![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to vmdk.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 vmdk.c CVS log | Up to [Qemu by Fabrice Bellard] / qemu / block |
1.1 root 1: /*
2: * Block driver for the VMDK format
3: *
4: * Copyright (c) 2004 Fabrice Bellard
5: * Copyright (c) 2005 Filip Navara
6: *
7: * Permission is hereby granted, free of charge, to any person obtaining a copy
8: * of this software and associated documentation files (the "Software"), to deal
9: * in the Software without restriction, including without limitation the rights
10: * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11: * copies of the Software, and to permit persons to whom the Software is
12: * furnished to do so, subject to the following conditions:
13: *
14: * The above copyright notice and this permission notice shall be included in
15: * all copies or substantial portions of the Software.
16: *
17: * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18: * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19: * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20: * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21: * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22: * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
23: * THE SOFTWARE.
24: */
25:
26: #include "qemu-common.h"
27: #include "block_int.h"
28: #include "module.h"
29:
30: #define VMDK3_MAGIC (('C' << 24) | ('O' << 16) | ('W' << 8) | 'D')
31: #define VMDK4_MAGIC (('K' << 24) | ('D' << 16) | ('M' << 8) | 'V')
32:
33: typedef struct {
34: uint32_t version;
35: uint32_t flags;
36: uint32_t disk_sectors;
37: uint32_t granularity;
38: uint32_t l1dir_offset;
39: uint32_t l1dir_size;
40: uint32_t file_sectors;
41: uint32_t cylinders;
42: uint32_t heads;
43: uint32_t sectors_per_track;
44: } VMDK3Header;
45:
46: typedef struct {
47: uint32_t version;
48: uint32_t flags;
49: int64_t capacity;
50: int64_t granularity;
51: int64_t desc_offset;
52: int64_t desc_size;
53: int32_t num_gtes_per_gte;
54: int64_t rgd_offset;
55: int64_t gd_offset;
56: int64_t grain_offset;
57: char filler[1];
58: char check_bytes[4];
59: } __attribute__((packed)) VMDK4Header;
60:
61: #define L2_CACHE_SIZE 16
62:
63: typedef struct BDRVVmdkState {
64: BlockDriverState *hd;
65: int64_t l1_table_offset;
66: int64_t l1_backup_table_offset;
67: uint32_t *l1_table;
68: uint32_t *l1_backup_table;
69: unsigned int l1_size;
70: uint32_t l1_entry_sectors;
71:
72: unsigned int l2_size;
73: uint32_t *l2_cache;
74: uint32_t l2_cache_offsets[L2_CACHE_SIZE];
75: uint32_t l2_cache_counts[L2_CACHE_SIZE];
76:
77: unsigned int cluster_sectors;
78: uint32_t parent_cid;
79: int is_parent;
80: } BDRVVmdkState;
81:
82: typedef struct VmdkMetaData {
83: uint32_t offset;
84: unsigned int l1_index;
85: unsigned int l2_index;
86: unsigned int l2_offset;
87: int valid;
88: } VmdkMetaData;
89:
90: static int vmdk_probe(const uint8_t *buf, int buf_size, const char *filename)
91: {
92: uint32_t magic;
93:
94: if (buf_size < 4)
95: return 0;
96: magic = be32_to_cpu(*(uint32_t *)buf);
97: if (magic == VMDK3_MAGIC ||
98: magic == VMDK4_MAGIC)
99: return 100;
100: else
101: return 0;
102: }
103:
104: #define CHECK_CID 1
105:
106: #define SECTOR_SIZE 512
107: #define DESC_SIZE 20*SECTOR_SIZE // 20 sectors of 512 bytes each
108: #define HEADER_SIZE 512 // first sector of 512 bytes
109:
110: static uint32_t vmdk_read_cid(BlockDriverState *bs, int parent)
111: {
112: BDRVVmdkState *s = bs->opaque;
113: char desc[DESC_SIZE];
114: uint32_t cid;
115: const char *p_name, *cid_str;
116: size_t cid_str_size;
117:
