]> git.proxmox.com Git - qemu.git/blame - block/vdi.c
qmp: fix handling of cmd with Equals in qmp-shell
[qemu.git] / block / vdi.c
CommitLineData
9aebd98a
SW
1/*
2 * Block driver for the Virtual Disk Image (VDI) format
3 *
641543b7 4 * Copyright (c) 2009, 2012 Stefan Weil
9aebd98a
SW
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"
737e150e 53#include "block/block_int.h"
1de7afc9 54#include "qemu/module.h"
caf71f86 55#include "migration/migration.h"
9aebd98a 56
ee682d27 57#if defined(CONFIG_UUID)
9aebd98a
SW
58#include <uuid/uuid.h>
59#else
60/* TODO: move uuid emulation to some central place in QEMU. */
9c17d615 61#include "sysemu/sysemu.h" /* UUID_FMT */
9aebd98a 62typedef unsigned char uuid_t[16];
9aebd98a
SW
63#endif
64
65/* Code configuration options. */
66
67/* Enable debug messages. */
68//~ #define CONFIG_VDI_DEBUG
69
70/* Support write operations on VDI images. */
71#define CONFIG_VDI_WRITE
72
73/* Support non-standard block (cluster) size. This is untested.
74 * Maybe it will be needed for very large images.
75 */
76//~ #define CONFIG_VDI_BLOCK_SIZE
77
78/* Support static (fixed, pre-allocated) images. */
79#define CONFIG_VDI_STATIC_IMAGE
80
81/* Command line option for static images. */
82#define BLOCK_OPT_STATIC "static"
83
84#define KiB 1024
85#define MiB (KiB * KiB)
86
87#define SECTOR_SIZE 512
99cce9fa 88#define DEFAULT_CLUSTER_SIZE (1 * MiB)
9aebd98a
SW
89
90#if defined(CONFIG_VDI_DEBUG)
91#define logout(fmt, ...) \
92 fprintf(stderr, "vdi\t%-24s" fmt, __func__, ##__VA_ARGS__)
93#else
94#define logout(fmt, ...) ((void)0)
95#endif
96
97/* Image signature. */
98#define VDI_SIGNATURE 0xbeda107f
99
100/* Image version. */
101#define VDI_VERSION_1_1 0x00010001
102
103/* Image type. */
104#define VDI_TYPE_DYNAMIC 1
105#define VDI_TYPE_STATIC 2
106
107/* Innotek / SUN images use these strings in header.text:
108 * "<<< innotek VirtualBox Disk Image >>>\n"
109 * "<<< Sun xVM VirtualBox Disk Image >>>\n"
110 * "<<< Sun VirtualBox Disk Image >>>\n"
111 * The value does not matter, so QEMU created images use a different text.
112 */
113#define VDI_TEXT "<<< QEMU VM Virtual Disk Image >>>\n"
114
c794b4e0
ES
115/* A never-allocated block; semantically arbitrary content. */
116#define VDI_UNALLOCATED 0xffffffffU
117
118/* A discarded (no longer allocated) block; semantically zero-filled. */
119#define VDI_DISCARDED 0xfffffffeU
120
121#define VDI_IS_ALLOCATED(X) ((X) < VDI_DISCARDED)
9aebd98a 122
ee682d27 123#if !defined(CONFIG_UUID)
8ba2aae3 124static inline void uuid_generate(uuid_t out)
9aebd98a 125{
4f3669ea 126 memset(out, 0, sizeof(uuid_t));
9aebd98a
SW
127}
128
8ba2aae3 129static inline int uuid_is_null(const uuid_t uu)
9aebd98a
SW
130{
131 uuid_t null_uuid = { 0 };
4f3669ea 132 return memcmp(uu, null_uuid, sizeof(uuid_t)) == 0;
9aebd98a
SW
133}
134
8ba2aae3 135static inline void uuid_unparse(const uuid_t uu, char *out)
9aebd98a
SW
136{
137 snprintf(out, 37, UUID_FMT,
138 uu[0], uu[1], uu[2], uu[3], uu[4], uu[5], uu[6], uu[7],
139 uu[8], uu[9], uu[10], uu[11], uu[12], uu[13], uu[14], uu[15]);
140}
141#endif
142
9aebd98a
SW
143typedef struct {
144 char text[0x40];
145 uint32_t signature;
146 uint32_t version;
147 uint32_t header_size;
148 uint32_t image_type;
