|
|
1.1 ! root 1: /* ! 2: * Block driver for the Virtual Disk Image (VDI) format ! 3: * ! 4: * Copyright (c) 2009 Stefan Weil ! 5: * ! 6: * This program is free software: you can redistribute it and/or modify ! 7: * it under the terms of the GNU General Public License as published by ! 8: * the Free Software Foundation, either version 2 of the License, or ! 9: * (at your option) version 3 or any later version. ! 10: * ! 11: * This program is distributed in the hope that it will be useful, ! 12: * but WITHOUT ANY WARRANTY; without even the implied warranty of ! 13: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the ! 14: * GNU General Public License for more details. ! 15: * ! 16: * You should have received a copy of the GNU General Public License ! 17: * along with this program. If not, see <http://www.gnu.org/licenses/>. ! 18: * ! 19: * Reference: ! 20: * http://forums.virtualbox.org/viewtopic.php?t=8046 ! 21: * ! 22: * This driver supports create / read / write operations on VDI images. ! 23: * ! 24: * Todo (see also TODO in code): ! 25: * ! 26: * Some features like snapshots are still missing. ! 27: * ! 28: * Deallocation of zero-filled blocks and shrinking images are missing, too ! 29: * (might be added to common block layer). ! 30: * ! 31: * Allocation of blocks could be optimized (less writes to block map and ! 32: * header). ! 33: * ! 34: * Read and write of adjacents blocks could be done in one operation ! 35: * (current code uses one operation per block (1 MiB). ! 36: * ! 37: * The code is not thread safe (missing locks for changes in header and ! 38: * block table, no problem with current QEMU). ! 39: * ! 40: * Hints: ! 41: * ! 42: * Blocks (VDI documentation) correspond to clusters (QEMU). ! 43: * QEMU's backing files could be implemented using VDI snapshot files (TODO). ! 44: * VDI snapshot files may also contain the complete machine state. ! 45: * Maybe this machine state can be converted to QEMU PC machine snapshot data. ! 46: * ! 47: * The driver keeps a block cache (little endian entries) in memory. ! 48: * For the standard block size (1 MiB), a 1 TiB disk will use 4 MiB RAM, ! 49: * so this seems to be reasonable. ! 50: */ ! 51: ! 52: #include "qemu-common.h" ! 53: #include "block_int.h" ! 54: #include "module.h" ! 55: ! 56: #if defined(CONFIG_UUID) ! 57: #include <uuid/uuid.h> ! 58: #else ! 59: /* TODO: move uuid emulation to some central place in QEMU. */ ! 60: #include "sysemu.h" /* UUID_FMT */ ! 61: typedef unsigned char uuid_t[16]; ! 62: void uuid_generate(uuid_t out); ! 63: int uuid_is_null(const uuid_t uu); ! 64: void uuid_unparse(const uuid_t uu, char *out); ! 65: #endif ! 66: ! 67: /* Code configuration options. */ ! 68: ! 69: /* Enable debug messages. */ ! 70: //~ #define CONFIG_VDI_DEBUG ! 71: ! 72: /* Support write operations on VDI images. */ ! 73: #define CONFIG_VDI_WRITE ! 74: ! 75: /* Support non-standard block (cluster) size. This is untested. ! 76: * Maybe it will be needed for very large images. ! 77: */ ! 78: //~ #define CONFIG_VDI_BLOCK_SIZE ! 79: ! 80: /* Support static (fixed, pre-allocated) images. */ ! 81: #define CONFIG_VDI_STATIC_IMAGE ! 82: ! 83: /* Command line option for static images. */ ! 84: #define BLOCK_OPT_STATIC "static" ! 85: ! 86: #define KiB 1024 ! 87: #define MiB (KiB * KiB) ! 88: ! 89: #define SECTOR_SIZE 512 ! 90: ! 91: #if defined(CONFIG_VDI_DEBUG) ! 92: #define logout(fmt, ...) \ ! 93: fprintf(stderr, "vdi\t%-24s" fmt, __func__, ##__VA_ARGS__) ! 94: #else ! 95: #define logout(fmt, ...) ((void)0) ! 96: #endif ! 97: ! 98: /* Image signature. */ ! 99: #define VDI_SIGNATURE 0xbeda107f ! 100: ! 101: /* Image version. */ ! 102: #define VDI_VERSION_1_1 0x00010001 ! 103: ! 104: /* Image type. */ ! 105: #define VDI_TYPE_DYNAMIC 1 ! 106: #define VDI_TYPE_STATIC 2 ! 107: ! 108: /* Innotek / SUN images use these strings in header.text: ! 109: * "<<< innotek VirtualBox Disk Image >>>\n" ! 