configure: Don't use shell arrays
[qemu-kvm/fedora.git] / block-vpc.c
blob71a171d0ebff5607dbdca3afd6e599986fbfd8fa
1 /*
2 * Block driver for Conectix/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"
28 /**************************************************************/
30 #define HEADER_SIZE 512
32 //#define CACHE
34 enum vhd_type {
35 VHD_FIXED = 2,
36 VHD_DYNAMIC = 3,
37 VHD_DIFFERENCING = 4,
40 // Seconds since Jan 1, 2000 0:00:00 (UTC)
41 #define VHD_TIMESTAMP_BASE 946684800
43 // always big-endian
44 struct vhd_footer {
45 char creator[8]; // "conectix"
46 uint32_t features;
47 uint32_t version;
49 // Offset of next header structure, 0xFFFFFFFF if none
50 uint64_t data_offset;
52 // Seconds since Jan 1, 2000 0:00:00 (UTC)
53 uint32_t timestamp;
55 char creator_app[4]; // "vpc "
56 uint16_t major;
57 uint16_t minor;
58 char creator_os[4]; // "Wi2k"
60 uint64_t orig_size;
61 uint64_t size;
63 uint16_t cyls;
64 uint8_t heads;
65 uint8_t secs_per_cyl;
67 uint32_t type;
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;
73 // UUID used to identify a parent hard disk (backing file)
74 uint8_t uuid[16];
76 uint8_t in_saved_state;
79 struct vhd_dyndisk_header {
80 char magic[8]; // "cxsparse"
82 // Offset of next header structure, 0xFFFFFFFF if none
83 uint64_t data_offset;
85 // Offset of the Block Allocation Table (BAT)
86 uint64_t table_offset;
88 uint32_t version;
89 uint32_t max_table_entries; // 32bit/entry
91 // 2 MB by default, must be a power of two
92 uint32_t block_size;
94 uint32_t checksum;
95 uint8_t parent_uuid[16];
96 uint32_t parent_timestamp;
97 uint32_t reserved;
99 // Backing file name (in UTF-16)
100 uint8_t parent_name[512];
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];
111 typedef struct BDRVVPCState {
112 BlockDriverState *hd;
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;
121 uint32_t block_size;
122 uint32_t bitmap_size;
124 #ifdef CACHE
125 uint8_t *pageentry_u8;
126 uint32_t *pageentry_u32;
127 uint16_t *pageentry_u16;
129 uint64_t last_bitmap;
130 #endif
131 } BDRVVPCState;
133 static uint32_t vpc_checksum(uint8_t* buf, size_t size)
135 uint32_t res = 0;
136 int i;
138 for (i = 0; i < size; i++)
139 res += buf[i];
141 return ~res;
145 static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename)
147 if (buf_size >= 8 && !strncmp((char *)buf, "conectix", 8))
148 return 100;
149 return 0;
152 static int vpc_open(BlockDriverState *bs, const char *filename, int flags)
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;
161 ret = bdrv_file_open(&s->hd, filename, flags);
162 if (ret < 0)
163 return ret;
165 if (bdrv_pread(s->hd, 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", 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 (bdrv_pread(s->hd, be64_to_cpu(footer->data_offset), buf, HEADER_SIZE)
185 != HEADER_SIZE)
186 goto fail;
188 dyndisk_header = (struct vhd_dyndisk_header*) buf;
190 if (strncmp(dyndisk_header->magic, "cxsparse", 8))
191 goto fail;
194 s->block_size = be32_to_cpu(dyndisk_header->block_size);
195 s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511;
197 s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries);
198 s->pagetable = qemu_malloc(s->max_table_entries * 4);
200 s->bat_offset = be64_to_cpu(dyndisk_header->table_offset);
201 if (bdrv_pread(s->hd, s->bat_offset, s->pagetable,
202 s->max_table_entries * 4) != s->max_table_entries * 4)
203 goto fail;
205 s->free_data_block_offset =
206 (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511;
208 for (i = 0; i < s->max_table_entries; i++) {
209 be32_to_cpus(&s->pagetable[i]);
210 if (s->pagetable[i] != 0xFFFFFFFF) {
211 int64_t next = (512 * (int64_t) s->pagetable[i]) +
212 s->bitmap_size + s->block_size;
214 if (next> s->free_data_block_offset)
215 s->free_data_block_offset = next;
219 s->last_bitmap_offset = (int64_t) -1;
221 #ifdef CACHE
222 s->pageentry_u8 = qemu_malloc(512);
223 s->pageentry_u32 = s->pageentry_u8;
224 s->pageentry_u16 = s->pageentry_u8;
225 s->last_pagetable = -1;
226 #endif
228 return 0;
229 fail:
230 bdrv_delete(s->hd);
231 return -1;
235 * Returns the absolute byte offset of the given sector in the image file.
