]>
Commit | Line | Data |
---|---|---|
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 | * | |
dc6fb73d | 34 | * Read and write of adjacent blocks could be done in one operation |
9aebd98a SW |
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 | ||
80c71a24 | 52 | #include "qemu/osdep.h" |
9aebd98a | 53 | #include "qemu-common.h" |
737e150e | 54 | #include "block/block_int.h" |
a08f0c3b | 55 | #include "sysemu/block-backend.h" |
1de7afc9 | 56 | #include "qemu/module.h" |
caf71f86 | 57 | #include "migration/migration.h" |
10817bf0 | 58 | #include "qemu/coroutine.h" |
9aebd98a | 59 | |
ee682d27 | 60 | #if defined(CONFIG_UUID) |
9aebd98a SW |
61 | #include <uuid/uuid.h> |
62 | #else | |
63 | /* TODO: move uuid emulation to some central place in QEMU. */ | |
9c17d615 | 64 | #include "sysemu/sysemu.h" /* UUID_FMT */ |
9aebd98a | 65 | typedef unsigned char uuid_t[16]; |
9aebd98a SW |
66 | #endif |
67 | ||
68 | /* Code configuration options. */ | |
69 | ||
70 | /* Enable debug messages. */ | |
71 | //~ #define CONFIG_VDI_DEBUG | |
72 | ||
73 | /* Support write operations on VDI images. */ | |
74 | #define CONFIG_VDI_WRITE | |
75 | ||
76 | /* Support non-standard block (cluster) size. This is untested. | |
77 | * Maybe it will be needed for very large images. | |
78 | */ | |
79 | //~ #define CONFIG_VDI_BLOCK_SIZE | |
80 | ||
81 | /* Support static (fixed, pre-allocated) images. */ | |
82 | #define CONFIG_VDI_STATIC_IMAGE | |
83 | ||
84 | /* Command line option for static images. */ | |
85 | #define BLOCK_OPT_STATIC "static" | |
86 | ||
87 | #define KiB 1024 | |
88 | #define MiB (KiB * KiB) | |
89 | ||
90 | #define SECTOR_SIZE 512 | |
99cce9fa | 91 | #define DEFAULT_CLUSTER_SIZE (1 * MiB) |
9aebd98a SW |
92 | |
93 | #if defined(CONFIG_VDI_DEBUG) | |
94 | #define logout(fmt, ...) \ | |
95 | fprintf(stderr, "vdi\t%-24s" fmt, __func__, ##__VA_ARGS__) | |
96 | #else | |
97 | #define logout(fmt, ...) ((void)0) | |
98 | #endif | |
99 | ||
100 | /* Image signature. */ | |
101 | #define VDI_SIGNATURE 0xbeda107f | |
102 | ||
103 | /* Image version. */ | |
104 | #define VDI_VERSION_1_1 0x00010001 | |
105 | ||
106 | /* Image type. */ | |
107 | #define VDI_TYPE_DYNAMIC 1 | |
108 | #define VDI_TYPE_STATIC 2 | |
109 | ||
110 | /* Innotek / SUN images use these strings in header.text: | |
111 | * "<<< innotek VirtualBox Disk Image >>>\n" | |
112 | * "<<< Sun xVM VirtualBox Disk Image >>>\n" | |
113 | * "<<< Sun VirtualBox Disk Image >>>\n" | |
114 | * The value does not matter, so QEMU created images use a different text. | |
115 | */ | |
116 | #define VDI_TEXT "<<< QEMU VM Virtual Disk Image >>>\n" | |
117 | ||
c794b4e0 ES |
118 | /* A never-allocated block; semantically arbitrary content. */ |
119 | #define VDI_UNALLOCATED 0xffffffffU | |
120 | ||
121 | /* A discarded (no longer allocated) block; semantically zero-filled. */ | |
122 | #define VDI_DISCARDED 0xfffffffeU | |
123 | ||
124 | #define VDI_IS_ALLOCATED(X) ((X) < VDI_DISCARDED) | |
9aebd98a | 125 | |
d20418ee HR |
126 | /* The bmap will take up VDI_BLOCKS_IN_IMAGE_MAX * sizeof(uint32_t) bytes; since |
127 | * the bmap is read and written in a single operation, its size needs to be | |
128 | * limited to INT_MAX; furthermore, when opening an image, the bmap size is | |
129 | * rounded up to be aligned on BDRV_SECTOR_SIZE. | |
130 | * Therefore this should satisfy the following: | |
131 | * VDI_BLOCKS_IN_IMAGE_MAX * sizeof(uint32_t) + BDRV_SECTOR_SIZE == INT_MAX + 1 | |
132 | * (INT_MAX + 1 is the first value not representable as an int) | |
133 | * This guarantees that any value below or equal to the constant will, when | |
134 | * multiplied by sizeof(uint32_t) and rounded up to a BDRV_SECTOR_SIZE boundary, | |
135 | * still be below or equal to INT_MAX. */ | |
136 | #define VDI_BLOCKS_IN_IMAGE_MAX \ | |
137 | ((unsigned)((INT_MAX + 1u - BDRV_SECTOR_SIZE) / sizeof(uint32_t))) | |
63fa06dc JC |
138 | #define VDI_DISK_SIZE_MAX ((uint64_t)VDI_BLOCKS_IN_IMAGE_MAX * \ |
139 | (uint64_t)DEFAULT_CLUSTER_SIZE) | |
140 | ||
ee682d27 | 141 | #if !defined(CONFIG_UUID) |
8ba2aae3 | 142 | static inline void uuid_generate(uuid_t out) |
9aebd98a | 143 | { |
4f3669ea | 144 | memset(out, 0, sizeof(uuid_t)); |
9aebd98a SW |
145 | } |
146 | ||
8ba2aae3 | 147 | static inline int uuid_is_null(const uuid_t uu) |
9aebd98a SW |
148 | { |
149 | uuid_t null_uuid = { 0 }; | |
4f3669ea | 150 | return memcmp(uu, null_uuid, sizeof(uuid_t)) == 0; |
9aebd98a SW |
151 | } |
152 | ||