110: * "<<< Sun xVM VirtualBox Disk Image >>>\n" ! 111: * "<<< Sun VirtualBox Disk Image >>>\n" ! 112: * The value does not matter, so QEMU created images use a different text. ! 113: */ ! 114: #define VDI_TEXT "<<< QEMU VM Virtual Disk Image >>>\n" ! 115: ! 116: /* Unallocated blocks use this index (no need to convert endianess). */ ! 117: #define VDI_UNALLOCATED UINT32_MAX ! 118: ! 119: #if !defined(CONFIG_UUID) ! 120: void uuid_generate(uuid_t out) ! 121: { ! 122: memset(out, 0, sizeof(out)); ! 123: } ! 124: ! 125: int uuid_is_null(const uuid_t uu) ! 126: { ! 127: uuid_t null_uuid = { 0 }; ! 128: return memcmp(uu, null_uuid, sizeof(uu)) == 0; ! 129: } ! 130: ! 131: void uuid_unparse(const uuid_t uu, char *out) ! 132: { ! 133: snprintf(out, 37, UUID_FMT, ! 134: uu[0], uu[1], uu[2], uu[3], uu[4], uu[5], uu[6], uu[7], ! 135: uu[8], uu[9], uu[10], uu[11], uu[12], uu[13], uu[14], uu[15]); ! 136: } ! 137: #endif ! 138: ! 139: typedef struct { ! 140: BlockDriverAIOCB common; ! 141: int64_t sector_num; ! 142: QEMUIOVector *qiov; ! 143: uint8_t *buf; ! 144: /* Total number of sectors. */ ! 145: int nb_sectors; ! 146: /* Number of sectors for current AIO. */ ! 147: int n_sectors; ! 148: /* New allocated block map entry. */ ! 149: uint32_t bmap_first; ! 150: uint32_t bmap_last; ! 151: /* Buffer for new allocated block. */ ! 152: void *block_buffer; ! 153: void *orig_buf; ! 154: int header_modified; ! 155: BlockDriverAIOCB *hd_aiocb; ! 156: struct iovec hd_iov; ! 157: QEMUIOVector hd_qiov; ! 158: QEMUBH *bh; ! 159: } VdiAIOCB; ! 160: ! 161: typedef struct { ! 162: char text[0x40]; ! 163: uint32_t signature; ! 164: uint32_t version; ! 165: uint32_t header_size; ! 166: uint32_t image_type; ! 167: uint32_t image_flags; ! 168: char description[256]; ! 169: uint32_t offset_bmap; ! 170: uint32_t offset_data; ! 171: uint32_t cylinders; /* disk geometry, unused here */ ! 172: uint32_t heads; /* disk geometry, unused here */ ! 173: uint32_t sectors; /* disk geometry, unused here */ ! 174: uint32_t sector_size; ! 175: uint32_t unused1; ! 176: uint64_t disk_size; ! 177: uint32_t block_size; ! 178: uint32_t block_extra; /* unused here */ ! 179: uint32_t blocks_in_image; ! 180: uint32_t blocks_allocated; ! 181: uuid_t uuid_image; ! 182: uuid_t uuid_last_snap; ! 183: uuid_t uuid_link; ! 184: uuid_t uuid_parent; ! 185: uint64_t unused2[7]; ! 186: } VdiHeader; ! 187: ! 188: typedef struct { ! 189: BlockDriverState *hd; ! 190: /* The block map entries are little endian (even in memory). */ ! 191: uint32_t *bmap; ! 192: /* Size of block (bytes). */ ! 193: uint32_t block_size; ! 194: /* Size of block (sectors). */ ! 195: uint32_t block_sectors; ! 196: /* First sector of block map. */ ! 197: uint32_t bmap_sector; ! 198: /* VDI header (converted to host endianess). */ ! 199: VdiHeader header; ! 200: } BDRVVdiState; ! 201: ! 202: /* Change UUID from little endian (IPRT = VirtualBox format) to big endian ! 203: * format (network byte order, standard, see RFC 4122) and vice versa. ! 204: */ ! 205: static void uuid_convert(uuid_t uuid) ! 206: { ! 207: bswap32s((uint32_t *)&uuid[0]); ! 208: bswap16s((uint16_t *)&uuid[4]); ! 209: bswap16s((uint16_t *)&uuid[6]); ! 210: } ! 211: ! 212: static void vdi_header_to_cpu(VdiHeader *header) ! 213: { ! 214: le32_to_cpus(&header->signature); ! 215: le32_to_cpus(&header->version); ! 216: le32_to_cpus(&header->header_size); ! 217: le32_to_cpus(&header->image_type); ! 218: le32_to_cpus(&header->image_flags); ! 219: le32_to_cpus(&header->offset_bmap); ! 220: le32_to_cpus(&header->offset_data); ! 221: le32_to_cpus(&header->cylinders); ! 222: le32_to_cpus(&header->heads); ! 223: le32_to_cpus(&header->sectors); ! 224: le32_to_cpus(&header->sector_size); ! 225: le64_to_cpus(&header->disk_size); ! 226: le32_to_cpus(&header->block_size); ! 227: le32_to_cpus(&header->block_extra); ! 228: le32_to_cpus(&header->blocks_in_image); ! 229: le32_to_cpus(&header->blocks_allocated); ! 