149 uint32_t image_flags;
150 char description[256];
151 uint32_t offset_bmap;
152 uint32_t offset_data;
153 uint32_t cylinders; /* disk geometry, unused here */
154 uint32_t heads; /* disk geometry, unused here */
155 uint32_t sectors; /* disk geometry, unused here */
156 uint32_t sector_size;
157 uint32_t unused1;
158 uint64_t disk_size;
159 uint32_t block_size;
160 uint32_t block_extra; /* unused here */
161 uint32_t blocks_in_image;
162 uint32_t blocks_allocated;
163 uuid_t uuid_image;
164 uuid_t uuid_last_snap;
165 uuid_t uuid_link;
166 uuid_t uuid_parent;
167 uint64_t unused2[7];
168} VdiHeader;
169
170typedef struct {
9aebd98a
SW
171 /* The block map entries are little endian (even in memory). */
172 uint32_t *bmap;
173 /* Size of block (bytes). */
174 uint32_t block_size;
175 /* Size of block (sectors). */
176 uint32_t block_sectors;
177 /* First sector of block map. */
178 uint32_t bmap_sector;
4ff9786c 179 /* VDI header (converted to host endianness). */
9aebd98a 180 VdiHeader header;
fc9d106c
KW
181
182 Error *migration_blocker;
9aebd98a
SW
183} BDRVVdiState;
184
185/* Change UUID from little endian (IPRT = VirtualBox format) to big endian
186 * format (network byte order, standard, see RFC 4122) and vice versa.
187 */
188static void uuid_convert(uuid_t uuid)
189{
190 bswap32s((uint32_t *)&uuid[0]);
191 bswap16s((uint16_t *)&uuid[4]);
192 bswap16s((uint16_t *)&uuid[6]);
193}
194
195static void vdi_header_to_cpu(VdiHeader *header)
196{
197 le32_to_cpus(&header->signature);
198 le32_to_cpus(&header->version);
199 le32_to_cpus(&header->header_size);
200 le32_to_cpus(&header->image_type);
201 le32_to_cpus(&header->image_flags);
202 le32_to_cpus(&header->offset_bmap);
203 le32_to_cpus(&header->offset_data);
204 le32_to_cpus(&header->cylinders);
205 le32_to_cpus(&header->heads);
206 le32_to_cpus(&header->sectors);
207 le32_to_cpus(&header->sector_size);
208 le64_to_cpus(&header->disk_size);
209 le32_to_cpus(&header->block_size);
210 le32_to_cpus(&header->block_extra);
211 le32_to_cpus(&header->blocks_in_image);
212 le32_to_cpus(&header->blocks_allocated);
213 uuid_convert(header->uuid_image);
214 uuid_convert(header->uuid_last_snap);
215 uuid_convert(header->uuid_link);
216 uuid_convert(header->uuid_parent);
217}
218
219static void vdi_header_to_le(VdiHeader *header)
220{
221 cpu_to_le32s(&header->signature);
222 cpu_to_le32s(&header->version);
223 cpu_to_le32s(&header->header_size);
224 cpu_to_le32s(&header->image_type);
225 cpu_to_le32s(&header->image_flags);
226 cpu_to_le32s(&header->offset_bmap);
227 cpu_to_le32s(&header->offset_data);
228 cpu_to_le32s(&header->cylinders);
229 cpu_to_le32s(&header->heads);
230 cpu_to_le32s(&header->sectors);
231 cpu_to_le32s(&header->sector_size);
232 cpu_to_le64s(&header->disk_size);
233 cpu_to_le32s(&header->block_size);
234 cpu_to_le32s(&header->block_extra);
235 cpu_to_le32s(&header->blocks_in_image);
236 cpu_to_le32s(&header->blocks_allocated);
237 cpu_to_le32s(&header->blocks_allocated);
238 uuid_convert(header->uuid_image);
239 uuid_convert(header->uuid_last_snap);
240 uuid_convert(header->uuid_link);
241 uuid_convert(header->uuid_parent);
242}
243
244#if defined(CONFIG_VDI_DEBUG)
