root / block-qcow2.c @ 94909d9f
History | View | Annotate | Download (88.9 kB)
1 | 585f8587 | bellard | /*
|
---|---|---|---|
2 | 585f8587 | bellard | * Block driver for the QCOW version 2 format
|
3 | 5fafdf24 | ths | *
|
4 | 585f8587 | bellard | * Copyright (c) 2004-2006 Fabrice Bellard
|
5 | 5fafdf24 | ths | *
|
6 | 585f8587 | bellard | * Permission is hereby granted, free of charge, to any person obtaining a copy
|
7 | 585f8587 | bellard | * of this software and associated documentation files (the "Software"), to deal
|
8 | 585f8587 | bellard | * in the Software without restriction, including without limitation the rights
|
9 | 585f8587 | bellard | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
10 | 585f8587 | bellard | * copies of the Software, and to permit persons to whom the Software is
|
11 | 585f8587 | bellard | * furnished to do so, subject to the following conditions:
|
12 | 585f8587 | bellard | *
|
13 | 585f8587 | bellard | * The above copyright notice and this permission notice shall be included in
|
14 | 585f8587 | bellard | * all copies or substantial portions of the Software.
|
15 | 585f8587 | bellard | *
|
16 | 585f8587 | bellard | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
17 | 585f8587 | bellard | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
18 | 585f8587 | bellard | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
19 | 585f8587 | bellard | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
20 | 585f8587 | bellard | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
21 | 585f8587 | bellard | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
22 | 585f8587 | bellard | * THE SOFTWARE.
|
23 | 585f8587 | bellard | */
|
24 | faf07963 | pbrook | #include "qemu-common.h" |
25 | 585f8587 | bellard | #include "block_int.h" |
26 | 585f8587 | bellard | #include <zlib.h> |
27 | 585f8587 | bellard | #include "aes.h" |
28 | 585f8587 | bellard | #include <assert.h> |
29 | 585f8587 | bellard | |
30 | 585f8587 | bellard | /*
|
31 | 585f8587 | bellard | Differences with QCOW:
|
32 | 585f8587 | bellard | |
33 | 585f8587 | bellard | - Support for multiple incremental snapshots.
|
34 | 585f8587 | bellard | - Memory management by reference counts.
|
35 | 585f8587 | bellard | - Clusters which have a reference count of one have the bit
|
36 | 585f8587 | bellard | QCOW_OFLAG_COPIED to optimize write performance.
|
37 | 5fafdf24 | ths | - Size of compressed clusters is stored in sectors to reduce bit usage
|
38 | 585f8587 | bellard | in the cluster offsets.
|
39 | 585f8587 | bellard | - Support for storing additional data (such as the VM state) in the
|
40 | 3b46e624 | ths | snapshots.
|
41 | 585f8587 | bellard | - If a backing store is used, the cluster size is not constrained
|
42 | 585f8587 | bellard | (could be backported to QCOW).
|
43 | 585f8587 | bellard | - L2 tables have always a size of one cluster.
|
44 | 585f8587 | bellard | */
|
45 | 585f8587 | bellard | |
46 | 585f8587 | bellard | //#define DEBUG_ALLOC
|
47 | 585f8587 | bellard | //#define DEBUG_ALLOC2
|
48 | 9b80ddf3 | aliguori | //#define DEBUG_EXT
|
49 | 5fafdf24 | ths | |
50 | 585f8587 | bellard | #define QCOW_MAGIC (('Q' << 24) | ('F' << 16) | ('I' << 8) | 0xfb) |
51 | 585f8587 | bellard | #define QCOW_VERSION 2 |
52 | 585f8587 | bellard | |
53 | 585f8587 | bellard | #define QCOW_CRYPT_NONE 0 |
54 | 585f8587 | bellard | #define QCOW_CRYPT_AES 1 |
55 | 585f8587 | bellard | |
56 | 095a9c58 | aliguori | #define QCOW_MAX_CRYPT_CLUSTERS 32 |
57 | 095a9c58 | aliguori | |
58 | 585f8587 | bellard | /* indicate that the refcount of the referenced cluster is exactly one. */
|
59 | 585f8587 | bellard | #define QCOW_OFLAG_COPIED (1LL << 63) |
60 | 585f8587 | bellard | /* indicate that the cluster is compressed (they never have the copied flag) */
|
61 | 585f8587 | bellard | #define QCOW_OFLAG_COMPRESSED (1LL << 62) |
62 | 585f8587 | bellard | |
63 | 585f8587 | bellard | #define REFCOUNT_SHIFT 1 /* refcount size is 2 bytes */ |
64 | 585f8587 | bellard | |
65 | 585f8587 | bellard | typedef struct QCowHeader { |
66 | 585f8587 | bellard | uint32_t magic; |
67 | 585f8587 | bellard | uint32_t version; |
68 | 585f8587 | bellard | uint64_t backing_file_offset; |
69 | 585f8587 | bellard | uint32_t backing_file_size; |
70 | 585f8587 | bellard | uint32_t cluster_bits; |
71 | 585f8587 | bellard | uint64_t size; /* in bytes */
|
72 | 585f8587 | bellard | uint32_t crypt_method; |
73 | 585f8587 | bellard | uint32_t l1_size; /* XXX: save number of clusters instead ? */
|
74 | 585f8587 | bellard | uint64_t l1_table_offset; |
75 | 585f8587 | bellard | uint64_t refcount_table_offset; |
76 | 585f8587 | bellard | uint32_t refcount_table_clusters; |
77 | 585f8587 | bellard | uint32_t nb_snapshots; |
78 | 585f8587 | bellard | uint64_t snapshots_offset; |
79 | 585f8587 | bellard | } QCowHeader; |
80 | 585f8587 | bellard | |
81 | 9b80ddf3 | aliguori | |
82 | 9b80ddf3 | aliguori | typedef struct { |
83 | 9b80ddf3 | aliguori | uint32_t magic; |
84 | 9b80ddf3 | aliguori | uint32_t len; |
85 | 9b80ddf3 | aliguori | } QCowExtension; |
86 | 9b80ddf3 | aliguori | #define QCOW_EXT_MAGIC_END 0 |
87 | f965509c | aliguori | #define QCOW_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA |
88 | 9b80ddf3 | aliguori | |
89 | 9b80ddf3 | aliguori | |
90 | 585f8587 | bellard | typedef struct __attribute__((packed)) QCowSnapshotHeader { |
91 | 585f8587 | bellard | /* header is 8 byte aligned */
|
92 | 585f8587 | bellard | uint64_t l1_table_offset; |
93 | 585f8587 | bellard | |
94 | 585f8587 | bellard | uint32_t l1_size; |
95 | 585f8587 | bellard | uint16_t id_str_size; |
96 | 585f8587 | bellard | uint16_t name_size; |
97 | 585f8587 | bellard | |
98 | 585f8587 | bellard | uint32_t date_sec; |
99 | 585f8587 | bellard | uint32_t date_nsec; |
100 | 585f8587 | bellard | |
101 | 585f8587 | bellard | uint64_t vm_clock_nsec; |
102 | 585f8587 | bellard | |
103 | 585f8587 | bellard | uint32_t vm_state_size; |
104 | 585f8587 | bellard | uint32_t extra_data_size; /* for extension */
|
105 | 585f8587 | bellard | /* extra data follows */
|
106 | 585f8587 | bellard | /* id_str follows */
|
107 | 585f8587 | bellard | /* name follows */
|
108 | 585f8587 | bellard | } QCowSnapshotHeader; |
109 | 585f8587 | bellard | |
110 | 585f8587 | bellard | #define L2_CACHE_SIZE 16 |
111 | 585f8587 | bellard | |
112 | 585f8587 | bellard | typedef struct QCowSnapshot { |
113 | 585f8587 | bellard | uint64_t l1_table_offset; |
114 | 585f8587 | bellard | uint32_t l1_size; |
115 | 585f8587 | bellard | char *id_str;
|
116 | 585f8587 | bellard | char *name;
|
117 | 585f8587 | bellard | uint32_t vm_state_size; |
118 | 585f8587 | bellard | uint32_t date_sec; |
119 | 585f8587 | bellard | uint32_t date_nsec; |
120 | 585f8587 | bellard | uint64_t vm_clock_nsec; |
121 | 585f8587 | bellard | } QCowSnapshot; |
122 | 585f8587 | bellard | |
123 | 585f8587 | bellard | typedef struct BDRVQcowState { |
124 | 585f8587 | bellard | BlockDriverState *hd; |
125 | 585f8587 | bellard | int cluster_bits;
|
126 | 585f8587 | bellard | int cluster_size;
|
127 | 585f8587 | bellard | int cluster_sectors;
|
128 | 585f8587 | bellard | int l2_bits;
|
129 | 585f8587 | bellard | int l2_size;
|
130 | 585f8587 | bellard | int l1_size;
|
131 | 585f8587 | bellard | int l1_vm_state_index;
|
132 | 585f8587 | bellard | int csize_shift;
|
133 | 585f8587 | bellard | int csize_mask;
|
134 | 585f8587 | bellard | uint64_t cluster_offset_mask; |
135 | 585f8587 | bellard | uint64_t l1_table_offset; |
136 | 585f8587 | bellard | uint64_t *l1_table; |
137 | 585f8587 | bellard | uint64_t *l2_cache; |
138 | 585f8587 | bellard | uint64_t l2_cache_offsets[L2_CACHE_SIZE]; |
139 | 585f8587 | bellard | uint32_t l2_cache_counts[L2_CACHE_SIZE]; |
140 | 585f8587 | bellard | uint8_t *cluster_cache; |
141 | 585f8587 | bellard | uint8_t *cluster_data; |
142 | 585f8587 | bellard | uint64_t cluster_cache_offset; |
143 | 585f8587 | bellard | |
144 | 585f8587 | bellard | uint64_t *refcount_table; |
145 | 585f8587 | bellard | uint64_t refcount_table_offset; |
146 | 585f8587 | bellard | uint32_t refcount_table_size; |
147 | 585f8587 | bellard | uint64_t refcount_block_cache_offset; |
148 | 585f8587 | bellard | uint16_t *refcount_block_cache; |
149 | 585f8587 | bellard | int64_t free_cluster_index; |
150 | 585f8587 | bellard | int64_t free_byte_offset; |
151 | 585f8587 | bellard | |
152 | 585f8587 | bellard | uint32_t crypt_method; /* current crypt method, 0 if no key yet */
|
153 | 585f8587 | bellard | uint32_t crypt_method_header; |
154 | 585f8587 | bellard | AES_KEY aes_encrypt_key; |
155 | 585f8587 | bellard | AES_KEY aes_decrypt_key; |
156 | 585f8587 | bellard | uint64_t snapshots_offset; |
157 | 585f8587 | bellard | int snapshots_size;
|
158 | 585f8587 | bellard | int nb_snapshots;
|
159 | 585f8587 | bellard | QCowSnapshot *snapshots; |
160 | 585f8587 | bellard | } BDRVQcowState; |
161 | 585f8587 | bellard | |
162 | 585f8587 | bellard | static int decompress_cluster(BDRVQcowState *s, uint64_t cluster_offset); |
163 | 5fafdf24 | ths | static int qcow_read(BlockDriverState *bs, int64_t sector_num, |
164 | 585f8587 | bellard | uint8_t *buf, int nb_sectors);
|
165 | 585f8587 | bellard | static int qcow_read_snapshots(BlockDriverState *bs); |
166 | 585f8587 | bellard | static void qcow_free_snapshots(BlockDriverState *bs); |
167 | 585f8587 | bellard | static int refcount_init(BlockDriverState *bs); |
168 | 585f8587 | bellard | static void refcount_close(BlockDriverState *bs); |
169 | 585f8587 | bellard | static int get_refcount(BlockDriverState *bs, int64_t cluster_index); |
170 | 5fafdf24 | ths | static int update_cluster_refcount(BlockDriverState *bs, |
171 | 585f8587 | bellard | int64_t cluster_index, |
172 | 585f8587 | bellard | int addend);
|
173 | 5fafdf24 | ths | static void update_refcount(BlockDriverState *bs, |
174 | 5fafdf24 | ths | int64_t offset, int64_t length, |
175 | 585f8587 | bellard | int addend);
|
176 | 585f8587 | bellard | static int64_t alloc_clusters(BlockDriverState *bs, int64_t size);
|
177 | 585f8587 | bellard | static int64_t alloc_bytes(BlockDriverState *bs, int size); |
178 | 5fafdf24 | ths | static void free_clusters(BlockDriverState *bs, |
179 | 585f8587 | bellard | int64_t offset, int64_t size); |
180 | e97fc193 | aliguori | static int check_refcounts(BlockDriverState *bs); |
181 | 585f8587 | bellard | |
182 | 585f8587 | bellard | static int qcow_probe(const uint8_t *buf, int buf_size, const char *filename) |
183 | 585f8587 | bellard | { |
184 | 585f8587 | bellard | const QCowHeader *cow_header = (const void *)buf; |
185 | 3b46e624 | ths | |
186 | 585f8587 | bellard | if (buf_size >= sizeof(QCowHeader) && |
187 | 585f8587 | bellard | be32_to_cpu(cow_header->magic) == QCOW_MAGIC && |
188 | 5fafdf24 | ths | be32_to_cpu(cow_header->version) == QCOW_VERSION) |
189 | 585f8587 | bellard | return 100; |
190 | 585f8587 | bellard | else
|
191 | 585f8587 | bellard | return 0; |
192 | 585f8587 | bellard | } |
193 | 585f8587 | bellard | |
194 | 9b80ddf3 | aliguori | |
195 | 9b80ddf3 | aliguori | /*
|
196 | 9b80ddf3 | aliguori | * read qcow2 extension and fill bs
|
197 | 9b80ddf3 | aliguori | * start reading from start_offset
|
198 | 9b80ddf3 | aliguori | * finish reading upon magic of value 0 or when end_offset reached
|
199 | 9b80ddf3 | aliguori | * unknown magic is skipped (future extension this version knows nothing about)
|
200 | 9b80ddf3 | aliguori | * return 0 upon success, non-0 otherwise
|
201 | 9b80ddf3 | aliguori | */
|
202 | 9b80ddf3 | aliguori | static int qcow_read_extensions(BlockDriverState *bs, uint64_t start_offset, |
203 | 9b80ddf3 | aliguori | uint64_t end_offset) |
204 | 9b80ddf3 | aliguori | { |
205 | 9b80ddf3 | aliguori | BDRVQcowState *s = bs->opaque; |
206 | 9b80ddf3 | aliguori | QCowExtension ext; |
207 | 9b80ddf3 | aliguori | uint64_t offset; |
208 | 9b80ddf3 | aliguori | |
209 | 9b80ddf3 | aliguori | #ifdef DEBUG_EXT
|
210 | 9b80ddf3 | aliguori | printf("qcow_read_extensions: start=%ld end=%ld\n", start_offset, end_offset);
|
211 | 9b80ddf3 | aliguori | #endif
|
212 | 9b80ddf3 | aliguori | offset = start_offset; |
213 | 9b80ddf3 | aliguori | while (offset < end_offset) {
|
214 | 9b80ddf3 | aliguori | |
215 | 9b80ddf3 | aliguori | #ifdef DEBUG_EXT
|
216 | 9b80ddf3 | aliguori | /* Sanity check */
|
217 | 9b80ddf3 | aliguori | if (offset > s->cluster_size)
|
218 | 9b80ddf3 | aliguori | printf("qcow_handle_extension: suspicious offset %lu\n", offset);
|
219 | 9b80ddf3 | aliguori | |
220 | 9b80ddf3 | aliguori | printf("attemting to read extended header in offset %lu\n", offset);
|
221 | 9b80ddf3 | aliguori | #endif
|
222 | 9b80ddf3 | aliguori | |
223 | 9b80ddf3 | aliguori | if (bdrv_pread(s->hd, offset, &ext, sizeof(ext)) != sizeof(ext)) { |
224 | 4c978075 | aliguori | fprintf(stderr, "qcow_handle_extension: ERROR: pread fail from offset %llu\n",
|
225 | 4c978075 | aliguori | (unsigned long long)offset); |
226 | 9b80ddf3 | aliguori | return 1; |
227 | 9b80ddf3 | aliguori | } |
228 | 9b80ddf3 | aliguori | be32_to_cpus(&ext.magic); |
229 | 9b80ddf3 | aliguori | be32_to_cpus(&ext.len); |
230 | 9b80ddf3 | aliguori | offset += sizeof(ext);
|
231 | 9b80ddf3 | aliguori | #ifdef DEBUG_EXT
|
232 | 9b80ddf3 | aliguori | printf("ext.magic = 0x%x\n", ext.magic);
|
233 | 9b80ddf3 | aliguori | #endif
|
234 | 9b80ddf3 | aliguori | switch (ext.magic) {
|
235 | 9b80ddf3 | aliguori | case QCOW_EXT_MAGIC_END:
|
236 | 9b80ddf3 | aliguori | return 0; |
237 | f965509c | aliguori | |
238 | f965509c | aliguori | case QCOW_EXT_MAGIC_BACKING_FORMAT:
|
239 | f965509c | aliguori | if (ext.len >= sizeof(bs->backing_format)) { |
240 | f965509c | aliguori | fprintf(stderr, "ERROR: ext_backing_format: len=%u too large"
|
241 | 4c978075 | aliguori | " (>=%zu)\n",
|
242 | f965509c | aliguori | ext.len, sizeof(bs->backing_format));
|
243 | f965509c | aliguori | return 2; |
244 | f965509c | aliguori | } |
245 | f965509c | aliguori | if (bdrv_pread(s->hd, offset , bs->backing_format,
|
246 | f965509c | aliguori | ext.len) != ext.len) |
247 | f965509c | aliguori | return 3; |
248 | f965509c | aliguori | bs->backing_format[ext.len] = '\0';
|
249 | f965509c | aliguori | #ifdef DEBUG_EXT
|
250 | f965509c | aliguori | printf("Qcow2: Got format extension %s\n", bs->backing_format);
|
251 | f965509c | aliguori | #endif
|
252 | f965509c | aliguori | offset += ((ext.len + 7) & ~7); |
253 | f965509c | aliguori | break;
|
254 | f965509c | aliguori | |
255 | 9b80ddf3 | aliguori | default:
|
256 | 9b80ddf3 | aliguori | /* unknown magic -- just skip it */
|
257 | 9b80ddf3 | aliguori | offset += ((ext.len + 7) & ~7); |
258 | 9b80ddf3 | aliguori | break;
|
259 | 9b80ddf3 | aliguori | } |
260 | 9b80ddf3 | aliguori | } |
261 | 9b80ddf3 | aliguori | |
262 | 9b80ddf3 | aliguori | return 0; |
263 | 9b80ddf3 | aliguori | } |
264 | 9b80ddf3 | aliguori | |
265 | 9b80ddf3 | aliguori | |
266 | 585f8587 | bellard | static int qcow_open(BlockDriverState *bs, const char *filename, int flags) |
267 | 585f8587 | bellard | { |
268 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
269 | 585f8587 | bellard | int len, i, shift, ret;
|
270 | 585f8587 | bellard | QCowHeader header; |
271 | 9b80ddf3 | aliguori | uint64_t ext_end; |
272 | 585f8587 | bellard | |
273 | 4dc822d7 | aliguori | /* Performance is terrible right now with cache=writethrough due mainly
|
274 | 4dc822d7 | aliguori | * to reference count updates. If the user does not explicitly specify
|
275 | 4dc822d7 | aliguori | * a caching type, force to writeback caching.
