]>
Commit | Line | Data |
---|---|---|
45aba42f KW |
1 | /* |
2 | * Block driver for the QCOW version 2 format | |
3 | * | |
4 | * Copyright (c) 2004-2006 Fabrice Bellard | |
5 | * | |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
24 | ||
25 | #include <zlib.h> | |
26 | ||
27 | #include "qemu-common.h" | |
28 | #include "block_int.h" | |
29 | #include "block/qcow2.h" | |
30 | ||
72893756 | 31 | int qcow2_grow_l1_table(BlockDriverState *bs, int min_size, bool exact_size) |
45aba42f KW |
32 | { |
33 | BDRVQcowState *s = bs->opaque; | |
34 | int new_l1_size, new_l1_size2, ret, i; | |
35 | uint64_t *new_l1_table; | |
5d757b56 | 36 | int64_t new_l1_table_offset; |
45aba42f KW |
37 | uint8_t data[12]; |
38 | ||
72893756 | 39 | if (min_size <= s->l1_size) |
45aba42f | 40 | return 0; |
72893756 SH |
41 | |
42 | if (exact_size) { | |
43 | new_l1_size = min_size; | |
44 | } else { | |
45 | /* Bump size up to reduce the number of times we have to grow */ | |
46 | new_l1_size = s->l1_size; | |
47 | if (new_l1_size == 0) { | |
48 | new_l1_size = 1; | |
49 | } | |
50 | while (min_size > new_l1_size) { | |
51 | new_l1_size = (new_l1_size * 3 + 1) / 2; | |
52 | } | |
45aba42f | 53 | } |
72893756 | 54 | |
45aba42f KW |
55 | #ifdef DEBUG_ALLOC2 |
56 | printf("grow l1_table from %d to %d\n", s->l1_size, new_l1_size); | |
57 | #endif | |
58 | ||
59 | new_l1_size2 = sizeof(uint64_t) * new_l1_size; | |
3f6a3ee5 | 60 | new_l1_table = qemu_mallocz(align_offset(new_l1_size2, 512)); |
45aba42f KW |
61 | memcpy(new_l1_table, s->l1_table, s->l1_size * sizeof(uint64_t)); |
62 | ||
63 | /* write new table (align to cluster) */ | |
66f82cee | 64 | BLKDBG_EVENT(bs->file, BLKDBG_L1_GROW_ALLOC_TABLE); |
ed6ccf0f | 65 | new_l1_table_offset = qcow2_alloc_clusters(bs, new_l1_size2); |
5d757b56 KW |
66 | if (new_l1_table_offset < 0) { |
67 | qemu_free(new_l1_table); | |
68 | return new_l1_table_offset; | |
69 | } | |
29c1a730 KW |
70 | |
71 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
72 | if (ret < 0) { | |
73 | return ret; | |
74 | } | |
45aba42f | 75 | |
66f82cee | 76 | BLKDBG_EVENT(bs->file, BLKDBG_L1_GROW_WRITE_TABLE); |
45aba42f KW |
77 | for(i = 0; i < s->l1_size; i++) |
78 | new_l1_table[i] = cpu_to_be64(new_l1_table[i]); | |
8b3b7206 KW |
79 | ret = bdrv_pwrite_sync(bs->file, new_l1_table_offset, new_l1_table, new_l1_size2); |
80 | if (ret < 0) | |
45aba42f KW |
81 | goto fail; |
82 | for(i = 0; i < s->l1_size; i++) | |
83 | new_l1_table[i] = be64_to_cpu(new_l1_table[i]); | |
84 | ||
85 | /* set new table */ | |
66f82cee | 86 | BLKDBG_EVENT(bs->file, BLKDBG_L1_GROW_ACTIVATE_TABLE); |
45aba42f | 87 | cpu_to_be32w((uint32_t*)data, new_l1_size); |
653df36b | 88 | cpu_to_be64wu((uint64_t*)(data + 4), new_l1_table_offset); |
8b3b7206 KW |
89 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, l1_size), data,sizeof(data)); |
90 | if (ret < 0) { | |
45aba42f | 91 | goto fail; |
fb8fa77c | 92 | } |
45aba42f | 93 | qemu_free(s->l1_table); |
ed6ccf0f | 94 | qcow2_free_clusters(bs, s->l1_table_offset, s->l1_size * sizeof(uint64_t)); |
45aba42f KW |
95 | s->l1_table_offset = new_l1_table_offset; |
96 | s->l1_table = new_l1_table; | |
97 | s->l1_size = new_l1_size; | |
98 | return 0; | |
99 | fail: | |
fb8fa77c KW |
100 | qemu_free(new_l1_table); |
101 | qcow2_free_clusters(bs, new_l1_table_offset, new_l1_size2); | |
8b3b7206 | 102 | return ret; |
45aba42f KW |
103 | } |
104 | ||
45aba42f KW |
105 | /* |
106 | * l2_load | |
107 | * | |
108 | * Loads a L2 table into memory. If the table is in the cache, the cache | |
109 | * is used; otherwise the L2 table is loaded from the image file. | |
110 | * | |
111 | * Returns a pointer to the L2 table on success, or NULL if the read from | |
112 | * the image file failed. | |
113 | */ | |
114 | ||
55c17e98 KW |
115 | static int l2_load(BlockDriverState *bs, uint64_t l2_offset, |
116 | uint64_t **l2_table) | |
45aba42f KW |
117 | { |
118 | BDRVQcowState *s = bs->opaque; | |
55c17e98 | 119 | int ret; |
45aba42f | 120 | |
