]>
Commit | Line | Data |
---|---|---|
6a0f9e82 | 1 | /* |
cc2040f8 | 2 | * Block driver for Connectix / Microsoft Virtual PC images |
5fafdf24 | 3 | * |
6a0f9e82 | 4 | * Copyright (c) 2005 Alex Beregszaszi |
15d35bc5 | 5 | * Copyright (c) 2009 Kevin Wolf <kwolf@suse.de> |
5fafdf24 | 6 | * |
6a0f9e82 FB |
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 | */ | |
faf07963 | 25 | #include "qemu-common.h" |
6a0f9e82 | 26 | #include "block_int.h" |
5efa9d5a | 27 | #include "module.h" |
612ff3d8 | 28 | #include "migration.h" |
6a0f9e82 FB |
29 | |
30 | /**************************************************************/ | |
31 | ||
32 | #define HEADER_SIZE 512 | |
33 | ||
34 | //#define CACHE | |
35 | ||
2cfacb62 AL |
36 | enum vhd_type { |
37 | VHD_FIXED = 2, | |
38 | VHD_DYNAMIC = 3, | |
39 | VHD_DIFFERENCING = 4, | |
40 | }; | |
41 | ||
57c7d9e5 AL |
42 | // Seconds since Jan 1, 2000 0:00:00 (UTC) |
43 | #define VHD_TIMESTAMP_BASE 946684800 | |
44 | ||
6a0f9e82 | 45 | // always big-endian |
b9fa33a6 | 46 | struct vhd_footer { |
2cfacb62 AL |
47 | char creator[8]; // "conectix" |
48 | uint32_t features; | |
49 | uint32_t version; | |
50 | ||
51 | // Offset of next header structure, 0xFFFFFFFF if none | |
52 | uint64_t data_offset; | |
53 | ||
54 | // Seconds since Jan 1, 2000 0:00:00 (UTC) | |
55 | uint32_t timestamp; | |
56 | ||
57 | char creator_app[4]; // "vpc " | |
58 | uint16_t major; | |
59 | uint16_t minor; | |
60 | char creator_os[4]; // "Wi2k" | |
61 | ||
62 | uint64_t orig_size; | |
63 | uint64_t size; | |
64 | ||
65 | uint16_t cyls; | |
66 | uint8_t heads; | |
67 | uint8_t secs_per_cyl; | |
68 | ||
69 | uint32_t type; | |
70 | ||
71 | // Checksum of the Hard Disk Footer ("one's complement of the sum of all | |
72 | // the bytes in the footer without the checksum field") | |
73 | uint32_t checksum; | |
74 | ||
75 | // UUID used to identify a parent hard disk (backing file) | |
76 | uint8_t uuid[16]; | |
77 | ||
78 | uint8_t in_saved_state; | |
b9fa33a6 AL |
79 | }; |
80 | ||
81 | struct vhd_dyndisk_header { | |
2cfacb62 AL |
82 | char magic[8]; // "cxsparse" |
83 | ||
84 | // Offset of next header structure, 0xFFFFFFFF if none | |
85 | uint64_t data_offset; | |
86 | ||
87 | // Offset of the Block Allocation Table (BAT) | |
88 | uint64_t table_offset; | |
89 | ||
90 | uint32_t version; | |
91 | uint32_t max_table_entries; // 32bit/entry | |
92 | ||
93 | // 2 MB by default, must be a power of two | |
94 | uint32_t block_size; | |
95 | ||
96 | uint32_t checksum; | |
97 | uint8_t parent_uuid[16]; | |
98 | uint32_t parent_timestamp; | |
99 | uint32_t reserved; | |
100 | ||
101 | // Backing file name (in UTF-16) | |
102 | uint8_t parent_name[512]; | |
103 | ||
104 | struct { | |
105 | uint32_t platform; | |
106 | uint32_t data_space; | |
107 | uint32_t data_length; | |
108 | uint32_t reserved; | |
109 | uint64_t data_offset; | |
110 | } parent_locator[8]; | |
6a0f9e82 FB |
111 | }; |
112 | ||
113 | typedef struct BDRVVPCState { | |
848c66e8 | 114 | CoMutex lock; |
15d35bc5 AL |
115 | uint8_t footer_buf[HEADER_SIZE]; |
116 | uint64_t free_data_block_offset; | |
2cfacb62 | 117 | int max_table_entries; |
6a0f9e82 | 118 | uint32_t *pagetable; |
15d35bc5 AL |
119 | uint64_t bat_offset; |
120 | uint64_t last_bitmap_offset; | |
6a0f9e82 | 121 | |
2cfacb62 | 122 | uint32_t block_size; |
15d35bc5 AL |
123 | uint32_t bitmap_size; |
124 | ||
6a0f9e82 FB |
125 | #ifdef CACHE |
126 | uint8_t *pageentry_u8; | |
127 | uint32_t *pageentry_u32; | |
128 | uint16_t *pageentry_u16; | |