230: uuid_convert(header->uuid_image); ! 231: uuid_convert(header->uuid_last_snap); ! 232: uuid_convert(header->uuid_link); ! 233: uuid_convert(header->uuid_parent); ! 234: } ! 235: ! 236: static void vdi_header_to_le(VdiHeader *header) ! 237: { ! 238: cpu_to_le32s(&header->signature); ! 239: cpu_to_le32s(&header->version); ! 240: cpu_to_le32s(&header->header_size); ! 241: cpu_to_le32s(&header->image_type); ! 242: cpu_to_le32s(&header->image_flags); ! 243: cpu_to_le32s(&header->offset_bmap); ! 244: cpu_to_le32s(&header->offset_data); ! 245: cpu_to_le32s(&header->cylinders); ! 246: cpu_to_le32s(&header->heads); ! 247: cpu_to_le32s(&header->sectors); ! 248: cpu_to_le32s(&header->sector_size); ! 249: cpu_to_le64s(&header->disk_size); ! 250: cpu_to_le32s(&header->block_size); ! 251: cpu_to_le32s(&header->block_extra); ! 252: cpu_to_le32s(&header->blocks_in_image); ! 253: cpu_to_le32s(&header->blocks_allocated); ! 254: cpu_to_le32s(&header->blocks_allocated); ! 255: uuid_convert(header->uuid_image); ! 256: uuid_convert(header->uuid_last_snap); ! 257: uuid_convert(header->uuid_link); ! 258: uuid_convert(header->uuid_parent); ! 259: } ! 260: ! 261: #if defined(CONFIG_VDI_DEBUG) ! 262: static void vdi_header_print(VdiHeader *header) ! 263: { ! 264: char uuid[37]; ! 265: logout("text %s", header->text); ! 266: logout("signature 0x%04x\n", header->signature); ! 267: logout("header size 0x%04x\n", header->header_size); ! 268: logout("image type 0x%04x\n", header->image_type); ! 269: logout("image flags 0x%04x\n", header->image_flags); ! 270: logout("description %s\n", header->description); ! 271: logout("offset bmap 0x%04x\n", header->offset_bmap); ! 272: logout("offset data 0x%04x\n", header->offset_data); ! 273: logout("cylinders 0x%04x\n", header->cylinders); ! 274: logout("heads 0x%04x\n", header->heads); ! 275: logout("sectors 0x%04x\n", header->sectors); ! 276: logout("sector size 0x%04x\n", header->sector_size); ! 277: logout("image size 0x%" PRIx64 " B (%" PRIu64 " MiB)\n", ! 278: header->disk_size, header->disk_size / MiB); ! 279: logout("block size 0x%04x\n", header->block_size); ! 280: logout("block extra 0x%04x\n", header->block_extra); ! 281: logout("blocks tot. 0x%04x\n", header->blocks_in_image); ! 282: logout("blocks all. 0x%04x\n", header->blocks_allocated); ! 283: uuid_unparse(header->uuid_image, uuid); ! 284: logout("uuid image %s\n", uuid); ! 285: uuid_unparse(header->uuid_last_snap, uuid); ! 286: logout("uuid snap %s\n", uuid); ! 287: uuid_unparse(header->uuid_link, uuid); ! 288: logout("uuid link %s\n", uuid); ! 289: uuid_unparse(header->uuid_parent, uuid); ! 290: logout("uuid parent %s\n", uuid); ! 291: } ! 292: #endif ! 293: ! 294: static int vdi_check(BlockDriverState *bs) ! 295: { ! 296: /* TODO: additional checks possible. */ ! 297: BDRVVdiState *s = (BDRVVdiState *)bs->opaque; ! 298: int n_errors = 0; ! 299: uint32_t blocks_allocated = 0; ! 300: uint32_t block; ! 301: uint32_t *bmap; ! 302: logout("\n"); ! 303: ! 304: bmap = qemu_malloc(s->header.blocks_in_image * sizeof(uint32_t)); ! 305: memset(bmap, 0xff, s->header.blocks_in_image * sizeof(uint32_t)); ! 306: ! 307: /* Check block map and value of blocks_allocated. */ ! 308: for (block = 0; block < s->header.blocks_in_image; block++) { ! 309: uint32_t bmap_entry = le32_to_cpu(s->bmap[block]); ! 310: if (bmap_entry != VDI_UNALLOCATED) { ! 311: if (bmap_entry < s->header.blocks_in_image) { ! 312: blocks_allocated++; ! 313: if (bmap[bmap_entry] == VDI_UNALLOCATED) { ! 314: bmap[bmap_entry] = bmap_entry; ! 315: } else { ! 316: fprintf(stderr, "ERROR: block index %" PRIu32 ! 317: " also used by %" PRIu32 "\n", bmap[bmap_entry], bmap_entry); ! 318: } ! 319: } else { ! 320: fprintf(stderr, "ERROR: block index %" PRIu32 ! 321: " too large, is %" PRIu32 "\n", block, bmap_entry); ! 322: n_errors++; ! 323: } ! 324: } ! 325: } ! 326: if (blocks_allocated != s->header.blocks_allocated) { ! 327: fprintf(stderr, "ERROR: allocated blocks mismatch, is %" PRIu32 ! 