blkif.h 8.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207
  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. #ifndef __XEN_PUBLIC_IO_BLKIF_H__
  9. #define __XEN_PUBLIC_IO_BLKIF_H__
  10. #include "ring.h"
  11. #include "../grant_table.h"
  12. /*
  13. * Front->back notifications: When enqueuing a new request, sending a
  14. * notification can be made conditional on req_event (i.e., the generic
  15. * hold-off mechanism provided by the ring macros). Backends must set
  16. * req_event appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
  17. *
  18. * Back->front notifications: When enqueuing a new response, sending a
  19. * notification can be made conditional on rsp_event (i.e., the generic
  20. * hold-off mechanism provided by the ring macros). Frontends must set
  21. * rsp_event appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
  22. */
  23. typedef uint16_t blkif_vdev_t;
  24. typedef uint64_t blkif_sector_t;
  25. /*
  26. * REQUEST CODES.
  27. */
  28. #define BLKIF_OP_READ 0
  29. #define BLKIF_OP_WRITE 1
  30. /*
  31. * Recognised only if "feature-barrier" is present in backend xenbus info.
  32. * The "feature_barrier" node contains a boolean indicating whether barrier
  33. * requests are likely to succeed or fail. Either way, a barrier request
  34. * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by
  35. * the underlying block-device hardware. The boolean simply indicates whether
  36. * or not it is worthwhile for the frontend to attempt barrier requests.
  37. * If a backend does not recognise BLKIF_OP_WRITE_BARRIER, it should *not*
  38. * create the "feature-barrier" node!
  39. */
  40. #define BLKIF_OP_WRITE_BARRIER 2
  41. /*
  42. * Recognised if "feature-flush-cache" is present in backend xenbus
  43. * info. A flush will ask the underlying storage hardware to flush its
  44. * non-volatile caches as appropriate. The "feature-flush-cache" node
  45. * contains a boolean indicating whether flush requests are likely to
  46. * succeed or fail. Either way, a flush request may fail at any time
  47. * with BLKIF_RSP_EOPNOTSUPP if it is unsupported by the underlying
  48. * block-device hardware. The boolean simply indicates whether or not it
  49. * is worthwhile for the frontend to attempt flushes. If a backend does
  50. * not recognise BLKIF_OP_WRITE_FLUSH_CACHE, it should *not* create the
  51. * "feature-flush-cache" node!
  52. */
  53. #define BLKIF_OP_FLUSH_DISKCACHE 3
  54. /*
  55. * Recognised only if "feature-discard" is present in backend xenbus info.
  56. * The "feature-discard" node contains a boolean indicating whether trim
  57. * (ATA) or unmap (SCSI) - conviently called discard requests are likely
  58. * to succeed or fail. Either way, a discard request
  59. * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by
  60. * the underlying block-device hardware. The boolean simply indicates whether
  61. * or not it is worthwhile for the frontend to attempt discard requests.
  62. * If a backend does not recognise BLKIF_OP_DISCARD, it should *not*
  63. * create the "feature-discard" node!
  64. *
  65. * Discard operation is a request for the underlying block device to mark
  66. * extents to be erased. However, discard does not guarantee that the blocks
  67. * will be erased from the device - it is just a hint to the device
  68. * controller that these blocks are no longer in use. What the device
  69. * controller does with that information is left to the controller.
  70. * Discard operations are passed with sector_number as the
  71. * sector index to begin discard operations at and nr_sectors as the number of
  72. * sectors to be discarded. The specified sectors should be discarded if the
  73. * underlying block device supports trim (ATA) or unmap (SCSI) operations,
  74. * or a BLKIF_RSP_EOPNOTSUPP should be returned.
  75. * More information about trim/unmap operations at:
  76. * http://t13.org/Documents/UploadedDocuments/docs2008/
  77. * e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc
  78. * http://www.seagate.com/staticfiles/support/disc/manuals/
  79. * Interface%20manuals/100293068c.pdf
  80. * The backend can optionally provide three extra XenBus attributes to
  81. * further optimize the discard functionality:
  82. * 'discard-aligment' - Devices that support discard functionality may
  83. * internally allocate space in units that are bigger than the exported
  84. * logical block size. The discard-alignment parameter indicates how many bytes
  85. * the beginning of the partition is offset from the internal allocation unit's
  86. * natural alignment.
  87. * 'discard-granularity' - Devices that support discard functionality may
  88. * internally allocate space using units that are bigger than the logical block
  89. * size. The discard-granularity parameter indicates the size of the internal
  90. * allocation unit in bytes if reported by the device. Otherwise the
  91. * discard-granularity will be set to match the device's physical block size.
  92. * 'discard-secure' - All copies of the discarded sectors (potentially created
  93. * by garbage collection) must also be erased. To use this feature, the flag
  94. * BLKIF_DISCARD_SECURE must be set in the blkif_request_trim.