3b46e624 | 129 | |
6a0f9e82 FB |
130 | uint64_t last_bitmap; |
131 | #endif | |
612ff3d8 KW |
132 | |
133 | Error *migration_blocker; | |
6a0f9e82 FB |
134 | } BDRVVPCState; |
135 | ||
57c7d9e5 AL |
136 | static uint32_t vpc_checksum(uint8_t* buf, size_t size) |
137 | { | |
138 | uint32_t res = 0; | |
139 | int i; | |
140 | ||
141 | for (i = 0; i < size; i++) | |
142 | res += buf[i]; | |
143 | ||
144 | return ~res; | |
145 | } | |
146 | ||
147 | ||
6a0f9e82 FB |
148 | static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename) |
149 | { | |
ffe8ab83 | 150 | if (buf_size >= 8 && !strncmp((char *)buf, "conectix", 8)) |
6a0f9e82 | 151 | return 100; |
6a0f9e82 FB |
152 | return 0; |
153 | } | |
154 | ||
66f82cee | 155 | static int vpc_open(BlockDriverState *bs, int flags) |
6a0f9e82 FB |
156 | { |
157 | BDRVVPCState *s = bs->opaque; | |
66f82cee | 158 | int i; |
b9fa33a6 AL |
159 | struct vhd_footer* footer; |
160 | struct vhd_dyndisk_header* dyndisk_header; | |
161 | uint8_t buf[HEADER_SIZE]; | |
57c7d9e5 | 162 | uint32_t checksum; |
efc8243d | 163 | int err = -1; |
24da78db | 164 | int disk_type = VHD_DYNAMIC; |
6a0f9e82 | 165 | |
66f82cee | 166 | if (bdrv_pread(bs->file, 0, s->footer_buf, HEADER_SIZE) != HEADER_SIZE) |
6a0f9e82 FB |
167 | goto fail; |
168 | ||
15d35bc5 | 169 | footer = (struct vhd_footer*) s->footer_buf; |
24da78db CA |
170 | if (strncmp(footer->creator, "conectix", 8)) { |
171 | int64_t offset = bdrv_getlength(bs->file); | |
172 | if (offset < HEADER_SIZE) { | |
173 | goto fail; | |
174 | } | |
175 | /* If a fixed disk, the footer is found only at the end of the file */ | |
176 | if (bdrv_pread(bs->file, offset-HEADER_SIZE, s->footer_buf, HEADER_SIZE) | |
177 | != HEADER_SIZE) { | |
178 | goto fail; | |
179 | } | |
180 | if (strncmp(footer->creator, "conectix", 8)) { | |
181 | goto fail; | |
182 | } | |
183 | disk_type = VHD_FIXED; | |
184 | } | |
6a0f9e82 | 185 | |
57c7d9e5 AL |
186 | checksum = be32_to_cpu(footer->checksum); |
187 | footer->checksum = 0; | |
188 | if (vpc_checksum(s->footer_buf, HEADER_SIZE) != checksum) | |
189 | fprintf(stderr, "block-vpc: The header checksum of '%s' is " | |
66f82cee | 190 | "incorrect.\n", bs->filename); |
57c7d9e5 | 191 | |
1fa79228 AL |
192 | // The visible size of a image in Virtual PC depends on the geometry |
193 | // rather than on the size stored in the footer (the size in the footer | |
194 | // is too large usually) | |
195 | bs->total_sectors = (int64_t) | |
196 | be16_to_cpu(footer->cyls) * footer->heads * footer->secs_per_cyl; | |
197 | ||
efc8243d SH |
198 | if (bs->total_sectors >= 65535 * 16 * 255) { |
199 | err = -EFBIG; | |
200 | goto fail; | |
201 | } | |
202 | ||
24da78db CA |
203 | if (disk_type == VHD_DYNAMIC) { |
204 | if (bdrv_pread(bs->file, be64_to_cpu(footer->data_offset), buf, | |
205 | HEADER_SIZE) != HEADER_SIZE) { | |
206 | goto fail; | |
207 | } | |
b9fa33a6 | 208 | |
24da78db | 209 | dyndisk_header = (struct vhd_dyndisk_header *) buf; |
6a0f9e82 | 210 | |
24da78db CA |
211 | if (strncmp(dyndisk_header->magic, "cxsparse", 8)) { |
212 | goto fail; | |
213 | } | |
6a0f9e82 | 214 | |
24da78db CA |
215 | s->block_size = be32_to_cpu(dyndisk_header->block_size); |
216 | s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511; | |
15d35bc5 | 217 | |
24da78db CA |
218 | s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries); |
219 | s->pagetable = g_malloc(s->max_table_entries * 4); | |
b71d1c2e | 220 | |
24da78db CA |
221 | s->bat_offset = be64_to_cpu(dyndisk_header->table_offset); |