118: /* the descriptor offset = 0x200 */
119: if (bdrv_pread(s->hd, 0x200, desc, DESC_SIZE) != DESC_SIZE)
120: return 0;
121:
122: if (parent) {
123: cid_str = "parentCID";
124: cid_str_size = sizeof("parentCID");
125: } else {
126: cid_str = "CID";
127: cid_str_size = sizeof("CID");
128: }
129:
130: if ((p_name = strstr(desc,cid_str)) != NULL) {
131: p_name += cid_str_size;
132: sscanf(p_name,"%x",&cid);
133: }
134:
135: return cid;
136: }
137:
138: static int vmdk_write_cid(BlockDriverState *bs, uint32_t cid)
139: {
140: BDRVVmdkState *s = bs->opaque;
141: char desc[DESC_SIZE], tmp_desc[DESC_SIZE];
142: char *p_name, *tmp_str;
143:
144: /* the descriptor offset = 0x200 */
145: if (bdrv_pread(s->hd, 0x200, desc, DESC_SIZE) != DESC_SIZE)
146: return -1;
147:
148: tmp_str = strstr(desc,"parentCID");
149: pstrcpy(tmp_desc, sizeof(tmp_desc), tmp_str);
150: if ((p_name = strstr(desc,"CID")) != NULL) {
151: p_name += sizeof("CID");
152: snprintf(p_name, sizeof(desc) - (p_name - desc), "%x\n", cid);
153: pstrcat(desc, sizeof(desc), tmp_desc);
154: }
155:
1.1.1.2 ! root 156: if (bdrv_pwrite_sync(s->hd, 0x200, desc, DESC_SIZE) < 0)
1.1 root 157: return -1;
158: return 0;
159: }
160:
161: static int vmdk_is_cid_valid(BlockDriverState *bs)
162: {
163: #ifdef CHECK_CID
164: BDRVVmdkState *s = bs->opaque;
165: BlockDriverState *p_bs = bs->backing_hd;
166: uint32_t cur_pcid;
167:
168: if (p_bs) {
169: cur_pcid = vmdk_read_cid(p_bs,0);
170: if (s->parent_cid != cur_pcid)
171: // CID not valid
172: return 0;
173: }
174: #endif
175: // CID valid
176: return 1;
177: }
178:
179: static int vmdk_snapshot_create(const char *filename, const char *backing_file)
180: {
181: int snp_fd, p_fd;
182: uint32_t p_cid;
183: char *p_name, *gd_buf, *rgd_buf;
184: const char *real_filename, *temp_str;
185: VMDK4Header header;
186: uint32_t gde_entries, gd_size;
187: int64_t gd_offset, rgd_offset, capacity, gt_size;
188: char p_desc[DESC_SIZE], s_desc[DESC_SIZE], hdr[HEADER_SIZE];
189: static const char desc_template[] =
190: "# Disk DescriptorFile\n"
191: "version=1\n"
192: "CID=%x\n"
193: "parentCID=%x\n"
194: "createType=\"monolithicSparse\"\n"
195: "parentFileNameHint=\"%s\"\n"
196: "\n"
197: "# Extent description\n"
198: "RW %u SPARSE \"%s\"\n"
199: "\n"
200: "# The Disk Data Base \n"
201: "#DDB\n"
202: "\n";
203:
204: snp_fd = open(filename, O_RDWR | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE, 0644);
205: if (snp_fd < 0)
206: return -1;
207: p_fd = open(backing_file, O_RDONLY | O_BINARY | O_LARGEFILE);
208: if (p_fd < 0) {
209: close(snp_fd);
210: return -1;
211: }
212:
213: /* read the header */
214: if (lseek(p_fd, 0x0, SEEK_SET) == -1)
215: goto fail;
216: if (read(p_fd, hdr, HEADER_SIZE) != HEADER_SIZE)
217: goto fail;
218:
219: /* write the header */
220: if (lseek(snp_fd, 0x0, SEEK_SET) == -1)
221: goto fail;
222: if (write(snp_fd, hdr, HEADER_SIZE) == -1)
223: goto fail;
224:
225: memset(&header, 0, sizeof(header));
226: memcpy(&header,&hdr[4], sizeof(header)); // skip the VMDK4_MAGIC
227:
228: ftruncate(snp_fd, header.grain_offset << 9);
229: /* the descriptor offset = 0x200 */
230: if (lseek(p_fd, 0x200, SEEK_SET) == -1)
231: goto fail;
232: if (read(p_fd, p_desc, DESC_SIZE) != DESC_SIZE)
233: goto fail;
234:
235: if ((p_name = strstr(p_desc,"CID")) != NULL) {
236: p_name += sizeof("CID");
237: sscanf(p_name,"%x",&p_cid);
238: }
239:
240: real_filename = filename;
241: if ((temp_str = strrchr(real_filename, '\\')) != NULL)
242: real_filename = temp_str + 1;
243: if ((temp_str = strrchr(real_filename, '/')) != NULL)
244: real_filename = temp_str + 1;
