![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to vdi.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Qemu by Fabrice Bellard] / qemu / block|
Return to vdi.c CVS log | Up to [Qemu by Fabrice Bellard] / qemu / block |
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:
1.1.1.2 ! root 402: if (header.disk_size % SECTOR_SIZE != 0) {
! 403: /* 'VBoxManage convertfromraw' can create images with odd disk sizes.
! 404: We accept them but round the disk size to the next multiple of
! 405: SECTOR_SIZE. */
! 406: logout("odd disk size %" PRIu64 " B, round up\n", header.disk_size);
! 407: header.disk_size += SECTOR_SIZE - 1;
! 408: header.disk_size &= ~(SECTOR_SIZE - 1);
! 409: }
! 410:
1.1 root 411: if (header.version != VDI_VERSION_1_1) {
412: logout("unsupported version %u.%u\n",
413: header.version >> 16, header.version & 0xffff);
414: goto fail;
415: } else if (header.offset_bmap % SECTOR_SIZE != 0) {
416: /* We only support block maps which start on a sector boundary. */
417: logout("unsupported block map offset 0x%x B\n", header.offset_bmap);
418: goto fail;
419: } else if (header.offset_data % SECTOR_SIZE != 0) {
420: /* We only support data blocks which start on a sector boundary. */
421: logout("unsupported data offset 0x%x B\n", header.offset_data);
422: goto fail;
423: } else if (header.sector_size != SECTOR_SIZE) {
424: logout("unsupported sector size %u B\n", header.sector_size);
425: goto fail;
426: } else if (header.block_size != 1 * MiB) {
427: logout("unsupported block size %u B\n", header.block_size);
428: goto fail;
1.1.1.2 ! root 429: } else if (header.disk_size >
1.1 root 430: (uint64_t)header.blocks_in_image * header.block_size) {
1.1.1.2 ! root 431: logout("unsupported disk size %" PRIu64 " B\n", header.disk_size);
1.1 root 432: goto fail;
433: } else if (!uuid_is_null(header.uuid_link)) {
434: logout("link uuid != 0, unsupported\n");
435: goto fail;
436: } else if (!uuid_is_null(header.uuid_parent)) {
437: logout("parent uuid != 0, unsupported\n");
438: goto fail;
439: }
440:
441: bs->total_sectors = header.disk_size / SECTOR_SIZE;
442:
443: s->block_size = header.block_size;
444: s->block_sectors = header.block_size / SECTOR_SIZE;
445: s->bmap_sector = header.offset_bmap / SECTOR_SIZE;
446: s->header = header;
447:
448: bmap_size = header.blocks_in_image * sizeof(uint32_t);
449: bmap_size = (bmap_size + SECTOR_SIZE - 1) / SECTOR_SIZE;
450: s->bmap = qemu_malloc(bmap_size * SECTOR_SIZE);
451: if (bdrv_read(s->hd, s->bmap_sector, (uint8_t *)s->bmap, bmap_size) < 0) {
452: goto fail_free_bmap;
453: }
454:
455: return 0;
456:
457: fail_free_bmap:
458: qemu_free(s->bmap);
459:
460: fail:
461: bdrv_delete(s->hd);
462: return -1;
463: }
464:
465: static int vdi_is_allocated(BlockDriverState *bs, int64_t sector_num,
466: int nb_sectors, int *pnum)
467: {
468: /* TODO: Check for too large sector_num (in bdrv_is_allocated or here). */
469: BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
470: size_t bmap_index = sector_num / s->block_sectors;
471: size_t sector_in_block = sector_num % s->block_sectors;
472: int n_sectors = s->block_sectors - sector_in_block;
473: uint32_t bmap_entry = le32_to_cpu(s->bmap[bmap_index]);
474: logout("%p, %" PRId64 ", %d, %p\n", bs, sector_num, nb_sectors, pnum);
475: if (n_sectors > nb_sectors) {
476: n_sectors = nb_sectors;
477: }
478: *pnum = n_sectors;
479: return bmap_entry != VDI_UNALLOCATED;
480: }
481:
482: static void vdi_aio_cancel(BlockDriverAIOCB *blockacb)
483: {