f18a768e | 153 | # if defined(CONFIG_VDI_DEBUG) |
8ba2aae3 | 154 | static inline void uuid_unparse(const uuid_t uu, char *out) |
9aebd98a SW |
155 | { |
156 | snprintf(out, 37, UUID_FMT, | |
157 | uu[0], uu[1], uu[2], uu[3], uu[4], uu[5], uu[6], uu[7], | |
158 | uu[8], uu[9], uu[10], uu[11], uu[12], uu[13], uu[14], uu[15]); | |
159 | } | |
f18a768e | 160 | # endif |
9aebd98a SW |
161 | #endif |
162 | ||
9aebd98a SW |
163 | typedef struct { |
164 | char text[0x40]; | |
165 | uint32_t signature; | |
166 | uint32_t version; | |
167 | uint32_t header_size; | |
168 | uint32_t image_type; | |
169 | uint32_t image_flags; | |
170 | char description[256]; | |
171 | uint32_t offset_bmap; | |
172 | uint32_t offset_data; | |
173 | uint32_t cylinders; /* disk geometry, unused here */ | |
174 | uint32_t heads; /* disk geometry, unused here */ | |
175 | uint32_t sectors; /* disk geometry, unused here */ | |
176 | uint32_t sector_size; | |
177 | uint32_t unused1; | |
178 | uint64_t disk_size; | |
179 | uint32_t block_size; | |
180 | uint32_t block_extra; /* unused here */ | |
181 | uint32_t blocks_in_image; | |
182 | uint32_t blocks_allocated; | |
183 | uuid_t uuid_image; | |
184 | uuid_t uuid_last_snap; | |
185 | uuid_t uuid_link; | |
186 | uuid_t uuid_parent; | |
187 | uint64_t unused2[7]; | |
8368febd | 188 | } QEMU_PACKED VdiHeader; |
9aebd98a SW |
189 | |
190 | typedef struct { | |
9aebd98a SW |
191 | /* The block map entries are little endian (even in memory). */ |
192 | uint32_t *bmap; | |
193 | /* Size of block (bytes). */ | |
194 | uint32_t block_size; | |
195 | /* Size of block (sectors). */ | |
196 | uint32_t block_sectors; | |
197 | /* First sector of block map. */ | |
198 | uint32_t bmap_sector; | |
4ff9786c | 199 | /* VDI header (converted to host endianness). */ |
9aebd98a | 200 | VdiHeader header; |
fc9d106c | 201 | |
f0ab6f10 HR |
202 | CoMutex write_lock; |
203 | ||
fc9d106c | 204 | Error *migration_blocker; |
9aebd98a SW |
205 | } BDRVVdiState; |
206 | ||
207 | /* Change UUID from little endian (IPRT = VirtualBox format) to big endian | |
208 | * format (network byte order, standard, see RFC 4122) and vice versa. | |
209 | */ | |
210 | static void uuid_convert(uuid_t uuid) | |
211 | { | |
212 | bswap32s((uint32_t *)&uuid[0]); | |
213 | bswap16s((uint16_t *)&uuid[4]); | |
214 | bswap16s((uint16_t *)&uuid[6]); | |
215 | } | |
216 | ||
217 | static void vdi_header_to_cpu(VdiHeader *header) | |
218 | { | |
219 | le32_to_cpus(&header->signature); | |
220 | le32_to_cpus(&header->version); | |
221 | le32_to_cpus(&header->header_size); | |
222 | le32_to_cpus(&header->image_type); | |
223 | le32_to_cpus(&header->image_flags); | |
224 | le32_to_cpus(&header->offset_bmap); | |
225 | le32_to_cpus(&header->offset_data); | |
226 | le32_to_cpus(&header->cylinders); | |
227 | le32_to_cpus(&header->heads); | |
228 | le32_to_cpus(&header->sectors); | |
229 | le32_to_cpus(&header->sector_size); | |
230 | le64_to_cpus(&header->disk_size); | |
231 | le32_to_cpus(&header->block_size); | |
232 | le32_to_cpus(&header->block_extra); | |
233 | le32_to_cpus(&header->blocks_in_image); | |
234 | le32_to_cpus(&header->blocks_allocated); | |
235 | uuid_convert(header->uuid_image); | |
236 | uuid_convert(header->uuid_last_snap); | |
237 | uuid_convert(header->uuid_link); | |
238 | uuid_convert(header->uuid_parent); | |
239 | } | |
240 | ||
241 | static void vdi_header_to_le(VdiHeader *header) | |
242 | { | |
243 | cpu_to_le32s(&header->signature); | |
244 | cpu_to_le32s(&header->version); | |
245 | cpu_to_le32s(&header->header_size); | |
246 | cpu_to_le32s(&header->image_type); | |
247 | cpu_to_le32s(&header->image_flags); | |
248 | cpu_to_le32s(&header->offset_bmap); | |
249 | cpu_to_le32s(&header->offset_data); | |
250 | cpu_to_le32s(&header->cylinders); | |
251 | cpu_to_le32s(&header->heads); | |
252 | cpu_to_le32s(&header->sectors); | |
253 | cpu_to_le32s(&header->sector_size); | |
254 | cpu_to_le64s(&header->disk_size); | |
255 | cpu_to_le32s(&header->block_size); | |
256 | cpu_to_le32s(&header->block_extra); | |
257 | cpu_to_le32s(&header->blocks_in_image); | |
258 | cpu_to_le32s(&header->blocks_allocated); | |
9aebd98a SW |
259 | uuid_convert(header->uuid_image); |
260 | uuid_convert(header->uuid_last_snap); | |
261 | uuid_convert(header->uuid_link); | |
262 | uuid_convert(header->uuid_parent); | |
263 | } | |
264 | ||
265 | #if defined(CONFIG_VDI_DEBUG) | |
266 | static void vdi_header_print(VdiHeader *header) | |
267 | { | |
268 | char uuid[37]; | |