245: if ((temp_str = strrchr(real_filename, ':')) != NULL)
246: real_filename = temp_str + 1;
247:
248: snprintf(s_desc, sizeof(s_desc), desc_template, p_cid, p_cid, backing_file,
249: (uint32_t)header.capacity, real_filename);
250:
251: /* write the descriptor */
252: if (lseek(snp_fd, 0x200, SEEK_SET) == -1)
253: goto fail;
254: if (write(snp_fd, s_desc, strlen(s_desc)) == -1)
255: goto fail;
256:
257: gd_offset = header.gd_offset * SECTOR_SIZE; // offset of GD table
258: rgd_offset = header.rgd_offset * SECTOR_SIZE; // offset of RGD table
259: capacity = header.capacity * SECTOR_SIZE; // Extent size
260: /*
261: * Each GDE span 32M disk, means:
262: * 512 GTE per GT, each GTE points to grain
263: */
264: gt_size = (int64_t)header.num_gtes_per_gte * header.granularity * SECTOR_SIZE;
265: if (!gt_size)
266: goto fail;
267: gde_entries = (uint32_t)(capacity / gt_size); // number of gde/rgde
268: gd_size = gde_entries * sizeof(uint32_t);
269:
270: /* write RGD */
271: rgd_buf = qemu_malloc(gd_size);
272: if (lseek(p_fd, rgd_offset, SEEK_SET) == -1)
273: goto fail_rgd;
274: if (read(p_fd, rgd_buf, gd_size) != gd_size)
275: goto fail_rgd;
276: if (lseek(snp_fd, rgd_offset, SEEK_SET) == -1)
277: goto fail_rgd;
278: if (write(snp_fd, rgd_buf, gd_size) == -1)
279: goto fail_rgd;
280:
281: /* write GD */
282: gd_buf = qemu_malloc(gd_size);
283: if (lseek(p_fd, gd_offset, SEEK_SET) == -1)
284: goto fail_gd;
285: if (read(p_fd, gd_buf, gd_size) != gd_size)
286: goto fail_gd;
287: if (lseek(snp_fd, gd_offset, SEEK_SET) == -1)
288: goto fail_gd;
289: if (write(snp_fd, gd_buf, gd_size) == -1)
290: goto fail_gd;
291: qemu_free(gd_buf);
1.1.1.2 ! root 292: qemu_free(rgd_buf);
1.1 root 293:
294: close(p_fd);
295: close(snp_fd);
296: return 0;
297:
298: fail_gd:
299: qemu_free(gd_buf);
300: fail_rgd:
301: qemu_free(rgd_buf);
302: fail:
303: close(p_fd);
304: close(snp_fd);
305: return -1;
306: }
307:
308: static void vmdk_parent_close(BlockDriverState *bs)
309: {
310: if (bs->backing_hd)
311: bdrv_close(bs->backing_hd);
312: }
313:
314: static int parent_open = 0;
315: static int vmdk_parent_open(BlockDriverState *bs, const char * filename)
316: {
317: BDRVVmdkState *s = bs->opaque;
318: char *p_name;
319: char desc[DESC_SIZE];
320: char parent_img_name[1024];
321:
322: /* the descriptor offset = 0x200 */
323: if (bdrv_pread(s->hd, 0x200, desc, DESC_SIZE) != DESC_SIZE)
324: return -1;
325:
326: if ((p_name = strstr(desc,"parentFileNameHint")) != NULL) {
327: char *end_name;
328: struct stat file_buf;
329:
330: p_name += sizeof("parentFileNameHint") + 1;
331: if ((end_name = strchr(p_name,'\"')) == NULL)
332: return -1;
333: if ((end_name - p_name) > sizeof (bs->backing_file) - 1)
334: return -1;
335:
336: pstrcpy(bs->backing_file, end_name - p_name + 1, p_name);
337: if (stat(bs->backing_file, &file_buf) != 0) {
338: path_combine(parent_img_name, sizeof(parent_img_name),
339: filename, bs->backing_file);
340: } else {
341: pstrcpy(parent_img_name, sizeof(parent_img_name),
342: bs->backing_file);
343: }
344:
345: bs->backing_hd = bdrv_new("");
346: if (!bs->backing_hd) {
347: failure:
348: bdrv_close(s->hd);
349: return -1;
350: }
351: parent_open = 1;
352: if (bdrv_open(bs->backing_hd, parent_img_name, BDRV_O_RDONLY) < 0)
353: goto failure;
354: parent_open = 0;
355: }
356:
357: return 0;
358: }
359:
360: static int vmdk_open(BlockDriverState *bs, const char *filename, int flags)
361: {
362: BDRVVmdkState *s = bs->opaque;
363: uint32_t magic;
364: int l1_size, i, ret;
365:
366: if (parent_open)
367: // Parent must be opened as RO.
