include/xen/interface/io/blkif.h

   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
  27 typedef uint16_t blkif_vdev_t;
  28 typedef uint64_t blkif_sector_t;
  29
  30 /*
  31  * Multiple hardware queues/rings:
  32  * If supported, the backend will write the key "multi-queue-max-queues" to
  33  * the directory for that vbd, and set its value to the maximum supported
  34  * number of queues.
  35  * Frontends that are aware of this feature and wish to use it can write the
  36  * key "multi-queue-num-queues" with the number they wish to use, which must be
  37  * greater than zero, and no more than the value reported by the backend in
  38  * "multi-queue-max-queues".
  39  *
  40  * For frontends requesting just one queue, the usual event-channel and
  41  * ring-ref keys are written as before, simplifying the backend processing
  42  * to avoid distinguishing between a frontend that doesn't understand the
  43  * multi-queue feature, and one that does, but requested only one queue.
  44  *
  45  * Frontends requesting two or more queues must not write the toplevel
  46  * event-channel and ring-ref keys, instead writing those keys under sub-keys
  47  * having the name "queue-N" where N is the integer ID of the queue/ring for
  48  * which those keys belong. Queues are indexed from zero.
  49  * For example, a frontend with two queues must write the following set of
  50  * queue-related keys:
  51  *
  52  * /local/domain/1/device/vbd/0/multi-queue-num-queues = "2"
  53  * /local/domain/1/device/vbd/0/queue-0 = ""
  54  * /local/domain/1/device/vbd/0/queue-0/ring-ref = "<ring-ref#0>"
  55  * /local/domain/1/device/vbd/0/queue-0/event-channel = "<evtchn#0>"
  56  * /local/domain/1/device/vbd/0/queue-1 = ""
  57  * /local/domain/1/device/vbd/0/queue-1/ring-ref = "<ring-ref#1>"
  58  * /local/domain/1/device/vbd/0/queue-1/event-channel = "<evtchn#1>"
  59  *
  60  * It is also possible to use multiple queues/rings together with
  61  * feature multi-page ring buffer.
  62  * For example, a frontend requests two queues/rings and the size of each ring
  63  * buffer is two pages must write the following set of related keys:
  64  *
  65  * /local/domain/1/device/vbd/0/multi-queue-num-queues = "2"
  66  * /local/domain/1/device/vbd/0/ring-page-order = "1"
  67  * /local/domain/1/device/vbd/0/queue-0 = ""
  68  * /local/domain/1/device/vbd/0/queue-0/ring-ref0 = "<ring-ref#0>"
  69  * /local/domain/1/device/vbd/0/queue-0/ring-ref1 = "<ring-ref#1>"
  70  * /local/domain/1/device/vbd/0/queue-0/event-channel = "<evtchn#0>"
  71  * /local/domain/1/device/vbd/0/queue-1 = ""
  72  * /local/domain/1/device/vbd/0/queue-1/ring-ref0 = "<ring-ref#2>"
  73  * /local/domain/1/device/vbd/0/queue-1/ring-ref1 = "<ring-ref#3>"
  74  * /local/domain/1/device/vbd/0/queue-1/event-channel = "<evtchn#1>"
  75  *
  76  */
  77
  78 /*
  79  * REQUEST CODES.
  80  */
  81 #define BLKIF_OP_READ              0
  82 #define BLKIF_OP_WRITE             1
  83 /*
  84  * Recognised only if "feature-barrier" is present in backend xenbus info.
  85  * The "feature_barrier" node contains a boolean indicating whether barrier
  86  * requests are likely to succeed or fail. Either way, a barrier request
  87  * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by
  88  * the underlying block-device hardware. The boolean simply indicates whether
  89  * or not it is worthwhile for the frontend to attempt barrier requests.
  90  * If a backend does not recognise BLKIF_OP_WRITE_BARRIER, it should *not*
  91  * create the "feature-barrier" node!
  92  */
  93 #define BLKIF_OP_WRITE_BARRIER     2
  94
  95 /*
  96  * Recognised if "feature-flush-cache" is present in backend xenbus
  97  * info.  A flush will ask the underlying storage hardware to flush its
  98  * non-volatile caches as appropriate.  The "feature-flush-cache" node
  99  * contains a boolean indicating whether flush requests are likely to
 100  * succeed or fail. Either way, a flush request may fail at any time
 101  * with BLKIF_RSP_EOPNOTSUPP if it is unsupported by the underlying
 102  * block-device hardware. The boolean simply indicates whether or not it
 103  * is worthwhile for the frontend to attempt flushes.  If a backend does
 104  * not recognise BLKIF_OP_WRITE_FLUSH_CACHE, it should *not* create the
 105  * "feature-flush-cache" node!
 106  */
 107 #define BLKIF_OP_FLUSH_DISKCACHE   3
 108
 109 /*
 110  * Recognised only if "feature-discard" is present in backend xenbus info.