269 | logout("text %s", header->text); | |
9f0470bb | 270 | logout("signature 0x%08x\n", header->signature); |
9aebd98a SW |
271 | logout("header size 0x%04x\n", header->header_size); |
272 | logout("image type 0x%04x\n", header->image_type); | |
273 | logout("image flags 0x%04x\n", header->image_flags); | |
274 | logout("description %s\n", header->description); | |
275 | logout("offset bmap 0x%04x\n", header->offset_bmap); | |
276 | logout("offset data 0x%04x\n", header->offset_data); | |
277 | logout("cylinders 0x%04x\n", header->cylinders); | |
278 | logout("heads 0x%04x\n", header->heads); | |
279 | logout("sectors 0x%04x\n", header->sectors); | |
280 | logout("sector size 0x%04x\n", header->sector_size); | |
281 | logout("image size 0x%" PRIx64 " B (%" PRIu64 " MiB)\n", | |
282 | header->disk_size, header->disk_size / MiB); | |
283 | logout("block size 0x%04x\n", header->block_size); | |
284 | logout("block extra 0x%04x\n", header->block_extra); | |
285 | logout("blocks tot. 0x%04x\n", header->blocks_in_image); | |
286 | logout("blocks all. 0x%04x\n", header->blocks_allocated); | |
287 | uuid_unparse(header->uuid_image, uuid); | |
288 | logout("uuid image %s\n", uuid); | |
289 | uuid_unparse(header->uuid_last_snap, uuid); | |
290 | logout("uuid snap %s\n", uuid); | |
291 | uuid_unparse(header->uuid_link, uuid); | |
292 | logout("uuid link %s\n", uuid); | |
293 | uuid_unparse(header->uuid_parent, uuid); | |
294 | logout("uuid parent %s\n", uuid); | |
295 | } | |
296 | #endif | |
297 | ||
4534ff54 KW |
298 | static int vdi_check(BlockDriverState *bs, BdrvCheckResult *res, |
299 | BdrvCheckMode fix) | |
9aebd98a SW |
300 | { |
301 | /* TODO: additional checks possible. */ | |
302 | BDRVVdiState *s = (BDRVVdiState *)bs->opaque; | |
9aebd98a SW |
303 | uint32_t blocks_allocated = 0; |
304 | uint32_t block; | |
305 | uint32_t *bmap; | |
306 | logout("\n"); | |
307 | ||
4534ff54 KW |
308 | if (fix) { |
309 | return -ENOTSUP; | |
310 | } | |
311 | ||
5839e53b | 312 | bmap = g_try_new(uint32_t, s->header.blocks_in_image); |
17cce735 KW |
313 | if (s->header.blocks_in_image && bmap == NULL) { |
314 | res->check_errors++; | |
315 | return -ENOMEM; | |
316 | } | |
317 | ||
9aebd98a SW |
318 | memset(bmap, 0xff, s->header.blocks_in_image * sizeof(uint32_t)); |
319 | ||
320 | /* Check block map and value of blocks_allocated. */ | |
321 | for (block = 0; block < s->header.blocks_in_image; block++) { | |
322 | uint32_t bmap_entry = le32_to_cpu(s->bmap[block]); | |
c794b4e0 | 323 | if (VDI_IS_ALLOCATED(bmap_entry)) { |
9aebd98a SW |
324 | if (bmap_entry < s->header.blocks_in_image) { |
325 | blocks_allocated++; | |
c794b4e0 | 326 | if (!VDI_IS_ALLOCATED(bmap[bmap_entry])) { |
9aebd98a SW |
327 | bmap[bmap_entry] = bmap_entry; |
328 | } else { | |
329 | fprintf(stderr, "ERROR: block index %" PRIu32 | |
330 | " also used by %" PRIu32 "\n", bmap[bmap_entry], bmap_entry); | |
9ac228e0 | 331 | res->corruptions++; |
9aebd98a SW |
332 | } |
333 | } else { | |
334 | fprintf(stderr, "ERROR: block index %" PRIu32 | |
335 | " too large, is %" PRIu32 "\n", block, bmap_entry); | |
9ac228e0 | 336 | res->corruptions++; |
9aebd98a SW |
337 | } |
338 | } | |
339 | } | |
340 | if (blocks_allocated != s->header.blocks_allocated) { | |
341 | fprintf(stderr, "ERROR: allocated blocks mismatch, is %" PRIu32 | |
342 | ", should be %" PRIu32 "\n", | |
343 | blocks_allocated, s->header.blocks_allocated); | |
9ac228e0 | 344 | res->corruptions++; |
9aebd98a SW |
345 | } |
346 | ||
7267c094 | 347 | g_free(bmap); |
9aebd98a | 348 | |
9ac228e0 | 349 | return 0; |
9aebd98a SW |
350 | } |
351 | ||
352 | static int vdi_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) | |
353 | { | |
354 | /* TODO: vdi_get_info would be needed for machine snapshots. | |
355 | vm_state_offset is still missing. */ | |
356 | BDRVVdiState *s = (BDRVVdiState *)bs->opaque; | |
357 | logout("\n"); | |
358 | bdi->cluster_size = s->block_size; | |
359 | bdi->vm_state_offset = 0; | |
95de6d70 | 360 | bdi->unallocated_blocks_are_zero = true; |
9aebd98a SW |
361 | return 0; |
362 | } | |
363 | ||
364 | static int vdi_make_empty(BlockDriverState *bs) | |
365 | { | |
366 | /* TODO: missing code. */ | |
367 | logout("\n"); | |
368 | /* The return value for missing code must be 0, see block.c. */ | |
369 | return 0; | |
370 | } | |
371 | ||
372 | static int vdi_probe(const uint8_t *buf, int buf_size, const char *filename) | |
373 | { | |
374 | const VdiHeader *header = (const VdiHeader *)buf; | |
