./configure: request pkg-config to provide private libs when static linking
[qemu/ar7.git] / block / vpc.c
blob89a5ee26686a08252765de544a8aac4078a4e8f5
1 /*
2 * Block driver for Connectix / Microsoft Virtual PC images
4 * Copyright (c) 2005 Alex Beregszaszi
5 * Copyright (c) 2009 Kevin Wolf <kwolf@suse.de>
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:
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
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.
25 #include "qemu-common.h"
26 #include "block_int.h"
27 #include "module.h"
28 #include "migration.h"
30 /**************************************************************/
32 #define HEADER_SIZE 512
34 //#define CACHE
36 enum vhd_type {
37 VHD_FIXED = 2,
38 VHD_DYNAMIC = 3,
39 VHD_DIFFERENCING = 4,
42 // Seconds since Jan 1, 2000 0:00:00 (UTC)
43 #define VHD_TIMESTAMP_BASE 946684800
45 // always big-endian
46 struct vhd_footer {
47 char creator[8]; // "conectix"
48 uint32_t features;
49 uint32_t version;
51 // Offset of next header structure, 0xFFFFFFFF if none
52 uint64_t data_offset;
54 // Seconds since Jan 1, 2000 0:00:00 (UTC)
55 uint32_t timestamp;
57 char creator_app[4]; // "vpc "
58 uint16_t major;
59 uint16_t minor;
60 char creator_os[4]; // "Wi2k"
62 uint64_t orig_size;
63 uint64_t size;
65 uint16_t cyls;
66 uint8_t heads;
67 uint8_t secs_per_cyl;
69 uint32_t type;
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;
75 // UUID used to identify a parent hard disk (backing file)
76 uint8_t uuid[16];
78 uint8_t in_saved_state;
81 struct vhd_dyndisk_header {
82 char magic[8]; // "cxsparse"
84 // Offset of next header structure, 0xFFFFFFFF if none
85 uint64_t data_offset;
87 // Offset of the Block Allocation Table (BAT)
88 uint64_t table_offset;
90 uint32_t version;
91 uint32_t max_table_entries; // 32bit/entry
93 // 2 MB by default, must be a power of two
94 uint32_t block_size;
96 uint32_t checksum;
97 uint8_t parent_uuid[16];
98 uint32_t parent_timestamp;
99 uint32_t reserved;
101 // Backing file name (in UTF-16)
102 uint8_t parent_name[512];
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];
113 typedef struct BDRVVPCState {
114 CoMutex lock;
115 uint8_t footer_buf[HEADER_SIZE];
116 uint64_t free_data_block_offset;
117 int max_table_entries;
118 uint32_t *pagetable;
119 uint64_t bat_offset;
120 uint64_t last_bitmap_offset;
122 uint32_t block_size;
123 uint32_t bitmap_size;
125 #ifdef CACHE
126 uint8_t *pageentry_u8;
127 uint32_t *pageentry_u32;
128 uint16_t *pageentry_u16;
130 uint64_t last_bitmap;
131 #endif
133 Error *migration_blocker;
134 } BDRVVPCState;
136 static uint32_t vpc_checksum(uint8_t* buf, size_t size)
138 uint32_t res = 0;
139 int i;
141 for (i = 0; i < size; i++)
142 res += buf[i];
144 return ~res;
148 static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename)
150 if (buf_size >= 8 && !strncmp((char *)buf, "conectix", 8))
151 return 100;
152 return 0;
155 static int vpc_open(BlockDriverState *bs, int flags)
157 BDRVVPCState *s = bs->opaque;
158 int i;
159 struct vhd_footer* footer;
160 struct vhd_dyndisk_header* dyndisk_header;
161 uint8_t buf[HEADER_SIZE];
162 uint32_t checksum;
163 int err = -1;
