]> git.proxmox.com Git - mirror_qemu.git/blob - block-vpc.c
block-vpc: Create images (Kevin Wolf)
[mirror_qemu.git] / block-vpc.c
1 /*
2 * Block driver for Conectix/Microsoft Virtual PC images
3 *
4 * Copyright (c) 2005 Alex Beregszaszi
5 * Copyright (c) 2009 Kevin Wolf <kwolf@suse.de>
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a copy
8 * of this software and associated documentation files (the "Software"), to deal
9 * in the Software without restriction, including without limitation the rights
10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11 * copies of the Software, and to permit persons to whom the Software is
12 * furnished to do so, subject to the following conditions:
13 *
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
23 * THE SOFTWARE.
24 */
25 #include "qemu-common.h"
26 #include "block_int.h"
27
28 /**************************************************************/
29
30 #define HEADER_SIZE 512
31
32 //#define CACHE
33
34 enum vhd_type {
35 VHD_FIXED = 2,
36 VHD_DYNAMIC = 3,
37 VHD_DIFFERENCING = 4,
38 };
39
40 // Seconds since Jan 1, 2000 0:00:00 (UTC)
41 #define VHD_TIMESTAMP_BASE 946684800
42
43 // always big-endian
44 struct vhd_footer {
45 char creator[8]; // "conectix"
46 uint32_t features;
47 uint32_t version;
48
49 // Offset of next header structure, 0xFFFFFFFF if none
50 uint64_t data_offset;
51
52 // Seconds since Jan 1, 2000 0:00:00 (UTC)
53 uint32_t timestamp;
54
55 char creator_app[4]; // "vpc "
56 uint16_t major;
57 uint16_t minor;
58 char creator_os[4]; // "Wi2k"
59
60 uint64_t orig_size;
61 uint64_t size;
62
63 uint16_t cyls;
64 uint8_t heads;
65 uint8_t secs_per_cyl;
66
67 uint32_t type;
68
69 // Checksum of the Hard Disk Footer ("one's complement of the sum of all
70 // the bytes in the footer without the checksum field")
71 uint32_t checksum;
72
73 // UUID used to identify a parent hard disk (backing file)
74 uint8_t uuid[16];
75
76 uint8_t in_saved_state;
77 };
78
79 struct vhd_dyndisk_header {
80 char magic[8]; // "cxsparse"
81
82 // Offset of next header structure, 0xFFFFFFFF if none
83 uint64_t data_offset;
84
85 // Offset of the Block Allocation Table (BAT)
86 uint64_t table_offset;
87
88 uint32_t version;
89 uint32_t max_table_entries; // 32bit/entry
90
91 // 2 MB by default, must be a power of two
92 uint32_t block_size;
93
94 uint32_t checksum;
95 uint8_t parent_uuid[16];
96 uint32_t parent_timestamp;
97 uint32_t reserved;
98
99 // Backing file name (in UTF-16)
100 uint8_t parent_name[512];
101
102 struct {
103 uint32_t platform;
104 uint32_t data_space;
105 uint32_t data_length;
106 uint32_t reserved;
107 uint64_t data_offset;
108 } parent_locator[8];
109 };
110
111 typedef struct BDRVVPCState {
112 BlockDriverState *hd;
113
114 uint8_t footer_buf[HEADER_SIZE];
115 uint64_t free_data_block_offset;
116 int max_table_entries;
117 uint32_t *pagetable;
118 uint64_t bat_offset;
119 uint64_t last_bitmap_offset;
120
121 uint32_t block_size;
122 uint32_t bitmap_size;
123
124 #ifdef CACHE
125 uint8_t *pageentry_u8;
126 uint32_t *pageentry_u32;
127 uint16_t *pageentry_u16;
128
129 uint64_t last_bitmap;
130 #endif
131 } BDRVVPCState;
132
133 static uint32_t vpc_checksum(uint8_t* buf, size_t size)