328: ", should be %" PRIu32 "\n", ! 329: blocks_allocated, s->header.blocks_allocated); ! 330: n_errors++; ! 331: } ! 332: ! 333: qemu_free(bmap); ! 334: ! 335: return n_errors; ! 336: } ! 337: ! 338: static int vdi_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) ! 339: { ! 340: /* TODO: vdi_get_info would be needed for machine snapshots. ! 341: vm_state_offset is still missing. */ ! 342: BDRVVdiState *s = (BDRVVdiState *)bs->opaque; ! 343: logout("\n"); ! 344: bdi->cluster_size = s->block_size; ! 345: bdi->vm_state_offset = 0; ! 346: return 0; ! 347: } ! 348: ! 349: static int vdi_make_empty(BlockDriverState *bs) ! 350: { ! 351: /* TODO: missing code. */ ! 352: logout("\n"); ! 353: /* The return value for missing code must be 0, see block.c. */ ! 354: return 0; ! 355: } ! 356: ! 357: static int vdi_probe(const uint8_t *buf, int buf_size, const char *filename) ! 358: { ! 359: const VdiHeader *header = (const VdiHeader *)buf; ! 360: int result = 0; ! 361: ! 362: logout("\n"); ! 363: ! 364: if (buf_size < sizeof(*header)) { ! 365: /* Header too small, no VDI. */ ! 366: } else if (le32_to_cpu(header->signature) == VDI_SIGNATURE) { ! 367: result = 100; ! 368: } ! 369: ! 370: if (result == 0) { ! 371: logout("no vdi image\n"); ! 372: } else { ! 373: logout("%s", header->text); ! 374: } ! 375: ! 376: return result; ! 377: } ! 378: ! 379: static int vdi_open(BlockDriverState *bs, const char *filename, int flags) ! 380: { ! 381: BDRVVdiState *s = bs->opaque; ! 382: VdiHeader header; ! 383: size_t bmap_size; ! 384: int ret; ! 385: ! 386: logout("\n"); ! 387: ! 388: ret = bdrv_file_open(&s->hd, filename, flags); ! 389: if (ret < 0) { ! 390: return ret; ! 391: } ! 392: ! 393: if (bdrv_read(s->hd, 0, (uint8_t *)&header, 1) < 0) { ! 394: goto fail; ! 395: } ! 396: ! 397: vdi_header_to_cpu(&header); ! 398: #if defined(CONFIG_VDI_DEBUG) ! 399: vdi_header_print(&header); ! 400: #endif ! 401: ! 402: if (header.version != VDI_VERSION_1_1) { ! 403: logout("unsupported version %u.%u\n", ! 404: header.version >> 16, header.version & 0xffff); ! 405: goto fail; ! 406: } else if (header.offset_bmap % SECTOR_SIZE != 0) { ! 407: /* We only support block maps which start on a sector boundary. */ ! 408: logout("unsupported block map offset 0x%x B\n", header.offset_bmap); ! 409: goto fail; ! 410: } else if (header.offset_data % SECTOR_SIZE != 0) { ! 411: /* We only support data blocks which start on a sector boundary. */ ! 412: logout("unsupported data offset 0x%x B\n", header.offset_data); ! 413: goto fail; ! 414: } else if (header.sector_size != SECTOR_SIZE) { ! 415: logout("unsupported sector size %u B\n", header.sector_size); ! 416: goto fail; ! 417: } else if (header.block_size != 1 * MiB) { ! 418: logout("unsupported block size %u B\n", header.block_size); ! 419: goto fail; ! 420: } else if (header.disk_size != ! 421: (uint64_t)header.blocks_in_image * header.block_size) { ! 422: logout("unexpected block number %u B\n", header.blocks_in_image); ! 423: goto fail; ! 424: } else if (!uuid_is_null(header.uuid_link)) { ! 425: logout("link uuid != 0, unsupported\n"); ! 426: goto fail; ! 427: } else if (!uuid_is_null(header.uuid_parent)) { ! 428: logout("parent uuid != 0, unsupported\n"); ! 429: goto fail; ! 430: } ! 431: ! 432: bs->total_sectors = header.disk_size / SECTOR_SIZE; ! 433: ! 434: s->block_size = header.block_size; ! 435: s->block_sectors = header.block_size / SECTOR_SIZE; ! 436: s->bmap_sector = header.offset_bmap / SECTOR_SIZE; ! 437: s->header = header; ! 438: ! 439: bmap_size = header.blocks_in_image * sizeof(uint32_t); ! 440: bmap_size = (bmap_size + SECTOR_SIZE - 1) / SECTOR_SIZE; ! 441: s->bmap = qemu_malloc(bmap_size * SECTOR_SIZE); ! 442: if (bdrv_read(s->hd, s->bmap_sector, (uint8_t *)s->bmap, bmap_size) < 0) { ! 443: goto fail_free_bmap; ! 444: } ! 445: ! 446: return 0; ! 447: ! 448: fail_free_bmap: ! 449: qemu_free(s->bmap); ! 450: ! 