29c1a730 | 121 | ret = qcow2_cache_get(bs, s->l2_table_cache, l2_offset, (void**) l2_table); |
45aba42f | 122 | |
29c1a730 | 123 | return ret; |
45aba42f KW |
124 | } |
125 | ||
6583e3c7 KW |
126 | /* |
127 | * Writes one sector of the L1 table to the disk (can't update single entries | |
128 | * and we really don't want bdrv_pread to perform a read-modify-write) | |
129 | */ | |
130 | #define L1_ENTRIES_PER_SECTOR (512 / 8) | |
66f82cee | 131 | static int write_l1_entry(BlockDriverState *bs, int l1_index) |
6583e3c7 | 132 | { |
66f82cee | 133 | BDRVQcowState *s = bs->opaque; |
6583e3c7 KW |
134 | uint64_t buf[L1_ENTRIES_PER_SECTOR]; |
135 | int l1_start_index; | |
f7defcb6 | 136 | int i, ret; |
6583e3c7 KW |
137 | |
138 | l1_start_index = l1_index & ~(L1_ENTRIES_PER_SECTOR - 1); | |
139 | for (i = 0; i < L1_ENTRIES_PER_SECTOR; i++) { | |
140 | buf[i] = cpu_to_be64(s->l1_table[l1_start_index + i]); | |
141 | } | |
142 | ||
66f82cee | 143 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); |
8b3b7206 | 144 | ret = bdrv_pwrite_sync(bs->file, s->l1_table_offset + 8 * l1_start_index, |
f7defcb6 KW |
145 | buf, sizeof(buf)); |
146 | if (ret < 0) { | |
147 | return ret; | |
6583e3c7 KW |
148 | } |
149 | ||
150 | return 0; | |
151 | } | |
152 | ||
45aba42f KW |
153 | /* |
154 | * l2_allocate | |
155 | * | |
156 | * Allocate a new l2 entry in the file. If l1_index points to an already | |
157 | * used entry in the L2 table (i.e. we are doing a copy on write for the L2 | |
158 | * table) copy the contents of the old L2 table into the newly allocated one. | |
159 | * Otherwise the new table is initialized with zeros. | |
160 | * | |
161 | */ | |
162 | ||
c46e1167 | 163 | static int l2_allocate(BlockDriverState *bs, int l1_index, uint64_t **table) |
45aba42f KW |
164 | { |
165 | BDRVQcowState *s = bs->opaque; | |
6583e3c7 | 166 | uint64_t old_l2_offset; |
f4f0d391 KW |
167 | uint64_t *l2_table; |
168 | int64_t l2_offset; | |
c46e1167 | 169 | int ret; |
45aba42f KW |
170 | |
171 | old_l2_offset = s->l1_table[l1_index]; | |
172 | ||
173 | /* allocate a new l2 entry */ | |
174 | ||
ed6ccf0f | 175 | l2_offset = qcow2_alloc_clusters(bs, s->l2_size * sizeof(uint64_t)); |
5d757b56 | 176 | if (l2_offset < 0) { |
c46e1167 | 177 | return l2_offset; |
5d757b56 | 178 | } |
29c1a730 KW |
179 | |
180 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
181 | if (ret < 0) { | |
182 | goto fail; | |
183 | } | |
45aba42f | 184 | |
45aba42f KW |
185 | /* allocate a new entry in the l2 cache */ |
186 | ||
29c1a730 KW |
187 | ret = qcow2_cache_get_empty(bs, s->l2_table_cache, l2_offset, (void**) table); |
188 | if (ret < 0) { | |
189 | return ret; | |
190 | } | |
191 | ||
192 | l2_table = *table; | |
45aba42f KW |
193 | |
194 | if (old_l2_offset == 0) { | |
195 | /* if there was no old l2 table, clear the new table */ | |
196 | memset(l2_table, 0, s->l2_size * sizeof(uint64_t)); | |
197 | } else { | |
29c1a730 KW |
198 | uint64_t* old_table; |
199 | ||
45aba42f | 200 | /* if there was an old l2 table, read it from the disk */ |
66f82cee | 201 | BLKDBG_EVENT(bs->file, BLKDBG_L2_ALLOC_COW_READ); |
29c1a730 KW |
202 | ret = qcow2_cache_get(bs, s->l2_table_cache, old_l2_offset, |
203 | (void**) &old_table); | |
204 | if (ret < 0) { | |
205 | goto fail; | |
206 | } | |
207 | ||
208 | memcpy(l2_table, old_table, s->cluster_size); | |
209 | ||
210 | ret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &old_table); | |
c46e1167 | 211 | if (ret < 0) { |
175e1152 | 212 | goto fail; |
c46e1167 | 213 | } |
45aba42f | 214 | } |
29c1a730 | 215 | |
45aba42f | 216 | /* write the l2 table to the file */ |
66f82cee | 217 | BLKDBG_EVENT(bs->file, BLKDBG_L2_ALLOC_WRITE); |
29c1a730 KW |
218 | |
219 | qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_table); | |
220 | ret = qcow2_cache_flush(bs, s->l2_table_cache); | |
c46e1167 | 221 | if (ret < 0) { |
175e1152 KW |
222 | goto fail; |
223 | } | |
224 | ||
225 | /* update the L1 entry */ | |
226 | s->l1_table[l1_index] = l2_offset | QCOW_OFLAG_COPIED; | |
227 | ret = write_l1_entry(bs, l1_index); | |