245static void vdi_header_print(VdiHeader *header)
246{
247 char uuid[37];
248 logout("text %s", header->text);
9f0470bb 249 logout("signature 0x%08x\n", header->signature);
9aebd98a
SW
250 logout("header size 0x%04x\n", header->header_size);
251 logout("image type 0x%04x\n", header->image_type);
252 logout("image flags 0x%04x\n", header->image_flags);
253 logout("description %s\n", header->description);
254 logout("offset bmap 0x%04x\n", header->offset_bmap);
255 logout("offset data 0x%04x\n", header->offset_data);
256 logout("cylinders 0x%04x\n", header->cylinders);
257 logout("heads 0x%04x\n", header->heads);
258 logout("sectors 0x%04x\n", header->sectors);
259 logout("sector size 0x%04x\n", header->sector_size);
260 logout("image size 0x%" PRIx64 " B (%" PRIu64 " MiB)\n",
261 header->disk_size, header->disk_size / MiB);
262 logout("block size 0x%04x\n", header->block_size);
263 logout("block extra 0x%04x\n", header->block_extra);
264 logout("blocks tot. 0x%04x\n", header->blocks_in_image);
265 logout("blocks all. 0x%04x\n", header->blocks_allocated);
266 uuid_unparse(header->uuid_image, uuid);
267 logout("uuid image %s\n", uuid);
268 uuid_unparse(header->uuid_last_snap, uuid);
269 logout("uuid snap %s\n", uuid);
270 uuid_unparse(header->uuid_link, uuid);
271 logout("uuid link %s\n", uuid);
272 uuid_unparse(header->uuid_parent, uuid);
273 logout("uuid parent %s\n", uuid);
274}
275#endif
276
4534ff54
KW
277static int vdi_check(BlockDriverState *bs, BdrvCheckResult *res,
278 BdrvCheckMode fix)
9aebd98a
SW
279{
280 /* TODO: additional checks possible. */
281 BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
9aebd98a
SW
282 uint32_t blocks_allocated = 0;
283 uint32_t block;
284 uint32_t *bmap;
285 logout("\n");
286
4534ff54
KW
287 if (fix) {
288 return -ENOTSUP;
289 }
290
7267c094 291 bmap = g_malloc(s->header.blocks_in_image * sizeof(uint32_t));
9aebd98a
SW
292 memset(bmap, 0xff, s->header.blocks_in_image * sizeof(uint32_t));
293
294 /* Check block map and value of blocks_allocated. */
295 for (block = 0; block < s->header.blocks_in_image; block++) {
296 uint32_t bmap_entry = le32_to_cpu(s->bmap[block]);
c794b4e0 297 if (VDI_IS_ALLOCATED(bmap_entry)) {
9aebd98a
SW
298 if (bmap_entry < s->header.blocks_in_image) {
299 blocks_allocated++;
c794b4e0 300 if (!VDI_IS_ALLOCATED(bmap[bmap_entry])) {
9aebd98a
SW
301 bmap[bmap_entry] = bmap_entry;
302 } else {
303 fprintf(stderr, "ERROR: block index %" PRIu32
304 " also used by %" PRIu32 "\n", bmap[bmap_entry], bmap_entry);
9ac228e0 305 res->corruptions++;
9aebd98a
SW
306 }
307 } else {
308 fprintf(stderr, "ERROR: block index %" PRIu32
309 " too large, is %" PRIu32 "\n", block, bmap_entry);
9ac228e0 310 res->corruptions++;
9aebd98a
SW
311 }
312 }
313 }
314 if (blocks_allocated != s->header.blocks_allocated) {
315 fprintf(stderr, "ERROR: allocated blocks mismatch, is %" PRIu32
316 ", should be %" PRIu32 "\n",
317 blocks_allocated, s->header.blocks_allocated);
9ac228e0 318 res->corruptions++;
9aebd98a
SW
319 }
320
7267c094 321 g_free(bmap);
9aebd98a 322
9ac228e0 323 return 0;
9aebd98a
SW
324}
325
326static int vdi_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
327{
328 /* TODO: vdi_get_info would be needed for machine snapshots.