484: /* TODO: This code is untested. How can I get it executed? */
485: VdiAIOCB *acb = (VdiAIOCB *)blockacb;
486: logout("\n");
487: if (acb->hd_aiocb) {
488: bdrv_aio_cancel(acb->hd_aiocb);
489: }
490: qemu_aio_release(acb);
491: }
492:
493: static AIOPool vdi_aio_pool = {
494: .aiocb_size = sizeof(VdiAIOCB),
495: .cancel = vdi_aio_cancel,
496: };
497:
498: static VdiAIOCB *vdi_aio_setup(BlockDriverState *bs, int64_t sector_num,
499: QEMUIOVector *qiov, int nb_sectors,
500: BlockDriverCompletionFunc *cb, void *opaque, int is_write)
501: {
502: VdiAIOCB *acb;
503:
504: logout("%p, %" PRId64 ", %p, %d, %p, %p, %d\n",
505: bs, sector_num, qiov, nb_sectors, cb, opaque, is_write);
506:
507: acb = qemu_aio_get(&vdi_aio_pool, bs, cb, opaque);
508: if (acb) {
509: acb->hd_aiocb = NULL;
510: acb->sector_num = sector_num;
511: acb->qiov = qiov;
512: if (qiov->niov > 1) {
513: acb->buf = qemu_blockalign(bs, qiov->size);
514: acb->orig_buf = acb->buf;
515: if (is_write) {
516: qemu_iovec_to_buffer(qiov, acb->buf);
517: }
518: } else {
519: acb->buf = (uint8_t *)qiov->iov->iov_base;
520: }
521: acb->nb_sectors = nb_sectors;
522: acb->n_sectors = 0;
523: acb->bmap_first = VDI_UNALLOCATED;
524: acb->bmap_last = VDI_UNALLOCATED;
525: acb->block_buffer = NULL;
526: acb->header_modified = 0;
527: }
528: return acb;
529: }
530:
531: static int vdi_schedule_bh(QEMUBHFunc *cb, VdiAIOCB *acb)
532: {
533: logout("\n");
534:
535: if (acb->bh) {
536: return -EIO;
537: }
538:
539: acb->bh = qemu_bh_new(cb, acb);
540: if (!acb->bh) {
541: return -EIO;
542: }
543:
544: qemu_bh_schedule(acb->bh);
545:
546: return 0;
547: }
548:
549: static void vdi_aio_read_cb(void *opaque, int ret);
550:
551: static void vdi_aio_read_bh(void *opaque)
552: {
553: VdiAIOCB *acb = opaque;
554: logout("\n");
555: qemu_bh_delete(acb->bh);
556: acb->bh = NULL;
557: vdi_aio_read_cb(opaque, 0);
558: }
559:
560: static void vdi_aio_read_cb(void *opaque, int ret)
561: {
562: VdiAIOCB *acb = opaque;
563: BlockDriverState *bs = acb->common.bs;
564: BDRVVdiState *s = bs->opaque;
565: uint32_t bmap_entry;
566: uint32_t block_index;
567: uint32_t sector_in_block;
568: uint32_t n_sectors;
569:
570: logout("%u sectors read\n", acb->n_sectors);
571:
572: acb->hd_aiocb = NULL;
573:
574: if (ret < 0) {
575: goto done;
576: }
577:
578: acb->nb_sectors -= acb->n_sectors;
579:
580: if (acb->nb_sectors == 0) {
581: /* request completed */
582: ret = 0;
583: goto done;
584: }
585:
586: acb->sector_num += acb->n_sectors;
587: acb->buf += acb->n_sectors * SECTOR_SIZE;
588:
589: block_index = acb->sector_num / s->block_sectors;
590: sector_in_block = acb->sector_num % s->block_sectors;
591: n_sectors = s->block_sectors - sector_in_block;
592: if (n_sectors > acb->nb_sectors) {
593: n_sectors = acb->nb_sectors;
594: }
595:
596: logout("will read %u sectors starting at sector %" PRIu64 "\n",
597: n_sectors, acb->sector_num);
598:
599: /* prepare next AIO request */
600: acb->n_sectors = n_sectors;
601: bmap_entry = le32_to_cpu(s->bmap[block_index]);
602: if (bmap_entry == VDI_UNALLOCATED) {
603: /* Block not allocated, return zeros, no need to wait. */
604: memset(acb->buf, 0, n_sectors * SECTOR_SIZE);
605: ret = vdi_schedule_bh(vdi_aio_read_bh, acb);
606: if (ret < 0) {
607: goto done;
608: }
609: } else {
610: uint64_t offset = s->header.offset_data / SECTOR_SIZE +
611: (uint64_t)bmap_entry * s->block_sectors +
612: sector_in_block;
613: acb->hd_iov.iov_base = (void *)acb->buf;
614: acb->hd_iov.iov_len = n_sectors * SECTOR_SIZE;