165 if (bdrv_pread(bs->file, 0, s->footer_buf, HEADER_SIZE) != HEADER_SIZE)
166 goto fail;
168 footer = (struct vhd_footer*) s->footer_buf;
169 if (strncmp(footer->creator, "conectix", 8))
170 goto fail;
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", bs->filename);
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;
184 if (bs->total_sectors >= 65535 * 16 * 255) {
185 err = -EFBIG;
186 goto fail;
189 if (bdrv_pread(bs->file, be64_to_cpu(footer->data_offset), buf, HEADER_SIZE)
190 != HEADER_SIZE)
191 goto fail;
193 dyndisk_header = (struct vhd_dyndisk_header*) buf;
195 if (strncmp(dyndisk_header->magic, "cxsparse", 8))
196 goto fail;
199 s->block_size = be32_to_cpu(dyndisk_header->block_size);
200 s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511;
202 s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries);
203 s->pagetable = g_malloc(s->max_table_entries * 4);
205 s->bat_offset = be64_to_cpu(dyndisk_header->table_offset);
206 if (bdrv_pread(bs->file, s->bat_offset, s->pagetable,
207 s->max_table_entries * 4) != s->max_table_entries * 4)
208 goto fail;
210 s->free_data_block_offset =
211 (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511;
213 for (i = 0; i < s->max_table_entries; i++) {
214 be32_to_cpus(&s->pagetable[i]);
215 if (s->pagetable[i] != 0xFFFFFFFF) {
216 int64_t next = (512 * (int64_t) s->pagetable[i]) +
217 s->bitmap_size + s->block_size;
219 if (next> s->free_data_block_offset)
220 s->free_data_block_offset = next;
224 s->last_bitmap_offset = (int64_t) -1;
226 #ifdef CACHE
227 s->pageentry_u8 = g_malloc(512);
228 s->pageentry_u32 = s->pageentry_u8;
229 s->pageentry_u16 = s->pageentry_u8;
230 s->last_pagetable = -1;
231 #endif
233 qemu_co_mutex_init(&s->lock);
235 /* Disable migration when VHD images are used */
236 error_set(&s->migration_blocker,
237 QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
238 "vpc", bs->device_name, "live migration");
239 migrate_add_blocker(s->migration_blocker);
241 return 0;
242 fail:
243 return err;
247 * Returns the absolute byte offset of the given sector in the image file.
248 * If the sector is not allocated, -1 is returned instead.
250 * The parameter write must be 1 if the offset will be used for a write
251 * operation (the block bitmaps is updated then), 0 otherwise.
253 static inline int64_t get_sector_offset(BlockDriverState *bs,
254 int64_t sector_num, int write)
256 BDRVVPCState *s = bs->opaque;
257 uint64_t offset = sector_num * 512;
258 uint64_t bitmap_offset, block_offset;
259 uint32_t pagetable_index, pageentry_index;
261 pagetable_index = offset / s->block_size;
262 pageentry_index = (offset % s->block_size) / 512;
264 if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff)
265 return -1; // not allocated
267 bitmap_offset = 512 * (uint64_t) s->pagetable[pagetable_index];
268 block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index);
270 // We must ensure that we don't write to any sectors which are marked as
271 // unused in the bitmap. We get away with setting all bits in the block
272 // bitmap each time we write to a new block. This might cause Virtual PC to
273 // miss sparse read optimization, but it's not a problem in terms of
274 // correctness.