228 | if (ret < 0) { | |
229 | goto fail; | |
c46e1167 | 230 | } |
45aba42f | 231 | |
c46e1167 KW |
232 | *table = l2_table; |
233 | return 0; | |
175e1152 KW |
234 | |
235 | fail: | |
29c1a730 | 236 | qcow2_cache_put(bs, s->l2_table_cache, (void**) table); |
68dba0bf | 237 | s->l1_table[l1_index] = old_l2_offset; |
175e1152 | 238 | return ret; |
45aba42f KW |
239 | } |
240 | ||
241 | static int count_contiguous_clusters(uint64_t nb_clusters, int cluster_size, | |
242 | uint64_t *l2_table, uint64_t start, uint64_t mask) | |
243 | { | |
244 | int i; | |
245 | uint64_t offset = be64_to_cpu(l2_table[0]) & ~mask; | |
246 | ||
247 | if (!offset) | |
248 | return 0; | |
249 | ||
250 | for (i = start; i < start + nb_clusters; i++) | |
80ee15a6 | 251 | if (offset + (uint64_t) i * cluster_size != (be64_to_cpu(l2_table[i]) & ~mask)) |
45aba42f KW |
252 | break; |
253 | ||
254 | return (i - start); | |
255 | } | |
256 | ||
257 | static int count_contiguous_free_clusters(uint64_t nb_clusters, uint64_t *l2_table) | |
258 | { | |
259 | int i = 0; | |
260 | ||
261 | while(nb_clusters-- && l2_table[i] == 0) | |
262 | i++; | |
263 | ||
264 | return i; | |
265 | } | |
266 | ||
267 | /* The crypt function is compatible with the linux cryptoloop | |
268 | algorithm for < 4 GB images. NOTE: out_buf == in_buf is | |
269 | supported */ | |
ed6ccf0f KW |
270 | void qcow2_encrypt_sectors(BDRVQcowState *s, int64_t sector_num, |
271 | uint8_t *out_buf, const uint8_t *in_buf, | |
272 | int nb_sectors, int enc, | |
273 | const AES_KEY *key) | |
45aba42f KW |
274 | { |
275 | union { | |
276 | uint64_t ll[2]; | |
277 | uint8_t b[16]; | |
278 | } ivec; | |
279 | int i; | |
280 | ||
281 | for(i = 0; i < nb_sectors; i++) { | |
282 | ivec.ll[0] = cpu_to_le64(sector_num); | |
283 | ivec.ll[1] = 0; | |
284 | AES_cbc_encrypt(in_buf, out_buf, 512, key, | |
285 | ivec.b, enc); | |
286 | sector_num++; | |
287 | in_buf += 512; | |
288 | out_buf += 512; | |
289 | } | |
290 | } | |
291 | ||
292 | ||
7c80ab3f JS |
293 | static int qcow2_read(BlockDriverState *bs, int64_t sector_num, |
294 | uint8_t *buf, int nb_sectors) | |
45aba42f KW |
295 | { |
296 | BDRVQcowState *s = bs->opaque; | |
297 | int ret, index_in_cluster, n, n1; | |
298 | uint64_t cluster_offset; | |
bd28f835 KW |
299 | struct iovec iov; |
300 | QEMUIOVector qiov; | |
45aba42f KW |
301 | |
302 | while (nb_sectors > 0) { | |
303 | n = nb_sectors; | |
1c46efaa KW |
304 | |
305 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, &n, | |
306 | &cluster_offset); | |
307 | if (ret < 0) { | |
308 | return ret; | |
309 | } | |
310 | ||
45aba42f KW |
311 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
312 | if (!cluster_offset) { | |
313 | if (bs->backing_hd) { | |
314 | /* read from the base image */ | |
bd28f835 KW |
315 | iov.iov_base = buf; |
316 | iov.iov_len = n * 512; | |
317 | qemu_iovec_init_external(&qiov, &iov, 1); | |
318 | ||
319 | n1 = qcow2_backing_read1(bs->backing_hd, &qiov, sector_num, n); | |
45aba42f | 320 | if (n1 > 0) { |
66f82cee | 321 | BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING); |
45aba42f KW |
322 | ret = bdrv_read(bs->backing_hd, sector_num, buf, n1); |
323 | if (ret < 0) | |
324 | return -1; | |
325 | } | |
326 | } else { | |
327 | memset(buf, 0, 512 * n); | |
328 | } | |
329 | } else if (cluster_offset & QCOW_OFLAG_COMPRESSED) { | |
66f82cee | 330 | if (qcow2_decompress_cluster(bs, cluster_offset) < 0) |
45aba42f KW |
331 | return -1; |
332 | memcpy(buf, s->cluster_cache + index_in_cluster * 512, 512 * n); | |
333 | } else { | |
66f82cee KW |
334 | BLKDBG_EVENT(bs->file, BLKDBG_READ); |
335 | ret = bdrv_pread(bs->file, cluster_offset + index_in_cluster * 512, buf, n * 512); | |
45aba42f KW |
336 | if (ret != n * 512) |
337 | return -1; | |
338 | if (s->crypt_method) { | |
ed6ccf0f | 339 | qcow2_encrypt_sectors(s, sector_num, buf, buf, n, 0, |
45aba42f KW |
340 | &s->aes_decrypt_key); |
341 | } | |
342 | } | |
343 | nb_sectors -= n; | |
344 | sector_num += n; | |
345 | buf += n * 512; | |
346 | } | |