134 {
135 uint32_t res = 0;
136 int i;
137
138 for (i = 0; i < size; i++)
139 res += buf[i];
140
141 return ~res;
142 }
143
144
145 static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename)
146 {
147 if (buf_size >= 8 && !strncmp((char *)buf, "conectix", 8))
148 return 100;
149 return 0;
150 }
151
152 static int vpc_open(BlockDriverState *bs, const char *filename, int flags)
153 {
154 BDRVVPCState *s = bs->opaque;
155 int ret, i;
156 struct vhd_footer* footer;
157 struct vhd_dyndisk_header* dyndisk_header;
158 uint8_t buf[HEADER_SIZE];
159 uint32_t checksum;
160
161 ret = bdrv_file_open(&s->hd, filename, flags);
162 if (ret < 0)
163 return ret;
164
165 if (bdrv_pread(s->hd, 0, s->footer_buf, HEADER_SIZE) != HEADER_SIZE)
166 goto fail;
167
168 footer = (struct vhd_footer*) s->footer_buf;
169 if (strncmp(footer->creator, "conectix", 8))
170 goto fail;
171
172 checksum = be32_to_cpu(footer->checksum);
173 footer->checksum = 0;
174 if (vpc_checksum(s->footer_buf, HEADER_SIZE) != checksum)
175 fprintf(stderr, "block-vpc: The header checksum of '%s' is "
176 "incorrect.\n", filename);
177
178 // The visible size of a image in Virtual PC depends on the geometry
179 // rather than on the size stored in the footer (the size in the footer
180 // is too large usually)
181 bs->total_sectors = (int64_t)
182 be16_to_cpu(footer->cyls) * footer->heads * footer->secs_per_cyl;
183
184 if (bdrv_pread(s->hd, be64_to_cpu(footer->data_offset), buf, HEADER_SIZE)
185 != HEADER_SIZE)
186 goto fail;
187
188 dyndisk_header = (struct vhd_dyndisk_header*) buf;
189
190 if (strncmp(dyndisk_header->magic, "cxsparse", 8))
191 goto fail;
192
193
194 s->block_size = be32_to_cpu(dyndisk_header->block_size);
195 s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511;
196
197 s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries);
198 s->pagetable = qemu_malloc(s->max_table_entries * 4);
199 if (!s->pagetable)
200 goto fail;
201
202 s->bat_offset = be64_to_cpu(dyndisk_header->table_offset);
203 if (bdrv_pread(s->hd, s->bat_offset, s->pagetable,
204 s->max_table_entries * 4) != s->max_table_entries * 4)
205 goto fail;
206
207 s->free_data_block_offset =
208 (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511;
209
210 for (i = 0; i < s->max_table_entries; i++) {
211 be32_to_cpus(&s->pagetable[i]);
212 if (s->pagetable[i] != 0xFFFFFFFF) {
213 int64_t next = (512 * (int64_t) s->pagetable[i]) +
214 s->bitmap_size + s->block_size;
215
216 if (next> s->free_data_block_offset)
217 s->free_data_block_offset = next;
218 }
219 }
220
221 s->last_bitmap_offset = (int64_t) -1;
222
223 #ifdef CACHE
224 s->pageentry_u8 = qemu_malloc(512);
225 if (!s->pageentry_u8)
226 goto fail;
227 s->pageentry_u32 = s->pageentry_u8;
228 s->pageentry_u16 = s->pageentry_u8;
229 s->last_pagetable = -1;
230 #endif
231
232 return 0;
233 fail:
234 bdrv_delete(s->hd);
235 return -1;
236 }
237
238 /*
239 * Returns the absolute byte offset of the given sector in the image file.
240 * If the sector is not allocated, -1 is returned instead.
241 *
242 * The parameter write must be 1 if the offset will be used for a write
243 * operation (the block bitmaps is updated then), 0 otherwise.
