1/****************************************************************************** 2 * blkif.h 3 * 4 * Unified block-device I/O interface for Xen guest OSes. 5 * 6 * Copyright (c) 2003-2004, Keir Fraser 7 */ 8 9#ifndef __XEN_PUBLIC_IO_BLKIF_H__ 10#define __XEN_PUBLIC_IO_BLKIF_H__ 11 12#include <xen/interface/io/ring.h> 13#include <xen/interface/grant_table.h> 14 15/* 16 * Front->back notifications: When enqueuing a new request, sending a 17 * notification can be made conditional on req_event (i.e., the generic 18 * hold-off mechanism provided by the ring macros). Backends must set 19 * req_event appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()). 20 * 21 * Back->front notifications: When enqueuing a new response, sending a 22 * notification can be made conditional on rsp_event (i.e., the generic 23 * hold-off mechanism provided by the ring macros). Frontends must set 24 * rsp_event appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()). 25 */ 26 27typedef uint16_t blkif_vdev_t; 28typedef uint64_t blkif_sector_t; 29 30/* 31 * REQUEST CODES. 32 */ 33#define BLKIF_OP_READ 0 34#define BLKIF_OP_WRITE 1 35/* 36 * Recognised only if "feature-barrier" is present in backend xenbus info. 37 * The "feature_barrier" node contains a boolean indicating whether barrier 38 * requests are likely to succeed or fail. Either way, a barrier request 39 * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by 40 * the underlying block-device hardware. The boolean simply indicates whether 41 * or not it is worthwhile for the frontend to attempt barrier requests. 42 * If a backend does not recognise BLKIF_OP_WRITE_BARRIER, it should *not* 43 * create the "feature-barrier" node! 44 */ 45#define BLKIF_OP_WRITE_BARRIER 2 46 47/* 48 * Recognised if "feature-flush-cache" is present in backend xenbus 49 * info. A flush will ask the underlying storage hardware to flush its 50 * non-volatile caches as appropriate. The "feature-flush-cache" node 51 * contains a boolean indicating whether flush requests are likely to 52 * succeed or fail. Either way, a flush request may fail at any time 53 * with BLKIF_RSP_EOPNOTSUPP if it is unsupported by the underlying 54 * block-device hardware. The boolean simply indicates whether or not it 55 * is worthwhile for the frontend to attempt flushes. If a backend does 56 * not recognise BLKIF_OP_WRITE_FLUSH_CACHE, it should *not* create the 57 * "feature-flush-cache" node! 58 */ 59#define BLKIF_OP_FLUSH_DISKCACHE 3 60 61/* 62 * Recognised only if "feature-discard" is present in backend xenbus info. 63 * The "feature-discard" node contains a boolean indicating whether trim 64 * (ATA) or unmap (SCSI) - conviently called discard requests are likely 65 * to succeed or fail. Either way, a discard request 66 * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by 67 * the underlying block-device hardware. The boolean simply indicates whether 68 * or not it is worthwhile for the frontend to attempt discard requests. 69 * If a backend does not recognise BLKIF_OP_DISCARD, it should *not* 70 * create the "feature-discard" node! 71 * 72 * Discard operation is a request for the underlying block device to mark 73 * extents to be erased. However, discard does not guarantee that the blocks 74 * will be erased from the device - it is just a hint to the device 75 * controller that these blocks are no longer in use. What the device 76 * controller does with that information is left to the controller. 