368: flags = BDRV_O_RDONLY;
369:
370: ret = bdrv_file_open(&s->hd, filename, flags);
371: if (ret < 0)
372: return ret;
373: if (bdrv_pread(s->hd, 0, &magic, sizeof(magic)) != sizeof(magic))
374: goto fail;
375:
376: magic = be32_to_cpu(magic);
377: if (magic == VMDK3_MAGIC) {
378: VMDK3Header header;
379:
380: if (bdrv_pread(s->hd, sizeof(magic), &header, sizeof(header)) != sizeof(header))
381: goto fail;
382: s->cluster_sectors = le32_to_cpu(header.granularity);
383: s->l2_size = 1 << 9;
384: s->l1_size = 1 << 6;
385: bs->total_sectors = le32_to_cpu(header.disk_sectors);
386: s->l1_table_offset = le32_to_cpu(header.l1dir_offset) << 9;
387: s->l1_backup_table_offset = 0;
388: s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
389: } else if (magic == VMDK4_MAGIC) {
390: VMDK4Header header;
391:
392: if (bdrv_pread(s->hd, sizeof(magic), &header, sizeof(header)) != sizeof(header))
393: goto fail;
394: bs->total_sectors = le64_to_cpu(header.capacity);
395: s->cluster_sectors = le64_to_cpu(header.granularity);
396: s->l2_size = le32_to_cpu(header.num_gtes_per_gte);
397: s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
398: if (s->l1_entry_sectors <= 0)
399: goto fail;
400: s->l1_size = (bs->total_sectors + s->l1_entry_sectors - 1)
401: / s->l1_entry_sectors;
402: s->l1_table_offset = le64_to_cpu(header.rgd_offset) << 9;
403: s->l1_backup_table_offset = le64_to_cpu(header.gd_offset) << 9;
404:
405: if (parent_open)
406: s->is_parent = 1;
407: else
408: s->is_parent = 0;
409:
410: // try to open parent images, if exist
411: if (vmdk_parent_open(bs, filename) != 0)
412: goto fail;
413: // write the CID once after the image creation
414: s->parent_cid = vmdk_read_cid(bs,1);
415: } else {
416: goto fail;
417: }
418:
419: /* read the L1 table */
420: l1_size = s->l1_size * sizeof(uint32_t);
421: s->l1_table = qemu_malloc(l1_size);
422: if (bdrv_pread(s->hd, s->l1_table_offset, s->l1_table, l1_size) != l1_size)
423: goto fail;
424: for(i = 0; i < s->l1_size; i++) {
425: le32_to_cpus(&s->l1_table[i]);
426: }
427:
428: if (s->l1_backup_table_offset) {
429: s->l1_backup_table = qemu_malloc(l1_size);
430: if (bdrv_pread(s->hd, s->l1_backup_table_offset, s->l1_backup_table, l1_size) != l1_size)
431: goto fail;
432: for(i = 0; i < s->l1_size; i++) {
433: le32_to_cpus(&s->l1_backup_table[i]);
434: }
435: }
436:
437: s->l2_cache = qemu_malloc(s->l2_size * L2_CACHE_SIZE * sizeof(uint32_t));
438: return 0;
439: fail:
440: qemu_free(s->l1_backup_table);
441: qemu_free(s->l1_table);
442: qemu_free(s->l2_cache);
443: bdrv_delete(s->hd);
444: return -1;
445: }
446:
447: static uint64_t get_cluster_offset(BlockDriverState *bs, VmdkMetaData *m_data,
448: uint64_t offset, int allocate);
449:
450: static int get_whole_cluster(BlockDriverState *bs, uint64_t cluster_offset,
451: uint64_t offset, int allocate)
452: {
453: BDRVVmdkState *s = bs->opaque;
454: uint8_t whole_grain[s->cluster_sectors*512]; // 128 sectors * 512 bytes each = grain size 64KB
455:
456: // we will be here if it's first write on non-exist grain(cluster).