275 if (write && (s->last_bitmap_offset != bitmap_offset)) {
276 uint8_t bitmap[s->bitmap_size];
278 s->last_bitmap_offset = bitmap_offset;
279 memset(bitmap, 0xff, s->bitmap_size);
280 bdrv_pwrite_sync(bs->file, bitmap_offset, bitmap, s->bitmap_size);
283 // printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n",
284 // sector_num, pagetable_index, pageentry_index,
285 // bitmap_offset, block_offset);
287 // disabled by reason
288 #if 0
289 #ifdef CACHE
290 if (bitmap_offset != s->last_bitmap)
292 lseek(s->fd, bitmap_offset, SEEK_SET);
294 s->last_bitmap = bitmap_offset;
296 // Scary! Bitmap is stored as big endian 32bit entries,
297 // while we used to look it up byte by byte
298 read(s->fd, s->pageentry_u8, 512);
299 for (i = 0; i < 128; i++)
300 be32_to_cpus(&s->pageentry_u32[i]);
303 if ((s->pageentry_u8[pageentry_index / 8] >> (pageentry_index % 8)) & 1)
304 return -1;
305 #else
306 lseek(s->fd, bitmap_offset + (pageentry_index / 8), SEEK_SET);
308 read(s->fd, &bitmap_entry, 1);
310 if ((bitmap_entry >> (pageentry_index % 8)) & 1)
311 return -1; // not allocated
312 #endif
313 #endif
315 return block_offset;
319 * Writes the footer to the end of the image file. This is needed when the
320 * file grows as it overwrites the old footer
322 * Returns 0 on success and < 0 on error
324 static int rewrite_footer(BlockDriverState* bs)
326 int ret;
327 BDRVVPCState *s = bs->opaque;
328 int64_t offset = s->free_data_block_offset;
330 ret = bdrv_pwrite_sync(bs->file, offset, s->footer_buf, HEADER_SIZE);
331 if (ret < 0)
332 return ret;
334 return 0;
338 * Allocates a new block. This involves writing a new footer and updating
339 * the Block Allocation Table to use the space at the old end of the image
340 * file (overwriting the old footer)
342 * Returns the sectors' offset in the image file on success and < 0 on error
344 static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num)
346 BDRVVPCState *s = bs->opaque;
347 int64_t bat_offset;
348 uint32_t index, bat_value;
349 int ret;
350 uint8_t bitmap[s->bitmap_size];
352 // Check if sector_num is valid
353 if ((sector_num < 0) || (sector_num > bs->total_sectors))
354 return -1;
356 // Write entry into in-memory BAT
357 index = (sector_num * 512) / s->block_size;
358 if (s->pagetable[index] != 0xFFFFFFFF)
359 return -1;
361 s->pagetable[index] = s->free_data_block_offset / 512;
363 // Initialize the block's bitmap
364 memset(bitmap, 0xff, s->bitmap_size);
365 ret = bdrv_pwrite_sync(bs->file, s->free_data_block_offset, bitmap,
366 s->bitmap_size);
367 if (ret < 0) {
368 return ret;
371 // Write new footer (the old one will be overwritten)
372 s->free_data_block_offset += s->block_size + s->bitmap_size;
373 ret = rewrite_footer(bs);
374 if (ret < 0)
375 goto fail;
377 // Write BAT entry to disk
378 bat_offset = s->bat_offset + (4 * index);
379 bat_value = be32_to_cpu(s->pagetable[index]);
380 ret = bdrv_pwrite_sync(bs->file, bat_offset, &bat_value, 4);
381 if (ret < 0)
382 goto fail;
384 return get_sector_offset(bs, sector_num, 0);
386 fail:
387 s->free_data_block_offset -= (s->block_size + s->bitmap_size);
388 return -1;
391 static int vpc_read(BlockDriverState *bs, int64_t sector_num,
392 uint8_t *buf, int nb_sectors)
394 BDRVVPCState *s = bs->opaque;
395 int ret;
396 int64_t offset;
397 int64_t sectors, sectors_per_block;
399 while (nb_sectors > 0) {
400 offset = get_sector_offset(bs, sector_num, 0);
402 sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
403 sectors = sectors_per_block - (sector_num % sectors_per_block);
404 if (sectors > nb_sectors) {
405 sectors = nb_sectors;
408 if (offset == -1) {
409 memset(buf, 0, sectors * BDRV_SECTOR_SIZE);