|
276 | 4dc822d7 | aliguori | */
|
277 | 4dc822d7 | aliguori | if ((flags & BDRV_O_CACHE_DEF)) {
|
278 | 4dc822d7 | aliguori | flags |= BDRV_O_CACHE_WB; |
279 | 4dc822d7 | aliguori | flags &= ~BDRV_O_CACHE_DEF; |
280 | 4dc822d7 | aliguori | } |
281 | b5eff355 | aurel32 | ret = bdrv_file_open(&s->hd, filename, flags); |
282 | 585f8587 | bellard | if (ret < 0) |
283 | 585f8587 | bellard | return ret;
|
284 | 585f8587 | bellard | if (bdrv_pread(s->hd, 0, &header, sizeof(header)) != sizeof(header)) |
285 | 585f8587 | bellard | goto fail;
|
286 | 585f8587 | bellard | be32_to_cpus(&header.magic); |
287 | 585f8587 | bellard | be32_to_cpus(&header.version); |
288 | 585f8587 | bellard | be64_to_cpus(&header.backing_file_offset); |
289 | 585f8587 | bellard | be32_to_cpus(&header.backing_file_size); |
290 | 585f8587 | bellard | be64_to_cpus(&header.size); |
291 | 585f8587 | bellard | be32_to_cpus(&header.cluster_bits); |
292 | 585f8587 | bellard | be32_to_cpus(&header.crypt_method); |
293 | 585f8587 | bellard | be64_to_cpus(&header.l1_table_offset); |
294 | 585f8587 | bellard | be32_to_cpus(&header.l1_size); |
295 | 585f8587 | bellard | be64_to_cpus(&header.refcount_table_offset); |
296 | 585f8587 | bellard | be32_to_cpus(&header.refcount_table_clusters); |
297 | 585f8587 | bellard | be64_to_cpus(&header.snapshots_offset); |
298 | 585f8587 | bellard | be32_to_cpus(&header.nb_snapshots); |
299 | 3b46e624 | ths | |
300 | 585f8587 | bellard | if (header.magic != QCOW_MAGIC || header.version != QCOW_VERSION)
|
301 | 585f8587 | bellard | goto fail;
|
302 | 5fafdf24 | ths | if (header.size <= 1 || |
303 | 5fafdf24 | ths | header.cluster_bits < 9 ||
|
304 | 585f8587 | bellard | header.cluster_bits > 16)
|
305 | 585f8587 | bellard | goto fail;
|
306 | 585f8587 | bellard | if (header.crypt_method > QCOW_CRYPT_AES)
|
307 | 585f8587 | bellard | goto fail;
|
308 | 585f8587 | bellard | s->crypt_method_header = header.crypt_method; |
309 | 585f8587 | bellard | if (s->crypt_method_header)
|
310 | 585f8587 | bellard | bs->encrypted = 1;
|
311 | 585f8587 | bellard | s->cluster_bits = header.cluster_bits; |
312 | 585f8587 | bellard | s->cluster_size = 1 << s->cluster_bits;
|
313 | 585f8587 | bellard | s->cluster_sectors = 1 << (s->cluster_bits - 9); |
314 | 585f8587 | bellard | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ |
315 | 585f8587 | bellard | s->l2_size = 1 << s->l2_bits;
|
316 | 585f8587 | bellard | bs->total_sectors = header.size / 512;
|
317 | 585f8587 | bellard | s->csize_shift = (62 - (s->cluster_bits - 8)); |
318 | 585f8587 | bellard | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; |
319 | 585f8587 | bellard | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; |
320 | 585f8587 | bellard | s->refcount_table_offset = header.refcount_table_offset; |
321 | 5fafdf24 | ths | s->refcount_table_size = |
322 | 585f8587 | bellard | header.refcount_table_clusters << (s->cluster_bits - 3);
|
323 | 585f8587 | bellard | |
324 | 585f8587 | bellard | s->snapshots_offset = header.snapshots_offset; |
325 | 585f8587 | bellard | s->nb_snapshots = header.nb_snapshots; |
326 | 585f8587 | bellard | |
327 | 585f8587 | bellard | /* read the level 1 table */
|
328 | 585f8587 | bellard | s->l1_size = header.l1_size; |
329 | 585f8587 | bellard | shift = s->cluster_bits + s->l2_bits; |
330 | 585f8587 | bellard | s->l1_vm_state_index = (header.size + (1LL << shift) - 1) >> shift; |
331 | 585f8587 | bellard | /* the L1 table must contain at least enough entries to put
|
332 | 585f8587 | bellard | header.size bytes */
|
333 | 585f8587 | bellard | if (s->l1_size < s->l1_vm_state_index)
|
334 | 585f8587 | bellard | goto fail;
|
335 | 585f8587 | bellard | s->l1_table_offset = header.l1_table_offset; |
336 | 585f8587 | bellard | s->l1_table = qemu_malloc(s->l1_size * sizeof(uint64_t));
|
337 | 5fafdf24 | ths | if (bdrv_pread(s->hd, s->l1_table_offset, s->l1_table, s->l1_size * sizeof(uint64_t)) != |
338 | 585f8587 | bellard | s->l1_size * sizeof(uint64_t))
|
339 | 585f8587 | bellard | goto fail;
|
340 | 585f8587 | bellard | for(i = 0;i < s->l1_size; i++) { |
341 | 585f8587 | bellard | be64_to_cpus(&s->l1_table[i]); |
342 | 585f8587 | bellard | } |
343 | 585f8587 | bellard | /* alloc L2 cache */
|
344 | 585f8587 | bellard | s->l2_cache = qemu_malloc(s->l2_size * L2_CACHE_SIZE * sizeof(uint64_t));
|
345 | 585f8587 | bellard | s->cluster_cache = qemu_malloc(s->cluster_size); |
346 | 585f8587 | bellard | /* one more sector for decompressed data alignment */
|
347 | 095a9c58 | aliguori | s->cluster_data = qemu_malloc(QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size |
348 | 095a9c58 | aliguori | + 512);
|
349 | 585f8587 | bellard | s->cluster_cache_offset = -1;
|
350 | 3b46e624 | ths | |
351 | 585f8587 | bellard | if (refcount_init(bs) < 0) |
352 | 585f8587 | bellard | goto fail;
|
353 | 585f8587 | bellard | |
354 | 9b80ddf3 | aliguori | /* read qcow2 extensions */
|
355 | 9b80ddf3 | aliguori | if (header.backing_file_offset)
|
356 | 9b80ddf3 | aliguori | ext_end = header.backing_file_offset; |
357 | 9b80ddf3 | aliguori | else
|
358 | 9b80ddf3 | aliguori | ext_end = s->cluster_size; |
359 | 9b80ddf3 | aliguori | if (qcow_read_extensions(bs, sizeof(header), ext_end)) |
360 | 9b80ddf3 | aliguori | goto fail;
|
361 | 9b80ddf3 | aliguori | |
362 | 585f8587 | bellard | /* read the backing file name */
|
363 | 585f8587 | bellard | if (header.backing_file_offset != 0) { |
364 | 585f8587 | bellard | len = header.backing_file_size; |
365 | 585f8587 | bellard | if (len > 1023) |
366 | 585f8587 | bellard | len = 1023;
|
367 | 585f8587 | bellard | if (bdrv_pread(s->hd, header.backing_file_offset, bs->backing_file, len) != len)
|
368 | 585f8587 | bellard | goto fail;
|
369 | 585f8587 | bellard | bs->backing_file[len] = '\0';
|
370 | 585f8587 | bellard | } |
371 | 585f8587 | bellard | if (qcow_read_snapshots(bs) < 0) |
372 | 585f8587 | bellard | goto fail;
|
373 | 585f8587 | bellard | |
374 | 585f8587 | bellard | #ifdef DEBUG_ALLOC
|
375 | 585f8587 | bellard | check_refcounts(bs); |
376 | 585f8587 | bellard | #endif
|
377 | 585f8587 | bellard | return 0; |
378 | 585f8587 | bellard | |
379 | 585f8587 | bellard | fail:
|
380 | 585f8587 | bellard | qcow_free_snapshots(bs); |
381 | 585f8587 | bellard | refcount_close(bs); |
382 | 585f8587 | bellard | qemu_free(s->l1_table); |
383 | 585f8587 | bellard | qemu_free(s->l2_cache); |
384 | 585f8587 | bellard | qemu_free(s->cluster_cache); |
385 | 585f8587 | bellard | qemu_free(s->cluster_data); |
386 | 585f8587 | bellard | bdrv_delete(s->hd); |
387 | 585f8587 | bellard | return -1; |
388 | 585f8587 | bellard | } |
389 | 585f8587 | bellard | |
390 | 585f8587 | bellard | static int qcow_set_key(BlockDriverState *bs, const char *key) |
391 | 585f8587 | bellard | { |
392 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
393 | 585f8587 | bellard | uint8_t keybuf[16];
|
394 | 585f8587 | bellard | int len, i;
|
395 | 3b46e624 | ths | |
396 | 585f8587 | bellard | memset(keybuf, 0, 16); |
397 | 585f8587 | bellard | len = strlen(key); |
398 | 585f8587 | bellard | if (len > 16) |
399 | 585f8587 | bellard | len = 16;
|
400 | 585f8587 | bellard | /* XXX: we could compress the chars to 7 bits to increase
|
401 | 585f8587 | bellard | entropy */
|
402 | 585f8587 | bellard | for(i = 0;i < len;i++) { |
403 | 585f8587 | bellard | keybuf[i] = key[i]; |
404 | 585f8587 | bellard | } |
405 | 585f8587 | bellard | s->crypt_method = s->crypt_method_header; |
406 | 585f8587 | bellard | |
407 | 585f8587 | bellard | if (AES_set_encrypt_key(keybuf, 128, &s->aes_encrypt_key) != 0) |
408 | 585f8587 | bellard | return -1; |
409 | 585f8587 | bellard | if (AES_set_decrypt_key(keybuf, 128, &s->aes_decrypt_key) != 0) |
410 | 585f8587 | bellard | return -1; |
411 | 585f8587 | bellard | #if 0
|
412 | 585f8587 | bellard | /* test */
|
413 | 585f8587 | bellard | {
|
414 | 585f8587 | bellard | uint8_t in[16];
|
415 | 585f8587 | bellard | uint8_t out[16];
|
416 | 585f8587 | bellard | uint8_t tmp[16];
|
417 | 585f8587 | bellard | for(i=0;i<16;i++)
|
418 | 585f8587 | bellard | in[i] = i;
|
419 | 585f8587 | bellard | AES_encrypt(in, tmp, &s->aes_encrypt_key);
|
420 | 585f8587 | bellard | AES_decrypt(tmp, out, &s->aes_decrypt_key);
|
421 | 585f8587 | bellard | for(i = 0; i < 16; i++)
|
422 | 585f8587 | bellard | printf(" %02x", tmp[i]);
|
423 | 585f8587 | bellard | printf("\n");
|
424 | 585f8587 | bellard | for(i = 0; i < 16; i++)
|
425 | 585f8587 | bellard | printf(" %02x", out[i]);
|
426 | 585f8587 | bellard | printf("\n");
|
427 | 585f8587 | bellard | }
|
428 | 585f8587 | bellard | #endif
|
429 | 585f8587 | bellard | return 0; |
430 | 585f8587 | bellard | } |
431 | 585f8587 | bellard | |
432 | 585f8587 | bellard | /* The crypt function is compatible with the linux cryptoloop
|
433 | 585f8587 | bellard | algorithm for < 4 GB images. NOTE: out_buf == in_buf is
|
434 | 585f8587 | bellard | supported */
|
435 | 585f8587 | bellard | static void encrypt_sectors(BDRVQcowState *s, int64_t sector_num, |
436 | 585f8587 | bellard | uint8_t *out_buf, const uint8_t *in_buf,
|
437 | 585f8587 | bellard | int nb_sectors, int enc, |
438 | 585f8587 | bellard | const AES_KEY *key)
|
439 | 585f8587 | bellard | { |
440 | 585f8587 | bellard | union {
|
441 | 585f8587 | bellard | uint64_t ll[2];
|
442 | 585f8587 | bellard | uint8_t b[16];
|
443 | 585f8587 | bellard | } ivec; |
444 | 585f8587 | bellard | int i;
|
445 | 585f8587 | bellard | |
446 | 585f8587 | bellard | for(i = 0; i < nb_sectors; i++) { |
447 | 585f8587 | bellard | ivec.ll[0] = cpu_to_le64(sector_num);
|
448 | 585f8587 | bellard | ivec.ll[1] = 0; |
449 | 5fafdf24 | ths | AES_cbc_encrypt(in_buf, out_buf, 512, key,
|
450 | 585f8587 | bellard | ivec.b, enc); |
451 | 585f8587 | bellard | sector_num++; |
452 | 585f8587 | bellard | in_buf += 512;
|
453 | 585f8587 | bellard | out_buf += 512;
|
454 | 585f8587 | bellard | } |
455 | 585f8587 | bellard | } |
456 | 585f8587 | bellard | |
457 | 585f8587 | bellard | static int copy_sectors(BlockDriverState *bs, uint64_t start_sect, |
458 | 585f8587 | bellard | uint64_t cluster_offset, int n_start, int n_end) |
459 | 585f8587 | bellard | { |
460 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
461 | 585f8587 | bellard | int n, ret;
|
462 | 585f8587 | bellard | |
463 | 585f8587 | bellard | n = n_end - n_start; |
464 | 585f8587 | bellard | if (n <= 0) |
465 | 585f8587 | bellard | return 0; |
466 | 585f8587 | bellard | ret = qcow_read(bs, start_sect + n_start, s->cluster_data, n); |
467 | 585f8587 | bellard | if (ret < 0) |
468 | 585f8587 | bellard | return ret;
|
469 | 585f8587 | bellard | if (s->crypt_method) {
|
470 | 5fafdf24 | ths | encrypt_sectors(s, start_sect + n_start, |
471 | 5fafdf24 | ths | s->cluster_data, |
472 | 585f8587 | bellard | s->cluster_data, n, 1,
|
473 | 585f8587 | bellard | &s->aes_encrypt_key); |
474 | 585f8587 | bellard | } |
475 | 5fafdf24 | ths | ret = bdrv_write(s->hd, (cluster_offset >> 9) + n_start,
|
476 | 585f8587 | bellard | s->cluster_data, n); |
477 | 585f8587 | bellard | if (ret < 0) |
478 | 585f8587 | bellard | return ret;
|
479 | 585f8587 | bellard | return 0; |
480 | 585f8587 | bellard | } |
481 | 585f8587 | bellard | |
482 | 585f8587 | bellard | static void l2_cache_reset(BlockDriverState *bs) |
483 | 585f8587 | bellard | { |
484 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
485 | 585f8587 | bellard | |
486 | 585f8587 | bellard | memset(s->l2_cache, 0, s->l2_size * L2_CACHE_SIZE * sizeof(uint64_t)); |
487 | 585f8587 | bellard | memset(s->l2_cache_offsets, 0, L2_CACHE_SIZE * sizeof(uint64_t)); |
488 | 585f8587 | bellard | memset(s->l2_cache_counts, 0, L2_CACHE_SIZE * sizeof(uint32_t)); |
489 | 585f8587 | bellard | } |
490 | 585f8587 | bellard | |
491 | 585f8587 | bellard | static inline int l2_cache_new_entry(BlockDriverState *bs) |
492 | 585f8587 | bellard | { |
493 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
494 | 585f8587 | bellard | uint32_t min_count; |
495 | 585f8587 | bellard | int min_index, i;
|
496 | 585f8587 | bellard | |
497 | 585f8587 | bellard | /* find a new entry in the least used one */
|
498 | 585f8587 | bellard | min_index = 0;
|
499 | 585f8587 | bellard | min_count = 0xffffffff;
|
500 | 585f8587 | bellard | for(i = 0; i < L2_CACHE_SIZE; i++) { |
501 | 585f8587 | bellard | if (s->l2_cache_counts[i] < min_count) {
|
502 | 585f8587 | bellard | min_count = s->l2_cache_counts[i]; |
503 | 585f8587 | bellard | min_index = i; |
504 | 585f8587 | bellard | } |
505 | 585f8587 | bellard | } |
506 | 585f8587 | bellard | return min_index;
|
507 | 585f8587 | bellard | } |
508 | 585f8587 | bellard | |
509 | 585f8587 | bellard | static int64_t align_offset(int64_t offset, int n) |
510 | 585f8587 | bellard | { |
511 | 585f8587 | bellard | offset = (offset + n - 1) & ~(n - 1); |
512 | 585f8587 | bellard | return offset;
|
513 | 585f8587 | bellard | } |
514 | 585f8587 | bellard | |
515 | 585f8587 | bellard | static int grow_l1_table(BlockDriverState *bs, int min_size) |
516 | 585f8587 | bellard | { |
517 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
518 | 585f8587 | bellard | int new_l1_size, new_l1_size2, ret, i;
|
519 | 585f8587 | bellard | uint64_t *new_l1_table; |
520 | 585f8587 | bellard | uint64_t new_l1_table_offset; |
521 | 643e5399 | aliguori | uint8_t data[12];
|
522 | 585f8587 | bellard | |
523 | 585f8587 | bellard | new_l1_size = s->l1_size; |
524 | 585f8587 | bellard | if (min_size <= new_l1_size)
|
525 | 585f8587 | bellard | return 0; |
526 | 585f8587 | bellard | while (min_size > new_l1_size) {
|
527 | 585f8587 | bellard | new_l1_size = (new_l1_size * 3 + 1) / 2; |
528 | 585f8587 | bellard | } |
529 | 585f8587 | bellard | #ifdef DEBUG_ALLOC2
|
530 | 585f8587 | bellard | printf("grow l1_table from %d to %d\n", s->l1_size, new_l1_size);
|
531 | 585f8587 | bellard | #endif
|
532 | 585f8587 | bellard | |
533 | 585f8587 | bellard | new_l1_size2 = sizeof(uint64_t) * new_l1_size;
|
534 | 585f8587 | bellard | new_l1_table = qemu_mallocz(new_l1_size2); |
535 | 585f8587 | bellard | memcpy(new_l1_table, s->l1_table, s->l1_size * sizeof(uint64_t));
|
536 | 585f8587 | bellard | |
537 | 585f8587 | bellard | /* write new table (align to cluster) */
|
538 | 585f8587 | bellard | new_l1_table_offset = alloc_clusters(bs, new_l1_size2); |
539 | 3b46e624 | ths | |
540 | 585f8587 | bellard | for(i = 0; i < s->l1_size; i++) |
541 | 585f8587 | bellard | new_l1_table[i] = cpu_to_be64(new_l1_table[i]); |
542 | 585f8587 | bellard | ret = bdrv_pwrite(s->hd, new_l1_table_offset, new_l1_table, new_l1_size2); |
543 | 585f8587 | bellard | if (ret != new_l1_size2)
|
544 | 585f8587 | bellard | goto fail;
|
545 | 585f8587 | bellard | for(i = 0; i < s->l1_size; i++) |
546 | 585f8587 | bellard | new_l1_table[i] = be64_to_cpu(new_l1_table[i]); |
547 | 3b46e624 | ths | |
548 | 585f8587 | bellard | /* set new table */
|
549 | 643e5399 | aliguori | cpu_to_be32w((uint32_t*)data, new_l1_size); |
550 | 643e5399 | aliguori | cpu_to_be64w((uint64_t*)(data + 4), new_l1_table_offset);
|
551 | 643e5399 | aliguori | if (bdrv_pwrite(s->hd, offsetof(QCowHeader, l1_size), data,
|
552 | 643e5399 | aliguori | sizeof(data)) != sizeof(data)) |
553 | 585f8587 | bellard | goto fail;
|
554 | 585f8587 | bellard | qemu_free(s->l1_table); |
555 | 585f8587 | bellard | free_clusters(bs, s->l1_table_offset, s->l1_size * sizeof(uint64_t));
|
556 | 585f8587 | bellard | s->l1_table_offset = new_l1_table_offset; |
557 | 585f8587 | bellard | s->l1_table = new_l1_table; |
558 | 585f8587 | bellard | s->l1_size = new_l1_size; |
559 | 585f8587 | bellard | return 0; |
560 | 585f8587 | bellard | fail:
|
561 | 585f8587 | bellard | qemu_free(s->l1_table); |
562 | 585f8587 | bellard | return -EIO;
|
563 | 585f8587 | bellard | } |
564 | 585f8587 | bellard | |
565 | 108534b9 | aliguori | /*
|
566 | 108534b9 | aliguori | * seek_l2_table
|
567 | 585f8587 | bellard | *
|
568 | 108534b9 | aliguori | * seek l2_offset in the l2_cache table
|
569 | 108534b9 | aliguori | * if not found, return NULL,
|
570 | 108534b9 | aliguori | * if found,
|
571 | 108534b9 | aliguori | * increments the l2 cache hit count of the entry,
|
572 | 108534b9 | aliguori | * if counter overflow, divide by two all counters
|
573 | 108534b9 | aliguori | * return the pointer to the l2 cache entry
|
574 | 585f8587 | bellard | *
|
575 | 108534b9 | aliguori | */
|
576 | 108534b9 | aliguori | |
577 | 108534b9 | aliguori | static uint64_t *seek_l2_table(BDRVQcowState *s, uint64_t l2_offset)
|
578 | 108534b9 | aliguori | { |
579 | 108534b9 | aliguori | int i, j;
|
580 | 108534b9 | aliguori | |
581 | 108534b9 | aliguori | for(i = 0; i < L2_CACHE_SIZE; i++) { |
582 | 108534b9 | aliguori | if (l2_offset == s->l2_cache_offsets[i]) {
|
583 | 108534b9 | aliguori | /* increment the hit count */
|
584 | 108534b9 | aliguori | if (++s->l2_cache_counts[i] == 0xffffffff) { |
585 | 108534b9 | aliguori | for(j = 0; j < L2_CACHE_SIZE; j++) { |
586 | 108534b9 | aliguori | s->l2_cache_counts[j] >>= 1;
|
587 | 108534b9 | aliguori | } |
588 | 108534b9 | aliguori | } |
589 | 108534b9 | aliguori | return s->l2_cache + (i << s->l2_bits);
|
590 | 108534b9 | aliguori | } |
591 | 108534b9 | aliguori | } |
592 | 108534b9 | aliguori | return NULL; |
593 | 108534b9 | aliguori | } |
594 | 108534b9 | aliguori | |
595 | 108534b9 | aliguori | /*
|
596 | 108534b9 | aliguori | * l2_load
|
597 | 108534b9 | aliguori | *
|
598 | 108534b9 | aliguori | * Loads a L2 table into memory. If the table is in the cache, the cache
|
599 | 108534b9 | aliguori | * is used; otherwise the L2 table is loaded from the image file.