222 | if (bdrv_pread(bs->file, s->bat_offset, s->pagetable, | |
223 | s->max_table_entries * 4) != s->max_table_entries * 4) { | |
224 | goto fail; | |
225 | } | |
b71d1c2e | 226 | |
24da78db CA |
227 | s->free_data_block_offset = |
228 | (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511; | |
15d35bc5 | 229 | |
24da78db CA |
230 | for (i = 0; i < s->max_table_entries; i++) { |
231 | be32_to_cpus(&s->pagetable[i]); | |
232 | if (s->pagetable[i] != 0xFFFFFFFF) { | |
233 | int64_t next = (512 * (int64_t) s->pagetable[i]) + | |
234 | s->bitmap_size + s->block_size; | |
15d35bc5 | 235 | |
24da78db CA |
236 | if (next > s->free_data_block_offset) { |
237 | s->free_data_block_offset = next; | |
238 | } | |
239 | } | |
15d35bc5 | 240 | } |
15d35bc5 | 241 | |
24da78db | 242 | s->last_bitmap_offset = (int64_t) -1; |
6a0f9e82 | 243 | |
6a0f9e82 | 244 | #ifdef CACHE |
24da78db CA |
245 | s->pageentry_u8 = g_malloc(512); |
246 | s->pageentry_u32 = s->pageentry_u8; | |
247 | s->pageentry_u16 = s->pageentry_u8; | |
248 | s->last_pagetable = -1; | |
6a0f9e82 | 249 | #endif |
24da78db | 250 | } |
6a0f9e82 | 251 | |
848c66e8 | 252 | qemu_co_mutex_init(&s->lock); |
612ff3d8 KW |
253 | |
254 | /* Disable migration when VHD images are used */ | |
255 | error_set(&s->migration_blocker, | |
256 | QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
257 | "vpc", bs->device_name, "live migration"); | |
258 | migrate_add_blocker(s->migration_blocker); | |
259 | ||
6a0f9e82 FB |
260 | return 0; |
261 | fail: | |
efc8243d | 262 | return err; |
6a0f9e82 FB |
263 | } |
264 | ||
b71d1c2e AL |
265 | /* |
266 | * Returns the absolute byte offset of the given sector in the image file. | |
267 | * If the sector is not allocated, -1 is returned instead. | |
15d35bc5 AL |
268 | * |
269 | * The parameter write must be 1 if the offset will be used for a write | |
270 | * operation (the block bitmaps is updated then), 0 otherwise. | |
b71d1c2e | 271 | */ |
15d35bc5 AL |
272 | static inline int64_t get_sector_offset(BlockDriverState *bs, |
273 | int64_t sector_num, int write) | |
6a0f9e82 FB |
274 | { |
275 | BDRVVPCState *s = bs->opaque; | |
276 | uint64_t offset = sector_num * 512; | |
277 | uint64_t bitmap_offset, block_offset; | |
278 | uint32_t pagetable_index, pageentry_index; | |
279 | ||
2cfacb62 AL |
280 | pagetable_index = offset / s->block_size; |
281 | pageentry_index = (offset % s->block_size) / 512; | |
3b46e624 | 282 | |
15d35bc5 AL |
283 | if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff) |
284 | return -1; // not allocated | |
6a0f9e82 | 285 | |
378e2aea | 286 | bitmap_offset = 512 * (uint64_t) s->pagetable[pagetable_index]; |
15d35bc5 AL |
287 | block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index); |
288 | ||
289 | // We must ensure that we don't write to any sectors which are marked as | |
290 | // unused in the bitmap. We get away with setting all bits in the block | |
291 | // bitmap each time we write to a new block. This might cause Virtual PC to | |
292 | // miss sparse read optimization, but it's not a problem in terms of | |
293 | // correctness. | |
294 | if (write && (s->last_bitmap_offset != bitmap_offset)) { | |
295 | uint8_t bitmap[s->bitmap_size]; | |
296 | ||
297 | s->last_bitmap_offset = bitmap_offset; | |
298 | memset(bitmap, 0xff, s->bitmap_size); | |
078a458e | 299 | bdrv_pwrite_sync(bs->file, bitmap_offset, bitmap, s->bitmap_size); |
15d35bc5 | 300 | } |
3b46e624 | 301 | |
26a76461 | 302 | // printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n", |