457: // try to read from parent image, if exist
458: if (bs->backing_hd) {
1.1.1.2 ! root 459: int ret;
1.1 root 460:
461: if (!vmdk_is_cid_valid(bs))
462: return -1;
463:
1.1.1.2 ! root 464: ret = bdrv_read(bs->backing_hd, offset >> 9, whole_grain,
! 465: s->cluster_sectors);
! 466: if (ret < 0) {
! 467: return -1;
! 468: }
1.1 root 469:
1.1.1.2 ! root 470: //Write grain only into the active image
! 471: ret = bdrv_write(s->hd, cluster_offset, whole_grain,
! 472: s->cluster_sectors);
! 473: if (ret < 0) {
! 474: return -1;
1.1 root 475: }
476: }
477: return 0;
478: }
479:
480: static int vmdk_L2update(BlockDriverState *bs, VmdkMetaData *m_data)
481: {
482: BDRVVmdkState *s = bs->opaque;
483:
484: /* update L2 table */
1.1.1.2 ! root 485: if (bdrv_pwrite_sync(s->hd, ((int64_t)m_data->l2_offset * 512) + (m_data->l2_index * sizeof(m_data->offset)),
! 486: &(m_data->offset), sizeof(m_data->offset)) < 0)
1.1 root 487: return -1;
488: /* update backup L2 table */
489: if (s->l1_backup_table_offset != 0) {
490: m_data->l2_offset = s->l1_backup_table[m_data->l1_index];
1.1.1.2 ! root 491: if (bdrv_pwrite_sync(s->hd, ((int64_t)m_data->l2_offset * 512) + (m_data->l2_index * sizeof(m_data->offset)),
! 492: &(m_data->offset), sizeof(m_data->offset)) < 0)
1.1 root 493: return -1;
494: }
495:
496: return 0;
497: }
498:
499: static uint64_t get_cluster_offset(BlockDriverState *bs, VmdkMetaData *m_data,
500: uint64_t offset, int allocate)
501: {
502: BDRVVmdkState *s = bs->opaque;
503: unsigned int l1_index, l2_offset, l2_index;
504: int min_index, i, j;
505: uint32_t min_count, *l2_table, tmp = 0;
506: uint64_t cluster_offset;
507:
508: if (m_data)
509: m_data->valid = 0;
510:
511: l1_index = (offset >> 9) / s->l1_entry_sectors;
512: if (l1_index >= s->l1_size)
513: return 0;
514: l2_offset = s->l1_table[l1_index];
515: if (!l2_offset)
516: return 0;
517: for(i = 0; i < L2_CACHE_SIZE; i++) {
518: if (l2_offset == s->l2_cache_offsets[i]) {
519: /* increment the hit count */
520: if (++s->l2_cache_counts[i] == 0xffffffff) {
521: for(j = 0; j < L2_CACHE_SIZE; j++) {
522: s->l2_cache_counts[j] >>= 1;
523: }
524: }
525: l2_table = s->l2_cache + (i * s->l2_size);
526: goto found;
527: }
528: }
529: /* not found: load a new entry in the least used one */
530: min_index = 0;
531: min_count = 0xffffffff;
532: for(i = 0; i < L2_CACHE_SIZE; i++) {
533: if (s->l2_cache_counts[i] < min_count) {
534: min_count = s->l2_cache_counts[i];
535: min_index = i;
536: }
537: }
538: l2_table = s->l2_cache + (min_index * s->l2_size);
539: if (bdrv_pread(s->hd, (int64_t)l2_offset * 512, l2_table, s->l2_size * sizeof(uint32_t)) !=
540: s->l2_size * sizeof(uint32_t))
541: return 0;
542:
543: s->l2_cache_offsets[min_index] = l2_offset;
544: s->l2_cache_counts[min_index] = 1;
545: found:
546: l2_index = ((offset >> 9) / s->cluster_sectors) % s->l2_size;
547: cluster_offset = le32_to_cpu(l2_table[l2_index]);
548:
549: if (!cluster_offset) {
550: if (!allocate)
551: return 0;
552: // Avoid the L2 tables update for the images that have snapshots.