451: fail: ! 452: bdrv_delete(s->hd); ! 453: return -1; ! 454: } ! 455: ! 456: static int vdi_is_allocated(BlockDriverState *bs, int64_t sector_num, ! 457: int nb_sectors, int *pnum) ! 458: { ! 459: /* TODO: Check for too large sector_num (in bdrv_is_allocated or here). */ ! 460: BDRVVdiState *s = (BDRVVdiState *)bs->opaque; ! 461: size_t bmap_index = sector_num / s->block_sectors; ! 462: size_t sector_in_block = sector_num % s->block_sectors; ! 463: int n_sectors = s->block_sectors - sector_in_block; ! 464: uint32_t bmap_entry = le32_to_cpu(s->bmap[bmap_index]); ! 465: logout("%p, %" PRId64 ", %d, %p\n", bs, sector_num, nb_sectors, pnum); ! 466: if (n_sectors > nb_sectors) { ! 467: n_sectors = nb_sectors; ! 468: } ! 469: *pnum = n_sectors; ! 470: return bmap_entry != VDI_UNALLOCATED; ! 471: } ! 472: ! 473: static void vdi_aio_cancel(BlockDriverAIOCB *blockacb) ! 474: { ! 475: /* TODO: This code is untested. How can I get it executed? */ ! 476: VdiAIOCB *acb = (VdiAIOCB *)blockacb; ! 477: logout("\n"); ! 478: if (acb->hd_aiocb) { ! 479: bdrv_aio_cancel(acb->hd_aiocb); ! 480: } ! 481: qemu_aio_release(acb); ! 482: } ! 483: ! 484: static AIOPool vdi_aio_pool = { ! 485: .aiocb_size = sizeof(VdiAIOCB), ! 486: .cancel = vdi_aio_cancel, ! 487: }; ! 488: ! 489: static VdiAIOCB *vdi_aio_setup(BlockDriverState *bs, int64_t sector_num, ! 490: QEMUIOVector *qiov, int nb_sectors, ! 491: BlockDriverCompletionFunc *cb, void *opaque, int is_write) ! 492: { ! 493: VdiAIOCB *acb; ! 494: ! 495: logout("%p, %" PRId64 ", %p, %d, %p, %p, %d\n", ! 496: bs, sector_num, qiov, nb_sectors, cb, opaque, is_write); ! 497: ! 498: acb = qemu_aio_get(&vdi_aio_pool, bs, cb, opaque); ! 499: if (acb) { ! 500: acb->hd_aiocb = NULL; ! 501: acb->sector_num = sector_num; ! 502: acb->qiov = qiov; ! 503: if (qiov->niov > 1) { ! 504: acb->buf = qemu_blockalign(bs, qiov->size); ! 505: acb->orig_buf = acb->buf; ! 506: if (is_write) { ! 507: qemu_iovec_to_buffer(qiov, acb->buf); ! 508: } ! 509: } else { ! 510: acb->buf = (uint8_t *)qiov->iov->iov_base; ! 511: } ! 512: acb->nb_sectors = nb_sectors; ! 513: acb->n_sectors = 0; ! 514: acb->bmap_first = VDI_UNALLOCATED; ! 515: acb->bmap_last = VDI_UNALLOCATED; ! 516: acb->block_buffer = NULL; ! 517: acb->header_modified = 0; ! 518: } ! 519: return acb; ! 520: } ! 521: ! 522: static int vdi_schedule_bh(QEMUBHFunc *cb, VdiAIOCB *acb) ! 523: { ! 524: logout("\n"); ! 525: ! 526: if (acb->bh) { ! 527: return -EIO; ! 528: } ! 529: ! 530: acb->bh = qemu_bh_new(cb, acb); ! 531: if (!acb->bh) { ! 532: return -EIO; ! 533: } ! 534: ! 535: qemu_bh_schedule(acb->bh); ! 536: ! 537: return 0; ! 538: } ! 539: ! 540: static void vdi_aio_read_cb(void *opaque, int ret); ! 541: ! 542: static void vdi_aio_read_bh(void *opaque) ! 543: { ! 544: VdiAIOCB *acb = opaque; ! 545: logout("\n"); ! 546: qemu_bh_delete(acb->bh); ! 547: acb->bh = NULL; ! 548: vdi_aio_read_cb(opaque, 0); ! 549: } ! 550: ! 551: static void vdi_aio_read_cb(void *opaque, int ret) ! 552: { ! 553: VdiAIOCB *acb = opaque; ! 554: BlockDriverState *bs = acb->common.bs; ! 555: BDRVVdiState *s = bs->opaque; ! 556: uint32_t bmap_entry; ! 557: uint32_t block_index; ! 558: uint32_t sector_in_block; ! 559: uint32_t n_sectors; ! 560: ! 561: logout("%u sectors read\n", acb->n_sectors); ! 562: ! 563: acb->hd_aiocb = NULL; ! 564: ! 565: if (ret < 0) { ! 566: goto done; ! 567: } ! 568: ! 569: acb->nb_sectors -= acb->n_sectors; ! 570: ! 571: if (acb->nb_sectors == 0) { ! 572: /* request completed */ ! 573: ret = 0; ! 574: goto done; ! 575: } ! 576: ! 577: acb->sector_num += acb->n_sectors; ! 578: acb->buf += acb->n_sectors * SECTOR_SIZE; ! 579: ! 580: block_index = acb->sector_num / s->block_sectors; ! 581: sector_in_block = acb->sector_num % s->block_sectors; ! 582: n_sectors = s->block_sectors - sector_in_block; ! 583: if (n_sectors > acb->nb_sectors) { ! 584: n_sectors = acb->nb_sectors; ! 585: } ! 586: ! 