329 vm_state_offset is still missing. */
330 BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
331 logout("\n");
332 bdi->cluster_size = s->block_size;
333 bdi->vm_state_offset = 0;
334 return 0;
335}
336
337static int vdi_make_empty(BlockDriverState *bs)
338{
339 /* TODO: missing code. */
340 logout("\n");
341 /* The return value for missing code must be 0, see block.c. */
342 return 0;
343}
344
345static int vdi_probe(const uint8_t *buf, int buf_size, const char *filename)
346{
347 const VdiHeader *header = (const VdiHeader *)buf;
348 int result = 0;
349
350 logout("\n");
351
352 if (buf_size < sizeof(*header)) {
353 /* Header too small, no VDI. */
354 } else if (le32_to_cpu(header->signature) == VDI_SIGNATURE) {
355 result = 100;
356 }
357
358 if (result == 0) {
359 logout("no vdi image\n");
360 } else {
361 logout("%s", header->text);
362 }
363
364 return result;
365}
366
1a86938f 367static int vdi_open(BlockDriverState *bs, QDict *options, int flags)
9aebd98a
SW
368{
369 BDRVVdiState *s = bs->opaque;
370 VdiHeader header;
371 size_t bmap_size;
8937f822 372 int ret;
9aebd98a
SW
373
374 logout("\n");
375
8937f822
SW
376 ret = bdrv_read(bs->file, 0, (uint8_t *)&header, 1);
377 if (ret < 0) {
9aebd98a
SW
378 goto fail;
379 }
380
381 vdi_header_to_cpu(&header);
382#if defined(CONFIG_VDI_DEBUG)
383 vdi_header_print(&header);
384#endif
385
f21dc3a4
SW
386 if (header.disk_size % SECTOR_SIZE != 0) {
387 /* 'VBoxManage convertfromraw' can create images with odd disk sizes.
388 We accept them but round the disk size to the next multiple of
389 SECTOR_SIZE. */
390 logout("odd disk size %" PRIu64 " B, round up\n", header.disk_size);
391 header.disk_size += SECTOR_SIZE - 1;
392 header.disk_size &= ~(SECTOR_SIZE - 1);
393 }
394
0e87ba2c
SW
395 if (header.signature != VDI_SIGNATURE) {
396 logout("bad vdi signature %08x\n", header.signature);
397 ret = -EMEDIUMTYPE;
398 goto fail;
399 } else if (header.version != VDI_VERSION_1_1) {
9aebd98a
SW
400 logout("unsupported version %u.%u\n",
401 header.version >> 16, header.version & 0xffff);
8937f822 402 ret = -ENOTSUP;
9aebd98a
SW
403 goto fail;
404 } else if (header.offset_bmap % SECTOR_SIZE != 0) {
405 /* We only support block maps which start on a sector boundary. */
406 logout("unsupported block map offset 0x%x B\n", header.offset_bmap);
8937f822 407 ret = -ENOTSUP;
9aebd98a
SW
408 goto fail;
409 } else if (header.offset_data % SECTOR_SIZE != 0) {
410 /* We only support data blocks which start on a sector boundary. */
411 logout("unsupported data offset 0x%x B\n", header.offset_data);
8937f822 412 ret = -ENOTSUP;
9aebd98a
SW
413 goto fail;
414 } else if (header.sector_size != SECTOR_SIZE) {
415 logout("unsupported sector size %u B\n", header.sector_size);
8937f822 416 ret = -ENOTSUP;
9aebd98a
SW
417 goto fail;
418 } else if (header.block_size != 1 * MiB) {
419 logout("unsupported block size %u B\n", header.block_size);
8937f822 420 ret = -ENOTSUP;
9aebd98a 421 goto fail;
f21dc3a4
SW
422 } else if (header.disk_size >