dddc7750 | 375 | int ret = 0; |
9aebd98a SW |
376 | |
377 | logout("\n"); | |
378 | ||
379 | if (buf_size < sizeof(*header)) { | |
380 | /* Header too small, no VDI. */ | |
381 | } else if (le32_to_cpu(header->signature) == VDI_SIGNATURE) { | |
dddc7750 | 382 | ret = 100; |
9aebd98a SW |
383 | } |
384 | ||
dddc7750 | 385 | if (ret == 0) { |
9aebd98a SW |
386 | logout("no vdi image\n"); |
387 | } else { | |
388 | logout("%s", header->text); | |
389 | } | |
390 | ||
dddc7750 | 391 | return ret; |
9aebd98a SW |
392 | } |
393 | ||
015a1036 HR |
394 | static int vdi_open(BlockDriverState *bs, QDict *options, int flags, |
395 | Error **errp) | |
9aebd98a SW |
396 | { |
397 | BDRVVdiState *s = bs->opaque; | |
398 | VdiHeader header; | |
399 | size_t bmap_size; | |
8937f822 | 400 | int ret; |
9aebd98a SW |
401 | |
402 | logout("\n"); | |
403 | ||
9a4f4c31 | 404 | ret = bdrv_read(bs->file->bs, 0, (uint8_t *)&header, 1); |
8937f822 | 405 | if (ret < 0) { |
9aebd98a SW |
406 | goto fail; |
407 | } | |
408 | ||
409 | vdi_header_to_cpu(&header); | |
410 | #if defined(CONFIG_VDI_DEBUG) | |
411 | vdi_header_print(&header); | |
412 | #endif | |
413 | ||
63fa06dc JC |
414 | if (header.disk_size > VDI_DISK_SIZE_MAX) { |
415 | error_setg(errp, "Unsupported VDI image size (size is 0x%" PRIx64 | |
416 | ", max supported is 0x%" PRIx64 ")", | |
417 | header.disk_size, VDI_DISK_SIZE_MAX); | |
418 | ret = -ENOTSUP; | |
419 | goto fail; | |
420 | } | |
421 | ||
f21dc3a4 SW |
422 | if (header.disk_size % SECTOR_SIZE != 0) { |
423 | /* 'VBoxManage convertfromraw' can create images with odd disk sizes. | |
424 | We accept them but round the disk size to the next multiple of | |
425 | SECTOR_SIZE. */ | |
426 | logout("odd disk size %" PRIu64 " B, round up\n", header.disk_size); | |
e9082e47 | 427 | header.disk_size = ROUND_UP(header.disk_size, SECTOR_SIZE); |
f21dc3a4 SW |
428 | } |
429 | ||
0e87ba2c | 430 | if (header.signature != VDI_SIGNATURE) { |
521b2b5d HR |
431 | error_setg(errp, "Image not in VDI format (bad signature %08" PRIx32 |
432 | ")", header.signature); | |
76abe407 | 433 | ret = -EINVAL; |
0e87ba2c SW |
434 | goto fail; |
435 | } else if (header.version != VDI_VERSION_1_1) { | |
521b2b5d HR |
436 | error_setg(errp, "unsupported VDI image (version %" PRIu32 ".%" PRIu32 |
437 | ")", header.version >> 16, header.version & 0xffff); | |
8937f822 | 438 | ret = -ENOTSUP; |
9aebd98a SW |
439 | goto fail; |
440 | } else if (header.offset_bmap % SECTOR_SIZE != 0) { | |
441 | /* We only support block maps which start on a sector boundary. */ | |
5b7aa9b5 | 442 | error_setg(errp, "unsupported VDI image (unaligned block map offset " |
521b2b5d | 443 | "0x%" PRIx32 ")", header.offset_bmap); |
8937f822 | 444 | ret = -ENOTSUP; |
9aebd98a SW |
445 | goto fail; |
446 | } else if (header.offset_data % SECTOR_SIZE != 0) { | |
447 | /* We only support data blocks which start on a sector boundary. */ | |
521b2b5d HR |
448 | error_setg(errp, "unsupported VDI image (unaligned data offset 0x%" |
449 | PRIx32 ")", header.offset_data); | |
8937f822 | 450 | ret = -ENOTSUP; |
9aebd98a SW |
451 | goto fail; |
452 | } else if (header.sector_size != SECTOR_SIZE) { | |
521b2b5d HR |
453 | error_setg(errp, "unsupported VDI image (sector size %" PRIu32 |
454 | " is not %u)", header.sector_size, SECTOR_SIZE); | |
8937f822 | 455 | ret = -ENOTSUP; |
9aebd98a | 456 | goto fail; |
63fa06dc | 457 | } else if (header.block_size != DEFAULT_CLUSTER_SIZE) { |
521b2b5d HR |
458 | error_setg(errp, "unsupported VDI image (block size %" PRIu32 |
459 | " is not %u)", header.block_size, DEFAULT_CLUSTER_SIZE); | |
8937f822 | 460 | ret = -ENOTSUP; |
9aebd98a | 461 | goto fail; |
f21dc3a4 SW |
462 | } else if (header.disk_size > |
463 | (uint64_t)header.blocks_in_image * header.block_size) { | |
5b7aa9b5 PB |
464 | error_setg(errp, "unsupported VDI image (disk size %" PRIu64 ", " |
465 | "image bitmap has room for %" PRIu64 ")", | |
466 | header.disk_size, | |
467 | (uint64_t)header.blocks_in_image * header.block_size); | |
8937f822 | 468 | ret = -ENOTSUP; |
9aebd98a SW |
469 | goto fail; |
470 | } else if (!uuid_is_null(header.uuid_link)) { | |
5b7aa9b5 | 471 | error_setg(errp, "unsupported VDI image (non-NULL link UUID)"); |
8937f822 | 472 | ret = -ENOTSUP; |
9aebd98a SW |
473 | goto fail; |
474 | } else if (!uuid_is_null(header.uuid_parent)) { | |
5b7aa9b5 | 475 | error_setg(errp, "unsupported VDI image (non-NULL parent UUID)"); |
8937f822 | 476 | ret = -ENOTSUP; |