587: logout("will read %u sectors starting at sector %" PRIu64 "\n", ! 588: n_sectors, acb->sector_num); ! 589: ! 590: /* prepare next AIO request */ ! 591: acb->n_sectors = n_sectors; ! 592: bmap_entry = le32_to_cpu(s->bmap[block_index]); ! 593: if (bmap_entry == VDI_UNALLOCATED) { ! 594: /* Block not allocated, return zeros, no need to wait. */ ! 595: memset(acb->buf, 0, n_sectors * SECTOR_SIZE); ! 596: ret = vdi_schedule_bh(vdi_aio_read_bh, acb); ! 597: if (ret < 0) { ! 598: goto done; ! 599: } ! 600: } else { ! 601: uint64_t offset = s->header.offset_data / SECTOR_SIZE + ! 602: (uint64_t)bmap_entry * s->block_sectors + ! 603: sector_in_block; ! 604: acb->hd_iov.iov_base = (void *)acb->buf; ! 605: acb->hd_iov.iov_len = n_sectors * SECTOR_SIZE; ! 606: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1); ! 607: acb->hd_aiocb = bdrv_aio_readv(s->hd, offset, &acb->hd_qiov, ! 608: n_sectors, vdi_aio_read_cb, acb); ! 609: if (acb->hd_aiocb == NULL) { ! 610: goto done; ! 611: } ! 612: } ! 613: return; ! 614: done: ! 615: if (acb->qiov->niov > 1) { ! 616: qemu_iovec_from_buffer(acb->qiov, acb->orig_buf, acb->qiov->size); ! 617: qemu_vfree(acb->orig_buf); ! 618: } ! 619: acb->common.cb(acb->common.opaque, ret); ! 620: qemu_aio_release(acb); ! 621: } ! 622: ! 623: static BlockDriverAIOCB *vdi_aio_readv(BlockDriverState *bs, ! 624: int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, ! 625: BlockDriverCompletionFunc *cb, void *opaque) ! 626: { ! 627: VdiAIOCB *acb; ! 628: logout("\n"); ! 629: acb = vdi_aio_setup(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); ! 630: if (!acb) { ! 631: return NULL; ! 632: } ! 633: vdi_aio_read_cb(acb, 0); ! 634: return &acb->common; ! 635: } ! 636: ! 637: static void vdi_aio_write_cb(void *opaque, int ret) ! 638: { ! 639: VdiAIOCB *acb = opaque; ! 640: BlockDriverState *bs = acb->common.bs; ! 641: BDRVVdiState *s = bs->opaque; ! 642: uint32_t bmap_entry; ! 643: uint32_t block_index; ! 644: uint32_t sector_in_block; ! 645: uint32_t n_sectors; ! 646: ! 647: acb->hd_aiocb = NULL; ! 648: ! 649: if (ret < 0) { ! 650: goto done; ! 651: } ! 652: ! 653: acb->nb_sectors -= acb->n_sectors; ! 654: acb->sector_num += acb->n_sectors; ! 655: acb->buf += acb->n_sectors * SECTOR_SIZE; ! 656: ! 657: if (acb->nb_sectors == 0) { ! 658: logout("finished data write\n"); ! 659: acb->n_sectors = 0; ! 660: if (acb->header_modified) { ! 661: VdiHeader *header = acb->block_buffer; ! 662: logout("now writing modified header\n"); ! 663: assert(acb->bmap_first != VDI_UNALLOCATED); ! 664: *header = s->header; ! 665: vdi_header_to_le(header); ! 666: acb->header_modified = 0; ! 667: acb->hd_iov.iov_base = acb->block_buffer; ! 668: acb->hd_iov.iov_len = SECTOR_SIZE; ! 669: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1); ! 670: acb->hd_aiocb = bdrv_aio_writev(s->hd, 0, &acb->hd_qiov, 1, ! 671: vdi_aio_write_cb, acb); ! 672: if (acb->hd_aiocb == NULL) { ! 673: goto done; ! 674: } ! 675: return; ! 676: } else if (acb->bmap_first != VDI_UNALLOCATED) { ! 677: /* One or more new blocks were allocated. */ ! 678: uint64_t offset; ! 679: uint32_t bmap_first; ! 680: uint32_t bmap_last; ! 681: qemu_free(acb->block_buffer); ! 682: acb->block_buffer = NULL; ! 683: bmap_first = acb->bmap_first; ! 684: bmap_last = acb->bmap_last; ! 685: logout("now writing modified block map entry %u...%u\n", ! 686: bmap_first, bmap_last); ! 687: /* Write modified sectors from block map. */ ! 688: bmap_first /= (SECTOR_SIZE / sizeof(uint32_t)); ! 689: bmap_last /= (SECTOR_SIZE / sizeof(uint32_t)); ! 690: n_sectors = bmap_last - bmap_first + 1; ! 691: offset = s->bmap_sector + bmap_first; ! 692: acb->bmap_first = VDI_UNALLOCATED; ! 693: acb->hd_iov.iov_base = (void *)((uint8_t *)&s->bmap[0] + ! 694: bmap_first * SECTOR_SIZE); ! 695: acb->hd_iov.iov_len = n_sectors * SECTOR_SIZE; ! 696: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1); ! 697: logout("will write %u block map sectors starting from entry %u\n", ! 