423 (uint64_t)header.blocks_in_image * header.block_size) {
424 logout("unsupported disk size %" PRIu64 " B\n", header.disk_size);
8937f822 425 ret = -ENOTSUP;
9aebd98a
SW
426 goto fail;
427 } else if (!uuid_is_null(header.uuid_link)) {
428 logout("link uuid != 0, unsupported\n");
8937f822 429 ret = -ENOTSUP;
9aebd98a
SW
430 goto fail;
431 } else if (!uuid_is_null(header.uuid_parent)) {
432 logout("parent uuid != 0, unsupported\n");
8937f822 433 ret = -ENOTSUP;
9aebd98a
SW
434 goto fail;
435 }
436
437 bs->total_sectors = header.disk_size / SECTOR_SIZE;
438
439 s->block_size = header.block_size;
440 s->block_sectors = header.block_size / SECTOR_SIZE;
441 s->bmap_sector = header.offset_bmap / SECTOR_SIZE;
442 s->header = header;
443
444 bmap_size = header.blocks_in_image * sizeof(uint32_t);
6eea90eb 445 bmap_size = (bmap_size + SECTOR_SIZE - 1) / SECTOR_SIZE;
6528499f 446 s->bmap = g_malloc(bmap_size * SECTOR_SIZE);
8937f822
SW
447 ret = bdrv_read(bs->file, s->bmap_sector, (uint8_t *)s->bmap, bmap_size);
448 if (ret < 0) {
9aebd98a
SW
449 goto fail_free_bmap;
450 }
451
fc9d106c
KW
452 /* Disable migration when vdi images are used */
453 error_set(&s->migration_blocker,
454 QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
455 "vdi", bs->device_name, "live migration");
456 migrate_add_blocker(s->migration_blocker);
457
9aebd98a
SW
458 return 0;
459
460 fail_free_bmap:
7267c094 461 g_free(s->bmap);
9aebd98a
SW
462
463 fail:
8937f822 464 return ret;
9aebd98a
SW
465}
466
ecfe2bba
JC
467static int vdi_reopen_prepare(BDRVReopenState *state,
468 BlockReopenQueue *queue, Error **errp)
469{
470 return 0;
471}
472
e850b35a
SH
473static int coroutine_fn vdi_co_is_allocated(BlockDriverState *bs,
474 int64_t sector_num, int nb_sectors, int *pnum)
9aebd98a
SW
475{
476 /* TODO: Check for too large sector_num (in bdrv_is_allocated or here). */
477 BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
478 size_t bmap_index = sector_num / s->block_sectors;
479 size_t sector_in_block = sector_num % s->block_sectors;
480 int n_sectors = s->block_sectors - sector_in_block;
481 uint32_t bmap_entry = le32_to_cpu(s->bmap[bmap_index]);
482 logout("%p, %" PRId64 ", %d, %p\n", bs, sector_num, nb_sectors, pnum);
483 if (n_sectors > nb_sectors) {
484 n_sectors = nb_sectors;
485 }
486 *pnum = n_sectors;
c794b4e0 487 return VDI_IS_ALLOCATED(bmap_entry);
9aebd98a
SW
488}
489
a7a43aa1
PB
490static int vdi_co_read(BlockDriverState *bs,
491 int64_t sector_num, uint8_t *buf, int nb_sectors)
9aebd98a 492{
9aebd98a
SW
493 BDRVVdiState *s = bs->opaque;
494 uint32_t bmap_entry;
495 uint32_t block_index;
496 uint32_t sector_in_block;
497 uint32_t n_sectors;
eb9566d1 498 int ret = 0;
4de659e8
PB
499
500 logout("\n");
9aebd98a 501
eb9566d1
PB
502 while (ret >= 0 && nb_sectors > 0) {
503 block_index = sector_num / s->block_sectors;
504 sector_in_block = sector_num % s->block_sectors;
505 n_sectors = s->block_sectors - sector_in_block;
506 if (n_sectors > nb_sectors) {