615: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1);
616: acb->hd_aiocb = bdrv_aio_readv(s->hd, offset, &acb->hd_qiov,
617: n_sectors, vdi_aio_read_cb, acb);
618: if (acb->hd_aiocb == NULL) {
619: goto done;
620: }
621: }
622: return;
623: done:
624: if (acb->qiov->niov > 1) {
625: qemu_iovec_from_buffer(acb->qiov, acb->orig_buf, acb->qiov->size);
626: qemu_vfree(acb->orig_buf);
627: }
628: acb->common.cb(acb->common.opaque, ret);
629: qemu_aio_release(acb);
630: }
631:
632: static BlockDriverAIOCB *vdi_aio_readv(BlockDriverState *bs,
633: int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
634: BlockDriverCompletionFunc *cb, void *opaque)
635: {
636: VdiAIOCB *acb;
637: logout("\n");
638: acb = vdi_aio_setup(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
639: if (!acb) {
640: return NULL;
641: }
642: vdi_aio_read_cb(acb, 0);
643: return &acb->common;
644: }
645:
646: static void vdi_aio_write_cb(void *opaque, int ret)
647: {
648: VdiAIOCB *acb = opaque;
649: BlockDriverState *bs = acb->common.bs;
650: BDRVVdiState *s = bs->opaque;
651: uint32_t bmap_entry;
652: uint32_t block_index;
653: uint32_t sector_in_block;
654: uint32_t n_sectors;
655:
656: acb->hd_aiocb = NULL;
657:
658: if (ret < 0) {
659: goto done;
660: }
661:
662: acb->nb_sectors -= acb->n_sectors;
663: acb->sector_num += acb->n_sectors;
664: acb->buf += acb->n_sectors * SECTOR_SIZE;
665:
666: if (acb->nb_sectors == 0) {
667: logout("finished data write\n");
668: acb->n_sectors = 0;
669: if (acb->header_modified) {
670: VdiHeader *header = acb->block_buffer;
671: logout("now writing modified header\n");
672: assert(acb->bmap_first != VDI_UNALLOCATED);
673: *header = s->header;
674: vdi_header_to_le(header);
675: acb->header_modified = 0;
676: acb->hd_iov.iov_base = acb->block_buffer;
677: acb->hd_iov.iov_len = SECTOR_SIZE;
678: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1);
679: acb->hd_aiocb = bdrv_aio_writev(s->hd, 0, &acb->hd_qiov, 1,
680: vdi_aio_write_cb, acb);
681: if (acb->hd_aiocb == NULL) {
682: goto done;
683: }
684: return;
685: } else if (acb->bmap_first != VDI_UNALLOCATED) {
686: /* One or more new blocks were allocated. */
687: uint64_t offset;
688: uint32_t bmap_first;
689: uint32_t bmap_last;
690: qemu_free(acb->block_buffer);
691: acb->block_buffer = NULL;
692: bmap_first = acb->bmap_first;
693: bmap_last = acb->bmap_last;
694: logout("now writing modified block map entry %u...%u\n",
695: bmap_first, bmap_last);
696: /* Write modified sectors from block map. */
697: bmap_first /= (SECTOR_SIZE / sizeof(uint32_t));
698: bmap_last /= (SECTOR_SIZE / sizeof(uint32_t));
699: n_sectors = bmap_last - bmap_first + 1;
700: offset = s->bmap_sector + bmap_first;
701: acb->bmap_first = VDI_UNALLOCATED;
702: acb->hd_iov.iov_base = (void *)((uint8_t *)&s->bmap[0] +
703: bmap_first * SECTOR_SIZE);
704: acb->hd_iov.iov_len = n_sectors * SECTOR_SIZE;
705: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1);
706: logout("will write %u block map sectors starting from entry %u\n",
707: n_sectors, bmap_first);
708: acb->hd_aiocb = bdrv_aio_writev(s->hd, offset, &acb->hd_qiov,
709: n_sectors, vdi_aio_write_cb, acb);
710: if (acb->hd_aiocb == NULL) {
711: goto done;
712: }
713: return;
714: }
715: ret = 0;
716: goto done;
717: }
718:
719: logout("%u sectors written\n", acb->n_sectors);
720:
721: block_index = acb->sector_num / s->block_sectors;
722: sector_in_block = acb->sector_num % s->block_sectors;
723: n_sectors = s->block_sectors - sector_in_block;