244 */
245 static inline int64_t get_sector_offset(BlockDriverState *bs,
246 int64_t sector_num, int write)
247 {
248 BDRVVPCState *s = bs->opaque;
249 uint64_t offset = sector_num * 512;
250 uint64_t bitmap_offset, block_offset;
251 uint32_t pagetable_index, pageentry_index;
252
253 pagetable_index = offset / s->block_size;
254 pageentry_index = (offset % s->block_size) / 512;
255
256 if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff)
257 return -1; // not allocated
258
259 bitmap_offset = 512 * s->pagetable[pagetable_index];
260 block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index);
261
262 // We must ensure that we don't write to any sectors which are marked as
263 // unused in the bitmap. We get away with setting all bits in the block
264 // bitmap each time we write to a new block. This might cause Virtual PC to
265 // miss sparse read optimization, but it's not a problem in terms of
266 // correctness.
267 if (write && (s->last_bitmap_offset != bitmap_offset)) {
268 uint8_t bitmap[s->bitmap_size];
269
270 s->last_bitmap_offset = bitmap_offset;
271 memset(bitmap, 0xff, s->bitmap_size);
272 bdrv_pwrite(s->hd, bitmap_offset, bitmap, s->bitmap_size);
273 }
274
275 // printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n",
276 // sector_num, pagetable_index, pageentry_index,
277 // bitmap_offset, block_offset);
278
279 // disabled by reason
280 #if 0
281 #ifdef CACHE
282 if (bitmap_offset != s->last_bitmap)
283 {
284 lseek(s->fd, bitmap_offset, SEEK_SET);
285
286 s->last_bitmap = bitmap_offset;
287
288 // Scary! Bitmap is stored as big endian 32bit entries,
289 // while we used to look it up byte by byte
290 read(s->fd, s->pageentry_u8, 512);
291 for (i = 0; i < 128; i++)
292 be32_to_cpus(&s->pageentry_u32[i]);
293 }
294
295 if ((s->pageentry_u8[pageentry_index / 8] >> (pageentry_index % 8)) & 1)
296 return -1;
297 #else
298 lseek(s->fd, bitmap_offset + (pageentry_index / 8), SEEK_SET);
299
300 read(s->fd, &bitmap_entry, 1);
301
302 if ((bitmap_entry >> (pageentry_index % 8)) & 1)
303 return -1; // not allocated
304 #endif
305 #endif
306
307 return block_offset;
308 }
309
310 /*
311 * Writes the footer to the end of the image file. This is needed when the
312 * file grows as it overwrites the old footer
313 *
314 * Returns 0 on success and < 0 on error
315 */
316 static int rewrite_footer(BlockDriverState* bs)
317 {
318 int ret;
319 BDRVVPCState *s = bs->opaque;
320 int64_t offset = s->free_data_block_offset;
321
322 ret = bdrv_pwrite(s->hd, offset, s->footer_buf, HEADER_SIZE);
323 if (ret < 0)
324 return ret;
325
326 return 0;
327 }
328
329 /*
330 * Allocates a new block. This involves writing a new footer and updating
331 * the Block Allocation Table to use the space at the old end of the image
332 * file (overwriting the old footer)
333 *
334 * Returns the sectors' offset in the image file on success and < 0 on error
335 */
336 static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num)
337 {
338 BDRVVPCState *s = bs->opaque;
339 int64_t bat_offset;
340 uint32_t index, bat_value;
341 int ret;
342 uint8_t bitmap[s->bitmap_size];
343
344 // Check if sector_num is valid
345 if ((sector_num < 0) || (sector_num > bs->total_sectors))
346 return -1;
347
348 // Write entry into in-memory BAT
349 index = (sector_num * 512) / s->block_size;
350 if (s->pagetable[index] != 0xFFFFFFFF)