9aebd98a | 477 | goto fail; |
63fa06dc JC |
478 | } else if (header.blocks_in_image > VDI_BLOCKS_IN_IMAGE_MAX) { |
479 | error_setg(errp, "unsupported VDI image " | |
480 | "(too many blocks %u, max is %u)", | |
481 | header.blocks_in_image, VDI_BLOCKS_IN_IMAGE_MAX); | |
482 | ret = -ENOTSUP; | |
483 | goto fail; | |
9aebd98a SW |
484 | } |
485 | ||
486 | bs->total_sectors = header.disk_size / SECTOR_SIZE; | |
487 | ||
488 | s->block_size = header.block_size; | |
489 | s->block_sectors = header.block_size / SECTOR_SIZE; | |
490 | s->bmap_sector = header.offset_bmap / SECTOR_SIZE; | |
491 | s->header = header; | |
492 | ||
493 | bmap_size = header.blocks_in_image * sizeof(uint32_t); | |
e9082e47 | 494 | bmap_size = DIV_ROUND_UP(bmap_size, SECTOR_SIZE); |
9a4f4c31 | 495 | s->bmap = qemu_try_blockalign(bs->file->bs, bmap_size * SECTOR_SIZE); |
17cce735 KW |
496 | if (s->bmap == NULL) { |
497 | ret = -ENOMEM; | |
498 | goto fail; | |
499 | } | |
500 | ||
9a4f4c31 KW |
501 | ret = bdrv_read(bs->file->bs, s->bmap_sector, (uint8_t *)s->bmap, |
502 | bmap_size); | |
8937f822 | 503 | if (ret < 0) { |
9aebd98a SW |
504 | goto fail_free_bmap; |
505 | } | |
506 | ||
fc9d106c | 507 | /* Disable migration when vdi images are used */ |
81e5f78a AG |
508 | error_setg(&s->migration_blocker, "The vdi format used by node '%s' " |
509 | "does not support live migration", | |
510 | bdrv_get_device_or_node_name(bs)); | |
fc9d106c KW |
511 | migrate_add_blocker(s->migration_blocker); |
512 | ||
f0ab6f10 HR |
513 | qemu_co_mutex_init(&s->write_lock); |
514 | ||
9aebd98a SW |
515 | return 0; |
516 | ||
517 | fail_free_bmap: | |
17cce735 | 518 | qemu_vfree(s->bmap); |
9aebd98a SW |
519 | |
520 | fail: | |
8937f822 | 521 | return ret; |
9aebd98a SW |
522 | } |
523 | ||
ecfe2bba JC |
524 | static int vdi_reopen_prepare(BDRVReopenState *state, |
525 | BlockReopenQueue *queue, Error **errp) | |
526 | { | |
527 | return 0; | |
528 | } | |
529 | ||
b6b8a333 | 530 | static int64_t coroutine_fn vdi_co_get_block_status(BlockDriverState *bs, |
67a0fd2a | 531 | int64_t sector_num, int nb_sectors, int *pnum, BlockDriverState **file) |
9aebd98a SW |
532 | { |
533 | /* TODO: Check for too large sector_num (in bdrv_is_allocated or here). */ | |
534 | BDRVVdiState *s = (BDRVVdiState *)bs->opaque; | |
535 | size_t bmap_index = sector_num / s->block_sectors; | |
536 | size_t sector_in_block = sector_num % s->block_sectors; | |
537 | int n_sectors = s->block_sectors - sector_in_block; | |
538 | uint32_t bmap_entry = le32_to_cpu(s->bmap[bmap_index]); | |
4bc74be9 PB |
539 | uint64_t offset; |
540 | int result; | |
541 | ||
9aebd98a SW |
542 | logout("%p, %" PRId64 ", %d, %p\n", bs, sector_num, nb_sectors, pnum); |
543 | if (n_sectors > nb_sectors) { | |
544 | n_sectors = nb_sectors; | |
545 | } | |
546 | *pnum = n_sectors; | |
4bc74be9 PB |
547 | result = VDI_IS_ALLOCATED(bmap_entry); |
548 | if (!result) { | |
549 | return 0; | |
550 | } | |
551 | ||
552 | offset = s->header.offset_data + | |
553 | (uint64_t)bmap_entry * s->block_size + | |
554 | sector_in_block * SECTOR_SIZE; | |
8bfb1371 | 555 | *file = bs->file->bs; |
4bc74be9 | 556 | return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID | offset; |
9aebd98a SW |
557 | } |
558 | ||
a7a43aa1 PB |
559 | static int vdi_co_read(BlockDriverState *bs, |
560 | int64_t sector_num, uint8_t *buf, int nb_sectors) | |
9aebd98a | 561 | { |
9aebd98a SW |
562 | BDRVVdiState *s = bs->opaque; |
563 | uint32_t bmap_entry; | |
564 | uint32_t block_index; | |
565 | uint32_t sector_in_block; | |
566 | uint32_t n_sectors; | |
eb9566d1 | 567 | int ret = 0; |
4de659e8 PB |
568 | |
569 | logout("\n"); | |
9aebd98a | 570 | |
eb9566d1 PB |
571 | while (ret >= 0 && nb_sectors > 0) { |
572 | block_index = sector_num / s->block_sectors; | |
573 | sector_in_block = sector_num % s->block_sectors; | |
574 | n_sectors = s->block_sectors - sector_in_block; | |
575 | if (n_sectors > nb_sectors) { | |
576 | n_sectors = nb_sectors; | |
577 | } | |
0c7bfc32 | 578 | |
eb9566d1 PB |
579 | logout("will read %u sectors starting at sector %" PRIu64 "\n", |
580 | n_sectors, sector_num); | |
581 | ||
582 | /* prepare next AIO request */ | |
583 | bmap_entry = le32_to_cpu(s->bmap[block_index]); | |
584 | if (!VDI_IS_ALLOCATED(bmap_entry)) { | |
585 | /* Block not allocated, return zeros, no need to wait. */ | |
586 | memset(buf, 0, n_sectors * SECTOR_SIZE); | |
587 | ret = 0; | |
588 | } else { | |
589 | uint64_t offset = s->header.offset_data / SECTOR_SIZE + | |