724: if (n_sectors > acb->nb_sectors) {
725: n_sectors = acb->nb_sectors;
726: }
727:
728: logout("will write %u sectors starting at sector %" PRIu64 "\n",
729: n_sectors, acb->sector_num);
730:
731: /* prepare next AIO request */
732: acb->n_sectors = n_sectors;
733: bmap_entry = le32_to_cpu(s->bmap[block_index]);
734: if (bmap_entry == VDI_UNALLOCATED) {
735: /* Allocate new block and write to it. */
736: uint64_t offset;
737: uint8_t *block;
738: bmap_entry = s->header.blocks_allocated;
739: s->bmap[block_index] = cpu_to_le32(bmap_entry);
740: s->header.blocks_allocated++;
741: offset = s->header.offset_data / SECTOR_SIZE +
742: (uint64_t)bmap_entry * s->block_sectors;
743: block = acb->block_buffer;
744: if (block == NULL) {
745: block = qemu_mallocz(s->block_size);
746: acb->block_buffer = block;
747: acb->bmap_first = block_index;
748: assert(!acb->header_modified);
749: acb->header_modified = 1;
750: }
751: acb->bmap_last = block_index;
752: memcpy(block + sector_in_block * SECTOR_SIZE,
753: acb->buf, n_sectors * SECTOR_SIZE);
754: acb->hd_iov.iov_base = (void *)block;
755: acb->hd_iov.iov_len = s->block_size;
756: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1);
757: acb->hd_aiocb = bdrv_aio_writev(s->hd, offset,
758: &acb->hd_qiov, s->block_sectors,
759: vdi_aio_write_cb, acb);
760: if (acb->hd_aiocb == NULL) {
761: goto done;
762: }
763: } else {
764: uint64_t offset = s->header.offset_data / SECTOR_SIZE +
765: (uint64_t)bmap_entry * s->block_sectors +
766: sector_in_block;
767: acb->hd_iov.iov_base = (void *)acb->buf;
768: acb->hd_iov.iov_len = n_sectors * SECTOR_SIZE;
769: qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1);
770: acb->hd_aiocb = bdrv_aio_writev(s->hd, offset, &acb->hd_qiov,
771: n_sectors, vdi_aio_write_cb, acb);
772: if (acb->hd_aiocb == NULL) {
773: goto done;
774: }
775: }
776:
777: return;
778:
779: done:
780: if (acb->qiov->niov > 1) {
781: qemu_vfree(acb->orig_buf);
782: }
783: acb->common.cb(acb->common.opaque, ret);
784: qemu_aio_release(acb);
785: }
786:
787: static BlockDriverAIOCB *vdi_aio_writev(BlockDriverState *bs,
788: int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
789: BlockDriverCompletionFunc *cb, void *opaque)
790: {
791: VdiAIOCB *acb;
792: logout("\n");
793: acb = vdi_aio_setup(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
794: if (!acb) {
795: return NULL;
796: }
797: vdi_aio_write_cb(acb, 0);
798: return &acb->common;
799: }
800:
801: static int vdi_create(const char *filename, QEMUOptionParameter *options)
802: {
803: int fd;
804: int result = 0;
805: uint64_t bytes = 0;
806: uint32_t blocks;
807: size_t block_size = 1 * MiB;
808: uint32_t image_type = VDI_TYPE_DYNAMIC;
809: VdiHeader header;
810: size_t i;
811: size_t bmap_size;
812: uint32_t *bmap;
813:
814: logout("\n");
815:
816: /* Read out options. */
817: while (options && options->name) {
818: if (!strcmp(options->name, BLOCK_OPT_SIZE)) {
819: bytes = options->value.n;
820: #if defined(CONFIG_VDI_BLOCK_SIZE)
821: } else if (!strcmp(options->name, BLOCK_OPT_CLUSTER_SIZE)) {
822: if (options->value.n) {
823: /* TODO: Additional checks (SECTOR_SIZE * 2^n, ...). */
824: block_size = options->value.n;
825: }
826: #endif
827: #if defined(CONFIG_VDI_STATIC_IMAGE)
828: } else if (!strcmp(options->name, BLOCK_OPT_STATIC)) {
829: if (options->value.n) {
830: image_type = VDI_TYPE_STATIC;
831: }
832: #endif
833: }
834: options++;
835: }
836:
837: fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE,
838: 0644);