347 | return 0; | |
348 | } | |
349 | ||
350 | static int copy_sectors(BlockDriverState *bs, uint64_t start_sect, | |
351 | uint64_t cluster_offset, int n_start, int n_end) | |
352 | { | |
353 | BDRVQcowState *s = bs->opaque; | |
354 | int n, ret; | |
355 | ||
356 | n = n_end - n_start; | |
357 | if (n <= 0) | |
358 | return 0; | |
66f82cee | 359 | BLKDBG_EVENT(bs->file, BLKDBG_COW_READ); |
7c80ab3f | 360 | ret = qcow2_read(bs, start_sect + n_start, s->cluster_data, n); |
45aba42f KW |
361 | if (ret < 0) |
362 | return ret; | |
363 | if (s->crypt_method) { | |
ed6ccf0f | 364 | qcow2_encrypt_sectors(s, start_sect + n_start, |
45aba42f KW |
365 | s->cluster_data, |
366 | s->cluster_data, n, 1, | |
367 | &s->aes_encrypt_key); | |
368 | } | |
66f82cee | 369 | BLKDBG_EVENT(bs->file, BLKDBG_COW_WRITE); |
9f8e668e | 370 | ret = bdrv_write(bs->file, (cluster_offset >> 9) + n_start, |
8b3b7206 | 371 | s->cluster_data, n); |
45aba42f KW |
372 | if (ret < 0) |
373 | return ret; | |
374 | return 0; | |
375 | } | |
376 | ||
377 | ||
378 | /* | |
379 | * get_cluster_offset | |
380 | * | |
1c46efaa KW |
381 | * For a given offset of the disk image, find the cluster offset in |
382 | * qcow2 file. The offset is stored in *cluster_offset. | |
45aba42f KW |
383 | * |
384 | * on entry, *num is the number of contiguous clusters we'd like to | |
385 | * access following offset. | |
386 | * | |
387 | * on exit, *num is the number of contiguous clusters we can read. | |
388 | * | |
1c46efaa KW |
389 | * Return 0, if the offset is found |
390 | * Return -errno, otherwise. | |
45aba42f KW |
391 | * |
392 | */ | |
393 | ||
1c46efaa KW |
394 | int qcow2_get_cluster_offset(BlockDriverState *bs, uint64_t offset, |
395 | int *num, uint64_t *cluster_offset) | |
45aba42f KW |
396 | { |
397 | BDRVQcowState *s = bs->opaque; | |
80ee15a6 | 398 | unsigned int l1_index, l2_index; |
1c46efaa | 399 | uint64_t l2_offset, *l2_table; |
45aba42f | 400 | int l1_bits, c; |
80ee15a6 KW |
401 | unsigned int index_in_cluster, nb_clusters; |
402 | uint64_t nb_available, nb_needed; | |
55c17e98 | 403 | int ret; |
45aba42f KW |
404 | |
405 | index_in_cluster = (offset >> 9) & (s->cluster_sectors - 1); | |
406 | nb_needed = *num + index_in_cluster; | |
407 | ||
408 | l1_bits = s->l2_bits + s->cluster_bits; | |
409 | ||
410 | /* compute how many bytes there are between the offset and | |
411 | * the end of the l1 entry | |
412 | */ | |
413 | ||
80ee15a6 | 414 | nb_available = (1ULL << l1_bits) - (offset & ((1ULL << l1_bits) - 1)); |
45aba42f KW |
415 | |
416 | /* compute the number of available sectors */ | |
417 | ||
418 | nb_available = (nb_available >> 9) + index_in_cluster; | |
419 | ||
420 | if (nb_needed > nb_available) { | |
421 | nb_needed = nb_available; | |
422 | } | |
423 | ||
1c46efaa | 424 | *cluster_offset = 0; |
45aba42f KW |
425 | |
426 | /* seek the the l2 offset in the l1 table */ | |
427 | ||
428 | l1_index = offset >> l1_bits; | |
429 | if (l1_index >= s->l1_size) | |
430 | goto out; | |
431 | ||
432 | l2_offset = s->l1_table[l1_index]; | |
433 | ||
434 | /* seek the l2 table of the given l2 offset */ | |
435 | ||
436 | if (!l2_offset) | |
437 | goto out; | |
438 | ||
439 | /* load the l2 table in memory */ | |
440 | ||
441 | l2_offset &= ~QCOW_OFLAG_COPIED; | |
55c17e98 KW |
442 | ret = l2_load(bs, l2_offset, &l2_table); |
443 | if (ret < 0) { | |
444 | return ret; | |
1c46efaa | 445 | } |
45aba42f KW |
446 | |
447 | /* find the cluster offset for the given disk offset */ | |
448 | ||
449 | l2_index = (offset >> s->cluster_bits) & (s->l2_size - 1); | |
1c46efaa | 450 | *cluster_offset = be64_to_cpu(l2_table[l2_index]); |
45aba42f KW |
451 | nb_clusters = size_to_clusters(s, nb_needed << 9); |
452 | ||
1c46efaa | 453 | if (!*cluster_offset) { |
45aba42f KW |
454 | /* how many empty clusters ? */ |
455 | c = count_contiguous_free_clusters(nb_clusters, &l2_table[l2_index]); | |
456 | } else { | |
457 | /* how many allocated clusters ? */ | |
458 | c = count_contiguous_clusters(nb_clusters, s->cluster_size, | |