553: if (!s->is_parent) {
554: cluster_offset = bdrv_getlength(s->hd);
555: bdrv_truncate(s->hd, cluster_offset + (s->cluster_sectors << 9));
556:
557: cluster_offset >>= 9;
558: tmp = cpu_to_le32(cluster_offset);
559: l2_table[l2_index] = tmp;
560: }
561: /* First of all we write grain itself, to avoid race condition
562: * that may to corrupt the image.
563: * This problem may occur because of insufficient space on host disk
564: * or inappropriate VM shutdown.
565: */
566: if (get_whole_cluster(bs, cluster_offset, offset, allocate) == -1)
567: return 0;
568:
569: if (m_data) {
570: m_data->offset = tmp;
571: m_data->l1_index = l1_index;
572: m_data->l2_index = l2_index;
573: m_data->l2_offset = l2_offset;
574: m_data->valid = 1;
575: }
576: }
577: cluster_offset <<= 9;
578: return cluster_offset;
579: }
580:
581: static int vmdk_is_allocated(BlockDriverState *bs, int64_t sector_num,
582: int nb_sectors, int *pnum)
583: {
584: BDRVVmdkState *s = bs->opaque;
585: int index_in_cluster, n;
586: uint64_t cluster_offset;
587:
588: cluster_offset = get_cluster_offset(bs, NULL, sector_num << 9, 0);
589: index_in_cluster = sector_num % s->cluster_sectors;
590: n = s->cluster_sectors - index_in_cluster;
591: if (n > nb_sectors)
592: n = nb_sectors;
593: *pnum = n;
594: return (cluster_offset != 0);
595: }
596:
597: static int vmdk_read(BlockDriverState *bs, int64_t sector_num,
598: uint8_t *buf, int nb_sectors)
599: {
600: BDRVVmdkState *s = bs->opaque;
601: int index_in_cluster, n, ret;
602: uint64_t cluster_offset;
603:
604: while (nb_sectors > 0) {
605: cluster_offset = get_cluster_offset(bs, NULL, sector_num << 9, 0);
606: index_in_cluster = sector_num % s->cluster_sectors;
607: n = s->cluster_sectors - index_in_cluster;
608: if (n > nb_sectors)
609: n = nb_sectors;
610: if (!cluster_offset) {
611: // try to read from parent image, if exist
612: if (bs->backing_hd) {
613: if (!vmdk_is_cid_valid(bs))
614: return -1;
615: ret = bdrv_read(bs->backing_hd, sector_num, buf, n);
616: if (ret < 0)
617: return -1;
618: } else {
619: memset(buf, 0, 512 * n);
620: }
621: } else {
622: if(bdrv_pread(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512) != n * 512)
623: return -1;
624: }
625: nb_sectors -= n;
626: sector_num += n;
627: buf += n * 512;
628: }
629: return 0;
630: }
631:
632: static int vmdk_write(BlockDriverState *bs, int64_t sector_num,
633: const uint8_t *buf, int nb_sectors)
634: {
635: BDRVVmdkState *s = bs->opaque;
636: VmdkMetaData m_data;
637: int index_in_cluster, n;
638: uint64_t cluster_offset;
639: static int cid_update = 0;
640:
641: if (sector_num > bs->total_sectors) {
642: fprintf(stderr,
643: "(VMDK) Wrong offset: sector_num=0x%" PRIx64
644: " total_sectors=0x%" PRIx64 "\n",
645: sector_num, bs->total_sectors);
646: return -1;
647: }
648:
649: while (nb_sectors > 0) {
650: index_in_cluster = sector_num & (s->cluster_sectors - 1);
651: n = s->cluster_sectors - index_in_cluster;
652: if (n > nb_sectors)
653: n = nb_sectors;
654: cluster_offset = get_cluster_offset(bs, &m_data, sector_num << 9, 1);
655: if (!cluster_offset)