|
600 | 108534b9 | aliguori | *
|
601 | 108534b9 | aliguori | * Returns a pointer to the L2 table on success, or NULL if the read from
|
602 | 108534b9 | aliguori | * the image file failed.
|
603 | 108534b9 | aliguori | */
|
604 | 108534b9 | aliguori | |
605 | 108534b9 | aliguori | static uint64_t *l2_load(BlockDriverState *bs, uint64_t l2_offset)
|
606 | 108534b9 | aliguori | { |
607 | 108534b9 | aliguori | BDRVQcowState *s = bs->opaque; |
608 | 108534b9 | aliguori | int min_index;
|
609 | 108534b9 | aliguori | uint64_t *l2_table; |
610 | 108534b9 | aliguori | |
611 | 108534b9 | aliguori | /* seek if the table for the given offset is in the cache */
|
612 | 108534b9 | aliguori | |
613 | 108534b9 | aliguori | l2_table = seek_l2_table(s, l2_offset); |
614 | 108534b9 | aliguori | if (l2_table != NULL) |
615 | 108534b9 | aliguori | return l2_table;
|
616 | 108534b9 | aliguori | |
617 | 108534b9 | aliguori | /* not found: load a new entry in the least used one */
|
618 | 108534b9 | aliguori | |
619 | 108534b9 | aliguori | min_index = l2_cache_new_entry(bs); |
620 | 108534b9 | aliguori | l2_table = s->l2_cache + (min_index << s->l2_bits); |
621 | 108534b9 | aliguori | if (bdrv_pread(s->hd, l2_offset, l2_table, s->l2_size * sizeof(uint64_t)) != |
622 | 108534b9 | aliguori | s->l2_size * sizeof(uint64_t))
|
623 | 108534b9 | aliguori | return NULL; |
624 | 108534b9 | aliguori | s->l2_cache_offsets[min_index] = l2_offset; |
625 | 108534b9 | aliguori | s->l2_cache_counts[min_index] = 1;
|
626 | 108534b9 | aliguori | |
627 | 108534b9 | aliguori | return l2_table;
|
628 | 108534b9 | aliguori | } |
629 | 108534b9 | aliguori | |
630 | 108534b9 | aliguori | /*
|
631 | 108534b9 | aliguori | * l2_allocate
|
632 | 585f8587 | bellard | *
|
633 | 108534b9 | aliguori | * Allocate a new l2 entry in the file. If l1_index points to an already
|
634 | 108534b9 | aliguori | * used entry in the L2 table (i.e. we are doing a copy on write for the L2
|
635 | 108534b9 | aliguori | * table) copy the contents of the old L2 table into the newly allocated one.
|
636 | 108534b9 | aliguori | * Otherwise the new table is initialized with zeros.
|
637 | 585f8587 | bellard | *
|
638 | 585f8587 | bellard | */
|
639 | 108534b9 | aliguori | |
640 | 108534b9 | aliguori | static uint64_t *l2_allocate(BlockDriverState *bs, int l1_index) |
641 | 108534b9 | aliguori | { |
642 | 108534b9 | aliguori | BDRVQcowState *s = bs->opaque; |
643 | 108534b9 | aliguori | int min_index;
|
644 | 108534b9 | aliguori | uint64_t old_l2_offset, tmp; |
645 | 108534b9 | aliguori | uint64_t *l2_table, l2_offset; |
646 | 108534b9 | aliguori | |
647 | 108534b9 | aliguori | old_l2_offset = s->l1_table[l1_index]; |
648 | 108534b9 | aliguori | |
649 | 108534b9 | aliguori | /* allocate a new l2 entry */
|
650 | 108534b9 | aliguori | |
651 | 108534b9 | aliguori | l2_offset = alloc_clusters(bs, s->l2_size * sizeof(uint64_t));
|
652 | 108534b9 | aliguori | |
653 | 108534b9 | aliguori | /* update the L1 entry */
|
654 | 108534b9 | aliguori | |
655 | 108534b9 | aliguori | s->l1_table[l1_index] = l2_offset | QCOW_OFLAG_COPIED; |
656 | 108534b9 | aliguori | |
657 | 108534b9 | aliguori | tmp = cpu_to_be64(l2_offset | QCOW_OFLAG_COPIED); |
658 | 108534b9 | aliguori | if (bdrv_pwrite(s->hd, s->l1_table_offset + l1_index * sizeof(tmp), |
659 | 108534b9 | aliguori | &tmp, sizeof(tmp)) != sizeof(tmp)) |
660 | 108534b9 | aliguori | return NULL; |
661 | 108534b9 | aliguori | |
662 | 108534b9 | aliguori | /* allocate a new entry in the l2 cache */
|
663 | 108534b9 | aliguori | |
664 | 108534b9 | aliguori | min_index = l2_cache_new_entry(bs); |
665 | 108534b9 | aliguori | l2_table = s->l2_cache + (min_index << s->l2_bits); |
666 | 108534b9 | aliguori | |
667 | 108534b9 | aliguori | if (old_l2_offset == 0) { |
668 | 108534b9 | aliguori | /* if there was no old l2 table, clear the new table */
|
669 | 108534b9 | aliguori | memset(l2_table, 0, s->l2_size * sizeof(uint64_t)); |
670 | 108534b9 | aliguori | } else {
|
671 | 108534b9 | aliguori | /* if there was an old l2 table, read it from the disk */
|
672 | 108534b9 | aliguori | if (bdrv_pread(s->hd, old_l2_offset,
|
673 | 108534b9 | aliguori | l2_table, s->l2_size * sizeof(uint64_t)) !=
|
674 | 108534b9 | aliguori | s->l2_size * sizeof(uint64_t))
|
675 | 108534b9 | aliguori | return NULL; |
676 | 108534b9 | aliguori | } |
677 | 108534b9 | aliguori | /* write the l2 table to the file */
|
678 | 108534b9 | aliguori | if (bdrv_pwrite(s->hd, l2_offset,
|
679 | 108534b9 | aliguori | l2_table, s->l2_size * sizeof(uint64_t)) !=
|
680 | 108534b9 | aliguori | s->l2_size * sizeof(uint64_t))
|
681 | 108534b9 | aliguori | return NULL; |
682 | 108534b9 | aliguori | |
683 | 108534b9 | aliguori | /* update the l2 cache entry */
|
684 | 108534b9 | aliguori | |
685 | 108534b9 | aliguori | s->l2_cache_offsets[min_index] = l2_offset; |
686 | 108534b9 | aliguori | s->l2_cache_counts[min_index] = 1;
|
687 | 108534b9 | aliguori | |
688 | 108534b9 | aliguori | return l2_table;
|
689 | 108534b9 | aliguori | } |
690 | 108534b9 | aliguori | |
691 | 6db6c638 | aliguori | static int size_to_clusters(BDRVQcowState *s, int64_t size) |
692 | 6db6c638 | aliguori | { |
693 | 6db6c638 | aliguori | return (size + (s->cluster_size - 1)) >> s->cluster_bits; |
694 | 6db6c638 | aliguori | } |
695 | 6db6c638 | aliguori | |
696 | 6db6c638 | aliguori | static int count_contiguous_clusters(uint64_t nb_clusters, int cluster_size, |
697 | ff4b91c2 | aliguori | uint64_t *l2_table, uint64_t start, uint64_t mask) |
698 | 6db6c638 | aliguori | { |
699 | 6db6c638 | aliguori | int i;
|
700 | 6db6c638 | aliguori | uint64_t offset = be64_to_cpu(l2_table[0]) & ~mask;
|
701 | 6db6c638 | aliguori | |
702 | ab5ccbd6 | aliguori | if (!offset)
|
703 | ab5ccbd6 | aliguori | return 0; |
704 | ab5ccbd6 | aliguori | |
705 | ff4b91c2 | aliguori | for (i = start; i < start + nb_clusters; i++)
|
706 | 6db6c638 | aliguori | if (offset + i * cluster_size != (be64_to_cpu(l2_table[i]) & ~mask))
|
707 | 6db6c638 | aliguori | break;
|
708 | 6db6c638 | aliguori | |
709 | ff4b91c2 | aliguori | return (i - start);
|
710 | 6db6c638 | aliguori | } |
711 | 6db6c638 | aliguori | |
712 | 6db6c638 | aliguori | static int count_contiguous_free_clusters(uint64_t nb_clusters, uint64_t *l2_table) |
713 | 6db6c638 | aliguori | { |
714 | 6db6c638 | aliguori | int i = 0; |
715 | 6db6c638 | aliguori | |
716 | 6db6c638 | aliguori | while(nb_clusters-- && l2_table[i] == 0) |
717 | 6db6c638 | aliguori | i++; |
718 | 6db6c638 | aliguori | |
719 | 6db6c638 | aliguori | return i;
|
720 | 6db6c638 | aliguori | } |
721 | 6db6c638 | aliguori | |
722 | 05203524 | aliguori | /*
|
723 | 05203524 | aliguori | * get_cluster_offset
|
724 | 05203524 | aliguori | *
|
725 | 05203524 | aliguori | * For a given offset of the disk image, return cluster offset in
|
726 | 05203524 | aliguori | * qcow2 file.
|
727 | 05203524 | aliguori | *
|
728 | 095a9c58 | aliguori | * on entry, *num is the number of contiguous clusters we'd like to
|
729 | 095a9c58 | aliguori | * access following offset.
|
730 | 095a9c58 | aliguori | *
|
731 | 095a9c58 | aliguori | * on exit, *num is the number of contiguous clusters we can read.
|
732 | 095a9c58 | aliguori | *
|
733 | 05203524 | aliguori | * Return 1, if the offset is found
|
734 | 05203524 | aliguori | * Return 0, otherwise.
|
735 | 05203524 | aliguori | *
|
736 | 05203524 | aliguori | */
|
737 | 05203524 | aliguori | |
738 | 095a9c58 | aliguori | static uint64_t get_cluster_offset(BlockDriverState *bs,
|
739 | 095a9c58 | aliguori | uint64_t offset, int *num)
|
740 | 05203524 | aliguori | { |
741 | 05203524 | aliguori | BDRVQcowState *s = bs->opaque; |
742 | 05203524 | aliguori | int l1_index, l2_index;
|
743 | 6db6c638 | aliguori | uint64_t l2_offset, *l2_table, cluster_offset; |
744 | 6db6c638 | aliguori | int l1_bits, c;
|
745 | 6db6c638 | aliguori | int index_in_cluster, nb_available, nb_needed, nb_clusters;
|
746 | 095a9c58 | aliguori | |
747 | 095a9c58 | aliguori | index_in_cluster = (offset >> 9) & (s->cluster_sectors - 1); |
748 | 095a9c58 | aliguori | nb_needed = *num + index_in_cluster; |
749 | 095a9c58 | aliguori | |
750 | 095a9c58 | aliguori | l1_bits = s->l2_bits + s->cluster_bits; |
751 | 095a9c58 | aliguori | |
752 | 095a9c58 | aliguori | /* compute how many bytes there are between the offset and
|
753 | 6db6c638 | aliguori | * the end of the l1 entry
|
754 | 095a9c58 | aliguori | */
|
755 | 095a9c58 | aliguori | |
756 | 095a9c58 | aliguori | nb_available = (1 << l1_bits) - (offset & ((1 << l1_bits) - 1)); |
757 | 095a9c58 | aliguori | |
758 | 095a9c58 | aliguori | /* compute the number of available sectors */
|
759 | 095a9c58 | aliguori | |
760 | 095a9c58 | aliguori | nb_available = (nb_available >> 9) + index_in_cluster;
|
761 | 095a9c58 | aliguori | |
762 | f8de1660 | aliguori | if (nb_needed > nb_available) {
|
763 | f8de1660 | aliguori | nb_needed = nb_available; |
764 | f8de1660 | aliguori | } |
765 | f8de1660 | aliguori | |
766 | 095a9c58 | aliguori | cluster_offset = 0;
|
767 | 05203524 | aliguori | |
768 | 05203524 | aliguori | /* seek the the l2 offset in the l1 table */
|
769 | 05203524 | aliguori | |
770 | 095a9c58 | aliguori | l1_index = offset >> l1_bits; |
771 | 05203524 | aliguori | if (l1_index >= s->l1_size)
|
772 | 095a9c58 | aliguori | goto out;
|
773 | 05203524 | aliguori | |
774 | 05203524 | aliguori | l2_offset = s->l1_table[l1_index]; |
775 | 05203524 | aliguori | |
776 | 05203524 | aliguori | /* seek the l2 table of the given l2 offset */
|
777 | 05203524 | aliguori | |
778 | 05203524 | aliguori | if (!l2_offset)
|
779 | 095a9c58 | aliguori | goto out;
|
780 | 05203524 | aliguori | |
781 | 05203524 | aliguori | /* load the l2 table in memory */
|
782 | 05203524 | aliguori | |
783 | 05203524 | aliguori | l2_offset &= ~QCOW_OFLAG_COPIED; |
784 | 05203524 | aliguori | l2_table = l2_load(bs, l2_offset); |
785 | 05203524 | aliguori | if (l2_table == NULL) |
786 | 768706a5 | aliguori | return 0; |
787 | 05203524 | aliguori | |
788 | 05203524 | aliguori | /* find the cluster offset for the given disk offset */
|
789 | 05203524 | aliguori | |
790 | 05203524 | aliguori | l2_index = (offset >> s->cluster_bits) & (s->l2_size - 1);
|
791 | 05203524 | aliguori | cluster_offset = be64_to_cpu(l2_table[l2_index]); |
792 | 6db6c638 | aliguori | nb_clusters = size_to_clusters(s, nb_needed << 9);
|
793 | 095a9c58 | aliguori | |
794 | 095a9c58 | aliguori | if (!cluster_offset) {
|
795 | 6db6c638 | aliguori | /* how many empty clusters ? */
|
796 | 6db6c638 | aliguori | c = count_contiguous_free_clusters(nb_clusters, &l2_table[l2_index]); |
797 | 095a9c58 | aliguori | } else {
|
798 | 6db6c638 | aliguori | /* how many allocated clusters ? */
|
799 | 6db6c638 | aliguori | c = count_contiguous_clusters(nb_clusters, s->cluster_size, |
800 | ff4b91c2 | aliguori | &l2_table[l2_index], 0, QCOW_OFLAG_COPIED);
|
801 | 6db6c638 | aliguori | } |
802 | 095a9c58 | aliguori | |
803 | 6db6c638 | aliguori | nb_available = (c * s->cluster_sectors); |
804 | 095a9c58 | aliguori | out:
|
805 | 095a9c58 | aliguori | if (nb_available > nb_needed)
|
806 | 095a9c58 | aliguori | nb_available = nb_needed; |
807 | 095a9c58 | aliguori | |
808 | 095a9c58 | aliguori | *num = nb_available - index_in_cluster; |
809 | 095a9c58 | aliguori | |
810 | 6db6c638 | aliguori | return cluster_offset & ~QCOW_OFLAG_COPIED;
|
811 | 05203524 | aliguori | } |
812 | 05203524 | aliguori | |
813 | 05203524 | aliguori | /*
|
814 | 52d893ec | aliguori | * free_any_clusters
|
815 | 05203524 | aliguori | *
|
816 | 52d893ec | aliguori | * free clusters according to its type: compressed or not
|
817 | 05203524 | aliguori | *
|
818 | 52d893ec | aliguori | */
|
819 | 52d893ec | aliguori | |
820 | 52d893ec | aliguori | static void free_any_clusters(BlockDriverState *bs, |
821 | 095a9c58 | aliguori | uint64_t cluster_offset, int nb_clusters)
|
822 | 52d893ec | aliguori | { |
823 | 52d893ec | aliguori | BDRVQcowState *s = bs->opaque; |
824 | 52d893ec | aliguori | |
825 | 52d893ec | aliguori | /* free the cluster */
|
826 | 52d893ec | aliguori | |
827 | 52d893ec | aliguori | if (cluster_offset & QCOW_OFLAG_COMPRESSED) {
|
828 | 52d893ec | aliguori | int nb_csectors;
|
829 | 52d893ec | aliguori | nb_csectors = ((cluster_offset >> s->csize_shift) & |
830 | 52d893ec | aliguori | s->csize_mask) + 1;
|
831 | 52d893ec | aliguori | free_clusters(bs, (cluster_offset & s->cluster_offset_mask) & ~511,
|
832 | 52d893ec | aliguori | nb_csectors * 512);
|
833 | 52d893ec | aliguori | return;
|
834 | 52d893ec | aliguori | } |
835 | 52d893ec | aliguori | |
836 | 095a9c58 | aliguori | free_clusters(bs, cluster_offset, nb_clusters << s->cluster_bits); |
837 | 095a9c58 | aliguori | |
838 | 095a9c58 | aliguori | return;
|
839 | 52d893ec | aliguori | } |
840 | 52d893ec | aliguori | |
841 | 52d893ec | aliguori | /*
|
842 | 52d893ec | aliguori | * get_cluster_table
|
843 | 05203524 | aliguori | *
|
844 | 52d893ec | aliguori | * for a given disk offset, load (and allocate if needed)
|
845 | 52d893ec | aliguori | * the l2 table.
|
846 | 52d893ec | aliguori | *
|
847 | 52d893ec | aliguori | * the l2 table offset in the qcow2 file and the cluster index
|
848 | 52d893ec | aliguori | * in the l2 table are given to the caller.