77 * Discard operations are passed with sector_number as the 78 * sector index to begin discard operations at and nr_sectors as the number of 79 * sectors to be discarded. The specified sectors should be discarded if the 80 * underlying block device supports trim (ATA) or unmap (SCSI) operations, 81 * or a BLKIF_RSP_EOPNOTSUPP should be returned. 82 * More information about trim/unmap operations at: 83 * http://t13.org/Documents/UploadedDocuments/docs2008/ 84 * e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc 85 * http://www.seagate.com/staticfiles/support/disc/manuals/ 86 * Interface%20manuals/100293068c.pdf 87 * The backend can optionally provide three extra XenBus attributes to 88 * further optimize the discard functionality: 89 * 'discard-aligment' - Devices that support discard functionality may 90 * internally allocate space in units that are bigger than the exported 91 * logical block size. The discard-alignment parameter indicates how many bytes 92 * the beginning of the partition is offset from the internal allocation unit's 93 * natural alignment. 94 * 'discard-granularity' - Devices that support discard functionality may 95 * internally allocate space using units that are bigger than the logical block 96 * size. The discard-granularity parameter indicates the size of the internal 97 * allocation unit in bytes if reported by the device. Otherwise the 98 * discard-granularity will be set to match the device's physical block size. 99 * 'discard-secure' - All copies of the discarded sectors (potentially created 100 * by garbage collection) must also be erased. To use this feature, the flag 101 * BLKIF_DISCARD_SECURE must be set in the blkif_request_trim. 102 */ 103#define BLKIF_OP_DISCARD 5 104 105/* 106 * Recognized if "feature-max-indirect-segments" in present in the backend 107 * xenbus info. The "feature-max-indirect-segments" node contains the maximum 108 * number of segments allowed by the backend per request. If the node is 109 * present, the frontend might use blkif_request_indirect structs in order to 110 * issue requests with more than BLKIF_MAX_SEGMENTS_PER_REQUEST (11). The 111 * maximum number of indirect segments is fixed by the backend, but the 112 * frontend can issue requests with any number of indirect segments as long as 113 * it's less than the number provided by the backend. The indirect_grefs field 114 * in blkif_request_indirect should be filled by the frontend with the 115 * grant references of the pages that are holding the indirect segments. 116 * These pages are filled with an array of blkif_request_segment that hold the 117 * information about the segments. The number of indirect pages to use is 118 * determined by the number of segments an indirect request contains. Every 119 * indirect page can contain a maximum of 120 * (PAGE_SIZE / sizeof(struct blkif_request_segment)) segments, so to 121 * calculate the number of indirect pages to use we have to do 122 * ceil(indirect_segments / (PAGE_SIZE / sizeof(struct blkif_request_segment))). 123 * 124 * If a backend does not recognize BLKIF_OP_INDIRECT, it should *not* 125 * create the "feature-max-indirect-segments" node! 126 */ 127#define BLKIF_OP_INDIRECT 6 128 129/* 130 * Maximum scatter/gather segments per request. 