656: return -1;
657:
658: if (bdrv_pwrite(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512) != n * 512)
659: return -1;
660: if (m_data.valid) {
661: /* update L2 tables */
662: if (vmdk_L2update(bs, &m_data) == -1)
663: return -1;
664: }
665: nb_sectors -= n;
666: sector_num += n;
667: buf += n * 512;
668:
669: // update CID on the first write every time the virtual disk is opened
670: if (!cid_update) {
671: vmdk_write_cid(bs, time(NULL));
672: cid_update++;
673: }
674: }
675: return 0;
676: }
677:
678: static int vmdk_create(const char *filename, QEMUOptionParameter *options)
679: {
680: int fd, i;
681: VMDK4Header header;
682: uint32_t tmp, magic, grains, gd_size, gt_size, gt_count;
683: static const char desc_template[] =
684: "# Disk DescriptorFile\n"
685: "version=1\n"
686: "CID=%x\n"
687: "parentCID=ffffffff\n"
688: "createType=\"monolithicSparse\"\n"
689: "\n"
690: "# Extent description\n"
691: "RW %" PRId64 " SPARSE \"%s\"\n"
692: "\n"
693: "# The Disk Data Base \n"
694: "#DDB\n"
695: "\n"
696: "ddb.virtualHWVersion = \"%d\"\n"
697: "ddb.geometry.cylinders = \"%" PRId64 "\"\n"
698: "ddb.geometry.heads = \"16\"\n"
699: "ddb.geometry.sectors = \"63\"\n"
700: "ddb.adapterType = \"ide\"\n";
701: char desc[1024];
702: const char *real_filename, *temp_str;
703: int64_t total_size = 0;
704: const char *backing_file = NULL;
705: int flags = 0;
706:
707: // Read out options
708: while (options && options->name) {
709: if (!strcmp(options->name, BLOCK_OPT_SIZE)) {
710: total_size = options->value.n / 512;
711: } else if (!strcmp(options->name, BLOCK_OPT_BACKING_FILE)) {
712: backing_file = options->value.s;
713: } else if (!strcmp(options->name, BLOCK_OPT_COMPAT6)) {
714: flags |= options->value.n ? BLOCK_FLAG_COMPAT6: 0;
715: }
716: options++;
717: }
718:
719: /* XXX: add support for backing file */
720: if (backing_file) {
721: return vmdk_snapshot_create(filename, backing_file);
722: }
723:
724: fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE,
725: 0644);
726: if (fd < 0)
727: return -1;
728: magic = cpu_to_be32(VMDK4_MAGIC);
729: memset(&header, 0, sizeof(header));
730: header.version = cpu_to_le32(1);
731: header.flags = cpu_to_le32(3); /* ?? */
732: header.capacity = cpu_to_le64(total_size);
733: header.granularity = cpu_to_le64(128);
734: header.num_gtes_per_gte = cpu_to_le32(512);
735:
736: grains = (total_size + header.granularity - 1) / header.granularity;
737: gt_size = ((header.num_gtes_per_gte * sizeof(uint32_t)) + 511) >> 9;
738: gt_count = (grains + header.num_gtes_per_gte - 1) / header.num_gtes_per_gte;
739: gd_size = (gt_count * sizeof(uint32_t) + 511) >> 9;
740:
741: header.desc_offset = 1;
742: header.desc_size = 20;
743: header.rgd_offset = header.desc_offset + header.desc_size;
744: header.gd_offset = header.rgd_offset + gd_size + (gt_size * gt_count);
745: header.grain_offset =
746: ((header.gd_offset + gd_size + (gt_size * gt_count) +
747: header.granularity - 1) / header.granularity) *
748: header.granularity;
749:
750: header.desc_offset = cpu_to_le64(header.desc_offset);
751: header.desc_size = cpu_to_le64(header.desc_size);