|
849 | 05203524 | aliguori | *
|
850 | 05203524 | aliguori | */
|
851 | 05203524 | aliguori | |
852 | 52d893ec | aliguori | static int get_cluster_table(BlockDriverState *bs, uint64_t offset, |
853 | 52d893ec | aliguori | uint64_t **new_l2_table, |
854 | 52d893ec | aliguori | uint64_t *new_l2_offset, |
855 | 52d893ec | aliguori | int *new_l2_index)
|
856 | 585f8587 | bellard | { |
857 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
858 | 108534b9 | aliguori | int l1_index, l2_index, ret;
|
859 | 52d893ec | aliguori | uint64_t l2_offset, *l2_table; |
860 | 108534b9 | aliguori | |
861 | 108534b9 | aliguori | /* seek the the l2 offset in the l1 table */
|
862 | 3b46e624 | ths | |
863 | 585f8587 | bellard | l1_index = offset >> (s->l2_bits + s->cluster_bits); |
864 | 585f8587 | bellard | if (l1_index >= s->l1_size) {
|
865 | 108534b9 | aliguori | ret = grow_l1_table(bs, l1_index + 1);
|
866 | 108534b9 | aliguori | if (ret < 0) |
867 | 585f8587 | bellard | return 0; |
868 | 585f8587 | bellard | } |
869 | 585f8587 | bellard | l2_offset = s->l1_table[l1_index]; |
870 | 108534b9 | aliguori | |
871 | 108534b9 | aliguori | /* seek the l2 table of the given l2 offset */
|
872 | 108534b9 | aliguori | |
873 | 05203524 | aliguori | if (l2_offset & QCOW_OFLAG_COPIED) {
|
874 | 05203524 | aliguori | /* load the l2 table in memory */
|
875 | 05203524 | aliguori | l2_offset &= ~QCOW_OFLAG_COPIED; |
876 | 05203524 | aliguori | l2_table = l2_load(bs, l2_offset); |
877 | 05203524 | aliguori | if (l2_table == NULL) |
878 | 585f8587 | bellard | return 0; |
879 | 05203524 | aliguori | } else {
|
880 | 05203524 | aliguori | if (l2_offset)
|
881 | 05203524 | aliguori | free_clusters(bs, l2_offset, s->l2_size * sizeof(uint64_t));
|
882 | 108534b9 | aliguori | l2_table = l2_allocate(bs, l1_index); |
883 | 108534b9 | aliguori | if (l2_table == NULL) |
884 | 585f8587 | bellard | return 0; |
885 | 108534b9 | aliguori | l2_offset = s->l1_table[l1_index] & ~QCOW_OFLAG_COPIED; |
886 | 585f8587 | bellard | } |
887 | 108534b9 | aliguori | |
888 | 108534b9 | aliguori | /* find the cluster offset for the given disk offset */
|
889 | 108534b9 | aliguori | |
890 | 585f8587 | bellard | l2_index = (offset >> s->cluster_bits) & (s->l2_size - 1);
|
891 | 05203524 | aliguori | |
892 | 52d893ec | aliguori | *new_l2_table = l2_table; |
893 | 52d893ec | aliguori | *new_l2_offset = l2_offset; |
894 | 52d893ec | aliguori | *new_l2_index = l2_index; |
895 | 52d893ec | aliguori | |
896 | 52d893ec | aliguori | return 1; |
897 | 52d893ec | aliguori | } |
898 | 52d893ec | aliguori | |
899 | 52d893ec | aliguori | /*
|
900 | 52d893ec | aliguori | * alloc_compressed_cluster_offset
|
901 | 52d893ec | aliguori | *
|
902 | 52d893ec | aliguori | * For a given offset of the disk image, return cluster offset in
|
903 | 52d893ec | aliguori | * qcow2 file.
|
904 | 52d893ec | aliguori | *
|
905 | 52d893ec | aliguori | * If the offset is not found, allocate a new compressed cluster.
|
906 | 52d893ec | aliguori | *
|
907 | 52d893ec | aliguori | * Return the cluster offset if successful,
|
908 | 52d893ec | aliguori | * Return 0, otherwise.
|
909 | 52d893ec | aliguori | *
|
910 | 52d893ec | aliguori | */
|
911 | 52d893ec | aliguori | |
912 | 52d893ec | aliguori | static uint64_t alloc_compressed_cluster_offset(BlockDriverState *bs,
|
913 | 52d893ec | aliguori | uint64_t offset, |
914 | 52d893ec | aliguori | int compressed_size)
|
915 | 52d893ec | aliguori | { |
916 | 52d893ec | aliguori | BDRVQcowState *s = bs->opaque; |
917 | 52d893ec | aliguori | int l2_index, ret;
|
918 | 52d893ec | aliguori | uint64_t l2_offset, *l2_table, cluster_offset; |
919 | 52d893ec | aliguori | int nb_csectors;
|
920 | 52d893ec | aliguori | |
921 | 52d893ec | aliguori | ret = get_cluster_table(bs, offset, &l2_table, &l2_offset, &l2_index); |
922 | 52d893ec | aliguori | if (ret == 0) |
923 | 52d893ec | aliguori | return 0; |
924 | 52d893ec | aliguori | |
925 | 52d893ec | aliguori | cluster_offset = be64_to_cpu(l2_table[l2_index]); |
926 | 05203524 | aliguori | if (cluster_offset & QCOW_OFLAG_COPIED)
|
927 | 05203524 | aliguori | return cluster_offset & ~QCOW_OFLAG_COPIED;
|
928 | 05203524 | aliguori | |
929 | 095a9c58 | aliguori | if (cluster_offset)
|
930 | 095a9c58 | aliguori | free_any_clusters(bs, cluster_offset, 1);
|
931 | 108534b9 | aliguori | |
932 | 52d893ec | aliguori | cluster_offset = alloc_bytes(bs, compressed_size); |
933 | 52d893ec | aliguori | nb_csectors = ((cluster_offset + compressed_size - 1) >> 9) - |
934 | 52d893ec | aliguori | (cluster_offset >> 9);
|
935 | 05203524 | aliguori | |
936 | 52d893ec | aliguori | cluster_offset |= QCOW_OFLAG_COMPRESSED | |
937 | 52d893ec | aliguori | ((uint64_t)nb_csectors << s->csize_shift); |
938 | 05203524 | aliguori | |
939 | 52d893ec | aliguori | /* update L2 table */
|
940 | 05203524 | aliguori | |
941 | 52d893ec | aliguori | /* compressed clusters never have the copied flag */
|
942 | 05203524 | aliguori | |
943 | 52d893ec | aliguori | l2_table[l2_index] = cpu_to_be64(cluster_offset); |
944 | 52d893ec | aliguori | if (bdrv_pwrite(s->hd,
|
945 | 52d893ec | aliguori | l2_offset + l2_index * sizeof(uint64_t),
|
946 | 52d893ec | aliguori | l2_table + l2_index, |
947 | 52d893ec | aliguori | sizeof(uint64_t)) != sizeof(uint64_t)) |
948 | 52d893ec | aliguori | return 0; |
949 | 05203524 | aliguori | |
950 | 52d893ec | aliguori | return cluster_offset;
|
951 | 52d893ec | aliguori | } |
952 | 05203524 | aliguori | |
953 | e976c6a1 | aliguori | typedef struct QCowL2Meta |
954 | e976c6a1 | aliguori | { |
955 | e976c6a1 | aliguori | uint64_t offset; |
956 | e976c6a1 | aliguori | int n_start;
|
957 | e976c6a1 | aliguori | int nb_available;
|
958 | e976c6a1 | aliguori | int nb_clusters;
|
959 | e976c6a1 | aliguori | } QCowL2Meta; |
960 | e976c6a1 | aliguori | |
961 | e976c6a1 | aliguori | static int alloc_cluster_link_l2(BlockDriverState *bs, uint64_t cluster_offset, |
962 | e976c6a1 | aliguori | QCowL2Meta *m) |
963 | e976c6a1 | aliguori | { |
964 | e976c6a1 | aliguori | BDRVQcowState *s = bs->opaque; |
965 | e976c6a1 | aliguori | int i, j = 0, l2_index, ret; |
966 | e976c6a1 | aliguori | uint64_t *old_cluster, start_sect, l2_offset, *l2_table; |
967 | e976c6a1 | aliguori | |
968 | e976c6a1 | aliguori | if (m->nb_clusters == 0) |
969 | e976c6a1 | aliguori | return 0; |
970 | e976c6a1 | aliguori | |
971 | 3ec88e80 | aliguori | old_cluster = qemu_malloc(m->nb_clusters * sizeof(uint64_t));
|
972 | e976c6a1 | aliguori | |
973 | e976c6a1 | aliguori | /* copy content of unmodified sectors */
|
974 | e976c6a1 | aliguori | start_sect = (m->offset & ~(s->cluster_size - 1)) >> 9; |
975 | e976c6a1 | aliguori | if (m->n_start) {
|
976 | e976c6a1 | aliguori | ret = copy_sectors(bs, start_sect, cluster_offset, 0, m->n_start);
|
977 | e976c6a1 | aliguori | if (ret < 0) |
978 | e976c6a1 | aliguori | goto err;
|
979 | e976c6a1 | aliguori | } |
980 | e976c6a1 | aliguori | |
981 | e976c6a1 | aliguori | if (m->nb_available & (s->cluster_sectors - 1)) { |
982 | e976c6a1 | aliguori | uint64_t end = m->nb_available & ~(uint64_t)(s->cluster_sectors - 1);
|
983 | e976c6a1 | aliguori | ret = copy_sectors(bs, start_sect + end, cluster_offset + (end << 9),
|
984 | e976c6a1 | aliguori | m->nb_available - end, s->cluster_sectors); |
985 | e976c6a1 | aliguori | if (ret < 0) |
986 | e976c6a1 | aliguori | goto err;
|
987 | e976c6a1 | aliguori | } |
988 | e976c6a1 | aliguori | |
989 | e976c6a1 | aliguori | ret = -EIO; |
990 | e976c6a1 | aliguori | /* update L2 table */
|
991 | e976c6a1 | aliguori | if (!get_cluster_table(bs, m->offset, &l2_table, &l2_offset, &l2_index))
|
992 | e976c6a1 | aliguori | goto err;
|
993 | e976c6a1 | aliguori | |
994 | e976c6a1 | aliguori | for (i = 0; i < m->nb_clusters; i++) { |
995 | e976c6a1 | aliguori | if(l2_table[l2_index + i] != 0) |
996 | e976c6a1 | aliguori | old_cluster[j++] = l2_table[l2_index + i]; |
997 | e976c6a1 | aliguori | |
998 | e976c6a1 | aliguori | l2_table[l2_index + i] = cpu_to_be64((cluster_offset + |
999 | e976c6a1 | aliguori | (i << s->cluster_bits)) | QCOW_OFLAG_COPIED); |
1000 | e976c6a1 | aliguori | } |
1001 | e976c6a1 | aliguori | |
1002 | e976c6a1 | aliguori | if (bdrv_pwrite(s->hd, l2_offset + l2_index * sizeof(uint64_t), |
1003 | e976c6a1 | aliguori | l2_table + l2_index, m->nb_clusters * sizeof(uint64_t)) !=
|
1004 | e976c6a1 | aliguori | m->nb_clusters * sizeof(uint64_t))
|
1005 | e976c6a1 | aliguori | goto err;
|
1006 | e976c6a1 | aliguori | |
1007 | e976c6a1 | aliguori | for (i = 0; i < j; i++) |
1008 | 641636d1 | aliguori | free_any_clusters(bs, be64_to_cpu(old_cluster[i]), 1);
|
1009 | e976c6a1 | aliguori | |
1010 | e976c6a1 | aliguori | ret = 0;
|
1011 | e976c6a1 | aliguori | err:
|
1012 | e976c6a1 | aliguori | qemu_free(old_cluster); |
1013 | e976c6a1 | aliguori | return ret;
|
1014 | e976c6a1 | aliguori | } |
1015 | e976c6a1 | aliguori | |
1016 | 52d893ec | aliguori | /*
|
1017 | 52d893ec | aliguori | * alloc_cluster_offset
|
1018 | 52d893ec | aliguori | *
|
1019 | 52d893ec | aliguori | * For a given offset of the disk image, return cluster offset in
|
1020 | 52d893ec | aliguori | * qcow2 file.
|
1021 | 52d893ec | aliguori | *
|
1022 | 52d893ec | aliguori | * If the offset is not found, allocate a new cluster.
|
1023 | 52d893ec | aliguori | *
|
1024 | 52d893ec | aliguori | * Return the cluster offset if successful,
|
1025 | 52d893ec | aliguori | * Return 0, otherwise.
|
1026 | 52d893ec | aliguori | *
|
1027 | 52d893ec | aliguori | */
|
1028 | 52d893ec | aliguori | |
1029 | 52d893ec | aliguori | static uint64_t alloc_cluster_offset(BlockDriverState *bs,
|
1030 | 52d893ec | aliguori | uint64_t offset, |
1031 | 095a9c58 | aliguori | int n_start, int n_end, |
1032 | e976c6a1 | aliguori | int *num, QCowL2Meta *m)
|
1033 | 52d893ec | aliguori | { |
1034 | 52d893ec | aliguori | BDRVQcowState *s = bs->opaque; |
1035 | 52d893ec | aliguori | int l2_index, ret;
|
1036 | 52d893ec | aliguori | uint64_t l2_offset, *l2_table, cluster_offset; |
1037 | e976c6a1 | aliguori | int nb_clusters, i = 0; |
1038 | 52d893ec | aliguori | |
1039 | 52d893ec | aliguori | ret = get_cluster_table(bs, offset, &l2_table, &l2_offset, &l2_index); |
1040 | 52d893ec | aliguori | if (ret == 0) |
1041 | 52d893ec | aliguori | return 0; |
1042 | 52d893ec | aliguori | |
1043 | 6db6c638 | aliguori | nb_clusters = size_to_clusters(s, n_end << 9);
|
1044 | 6db6c638 | aliguori | |
1045 | e976c6a1 | aliguori | nb_clusters = MIN(nb_clusters, s->l2_size - l2_index); |
1046 | 095a9c58 | aliguori | |
1047 | 52d893ec | aliguori | cluster_offset = be64_to_cpu(l2_table[l2_index]); |
1048 | 52d893ec | aliguori | |
1049 | 095a9c58 | aliguori | /* We keep all QCOW_OFLAG_COPIED clusters */
|
1050 | 095a9c58 | aliguori | |
1051 | 095a9c58 | aliguori | if (cluster_offset & QCOW_OFLAG_COPIED) {
|
1052 | 6db6c638 | aliguori | nb_clusters = count_contiguous_clusters(nb_clusters, s->cluster_size, |
1053 | ff4b91c2 | aliguori | &l2_table[l2_index], 0, 0); |
1054 | 095a9c58 | aliguori | |
1055 | 095a9c58 | aliguori | cluster_offset &= ~QCOW_OFLAG_COPIED; |
1056 | e976c6a1 | aliguori | m->nb_clusters = 0;
|
1057 | 095a9c58 | aliguori | |
1058 | 095a9c58 | aliguori | goto out;
|
1059 | 095a9c58 | aliguori | } |
1060 | 095a9c58 | aliguori | |
1061 | 095a9c58 | aliguori | /* for the moment, multiple compressed clusters are not managed */
|
1062 | 095a9c58 | aliguori | |
1063 | 095a9c58 | aliguori | if (cluster_offset & QCOW_OFLAG_COMPRESSED)
|
1064 | 095a9c58 | aliguori | nb_clusters = 1;
|
1065 | 095a9c58 | aliguori | |
1066 | bc352085 | aliguori | /* how many available clusters ? */
|
1067 | 095a9c58 | aliguori | |
1068 | bc352085 | aliguori | while (i < nb_clusters) {
|
1069 | ab5ccbd6 | aliguori | i += count_contiguous_clusters(nb_clusters - i, s->cluster_size, |
1070 | ff4b91c2 | aliguori | &l2_table[l2_index], i, 0);
|
1071 | ab5ccbd6 | aliguori | |
1072 | ab5ccbd6 | aliguori | if(be64_to_cpu(l2_table[l2_index + i]))
|
1073 | ab5ccbd6 | aliguori | break;
|
1074 | ab5ccbd6 | aliguori | |
1075 | 6db6c638 | aliguori | i += count_contiguous_free_clusters(nb_clusters - i, |
1076 | 6db6c638 | aliguori | &l2_table[l2_index + i]); |
1077 | 095a9c58 | aliguori | |
1078 | 6db6c638 | aliguori | cluster_offset = be64_to_cpu(l2_table[l2_index + i]); |
1079 | 095a9c58 | aliguori | |
1080 | 6db6c638 | aliguori | if ((cluster_offset & QCOW_OFLAG_COPIED) ||
|
1081 | bc352085 | aliguori | (cluster_offset & QCOW_OFLAG_COMPRESSED)) |
1082 | 6db6c638 | aliguori | break;
|
1083 | 095a9c58 | aliguori | } |
1084 | bc352085 | aliguori | nb_clusters = i; |
1085 | 05203524 | aliguori | |
1086 | 05203524 | aliguori | /* allocate a new cluster */
|
1087 | 05203524 | aliguori | |
1088 | 095a9c58 | aliguori | cluster_offset = alloc_clusters(bs, nb_clusters * s->cluster_size); |
1089 | 05203524 | aliguori | |
1090 | e976c6a1 | aliguori | /* save info needed for meta data update */
|
1091 | e976c6a1 | aliguori | m->offset = offset; |
1092 | e976c6a1 | aliguori | m->n_start = n_start; |
1093 | e976c6a1 | aliguori | m->nb_clusters = nb_clusters; |
1094 | 05203524 | aliguori | |
1095 | 095a9c58 | aliguori | out:
|
1096 | e976c6a1 | aliguori | m->nb_available = MIN(nb_clusters << (s->cluster_bits - 9), n_end);
|
1097 | e976c6a1 | aliguori | |
1098 | e976c6a1 | aliguori | *num = m->nb_available - n_start; |
1099 | 095a9c58 | aliguori | |
1100 | 585f8587 | bellard | return cluster_offset;
|
1101 | 585f8587 | bellard | } |
1102 | 585f8587 | bellard | |
1103 | 5fafdf24 | ths | static int qcow_is_allocated(BlockDriverState *bs, int64_t sector_num, |
1104 | 585f8587 | bellard | int nb_sectors, int *pnum) |
1105 | 585f8587 | bellard | { |
1106 | 585f8587 | bellard | uint64_t cluster_offset; |
1107 | 585f8587 | bellard | |
1108 | 095a9c58 | aliguori | *pnum = nb_sectors; |
1109 | 095a9c58 | aliguori | cluster_offset = get_cluster_offset(bs, sector_num << 9, pnum);
|
1110 | 095a9c58 | aliguori | |
1111 | 585f8587 | bellard | return (cluster_offset != 0); |
1112 | 585f8587 | bellard | } |
1113 | 585f8587 | bellard | |
1114 | 585f8587 | bellard | static int decompress_buffer(uint8_t *out_buf, int out_buf_size, |
1115 | 585f8587 | bellard | const uint8_t *buf, int buf_size) |
1116 | 585f8587 | bellard | { |
1117 | 585f8587 | bellard | z_stream strm1, *strm = &strm1; |
1118 | 585f8587 | bellard | int ret, out_len;
|
1119 | 585f8587 | bellard | |
1120 | 585f8587 | bellard | memset(strm, 0, sizeof(*strm)); |
1121 | 585f8587 | bellard | |
1122 | 585f8587 | bellard | strm->next_in = (uint8_t *)buf; |
1123 | 585f8587 | bellard | strm->avail_in = buf_size; |
1124 | 585f8587 | bellard | strm->next_out = out_buf; |
1125 | 585f8587 | bellard | strm->avail_out = out_buf_size; |
1126 | 585f8587 | bellard | |
1127 | 585f8587 | bellard | ret = inflateInit2(strm, -12);
|
1128 | 585f8587 | bellard | if (ret != Z_OK)
|
1129 | 585f8587 | bellard | return -1; |
1130 | 585f8587 | bellard | ret = inflate(strm, Z_FINISH); |
1131 | 585f8587 | bellard | out_len = strm->next_out - out_buf; |
1132 | 585f8587 | bellard | if ((ret != Z_STREAM_END && ret != Z_BUF_ERROR) ||
|
1133 | 585f8587 | bellard | out_len != out_buf_size) { |
1134 | 585f8587 | bellard | inflateEnd(strm); |
1135 | 585f8587 | bellard | return -1; |
1136 | 585f8587 | bellard | } |
1137 | 585f8587 | bellard | inflateEnd(strm); |
1138 | 585f8587 | bellard | return 0; |
1139 | 585f8587 | bellard | } |
1140 | 3b46e624 | ths | |
1141 | 585f8587 | bellard | static int decompress_cluster(BDRVQcowState *s, uint64_t cluster_offset) |
1142 | 585f8587 | bellard | { |
1143 | 585f8587 | bellard | int ret, csize, nb_csectors, sector_offset;
|
1144 | 585f8587 | bellard | uint64_t coffset; |
1145 | 585f8587 | bellard | |
1146 | 585f8587 | bellard | coffset = cluster_offset & s->cluster_offset_mask; |
1147 | 585f8587 | bellard | if (s->cluster_cache_offset != coffset) {
|
1148 | 585f8587 | bellard | nb_csectors = ((cluster_offset >> s->csize_shift) & s->csize_mask) + 1;
|
1149 | 585f8587 | bellard | sector_offset = coffset & 511;