  95. */
  96. #define BLKIF_OP_DISCARD 5
  97. /*
  98. * Maximum scatter/gather segments per request.
  99. * This is carefully chosen so that sizeof(struct blkif_ring) <= PAGE_SIZE.
  100. * NB. This could be 12 if the ring indexes weren't stored in the same page.
  101. */
  102. #define BLKIF_MAX_SEGMENTS_PER_REQUEST 11
  103. struct blkif_request_rw {
  104. uint8_t nr_segments; /* number of segments */
  105. blkif_vdev_t handle; /* only for read/write requests */
  106. #ifdef CONFIG_X86_64
  107. uint32_t _pad1; /* offsetof(blkif_request,u.rw.id) == 8 */
  108. #endif
  109. uint64_t id; /* private guest value, echoed in resp */
  110. blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
  111. struct blkif_request_segment {
  112. grant_ref_t gref; /* reference to I/O buffer frame */
  113. /* @first_sect: first sector in frame to transfer (inclusive). */
  114. /* @last_sect: last sector in frame to transfer (inclusive). */
  115. uint8_t first_sect, last_sect;
  116. } seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
  117. } __attribute__((__packed__));
  118. struct blkif_request_discard {
  119. uint8_t flag; /* BLKIF_DISCARD_SECURE or zero. */
  120. #define BLKIF_DISCARD_SECURE (1<<0) /* ignored if discard-secure=0 */
  121. blkif_vdev_t _pad1; /* only for read/write requests */
  122. #ifdef CONFIG_X86_64
  123. uint32_t _pad2; /* offsetof(blkif_req..,u.discard.id)==8*/
  124. #endif
  125. uint64_t id; /* private guest value, echoed in resp */
  126. blkif_sector_t sector_number;
  127. uint64_t nr_sectors;
  128. uint8_t _pad3;
  129. } __attribute__((__packed__));
  130. struct blkif_request_other {
  131. uint8_t _pad1;
  132. blkif_vdev_t _pad2; /* only for read/write requests */
  133. #ifdef CONFIG_X86_64
  134. uint32_t _pad3; /* offsetof(blkif_req..,u.other.id)==8*/
  135. #endif
  136. uint64_t id; /* private guest value, echoed in resp */
  137. } __attribute__((__packed__));
  138. struct blkif_request {
  139. uint8_t operation; /* BLKIF_OP_??? */
  140. union {
  141. struct blkif_request_rw rw;
  142. struct blkif_request_discard discard;
  143. struct blkif_request_other other;
  144. } u;
  145. } __attribute__((__packed__));
  146. struct blkif_response {
  147. uint64_t id; /* copied from request */
  148. uint8_t operation; /* copied from request */
  149. int16_t status; /* BLKIF_RSP_??? */
  150. };
  151. /*
  152. * STATUS RETURN CODES.
  153. */
  154. /* Operation not supported (only happens on barrier writes). */
  155. #define BLKIF_RSP_EOPNOTSUPP -2
  156. /* Operation failed for some unspecified reason (-EIO). */
  157. #define BLKIF_RSP_ERROR -1
  158. /* Operation completed successfully. */
  159. #define BLKIF_RSP_OKAY 0
  160. /*
  161. * Generate blkif ring structures and types.
  162. */
  163. DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response);
  164. #define VDISK_CDROM 0x1
  165. #define VDISK_REMOVABLE 0x2
  166. #define VDISK_READONLY 0x4
  167. /* Xen-defined major numbers for virtual disks, they look strangely
  168. * familiar */
  169. #define XEN_IDE0_MAJOR 3
  170. #define XEN_IDE1_MAJOR 22
  171. #define XEN_SCSI_DISK0_MAJOR 8
  172. #define XEN_SCSI_DISK1_MAJOR 65
  173. #define XEN_SCSI_DISK2_MAJOR 66
  174. #define XEN_SCSI_DISK3_MAJOR 67
  175. #define XEN_SCSI_DISK4_MAJOR 68
  176. #define XEN_SCSI_DISK5_MAJOR 69
  177. #define XEN_SCSI_DISK6_MAJOR 70
  178. #define XEN_SCSI_DISK7_MAJOR 71
  179. #define XEN_SCSI_DISK8_MAJOR 128
  180. #define XEN_SCSI_DISK9_MAJOR 129
  181. #define XEN_SCSI_DISK10_MAJOR 130
  182. #define XEN_SCSI_DISK11_MAJOR 131
  183. #define XEN_SCSI_DISK12_MAJOR 132
  184. #define XEN_SCSI_DISK13_MAJOR 133
  185. #define XEN_SCSI_DISK14_MAJOR 134
  186. #define XEN_SCSI_DISK15_MAJOR 135
  187. #endif /* __XEN_PUBLIC_IO_BLKIF_H__ */