590 | (uint64_t)bmap_entry * s->block_sectors + | |
591 | sector_in_block; | |
9a4f4c31 | 592 | ret = bdrv_read(bs->file->bs, offset, buf, n_sectors); |
eb9566d1 PB |
593 | } |
594 | logout("%u sectors read\n", n_sectors); | |
0c7bfc32 | 595 | |
eb9566d1 PB |
596 | nb_sectors -= n_sectors; |
597 | sector_num += n_sectors; | |
598 | buf += n_sectors * SECTOR_SIZE; | |
9aebd98a | 599 | } |
3d46a75a | 600 | |
3d46a75a | 601 | return ret; |
9aebd98a SW |
602 | } |
603 | ||
a7a43aa1 PB |
604 | static int vdi_co_write(BlockDriverState *bs, |
605 | int64_t sector_num, const uint8_t *buf, int nb_sectors) | |
9aebd98a | 606 | { |
9aebd98a SW |
607 | BDRVVdiState *s = bs->opaque; |
608 | uint32_t bmap_entry; | |
609 | uint32_t block_index; | |
610 | uint32_t sector_in_block; | |
611 | uint32_t n_sectors; | |
bfc45fc1 PB |
612 | uint32_t bmap_first = VDI_UNALLOCATED; |
613 | uint32_t bmap_last = VDI_UNALLOCATED; | |
bfc45fc1 | 614 | uint8_t *block = NULL; |
eb9566d1 | 615 | int ret = 0; |
4de659e8 PB |
616 | |
617 | logout("\n"); | |
9aebd98a | 618 | |
eb9566d1 PB |
619 | while (ret >= 0 && nb_sectors > 0) { |
620 | block_index = sector_num / s->block_sectors; | |
621 | sector_in_block = sector_num % s->block_sectors; | |
622 | n_sectors = s->block_sectors - sector_in_block; | |
623 | if (n_sectors > nb_sectors) { | |
624 | n_sectors = nb_sectors; | |
625 | } | |
9aebd98a | 626 | |
eb9566d1 PB |
627 | logout("will write %u sectors starting at sector %" PRIu64 "\n", |
628 | n_sectors, sector_num); | |
629 | ||
630 | /* prepare next AIO request */ | |
631 | bmap_entry = le32_to_cpu(s->bmap[block_index]); | |
632 | if (!VDI_IS_ALLOCATED(bmap_entry)) { | |
633 | /* Allocate new block and write to it. */ | |
634 | uint64_t offset; | |
635 | bmap_entry = s->header.blocks_allocated; | |
636 | s->bmap[block_index] = cpu_to_le32(bmap_entry); | |
637 | s->header.blocks_allocated++; | |
638 | offset = s->header.offset_data / SECTOR_SIZE + | |
639 | (uint64_t)bmap_entry * s->block_sectors; | |
640 | if (block == NULL) { | |
641 | block = g_malloc(s->block_size); | |
642 | bmap_first = block_index; | |
643 | } | |
644 | bmap_last = block_index; | |
645 | /* Copy data to be written to new block and zero unused parts. */ | |
646 | memset(block, 0, sector_in_block * SECTOR_SIZE); | |
647 | memcpy(block + sector_in_block * SECTOR_SIZE, | |
648 | buf, n_sectors * SECTOR_SIZE); | |
649 | memset(block + (sector_in_block + n_sectors) * SECTOR_SIZE, 0, | |
650 | (s->block_sectors - n_sectors - sector_in_block) * SECTOR_SIZE); | |
f0ab6f10 HR |
651 | |
652 | /* Note that this coroutine does not yield anywhere from reading the | |
653 | * bmap entry until here, so in regards to all the coroutines trying | |
654 | * to write to this cluster, the one doing the allocation will | |
655 | * always be the first to try to acquire the lock. | |
656 | * Therefore, it is also the first that will actually be able to | |
657 | * acquire the lock and thus the padded cluster is written before | |
658 | * the other coroutines can write to the affected area. */ | |
659 | qemu_co_mutex_lock(&s->write_lock); | |
9a4f4c31 | 660 | ret = bdrv_write(bs->file->bs, offset, block, s->block_sectors); |
f0ab6f10 | 661 | qemu_co_mutex_unlock(&s->write_lock); |
eb9566d1 PB |
662 | } else { |
663 | uint64_t offset = s->header.offset_data / SECTOR_SIZE + | |
664 | (uint64_t)bmap_entry * s->block_sectors + | |
665 | sector_in_block; | |
f0ab6f10 HR |
666 | qemu_co_mutex_lock(&s->write_lock); |
667 | /* This lock is only used to make sure the following write operation | |
668 | * is executed after the write issued by the coroutine allocating | |
669 | * this cluster, therefore we do not need to keep it locked. | |
670 | * As stated above, the allocating coroutine will always try to lock | |
671 | * the mutex before all the other concurrent accesses to that | |
672 | * cluster, therefore at this point we can be absolutely certain | |
673 | * that that write operation has returned (there may be other writes | |
674 | * in flight, but they do not concern this very operation). */ | |
675 | qemu_co_mutex_unlock(&s->write_lock); | |
9a4f4c31 | 676 | ret = bdrv_write(bs->file->bs, offset, buf, n_sectors); |
9aebd98a | 677 | } |
0c7bfc32 | 678 | |
eb9566d1 PB |
679 | nb_sectors -= n_sectors; |
680 | sector_num += n_sectors; | |
681 | buf += n_sectors * SECTOR_SIZE; | |
0c7bfc32 | 682 | |
eb9566d1 | 683 | logout("%u sectors written\n", n_sectors); |
9aebd98a | 684 | } |