459 | &l2_table[l2_index], 0, QCOW_OFLAG_COPIED); | |
460 | } | |
461 | ||
29c1a730 KW |
462 | qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table); |
463 | ||
45aba42f KW |
464 | nb_available = (c * s->cluster_sectors); |
465 | out: | |
466 | if (nb_available > nb_needed) | |
467 | nb_available = nb_needed; | |
468 | ||
469 | *num = nb_available - index_in_cluster; | |
470 | ||
1c46efaa KW |
471 | *cluster_offset &=~QCOW_OFLAG_COPIED; |
472 | return 0; | |
45aba42f KW |
473 | } |
474 | ||
475 | /* | |
476 | * get_cluster_table | |
477 | * | |
478 | * for a given disk offset, load (and allocate if needed) | |
479 | * the l2 table. | |
480 | * | |
481 | * the l2 table offset in the qcow2 file and the cluster index | |
482 | * in the l2 table are given to the caller. | |
483 | * | |
1e3e8f1a | 484 | * Returns 0 on success, -errno in failure case |
45aba42f | 485 | */ |
45aba42f KW |
486 | static int get_cluster_table(BlockDriverState *bs, uint64_t offset, |
487 | uint64_t **new_l2_table, | |
488 | uint64_t *new_l2_offset, | |
489 | int *new_l2_index) | |
490 | { | |
491 | BDRVQcowState *s = bs->opaque; | |
80ee15a6 | 492 | unsigned int l1_index, l2_index; |
c46e1167 KW |
493 | uint64_t l2_offset; |
494 | uint64_t *l2_table = NULL; | |
80ee15a6 | 495 | int ret; |
45aba42f KW |
496 | |
497 | /* seek the the l2 offset in the l1 table */ | |
498 | ||
499 | l1_index = offset >> (s->l2_bits + s->cluster_bits); | |
500 | if (l1_index >= s->l1_size) { | |
72893756 | 501 | ret = qcow2_grow_l1_table(bs, l1_index + 1, false); |
1e3e8f1a KW |
502 | if (ret < 0) { |
503 | return ret; | |
504 | } | |
45aba42f KW |
505 | } |
506 | l2_offset = s->l1_table[l1_index]; | |
507 | ||
508 | /* seek the l2 table of the given l2 offset */ | |
509 | ||
510 | if (l2_offset & QCOW_OFLAG_COPIED) { | |
511 | /* load the l2 table in memory */ | |
512 | l2_offset &= ~QCOW_OFLAG_COPIED; | |
55c17e98 KW |
513 | ret = l2_load(bs, l2_offset, &l2_table); |
514 | if (ret < 0) { | |
515 | return ret; | |
1e3e8f1a | 516 | } |
45aba42f | 517 | } else { |
29c1a730 | 518 | /* FIXME Order */ |
45aba42f | 519 | if (l2_offset) |
ed6ccf0f | 520 | qcow2_free_clusters(bs, l2_offset, s->l2_size * sizeof(uint64_t)); |
c46e1167 KW |
521 | ret = l2_allocate(bs, l1_index, &l2_table); |
522 | if (ret < 0) { | |
523 | return ret; | |
1e3e8f1a | 524 | } |
45aba42f KW |
525 | l2_offset = s->l1_table[l1_index] & ~QCOW_OFLAG_COPIED; |
526 | } | |
527 | ||
528 | /* find the cluster offset for the given disk offset */ | |
529 | ||
530 | l2_index = (offset >> s->cluster_bits) & (s->l2_size - 1); | |
531 | ||
532 | *new_l2_table = l2_table; | |
533 | *new_l2_offset = l2_offset; | |
534 | *new_l2_index = l2_index; | |
535 | ||
1e3e8f1a | 536 | return 0; |
45aba42f KW |
537 | } |
538 | ||
539 | /* | |
540 | * alloc_compressed_cluster_offset | |
541 | * | |
542 | * For a given offset of the disk image, return cluster offset in | |
543 | * qcow2 file. | |
544 | * | |
545 | * If the offset is not found, allocate a new compressed cluster. | |
546 | * | |
547 | * Return the cluster offset if successful, | |
548 | * Return 0, otherwise. | |
549 | * | |
550 | */ | |
551 | ||
ed6ccf0f KW |
552 | uint64_t qcow2_alloc_compressed_cluster_offset(BlockDriverState *bs, |
553 | uint64_t offset, | |
554 | int compressed_size) | |
45aba42f KW |
555 | { |
556 | BDRVQcowState *s = bs->opaque; | |
557 | int l2_index, ret; | |
f4f0d391 KW |
558 | uint64_t l2_offset, *l2_table; |
559 | int64_t cluster_offset; | |
45aba42f KW |
560 | int nb_csectors; |
561 | ||
562 | ret = get_cluster_table(bs, offset, &l2_table, &l2_offset, &l2_index); | |
1e3e8f1a | 563 | if (ret < 0) { |
45aba42f | 564 | return 0; |
1e3e8f1a | 565 | } |
45aba42f KW |
566 | |
567 | cluster_offset = be64_to_cpu(l2_table[l2_index]); | |
568 | if (cluster_offset & QCOW_OFLAG_COPIED) | |
569 | return cluster_offset & ~QCOW_OFLAG_COPIED; | |
570 | ||
571 | if (cluster_offset) | |
ed6ccf0f | 572 | qcow2_free_any_clusters(bs, cluster_offset, 1); |
45aba42f | 573 | |