|
1150 | 585f8587 | bellard | csize = nb_csectors * 512 - sector_offset;
|
1151 | 585f8587 | bellard | ret = bdrv_read(s->hd, coffset >> 9, s->cluster_data, nb_csectors);
|
1152 | 585f8587 | bellard | if (ret < 0) { |
1153 | 585f8587 | bellard | return -1; |
1154 | 585f8587 | bellard | } |
1155 | 585f8587 | bellard | if (decompress_buffer(s->cluster_cache, s->cluster_size,
|
1156 | 585f8587 | bellard | s->cluster_data + sector_offset, csize) < 0) {
|
1157 | 585f8587 | bellard | return -1; |
1158 | 585f8587 | bellard | } |
1159 | 585f8587 | bellard | s->cluster_cache_offset = coffset; |
1160 | 585f8587 | bellard | } |
1161 | 585f8587 | bellard | return 0; |
1162 | 585f8587 | bellard | } |
1163 | 585f8587 | bellard | |
1164 | a9465922 | bellard | /* handle reading after the end of the backing file */
|
1165 | 5fafdf24 | ths | static int backing_read1(BlockDriverState *bs, |
1166 | a9465922 | bellard | int64_t sector_num, uint8_t *buf, int nb_sectors)
|
1167 | a9465922 | bellard | { |
1168 | a9465922 | bellard | int n1;
|
1169 | a9465922 | bellard | if ((sector_num + nb_sectors) <= bs->total_sectors)
|
1170 | a9465922 | bellard | return nb_sectors;
|
1171 | a9465922 | bellard | if (sector_num >= bs->total_sectors)
|
1172 | a9465922 | bellard | n1 = 0;
|
1173 | a9465922 | bellard | else
|
1174 | a9465922 | bellard | n1 = bs->total_sectors - sector_num; |
1175 | a9465922 | bellard | memset(buf + n1 * 512, 0, 512 * (nb_sectors - n1)); |
1176 | a9465922 | bellard | return n1;
|
1177 | a9465922 | bellard | } |
1178 | a9465922 | bellard | |
1179 | 5fafdf24 | ths | static int qcow_read(BlockDriverState *bs, int64_t sector_num, |
1180 | 585f8587 | bellard | uint8_t *buf, int nb_sectors)
|
1181 | 585f8587 | bellard | { |
1182 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1183 | a9465922 | bellard | int ret, index_in_cluster, n, n1;
|
1184 | 585f8587 | bellard | uint64_t cluster_offset; |
1185 | 3b46e624 | ths | |
1186 | 585f8587 | bellard | while (nb_sectors > 0) { |
1187 | 095a9c58 | aliguori | n = nb_sectors; |
1188 | 095a9c58 | aliguori | cluster_offset = get_cluster_offset(bs, sector_num << 9, &n);
|
1189 | 585f8587 | bellard | index_in_cluster = sector_num & (s->cluster_sectors - 1);
|
1190 | 585f8587 | bellard | if (!cluster_offset) {
|
1191 | 585f8587 | bellard | if (bs->backing_hd) {
|
1192 | 585f8587 | bellard | /* read from the base image */
|
1193 | a9465922 | bellard | n1 = backing_read1(bs->backing_hd, sector_num, buf, n); |
1194 | a9465922 | bellard | if (n1 > 0) { |
1195 | a9465922 | bellard | ret = bdrv_read(bs->backing_hd, sector_num, buf, n1); |
1196 | a9465922 | bellard | if (ret < 0) |
1197 | a9465922 | bellard | return -1; |
1198 | a9465922 | bellard | } |
1199 | 585f8587 | bellard | } else {
|
1200 | 585f8587 | bellard | memset(buf, 0, 512 * n); |
1201 | 585f8587 | bellard | } |
1202 | 585f8587 | bellard | } else if (cluster_offset & QCOW_OFLAG_COMPRESSED) { |
1203 | 585f8587 | bellard | if (decompress_cluster(s, cluster_offset) < 0) |
1204 | 585f8587 | bellard | return -1; |
1205 | 585f8587 | bellard | memcpy(buf, s->cluster_cache + index_in_cluster * 512, 512 * n); |
1206 | 585f8587 | bellard | } else {
|
1207 | 585f8587 | bellard | ret = bdrv_pread(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512); |
1208 | 5fafdf24 | ths | if (ret != n * 512) |
1209 | 585f8587 | bellard | return -1; |
1210 | 585f8587 | bellard | if (s->crypt_method) {
|
1211 | 5fafdf24 | ths | encrypt_sectors(s, sector_num, buf, buf, n, 0,
|
1212 | 585f8587 | bellard | &s->aes_decrypt_key); |
1213 | 585f8587 | bellard | } |
1214 | 585f8587 | bellard | } |
1215 | 585f8587 | bellard | nb_sectors -= n; |
1216 | 585f8587 | bellard | sector_num += n; |
1217 | 585f8587 | bellard | buf += n * 512;
|
1218 | 585f8587 | bellard | } |
1219 | 585f8587 | bellard | return 0; |
1220 | 585f8587 | bellard | } |
1221 | 585f8587 | bellard | |
1222 | 5fafdf24 | ths | static int qcow_write(BlockDriverState *bs, int64_t sector_num, |
1223 | 585f8587 | bellard | const uint8_t *buf, int nb_sectors) |
1224 | 585f8587 | bellard | { |
1225 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1226 | 585f8587 | bellard | int ret, index_in_cluster, n;
|
1227 | 585f8587 | bellard | uint64_t cluster_offset; |
1228 | 095a9c58 | aliguori | int n_end;
|
1229 | e976c6a1 | aliguori | QCowL2Meta l2meta; |
1230 | 3b46e624 | ths | |
1231 | 585f8587 | bellard | while (nb_sectors > 0) { |
1232 | 585f8587 | bellard | index_in_cluster = sector_num & (s->cluster_sectors - 1);
|
1233 | 095a9c58 | aliguori | n_end = index_in_cluster + nb_sectors; |
1234 | 095a9c58 | aliguori | if (s->crypt_method &&
|
1235 | 095a9c58 | aliguori | n_end > QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors) |
1236 | 095a9c58 | aliguori | n_end = QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors; |
1237 | 52d893ec | aliguori | cluster_offset = alloc_cluster_offset(bs, sector_num << 9,
|
1238 | 05203524 | aliguori | index_in_cluster, |
1239 | e976c6a1 | aliguori | n_end, &n, &l2meta); |
1240 | 585f8587 | bellard | if (!cluster_offset)
|
1241 | 585f8587 | bellard | return -1; |
1242 | 585f8587 | bellard | if (s->crypt_method) {
|
1243 | 585f8587 | bellard | encrypt_sectors(s, sector_num, s->cluster_data, buf, n, 1,
|
1244 | 585f8587 | bellard | &s->aes_encrypt_key); |
1245 | 5fafdf24 | ths | ret = bdrv_pwrite(s->hd, cluster_offset + index_in_cluster * 512,
|
1246 | 585f8587 | bellard | s->cluster_data, n * 512);
|
1247 | 585f8587 | bellard | } else {
|
1248 | 585f8587 | bellard | ret = bdrv_pwrite(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512); |
1249 | 585f8587 | bellard | } |
1250 | e976c6a1 | aliguori | if (ret != n * 512 || alloc_cluster_link_l2(bs, cluster_offset, &l2meta) < 0) { |
1251 | e976c6a1 | aliguori | free_any_clusters(bs, cluster_offset, l2meta.nb_clusters); |
1252 | 585f8587 | bellard | return -1; |
1253 | e976c6a1 | aliguori | } |
1254 | 585f8587 | bellard | nb_sectors -= n; |
1255 | 585f8587 | bellard | sector_num += n; |
1256 | 585f8587 | bellard | buf += n * 512;
|
1257 | 585f8587 | bellard | } |
1258 | 585f8587 | bellard | s->cluster_cache_offset = -1; /* disable compressed cache */ |
1259 | 585f8587 | bellard | return 0; |
1260 | 585f8587 | bellard | } |
1261 | 585f8587 | bellard | |
1262 | ce1a14dc | pbrook | typedef struct QCowAIOCB { |
1263 | ce1a14dc | pbrook | BlockDriverAIOCB common; |
1264 | 585f8587 | bellard | int64_t sector_num; |
1265 | f141eafe | aliguori | QEMUIOVector *qiov; |
1266 | 585f8587 | bellard | uint8_t *buf; |
1267 | f141eafe | aliguori | void *orig_buf;
|
1268 | 585f8587 | bellard | int nb_sectors;
|
1269 | 585f8587 | bellard | int n;
|
1270 | 585f8587 | bellard | uint64_t cluster_offset; |
1271 | 5fafdf24 | ths | uint8_t *cluster_data; |
1272 | 585f8587 | bellard | BlockDriverAIOCB *hd_aiocb; |
1273 | c87c0672 | aliguori | struct iovec hd_iov;
|
1274 | c87c0672 | aliguori | QEMUIOVector hd_qiov; |
1275 | 1490791f | aliguori | QEMUBH *bh; |
1276 | e976c6a1 | aliguori | QCowL2Meta l2meta; |
1277 | 585f8587 | bellard | } QCowAIOCB; |
1278 | 585f8587 | bellard | |
1279 | 1490791f | aliguori | static void qcow_aio_read_cb(void *opaque, int ret); |
1280 | 1490791f | aliguori | static void qcow_aio_read_bh(void *opaque) |
1281 | 1490791f | aliguori | { |
1282 | 1490791f | aliguori | QCowAIOCB *acb = opaque; |
1283 | 1490791f | aliguori | qemu_bh_delete(acb->bh); |
1284 | 1490791f | aliguori | acb->bh = NULL;
|
1285 | 1490791f | aliguori | qcow_aio_read_cb(opaque, 0);
|
1286 | 1490791f | aliguori | } |
1287 | 1490791f | aliguori | |
1288 | a32ef786 | aliguori | static int qcow_schedule_bh(QEMUBHFunc *cb, QCowAIOCB *acb) |
1289 | a32ef786 | aliguori | { |
1290 | a32ef786 | aliguori | if (acb->bh)
|
1291 | a32ef786 | aliguori | return -EIO;
|
1292 | a32ef786 | aliguori | |
1293 | a32ef786 | aliguori | acb->bh = qemu_bh_new(cb, acb); |
1294 | a32ef786 | aliguori | if (!acb->bh)
|
1295 | a32ef786 | aliguori | return -EIO;
|
1296 | a32ef786 | aliguori | |
1297 | a32ef786 | aliguori | qemu_bh_schedule(acb->bh); |
1298 | a32ef786 | aliguori | |
1299 | a32ef786 | aliguori | return 0; |
1300 | a32ef786 | aliguori | } |
1301 | a32ef786 | aliguori | |
1302 | 585f8587 | bellard | static void qcow_aio_read_cb(void *opaque, int ret) |
1303 | 585f8587 | bellard | { |
1304 | ce1a14dc | pbrook | QCowAIOCB *acb = opaque; |
1305 | ce1a14dc | pbrook | BlockDriverState *bs = acb->common.bs; |
1306 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1307 | a9465922 | bellard | int index_in_cluster, n1;
|
1308 | 585f8587 | bellard | |
1309 | ce1a14dc | pbrook | acb->hd_aiocb = NULL;
|
1310 | f141eafe | aliguori | if (ret < 0) |
1311 | f141eafe | aliguori | goto done;
|
1312 | 585f8587 | bellard | |
1313 | 585f8587 | bellard | /* post process the read buffer */
|
1314 | ce1a14dc | pbrook | if (!acb->cluster_offset) {
|
1315 | 585f8587 | bellard | /* nothing to do */
|
1316 | ce1a14dc | pbrook | } else if (acb->cluster_offset & QCOW_OFLAG_COMPRESSED) { |
1317 | 585f8587 | bellard | /* nothing to do */
|
1318 | 585f8587 | bellard | } else {
|
1319 | 585f8587 | bellard | if (s->crypt_method) {
|
1320 | 5fafdf24 | ths | encrypt_sectors(s, acb->sector_num, acb->buf, acb->buf, |
1321 | 5fafdf24 | ths | acb->n, 0,
|
1322 | 585f8587 | bellard | &s->aes_decrypt_key); |
1323 | 585f8587 | bellard | } |
1324 | 585f8587 | bellard | } |
1325 | 585f8587 | bellard | |
1326 | ce1a14dc | pbrook | acb->nb_sectors -= acb->n; |
1327 | ce1a14dc | pbrook | acb->sector_num += acb->n; |
1328 | ce1a14dc | pbrook | acb->buf += acb->n * 512;
|
1329 | 585f8587 | bellard | |
1330 | ce1a14dc | pbrook | if (acb->nb_sectors == 0) { |
1331 | 585f8587 | bellard | /* request completed */
|
1332 | f141eafe | aliguori | ret = 0;
|
1333 | f141eafe | aliguori | goto done;
|
1334 | 585f8587 | bellard | } |
1335 | 3b46e624 | ths | |
1336 | 585f8587 | bellard | /* prepare next AIO request */
|
1337 | 095a9c58 | aliguori | acb->n = acb->nb_sectors; |
1338 | 095a9c58 | aliguori | acb->cluster_offset = get_cluster_offset(bs, acb->sector_num << 9, &acb->n);
|
1339 | ce1a14dc | pbrook | index_in_cluster = acb->sector_num & (s->cluster_sectors - 1);
|
1340 | ce1a14dc | pbrook | |
1341 | ce1a14dc | pbrook | if (!acb->cluster_offset) {
|
1342 | 585f8587 | bellard | if (bs->backing_hd) {
|
1343 | 585f8587 | bellard | /* read from the base image */
|
1344 | 5fafdf24 | ths | n1 = backing_read1(bs->backing_hd, acb->sector_num, |
1345 | ce1a14dc | pbrook | acb->buf, acb->n); |
1346 | a9465922 | bellard | if (n1 > 0) { |
1347 | 3f4cb3d3 | blueswir1 | acb->hd_iov.iov_base = (void *)acb->buf;
|
1348 | c87c0672 | aliguori | acb->hd_iov.iov_len = acb->n * 512;
|
1349 | c87c0672 | aliguori | qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1);
|
1350 | c87c0672 | aliguori | acb->hd_aiocb = bdrv_aio_readv(bs->backing_hd, acb->sector_num, |
1351 | c87c0672 | aliguori | &acb->hd_qiov, acb->n, |
1352 | c87c0672 | aliguori | qcow_aio_read_cb, acb); |
1353 | ce1a14dc | pbrook | if (acb->hd_aiocb == NULL) |
1354 | f141eafe | aliguori | goto done;
|
1355 | a9465922 | bellard | } else {
|
1356 | a32ef786 | aliguori | ret = qcow_schedule_bh(qcow_aio_read_bh, acb); |
1357 | a32ef786 | aliguori | if (ret < 0) |
1358 | f141eafe | aliguori | goto done;
|
1359 | a9465922 | bellard | } |
1360 | 585f8587 | bellard | } else {
|
1361 | 585f8587 | bellard | /* Note: in this case, no need to wait */
|
1362 | ce1a14dc | pbrook | memset(acb->buf, 0, 512 * acb->n); |
1363 | a32ef786 | aliguori | ret = qcow_schedule_bh(qcow_aio_read_bh, acb); |
1364 | a32ef786 | aliguori | if (ret < 0) |
1365 | f141eafe | aliguori | goto done;
|
1366 | 585f8587 | bellard | } |
1367 | ce1a14dc | pbrook | } else if (acb->cluster_offset & QCOW_OFLAG_COMPRESSED) { |
1368 | 585f8587 | bellard | /* add AIO support for compressed blocks ? */
|
1369 | ce1a14dc | pbrook | if (decompress_cluster(s, acb->cluster_offset) < 0) |
1370 | f141eafe | aliguori | goto done;
|
1371 | 5fafdf24 | ths | memcpy(acb->buf, |
1372 | ce1a14dc | pbrook | s->cluster_cache + index_in_cluster * 512, 512 * acb->n); |
1373 | a32ef786 | aliguori | ret = qcow_schedule_bh(qcow_aio_read_bh, acb); |
1374 | a32ef786 | aliguori | if (ret < 0) |
1375 | f141eafe | aliguori | goto done;
|
1376 | 585f8587 | bellard | } else {
|
1377 | ce1a14dc | pbrook | if ((acb->cluster_offset & 511) != 0) { |
1378 | 585f8587 | bellard | ret = -EIO; |
1379 | f141eafe | aliguori | goto done;
|
1380 | 585f8587 | bellard | } |
1381 | c87c0672 | aliguori | |
1382 | 3f4cb3d3 | blueswir1 | acb->hd_iov.iov_base = (void *)acb->buf;
|
1383 | c87c0672 | aliguori | acb->hd_iov.iov_len = acb->n * 512;
|
1384 | c87c0672 | aliguori | qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1);
|
1385 | c87c0672 | aliguori | acb->hd_aiocb = bdrv_aio_readv(s->hd, |
1386 | 5fafdf24 | ths | (acb->cluster_offset >> 9) + index_in_cluster,
|
1387 | c87c0672 | aliguori | &acb->hd_qiov, acb->n, qcow_aio_read_cb, acb); |
1388 | ce1a14dc | pbrook | if (acb->hd_aiocb == NULL) |
1389 | f141eafe | aliguori | goto done;
|
1390 | f141eafe | aliguori | } |
1391 | f141eafe | aliguori | |
1392 | f141eafe | aliguori | return;
|
1393 | f141eafe | aliguori | done:
|
1394 | f141eafe | aliguori | if (acb->qiov->niov > 1) { |
1395 | f141eafe | aliguori | qemu_iovec_from_buffer(acb->qiov, acb->orig_buf, acb->qiov->size); |
1396 | f141eafe | aliguori | qemu_vfree(acb->orig_buf); |
1397 | 585f8587 | bellard | } |
1398 | f141eafe | aliguori | acb->common.cb(acb->common.opaque, ret); |
1399 | f141eafe | aliguori | qemu_aio_release(acb); |
1400 | 585f8587 | bellard | } |
1401 | 585f8587 | bellard | |
1402 | ce1a14dc | pbrook | static QCowAIOCB *qcow_aio_setup(BlockDriverState *bs,
|
1403 | f141eafe | aliguori | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
|
1404 | f141eafe | aliguori | BlockDriverCompletionFunc *cb, void *opaque, int is_write) |
1405 | 585f8587 | bellard | { |
1406 | ce1a14dc | pbrook | QCowAIOCB *acb; |
1407 | ce1a14dc | pbrook | |
1408 | ce1a14dc | pbrook | acb = qemu_aio_get(bs, cb, opaque); |
1409 | ce1a14dc | pbrook | if (!acb)
|
1410 | ce1a14dc | pbrook | return NULL; |
1411 | ce1a14dc | pbrook | acb->hd_aiocb = NULL;
|
1412 | ce1a14dc | pbrook | acb->sector_num = sector_num; |
1413 | f141eafe | aliguori | acb->qiov = qiov; |
1414 | f141eafe | aliguori | if (qiov->niov > 1) { |
1415 | f141eafe | aliguori | acb->buf = acb->orig_buf = qemu_memalign(512, qiov->size);
|
1416 | f141eafe | aliguori | if (is_write)
|
1417 | f141eafe | aliguori | qemu_iovec_to_buffer(qiov, acb->buf); |
1418 | 3f4cb3d3 | blueswir1 | } else {
|
1419 | 3f4cb3d3 | blueswir1 | acb->buf = (uint8_t *)qiov->iov->iov_base; |
1420 | 3f4cb3d3 | blueswir1 | } |
1421 | ce1a14dc | pbrook | acb->nb_sectors = nb_sectors; |
1422 | ce1a14dc | pbrook | acb->n = 0;
|
1423 | ce1a14dc | pbrook | acb->cluster_offset = 0;
|
1424 | e976c6a1 | aliguori | acb->l2meta.nb_clusters = 0;
|
1425 | ce1a14dc | pbrook | return acb;
|
1426 | ce1a14dc | pbrook | } |
1427 | ce1a14dc | pbrook | |
1428 | f141eafe | aliguori | static BlockDriverAIOCB *qcow_aio_readv(BlockDriverState *bs,
|
1429 | f141eafe | aliguori | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
|
1430 | ce1a14dc | pbrook | BlockDriverCompletionFunc *cb, void *opaque)
|
1431 | ce1a14dc | pbrook | { |
1432 | ce1a14dc | pbrook | QCowAIOCB *acb; |
1433 | ce1a14dc | pbrook | |
1434 | f141eafe | aliguori | acb = qcow_aio_setup(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
|
1435 | ce1a14dc | pbrook | if (!acb)
|
1436 | ce1a14dc | pbrook | return NULL; |
1437 | 585f8587 | bellard | |
1438 | 585f8587 | bellard | qcow_aio_read_cb(acb, 0);
|
1439 | ce1a14dc | pbrook | return &acb->common;
|
1440 | 585f8587 | bellard | } |
1441 | 585f8587 | bellard | |
1442 | 585f8587 | bellard | static void qcow_aio_write_cb(void *opaque, int ret) |
1443 | 585f8587 | bellard | { |
1444 | ce1a14dc | pbrook | QCowAIOCB *acb = opaque; |
1445 | ce1a14dc | pbrook | BlockDriverState *bs = acb->common.bs; |