6a0f9e82 FB |
303 | // sector_num, pagetable_index, pageentry_index, |
304 | // bitmap_offset, block_offset); | |
305 | ||
306 | // disabled by reason | |
307 | #if 0 | |
308 | #ifdef CACHE | |
309 | if (bitmap_offset != s->last_bitmap) | |
310 | { | |
311 | lseek(s->fd, bitmap_offset, SEEK_SET); | |
312 | ||
313 | s->last_bitmap = bitmap_offset; | |
5fafdf24 | 314 | |
6a0f9e82 FB |
315 | // Scary! Bitmap is stored as big endian 32bit entries, |
316 | // while we used to look it up byte by byte | |
317 | read(s->fd, s->pageentry_u8, 512); | |
318 | for (i = 0; i < 128; i++) | |
319 | be32_to_cpus(&s->pageentry_u32[i]); | |
320 | } | |
321 | ||
322 | if ((s->pageentry_u8[pageentry_index / 8] >> (pageentry_index % 8)) & 1) | |
323 | return -1; | |
324 | #else | |
325 | lseek(s->fd, bitmap_offset + (pageentry_index / 8), SEEK_SET); | |
5fafdf24 | 326 | |
6a0f9e82 FB |
327 | read(s->fd, &bitmap_entry, 1); |
328 | ||
329 | if ((bitmap_entry >> (pageentry_index % 8)) & 1) | |
330 | return -1; // not allocated | |
331 | #endif | |
332 | #endif | |
6a0f9e82 | 333 | |
b71d1c2e | 334 | return block_offset; |
6a0f9e82 FB |
335 | } |
336 | ||
15d35bc5 AL |
337 | /* |
338 | * Writes the footer to the end of the image file. This is needed when the | |
339 | * file grows as it overwrites the old footer | |
340 | * | |
341 | * Returns 0 on success and < 0 on error | |
342 | */ | |
343 | static int rewrite_footer(BlockDriverState* bs) | |
344 | { | |
345 | int ret; | |
346 | BDRVVPCState *s = bs->opaque; | |
347 | int64_t offset = s->free_data_block_offset; | |
348 | ||
078a458e | 349 | ret = bdrv_pwrite_sync(bs->file, offset, s->footer_buf, HEADER_SIZE); |
15d35bc5 AL |
350 | if (ret < 0) |
351 | return ret; | |
352 | ||
353 | return 0; | |
354 | } | |
355 | ||
356 | /* | |
357 | * Allocates a new block. This involves writing a new footer and updating | |
358 | * the Block Allocation Table to use the space at the old end of the image | |
359 | * file (overwriting the old footer) | |
360 | * | |
361 | * Returns the sectors' offset in the image file on success and < 0 on error | |
362 | */ | |
363 | static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num) | |
364 | { | |
365 | BDRVVPCState *s = bs->opaque; | |
366 | int64_t bat_offset; | |
367 | uint32_t index, bat_value; | |
368 | int ret; | |
369 | uint8_t bitmap[s->bitmap_size]; | |
370 | ||
371 | // Check if sector_num is valid | |
372 | if ((sector_num < 0) || (sector_num > bs->total_sectors)) | |
373 | return -1; | |
374 | ||
375 | // Write entry into in-memory BAT | |
376 | index = (sector_num * 512) / s->block_size; | |
377 | if (s->pagetable[index] != 0xFFFFFFFF) | |
378 | return -1; | |
379 | ||
380 | s->pagetable[index] = s->free_data_block_offset / 512; | |
381 | ||
382 | // Initialize the block's bitmap | |
383 | memset(bitmap, 0xff, s->bitmap_size); | |
5bb1cbac | 384 | ret = bdrv_pwrite_sync(bs->file, s->free_data_block_offset, bitmap, |
078a458e | 385 | s->bitmap_size); |
5bb1cbac KW |
386 | if (ret < 0) { |
387 | return ret; | |
388 | } | |
15d35bc5 AL |
389 | |
390 | // Write new footer (the old one will be overwritten) | |
391 | s->free_data_block_offset += s->block_size + s->bitmap_size; | |
392 | ret = rewrite_footer(bs); | |
393 | if (ret < 0) | |
394 | goto fail; | |
395 | ||
396 | // Write BAT entry to disk | |
397 | bat_offset = s->bat_offset + (4 * index); | |
398 | bat_value = be32_to_cpu(s->pagetable[index]); | |
078a458e | 399 | ret = bdrv_pwrite_sync(bs->file, bat_offset, &bat_value, 4); |