752: header.rgd_offset = cpu_to_le64(header.rgd_offset);
753: header.gd_offset = cpu_to_le64(header.gd_offset);
754: header.grain_offset = cpu_to_le64(header.grain_offset);
755:
756: header.check_bytes[0] = 0xa;
757: header.check_bytes[1] = 0x20;
758: header.check_bytes[2] = 0xd;
759: header.check_bytes[3] = 0xa;
760:
761: /* write all the data */
762: write(fd, &magic, sizeof(magic));
763: write(fd, &header, sizeof(header));
764:
765: ftruncate(fd, header.grain_offset << 9);
766:
767: /* write grain directory */
768: lseek(fd, le64_to_cpu(header.rgd_offset) << 9, SEEK_SET);
769: for (i = 0, tmp = header.rgd_offset + gd_size;
770: i < gt_count; i++, tmp += gt_size)
771: write(fd, &tmp, sizeof(tmp));
772:
773: /* write backup grain directory */
774: lseek(fd, le64_to_cpu(header.gd_offset) << 9, SEEK_SET);
775: for (i = 0, tmp = header.gd_offset + gd_size;
776: i < gt_count; i++, tmp += gt_size)
777: write(fd, &tmp, sizeof(tmp));
778:
779: /* compose the descriptor */
780: real_filename = filename;
781: if ((temp_str = strrchr(real_filename, '\\')) != NULL)
782: real_filename = temp_str + 1;
783: if ((temp_str = strrchr(real_filename, '/')) != NULL)
784: real_filename = temp_str + 1;
785: if ((temp_str = strrchr(real_filename, ':')) != NULL)
786: real_filename = temp_str + 1;
787: snprintf(desc, sizeof(desc), desc_template, (unsigned int)time(NULL),
788: total_size, real_filename,
789: (flags & BLOCK_FLAG_COMPAT6 ? 6 : 4),
790: total_size / (int64_t)(63 * 16));
791:
792: /* write the descriptor */
793: lseek(fd, le64_to_cpu(header.desc_offset) << 9, SEEK_SET);
794: write(fd, desc, strlen(desc));
795:
796: close(fd);
797: return 0;
798: }
799:
800: static void vmdk_close(BlockDriverState *bs)
801: {
802: BDRVVmdkState *s = bs->opaque;
803:
804: qemu_free(s->l1_table);
805: qemu_free(s->l2_cache);
806: // try to close parent image, if exist
807: vmdk_parent_close(s->hd);
808: bdrv_delete(s->hd);
809: }
810:
811: static void vmdk_flush(BlockDriverState *bs)
812: {
813: BDRVVmdkState *s = bs->opaque;
814: bdrv_flush(s->hd);
815: }
816:
817:
818: static QEMUOptionParameter vmdk_create_options[] = {
819: {
820: .name = BLOCK_OPT_SIZE,
821: .type = OPT_SIZE,
822: .help = "Virtual disk size"
823: },
824: {
825: .name = BLOCK_OPT_BACKING_FILE,
826: .type = OPT_STRING,
827: .help = "File name of a base image"
828: },
829: {
830: .name = BLOCK_OPT_COMPAT6,
831: .type = OPT_FLAG,
832: .help = "VMDK version 6 image"
833: },
834: { NULL }
835: };
836:
837: static BlockDriver bdrv_vmdk = {
838: .format_name = "vmdk",
839: .instance_size = sizeof(BDRVVmdkState),
840: .bdrv_probe = vmdk_probe,
841: .bdrv_open = vmdk_open,
842: .bdrv_read = vmdk_read,
843: .bdrv_write = vmdk_write,
844: .bdrv_close = vmdk_close,
845: .bdrv_create = vmdk_create,
846: .bdrv_flush = vmdk_flush,
847: .bdrv_is_allocated = vmdk_is_allocated,
848:
849: .create_options = vmdk_create_options,
850: };
851:
852: static void bdrv_vmdk_init(void)
853: {
854: bdrv_register(&bdrv_vmdk);
855: }
856:
857: block_init(bdrv_vmdk_init);
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.