 111  * The "feature-discard" node contains a boolean indicating whether trim
 112  * (ATA) or unmap (SCSI) - conviently called discard requests are likely
 113  * to succeed or fail. Either way, a discard request
 114  * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by
 115  * the underlying block-device hardware. The boolean simply indicates whether
 116  * or not it is worthwhile for the frontend to attempt discard requests.
 117  * If a backend does not recognise BLKIF_OP_DISCARD, it should *not*
 118  * create the "feature-discard" node!
 119  *
 120  * Discard operation is a request for the underlying block device to mark
 121  * extents to be erased. However, discard does not guarantee that the blocks
 122  * will be erased from the device - it is just a hint to the device
 123  * controller that these blocks are no longer in use. What the device
 124  * controller does with that information is left to the controller.
 125  * Discard operations are passed with sector_number as the
 126  * sector index to begin discard operations at and nr_sectors as the number of
 127  * sectors to be discarded. The specified sectors should be discarded if the
 128  * underlying block device supports trim (ATA) or unmap (SCSI) operations,
 129  * or a BLKIF_RSP_EOPNOTSUPP  should be returned.
 130  * More information about trim/unmap operations at:
 131  * http://t13.org/Documents/UploadedDocuments/docs2008/
 132  *     e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc
 133  * http://www.seagate.com/staticfiles/support/disc/manuals/
 134  *     Interface%20manuals/100293068c.pdf
 135  * The backend can optionally provide three extra XenBus attributes to
 136  * further optimize the discard functionality:
 137  * 'discard-alignment' - Devices that support discard functionality may
 138  * internally allocate space in units that are bigger than the exported
 139  * logical block size. The discard-alignment parameter indicates how many bytes
 140  * the beginning of the partition is offset from the internal allocation unit's
 141  * natural alignment.
 142  * 'discard-granularity'  - Devices that support discard functionality may
 143  * internally allocate space using units that are bigger than the logical block
 144  * size. The discard-granularity parameter indicates the size of the internal
 145  * allocation unit in bytes if reported by the device. Otherwise the
 146  * discard-granularity will be set to match the device's physical block size.
 147  * 'discard-secure' - All copies of the discarded sectors (potentially created
 148  * by garbage collection) must also be erased.  To use this feature, the flag
 149  * BLKIF_DISCARD_SECURE must be set in the blkif_request_trim.
 150  */
 151 #define BLKIF_OP_DISCARD           5
 152
 153 /*
 154  * Recognized if "feature-max-indirect-segments" in present in the backend
 155  * xenbus info. The "feature-max-indirect-segments" node contains the maximum
 156  * number of segments allowed by the backend per request. If the node is
 157  * present, the frontend might use blkif_request_indirect structs in order to
 158  * issue requests with more than BLKIF_MAX_SEGMENTS_PER_REQUEST (11). The
 159  * maximum number of indirect segments is fixed by the backend, but the
 160  * frontend can issue requests with any number of indirect segments as long as
 161  * it's less than the number provided by the backend. The indirect_grefs field
 162  * in blkif_request_indirect should be filled by the frontend with the
 163  * grant references of the pages that are holding the indirect segments.
 164  * These pages are filled with an array of blkif_request_segment that hold the
 165  * information about the segments. The number of indirect pages to use is
 166  * determined by the number of segments an indirect request contains. Every
 167  * indirect page can contain a maximum of
 168  * (PAGE_SIZE / sizeof(struct blkif_request_segment)) segments, so to
 169  * calculate the number of indirect pages to use we have to do
 170  * ceil(indirect_segments / (PAGE_SIZE / sizeof(struct blkif_request_segment))).
 171  *
 172  * If a backend does not recognize BLKIF_OP_INDIRECT, it should *not*
 173  * create the "feature-max-indirect-segments" node!
 174  */
 175 #define BLKIF_OP_INDIRECT          6
 176
 177 /*
 178  * Maximum scatter/gather segments per request.
 179  * This is carefully chosen so that sizeof(struct blkif_ring) <= PAGE_SIZE.
 180  * NB. This could be 12 if the ring indexes weren't stored in the same page.