15d35bc5 AL |
400 | if (ret < 0) |
401 | goto fail; | |
402 | ||
403 | return get_sector_offset(bs, sector_num, 0); | |
404 | ||
405 | fail: | |
406 | s->free_data_block_offset -= (s->block_size + s->bitmap_size); | |
407 | return -1; | |
408 | } | |
409 | ||
5fafdf24 | 410 | static int vpc_read(BlockDriverState *bs, int64_t sector_num, |
6a0f9e82 FB |
411 | uint8_t *buf, int nb_sectors) |
412 | { | |
6c6ea921 | 413 | BDRVVPCState *s = bs->opaque; |
6a0f9e82 | 414 | int ret; |
b71d1c2e | 415 | int64_t offset; |
6c6ea921 | 416 | int64_t sectors, sectors_per_block; |
24da78db | 417 | struct vhd_footer *footer = (struct vhd_footer *) s->footer_buf; |
6a0f9e82 | 418 | |
24da78db CA |
419 | if (cpu_to_be32(footer->type) == VHD_FIXED) { |
420 | return bdrv_read(bs->file, sector_num, buf, nb_sectors); | |
421 | } | |
6a0f9e82 | 422 | while (nb_sectors > 0) { |
15d35bc5 | 423 | offset = get_sector_offset(bs, sector_num, 0); |
b71d1c2e | 424 | |
6c6ea921 KW |
425 | sectors_per_block = s->block_size >> BDRV_SECTOR_BITS; |
426 | sectors = sectors_per_block - (sector_num % sectors_per_block); | |
427 | if (sectors > nb_sectors) { | |
428 | sectors = nb_sectors; | |
429 | } | |
430 | ||
b71d1c2e | 431 | if (offset == -1) { |
6c6ea921 | 432 | memset(buf, 0, sectors * BDRV_SECTOR_SIZE); |
b71d1c2e | 433 | } else { |
6c6ea921 KW |
434 | ret = bdrv_pread(bs->file, offset, buf, |
435 | sectors * BDRV_SECTOR_SIZE); | |
436 | if (ret != sectors * BDRV_SECTOR_SIZE) { | |
b71d1c2e | 437 | return -1; |
6c6ea921 | 438 | } |
b71d1c2e AL |
439 | } |
440 | ||
6c6ea921 KW |
441 | nb_sectors -= sectors; |
442 | sector_num += sectors; | |
443 | buf += sectors * BDRV_SECTOR_SIZE; | |
6a0f9e82 FB |
444 | } |
445 | return 0; | |
446 | } | |
447 | ||
2914caa0 PB |
448 | static coroutine_fn int vpc_co_read(BlockDriverState *bs, int64_t sector_num, |
449 | uint8_t *buf, int nb_sectors) | |
450 | { | |
451 | int ret; | |
452 | BDRVVPCState *s = bs->opaque; | |
453 | qemu_co_mutex_lock(&s->lock); | |
454 | ret = vpc_read(bs, sector_num, buf, nb_sectors); | |
455 | qemu_co_mutex_unlock(&s->lock); | |
456 | return ret; | |
457 | } | |
458 | ||
15d35bc5 AL |
459 | static int vpc_write(BlockDriverState *bs, int64_t sector_num, |
460 | const uint8_t *buf, int nb_sectors) | |
461 | { | |
6c6ea921 | 462 | BDRVVPCState *s = bs->opaque; |
15d35bc5 | 463 | int64_t offset; |
6c6ea921 | 464 | int64_t sectors, sectors_per_block; |
15d35bc5 | 465 | int ret; |
24da78db | 466 | struct vhd_footer *footer = (struct vhd_footer *) s->footer_buf; |
15d35bc5 | 467 | |
24da78db CA |
468 | if (cpu_to_be32(footer->type) == VHD_FIXED) { |
469 | return bdrv_write(bs->file, sector_num, buf, nb_sectors); | |
470 | } | |
15d35bc5 AL |
471 | while (nb_sectors > 0) { |
472 | offset = get_sector_offset(bs, sector_num, 1); | |
473 | ||
6c6ea921 KW |
474 | sectors_per_block = s->block_size >> BDRV_SECTOR_BITS; |
475 | sectors = sectors_per_block - (sector_num % sectors_per_block); | |
476 | if (sectors > nb_sectors) { | |
477 | sectors = nb_sectors; | |
478 | } | |
479 | ||
15d35bc5 AL |
480 | if (offset == -1) { |
481 | offset = alloc_block(bs, sector_num); | |
482 | if (offset < 0) | |
483 | return -1; | |
484 | } | |
485 | ||
6c6ea921 KW |
486 | ret = bdrv_pwrite(bs->file, offset, buf, sectors * BDRV_SECTOR_SIZE); |
487 | if (ret != sectors * BDRV_SECTOR_SIZE) { | |
15d35bc5 | 488 | return -1; |
6c6ea921 | 489 | } |
15d35bc5 | 490 | |
6c6ea921 KW |
491 | nb_sectors -= sectors; |
492 | sector_num += sectors; | |