410 } else {
411 ret = bdrv_pread(bs->file, offset, buf,
412 sectors * BDRV_SECTOR_SIZE);
413 if (ret != sectors * BDRV_SECTOR_SIZE) {
414 return -1;
418 nb_sectors -= sectors;
419 sector_num += sectors;
420 buf += sectors * BDRV_SECTOR_SIZE;
422 return 0;
425 static coroutine_fn int vpc_co_read(BlockDriverState *bs, int64_t sector_num,
426 uint8_t *buf, int nb_sectors)
428 int ret;
429 BDRVVPCState *s = bs->opaque;
430 qemu_co_mutex_lock(&s->lock);
431 ret = vpc_read(bs, sector_num, buf, nb_sectors);
432 qemu_co_mutex_unlock(&s->lock);
433 return ret;
436 static int vpc_write(BlockDriverState *bs, int64_t sector_num,
437 const uint8_t *buf, int nb_sectors)
439 BDRVVPCState *s = bs->opaque;
440 int64_t offset;
441 int64_t sectors, sectors_per_block;
442 int ret;
444 while (nb_sectors > 0) {
445 offset = get_sector_offset(bs, sector_num, 1);
447 sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
448 sectors = sectors_per_block - (sector_num % sectors_per_block);
449 if (sectors > nb_sectors) {
450 sectors = nb_sectors;
453 if (offset == -1) {
454 offset = alloc_block(bs, sector_num);
455 if (offset < 0)
456 return -1;
459 ret = bdrv_pwrite(bs->file, offset, buf, sectors * BDRV_SECTOR_SIZE);
460 if (ret != sectors * BDRV_SECTOR_SIZE) {
461 return -1;
464 nb_sectors -= sectors;
465 sector_num += sectors;
466 buf += sectors * BDRV_SECTOR_SIZE;
469 return 0;
472 static coroutine_fn int vpc_co_write(BlockDriverState *bs, int64_t sector_num,
473 const uint8_t *buf, int nb_sectors)
475 int ret;
476 BDRVVPCState *s = bs->opaque;
477 qemu_co_mutex_lock(&s->lock);
478 ret = vpc_write(bs, sector_num, buf, nb_sectors);
479 qemu_co_mutex_unlock(&s->lock);
480 return ret;
483 static coroutine_fn int vpc_co_flush(BlockDriverState *bs)
485 return bdrv_co_flush(bs->file);
489 * Calculates the number of cylinders, heads and sectors per cylinder
490 * based on a given number of sectors. This is the algorithm described
491 * in the VHD specification.
493 * Note that the geometry doesn't always exactly match total_sectors but
494 * may round it down.
496 * Returns 0 on success, -EFBIG if the size is larger than 127 GB
498 static int calculate_geometry(int64_t total_sectors, uint16_t* cyls,
499 uint8_t* heads, uint8_t* secs_per_cyl)
501 uint32_t cyls_times_heads;
503 if (total_sectors > 65535 * 16 * 255)
504 return -EFBIG;
506 if (total_sectors > 65535 * 16 * 63) {
507 *secs_per_cyl = 255;
508 *heads = 16;
509 cyls_times_heads = total_sectors / *secs_per_cyl;
510 } else {
511 *secs_per_cyl = 17;
512 cyls_times_heads = total_sectors / *secs_per_cyl;
513 *heads = (cyls_times_heads + 1023) / 1024;
515 if (*heads < 4)
516 *heads = 4;
518 if (cyls_times_heads >= (*heads * 1024) || *heads > 16) {
519 *secs_per_cyl = 31;
520 *heads = 16;
521 cyls_times_heads = total_sectors / *secs_per_cyl;
524 if (cyls_times_heads >= (*heads * 1024)) {
525 *secs_per_cyl = 63;
526 *heads = 16;
527 cyls_times_heads = total_sectors / *secs_per_cyl;
531 *cyls = cyls_times_heads / *heads;
533 return 0;
536 static int vpc_create(const char *filename, QEMUOptionParameter *options)
538 uint8_t buf[1024];
539 struct vhd_footer* footer = (struct vhd_footer*) buf;
540 struct vhd_dyndisk_header* dyndisk_header =
541 (struct vhd_dyndisk_header*) buf;
542 int fd, i;
543 uint16_t cyls = 0;
544 uint8_t heads = 0;
545 uint8_t secs_per_cyl = 0;
546 size_t block_size, num_bat_entries;
547 int64_t total_sectors = 0;
548 int ret = -EIO;
550 // Read out options
551 total_sectors = get_option_parameter(options, BLOCK_OPT_SIZE)->value.n /
552 BDRV_SECTOR_SIZE;
554 // Create the file
555 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644);
