interface/io/blkif.h

0001 /* SPDX-License-Identifier: MIT */
0002 /******************************************************************************
0003  * blkif.h
0004  *
0005  * Unified block-device I/O interface for Xen guest OSes.
0006  *
0007  * Copyright (c) 2003-2004, Keir Fraser
0008  */
0009
0010 #ifndef __XEN_PUBLIC_IO_BLKIF_H__
0011 #define __XEN_PUBLIC_IO_BLKIF_H__
0012
0013 #include <xen/interface/io/ring.h>
0014 #include <xen/interface/grant_table.h>
0015
0016 /*
0017  * Front->back notifications: When enqueuing a new request, sending a
0018  * notification can be made conditional on req_event (i.e., the generic
0019  * hold-off mechanism provided by the ring macros). Backends must set
0020  * req_event appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
0021  *
0022  * Back->front notifications: When enqueuing a new response, sending a
0023  * notification can be made conditional on rsp_event (i.e., the generic
0024  * hold-off mechanism provided by the ring macros). Frontends must set
0025  * rsp_event appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
0026  */
0027
0028 typedef uint16_t blkif_vdev_t;
0029 typedef uint64_t blkif_sector_t;
0030
0031 /*
0032  * Multiple hardware queues/rings:
0033  * If supported, the backend will write the key "multi-queue-max-queues" to
0034  * the directory for that vbd, and set its value to the maximum supported
0035  * number of queues.
0036  * Frontends that are aware of this feature and wish to use it can write the
0037  * key "multi-queue-num-queues" with the number they wish to use, which must be
0038  * greater than zero, and no more than the value reported by the backend in
0039  * "multi-queue-max-queues".
0040  *
0041  * For frontends requesting just one queue, the usual event-channel and
0042  * ring-ref keys are written as before, simplifying the backend processing
0043  * to avoid distinguishing between a frontend that doesn't understand the
0044  * multi-queue feature, and one that does, but requested only one queue.
0045  *
0046  * Frontends requesting two or more queues must not write the toplevel
0047  * event-channel and ring-ref keys, instead writing those keys under sub-keys
0048  * having the name "queue-N" where N is the integer ID of the queue/ring for
0049  * which those keys belong. Queues are indexed from zero.
0050  * For example, a frontend with two queues must write the following set of
0051  * queue-related keys:
0052  *
0053  * /local/domain/1/device/vbd/0/multi-queue-num-queues = "2"
0054  * /local/domain/1/device/vbd/0/queue-0 = ""
0055  * /local/domain/1/device/vbd/0/queue-0/ring-ref = "<ring-ref#0>"
0056  * /local/domain/1/device/vbd/0/queue-0/event-channel = "<evtchn#0>"
0057  * /local/domain/1/device/vbd/0/queue-1 = ""
0058  * /local/domain/1/device/vbd/0/queue-1/ring-ref = "<ring-ref#1>"
0059  * /local/domain/1/device/vbd/0/queue-1/event-channel = "<evtchn#1>"
0060  *
0061  * It is also possible to use multiple queues/rings together with
0062  * feature multi-page ring buffer.
0063  * For example, a frontend requests two queues/rings and the size of each ring
0064  * buffer is two pages must write the following set of related keys:
0065  *
0066  * /local/domain/1/device/vbd/0/multi-queue-num-queues = "2"
0067  * /local/domain/1/device/vbd/0/ring-page-order = "1"
0068  * /local/domain/1/device/vbd/0/queue-0 = ""
0069  * /local/domain/1/device/vbd/0/queue-0/ring-ref0 = "<ring-ref#0>"
0070  * /local/domain/1/device/vbd/0/queue-0/ring-ref1 = "<ring-ref#1>"
0071  * /local/domain/1/device/vbd/0/queue-0/event-channel = "<evtchn#0>"
0072  * /local/domain/1/device/vbd/0/queue-1 = ""
0073  * /local/domain/1/device/vbd/0/queue-1/ring-ref0 = "<ring-ref#2>"
0074  * /local/domain/1/device/vbd/0/queue-1/ring-ref1 = "<ring-ref#3>"
0075  * /local/domain/1/device/vbd/0/queue-1/event-channel = "<evtchn#1>"
0076  *
0077  */
0078
0079 /*
0080  * REQUEST CODES.
