony PERARD /* @first_sect: first sector in frame to transfer (inclusive). */ 631a3434a2dSAnthony PERARD /* @last_sect: last sector in frame to transfer (inclusive). */ 632a3434a2dSAnthony PERARD uint8_t first_sect, last_sect; 633a3434a2dSAnthony PERARD }; 634a3434a2dSAnthony PERARD 635a3434a2dSAnthony PERARD /* 636a3434a2dSAnthony PERARD * Starting ring element for any I/O request. 637a3434a2dSAnthony PERARD */ 638a3434a2dSAnthony PERARD struct blkif_request { 639a3434a2dSAnthony PERARD uint8_t operation; /* BLKIF_OP_??? */ 640a3434a2dSAnthony PERARD uint8_t nr_segments; /* number of segments */ 641a3434a2dSAnthony PERARD blkif_vdev_t handle; /* only for read/write requests */ 642a3434a2dSAnthony PERARD uint64_t id; /* private guest value, echoed in resp */ 643a3434a2dSAnthony PERARD blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 644a3434a2dSAnthony PERARD struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 645a3434a2dSAnthony PERARD }; 646a3434a2dSAnthony PERARD typedef struct blkif_request blkif_request_t; 647a3434a2dSAnthony PERARD 648a3434a2dSAnthony PERARD /* 649a3434a2dSAnthony PERARD * Cast to this structure when blkif_request.operation == BLKIF_OP_DISCARD 650a3434a2dSAnthony PERARD * sizeof(struct blkif_request_discard) <= sizeof(struct blkif_request) 651a3434a2dSAnthony PERARD */ 652a3434a2dSAnthony PERARD struct blkif_request_discard { 653a3434a2dSAnthony PERARD uint8_t operation; /* BLKIF_OP_DISCARD */ 654a3434a2dSAnthony PERARD uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */ 655a3434a2dSAnthony PERARD #define BLKIF_DISCARD_SECURE (1<<0) /* ignored if discard-secure=0 */ 656a3434a2dSAnthony PERARD blkif_vdev_t handle; /* same as for read/write requests */ 657a3434a2dSAnthony PERARD uint64_t id; /* private guest value, echoed in resp */ 658a3434a2dSAnthony PERARD blkif_sector_t sector_number;/* start sector idx on disk */ 659a3434a2dSAnthony PERARD uint64_t nr_sectors; /* number of contiguous sectors to discard*/ 660a3434a2dSAnthony PERARD }; 661a3434a2dSAnthony PERARD typedef struct blkif_request_discard blkif_request_discard_t; 662a3434a2dSAnthony PERARD 663a3434a2dSAnthony PERARD struct blkif_request_indirect { 664a3434a2dSAnthony PERARD uint8_t operation; /* BLKIF_OP_INDIRECT */ 665a3434a2dSAnthony PERARD uint8_t indirect_op; /* BLKIF_OP_{READ/WRITE} */ 666a3434a2dSAnthony PERARD uint16_t nr_segments; /* number of segments */ 667a3434a2dSAnthony PERARD uint64_t id; /* private guest value, echoed in resp */ 668a3434a2dSAnthony PERARD blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 669a3434a2dSAnthony PERARD blkif_vdev_t handle; /* same as for read/write requests */ 670a3434a2dSAnthony PERARD grant_ref_t indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST]; 671a3434a2dSAnthony PERARD #ifdef __i386__ 672a3434a2dSAnthony PERARD uint64_t pad; /* Make it 64 byte aligned on i386 */ 673a3434a2dSAnthony PERARD #endif 674a3434a2dSAnthony PERARD }; 675a3434a2dSAnthony PERARD typedef struct blkif_request_indirect blkif_request_indirect_t; 676a3434a2dSAnthony PERARD 677a3434a2dSAnthony PERARD struct blkif_response { 678a3434a2dSAnthony PERARD uint64_t id; /* copied from request */ 679a3434a2dSAnthony PERARD uint8_t operation; /* copied from request */ 680a3434a2dSAnthony PERARD int16_t status; /* BLKIF_RSP_??? */ 681a3434a2dSAnthony PERARD }; 682a3434a2dSAnthony PERARD typedef struct blkif_response blkif_response_t; 683a3434a2dSAnthony PERARD 684a3434a2dSAnthony PERARD /* 685a3434a2dSAnthony PERARD * STATUS RETURN CODES. 686a3434a2dSAnthony PERARD */ 687a3434a2dSAnthony PERARD /* Operation not supported (only happens on barrier writes). */ 688a3434a2dSAnthony PERARD #define BLKIF_RSP_EOPNOTSUPP -2 689a3434a2dSAnthony PERARD /* Operation failed for some unspecified reason (-EIO). */ 690a3434a2dSAnthony PERARD #define BLKIF_RSP_ERROR -1 691a3434a2dSAnthony PERARD /* Operation completed successfully. */ 692a3434a2dSAnthony PERARD #define BLKIF_RSP_OKAY 0 693a3434a2dSAnthony PERARD 694a3434a2dSAnthony PERARD /* 695a3434a2dSAnthony PERARD * Generate blkif ring structures and types. 696a3434a2dSAnthony PERARD */ 697a3434a2dSAnthony PERARD DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response); 698a3434a2dSAnthony PERARD 699a3434a2dSAnthony PERARD #define VDISK_CDROM 0x1 700a3434a2dSAnthony PERARD #define VDISK_REMOVABLE 0x2 701a3434a2dSAnthony PERARD #define VDISK_READONLY 0x4 702a3434a2dSAnthony PERARD 703a3434a2dSAnthony PERARD #endif /* __XEN_PUBLIC_IO_BLKIF_H__ */ 70450c88402SJoao Martins 70550c88402SJoao Martins /* 70650c88402SJoao Martins * Local variables: 70750c88402SJoao Martins * mode: C 70850c88402SJoao Martins * c-file-style: "BSD" 70950c88402SJoao Martins * c-basic-offset: 4 71050c88402SJoao Martins * tab-width: 4 71150c88402SJoao Martins * indent-tabs-mode: nil 71250c88402SJoao Martins * End: 71350c88402SJoao Martins */ 714