507 n_sectors = nb_sectors;
508 }
0c7bfc32 509
eb9566d1
PB
510 logout("will read %u sectors starting at sector %" PRIu64 "\n",
511 n_sectors, sector_num);
512
513 /* prepare next AIO request */
514 bmap_entry = le32_to_cpu(s->bmap[block_index]);
515 if (!VDI_IS_ALLOCATED(bmap_entry)) {
516 /* Block not allocated, return zeros, no need to wait. */
517 memset(buf, 0, n_sectors * SECTOR_SIZE);
518 ret = 0;
519 } else {
520 uint64_t offset = s->header.offset_data / SECTOR_SIZE +
521 (uint64_t)bmap_entry * s->block_sectors +
522 sector_in_block;
523 ret = bdrv_read(bs->file, offset, buf, n_sectors);
524 }
525 logout("%u sectors read\n", n_sectors);
0c7bfc32 526
eb9566d1
PB
527 nb_sectors -= n_sectors;
528 sector_num += n_sectors;
529 buf += n_sectors * SECTOR_SIZE;
9aebd98a 530 }
3d46a75a 531
3d46a75a 532 return ret;
9aebd98a
SW
533}
534
a7a43aa1
PB
535static int vdi_co_write(BlockDriverState *bs,
536 int64_t sector_num, const uint8_t *buf, int nb_sectors)
9aebd98a 537{
9aebd98a
SW
538 BDRVVdiState *s = bs->opaque;
539 uint32_t bmap_entry;
540 uint32_t block_index;
541 uint32_t sector_in_block;
542 uint32_t n_sectors;
bfc45fc1
PB
543 uint32_t bmap_first = VDI_UNALLOCATED;
544 uint32_t bmap_last = VDI_UNALLOCATED;
bfc45fc1 545 uint8_t *block = NULL;
eb9566d1 546 int ret = 0;
4de659e8
PB
547
548 logout("\n");
9aebd98a 549
eb9566d1
PB
550 while (ret >= 0 && nb_sectors > 0) {
551 block_index = sector_num / s->block_sectors;
552 sector_in_block = sector_num % s->block_sectors;
553 n_sectors = s->block_sectors - sector_in_block;
554 if (n_sectors > nb_sectors) {
555 n_sectors = nb_sectors;
556 }
9aebd98a 557
eb9566d1
PB
558 logout("will write %u sectors starting at sector %" PRIu64 "\n",
559 n_sectors, sector_num);
560
561 /* prepare next AIO request */
562 bmap_entry = le32_to_cpu(s->bmap[block_index]);
563 if (!VDI_IS_ALLOCATED(bmap_entry)) {
564 /* Allocate new block and write to it. */
565 uint64_t offset;
566 bmap_entry = s->header.blocks_allocated;
567 s->bmap[block_index] = cpu_to_le32(bmap_entry);
568 s->header.blocks_allocated++;
569 offset = s->header.offset_data / SECTOR_SIZE +
570 (uint64_t)bmap_entry * s->block_sectors;
571 if (block == NULL) {
572 block = g_malloc(s->block_size);
573 bmap_first = block_index;
574 }
575 bmap_last = block_index;
576 /* Copy data to be written to new block and zero unused parts. */
577 memset(block, 0, sector_in_block * SECTOR_SIZE);
578 memcpy(block + sector_in_block * SECTOR_SIZE,
579 buf, n_sectors * SECTOR_SIZE);
580 memset(block + (sector_in_block + n_sectors) * SECTOR_SIZE, 0,
581 (s->block_sectors - n_sectors - sector_in_block) * SECTOR_SIZE);
582 ret = bdrv_write(bs->file, offset, block, s->block_sectors);
583 } else {
584 uint64_t offset = s->header.offset_data / SECTOR_SIZE +
585 (uint64_t)bmap_entry * s->block_sectors +
586 sector_in_block;
587 ret = bdrv_write(bs->file, offset, buf, n_sectors);
9aebd98a 588 }
0c7bfc32 589
eb9566d1
PB
590 nb_sectors -= n_sectors;
591 sector_num += n_sectors;
592 buf += n_sectors * SECTOR_SIZE;