1446 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1447 | 585f8587 | bellard | int index_in_cluster;
|
1448 | 585f8587 | bellard | const uint8_t *src_buf;
|
1449 | 095a9c58 | aliguori | int n_end;
|
1450 | ce1a14dc | pbrook | |
1451 | ce1a14dc | pbrook | acb->hd_aiocb = NULL;
|
1452 | ce1a14dc | pbrook | |
1453 | f141eafe | aliguori | if (ret < 0) |
1454 | f141eafe | aliguori | goto done;
|
1455 | 585f8587 | bellard | |
1456 | e976c6a1 | aliguori | if (alloc_cluster_link_l2(bs, acb->cluster_offset, &acb->l2meta) < 0) { |
1457 | e976c6a1 | aliguori | free_any_clusters(bs, acb->cluster_offset, acb->l2meta.nb_clusters); |
1458 | f141eafe | aliguori | goto done;
|
1459 | e976c6a1 | aliguori | } |
1460 | e976c6a1 | aliguori | |
1461 | ce1a14dc | pbrook | acb->nb_sectors -= acb->n; |
1462 | ce1a14dc | pbrook | acb->sector_num += acb->n; |
1463 | ce1a14dc | pbrook | acb->buf += acb->n * 512;
|
1464 | 585f8587 | bellard | |
1465 | ce1a14dc | pbrook | if (acb->nb_sectors == 0) { |
1466 | 585f8587 | bellard | /* request completed */
|
1467 | f141eafe | aliguori | ret = 0;
|
1468 | f141eafe | aliguori | goto done;
|
1469 | 585f8587 | bellard | } |
1470 | 3b46e624 | ths | |
1471 | ce1a14dc | pbrook | index_in_cluster = acb->sector_num & (s->cluster_sectors - 1);
|
1472 | 095a9c58 | aliguori | n_end = index_in_cluster + acb->nb_sectors; |
1473 | 095a9c58 | aliguori | if (s->crypt_method &&
|
1474 | 095a9c58 | aliguori | n_end > QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors) |
1475 | 095a9c58 | aliguori | n_end = QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors; |
1476 | 095a9c58 | aliguori | |
1477 | e976c6a1 | aliguori | acb->cluster_offset = alloc_cluster_offset(bs, acb->sector_num << 9,
|
1478 | 05203524 | aliguori | index_in_cluster, |
1479 | e976c6a1 | aliguori | n_end, &acb->n, &acb->l2meta); |
1480 | e976c6a1 | aliguori | if (!acb->cluster_offset || (acb->cluster_offset & 511) != 0) { |
1481 | 585f8587 | bellard | ret = -EIO; |
1482 | f141eafe | aliguori | goto done;
|
1483 | 585f8587 | bellard | } |
1484 | 585f8587 | bellard | if (s->crypt_method) {
|
1485 | ce1a14dc | pbrook | if (!acb->cluster_data) {
|
1486 | 095a9c58 | aliguori | acb->cluster_data = qemu_mallocz(QCOW_MAX_CRYPT_CLUSTERS * |
1487 | 095a9c58 | aliguori | s->cluster_size); |
1488 | 585f8587 | bellard | } |
1489 | 5fafdf24 | ths | encrypt_sectors(s, acb->sector_num, acb->cluster_data, acb->buf, |
1490 | ce1a14dc | pbrook | acb->n, 1, &s->aes_encrypt_key);
|
1491 | ce1a14dc | pbrook | src_buf = acb->cluster_data; |
1492 | 585f8587 | bellard | } else {
|
1493 | ce1a14dc | pbrook | src_buf = acb->buf; |
1494 | 585f8587 | bellard | } |
1495 | c87c0672 | aliguori | acb->hd_iov.iov_base = (void *)src_buf;
|
1496 | c87c0672 | aliguori | acb->hd_iov.iov_len = acb->n * 512;
|
1497 | c87c0672 | aliguori | qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1);
|
1498 | c87c0672 | aliguori | acb->hd_aiocb = bdrv_aio_writev(s->hd, |
1499 | c87c0672 | aliguori | (acb->cluster_offset >> 9) + index_in_cluster,
|
1500 | c87c0672 | aliguori | &acb->hd_qiov, acb->n, |
1501 | c87c0672 | aliguori | qcow_aio_write_cb, acb); |
1502 | ce1a14dc | pbrook | if (acb->hd_aiocb == NULL) |
1503 | f141eafe | aliguori | goto done;
|
1504 | f141eafe | aliguori | |
1505 | f141eafe | aliguori | return;
|
1506 | f141eafe | aliguori | |
1507 | f141eafe | aliguori | done:
|
1508 | f141eafe | aliguori | if (acb->qiov->niov > 1) |
1509 | f141eafe | aliguori | qemu_vfree(acb->orig_buf); |
1510 | f141eafe | aliguori | acb->common.cb(acb->common.opaque, ret); |
1511 | f141eafe | aliguori | qemu_aio_release(acb); |
1512 | 585f8587 | bellard | } |
1513 | 585f8587 | bellard | |
1514 | f141eafe | aliguori | static BlockDriverAIOCB *qcow_aio_writev(BlockDriverState *bs,
|
1515 | f141eafe | aliguori | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
|
1516 | ce1a14dc | pbrook | BlockDriverCompletionFunc *cb, void *opaque)
|
1517 | 585f8587 | bellard | { |
1518 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1519 | ce1a14dc | pbrook | QCowAIOCB *acb; |
1520 | 3b46e624 | ths | |
1521 | 585f8587 | bellard | s->cluster_cache_offset = -1; /* disable compressed cache */ |
1522 | 585f8587 | bellard | |
1523 | f141eafe | aliguori | acb = qcow_aio_setup(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
|
1524 | ce1a14dc | pbrook | if (!acb)
|
1525 | ce1a14dc | pbrook | return NULL; |
1526 | 3b46e624 | ths | |
1527 | 585f8587 | bellard | qcow_aio_write_cb(acb, 0);
|
1528 | ce1a14dc | pbrook | return &acb->common;
|
1529 | 585f8587 | bellard | } |
1530 | 585f8587 | bellard | |
1531 | ce1a14dc | pbrook | static void qcow_aio_cancel(BlockDriverAIOCB *blockacb) |
1532 | 585f8587 | bellard | { |
1533 | ce1a14dc | pbrook | QCowAIOCB *acb = (QCowAIOCB *)blockacb; |
1534 | ce1a14dc | pbrook | if (acb->hd_aiocb)
|
1535 | ce1a14dc | pbrook | bdrv_aio_cancel(acb->hd_aiocb); |
1536 | ce1a14dc | pbrook | qemu_aio_release(acb); |
1537 | 585f8587 | bellard | } |
1538 | 585f8587 | bellard | |
1539 | 585f8587 | bellard | static void qcow_close(BlockDriverState *bs) |
1540 | 585f8587 | bellard | { |
1541 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1542 | 585f8587 | bellard | qemu_free(s->l1_table); |
1543 | 585f8587 | bellard | qemu_free(s->l2_cache); |
1544 | 585f8587 | bellard | qemu_free(s->cluster_cache); |
1545 | 585f8587 | bellard | qemu_free(s->cluster_data); |
1546 | 585f8587 | bellard | refcount_close(bs); |
1547 | 585f8587 | bellard | bdrv_delete(s->hd); |
1548 | 585f8587 | bellard | } |
1549 | 585f8587 | bellard | |
1550 | 585f8587 | bellard | /* XXX: use std qcow open function ? */
|
1551 | 585f8587 | bellard | typedef struct QCowCreateState { |
1552 | 585f8587 | bellard | int cluster_size;
|
1553 | 585f8587 | bellard | int cluster_bits;
|
1554 | 585f8587 | bellard | uint16_t *refcount_block; |
1555 | 585f8587 | bellard | uint64_t *refcount_table; |
1556 | 585f8587 | bellard | int64_t l1_table_offset; |
1557 | 585f8587 | bellard | int64_t refcount_table_offset; |
1558 | 585f8587 | bellard | int64_t refcount_block_offset; |
1559 | 585f8587 | bellard | } QCowCreateState; |
1560 | 585f8587 | bellard | |
1561 | 585f8587 | bellard | static void create_refcount_update(QCowCreateState *s, |
1562 | 585f8587 | bellard | int64_t offset, int64_t size) |
1563 | 585f8587 | bellard | { |
1564 | 585f8587 | bellard | int refcount;
|
1565 | 585f8587 | bellard | int64_t start, last, cluster_offset; |
1566 | 585f8587 | bellard | uint16_t *p; |
1567 | 585f8587 | bellard | |
1568 | 585f8587 | bellard | start = offset & ~(s->cluster_size - 1);
|
1569 | 585f8587 | bellard | last = (offset + size - 1) & ~(s->cluster_size - 1); |
1570 | 5fafdf24 | ths | for(cluster_offset = start; cluster_offset <= last;
|
1571 | 585f8587 | bellard | cluster_offset += s->cluster_size) { |
1572 | 585f8587 | bellard | p = &s->refcount_block[cluster_offset >> s->cluster_bits]; |
1573 | 585f8587 | bellard | refcount = be16_to_cpu(*p); |
1574 | 585f8587 | bellard | refcount++; |
1575 | 585f8587 | bellard | *p = cpu_to_be16(refcount); |
1576 | 585f8587 | bellard | } |
1577 | 585f8587 | bellard | } |
1578 | 585f8587 | bellard | |
1579 | f965509c | aliguori | static int qcow_create2(const char *filename, int64_t total_size, |
1580 | f965509c | aliguori | const char *backing_file, const char *backing_format, |
1581 | f965509c | aliguori | int flags)
|
1582 | 585f8587 | bellard | { |
1583 | f965509c | aliguori | |
1584 | 585f8587 | bellard | int fd, header_size, backing_filename_len, l1_size, i, shift, l2_bits;
|
1585 | 2d2431f0 | aliguori | int ref_clusters, backing_format_len = 0; |
1586 | 585f8587 | bellard | QCowHeader header; |
1587 | 585f8587 | bellard | uint64_t tmp, offset; |
1588 | 585f8587 | bellard | QCowCreateState s1, *s = &s1; |
1589 | f965509c | aliguori | QCowExtension ext_bf = {0, 0}; |
1590 | f965509c | aliguori | |
1591 | 3b46e624 | ths | |
1592 | 585f8587 | bellard | memset(s, 0, sizeof(*s)); |
1593 | 585f8587 | bellard | |
1594 | 585f8587 | bellard | fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644);
|
1595 | 585f8587 | bellard | if (fd < 0) |
1596 | 585f8587 | bellard | return -1; |
1597 | 585f8587 | bellard | memset(&header, 0, sizeof(header)); |
1598 | 585f8587 | bellard | header.magic = cpu_to_be32(QCOW_MAGIC); |
1599 | 585f8587 | bellard | header.version = cpu_to_be32(QCOW_VERSION); |
1600 | 585f8587 | bellard | header.size = cpu_to_be64(total_size * 512);
|
1601 | 585f8587 | bellard | header_size = sizeof(header);
|
1602 | 585f8587 | bellard | backing_filename_len = 0;
|
1603 | 585f8587 | bellard | if (backing_file) {
|
1604 | f965509c | aliguori | if (backing_format) {
|
1605 | f965509c | aliguori | ext_bf.magic = QCOW_EXT_MAGIC_BACKING_FORMAT; |
1606 | f965509c | aliguori | backing_format_len = strlen(backing_format); |
1607 | f965509c | aliguori | ext_bf.len = (backing_format_len + 7) & ~7; |
1608 | f965509c | aliguori | header_size += ((sizeof(ext_bf) + ext_bf.len + 7) & ~7); |
1609 | f965509c | aliguori | } |
1610 | 585f8587 | bellard | header.backing_file_offset = cpu_to_be64(header_size); |
1611 | 585f8587 | bellard | backing_filename_len = strlen(backing_file); |
1612 | 585f8587 | bellard | header.backing_file_size = cpu_to_be32(backing_filename_len); |
1613 | 585f8587 | bellard | header_size += backing_filename_len; |
1614 | 585f8587 | bellard | } |
1615 | 585f8587 | bellard | s->cluster_bits = 12; /* 4 KB clusters */ |
1616 | 585f8587 | bellard | s->cluster_size = 1 << s->cluster_bits;
|
1617 | 585f8587 | bellard | header.cluster_bits = cpu_to_be32(s->cluster_bits); |
1618 | 585f8587 | bellard | header_size = (header_size + 7) & ~7; |
1619 | ec36ba14 | ths | if (flags & BLOCK_FLAG_ENCRYPT) {
|
1620 | 585f8587 | bellard | header.crypt_method = cpu_to_be32(QCOW_CRYPT_AES); |
1621 | 585f8587 | bellard | } else {
|
1622 | 585f8587 | bellard | header.crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); |
1623 | 585f8587 | bellard | } |
1624 | 585f8587 | bellard | l2_bits = s->cluster_bits - 3;
|
1625 | 585f8587 | bellard | shift = s->cluster_bits + l2_bits; |
1626 | 585f8587 | bellard | l1_size = (((total_size * 512) + (1LL << shift) - 1) >> shift); |
1627 | 585f8587 | bellard | offset = align_offset(header_size, s->cluster_size); |
1628 | 585f8587 | bellard | s->l1_table_offset = offset; |
1629 | 585f8587 | bellard | header.l1_table_offset = cpu_to_be64(s->l1_table_offset); |
1630 | 585f8587 | bellard | header.l1_size = cpu_to_be32(l1_size); |
1631 | 15e6690a | bellard | offset += align_offset(l1_size * sizeof(uint64_t), s->cluster_size);
|
1632 | 585f8587 | bellard | |
1633 | 585f8587 | bellard | s->refcount_table = qemu_mallocz(s->cluster_size); |
1634 | 3b46e624 | ths | |
1635 | 585f8587 | bellard | s->refcount_table_offset = offset; |
1636 | 585f8587 | bellard | header.refcount_table_offset = cpu_to_be64(offset); |
1637 | 585f8587 | bellard | header.refcount_table_clusters = cpu_to_be32(1);
|
1638 | 585f8587 | bellard | offset += s->cluster_size; |
1639 | 585f8587 | bellard | s->refcount_block_offset = offset; |
1640 | 2d2431f0 | aliguori | |
1641 | 2d2431f0 | aliguori | /* count how many refcount blocks needed */
|
1642 | 2d2431f0 | aliguori | tmp = offset >> s->cluster_bits; |
1643 | 2d2431f0 | aliguori | ref_clusters = (tmp >> (s->cluster_bits - REFCOUNT_SHIFT)) + 1;
|
1644 | 2d2431f0 | aliguori | for (i=0; i < ref_clusters; i++) { |
1645 | 2d2431f0 | aliguori | s->refcount_table[i] = cpu_to_be64(offset); |
1646 | 2d2431f0 | aliguori | offset += s->cluster_size; |
1647 | 2d2431f0 | aliguori | } |
1648 | 2d2431f0 | aliguori | |
1649 | 2d2431f0 | aliguori | s->refcount_block = qemu_mallocz(ref_clusters * s->cluster_size); |
1650 | 585f8587 | bellard | |
1651 | 585f8587 | bellard | /* update refcounts */
|
1652 | 585f8587 | bellard | create_refcount_update(s, 0, header_size);
|
1653 | 15e6690a | bellard | create_refcount_update(s, s->l1_table_offset, l1_size * sizeof(uint64_t));
|
1654 | 585f8587 | bellard | create_refcount_update(s, s->refcount_table_offset, s->cluster_size); |
1655 | 2d2431f0 | aliguori | create_refcount_update(s, s->refcount_block_offset, ref_clusters * s->cluster_size); |
1656 | 3b46e624 | ths | |
1657 | 585f8587 | bellard | /* write all the data */
|
1658 | 585f8587 | bellard | write(fd, &header, sizeof(header));
|
1659 | 585f8587 | bellard | if (backing_file) {
|
1660 | f965509c | aliguori | if (backing_format_len) {
|
1661 | f965509c | aliguori | char zero[16]; |
1662 | f965509c | aliguori | int d = ext_bf.len - backing_format_len;
|
1663 | f965509c | aliguori | |
1664 | f965509c | aliguori | memset(zero, 0, sizeof(zero)); |
1665 | f965509c | aliguori | cpu_to_be32s(&ext_bf.magic); |
1666 | f965509c | aliguori | cpu_to_be32s(&ext_bf.len); |
1667 | f965509c | aliguori | write(fd, &ext_bf, sizeof(ext_bf));
|
1668 | f965509c | aliguori | write(fd, backing_format, backing_format_len); |
1669 | f965509c | aliguori | if (d>0) { |
1670 | f965509c | aliguori | write(fd, zero, d); |
1671 | f965509c | aliguori | } |
1672 | f965509c | aliguori | } |
1673 | 585f8587 | bellard | write(fd, backing_file, backing_filename_len); |
1674 | 585f8587 | bellard | } |
1675 | 585f8587 | bellard | lseek(fd, s->l1_table_offset, SEEK_SET); |
1676 | 585f8587 | bellard | tmp = 0;
|
1677 | 585f8587 | bellard | for(i = 0;i < l1_size; i++) { |
1678 | 585f8587 | bellard | write(fd, &tmp, sizeof(tmp));
|
1679 | 585f8587 | bellard | } |
1680 | 585f8587 | bellard | lseek(fd, s->refcount_table_offset, SEEK_SET); |
1681 | 585f8587 | bellard | write(fd, s->refcount_table, s->cluster_size); |
1682 | 3b46e624 | ths | |
1683 | 585f8587 | bellard | lseek(fd, s->refcount_block_offset, SEEK_SET); |
1684 | 2d2431f0 | aliguori | write(fd, s->refcount_block, ref_clusters * s->cluster_size); |
1685 | 585f8587 | bellard | |
1686 | 585f8587 | bellard | qemu_free(s->refcount_table); |
1687 | 585f8587 | bellard | qemu_free(s->refcount_block); |
1688 | 585f8587 | bellard | close(fd); |
1689 | 585f8587 | bellard | return 0; |
1690 | 585f8587 | bellard | } |
1691 | 585f8587 | bellard | |
1692 | f965509c | aliguori | static int qcow_create(const char *filename, int64_t total_size, |
1693 | f965509c | aliguori | const char *backing_file, int flags) |
1694 | f965509c | aliguori | { |
1695 | f965509c | aliguori | return qcow_create2(filename, total_size, backing_file, NULL, flags); |
1696 | f965509c | aliguori | } |
1697 | f965509c | aliguori | |
1698 | 585f8587 | bellard | static int qcow_make_empty(BlockDriverState *bs) |
1699 | 585f8587 | bellard | { |
1700 | 585f8587 | bellard | #if 0
|
1701 | 585f8587 | bellard | /* XXX: not correct */
|
1702 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque;
|
1703 | 585f8587 | bellard | uint32_t l1_length = s->l1_size * sizeof(uint64_t);
|
1704 | 585f8587 | bellard | int ret;
|
1705 | 585f8587 | bellard | |
1706 | 585f8587 | bellard | memset(s->l1_table, 0, l1_length);
|
1707 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, s->l1_table_offset, s->l1_table, l1_length) < 0)
|
1708 | ac674887 | aliguori | return -1;
|
1709 | 585f8587 | bellard | ret = bdrv_truncate(s->hd, s->l1_table_offset + l1_length);
|
1710 | 585f8587 | bellard | if (ret < 0)
|
1711 | 585f8587 | bellard | return ret;
|
1712 | 3b46e624 | ths | |
1713 | 585f8587 | bellard | l2_cache_reset(bs);
|
1714 | 585f8587 | bellard | #endif
|
1715 | 585f8587 | bellard | return 0; |
1716 | 585f8587 | bellard | } |
1717 | 585f8587 | bellard | |
1718 | 585f8587 | bellard | /* XXX: put compressed sectors first, then all the cluster aligned
|
1719 | 585f8587 | bellard | tables to avoid losing bytes in alignment */
|
1720 | 5fafdf24 | ths | static int qcow_write_compressed(BlockDriverState *bs, int64_t sector_num, |
1721 | 585f8587 | bellard | const uint8_t *buf, int nb_sectors) |
1722 | 585f8587 | bellard | { |
1723 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1724 | 585f8587 | bellard | z_stream strm; |
1725 | 585f8587 | bellard | int ret, out_len;
|
1726 | 585f8587 | bellard | uint8_t *out_buf; |
1727 | 585f8587 | bellard | uint64_t cluster_offset; |