556 if (fd < 0)
557 return -EIO;
559 /* Calculate matching total_size and geometry. Increase the number of
560 sectors requested until we get enough (or fail). */
561 for (i = 0; total_sectors > (int64_t)cyls * heads * secs_per_cyl; i++) {
562 if (calculate_geometry(total_sectors + i,
563 &cyls, &heads, &secs_per_cyl)) {
564 ret = -EFBIG;
565 goto fail;
568 total_sectors = (int64_t) cyls * heads * secs_per_cyl;
570 // Prepare the Hard Disk Footer
571 memset(buf, 0, 1024);
573 memcpy(footer->creator, "conectix", 8);
574 // TODO Check if "qemu" creator_app is ok for VPC
575 memcpy(footer->creator_app, "qemu", 4);
576 memcpy(footer->creator_os, "Wi2k", 4);
578 footer->features = be32_to_cpu(0x02);
579 footer->version = be32_to_cpu(0x00010000);
580 footer->data_offset = be64_to_cpu(HEADER_SIZE);
581 footer->timestamp = be32_to_cpu(time(NULL) - VHD_TIMESTAMP_BASE);
583 // Version of Virtual PC 2007
584 footer->major = be16_to_cpu(0x0005);
585 footer->minor =be16_to_cpu(0x0003);
587 footer->orig_size = be64_to_cpu(total_sectors * 512);
588 footer->size = be64_to_cpu(total_sectors * 512);
590 footer->cyls = be16_to_cpu(cyls);
591 footer->heads = heads;
592 footer->secs_per_cyl = secs_per_cyl;
594 footer->type = be32_to_cpu(VHD_DYNAMIC);
596 // TODO uuid is missing
598 footer->checksum = be32_to_cpu(vpc_checksum(buf, HEADER_SIZE));
600 // Write the footer (twice: at the beginning and at the end)
601 block_size = 0x200000;
602 num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512);
604 if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
605 goto fail;
608 if (lseek(fd, 1536 + ((num_bat_entries * 4 + 511) & ~511), SEEK_SET) < 0) {
609 goto fail;
611 if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
612 goto fail;
615 // Write the initial BAT
616 if (lseek(fd, 3 * 512, SEEK_SET) < 0) {
617 goto fail;
620 memset(buf, 0xFF, 512);
621 for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++) {
622 if (write(fd, buf, 512) != 512) {
623 goto fail;
628 // Prepare the Dynamic Disk Header
629 memset(buf, 0, 1024);
631 memcpy(dyndisk_header->magic, "cxsparse", 8);
634 * Note: The spec is actually wrong here for data_offset, it says
635 * 0xFFFFFFFF, but MS tools expect all 64 bits to be set.
637 dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
638 dyndisk_header->table_offset = be64_to_cpu(3 * 512);
639 dyndisk_header->version = be32_to_cpu(0x00010000);
640 dyndisk_header->block_size = be32_to_cpu(block_size);
641 dyndisk_header->max_table_entries = be32_to_cpu(num_bat_entries);
643 dyndisk_header->checksum = be32_to_cpu(vpc_checksum(buf, 1024));
645 // Write the header
646 if (lseek(fd, 512, SEEK_SET) < 0) {
647 goto fail;
650 if (write(fd, buf, 1024) != 1024) {
651 goto fail;
653 ret = 0;
655 fail:
656 close(fd);
657 return ret;
660 static void vpc_close(BlockDriverState *bs)
662 BDRVVPCState *s = bs->opaque;
663 g_free(s->pagetable);
664 #ifdef CACHE
665 g_free(s->pageentry_u8);
666 #endif
668 migrate_del_blocker(s->migration_blocker);
669 error_free(s->migration_blocker);
672 static QEMUOptionParameter vpc_create_options[] = {
674 .name = BLOCK_OPT_SIZE,
675 .type = OPT_SIZE,
676 .help = "Virtual disk size"
678 { NULL }
681 static BlockDriver bdrv_vpc = {
682 .format_name = "vpc",
683 .instance_size = sizeof(BDRVVPCState),
685 .bdrv_probe = vpc_probe,
686 .bdrv_open = vpc_open,
687 .bdrv_close = vpc_close,
688 .bdrv_create = vpc_create,
690 .bdrv_read = vpc_co_read,
691 .bdrv_write = vpc_co_write,
692 .bdrv_co_flush_to_disk = vpc_co_flush,
694 .create_options = vpc_create_options,
697 static void bdrv_vpc_init(void)
699 bdrv_register(&bdrv_vpc);
702 block_init(bdrv_vpc_init);