0081  */
0082 #define BLKIF_OP_READ              0
0083 #define BLKIF_OP_WRITE             1
0084 /*
0085  * Recognised only if "feature-barrier" is present in backend xenbus info.
0086  * The "feature_barrier" node contains a boolean indicating whether barrier
0087  * requests are likely to succeed or fail. Either way, a barrier request
0088  * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by
0089  * the underlying block-device hardware. The boolean simply indicates whether
0090  * or not it is worthwhile for the frontend to attempt barrier requests.
0091  * If a backend does not recognise BLKIF_OP_WRITE_BARRIER, it should *not*
0092  * create the "feature-barrier" node!
0093  */
0094 #define BLKIF_OP_WRITE_BARRIER     2
0095
0096 /*
0097  * Recognised if "feature-flush-cache" is present in backend xenbus
0098  * info.  A flush will ask the underlying storage hardware to flush its
0099  * non-volatile caches as appropriate.  The "feature-flush-cache" node
0100  * contains a boolean indicating whether flush requests are likely to
0101  * succeed or fail. Either way, a flush request may fail at any time
0102  * with BLKIF_RSP_EOPNOTSUPP if it is unsupported by the underlying
0103  * block-device hardware. The boolean simply indicates whether or not it
0104  * is worthwhile for the frontend to attempt flushes.  If a backend does
0105  * not recognise BLKIF_OP_WRITE_FLUSH_CACHE, it should *not* create the
0106  * "feature-flush-cache" node!
0107  */
0108 #define BLKIF_OP_FLUSH_DISKCACHE   3
0109
0110 /*
0111  * Recognised only if "feature-discard" is present in backend xenbus info.
0112  * The "feature-discard" node contains a boolean indicating whether trim
0113  * (ATA) or unmap (SCSI) - conviently called discard requests are likely
0114  * to succeed or fail. Either way, a discard request
0115  * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by
0116  * the underlying block-device hardware. The boolean simply indicates whether
0117  * or not it is worthwhile for the frontend to attempt discard requests.
0118  * If a backend does not recognise BLKIF_OP_DISCARD, it should *not*
0119  * create the "feature-discard" node!
0120  *
0121  * Discard operation is a request for the underlying block device to mark
0122  * extents to be erased. However, discard does not guarantee that the blocks
0123  * will be erased from the device - it is just a hint to the device
0124  * controller that these blocks are no longer in use. What the device
0125  * controller does with that information is left to the controller.
0126  * Discard operations are passed with sector_number as the
0127  * sector index to begin discard operations at and nr_sectors as the number of
0128  * sectors to be discarded. The specified sectors should be discarded if the
0129  * underlying block device supports trim (ATA) or unmap (SCSI) operations,
0130  * or a BLKIF_RSP_EOPNOTSUPP  should be returned.
0131  * More information about trim/unmap operations at:
0132  * http://t13.org/Documents/UploadedDocuments/docs2008/
0133  *     e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc
0134  * http://www.seagate.com/staticfiles/support/disc/manuals/
0135  *     Interface%20manuals/100293068c.pdf
0136  * The backend can optionally provide three extra XenBus attributes to
0137  * further optimize the discard functionality:
0138  * 'discard-alignment' - Devices that support discard functionality may
0139  * internally allocate space in units that are bigger than the exported
0140  * logical block size. The discard-alignment parameter indicates how many bytes
0141  * the beginning of the partition is offset from the internal allocation unit's
0142  * natural alignment.