493 | buf += sectors * BDRV_SECTOR_SIZE; | |
15d35bc5 AL |
494 | } |
495 | ||
496 | return 0; | |
497 | } | |
498 | ||
e183ef75 PB |
499 | static coroutine_fn int vpc_co_write(BlockDriverState *bs, int64_t sector_num, |
500 | const uint8_t *buf, int nb_sectors) | |
501 | { | |
502 | int ret; | |
503 | BDRVVPCState *s = bs->opaque; | |
504 | qemu_co_mutex_lock(&s->lock); | |
505 | ret = vpc_write(bs, sector_num, buf, nb_sectors); | |
506 | qemu_co_mutex_unlock(&s->lock); | |
507 | return ret; | |
508 | } | |
509 | ||
57c7d9e5 AL |
510 | /* |
511 | * Calculates the number of cylinders, heads and sectors per cylinder | |
512 | * based on a given number of sectors. This is the algorithm described | |
513 | * in the VHD specification. | |
514 | * | |
515 | * Note that the geometry doesn't always exactly match total_sectors but | |
516 | * may round it down. | |
6e9ea0c0 AJ |
517 | * |
518 | * Returns 0 on success, -EFBIG if the size is larger than 127 GB | |
57c7d9e5 | 519 | */ |
6e9ea0c0 | 520 | static int calculate_geometry(int64_t total_sectors, uint16_t* cyls, |
57c7d9e5 AL |
521 | uint8_t* heads, uint8_t* secs_per_cyl) |
522 | { | |
523 | uint32_t cyls_times_heads; | |
524 | ||
525 | if (total_sectors > 65535 * 16 * 255) | |
6e9ea0c0 | 526 | return -EFBIG; |
57c7d9e5 AL |
527 | |
528 | if (total_sectors > 65535 * 16 * 63) { | |
529 | *secs_per_cyl = 255; | |
530 | *heads = 16; | |
531 | cyls_times_heads = total_sectors / *secs_per_cyl; | |
532 | } else { | |
533 | *secs_per_cyl = 17; | |
534 | cyls_times_heads = total_sectors / *secs_per_cyl; | |
535 | *heads = (cyls_times_heads + 1023) / 1024; | |
536 | ||
537 | if (*heads < 4) | |
538 | *heads = 4; | |
539 | ||
540 | if (cyls_times_heads >= (*heads * 1024) || *heads > 16) { | |
541 | *secs_per_cyl = 31; | |
542 | *heads = 16; | |
543 | cyls_times_heads = total_sectors / *secs_per_cyl; | |
544 | } | |
545 | ||
546 | if (cyls_times_heads >= (*heads * 1024)) { | |
547 | *secs_per_cyl = 63; | |
548 | *heads = 16; | |
549 | cyls_times_heads = total_sectors / *secs_per_cyl; | |
550 | } | |
551 | } | |
552 | ||
dede4188 | 553 | *cyls = cyls_times_heads / *heads; |
6e9ea0c0 AJ |
554 | |
555 | return 0; | |
57c7d9e5 AL |
556 | } |
557 | ||
24da78db | 558 | static int create_dynamic_disk(int fd, uint8_t *buf, int64_t total_sectors) |
57c7d9e5 | 559 | { |
57c7d9e5 AL |
560 | struct vhd_dyndisk_header* dyndisk_header = |
561 | (struct vhd_dyndisk_header*) buf; | |
57c7d9e5 | 562 | size_t block_size, num_bat_entries; |
24da78db | 563 | int i; |
f0ff243a | 564 | int ret = -EIO; |
57c7d9e5 | 565 | |
57c7d9e5 AL |
566 | // Write the footer (twice: at the beginning and at the end) |
567 | block_size = 0x200000; | |
568 | num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512); | |
569 | ||
f0ff243a BS |
570 | if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) { |
571 | goto fail; | |
572 | } | |
57c7d9e5 | 573 | |
f0ff243a BS |
574 | if (lseek(fd, 1536 + ((num_bat_entries * 4 + 511) & ~511), SEEK_SET) < 0) { |
575 | goto fail; | |
576 | } | |
577 | if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) { | |
578 | goto fail; | |
579 | } | |
57c7d9e5 AL |
580 | |
581 | // Write the initial BAT | |
f0ff243a BS |
582 | if (lseek(fd, 3 * 512, SEEK_SET) < 0) { |
583 | goto fail; | |
584 | } | |
57c7d9e5 AL |
585 | |
586 | memset(buf, 0xFF, 512); | |
f0ff243a BS |
587 | for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++) { |
588 | if (write(fd, buf, 512) != 512) { | |
589 | goto fail; | |
590 | } | |
591 | } | |