0c7bfc32 593
eb9566d1 594 logout("%u sectors written\n", n_sectors);
9aebd98a 595 }
9aebd98a 596
0c7bfc32 597 logout("finished data write\n");
4eea78e6
PB
598 if (ret < 0) {
599 return ret;
600 }
601
602 if (block) {
603 /* One or more new blocks were allocated. */
604 VdiHeader *header = (VdiHeader *) block;
605 uint8_t *base;
606 uint64_t offset;
607
608 logout("now writing modified header\n");
609 assert(VDI_IS_ALLOCATED(bmap_first));
610 *header = s->header;
611 vdi_header_to_le(header);
612 ret = bdrv_write(bs->file, 0, block, 1);
bfc45fc1
PB
613 g_free(block);
614 block = NULL;
4eea78e6
PB
615
616 if (ret < 0) {
617 return ret;
0c7bfc32 618 }
4eea78e6
PB
619
620 logout("now writing modified block map entry %u...%u\n",
621 bmap_first, bmap_last);
622 /* Write modified sectors from block map. */
623 bmap_first /= (SECTOR_SIZE / sizeof(uint32_t));
624 bmap_last /= (SECTOR_SIZE / sizeof(uint32_t));
625 n_sectors = bmap_last - bmap_first + 1;
626 offset = s->bmap_sector + bmap_first;
627 base = ((uint8_t *)&s->bmap[0]) + bmap_first * SECTOR_SIZE;
628 logout("will write %u block map sectors starting from entry %u\n",
629 n_sectors, bmap_first);
630 ret = bdrv_write(bs->file, offset, base, n_sectors);
0c7bfc32
PB
631 }
632
3d46a75a 633 return ret;
9aebd98a
SW
634}
635
9aebd98a
SW
636static int vdi_create(const char *filename, QEMUOptionParameter *options)
637{
638 int fd;
639 int result = 0;
640 uint64_t bytes = 0;
641 uint32_t blocks;
99cce9fa 642 size_t block_size = DEFAULT_CLUSTER_SIZE;
9aebd98a
SW
643 uint32_t image_type = VDI_TYPE_DYNAMIC;
644 VdiHeader header;
645 size_t i;
646 size_t bmap_size;
9aebd98a
SW
647
648 logout("\n");
649
650 /* Read out options. */
651 while (options && options->name) {
652 if (!strcmp(options->name, BLOCK_OPT_SIZE)) {
653 bytes = options->value.n;
654#if defined(CONFIG_VDI_BLOCK_SIZE)
655 } else if (!strcmp(options->name, BLOCK_OPT_CLUSTER_SIZE)) {
656 if (options->value.n) {
657 /* TODO: Additional checks (SECTOR_SIZE * 2^n, ...). */
658 block_size = options->value.n;
659 }
660#endif
661#if defined(CONFIG_VDI_STATIC_IMAGE)
662 } else if (!strcmp(options->name, BLOCK_OPT_STATIC)) {
6eea90eb
SW
663 if (options->value.n) {
664 image_type = VDI_TYPE_STATIC;
665 }
9aebd98a
SW
666#endif
667 }
668 options++;
669 }
670
6165f4d8
CB
671 fd = qemu_open(filename,
672 O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE,
673 0644);
9aebd98a
SW
674 if (fd < 0) {
675 return -errno;
676 }
677
f21dc3a4
SW
678 /* We need enough blocks to store the given disk size,
679 so always round up. */
680 blocks = (bytes + block_size - 1) / block_size;
681
9aebd98a
SW
682 bmap_size = blocks * sizeof(uint32_t);
683 bmap_size = ((bmap_size + SECTOR_SIZE - 1) & ~(SECTOR_SIZE -1));
684
685 memset(&header, 0, sizeof(header));
1786dc15 686 pstrcpy(header.text, sizeof(header.text), VDI_TEXT);
9aebd98a
SW
687 header.signature = VDI_SIGNATURE;
688 header.version = VDI_VERSION_1_1;
689 header.header_size = 0x180;
690 header.image_type = image_type;