0143  * 'discard-granularity'  - Devices that support discard functionality may
0144  * internally allocate space using units that are bigger than the logical block
0145  * size. The discard-granularity parameter indicates the size of the internal
0146  * allocation unit in bytes if reported by the device. Otherwise the
0147  * discard-granularity will be set to match the device's physical block size.
0148  * 'discard-secure' - All copies of the discarded sectors (potentially created
0149  * by garbage collection) must also be erased.  To use this feature, the flag
0150  * BLKIF_DISCARD_SECURE must be set in the blkif_request_trim.
0151  */
0152 #define BLKIF_OP_DISCARD           5
0153
0154 /*
0155  * Recognized if "feature-max-indirect-segments" in present in the backend
0156  * xenbus info. The "feature-max-indirect-segments" node contains the maximum
0157  * number of segments allowed by the backend per request. If the node is
0158  * present, the frontend might use blkif_request_indirect structs in order to
0159  * issue requests with more than BLKIF_MAX_SEGMENTS_PER_REQUEST (11). The
0160  * maximum number of indirect segments is fixed by the backend, but the
0161  * frontend can issue requests with any number of indirect segments as long as
0162  * it's less than the number provided by the backend. The indirect_grefs field
0163  * in blkif_request_indirect should be filled by the frontend with the
0164  * grant references of the pages that are holding the indirect segments.
0165  * These pages are filled with an array of blkif_request_segment that hold the
0166  * information about the segments. The number of indirect pages to use is
0167  * determined by the number of segments an indirect request contains. Every
0168  * indirect page can contain a maximum of
0169  * (PAGE_SIZE / sizeof(struct blkif_request_segment)) segments, so to
0170  * calculate the number of indirect pages to use we have to do
0171  * ceil(indirect_segments / (PAGE_SIZE / sizeof(struct blkif_request_segment))).
0172  *
0173  * If a backend does not recognize BLKIF_OP_INDIRECT, it should *not*
0174  * create the "feature-max-indirect-segments" node!
0175  */
0176 #define BLKIF_OP_INDIRECT          6
0177
0178 /*
0179  * Maximum scatter/gather segments per request.
0180  * This is carefully chosen so that sizeof(struct blkif_ring) <= PAGE_SIZE.
0181  * NB. This could be 12 if the ring indexes weren't stored in the same page.
0182  */
0183 #define BLKIF_MAX_SEGMENTS_PER_REQUEST 11
0184
0185 #define BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST 8
0186
0187 struct blkif_request_segment {
0188         grant_ref_t gref;        /* reference to I/O buffer frame        */
0189         /* @first_sect: first sector in frame to transfer (inclusive).   */
0190         /* @last_sect: last sector in frame to transfer (inclusive).     */
0191         uint8_t     first_sect, last_sect;
0192 };
0193
0194 struct blkif_request_rw {
0195     uint8_t        nr_segments;  /* number of segments                   */
0196     blkif_vdev_t   handle;       /* only for read/write requests         */
0197 #ifndef CONFIG_X86_32
0198     uint32_t       _pad1;        /* offsetof(blkif_request,u.rw.id) == 8 */
0199 #endif
0200     uint64_t       id;           /* private guest value, echoed in resp  */
0201     blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
0202     struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
0203 } __attribute__((__packed__));
0204
0205 struct blkif_request_discard {
0206     uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero.        */