236 * If the sector is not allocated, -1 is returned instead.
238 * The parameter write must be 1 if the offset will be used for a write
239 * operation (the block bitmaps is updated then), 0 otherwise.
241 static inline int64_t get_sector_offset(BlockDriverState *bs,
242 int64_t sector_num, int write)
244 BDRVVPCState *s = bs->opaque;
245 uint64_t offset = sector_num * 512;
246 uint64_t bitmap_offset, block_offset;
247 uint32_t pagetable_index, pageentry_index;
249 pagetable_index = offset / s->block_size;
250 pageentry_index = (offset % s->block_size) / 512;
252 if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff)
253 return -1; // not allocated
255 bitmap_offset = 512 * (uint64_t) s->pagetable[pagetable_index];
256 block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index);
258 // We must ensure that we don't write to any sectors which are marked as
259 // unused in the bitmap. We get away with setting all bits in the block
260 // bitmap each time we write to a new block. This might cause Virtual PC to
261 // miss sparse read optimization, but it's not a problem in terms of
262 // correctness.
263 if (write && (s->last_bitmap_offset != bitmap_offset)) {
264 uint8_t bitmap[s->bitmap_size];
266 s->last_bitmap_offset = bitmap_offset;
267 memset(bitmap, 0xff, s->bitmap_size);
268 bdrv_pwrite(s->hd, bitmap_offset, bitmap, s->bitmap_size);
271 // printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n",
272 // sector_num, pagetable_index, pageentry_index,
273 // bitmap_offset, block_offset);
275 // disabled by reason
276 #if 0
277 #ifdef CACHE
278 if (bitmap_offset != s->last_bitmap)
280 lseek(s->fd, bitmap_offset, SEEK_SET);
282 s->last_bitmap = bitmap_offset;
284 // Scary! Bitmap is stored as big endian 32bit entries,
285 // while we used to look it up byte by byte
286 read(s->fd, s->pageentry_u8, 512);
287 for (i = 0; i < 128; i++)
288 be32_to_cpus(&s->pageentry_u32[i]);
291 if ((s->pageentry_u8[pageentry_index / 8] >> (pageentry_index % 8)) & 1)
292 return -1;
293 #else
294 lseek(s->fd, bitmap_offset + (pageentry_index / 8), SEEK_SET);
296 read(s->fd, &bitmap_entry, 1);
298 if ((bitmap_entry >> (pageentry_index % 8)) & 1)
299 return -1; // not allocated
300 #endif
301 #endif
303 return block_offset;
307 * Writes the footer to the end of the image file. This is needed when the
308 * file grows as it overwrites the old footer
310 * Returns 0 on success and < 0 on error
312 static int rewrite_footer(BlockDriverState* bs)
314 int ret;
315 BDRVVPCState *s = bs->opaque;
316 int64_t offset = s->free_data_block_offset;
318 ret = bdrv_pwrite(s->hd, offset, s->footer_buf, HEADER_SIZE);
319 if (ret < 0)
320 return ret;
322 return 0;