131 * This is carefully chosen so that sizeof(struct blkif_ring) <= PAGE_SIZE. 132 * NB. This could be 12 if the ring indexes weren't stored in the same page. 133 */ 134#define BLKIF_MAX_SEGMENTS_PER_REQUEST 11 135 136#define BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST 8 137 138struct blkif_request_segment { 139 grant_ref_t gref; /* reference to I/O buffer frame */ 140 /* @first_sect: first sector in frame to transfer (inclusive). */ 141 /* @last_sect: last sector in frame to transfer (inclusive). */ 142 uint8_t first_sect, last_sect; 143}; 144 145struct blkif_request_rw { 146 uint8_t nr_segments; /* number of segments */ 147 blkif_vdev_t handle; /* only for read/write requests */ 148#ifdef CONFIG_X86_64 149 uint32_t _pad1; /* offsetof(blkif_request,u.rw.id) == 8 */ 150#endif 151 uint64_t id; /* private guest value, echoed in resp */ 152 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 153 struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 154} __attribute__((__packed__)); 155 156struct blkif_request_discard { 157 uint8_t flag; /* BLKIF_DISCARD_SECURE or zero. */ 158#define BLKIF_DISCARD_SECURE (1<<0) /* ignored if discard-secure=0 */ 159 blkif_vdev_t _pad1; /* only for read/write requests */ 160#ifdef CONFIG_X86_64 161 uint32_t _pad2; /* offsetof(blkif_req..,u.discard.id)==8*/ 162#endif 163 uint64_t id; /* private guest value, echoed in resp */ 164 blkif_sector_t sector_number; 165 uint64_t nr_sectors; 166 uint8_t _pad3; 167} __attribute__((__packed__)); 168 169struct blkif_request_other { 170 uint8_t _pad1; 171 blkif_vdev_t _pad2; /* only for read/write requests */ 172#ifdef CONFIG_X86_64 173 uint32_t _pad3; /* offsetof(blkif_req..,u.other.id)==8*/ 174#endif 175 uint64_t id; /* private guest value, echoed in resp */ 176} __attribute__((__packed__)); 177 178struct blkif_request_indirect { 179 uint8_t indirect_op; 180 uint16_t nr_segments; 181#ifdef CONFIG_X86_64 182 uint32_t _pad1; /* offsetof(blkif_...,u.indirect.id) == 8 */ 183#endif 184 uint64_t id; 185 blkif_sector_t sector_number; 186 blkif_vdev_t handle; 187 uint16_t _pad2; 188 grant_ref_t indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST]; 189#ifdef CONFIG_X86_64 190 uint32_t _pad3; /* make it 64 byte aligned */ 191#else 192 uint64_t _pad3; /* make it 64 byte aligned */ 193#endif 194} __attribute__((__packed__)); 195 196struct blkif_request { 197 uint8_t operation; /* BLKIF_OP_??? */ 198 union { 199 struct blkif_request_rw rw; 200 struct blkif_request_discard discard; 201 struct blkif_request_other other; 202 struct blkif_request_indirect indirect; 203 } u; 204} __attribute__((__packed__)); 205 206struct blkif_response { 207 uint64_t id; /* copied from request */ 208 uint8_t operation; /* copied from request */ 209 int16_t status; /* BLKIF_RSP_??? */ 210}; 211 212/* 213 * STATUS RETURN CODES. 214 */ 215 /* Operation not supported (only happens on barrier writes). */ 216#define BLKIF_RSP_EOPNOTSUPP -2 217 /* Operation failed for some unspecified reason (-EIO). */ 218#define BLKIF_RSP_ERROR -1 219 /* Operation completed successfully. */ 220#define BLKIF_RSP_OKAY 0 221 222/* 223 * Generate blkif ring structures and types. 224 */ 225 226DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response); 227 228#define VDISK_CDROM 0x1 229#define VDISK_REMOVABLE 0x2 230#define VDISK_READONLY 0x4 231 232/* Xen-defined major numbers for virtual disks, they look strangely 233 * familiar */ 234#define XEN_IDE0_MAJOR 3 235#define XEN_IDE1_MAJOR 22 236#define XEN_SCSI_DISK0_MAJOR 8 237#define XEN_SCSI_DISK1_MAJOR 65 238#define XEN_SCSI_DISK2_MAJOR 66 239#define XEN_SCSI_DISK3_MAJOR 67 240#define XEN_SCSI_DISK4_MAJOR 68 241#define XEN_SCSI_DISK5_MAJOR 69 242#define XEN_SCSI_DISK6_MAJOR 70 243#define XEN_SCSI_DISK7_MAJOR 71 244#define XEN_SCSI_DISK8_MAJOR 128 245#define XEN_SCSI_DISK9_MAJOR 129 246#define XEN_SCSI_DISK10_MAJOR 130 247#define XEN_SCSI_DISK11_MAJOR 131 248#define XEN_SCSI_DISK12_MAJOR 132 249#define XEN_SCSI_DISK13_MAJOR 133 250#define XEN_SCSI_DISK14_MAJOR 134 251#define XEN_SCSI_DISK15_MAJOR 135 252 253#endif /* __XEN_PUBLIC_IO_BLKIF_H__ */ 254