0207 #define BLKIF_DISCARD_SECURE (1<<0)  /* ignored if discard-secure=0          */
0208     blkif_vdev_t   _pad1;        /* only for read/write requests         */
0209 #ifndef CONFIG_X86_32
0210     uint32_t       _pad2;        /* offsetof(blkif_req..,u.discard.id)==8*/
0211 #endif
0212     uint64_t       id;           /* private guest value, echoed in resp  */
0213     blkif_sector_t sector_number;
0214     uint64_t       nr_sectors;
0215     uint8_t        _pad3;
0216 } __attribute__((__packed__));
0217
0218 struct blkif_request_other {
0219     uint8_t      _pad1;
0220     blkif_vdev_t _pad2;        /* only for read/write requests         */
0221 #ifndef CONFIG_X86_32
0222     uint32_t     _pad3;        /* offsetof(blkif_req..,u.other.id)==8*/
0223 #endif
0224     uint64_t     id;           /* private guest value, echoed in resp  */
0225 } __attribute__((__packed__));
0226
0227 struct blkif_request_indirect {
0228     uint8_t        indirect_op;
0229     uint16_t       nr_segments;
0230 #ifndef CONFIG_X86_32
0231     uint32_t       _pad1;        /* offsetof(blkif_...,u.indirect.id) == 8 */
0232 #endif
0233     uint64_t       id;
0234     blkif_sector_t sector_number;
0235     blkif_vdev_t   handle;
0236     uint16_t       _pad2;
0237     grant_ref_t    indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST];
0238 #ifndef CONFIG_X86_32
0239     uint32_t      _pad3;         /* make it 64 byte aligned */
0240 #else
0241     uint64_t      _pad3;         /* make it 64 byte aligned */
0242 #endif
0243 } __attribute__((__packed__));
0244
0245 struct blkif_request {
0246     uint8_t        operation;    /* BLKIF_OP_???                         */
0247     union {
0248         struct blkif_request_rw rw;
0249         struct blkif_request_discard discard;
0250         struct blkif_request_other other;
0251         struct blkif_request_indirect indirect;
0252     } u;
0253 } __attribute__((__packed__));
0254
0255 struct blkif_response {
0256     uint64_t        id;              /* copied from request */
0257     uint8_t         operation;       /* copied from request */
0258     int16_t         status;          /* BLKIF_RSP_???       */
0259 };
0260
0261 /*
0262  * STATUS RETURN CODES.
0263  */
0264  /* Operation not supported (only happens on barrier writes). */
0265 #define BLKIF_RSP_EOPNOTSUPP  -2
0266  /* Operation failed for some unspecified reason (-EIO). */
0267 #define BLKIF_RSP_ERROR       -1
0268  /* Operation completed successfully. */
0269 #define BLKIF_RSP_OKAY         0
0270
0271 /*
0272  * Generate blkif ring structures and types.
0273  */
0274
0275 DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response);
0276
0277 #define VDISK_CDROM        0x1
0278 #define VDISK_REMOVABLE    0x2
0279 #define VDISK_READONLY     0x4
0280
0281 /* Xen-defined major numbers for virtual disks, they look strangely
0282  * familiar */
0283 #define XEN_IDE0_MAJOR  3
0284 #define XEN_IDE1_MAJOR  22
0285 #define XEN_SCSI_DISK0_MAJOR    8
0286 #define XEN_SCSI_DISK1_MAJOR    65
0287 #define XEN_SCSI_DISK2_MAJOR    66
0288 #define XEN_SCSI_DISK3_MAJOR    67
0289 #define XEN_SCSI_DISK4_MAJOR    68
0290 #define XEN_SCSI_DISK5_MAJOR    69
0291 #define XEN_SCSI_DISK6_MAJOR    70
0292 #define XEN_SCSI_DISK7_MAJOR    71
0293 #define XEN_SCSI_DISK8_MAJOR    128
0294 #define XEN_SCSI_DISK9_MAJOR    129
0295 #define XEN_SCSI_DISK10_MAJOR   130
0296 #define XEN_SCSI_DISK11_MAJOR   131
0297 #define XEN_SCSI_DISK12_MAJOR   132
0298 #define XEN_SCSI_DISK13_MAJOR   133
0299 #define XEN_SCSI_DISK14_MAJOR   134
0300 #define XEN_SCSI_DISK15_MAJOR   135
0301
0302 #endif /* __XEN_PUBLIC_IO_BLKIF_H__ */