57c7d9e5 | 592 | |
57c7d9e5 AL |
593 | // Prepare the Dynamic Disk Header |
594 | memset(buf, 0, 1024); | |
595 | ||
5ec4d682 | 596 | memcpy(dyndisk_header->magic, "cxsparse", 8); |
57c7d9e5 | 597 | |
78439f6a CA |
598 | /* |
599 | * Note: The spec is actually wrong here for data_offset, it says | |
600 | * 0xFFFFFFFF, but MS tools expect all 64 bits to be set. | |
601 | */ | |
602 | dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL); | |
57c7d9e5 AL |
603 | dyndisk_header->table_offset = be64_to_cpu(3 * 512); |
604 | dyndisk_header->version = be32_to_cpu(0x00010000); | |
605 | dyndisk_header->block_size = be32_to_cpu(block_size); | |
606 | dyndisk_header->max_table_entries = be32_to_cpu(num_bat_entries); | |
607 | ||
608 | dyndisk_header->checksum = be32_to_cpu(vpc_checksum(buf, 1024)); | |
609 | ||
610 | // Write the header | |
f0ff243a BS |
611 | if (lseek(fd, 512, SEEK_SET) < 0) { |
612 | goto fail; | |
613 | } | |
57c7d9e5 | 614 | |
f0ff243a BS |
615 | if (write(fd, buf, 1024) != 1024) { |
616 | goto fail; | |
617 | } | |
618 | ret = 0; | |
619 | ||
24da78db CA |
620 | fail: |
621 | return ret; | |
622 | } | |
623 | ||
624 | static int create_fixed_disk(int fd, uint8_t *buf, int64_t total_size) | |
625 | { | |
626 | int ret = -EIO; | |
627 | ||
628 | /* Add footer to total size */ | |
629 | total_size += 512; | |
630 | if (ftruncate(fd, total_size) != 0) { | |
631 | ret = -errno; | |
632 | goto fail; | |
633 | } | |
634 | if (lseek(fd, -512, SEEK_END) < 0) { | |
635 | goto fail; | |
636 | } | |
637 | if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) { | |
638 | goto fail; | |
639 | } | |
640 | ||
641 | ret = 0; | |
642 | ||
643 | fail: | |
644 | return ret; | |
645 | } | |
646 | ||
647 | static int vpc_create(const char *filename, QEMUOptionParameter *options) | |
648 | { | |
649 | uint8_t buf[1024]; | |
650 | struct vhd_footer *footer = (struct vhd_footer *) buf; | |
651 | QEMUOptionParameter *disk_type_param; | |
652 | int fd, i; | |
653 | uint16_t cyls = 0; | |
654 | uint8_t heads = 0; | |
655 | uint8_t secs_per_cyl = 0; | |
656 | int64_t total_sectors; | |
657 | int64_t total_size; | |
658 | int disk_type; | |
659 | int ret = -EIO; | |
660 | ||
661 | /* Read out options */ | |
662 | total_size = get_option_parameter(options, BLOCK_OPT_SIZE)->value.n; | |
663 | ||
664 | disk_type_param = get_option_parameter(options, BLOCK_OPT_SUBFMT); | |
665 | if (disk_type_param && disk_type_param->value.s) { | |
666 | if (!strcmp(disk_type_param->value.s, "dynamic")) { | |
667 | disk_type = VHD_DYNAMIC; | |
668 | } else if (!strcmp(disk_type_param->value.s, "fixed")) { | |
669 | disk_type = VHD_FIXED; | |
670 | } else { | |
671 | return -EINVAL; | |
672 | } | |
673 | } else { | |
674 | disk_type = VHD_DYNAMIC; | |
675 | } | |
676 | ||
677 | /* Create the file */ | |
678 | fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644); | |
679 | if (fd < 0) { | |
680 | return -EIO; | |
681 | } | |
682 | ||
ecd880d9 KW |
683 | /* |
684 | * Calculate matching total_size and geometry. Increase the number of | |
685 | * sectors requested until we get enough (or fail). This ensures that | |
686 | * qemu-img convert doesn't truncate images, but rather rounds up. | |
687 | */ | |
24da78db | 688 | total_sectors = total_size / BDRV_SECTOR_SIZE; |
ecd880d9 KW |
689 | for (i = 0; total_sectors > (int64_t)cyls * heads * secs_per_cyl; i++) { |
690 | if (calculate_geometry(total_sectors + i, &cyls, &heads, | |
691 | &secs_per_cyl)) | |
692 | { | |
24da78db CA |
693 | ret = -EFBIG; |