9aebd98a | 685 | |
0c7bfc32 | 686 | logout("finished data write\n"); |
4eea78e6 PB |
687 | if (ret < 0) { |
688 | return ret; | |
689 | } | |
690 | ||
691 | if (block) { | |
692 | /* One or more new blocks were allocated. */ | |
693 | VdiHeader *header = (VdiHeader *) block; | |
694 | uint8_t *base; | |
695 | uint64_t offset; | |
696 | ||
697 | logout("now writing modified header\n"); | |
698 | assert(VDI_IS_ALLOCATED(bmap_first)); | |
699 | *header = s->header; | |
700 | vdi_header_to_le(header); | |
9a4f4c31 | 701 | ret = bdrv_write(bs->file->bs, 0, block, 1); |
bfc45fc1 PB |
702 | g_free(block); |
703 | block = NULL; | |
4eea78e6 PB |
704 | |
705 | if (ret < 0) { | |
706 | return ret; | |
0c7bfc32 | 707 | } |
4eea78e6 PB |
708 | |
709 | logout("now writing modified block map entry %u...%u\n", | |
710 | bmap_first, bmap_last); | |
711 | /* Write modified sectors from block map. */ | |
712 | bmap_first /= (SECTOR_SIZE / sizeof(uint32_t)); | |
713 | bmap_last /= (SECTOR_SIZE / sizeof(uint32_t)); | |
714 | n_sectors = bmap_last - bmap_first + 1; | |
715 | offset = s->bmap_sector + bmap_first; | |
716 | base = ((uint8_t *)&s->bmap[0]) + bmap_first * SECTOR_SIZE; | |
717 | logout("will write %u block map sectors starting from entry %u\n", | |
718 | n_sectors, bmap_first); | |
9a4f4c31 | 719 | ret = bdrv_write(bs->file->bs, offset, base, n_sectors); |
0c7bfc32 PB |
720 | } |
721 | ||
3d46a75a | 722 | return ret; |
9aebd98a SW |
723 | } |
724 | ||
004b7f25 | 725 | static int vdi_create(const char *filename, QemuOpts *opts, Error **errp) |
9aebd98a | 726 | { |
dddc7750 | 727 | int ret = 0; |
9aebd98a SW |
728 | uint64_t bytes = 0; |
729 | uint32_t blocks; | |
99cce9fa | 730 | size_t block_size = DEFAULT_CLUSTER_SIZE; |
9aebd98a SW |
731 | uint32_t image_type = VDI_TYPE_DYNAMIC; |
732 | VdiHeader header; | |
733 | size_t i; | |
734 | size_t bmap_size; | |
70747862 JC |
735 | int64_t offset = 0; |
736 | Error *local_err = NULL; | |
a08f0c3b | 737 | BlockBackend *blk = NULL; |
70747862 | 738 | uint32_t *bmap = NULL; |
9aebd98a SW |
739 | |
740 | logout("\n"); | |
741 | ||
742 | /* Read out options. */ | |
c2eb918e HT |
743 | bytes = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
744 | BDRV_SECTOR_SIZE); | |
9aebd98a | 745 | #if defined(CONFIG_VDI_BLOCK_SIZE) |
004b7f25 CL |
746 | /* TODO: Additional checks (SECTOR_SIZE * 2^n, ...). */ |
747 | block_size = qemu_opt_get_size_del(opts, | |
748 | BLOCK_OPT_CLUSTER_SIZE, | |
749 | DEFAULT_CLUSTER_SIZE); | |
9aebd98a SW |
750 | #endif |
751 | #if defined(CONFIG_VDI_STATIC_IMAGE) | |
004b7f25 CL |
752 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_STATIC, false)) { |
753 | image_type = VDI_TYPE_STATIC; | |
9aebd98a | 754 | } |
004b7f25 | 755 | #endif |
9aebd98a | 756 | |
63fa06dc | 757 | if (bytes > VDI_DISK_SIZE_MAX) { |
dddc7750 | 758 | ret = -ENOTSUP; |
63fa06dc JC |
759 | error_setg(errp, "Unsupported VDI image size (size is 0x%" PRIx64 |
760 | ", max supported is 0x%" PRIx64 ")", | |
761 | bytes, VDI_DISK_SIZE_MAX); | |
762 | goto exit; | |
763 | } | |
764 | ||
dddc7750 JC |
765 | ret = bdrv_create_file(filename, opts, &local_err); |
766 | if (ret < 0) { | |
70747862 | 767 | error_propagate(errp, local_err); |
63fa06dc | 768 | goto exit; |
9aebd98a | 769 | } |
a08f0c3b KW |
770 | |
771 | blk = blk_new_open("image", filename, NULL, NULL, | |
772 | BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_PROTOCOL, | |
773 | &local_err); | |
774 | if (blk == NULL) { | |
70747862 | 775 | error_propagate(errp, local_err); |
a08f0c3b | 776 | ret = -EIO; |
70747862 | 777 | goto exit; |
4ab15590 CL |
778 | } |
779 | ||
a08f0c3b KW |
780 | blk_set_allow_write_beyond_eof(blk, true); |
781 | ||
f21dc3a4 SW |
782 | /* We need enough blocks to store the given disk size, |
783 | so always round up. */ | |
e9082e47 | 784 | blocks = DIV_ROUND_UP(bytes, block_size); |
f21dc3a4 | 785 | |
9aebd98a | 786 | bmap_size = blocks * sizeof(uint32_t); |
e9082e47 | 787 | bmap_size = ROUND_UP(bmap_size, SECTOR_SIZE); |
9aebd98a SW |
788 | |
789 | memset(&header, 0, sizeof(header)); | |
1786dc15 | 790 | pstrcpy(header.text, sizeof(header.text), VDI_TEXT); |
9aebd98a SW |
791 | header.signature = VDI_SIGNATURE; |
792 | header.version = VDI_VERSION_1_1; | |
793 | header.header_size = 0x180; | |
794 | header.image_type = image_type; | |
795 | header.offset_bmap = 0x200; | |
796 | header.offset_data = 0x200 + bmap_size; | |
797 | header.sector_size = SECTOR_SIZE; | |
798 | header.disk_size = bytes; | |