ed6ccf0f | 574 | cluster_offset = qcow2_alloc_bytes(bs, compressed_size); |
5d757b56 | 575 | if (cluster_offset < 0) { |
29c1a730 | 576 | qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table); |
5d757b56 KW |
577 | return 0; |
578 | } | |
579 | ||
45aba42f KW |
580 | nb_csectors = ((cluster_offset + compressed_size - 1) >> 9) - |
581 | (cluster_offset >> 9); | |
582 | ||
583 | cluster_offset |= QCOW_OFLAG_COMPRESSED | | |
584 | ((uint64_t)nb_csectors << s->csize_shift); | |
585 | ||
586 | /* update L2 table */ | |
587 | ||
588 | /* compressed clusters never have the copied flag */ | |
589 | ||
66f82cee | 590 | BLKDBG_EVENT(bs->file, BLKDBG_L2_UPDATE_COMPRESSED); |
29c1a730 | 591 | qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_table); |
45aba42f | 592 | l2_table[l2_index] = cpu_to_be64(cluster_offset); |
29c1a730 | 593 | ret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table); |
79a31189 | 594 | if (ret < 0) { |
29c1a730 | 595 | return 0; |
4c1612d9 KW |
596 | } |
597 | ||
29c1a730 | 598 | return cluster_offset; |
4c1612d9 KW |
599 | } |
600 | ||
148da7ea | 601 | int qcow2_alloc_cluster_link_l2(BlockDriverState *bs, QCowL2Meta *m) |
45aba42f KW |
602 | { |
603 | BDRVQcowState *s = bs->opaque; | |
604 | int i, j = 0, l2_index, ret; | |
605 | uint64_t *old_cluster, start_sect, l2_offset, *l2_table; | |
148da7ea | 606 | uint64_t cluster_offset = m->cluster_offset; |
29c1a730 | 607 | bool cow = false; |
45aba42f KW |
608 | |
609 | if (m->nb_clusters == 0) | |
610 | return 0; | |
611 | ||
612 | old_cluster = qemu_malloc(m->nb_clusters * sizeof(uint64_t)); | |
613 | ||
614 | /* copy content of unmodified sectors */ | |
615 | start_sect = (m->offset & ~(s->cluster_size - 1)) >> 9; | |
616 | if (m->n_start) { | |
29c1a730 | 617 | cow = true; |
45aba42f KW |
618 | ret = copy_sectors(bs, start_sect, cluster_offset, 0, m->n_start); |
619 | if (ret < 0) | |
620 | goto err; | |
621 | } | |
622 | ||
623 | if (m->nb_available & (s->cluster_sectors - 1)) { | |
624 | uint64_t end = m->nb_available & ~(uint64_t)(s->cluster_sectors - 1); | |
29c1a730 | 625 | cow = true; |
45aba42f KW |
626 | ret = copy_sectors(bs, start_sect + end, cluster_offset + (end << 9), |
627 | m->nb_available - end, s->cluster_sectors); | |
628 | if (ret < 0) | |
629 | goto err; | |
630 | } | |
631 | ||
29c1a730 KW |
632 | /* |
633 | * Update L2 table. | |
634 | * | |
635 | * Before we update the L2 table to actually point to the new cluster, we | |
636 | * need to be sure that the refcounts have been increased and COW was | |
637 | * handled. | |
638 | */ | |
639 | if (cow) { | |
640 | bdrv_flush(bs->file); | |
641 | } | |
642 | ||
643 | qcow2_cache_set_dependency(bs, s->l2_table_cache, s->refcount_block_cache); | |
1e3e8f1a KW |
644 | ret = get_cluster_table(bs, m->offset, &l2_table, &l2_offset, &l2_index); |
645 | if (ret < 0) { | |
45aba42f | 646 | goto err; |
1e3e8f1a | 647 | } |
29c1a730 | 648 | qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_table); |
45aba42f KW |
649 | |
650 | for (i = 0; i < m->nb_clusters; i++) { | |
651 | /* if two concurrent writes happen to the same unallocated cluster | |
652 | * each write allocates separate cluster and writes data concurrently. | |
653 | * The first one to complete updates l2 table with pointer to its | |
654 | * cluster the second one has to do RMW (which is done above by | |
655 | * copy_sectors()), update l2 table with its cluster pointer and free | |
656 | * old cluster. This is what this loop does */ | |
657 | if(l2_table[l2_index + i] != 0) | |
658 | old_cluster[j++] = l2_table[l2_index + i]; | |
659 | ||
660 | l2_table[l2_index + i] = cpu_to_be64((cluster_offset + | |
661 | (i << s->cluster_bits)) | QCOW_OFLAG_COPIED); | |
662 | } | |
663 | ||
9f8e668e | 664 | |
29c1a730 | 665 | ret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table); |
c835d00f | 666 | if (ret < 0) { |
45aba42f | 667 | goto err; |
4c1612d9 | 668 | } |
45aba42f | 669 | |
7ec5e6a4 KW |
670 | /* |
671 | * If this was a COW, we need to decrease the refcount of the old cluster. | |