351 return -1;
352
353 s->pagetable[index] = s->free_data_block_offset / 512;
354
355 // Initialize the block's bitmap
356 memset(bitmap, 0xff, s->bitmap_size);
357 bdrv_pwrite(s->hd, s->free_data_block_offset, bitmap, s->bitmap_size);
358
359 // Write new footer (the old one will be overwritten)
360 s->free_data_block_offset += s->block_size + s->bitmap_size;
361 ret = rewrite_footer(bs);
362 if (ret < 0)
363 goto fail;
364
365 // Write BAT entry to disk
366 bat_offset = s->bat_offset + (4 * index);
367 bat_value = be32_to_cpu(s->pagetable[index]);
368 ret = bdrv_pwrite(s->hd, bat_offset, &bat_value, 4);
369 if (ret < 0)
370 goto fail;
371
372 return get_sector_offset(bs, sector_num, 0);
373
374 fail:
375 s->free_data_block_offset -= (s->block_size + s->bitmap_size);
376 return -1;
377 }
378
379 static int vpc_read(BlockDriverState *bs, int64_t sector_num,
380 uint8_t *buf, int nb_sectors)
381 {
382 BDRVVPCState *s = bs->opaque;
383 int ret;
384 int64_t offset;
385
386 while (nb_sectors > 0) {
387 offset = get_sector_offset(bs, sector_num, 0);
388
389 if (offset == -1) {
390 memset(buf, 0, 512);
391 } else {
392 ret = bdrv_pread(s->hd, offset, buf, 512);
393 if (ret != 512)
394 return -1;
395 }
396
397 nb_sectors--;
398 sector_num++;
399 buf += 512;
400 }
401 return 0;
402 }
403
404 static int vpc_write(BlockDriverState *bs, int64_t sector_num,
405 const uint8_t *buf, int nb_sectors)
406 {
407 BDRVVPCState *s = bs->opaque;
408 int64_t offset;
409 int ret;
410
411 while (nb_sectors > 0) {
412 offset = get_sector_offset(bs, sector_num, 1);
413
414 if (offset == -1) {
415 offset = alloc_block(bs, sector_num);
416 if (offset < 0)
417 return -1;
418 }
419
420 ret = bdrv_pwrite(s->hd, offset, buf, 512);
421 if (ret != 512)
422 return -1;
423
424 nb_sectors--;
425 sector_num++;
426 buf += 512;
427 }
428
429 return 0;
430 }
431
432
433 /*
434 * Calculates the number of cylinders, heads and sectors per cylinder
435 * based on a given number of sectors. This is the algorithm described
436 * in the VHD specification.
437 *
438 * Note that the geometry doesn't always exactly match total_sectors but
439 * may round it down.
440 */
441 static void calculate_geometry(int64_t total_sectors, uint16_t* cyls,
442 uint8_t* heads, uint8_t* secs_per_cyl)
443 {
444 uint32_t cyls_times_heads;
445
446 if (total_sectors > 65535 * 16 * 255)
447 total_sectors = 65535 * 16 * 255;
448
449 if (total_sectors > 65535 * 16 * 63) {
450 *secs_per_cyl = 255;
451 *heads = 16;
452 cyls_times_heads = total_sectors / *secs_per_cyl;
453 } else {
454 *secs_per_cyl = 17;
455 cyls_times_heads = total_sectors / *secs_per_cyl;
456 *heads = (cyls_times_heads + 1023) / 1024;
457
458 if (*heads < 4)
459 *heads = 4;
460
461 if (cyls_times_heads >= (*heads * 1024) || *heads > 16) {
462 *secs_per_cyl = 31;
463 *heads = 16;
464 cyls_times_heads = total_sectors / *secs_per_cyl;
465 }
466
467 if (cyls_times_heads >= (*heads * 1024)) {
468 *secs_per_cyl = 63;
469 *heads = 16;
470 cyls_times_heads = total_sectors / *secs_per_cyl;
471 }
472 }
473
474 // Note: Rounding up deviates from the Virtual PC behaviour
475 // However, we need this to avoid truncating images in qemu-img convert
476 *cyls = (cyls_times_heads + *heads - 1) / *heads;
477 }
478
479 static int vpc_create(const char *filename, int64_t total_sectors,