839: if (fd < 0) {
840: return -errno;
841: }
842:
1.1.1.2 ! root 843: /* We need enough blocks to store the given disk size,
! 844: so always round up. */
! 845: blocks = (bytes + block_size - 1) / block_size;
! 846:
1.1 root 847: bmap_size = blocks * sizeof(uint32_t);
848: bmap_size = ((bmap_size + SECTOR_SIZE - 1) & ~(SECTOR_SIZE -1));
849:
850: memset(&header, 0, sizeof(header));
851: pstrcpy(header.text, sizeof(header.text), VDI_TEXT);
852: header.signature = VDI_SIGNATURE;
853: header.version = VDI_VERSION_1_1;
854: header.header_size = 0x180;
855: header.image_type = image_type;
856: header.offset_bmap = 0x200;
857: header.offset_data = 0x200 + bmap_size;
858: header.sector_size = SECTOR_SIZE;
859: header.disk_size = bytes;
860: header.block_size = block_size;
861: header.blocks_in_image = blocks;
862: if (image_type == VDI_TYPE_STATIC) {
863: header.blocks_allocated = blocks;
864: }
865: uuid_generate(header.uuid_image);
866: uuid_generate(header.uuid_last_snap);
867: /* There is no need to set header.uuid_link or header.uuid_parent here. */
868: #if defined(CONFIG_VDI_DEBUG)
869: vdi_header_print(&header);
870: #endif
871: vdi_header_to_le(&header);
872: if (write(fd, &header, sizeof(header)) < 0) {
873: result = -errno;
874: }
875:
876: bmap = (uint32_t *)qemu_mallocz(bmap_size);
877: for (i = 0; i < blocks; i++) {
878: if (image_type == VDI_TYPE_STATIC) {
879: bmap[i] = i;
880: } else {
881: bmap[i] = VDI_UNALLOCATED;
882: }
883: }
884: if (write(fd, bmap, bmap_size) < 0) {
885: result = -errno;
886: }
887: qemu_free(bmap);
888: if (image_type == VDI_TYPE_STATIC) {
889: if (ftruncate(fd, sizeof(header) + bmap_size + blocks * block_size)) {
890: result = -errno;
891: }
892: }
893:
894: if (close(fd) < 0) {
895: result = -errno;
896: }
897:
898: return result;
899: }
900:
901: static void vdi_close(BlockDriverState *bs)
902: {
903: BDRVVdiState *s = bs->opaque;
904: logout("\n");
905: bdrv_delete(s->hd);
906: }
907:
908: static void vdi_flush(BlockDriverState *bs)
909: {
910: BDRVVdiState *s = bs->opaque;
911: logout("\n");
912: bdrv_flush(s->hd);
913: }
914:
915:
916: static QEMUOptionParameter vdi_create_options[] = {
917: {
918: .name = BLOCK_OPT_SIZE,
919: .type = OPT_SIZE,
920: .help = "Virtual disk size"
921: },
922: #if defined(CONFIG_VDI_BLOCK_SIZE)
923: {
924: .name = BLOCK_OPT_CLUSTER_SIZE,
925: .type = OPT_SIZE,
926: .help = "VDI cluster (block) size"
927: },
928: #endif
929: #if defined(CONFIG_VDI_STATIC_IMAGE)
930: {
931: .name = BLOCK_OPT_STATIC,
932: .type = OPT_FLAG,
933: .help = "VDI static (pre-allocated) image"
934: },
935: #endif
936: /* TODO: An additional option to set UUID values might be useful. */
937: { NULL }
938: };
939:
940: static BlockDriver bdrv_vdi = {
941: .format_name = "vdi",
942: .instance_size = sizeof(BDRVVdiState),
943: .bdrv_probe = vdi_probe,
944: .bdrv_open = vdi_open,
945: .bdrv_close = vdi_close,
946: .bdrv_create = vdi_create,
947: .bdrv_flush = vdi_flush,
948: .bdrv_is_allocated = vdi_is_allocated,
949: .bdrv_make_empty = vdi_make_empty,
950:
951: .bdrv_aio_readv = vdi_aio_readv,
952: #if defined(CONFIG_VDI_WRITE)
953: .bdrv_aio_writev = vdi_aio_writev,
954: #endif
955:
956: .bdrv_get_info = vdi_get_info,
957:
958: .create_options = vdi_create_options,
959: .bdrv_check = vdi_check,
960: };
961:
962: static void bdrv_vdi_init(void)
963: {
964: logout("\n");
965: bdrv_register(&bdrv_vdi);
966: }
967:
968: block_init(bdrv_vdi_init);
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.