672 | * Also flush bs->file to get the right order for L2 and refcount update. | |
673 | */ | |
674 | if (j != 0) { | |
7ec5e6a4 KW |
675 | for (i = 0; i < j; i++) { |
676 | qcow2_free_any_clusters(bs, | |
677 | be64_to_cpu(old_cluster[i]) & ~QCOW_OFLAG_COPIED, 1); | |
678 | } | |
679 | } | |
45aba42f KW |
680 | |
681 | ret = 0; | |
682 | err: | |
683 | qemu_free(old_cluster); | |
684 | return ret; | |
685 | } | |
686 | ||
687 | /* | |
688 | * alloc_cluster_offset | |
689 | * | |
148da7ea | 690 | * For a given offset of the disk image, return cluster offset in qcow2 file. |
45aba42f KW |
691 | * If the offset is not found, allocate a new cluster. |
692 | * | |
148da7ea KW |
693 | * If the cluster was already allocated, m->nb_clusters is set to 0, |
694 | * m->depends_on is set to NULL and the other fields in m are meaningless. | |
695 | * | |
696 | * If the cluster is newly allocated, m->nb_clusters is set to the number of | |
697 | * contiguous clusters that have been allocated. This may be 0 if the request | |
698 | * conflict with another write request in flight; in this case, m->depends_on | |
699 | * is set and the remaining fields of m are meaningless. | |
45aba42f | 700 | * |
148da7ea KW |
701 | * If m->nb_clusters is non-zero, the other fields of m are valid and contain |
702 | * information about the first allocated cluster. | |
703 | * | |
704 | * Return 0 on success and -errno in error cases | |
45aba42f | 705 | */ |
f4f0d391 KW |
706 | int qcow2_alloc_cluster_offset(BlockDriverState *bs, uint64_t offset, |
707 | int n_start, int n_end, int *num, QCowL2Meta *m) | |
45aba42f KW |
708 | { |
709 | BDRVQcowState *s = bs->opaque; | |
710 | int l2_index, ret; | |
5d757b56 KW |
711 | uint64_t l2_offset, *l2_table; |
712 | int64_t cluster_offset; | |
80ee15a6 | 713 | unsigned int nb_clusters, i = 0; |
f214978a | 714 | QCowL2Meta *old_alloc; |
45aba42f KW |
715 | |
716 | ret = get_cluster_table(bs, offset, &l2_table, &l2_offset, &l2_index); | |
1e3e8f1a | 717 | if (ret < 0) { |
148da7ea | 718 | return ret; |
1e3e8f1a | 719 | } |
45aba42f KW |
720 | |
721 | nb_clusters = size_to_clusters(s, n_end << 9); | |
722 | ||
723 | nb_clusters = MIN(nb_clusters, s->l2_size - l2_index); | |
724 | ||
725 | cluster_offset = be64_to_cpu(l2_table[l2_index]); | |
726 | ||
727 | /* We keep all QCOW_OFLAG_COPIED clusters */ | |
728 | ||
729 | if (cluster_offset & QCOW_OFLAG_COPIED) { | |
730 | nb_clusters = count_contiguous_clusters(nb_clusters, s->cluster_size, | |
731 | &l2_table[l2_index], 0, 0); | |
732 | ||
733 | cluster_offset &= ~QCOW_OFLAG_COPIED; | |
734 | m->nb_clusters = 0; | |
148da7ea | 735 | m->depends_on = NULL; |
45aba42f KW |
736 | |
737 | goto out; | |
738 | } | |
739 | ||
740 | /* for the moment, multiple compressed clusters are not managed */ | |
741 | ||
742 | if (cluster_offset & QCOW_OFLAG_COMPRESSED) | |
743 | nb_clusters = 1; | |
744 | ||
745 | /* how many available clusters ? */ | |
746 | ||
747 | while (i < nb_clusters) { | |
748 | i += count_contiguous_clusters(nb_clusters - i, s->cluster_size, | |
749 | &l2_table[l2_index], i, 0); | |
4805bb66 | 750 | if ((i >= nb_clusters) || be64_to_cpu(l2_table[l2_index + i])) { |
45aba42f | 751 | break; |
4805bb66 | 752 | } |
45aba42f KW |
753 | |
754 | i += count_contiguous_free_clusters(nb_clusters - i, | |
755 | &l2_table[l2_index + i]); | |
4805bb66 KW |
756 | if (i >= nb_clusters) { |
757 | break; | |
758 | } | |
45aba42f KW |
759 | |
760 | cluster_offset = be64_to_cpu(l2_table[l2_index + i]); | |
761 | ||
762 | if ((cluster_offset & QCOW_OFLAG_COPIED) || | |
763 | (cluster_offset & QCOW_OFLAG_COMPRESSED)) | |
764 | break; | |
765 | } | |
4805bb66 | 766 | assert(i <= nb_clusters); |
45aba42f KW |
767 | nb_clusters = i; |
768 | ||
f214978a KW |
769 | /* |
770 | * Check if there already is an AIO write request in flight which allocates | |
771 | * the same cluster. In this case we need to wait until the previous | |
772 | * request has completed and updated the L2 table accordingly. | |
773 | */ | |