480 const char *backing_file, int flags)
481 {
482 uint8_t buf[1024];
483 struct vhd_footer* footer = (struct vhd_footer*) buf;
484 struct vhd_dyndisk_header* dyndisk_header =
485 (struct vhd_dyndisk_header*) buf;
486 int fd, i;
487 uint16_t cyls;
488 uint8_t heads;
489 uint8_t secs_per_cyl;
490 size_t block_size, num_bat_entries;
491
492 if (backing_file != NULL)
493 return -ENOTSUP;
494
495 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644);
496 if (fd < 0)
497 return -EIO;
498
499 // Calculate matching total_size and geometry
500 calculate_geometry(total_sectors, &cyls, &heads, &secs_per_cyl);
501 total_sectors = (int64_t) cyls * heads * secs_per_cyl;
502
503 // Prepare the Hard Disk Footer
504 memset(buf, 0, 1024);
505
506 strncpy(footer->creator, "conectix", 8);
507 // TODO Check if "qemu" creator_app is ok for VPC
508 strncpy(footer->creator_app, "qemu", 4);
509 strncpy(footer->creator_os, "Wi2k", 4);
510
511 footer->features = be32_to_cpu(0x02);
512 footer->version = be32_to_cpu(0x00010000);
513 footer->data_offset = be64_to_cpu(HEADER_SIZE);
514 footer->timestamp = be32_to_cpu(time(NULL) - VHD_TIMESTAMP_BASE);
515
516 // Version of Virtual PC 2007
517 footer->major = be16_to_cpu(0x0005);
518 footer->minor =be16_to_cpu(0x0003);
519
520 footer->orig_size = be64_to_cpu(total_sectors * 512);
521 footer->size = be64_to_cpu(total_sectors * 512);
522
523 footer->cyls = be16_to_cpu(cyls);
524 footer->heads = heads;
525 footer->secs_per_cyl = secs_per_cyl;
526
527 footer->type = be32_to_cpu(VHD_DYNAMIC);
528
529 // TODO uuid is missing
530
531 footer->checksum = be32_to_cpu(vpc_checksum(buf, HEADER_SIZE));
532
533 // Write the footer (twice: at the beginning and at the end)
534 block_size = 0x200000;
535 num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512);
536
537 if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE)
538 return -EIO;
539
540 if (lseek(fd, 1536 + ((num_bat_entries * 4 + 511) & ~511), SEEK_SET) < 0)
541 return -EIO;
542 if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE)
543 return -EIO;
544
545 // Write the initial BAT
546 if (lseek(fd, 3 * 512, SEEK_SET) < 0)
547 return -EIO;
548
549 memset(buf, 0xFF, 512);
550 for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++)
551 if (write(fd, buf, 512) != 512)
552 return -EIO;
553
554
555 // Prepare the Dynamic Disk Header
556 memset(buf, 0, 1024);
557
558 strncpy(dyndisk_header->magic, "cxsparse", 8);
559
560 dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFF);
561 dyndisk_header->table_offset = be64_to_cpu(3 * 512);
562 dyndisk_header->version = be32_to_cpu(0x00010000);
563 dyndisk_header->block_size = be32_to_cpu(block_size);
564 dyndisk_header->max_table_entries = be32_to_cpu(num_bat_entries);
565
566 dyndisk_header->checksum = be32_to_cpu(vpc_checksum(buf, 1024));
567
568 // Write the header
569 if (lseek(fd, 512, SEEK_SET) < 0)
570 return -EIO;
571 if (write(fd, buf, 1024) != 1024)
572 return -EIO;
573
574 close(fd);
575 return 0;
576 }
577
578 static void vpc_close(BlockDriverState *bs)
579 {
580 BDRVVPCState *s = bs->opaque;
581 qemu_free(s->pagetable);
582 #ifdef CACHE
583 qemu_free(s->pageentry_u8);
584 #endif
585 bdrv_delete(s->hd);
586 }
587
588 BlockDriver bdrv_vpc = {
589 "vpc",
590 sizeof(BDRVVPCState),
591 vpc_probe,
592 vpc_open,
593 vpc_read,
594 vpc_write,
595 vpc_close,
596 vpc_create,
597 };