698: n_sectors, bmap_first); ! 699: acb->hd_aiocb = bdrv_aio_writev(s->hd, offset, &acb->hd_qiov, ! 700: n_sectors, vdi_aio_write_cb, acb); ! 701: if (acb->hd_aiocb == NULL) { ! 702: goto done; ! 703: } ! 704: return; ! 705: } ! 706: ret = 0; ! 707: goto done; ! 708: } ! 709: ! 710: logout("%u sectors written\n", acb->n_sectors); ! 711: ! 712: block_index = acb->sector_num / s->block_sectors; ! 713: sector_in_block = acb->sector_num % s->block_sectors; ! 714: n_sectors = s->block_sectors - sector_in_block; ! 715: if (n_sectors > acb->nb_sectors) { ! 716: n_sectors = acb->nb_sectors; ! 717: } ! 718: ! 719: logout("will write %u sectors starting at sector %" PRIu64 "\n", ! 720: n_sectors, acb->sector_num); ! 721: ! 722: /* prepare next AIO request */ ! 723: acb->n_sectors = n_sectors; ! 724: bmap_entry = le32_to_cpu(s->bmap[block_index]); ! 725: if (bmap_entry == VDI_UNALLOCATED) { ! 726: /* Allocate new block and write to it. */ ! 727: uint64_t offset; ! 728: uint8_t *block; ! 729: bmap_entry = s->header.blocks_allocated; ! 730: s->bmap[block_index] = cpu_to_le32(bmap_entry); ! 731: s->header.blocks_allocated++; ! 732: offset = s->header.offset_data / SECTOR_SIZE + ! 733: (uint64_t)bmap_entry * s->block_sectors; ! 734: block = acb->block_buffer; ! 735: if (block == NULL) { ! 736: block = qemu_mallocz(s->block_size); ! 737: acb->block_buffer = block; ! 738: acb->bmap_first = block_index; ! 739: assert(!acb->header_modified); ! 740: acb->header_modified = 1; ! 741: } ! 742: acb->bmap_last = block_index; ! 743: memcpy(block + sector_in_block * SECTOR_SIZE, ! 744: acb->buf, n_sectors * SECTOR_SIZE); ! 745: acb->hd_iov.iov_base = (void *)block; ! 746: acb->hd_iov.iov_len = s->block_size; ! 747: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1); ! 748: acb->hd_aiocb = bdrv_aio_writev(s->hd, offset, ! 749: &acb->hd_qiov, s->block_sectors, ! 750: vdi_aio_write_cb, acb); ! 751: if (acb->hd_aiocb == NULL) { ! 752: goto done; ! 753: } ! 754: } else { ! 755: uint64_t offset = s->header.offset_data / SECTOR_SIZE + ! 756: (uint64_t)bmap_entry * s->block_sectors + ! 757: sector_in_block; ! 758: acb->hd_iov.iov_base = (void *)acb->buf; ! 759: acb->hd_iov.iov_len = n_sectors * SECTOR_SIZE; ! 760: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1); ! 761: acb->hd_aiocb = bdrv_aio_writev(s->hd, offset, &acb->hd_qiov, ! 762: n_sectors, vdi_aio_write_cb, acb); ! 763: if (acb->hd_aiocb == NULL) { ! 764: goto done; ! 765: } ! 766: } ! 767: ! 768: return; ! 769: ! 770: done: ! 771: if (acb->qiov->niov > 1) { ! 772: qemu_vfree(acb->orig_buf); ! 773: } ! 774: acb->common.cb(acb->common.opaque, ret); ! 775: qemu_aio_release(acb); ! 776: } ! 777: ! 778: static BlockDriverAIOCB *vdi_aio_writev(BlockDriverState *bs, ! 779: int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, ! 780: BlockDriverCompletionFunc *cb, void *opaque) ! 781: { ! 782: VdiAIOCB *acb; ! 783: logout("\n"); ! 784: acb = vdi_aio_setup(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); ! 785: if (!acb) { ! 786: return NULL; ! 787: } ! 788: vdi_aio_write_cb(acb, 0); ! 789: return &acb->common; ! 790: } ! 791: ! 792: static int vdi_create(const char *filename, QEMUOptionParameter *options) ! 793: { ! 794: int fd; ! 795: int result = 0; ! 796: uint64_t bytes = 0; ! 797: uint32_t blocks; ! 798: size_t block_size = 1 * MiB; ! 799: uint32_t image_type = VDI_TYPE_DYNAMIC; ! 800: VdiHeader header; ! 801: size_t i; ! 802: size_t bmap_size; ! 803: uint32_t *bmap; ! 804: ! 805: logout("\n"); ! 806: ! 807: /* Read out options. */ ! 808: while (options && options->name) { ! 809: if (!strcmp(options->name, BLOCK_OPT_SIZE)) { ! 810: bytes = options->value.n; ! 811: #if defined(CONFIG_VDI_BLOCK_SIZE) ! 812: } else if (!strcmp(options->name, BLOCK_OPT_CLUSTER_SIZE)) { ! 813: if (options->value.n) { ! 814: /* TODO: Additional checks (SECTOR_SIZE * 2^n, ...). */ ! 815: block_size = options->value.n; ! 816: } ! 