1728 | 585f8587 | bellard | |
1729 | 585f8587 | bellard | if (nb_sectors == 0) { |
1730 | 585f8587 | bellard | /* align end of file to a sector boundary to ease reading with
|
1731 | 585f8587 | bellard | sector based I/Os */
|
1732 | 585f8587 | bellard | cluster_offset = bdrv_getlength(s->hd); |
1733 | 585f8587 | bellard | cluster_offset = (cluster_offset + 511) & ~511; |
1734 | 585f8587 | bellard | bdrv_truncate(s->hd, cluster_offset); |
1735 | 585f8587 | bellard | return 0; |
1736 | 585f8587 | bellard | } |
1737 | 585f8587 | bellard | |
1738 | 585f8587 | bellard | if (nb_sectors != s->cluster_sectors)
|
1739 | 585f8587 | bellard | return -EINVAL;
|
1740 | 585f8587 | bellard | |
1741 | 585f8587 | bellard | out_buf = qemu_malloc(s->cluster_size + (s->cluster_size / 1000) + 128); |
1742 | 585f8587 | bellard | |
1743 | 585f8587 | bellard | /* best compression, small window, no zlib header */
|
1744 | 585f8587 | bellard | memset(&strm, 0, sizeof(strm)); |
1745 | 585f8587 | bellard | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, |
1746 | 5fafdf24 | ths | Z_DEFLATED, -12,
|
1747 | 585f8587 | bellard | 9, Z_DEFAULT_STRATEGY);
|
1748 | 585f8587 | bellard | if (ret != 0) { |
1749 | 585f8587 | bellard | qemu_free(out_buf); |
1750 | 585f8587 | bellard | return -1; |
1751 | 585f8587 | bellard | } |
1752 | 585f8587 | bellard | |
1753 | 585f8587 | bellard | strm.avail_in = s->cluster_size; |
1754 | 585f8587 | bellard | strm.next_in = (uint8_t *)buf; |
1755 | 585f8587 | bellard | strm.avail_out = s->cluster_size; |
1756 | 585f8587 | bellard | strm.next_out = out_buf; |
1757 | 585f8587 | bellard | |
1758 | 585f8587 | bellard | ret = deflate(&strm, Z_FINISH); |
1759 | 585f8587 | bellard | if (ret != Z_STREAM_END && ret != Z_OK) {
|
1760 | 585f8587 | bellard | qemu_free(out_buf); |
1761 | 585f8587 | bellard | deflateEnd(&strm); |
1762 | 585f8587 | bellard | return -1; |
1763 | 585f8587 | bellard | } |
1764 | 585f8587 | bellard | out_len = strm.next_out - out_buf; |
1765 | 585f8587 | bellard | |
1766 | 585f8587 | bellard | deflateEnd(&strm); |
1767 | 585f8587 | bellard | |
1768 | 585f8587 | bellard | if (ret != Z_STREAM_END || out_len >= s->cluster_size) {
|
1769 | 585f8587 | bellard | /* could not compress: write normal cluster */
|
1770 | 585f8587 | bellard | qcow_write(bs, sector_num, buf, s->cluster_sectors); |
1771 | 585f8587 | bellard | } else {
|
1772 | 52d893ec | aliguori | cluster_offset = alloc_compressed_cluster_offset(bs, sector_num << 9,
|
1773 | 52d893ec | aliguori | out_len); |
1774 | 52d893ec | aliguori | if (!cluster_offset)
|
1775 | 52d893ec | aliguori | return -1; |
1776 | 585f8587 | bellard | cluster_offset &= s->cluster_offset_mask; |
1777 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, cluster_offset, out_buf, out_len) != out_len) {
|
1778 | 585f8587 | bellard | qemu_free(out_buf); |
1779 | 585f8587 | bellard | return -1; |
1780 | 585f8587 | bellard | } |
1781 | 585f8587 | bellard | } |
1782 | 3b46e624 | ths | |
1783 | 585f8587 | bellard | qemu_free(out_buf); |
1784 | 585f8587 | bellard | return 0; |
1785 | 585f8587 | bellard | } |
1786 | 585f8587 | bellard | |
1787 | 585f8587 | bellard | static void qcow_flush(BlockDriverState *bs) |
1788 | 585f8587 | bellard | { |
1789 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1790 | 585f8587 | bellard | bdrv_flush(s->hd); |
1791 | 585f8587 | bellard | } |
1792 | 585f8587 | bellard | |
1793 | 585f8587 | bellard | static int qcow_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
1794 | 585f8587 | bellard | { |
1795 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1796 | 585f8587 | bellard | bdi->cluster_size = s->cluster_size; |
1797 | 5fafdf24 | ths | bdi->vm_state_offset = (int64_t)s->l1_vm_state_index << |
1798 | 585f8587 | bellard | (s->cluster_bits + s->l2_bits); |
1799 | 585f8587 | bellard | return 0; |
1800 | 585f8587 | bellard | } |
1801 | 585f8587 | bellard | |
1802 | 585f8587 | bellard | /*********************************************************/
|
1803 | 585f8587 | bellard | /* snapshot support */
|
1804 | 585f8587 | bellard | |
1805 | 585f8587 | bellard | /* update the refcounts of snapshots and the copied flag */
|
1806 | 5fafdf24 | ths | static int update_snapshot_refcount(BlockDriverState *bs, |
1807 | 585f8587 | bellard | int64_t l1_table_offset, |
1808 | 585f8587 | bellard | int l1_size,
|
1809 | 585f8587 | bellard | int addend)
|
1810 | 585f8587 | bellard | { |
1811 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1812 | 585f8587 | bellard | uint64_t *l1_table, *l2_table, l2_offset, offset, l1_size2, l1_allocated; |
1813 | 585f8587 | bellard | int64_t old_offset, old_l2_offset; |
1814 | 585f8587 | bellard | int l2_size, i, j, l1_modified, l2_modified, nb_csectors, refcount;
|
1815 | 3b46e624 | ths | |
1816 | 585f8587 | bellard | l2_cache_reset(bs); |
1817 | 585f8587 | bellard | |
1818 | 585f8587 | bellard | l2_table = NULL;
|
1819 | 585f8587 | bellard | l1_table = NULL;
|
1820 | 585f8587 | bellard | l1_size2 = l1_size * sizeof(uint64_t);
|
1821 | 585f8587 | bellard | l1_allocated = 0;
|
1822 | 585f8587 | bellard | if (l1_table_offset != s->l1_table_offset) {
|
1823 | 585f8587 | bellard | l1_table = qemu_malloc(l1_size2); |
1824 | 585f8587 | bellard | l1_allocated = 1;
|
1825 | 5fafdf24 | ths | if (bdrv_pread(s->hd, l1_table_offset,
|
1826 | 585f8587 | bellard | l1_table, l1_size2) != l1_size2) |
1827 | 585f8587 | bellard | goto fail;
|
1828 | 585f8587 | bellard | for(i = 0;i < l1_size; i++) |
1829 | 585f8587 | bellard | be64_to_cpus(&l1_table[i]); |
1830 | 585f8587 | bellard | } else {
|
1831 | 585f8587 | bellard | assert(l1_size == s->l1_size); |
1832 | 585f8587 | bellard | l1_table = s->l1_table; |
1833 | 585f8587 | bellard | l1_allocated = 0;
|
1834 | 585f8587 | bellard | } |
1835 | 3b46e624 | ths | |
1836 | 585f8587 | bellard | l2_size = s->l2_size * sizeof(uint64_t);
|
1837 | 585f8587 | bellard | l2_table = qemu_malloc(l2_size); |
1838 | 585f8587 | bellard | l1_modified = 0;
|
1839 | 585f8587 | bellard | for(i = 0; i < l1_size; i++) { |
1840 | 585f8587 | bellard | l2_offset = l1_table[i]; |
1841 | 585f8587 | bellard | if (l2_offset) {
|
1842 | 585f8587 | bellard | old_l2_offset = l2_offset; |
1843 | 585f8587 | bellard | l2_offset &= ~QCOW_OFLAG_COPIED; |
1844 | 585f8587 | bellard | l2_modified = 0;
|
1845 | 585f8587 | bellard | if (bdrv_pread(s->hd, l2_offset, l2_table, l2_size) != l2_size)
|
1846 | 585f8587 | bellard | goto fail;
|
1847 | 585f8587 | bellard | for(j = 0; j < s->l2_size; j++) { |
1848 | 585f8587 | bellard | offset = be64_to_cpu(l2_table[j]); |
1849 | 585f8587 | bellard | if (offset != 0) { |
1850 | 585f8587 | bellard | old_offset = offset; |
1851 | 585f8587 | bellard | offset &= ~QCOW_OFLAG_COPIED; |
1852 | 585f8587 | bellard | if (offset & QCOW_OFLAG_COMPRESSED) {
|
1853 | 5fafdf24 | ths | nb_csectors = ((offset >> s->csize_shift) & |
1854 | 585f8587 | bellard | s->csize_mask) + 1;
|
1855 | 585f8587 | bellard | if (addend != 0) |
1856 | 585f8587 | bellard | update_refcount(bs, (offset & s->cluster_offset_mask) & ~511,
|
1857 | 585f8587 | bellard | nb_csectors * 512, addend);
|
1858 | 585f8587 | bellard | /* compressed clusters are never modified */
|
1859 | 5fafdf24 | ths | refcount = 2;
|
1860 | 585f8587 | bellard | } else {
|
1861 | 585f8587 | bellard | if (addend != 0) { |
1862 | 585f8587 | bellard | refcount = update_cluster_refcount(bs, offset >> s->cluster_bits, addend); |
1863 | 585f8587 | bellard | } else {
|
1864 | 585f8587 | bellard | refcount = get_refcount(bs, offset >> s->cluster_bits); |
1865 | 585f8587 | bellard | } |
1866 | 585f8587 | bellard | } |
1867 | 585f8587 | bellard | |
1868 | 585f8587 | bellard | if (refcount == 1) { |
1869 | 585f8587 | bellard | offset |= QCOW_OFLAG_COPIED; |
1870 | 585f8587 | bellard | } |
1871 | 585f8587 | bellard | if (offset != old_offset) {
|
1872 | 585f8587 | bellard | l2_table[j] = cpu_to_be64(offset); |
1873 | 585f8587 | bellard | l2_modified = 1;
|
1874 | 585f8587 | bellard | } |
1875 | 585f8587 | bellard | } |
1876 | 585f8587 | bellard | } |
1877 | 585f8587 | bellard | if (l2_modified) {
|
1878 | 5fafdf24 | ths | if (bdrv_pwrite(s->hd,
|
1879 | 585f8587 | bellard | l2_offset, l2_table, l2_size) != l2_size) |
1880 | 585f8587 | bellard | goto fail;
|
1881 | 585f8587 | bellard | } |
1882 | 585f8587 | bellard | |
1883 | 585f8587 | bellard | if (addend != 0) { |
1884 | 585f8587 | bellard | refcount = update_cluster_refcount(bs, l2_offset >> s->cluster_bits, addend); |
1885 | 585f8587 | bellard | } else {
|
1886 | 585f8587 | bellard | refcount = get_refcount(bs, l2_offset >> s->cluster_bits); |
1887 | 585f8587 | bellard | } |
1888 | 585f8587 | bellard | if (refcount == 1) { |
1889 | 585f8587 | bellard | l2_offset |= QCOW_OFLAG_COPIED; |
1890 | 585f8587 | bellard | } |
1891 | 585f8587 | bellard | if (l2_offset != old_l2_offset) {
|
1892 | 585f8587 | bellard | l1_table[i] = l2_offset; |
1893 | 585f8587 | bellard | l1_modified = 1;
|
1894 | 585f8587 | bellard | } |
1895 | 585f8587 | bellard | } |
1896 | 585f8587 | bellard | } |
1897 | 585f8587 | bellard | if (l1_modified) {
|
1898 | 585f8587 | bellard | for(i = 0; i < l1_size; i++) |
1899 | 585f8587 | bellard | cpu_to_be64s(&l1_table[i]); |
1900 | 5fafdf24 | ths | if (bdrv_pwrite(s->hd, l1_table_offset, l1_table,
|
1901 | 585f8587 | bellard | l1_size2) != l1_size2) |
1902 | 585f8587 | bellard | goto fail;
|
1903 | 585f8587 | bellard | for(i = 0; i < l1_size; i++) |
1904 | 585f8587 | bellard | be64_to_cpus(&l1_table[i]); |
1905 | 585f8587 | bellard | } |
1906 | 585f8587 | bellard | if (l1_allocated)
|
1907 | 585f8587 | bellard | qemu_free(l1_table); |
1908 | 585f8587 | bellard | qemu_free(l2_table); |
1909 | 585f8587 | bellard | return 0; |
1910 | 585f8587 | bellard | fail:
|
1911 | 585f8587 | bellard | if (l1_allocated)
|
1912 | 585f8587 | bellard | qemu_free(l1_table); |
1913 | 585f8587 | bellard | qemu_free(l2_table); |
1914 | 585f8587 | bellard | return -EIO;
|
1915 | 585f8587 | bellard | } |
1916 | 585f8587 | bellard | |
1917 | 585f8587 | bellard | static void qcow_free_snapshots(BlockDriverState *bs) |
1918 | 585f8587 | bellard | { |
1919 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1920 | 585f8587 | bellard | int i;
|
1921 | 585f8587 | bellard | |
1922 | 585f8587 | bellard | for(i = 0; i < s->nb_snapshots; i++) { |
1923 | 585f8587 | bellard | qemu_free(s->snapshots[i].name); |
1924 | 585f8587 | bellard | qemu_free(s->snapshots[i].id_str); |
1925 | 585f8587 | bellard | } |
1926 | 585f8587 | bellard | qemu_free(s->snapshots); |
1927 | 585f8587 | bellard | s->snapshots = NULL;
|
1928 | 585f8587 | bellard | s->nb_snapshots = 0;
|
1929 | 585f8587 | bellard | } |
1930 | 585f8587 | bellard | |
1931 | 585f8587 | bellard | static int qcow_read_snapshots(BlockDriverState *bs) |
1932 | 585f8587 | bellard | { |
1933 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1934 | 585f8587 | bellard | QCowSnapshotHeader h; |
1935 | 585f8587 | bellard | QCowSnapshot *sn; |
1936 | 585f8587 | bellard | int i, id_str_size, name_size;
|
1937 | 585f8587 | bellard | int64_t offset; |
1938 | 585f8587 | bellard | uint32_t extra_data_size; |
1939 | 585f8587 | bellard | |
1940 | 63c75dcd | malc | if (!s->nb_snapshots) {
|
1941 | 63c75dcd | malc | s->snapshots = NULL;
|
1942 | 63c75dcd | malc | s->snapshots_size = 0;
|
1943 | 63c75dcd | malc | return 0; |
1944 | 63c75dcd | malc | } |
1945 | 63c75dcd | malc | |
1946 | 585f8587 | bellard | offset = s->snapshots_offset; |
1947 | 585f8587 | bellard | s->snapshots = qemu_mallocz(s->nb_snapshots * sizeof(QCowSnapshot));
|
1948 | 585f8587 | bellard | for(i = 0; i < s->nb_snapshots; i++) { |
1949 | 585f8587 | bellard | offset = align_offset(offset, 8);
|
1950 | 585f8587 | bellard | if (bdrv_pread(s->hd, offset, &h, sizeof(h)) != sizeof(h)) |
1951 | 585f8587 | bellard | goto fail;
|
1952 | 585f8587 | bellard | offset += sizeof(h);
|
1953 | 585f8587 | bellard | sn = s->snapshots + i; |
1954 | 585f8587 | bellard | sn->l1_table_offset = be64_to_cpu(h.l1_table_offset); |
1955 | 585f8587 | bellard | sn->l1_size = be32_to_cpu(h.l1_size); |
1956 | 585f8587 | bellard | sn->vm_state_size = be32_to_cpu(h.vm_state_size); |
1957 | 585f8587 | bellard | sn->date_sec = be32_to_cpu(h.date_sec); |
1958 | 585f8587 | bellard | sn->date_nsec = be32_to_cpu(h.date_nsec); |
1959 | 585f8587 | bellard | sn->vm_clock_nsec = be64_to_cpu(h.vm_clock_nsec); |
1960 | 585f8587 | bellard | extra_data_size = be32_to_cpu(h.extra_data_size); |
1961 | 585f8587 | bellard | |
1962 | 585f8587 | bellard | id_str_size = be16_to_cpu(h.id_str_size); |
1963 | 585f8587 | bellard | name_size = be16_to_cpu(h.name_size); |
1964 | 585f8587 | bellard | |
1965 | 585f8587 | bellard | offset += extra_data_size; |
1966 | 585f8587 | bellard | |
1967 | 585f8587 | bellard | sn->id_str = qemu_malloc(id_str_size + 1);
|
1968 | 585f8587 | bellard | if (bdrv_pread(s->hd, offset, sn->id_str, id_str_size) != id_str_size)
|
1969 | 585f8587 | bellard | goto fail;
|
1970 | 585f8587 | bellard | offset += id_str_size; |
1971 | 585f8587 | bellard | sn->id_str[id_str_size] = '\0';
|
1972 | 585f8587 | bellard | |
1973 | 585f8587 | bellard | sn->name = qemu_malloc(name_size + 1);
|
1974 | 585f8587 | bellard | if (bdrv_pread(s->hd, offset, sn->name, name_size) != name_size)
|
1975 | 585f8587 | bellard | goto fail;
|
1976 | 585f8587 | bellard | offset += name_size; |
1977 | 585f8587 | bellard | sn->name[name_size] = '\0';
|
1978 | 585f8587 | bellard | } |
1979 | 585f8587 | bellard | s->snapshots_size = offset - s->snapshots_offset; |
1980 | 585f8587 | bellard | return 0; |
1981 | 585f8587 | bellard | fail:
|
1982 | 585f8587 | bellard | qcow_free_snapshots(bs); |
1983 | 585f8587 | bellard | return -1; |
1984 | 585f8587 | bellard | } |
1985 | 585f8587 | bellard | |
1986 | 585f8587 | bellard | /* add at the end of the file a new list of snapshots */
|
1987 | 585f8587 | bellard | static int qcow_write_snapshots(BlockDriverState *bs) |
1988 | 585f8587 | bellard | { |
1989 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
1990 | 585f8587 | bellard | QCowSnapshot *sn; |
1991 | 585f8587 | bellard | QCowSnapshotHeader h; |
1992 | 585f8587 | bellard | int i, name_size, id_str_size, snapshots_size;
|
1993 | 585f8587 | bellard | uint64_t data64; |
1994 | 585f8587 | bellard | uint32_t data32; |
1995 | 585f8587 | bellard | int64_t offset, snapshots_offset; |
1996 | 585f8587 | bellard | |
1997 | 585f8587 | bellard | /* compute the size of the snapshots */
|
1998 | 585f8587 | bellard | offset = 0;
|
1999 | 585f8587 | bellard | for(i = 0; i < s->nb_snapshots; i++) { |
2000 | 585f8587 | bellard | sn = s->snapshots + i; |
2001 | 585f8587 | bellard | offset = align_offset(offset, 8);
|
2002 | 585f8587 | bellard | offset += sizeof(h);
|
2003 | 585f8587 | bellard | offset += strlen(sn->id_str); |
2004 | 585f8587 | bellard | offset += strlen(sn->name); |
2005 | 585f8587 | bellard | } |
2006 | 585f8587 | bellard | snapshots_size = offset; |
2007 | 585f8587 | bellard | |
2008 | 585f8587 | bellard | snapshots_offset = alloc_clusters(bs, snapshots_size); |
2009 | 585f8587 | bellard | offset = snapshots_offset; |
2010 | 3b46e624 | ths | |
2011 | 585f8587 | bellard | for(i = 0; i < s->nb_snapshots; i++) { |
2012 | 585f8587 | bellard | sn = s->snapshots + i; |
2013 | 585f8587 | bellard | memset(&h, 0, sizeof(h)); |
2014 | 585f8587 | bellard | h.l1_table_offset = cpu_to_be64(sn->l1_table_offset); |
2015 | 585f8587 | bellard | h.l1_size = cpu_to_be32(sn->l1_size); |
2016 | 585f8587 | bellard | h.vm_state_size = cpu_to_be32(sn->vm_state_size); |
2017 | 585f8587 | bellard | h.date_sec = cpu_to_be32(sn->date_sec); |
2018 | 585f8587 | bellard | h.date_nsec = cpu_to_be32(sn->date_nsec); |
2019 | 585f8587 | bellard | h.vm_clock_nsec = cpu_to_be64(sn->vm_clock_nsec); |
2020 | 3b46e624 | ths | |
2021 | 585f8587 | bellard | id_str_size = strlen(sn->id_str); |
2022 | 585f8587 | bellard | name_size = strlen(sn->name); |
2023 | 585f8587 | bellard | h.id_str_size = cpu_to_be16(id_str_size); |