694 | goto fail; | |
695 | } | |
696 | } | |
ecd880d9 | 697 | |
24da78db CA |
698 | total_sectors = (int64_t) cyls * heads * secs_per_cyl; |
699 | ||
700 | /* Prepare the Hard Disk Footer */ | |
701 | memset(buf, 0, 1024); | |
702 | ||
703 | memcpy(footer->creator, "conectix", 8); | |
704 | /* TODO Check if "qemu" creator_app is ok for VPC */ | |
705 | memcpy(footer->creator_app, "qemu", 4); | |
706 | memcpy(footer->creator_os, "Wi2k", 4); | |
707 | ||
708 | footer->features = be32_to_cpu(0x02); | |
709 | footer->version = be32_to_cpu(0x00010000); | |
710 | if (disk_type == VHD_DYNAMIC) { | |
711 | footer->data_offset = be64_to_cpu(HEADER_SIZE); | |
712 | } else { | |
713 | footer->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL); | |
714 | } | |
715 | footer->timestamp = be32_to_cpu(time(NULL) - VHD_TIMESTAMP_BASE); | |
716 | ||
717 | /* Version of Virtual PC 2007 */ | |
718 | footer->major = be16_to_cpu(0x0005); | |
719 | footer->minor = be16_to_cpu(0x0003); | |
720 | if (disk_type == VHD_DYNAMIC) { | |
721 | footer->orig_size = be64_to_cpu(total_sectors * 512); | |
722 | footer->size = be64_to_cpu(total_sectors * 512); | |
723 | } else { | |
724 | footer->orig_size = be64_to_cpu(total_size); | |
725 | footer->size = be64_to_cpu(total_size); | |
726 | } | |
727 | footer->cyls = be16_to_cpu(cyls); | |
728 | footer->heads = heads; | |
729 | footer->secs_per_cyl = secs_per_cyl; | |
730 | ||
731 | footer->type = be32_to_cpu(disk_type); | |
732 | ||
733 | /* TODO uuid is missing */ | |
734 | ||
735 | footer->checksum = be32_to_cpu(vpc_checksum(buf, HEADER_SIZE)); | |
736 | ||
737 | if (disk_type == VHD_DYNAMIC) { | |
738 | ret = create_dynamic_disk(fd, buf, total_sectors); | |
739 | } else { | |
740 | ret = create_fixed_disk(fd, buf, total_size); | |
741 | } | |
742 | ||
f0ff243a | 743 | fail: |
57c7d9e5 | 744 | close(fd); |
f0ff243a | 745 | return ret; |
57c7d9e5 AL |
746 | } |
747 | ||
6a0f9e82 FB |
748 | static void vpc_close(BlockDriverState *bs) |
749 | { | |
750 | BDRVVPCState *s = bs->opaque; | |
7267c094 | 751 | g_free(s->pagetable); |
6a0f9e82 | 752 | #ifdef CACHE |
7267c094 | 753 | g_free(s->pageentry_u8); |
6a0f9e82 | 754 | #endif |
612ff3d8 KW |
755 | |
756 | migrate_del_blocker(s->migration_blocker); | |
757 | error_free(s->migration_blocker); | |
6a0f9e82 FB |
758 | } |
759 | ||
0e7e1989 | 760 | static QEMUOptionParameter vpc_create_options[] = { |
db08adf5 KW |
761 | { |
762 | .name = BLOCK_OPT_SIZE, | |
763 | .type = OPT_SIZE, | |
764 | .help = "Virtual disk size" | |
765 | }, | |
24da78db CA |
766 | { |
767 | .name = BLOCK_OPT_SUBFMT, | |
768 | .type = OPT_STRING, | |
769 | .help = | |
770 | "Type of virtual hard disk format. Supported formats are " | |
771 | "{dynamic (default) | fixed} " | |
772 | }, | |
0e7e1989 KW |
773 | { NULL } |
774 | }; | |
775 | ||
5efa9d5a | 776 | static BlockDriver bdrv_vpc = { |
4a411185 KW |
777 | .format_name = "vpc", |
778 | .instance_size = sizeof(BDRVVPCState), | |
c68b89ac | 779 | |
4a411185 KW |
780 | .bdrv_probe = vpc_probe, |
781 | .bdrv_open = vpc_open, | |
4a411185 KW |
782 | .bdrv_close = vpc_close, |
783 | .bdrv_create = vpc_create, | |
0e7e1989 | 784 | |
c68b89ac KW |
785 | .bdrv_read = vpc_co_read, |
786 | .bdrv_write = vpc_co_write, | |
c68b89ac | 787 | |
0e7e1989 | 788 | .create_options = vpc_create_options, |
6a0f9e82 | 789 | }; |
5efa9d5a AL |
790 | |
791 | static void bdrv_vpc_init(void) | |
792 | { | |
793 | bdrv_register(&bdrv_vpc); | |
794 | } | |
795 | ||
796 | block_init(bdrv_vpc_init); |