817: #endif ! 818: #if defined(CONFIG_VDI_STATIC_IMAGE) ! 819: } else if (!strcmp(options->name, BLOCK_OPT_STATIC)) { ! 820: if (options->value.n) { ! 821: image_type = VDI_TYPE_STATIC; ! 822: } ! 823: #endif ! 824: } ! 825: options++; ! 826: } ! 827: ! 828: fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE, ! 829: 0644); ! 830: if (fd < 0) { ! 831: return -errno; ! 832: } ! 833: ! 834: blocks = bytes / block_size; ! 835: bmap_size = blocks * sizeof(uint32_t); ! 836: bmap_size = ((bmap_size + SECTOR_SIZE - 1) & ~(SECTOR_SIZE -1)); ! 837: ! 838: memset(&header, 0, sizeof(header)); ! 839: pstrcpy(header.text, sizeof(header.text), VDI_TEXT); ! 840: header.signature = VDI_SIGNATURE; ! 841: header.version = VDI_VERSION_1_1; ! 842: header.header_size = 0x180; ! 843: header.image_type = image_type; ! 844: header.offset_bmap = 0x200; ! 845: header.offset_data = 0x200 + bmap_size; ! 846: header.sector_size = SECTOR_SIZE; ! 847: header.disk_size = bytes; ! 848: header.block_size = block_size; ! 849: header.blocks_in_image = blocks; ! 850: if (image_type == VDI_TYPE_STATIC) { ! 851: header.blocks_allocated = blocks; ! 852: } ! 853: uuid_generate(header.uuid_image); ! 854: uuid_generate(header.uuid_last_snap); ! 855: /* There is no need to set header.uuid_link or header.uuid_parent here. */ ! 856: #if defined(CONFIG_VDI_DEBUG) ! 857: vdi_header_print(&header); ! 858: #endif ! 859: vdi_header_to_le(&header); ! 860: if (write(fd, &header, sizeof(header)) < 0) { ! 861: result = -errno; ! 862: } ! 863: ! 864: bmap = (uint32_t *)qemu_mallocz(bmap_size); ! 865: for (i = 0; i < blocks; i++) { ! 866: if (image_type == VDI_TYPE_STATIC) { ! 867: bmap[i] = i; ! 868: } else { ! 869: bmap[i] = VDI_UNALLOCATED; ! 870: } ! 871: } ! 872: if (write(fd, bmap, bmap_size) < 0) { ! 873: result = -errno; ! 874: } ! 875: qemu_free(bmap); ! 876: if (image_type == VDI_TYPE_STATIC) { ! 877: if (ftruncate(fd, sizeof(header) + bmap_size + blocks * block_size)) { ! 878: result = -errno; ! 879: } ! 880: } ! 881: ! 882: if (close(fd) < 0) { ! 883: result = -errno; ! 884: } ! 885: ! 886: return result; ! 887: } ! 888: ! 889: static void vdi_close(BlockDriverState *bs) ! 890: { ! 891: BDRVVdiState *s = bs->opaque; ! 892: logout("\n"); ! 893: bdrv_delete(s->hd); ! 894: } ! 895: ! 896: static void vdi_flush(BlockDriverState *bs) ! 897: { ! 898: BDRVVdiState *s = bs->opaque; ! 899: logout("\n"); ! 900: bdrv_flush(s->hd); ! 901: } ! 902: ! 903: ! 904: static QEMUOptionParameter vdi_create_options[] = { ! 905: { ! 906: .name = BLOCK_OPT_SIZE, ! 907: .type = OPT_SIZE, ! 908: .help = "Virtual disk size" ! 909: }, ! 910: #if defined(CONFIG_VDI_BLOCK_SIZE) ! 911: { ! 912: .name = BLOCK_OPT_CLUSTER_SIZE, ! 913: .type = OPT_SIZE, ! 914: .help = "VDI cluster (block) size" ! 915: }, ! 916: #endif ! 917: #if defined(CONFIG_VDI_STATIC_IMAGE) ! 918: { ! 919: .name = BLOCK_OPT_STATIC, ! 920: .type = OPT_FLAG, ! 921: .help = "VDI static (pre-allocated) image" ! 922: }, ! 923: #endif ! 924: /* TODO: An additional option to set UUID values might be useful. */ ! 925: { NULL } ! 926: }; ! 927: ! 928: static BlockDriver bdrv_vdi = { ! 929: .format_name = "vdi", ! 930: .instance_size = sizeof(BDRVVdiState), ! 931: .bdrv_probe = vdi_probe, ! 932: .bdrv_open = vdi_open, ! 933: .bdrv_close = vdi_close, ! 934: .bdrv_create = vdi_create, ! 935: .bdrv_flush = vdi_flush, ! 936: .bdrv_is_allocated = vdi_is_allocated, ! 937: .bdrv_make_empty = vdi_make_empty, ! 938: ! 939: .bdrv_aio_readv = vdi_aio_readv, ! 940: #if defined(CONFIG_VDI_WRITE) ! 941: .bdrv_aio_writev = vdi_aio_writev, ! 942: #endif ! 943: ! 944: .bdrv_get_info = vdi_get_info, ! 945: ! 946: .create_options = vdi_create_options, ! 947: .bdrv_check = vdi_check, ! 948: }; ! 949: ! 950: static void bdrv_vdi_init(void) ! 951: { ! 952: logout("\n"); ! 953: bdrv_register(&bdrv_vdi); ! 954: } ! 955: ! 956: block_init(bdrv_vdi_init);
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.