72cf2d4f | 774 | QLIST_FOREACH(old_alloc, &s->cluster_allocs, next_in_flight) { |
f214978a KW |
775 | |
776 | uint64_t end_offset = offset + nb_clusters * s->cluster_size; | |
777 | uint64_t old_offset = old_alloc->offset; | |
778 | uint64_t old_end_offset = old_alloc->offset + | |
779 | old_alloc->nb_clusters * s->cluster_size; | |
780 | ||
781 | if (end_offset < old_offset || offset > old_end_offset) { | |
782 | /* No intersection */ | |
783 | } else { | |
784 | if (offset < old_offset) { | |
785 | /* Stop at the start of a running allocation */ | |
786 | nb_clusters = (old_offset - offset) >> s->cluster_bits; | |
787 | } else { | |
788 | nb_clusters = 0; | |
789 | } | |
790 | ||
791 | if (nb_clusters == 0) { | |
792 | /* Set dependency and wait for a callback */ | |
793 | m->depends_on = old_alloc; | |
794 | m->nb_clusters = 0; | |
795 | *num = 0; | |
29c1a730 KW |
796 | ret = 0; |
797 | goto fail; | |
f214978a KW |
798 | } |
799 | } | |
800 | } | |
801 | ||
802 | if (!nb_clusters) { | |
803 | abort(); | |
804 | } | |
805 | ||
72cf2d4f | 806 | QLIST_INSERT_HEAD(&s->cluster_allocs, m, next_in_flight); |
f214978a | 807 | |
45aba42f KW |
808 | /* allocate a new cluster */ |
809 | ||
ed6ccf0f | 810 | cluster_offset = qcow2_alloc_clusters(bs, nb_clusters * s->cluster_size); |
5d757b56 | 811 | if (cluster_offset < 0) { |
c644db3d | 812 | QLIST_REMOVE(m, next_in_flight); |
29c1a730 KW |
813 | ret = cluster_offset; |
814 | goto fail; | |
5d757b56 | 815 | } |
45aba42f KW |
816 | |
817 | /* save info needed for meta data update */ | |
818 | m->offset = offset; | |
819 | m->n_start = n_start; | |
820 | m->nb_clusters = nb_clusters; | |
821 | ||
822 | out: | |
29c1a730 KW |
823 | ret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table); |
824 | if (ret < 0) { | |
825 | return ret; | |
826 | } | |
827 | ||
45aba42f | 828 | m->nb_available = MIN(nb_clusters << (s->cluster_bits - 9), n_end); |
148da7ea | 829 | m->cluster_offset = cluster_offset; |
45aba42f KW |
830 | |
831 | *num = m->nb_available - n_start; | |
832 | ||
148da7ea | 833 | return 0; |
29c1a730 KW |
834 | |
835 | fail: | |
836 | qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table); | |
837 | return ret; | |
45aba42f KW |
838 | } |
839 | ||
840 | static int decompress_buffer(uint8_t *out_buf, int out_buf_size, | |
841 | const uint8_t *buf, int buf_size) | |
842 | { | |
843 | z_stream strm1, *strm = &strm1; | |
844 | int ret, out_len; | |
845 | ||
846 | memset(strm, 0, sizeof(*strm)); | |
847 | ||
848 | strm->next_in = (uint8_t *)buf; | |
849 | strm->avail_in = buf_size; | |
850 | strm->next_out = out_buf; | |
851 | strm->avail_out = out_buf_size; | |
852 | ||
853 | ret = inflateInit2(strm, -12); | |
854 | if (ret != Z_OK) | |
855 | return -1; | |
856 | ret = inflate(strm, Z_FINISH); | |
857 | out_len = strm->next_out - out_buf; | |
858 | if ((ret != Z_STREAM_END && ret != Z_BUF_ERROR) || | |
859 | out_len != out_buf_size) { | |
860 | inflateEnd(strm); | |
861 | return -1; | |
862 | } | |
863 | inflateEnd(strm); | |
864 | return 0; | |
865 | } | |
866 | ||
66f82cee | 867 | int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset) |
45aba42f | 868 | { |
66f82cee | 869 | BDRVQcowState *s = bs->opaque; |
45aba42f KW |
870 | int ret, csize, nb_csectors, sector_offset; |
871 | uint64_t coffset; | |
872 | ||
873 | coffset = cluster_offset & s->cluster_offset_mask; | |
874 | if (s->cluster_cache_offset != coffset) { | |
875 | nb_csectors = ((cluster_offset >> s->csize_shift) & s->csize_mask) + 1; | |
876 | sector_offset = coffset & 511; | |
877 | csize = nb_csectors * 512 - sector_offset; | |
66f82cee KW |
878 | BLKDBG_EVENT(bs->file, BLKDBG_READ_COMPRESSED); |
879 | ret = bdrv_read(bs->file, coffset >> 9, s->cluster_data, nb_csectors); | |
45aba42f KW |
880 | if (ret < 0) { |
881 | return -1; | |
882 | } | |
883 | if (decompress_buffer(s->cluster_cache, s->cluster_size, | |
884 | s->cluster_data + sector_offset, csize) < 0) { | |
885 | return -1; | |
886 | } | |
887 | s->cluster_cache_offset = coffset; | |
888 | } | |
889 | return 0; | |
890 | } |