 181  */
 182 #define BLKIF_MAX_SEGMENTS_PER_REQUEST 11
 183
 184 #define BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST 8
 185
 186 struct blkif_request_segment {
 187                 grant_ref_t gref;        /* reference to I/O buffer frame        */
 188                 /* @first_sect: first sector in frame to transfer (inclusive).   */
 189                 /* @last_sect: last sector in frame to transfer (inclusive).     */
 190                 uint8_t     first_sect, last_sect;
 191 };
 192
 193 struct blkif_request_rw {
 194         uint8_t        nr_segments;  /* number of segments                   */
 195         blkif_vdev_t   handle;       /* only for read/write requests         */
 196 #ifndef CONFIG_X86_32
 197         uint32_t       _pad1;        /* offsetof(blkif_request,u.rw.id) == 8 */
 198 #endif
 199         uint64_t       id;           /* private guest value, echoed in resp  */
 200         blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
 201         struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
 202 } __attribute__((__packed__));
 203
 204 struct blkif_request_discard {
 205         uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero.        */
 206 #define BLKIF_DISCARD_SECURE (1<<0)  /* ignored if discard-secure=0          */
 207         blkif_vdev_t   _pad1;        /* only for read/write requests         */
 208 #ifndef CONFIG_X86_32
 209         uint32_t       _pad2;        /* offsetof(blkif_req..,u.discard.id)==8*/
 210 #endif
 211         uint64_t       id;           /* private guest value, echoed in resp  */
 212         blkif_sector_t sector_number;
 213         uint64_t       nr_sectors;
 214         uint8_t        _pad3;
 215 } __attribute__((__packed__));
 216
 217 struct blkif_request_other {
 218         uint8_t      _pad1;
 219         blkif_vdev_t _pad2;        /* only for read/write requests         */
 220 #ifndef CONFIG_X86_32
 221         uint32_t     _pad3;        /* offsetof(blkif_req..,u.other.id)==8*/
 222 #endif
 223         uint64_t     id;           /* private guest value, echoed in resp  */
 224 } __attribute__((__packed__));
 225
 226 struct blkif_request_indirect {
 227         uint8_t        indirect_op;
 228         uint16_t       nr_segments;
 229 #ifndef CONFIG_X86_32
 230         uint32_t       _pad1;        /* offsetof(blkif_...,u.indirect.id) == 8 */
 231 #endif
 232         uint64_t       id;
 233         blkif_sector_t sector_number;
 234         blkif_vdev_t   handle;
 235         uint16_t       _pad2;
 236         grant_ref_t    indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST];
 237 #ifndef CONFIG_X86_32
 238         uint32_t      _pad3;         /* make it 64 byte aligned */
 239 #else
 240         uint64_t      _pad3;         /* make it 64 byte aligned */
 241 #endif
 242 } __attribute__((__packed__));
 243
 244 struct blkif_request {
 245         uint8_t        operation;    /* BLKIF_OP_???                         */
 246         union {
 247                 struct blkif_request_rw rw;
 248                 struct blkif_request_discard discard;
 249                 struct blkif_request_other other;
 250                 struct blkif_request_indirect indirect;
 251         } u;
 252 } __attribute__((__packed__));
 253
 254 struct blkif_response {
 255         uint64_t        id;              /* copied from request */
 256         uint8_t         operation;       /* copied from request */
 257         int16_t         status;          /* BLKIF_RSP_???       */
 258 };
 259
 260 /*
 261  * STATUS RETURN CODES.
 262  */
 263  /* Operation not supported (only happens on barrier writes). */
 264 #define BLKIF_RSP_EOPNOTSUPP  -2
 265  /* Operation failed for some unspecified reason (-EIO). */
 266 #define BLKIF_RSP_ERROR       -1
 267  /* Operation completed successfully. */
 268 #define BLKIF_RSP_OKAY         0
 269
 270 /*
 271  * Generate blkif ring structures and types.
 272  */
 273
 274 DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response);
 275
 276 #define VDISK_CDROM        0x1
 277 #define VDISK_REMOVABLE    0x2
 278 #define VDISK_READONLY     0x4
 279
 280 /* Xen-defined major numbers for virtual disks, they look strangely
 281  * familiar */
 282 #define XEN_IDE0_MAJOR  3
 283 #define XEN_IDE1_MAJOR  22
 284 #define XEN_SCSI_DISK0_MAJOR    8
 285 #define XEN_SCSI_DISK1_MAJOR    65
 286 #define XEN_SCSI_DISK2_MAJOR    66
 287 #define XEN_SCSI_DISK3_MAJOR    67
 288 #define XEN_SCSI_DISK4_MAJOR    68
 289 #define XEN_SCSI_DISK5_MAJOR    69
 290 #define XEN_SCSI_DISK6_MAJOR    70
 291 #define XEN_SCSI_DISK7_MAJOR    71
 292 #define XEN_SCSI_DISK8_MAJOR    128
 293 #define XEN_SCSI_DISK9_MAJOR    129
 294 #define XEN_SCSI_DISK10_MAJOR   130
 295 #define XEN_SCSI_DISK11_MAJOR   131
 296 #define XEN_SCSI_DISK12_MAJOR   132
 297 #define XEN_SCSI_DISK13_MAJOR   133
 298 #define XEN_SCSI_DISK14_MAJOR   134
 299 #define XEN_SCSI_DISK15_MAJOR   135
 300
 301 #endif /* __XEN_PUBLIC_IO_BLKIF_H__ */