799 | header.block_size = block_size; | |
800 | header.blocks_in_image = blocks; | |
6eea90eb SW |
801 | if (image_type == VDI_TYPE_STATIC) { |
802 | header.blocks_allocated = blocks; | |
803 | } | |
9aebd98a SW |
804 | uuid_generate(header.uuid_image); |
805 | uuid_generate(header.uuid_last_snap); | |
806 | /* There is no need to set header.uuid_link or header.uuid_parent here. */ | |
807 | #if defined(CONFIG_VDI_DEBUG) | |
808 | vdi_header_print(&header); | |
809 | #endif | |
810 | vdi_header_to_le(&header); | |
a08f0c3b | 811 | ret = blk_pwrite(blk, offset, &header, sizeof(header)); |
dddc7750 | 812 | if (ret < 0) { |
70747862 JC |
813 | error_setg(errp, "Error writing header to %s", filename); |
814 | goto exit; | |
9aebd98a | 815 | } |
70747862 | 816 | offset += sizeof(header); |
9aebd98a | 817 | |
b76b6e95 | 818 | if (bmap_size > 0) { |
17cce735 KW |
819 | bmap = g_try_malloc0(bmap_size); |
820 | if (bmap == NULL) { | |
821 | ret = -ENOMEM; | |
822 | error_setg(errp, "Could not allocate bmap"); | |
823 | goto exit; | |
824 | } | |
514f21a5 SW |
825 | for (i = 0; i < blocks; i++) { |
826 | if (image_type == VDI_TYPE_STATIC) { | |
827 | bmap[i] = i; | |
828 | } else { | |
829 | bmap[i] = VDI_UNALLOCATED; | |
830 | } | |
9aebd98a | 831 | } |
a08f0c3b | 832 | ret = blk_pwrite(blk, offset, bmap, bmap_size); |
dddc7750 | 833 | if (ret < 0) { |
70747862 JC |
834 | error_setg(errp, "Error writing bmap to %s", filename); |
835 | goto exit; | |
514f21a5 | 836 | } |
70747862 | 837 | offset += bmap_size; |
9aebd98a | 838 | } |
514f21a5 | 839 | |
9aebd98a | 840 | if (image_type == VDI_TYPE_STATIC) { |
a08f0c3b | 841 | ret = blk_truncate(blk, offset + blocks * block_size); |
dddc7750 | 842 | if (ret < 0) { |
70747862 JC |
843 | error_setg(errp, "Failed to statically allocate %s", filename); |
844 | goto exit; | |
9aebd98a SW |
845 | } |
846 | } | |
847 | ||
63fa06dc | 848 | exit: |
a08f0c3b | 849 | blk_unref(blk); |
70747862 | 850 | g_free(bmap); |
dddc7750 | 851 | return ret; |
9aebd98a SW |
852 | } |
853 | ||
854 | static void vdi_close(BlockDriverState *bs) | |
855 | { | |
fc9d106c | 856 | BDRVVdiState *s = bs->opaque; |
6ac5f388 | 857 | |
17cce735 | 858 | qemu_vfree(s->bmap); |
6ac5f388 | 859 | |
fc9d106c KW |
860 | migrate_del_blocker(s->migration_blocker); |
861 | error_free(s->migration_blocker); | |
9aebd98a SW |
862 | } |
863 | ||
004b7f25 CL |
864 | static QemuOptsList vdi_create_opts = { |
865 | .name = "vdi-create-opts", | |
866 | .head = QTAILQ_HEAD_INITIALIZER(vdi_create_opts.head), | |
867 | .desc = { | |
868 | { | |
869 | .name = BLOCK_OPT_SIZE, | |
870 | .type = QEMU_OPT_SIZE, | |
871 | .help = "Virtual disk size" | |
872 | }, | |
9aebd98a | 873 | #if defined(CONFIG_VDI_BLOCK_SIZE) |
004b7f25 CL |
874 | { |
875 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
876 | .type = QEMU_OPT_SIZE, | |
877 | .help = "VDI cluster (block) size", | |
878 | .def_value_str = stringify(DEFAULT_CLUSTER_SIZE) | |
879 | }, | |
9aebd98a SW |
880 | #endif |
881 | #if defined(CONFIG_VDI_STATIC_IMAGE) | |
004b7f25 CL |
882 | { |
883 | .name = BLOCK_OPT_STATIC, | |
884 | .type = QEMU_OPT_BOOL, | |
885 | .help = "VDI static (pre-allocated) image", | |
886 | .def_value_str = "off" | |
887 | }, | |
9aebd98a | 888 | #endif |
004b7f25 CL |
889 | /* TODO: An additional option to set UUID values might be useful. */ |
890 | { /* end of list */ } | |
891 | } | |
9aebd98a SW |
892 | }; |
893 | ||
894 | static BlockDriver bdrv_vdi = { | |
895 | .format_name = "vdi", | |
896 | .instance_size = sizeof(BDRVVdiState), | |
897 | .bdrv_probe = vdi_probe, | |
898 | .bdrv_open = vdi_open, | |
899 | .bdrv_close = vdi_close, | |
ecfe2bba | 900 | .bdrv_reopen_prepare = vdi_reopen_prepare, |
c282e1fd | 901 | .bdrv_create = vdi_create, |
3ac21627 | 902 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
b6b8a333 | 903 | .bdrv_co_get_block_status = vdi_co_get_block_status, |
9aebd98a SW |
904 | .bdrv_make_empty = vdi_make_empty, |
905 | ||
a7a43aa1 | 906 | .bdrv_read = vdi_co_read, |
9aebd98a | 907 | #if defined(CONFIG_VDI_WRITE) |
a7a43aa1 | 908 | .bdrv_write = vdi_co_write, |
9aebd98a SW |
909 | #endif |
910 | ||
911 | .bdrv_get_info = vdi_get_info, | |
912 | ||
004b7f25 | 913 | .create_opts = &vdi_create_opts, |
9aebd98a SW |
914 | .bdrv_check = vdi_check, |
915 | }; | |
916 | ||
917 | static void bdrv_vdi_init(void) | |
918 | { | |
919 | logout("\n"); | |
920 | bdrv_register(&bdrv_vdi); | |
921 | } | |
922 | ||
923 | block_init(bdrv_vdi_init); |