691 header.offset_bmap = 0x200;
692 header.offset_data = 0x200 + bmap_size;
693 header.sector_size = SECTOR_SIZE;
694 header.disk_size = bytes;
695 header.block_size = block_size;
696 header.blocks_in_image = blocks;
6eea90eb
SW
697 if (image_type == VDI_TYPE_STATIC) {
698 header.blocks_allocated = blocks;
699 }
9aebd98a
SW
700 uuid_generate(header.uuid_image);
701 uuid_generate(header.uuid_last_snap);
702 /* There is no need to set header.uuid_link or header.uuid_parent here. */
703#if defined(CONFIG_VDI_DEBUG)
704 vdi_header_print(&header);
705#endif
706 vdi_header_to_le(&header);
707 if (write(fd, &header, sizeof(header)) < 0) {
708 result = -errno;
709 }
710
b76b6e95 711 if (bmap_size > 0) {
514f21a5
SW
712 uint32_t *bmap = g_malloc0(bmap_size);
713 for (i = 0; i < blocks; i++) {
714 if (image_type == VDI_TYPE_STATIC) {
715 bmap[i] = i;
716 } else {
717 bmap[i] = VDI_UNALLOCATED;
718 }
9aebd98a 719 }
514f21a5
SW
720 if (write(fd, bmap, bmap_size) < 0) {
721 result = -errno;
722 }
723 g_free(bmap);
9aebd98a 724 }
514f21a5 725
9aebd98a
SW
726 if (image_type == VDI_TYPE_STATIC) {
727 if (ftruncate(fd, sizeof(header) + bmap_size + blocks * block_size)) {
728 result = -errno;
729 }
730 }
731
732 if (close(fd) < 0) {
733 result = -errno;
734 }
735
736 return result;
737}
738
739static void vdi_close(BlockDriverState *bs)
740{
fc9d106c 741 BDRVVdiState *s = bs->opaque;
6ac5f388
KW
742
743 g_free(s->bmap);
744
fc9d106c
KW
745 migrate_del_blocker(s->migration_blocker);
746 error_free(s->migration_blocker);
9aebd98a
SW
747}
748
9aebd98a
SW
749static QEMUOptionParameter vdi_create_options[] = {
750 {
751 .name = BLOCK_OPT_SIZE,
752 .type = OPT_SIZE,
753 .help = "Virtual disk size"
754 },
755#if defined(CONFIG_VDI_BLOCK_SIZE)
756 {
757 .name = BLOCK_OPT_CLUSTER_SIZE,
758 .type = OPT_SIZE,
99cce9fa
KW
759 .help = "VDI cluster (block) size",
760 .value = { .n = DEFAULT_CLUSTER_SIZE },
9aebd98a
SW
761 },
762#endif
763#if defined(CONFIG_VDI_STATIC_IMAGE)
764 {
765 .name = BLOCK_OPT_STATIC,
766 .type = OPT_FLAG,
767 .help = "VDI static (pre-allocated) image"
768 },
769#endif
770 /* TODO: An additional option to set UUID values might be useful. */
771 { NULL }
772};
773
774static BlockDriver bdrv_vdi = {
775 .format_name = "vdi",
776 .instance_size = sizeof(BDRVVdiState),
777 .bdrv_probe = vdi_probe,
778 .bdrv_open = vdi_open,
779 .bdrv_close = vdi_close,
ecfe2bba 780 .bdrv_reopen_prepare = vdi_reopen_prepare,
9aebd98a 781 .bdrv_create = vdi_create,
e850b35a 782 .bdrv_co_is_allocated = vdi_co_is_allocated,
9aebd98a
SW
783 .bdrv_make_empty = vdi_make_empty,
784
a7a43aa1 785 .bdrv_read = vdi_co_read,
9aebd98a 786#if defined(CONFIG_VDI_WRITE)
a7a43aa1 787 .bdrv_write = vdi_co_write,
9aebd98a
SW
788#endif
789
790 .bdrv_get_info = vdi_get_info,
791
792 .create_options = vdi_create_options,
793 .bdrv_check = vdi_check,
794};
795
796static void bdrv_vdi_init(void)
797{
798 logout("\n");
799 bdrv_register(&bdrv_vdi);
800}
801
802block_init(bdrv_vdi_init);