326 * Allocates a new block. This involves writing a new footer and updating
327 * the Block Allocation Table to use the space at the old end of the image
328 * file (overwriting the old footer)
330 * Returns the sectors' offset in the image file on success and < 0 on error
332 static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num)
334 BDRVVPCState *s = bs->opaque;
335 int64_t bat_offset;
336 uint32_t index, bat_value;
337 int ret;
338 uint8_t bitmap[s->bitmap_size];
340 // Check if sector_num is valid
341 if ((sector_num < 0) || (sector_num > bs->total_sectors))
342 return -1;
344 // Write entry into in-memory BAT
345 index = (sector_num * 512) / s->block_size;
346 if (s->pagetable[index] != 0xFFFFFFFF)
347 return -1;
349 s->pagetable[index] = s->free_data_block_offset / 512;
351 // Initialize the block's bitmap
352 memset(bitmap, 0xff, s->bitmap_size);
353 bdrv_pwrite(s->hd, s->free_data_block_offset, bitmap, s->bitmap_size);
355 // Write new footer (the old one will be overwritten)
356 s->free_data_block_offset += s->block_size + s->bitmap_size;
357 ret = rewrite_footer(bs);
358 if (ret < 0)
359 goto fail;
361 // Write BAT entry to disk
362 bat_offset = s->bat_offset + (4 * index);
363 bat_value = be32_to_cpu(s->pagetable[index]);
364 ret = bdrv_pwrite(s->hd, bat_offset, &bat_value, 4);
365 if (ret < 0)
366 goto fail;
368 return get_sector_offset(bs, sector_num, 0);
370 fail:
371 s->free_data_block_offset -= (s->block_size + s->bitmap_size);
372 return -1;
375 static int vpc_read(BlockDriverState *bs, int64_t sector_num,
376 uint8_t *buf, int nb_sectors)
378 BDRVVPCState *s = bs->opaque;
379 int ret;
380 int64_t offset;
382 while (nb_sectors > 0) {
383 offset = get_sector_offset(bs, sector_num, 0);
385 if (offset == -1) {
386 memset(buf, 0, 512);
387 } else {
388 ret = bdrv_pread(s->hd, offset, buf, 512);
389 if (ret != 512)
390 return -1;
393 nb_sectors--;
394 sector_num++;
395 buf += 512;
397 return 0;
400 static int vpc_write(BlockDriverState *bs, int64_t sector_num,
401 const uint8_t *buf, int nb_sectors)
403 BDRVVPCState *s = bs->opaque;
404 int64_t offset;
405 int ret;
407 while (nb_sectors > 0) {
408 offset = get_sector_offset(bs, sector_num, 1);
410 if (offset == -1) {
411 offset = alloc_block(bs, sector_num);
412 if (offset < 0)
413 return -1;
416 ret = bdrv_pwrite(s->hd, offset, buf, 512);
417 if (ret != 512)
418 return -1;
420 nb_sectors--;
421 sector_num++;
422 buf += 512;
425 return 0;
430 * Calculates the number of cylinders, heads and sectors per cylinder
431 * based on a given number of sectors. This is the algorithm described
432 * in the VHD specification.
434 * Note that the geometry doesn't always exactly match total_sectors but
435 * may round it down.
437 * Returns 0 on success, -EFBIG if the size is larger than 127 GB
439 static int calculate_geometry(int64_t total_sectors, uint16_t* cyls,
440 uint8_t* heads, uint8_t* secs_per_cyl)
442 uint32_t cyls_times_heads;
444 if (total_sectors > 65535 * 16 * 255)
445 return -EFBIG;
447 if (total_sectors > 65535 * 16 * 63) {
448 *secs_per_cyl = 255;
449 *heads = 16;
450 cyls_times_heads = total_sectors / *secs_per_cyl;
451 } else {
452 *secs_per_cyl = 17;
453 cyls_times_heads = total_sectors / *secs_per_cyl;
454 *heads = (cyls_times_heads + 1023) / 1024;
456 if (*heads < 4)
457 *heads = 4;
459 if (cyls_times_heads >= (*heads * 1024) || *heads > 16) {
460 *secs_per_cyl = 31;
461 *heads = 16;
462 cyls_times_heads = total_sectors / *secs_per_cyl;
465 if (cyls_times_heads >= (*heads * 1024)) {
466 *secs_per_cyl = 63;
467 *heads = 16;
468 cyls_times_heads = total_sectors / *secs_per_cyl;
472 // Note: Rounding up deviates from the Virtual PC behaviour
473 // However, we need this to avoid truncating images in qemu-img convert
474 *cyls = (cyls_times_heads + *heads - 1) / *heads;
476 return 0;
479 static int vpc_create(const char *filename, int64_t total_sectors,
480 const char *backing_file, int flags)
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;
492 if (backing_file != NULL)
493 return -ENOTSUP;
495 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644);
496 if (fd < 0)
497 return -EIO;
499 // Calculate matching total_size and geometry
500 if (calculate_geometry(total_sectors, &cyls, &heads, &secs_per_cyl))
501 return -EFBIG;
502 total_sectors = (int64_t) cyls * heads * secs_per_cyl;
504 // Prepare the Hard Disk Footer
505 memset(buf, 0, 1024);
507 strncpy(footer->creator, "conectix", 8);
508 // TODO Check if "qemu" creator_app is ok for VPC
509 strncpy(footer->creator_app, "qemu", 4);
510 strncpy(footer->creator_os, "Wi2k", 4);
512 footer->features = be32_to_cpu(0x02);
513 footer->version = be32_to_cpu(0x00010000);
514 footer->data_offset = be64_to_cpu(HEADER_SIZE);
515 footer->timestamp = be32_to_cpu(time(NULL) - VHD_TIMESTAMP_BASE);
517 // Version of Virtual PC 2007
518 footer->major = be16_to_cpu(0x0005);
519 footer->minor =be16_to_cpu(0x0003);
521 footer->orig_size = be64_to_cpu(total_sectors * 512);
522 footer->size = be64_to_cpu(total_sectors * 512);
524 footer->cyls = be16_to_cpu(cyls);
525 footer->heads = heads;
526 footer->secs_per_cyl = secs_per_cyl;
528 footer->type = be32_to_cpu(VHD_DYNAMIC);
530 // TODO uuid is missing
532 footer->checksum = be32_to_cpu(vpc_checksum(buf, HEADER_SIZE));
534 // Write the footer (twice: at the beginning and at the end)
535 block_size = 0x200000;
536 num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512);
538 if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE)
539 return -EIO;
541 if (lseek(fd, 1536 + ((num_bat_entries * 4 + 511) & ~511), SEEK_SET) < 0)
542 return -EIO;
543 if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE)
544 return -EIO;
546 // Write the initial BAT
547 if (lseek(fd, 3 * 512, SEEK_SET) < 0)
548 return -EIO;
550 memset(buf, 0xFF, 512);
551 for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++)
552 if (write(fd, buf, 512) != 512)
553 return -EIO;
556 // Prepare the Dynamic Disk Header
557 memset(buf, 0, 1024);
559 strncpy(dyndisk_header->magic, "cxsparse", 8);
561 dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFF);
562 dyndisk_header->table_offset = be64_to_cpu(3 * 512);
563 dyndisk_header->version = be32_to_cpu(0x00010000);
564 dyndisk_header->block_size = be32_to_cpu(block_size);
565 dyndisk_header->max_table_entries = be32_to_cpu(num_bat_entries);
567 dyndisk_header->checksum = be32_to_cpu(vpc_checksum(buf, 1024));
569 // Write the header
570 if (lseek(fd, 512, SEEK_SET) < 0)
571 return -EIO;
572 if (write(fd, buf, 1024) != 1024)
573 return -EIO;
575 close(fd);
576 return 0;
579 static void vpc_close(BlockDriverState *bs)
581 BDRVVPCState *s = bs->opaque;
582 qemu_free(s->pagetable);
583 #ifdef CACHE
584 qemu_free(s->pageentry_u8);
585 #endif
586 bdrv_delete(s->hd);
589 BlockDriver bdrv_vpc = {
590 .format_name = "vpc",
591 .instance_size = sizeof(BDRVVPCState),
592 .bdrv_probe = vpc_probe,
593 .bdrv_open = vpc_open,
594 .bdrv_read = vpc_read,
595 .bdrv_write = vpc_write,
596 .bdrv_close = vpc_close,
597 .bdrv_create = vpc_create,