2024 | 585f8587 | bellard | h.name_size = cpu_to_be16(name_size); |
2025 | 585f8587 | bellard | offset = align_offset(offset, 8);
|
2026 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, offset, &h, sizeof(h)) != sizeof(h)) |
2027 | 585f8587 | bellard | goto fail;
|
2028 | 585f8587 | bellard | offset += sizeof(h);
|
2029 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, offset, sn->id_str, id_str_size) != id_str_size)
|
2030 | 585f8587 | bellard | goto fail;
|
2031 | 585f8587 | bellard | offset += id_str_size; |
2032 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, offset, sn->name, name_size) != name_size)
|
2033 | 585f8587 | bellard | goto fail;
|
2034 | 585f8587 | bellard | offset += name_size; |
2035 | 585f8587 | bellard | } |
2036 | 585f8587 | bellard | |
2037 | 585f8587 | bellard | /* update the various header fields */
|
2038 | 585f8587 | bellard | data64 = cpu_to_be64(snapshots_offset); |
2039 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, offsetof(QCowHeader, snapshots_offset),
|
2040 | 585f8587 | bellard | &data64, sizeof(data64)) != sizeof(data64)) |
2041 | 585f8587 | bellard | goto fail;
|
2042 | 585f8587 | bellard | data32 = cpu_to_be32(s->nb_snapshots); |
2043 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, offsetof(QCowHeader, nb_snapshots),
|
2044 | 585f8587 | bellard | &data32, sizeof(data32)) != sizeof(data32)) |
2045 | 585f8587 | bellard | goto fail;
|
2046 | 585f8587 | bellard | |
2047 | 585f8587 | bellard | /* free the old snapshot table */
|
2048 | 585f8587 | bellard | free_clusters(bs, s->snapshots_offset, s->snapshots_size); |
2049 | 585f8587 | bellard | s->snapshots_offset = snapshots_offset; |
2050 | 585f8587 | bellard | s->snapshots_size = snapshots_size; |
2051 | 585f8587 | bellard | return 0; |
2052 | 585f8587 | bellard | fail:
|
2053 | 585f8587 | bellard | return -1; |
2054 | 585f8587 | bellard | } |
2055 | 585f8587 | bellard | |
2056 | 585f8587 | bellard | static void find_new_snapshot_id(BlockDriverState *bs, |
2057 | 585f8587 | bellard | char *id_str, int id_str_size) |
2058 | 585f8587 | bellard | { |
2059 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2060 | 585f8587 | bellard | QCowSnapshot *sn; |
2061 | 585f8587 | bellard | int i, id, id_max = 0; |
2062 | 585f8587 | bellard | |
2063 | 585f8587 | bellard | for(i = 0; i < s->nb_snapshots; i++) { |
2064 | 585f8587 | bellard | sn = s->snapshots + i; |
2065 | 585f8587 | bellard | id = strtoul(sn->id_str, NULL, 10); |
2066 | 585f8587 | bellard | if (id > id_max)
|
2067 | 585f8587 | bellard | id_max = id; |
2068 | 585f8587 | bellard | } |
2069 | 585f8587 | bellard | snprintf(id_str, id_str_size, "%d", id_max + 1); |
2070 | 585f8587 | bellard | } |
2071 | 585f8587 | bellard | |
2072 | 585f8587 | bellard | static int find_snapshot_by_id(BlockDriverState *bs, const char *id_str) |
2073 | 585f8587 | bellard | { |
2074 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2075 | 585f8587 | bellard | int i;
|
2076 | 585f8587 | bellard | |
2077 | 585f8587 | bellard | for(i = 0; i < s->nb_snapshots; i++) { |
2078 | 585f8587 | bellard | if (!strcmp(s->snapshots[i].id_str, id_str))
|
2079 | 585f8587 | bellard | return i;
|
2080 | 585f8587 | bellard | } |
2081 | 585f8587 | bellard | return -1; |
2082 | 585f8587 | bellard | } |
2083 | 585f8587 | bellard | |
2084 | 585f8587 | bellard | static int find_snapshot_by_id_or_name(BlockDriverState *bs, const char *name) |
2085 | 585f8587 | bellard | { |
2086 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2087 | 585f8587 | bellard | int i, ret;
|
2088 | 3b46e624 | ths | |
2089 | 585f8587 | bellard | ret = find_snapshot_by_id(bs, name); |
2090 | 585f8587 | bellard | if (ret >= 0) |
2091 | 585f8587 | bellard | return ret;
|
2092 | 585f8587 | bellard | for(i = 0; i < s->nb_snapshots; i++) { |
2093 | 585f8587 | bellard | if (!strcmp(s->snapshots[i].name, name))
|
2094 | 585f8587 | bellard | return i;
|
2095 | 585f8587 | bellard | } |
2096 | 585f8587 | bellard | return -1; |
2097 | 585f8587 | bellard | } |
2098 | 585f8587 | bellard | |
2099 | 585f8587 | bellard | /* if no id is provided, a new one is constructed */
|
2100 | 5fafdf24 | ths | static int qcow_snapshot_create(BlockDriverState *bs, |
2101 | 585f8587 | bellard | QEMUSnapshotInfo *sn_info) |
2102 | 585f8587 | bellard | { |
2103 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2104 | 585f8587 | bellard | QCowSnapshot *snapshots1, sn1, *sn = &sn1; |
2105 | 585f8587 | bellard | int i, ret;
|
2106 | 585f8587 | bellard | uint64_t *l1_table = NULL;
|
2107 | 3b46e624 | ths | |
2108 | 585f8587 | bellard | memset(sn, 0, sizeof(*sn)); |
2109 | 585f8587 | bellard | |
2110 | 585f8587 | bellard | if (sn_info->id_str[0] == '\0') { |
2111 | 585f8587 | bellard | /* compute a new id */
|
2112 | 585f8587 | bellard | find_new_snapshot_id(bs, sn_info->id_str, sizeof(sn_info->id_str));
|
2113 | 585f8587 | bellard | } |
2114 | 585f8587 | bellard | |
2115 | 585f8587 | bellard | /* check that the ID is unique */
|
2116 | 585f8587 | bellard | if (find_snapshot_by_id(bs, sn_info->id_str) >= 0) |
2117 | 585f8587 | bellard | return -ENOENT;
|
2118 | 585f8587 | bellard | |
2119 | 585f8587 | bellard | sn->id_str = qemu_strdup(sn_info->id_str); |
2120 | 585f8587 | bellard | if (!sn->id_str)
|
2121 | 585f8587 | bellard | goto fail;
|
2122 | 585f8587 | bellard | sn->name = qemu_strdup(sn_info->name); |
2123 | 585f8587 | bellard | if (!sn->name)
|
2124 | 585f8587 | bellard | goto fail;
|
2125 | 585f8587 | bellard | sn->vm_state_size = sn_info->vm_state_size; |
2126 | 585f8587 | bellard | sn->date_sec = sn_info->date_sec; |
2127 | 585f8587 | bellard | sn->date_nsec = sn_info->date_nsec; |
2128 | 585f8587 | bellard | sn->vm_clock_nsec = sn_info->vm_clock_nsec; |
2129 | 585f8587 | bellard | |
2130 | 585f8587 | bellard | ret = update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, 1);
|
2131 | 585f8587 | bellard | if (ret < 0) |
2132 | 585f8587 | bellard | goto fail;
|
2133 | 585f8587 | bellard | |
2134 | 585f8587 | bellard | /* create the L1 table of the snapshot */
|
2135 | 585f8587 | bellard | sn->l1_table_offset = alloc_clusters(bs, s->l1_size * sizeof(uint64_t));
|
2136 | 585f8587 | bellard | sn->l1_size = s->l1_size; |
2137 | 585f8587 | bellard | |
2138 | 585f8587 | bellard | l1_table = qemu_malloc(s->l1_size * sizeof(uint64_t));
|
2139 | 585f8587 | bellard | for(i = 0; i < s->l1_size; i++) { |
2140 | 585f8587 | bellard | l1_table[i] = cpu_to_be64(s->l1_table[i]); |
2141 | 585f8587 | bellard | } |
2142 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, sn->l1_table_offset,
|
2143 | 5fafdf24 | ths | l1_table, s->l1_size * sizeof(uint64_t)) !=
|
2144 | 585f8587 | bellard | (s->l1_size * sizeof(uint64_t)))
|
2145 | 585f8587 | bellard | goto fail;
|
2146 | 585f8587 | bellard | qemu_free(l1_table); |
2147 | 585f8587 | bellard | l1_table = NULL;
|
2148 | 585f8587 | bellard | |
2149 | 585f8587 | bellard | snapshots1 = qemu_malloc((s->nb_snapshots + 1) * sizeof(QCowSnapshot)); |
2150 | 63c75dcd | malc | if (s->snapshots) {
|
2151 | 63c75dcd | malc | memcpy(snapshots1, s->snapshots, s->nb_snapshots * sizeof(QCowSnapshot));
|
2152 | 63c75dcd | malc | qemu_free(s->snapshots); |
2153 | 63c75dcd | malc | } |
2154 | 585f8587 | bellard | s->snapshots = snapshots1; |
2155 | 585f8587 | bellard | s->snapshots[s->nb_snapshots++] = *sn; |
2156 | 585f8587 | bellard | |
2157 | 585f8587 | bellard | if (qcow_write_snapshots(bs) < 0) |
2158 | 585f8587 | bellard | goto fail;
|
2159 | 585f8587 | bellard | #ifdef DEBUG_ALLOC
|
2160 | 585f8587 | bellard | check_refcounts(bs); |
2161 | 585f8587 | bellard | #endif
|
2162 | 585f8587 | bellard | return 0; |
2163 | 585f8587 | bellard | fail:
|
2164 | 585f8587 | bellard | qemu_free(sn->name); |
2165 | 585f8587 | bellard | qemu_free(l1_table); |
2166 | 585f8587 | bellard | return -1; |
2167 | 585f8587 | bellard | } |
2168 | 585f8587 | bellard | |
2169 | 585f8587 | bellard | /* copy the snapshot 'snapshot_name' into the current disk image */
|
2170 | 5fafdf24 | ths | static int qcow_snapshot_goto(BlockDriverState *bs, |
2171 | 585f8587 | bellard | const char *snapshot_id) |
2172 | 585f8587 | bellard | { |
2173 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2174 | 585f8587 | bellard | QCowSnapshot *sn; |
2175 | 585f8587 | bellard | int i, snapshot_index, l1_size2;
|
2176 | 585f8587 | bellard | |
2177 | 585f8587 | bellard | snapshot_index = find_snapshot_by_id_or_name(bs, snapshot_id); |
2178 | 585f8587 | bellard | if (snapshot_index < 0) |
2179 | 585f8587 | bellard | return -ENOENT;
|
2180 | 585f8587 | bellard | sn = &s->snapshots[snapshot_index]; |
2181 | 585f8587 | bellard | |
2182 | 585f8587 | bellard | if (update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, -1) < 0) |
2183 | 585f8587 | bellard | goto fail;
|
2184 | 585f8587 | bellard | |
2185 | 585f8587 | bellard | if (grow_l1_table(bs, sn->l1_size) < 0) |
2186 | 585f8587 | bellard | goto fail;
|
2187 | 585f8587 | bellard | |
2188 | 585f8587 | bellard | s->l1_size = sn->l1_size; |
2189 | 585f8587 | bellard | l1_size2 = s->l1_size * sizeof(uint64_t);
|
2190 | 585f8587 | bellard | /* copy the snapshot l1 table to the current l1 table */
|
2191 | 5fafdf24 | ths | if (bdrv_pread(s->hd, sn->l1_table_offset,
|
2192 | 585f8587 | bellard | s->l1_table, l1_size2) != l1_size2) |
2193 | 585f8587 | bellard | goto fail;
|
2194 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, s->l1_table_offset,
|
2195 | 585f8587 | bellard | s->l1_table, l1_size2) != l1_size2) |
2196 | 585f8587 | bellard | goto fail;
|
2197 | 585f8587 | bellard | for(i = 0;i < s->l1_size; i++) { |
2198 | 585f8587 | bellard | be64_to_cpus(&s->l1_table[i]); |
2199 | 585f8587 | bellard | } |
2200 | 585f8587 | bellard | |
2201 | 585f8587 | bellard | if (update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, 1) < 0) |
2202 | 585f8587 | bellard | goto fail;
|
2203 | 585f8587 | bellard | |
2204 | 585f8587 | bellard | #ifdef DEBUG_ALLOC
|
2205 | 585f8587 | bellard | check_refcounts(bs); |
2206 | 585f8587 | bellard | #endif
|
2207 | 585f8587 | bellard | return 0; |
2208 | 585f8587 | bellard | fail:
|
2209 | 585f8587 | bellard | return -EIO;
|
2210 | 585f8587 | bellard | } |
2211 | 585f8587 | bellard | |
2212 | 585f8587 | bellard | static int qcow_snapshot_delete(BlockDriverState *bs, const char *snapshot_id) |
2213 | 585f8587 | bellard | { |
2214 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2215 | 585f8587 | bellard | QCowSnapshot *sn; |
2216 | 585f8587 | bellard | int snapshot_index, ret;
|
2217 | 3b46e624 | ths | |
2218 | 585f8587 | bellard | snapshot_index = find_snapshot_by_id_or_name(bs, snapshot_id); |
2219 | 585f8587 | bellard | if (snapshot_index < 0) |
2220 | 585f8587 | bellard | return -ENOENT;
|
2221 | 585f8587 | bellard | sn = &s->snapshots[snapshot_index]; |
2222 | 585f8587 | bellard | |
2223 | 585f8587 | bellard | ret = update_snapshot_refcount(bs, sn->l1_table_offset, sn->l1_size, -1);
|
2224 | 585f8587 | bellard | if (ret < 0) |
2225 | 585f8587 | bellard | return ret;
|
2226 | 585f8587 | bellard | /* must update the copied flag on the current cluster offsets */
|
2227 | 585f8587 | bellard | ret = update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, 0);
|
2228 | 585f8587 | bellard | if (ret < 0) |
2229 | 585f8587 | bellard | return ret;
|
2230 | 585f8587 | bellard | free_clusters(bs, sn->l1_table_offset, sn->l1_size * sizeof(uint64_t));
|
2231 | 585f8587 | bellard | |
2232 | 585f8587 | bellard | qemu_free(sn->id_str); |
2233 | 585f8587 | bellard | qemu_free(sn->name); |
2234 | 585f8587 | bellard | memmove(sn, sn + 1, (s->nb_snapshots - snapshot_index - 1) * sizeof(*sn)); |
2235 | 585f8587 | bellard | s->nb_snapshots--; |
2236 | 585f8587 | bellard | ret = qcow_write_snapshots(bs); |
2237 | 585f8587 | bellard | if (ret < 0) { |
2238 | 585f8587 | bellard | /* XXX: restore snapshot if error ? */
|
2239 | 585f8587 | bellard | return ret;
|
2240 | 585f8587 | bellard | } |
2241 | 585f8587 | bellard | #ifdef DEBUG_ALLOC
|
2242 | 585f8587 | bellard | check_refcounts(bs); |
2243 | 585f8587 | bellard | #endif
|
2244 | 585f8587 | bellard | return 0; |
2245 | 585f8587 | bellard | } |
2246 | 585f8587 | bellard | |
2247 | 5fafdf24 | ths | static int qcow_snapshot_list(BlockDriverState *bs, |
2248 | 585f8587 | bellard | QEMUSnapshotInfo **psn_tab) |
2249 | 585f8587 | bellard | { |
2250 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2251 | 585f8587 | bellard | QEMUSnapshotInfo *sn_tab, *sn_info; |
2252 | 585f8587 | bellard | QCowSnapshot *sn; |
2253 | 585f8587 | bellard | int i;
|
2254 | 585f8587 | bellard | |
2255 | 585f8587 | bellard | sn_tab = qemu_mallocz(s->nb_snapshots * sizeof(QEMUSnapshotInfo));
|
2256 | 585f8587 | bellard | for(i = 0; i < s->nb_snapshots; i++) { |
2257 | 585f8587 | bellard | sn_info = sn_tab + i; |
2258 | 585f8587 | bellard | sn = s->snapshots + i; |
2259 | 585f8587 | bellard | pstrcpy(sn_info->id_str, sizeof(sn_info->id_str),
|
2260 | 585f8587 | bellard | sn->id_str); |
2261 | 585f8587 | bellard | pstrcpy(sn_info->name, sizeof(sn_info->name),
|
2262 | 585f8587 | bellard | sn->name); |
2263 | 585f8587 | bellard | sn_info->vm_state_size = sn->vm_state_size; |
2264 | 585f8587 | bellard | sn_info->date_sec = sn->date_sec; |
2265 | 585f8587 | bellard | sn_info->date_nsec = sn->date_nsec; |
2266 | 585f8587 | bellard | sn_info->vm_clock_nsec = sn->vm_clock_nsec; |
2267 | 585f8587 | bellard | } |
2268 | 585f8587 | bellard | *psn_tab = sn_tab; |
2269 | 585f8587 | bellard | return s->nb_snapshots;
|
2270 | 585f8587 | bellard | } |
2271 | 585f8587 | bellard | |
2272 | 585f8587 | bellard | /*********************************************************/
|
2273 | 585f8587 | bellard | /* refcount handling */
|
2274 | 585f8587 | bellard | |
2275 | 585f8587 | bellard | static int refcount_init(BlockDriverState *bs) |
2276 | 585f8587 | bellard | { |
2277 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2278 | 585f8587 | bellard | int ret, refcount_table_size2, i;
|
2279 | 3b46e624 | ths | |
2280 | 585f8587 | bellard | s->refcount_block_cache = qemu_malloc(s->cluster_size); |
2281 | 585f8587 | bellard | refcount_table_size2 = s->refcount_table_size * sizeof(uint64_t);
|
2282 | 585f8587 | bellard | s->refcount_table = qemu_malloc(refcount_table_size2); |
2283 | 585f8587 | bellard | if (s->refcount_table_size > 0) { |
2284 | 585f8587 | bellard | ret = bdrv_pread(s->hd, s->refcount_table_offset, |
2285 | 585f8587 | bellard | s->refcount_table, refcount_table_size2); |
2286 | 585f8587 | bellard | if (ret != refcount_table_size2)
|
2287 | 585f8587 | bellard | goto fail;
|
2288 | 585f8587 | bellard | for(i = 0; i < s->refcount_table_size; i++) |
2289 | 585f8587 | bellard | be64_to_cpus(&s->refcount_table[i]); |
2290 | 585f8587 | bellard | } |
2291 | 585f8587 | bellard | return 0; |
2292 | 585f8587 | bellard | fail:
|
2293 | 585f8587 | bellard | return -ENOMEM;
|
2294 | 585f8587 | bellard | } |
2295 | 585f8587 | bellard | |
2296 | 585f8587 | bellard | static void refcount_close(BlockDriverState *bs) |
2297 | 585f8587 | bellard | { |
2298 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2299 | 585f8587 | bellard | qemu_free(s->refcount_block_cache); |
2300 | 585f8587 | bellard | qemu_free(s->refcount_table); |
2301 | 585f8587 | bellard | } |
2302 | 585f8587 | bellard | |
2303 | 585f8587 | bellard | |
2304 | 5fafdf24 | ths | static int load_refcount_block(BlockDriverState *bs, |
2305 | 585f8587 | bellard | int64_t refcount_block_offset) |
2306 | 585f8587 | bellard | { |
2307 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2308 | 585f8587 | bellard | int ret;
|
2309 | 5fafdf24 | ths | ret = bdrv_pread(s->hd, refcount_block_offset, s->refcount_block_cache, |
2310 | 585f8587 | bellard | s->cluster_size); |
2311 | 585f8587 | bellard | if (ret != s->cluster_size)
|
2312 | 585f8587 | bellard | return -EIO;
|
2313 | 585f8587 | bellard | s->refcount_block_cache_offset = refcount_block_offset; |
2314 | 585f8587 | bellard | return 0; |
2315 | 585f8587 | bellard | } |
2316 | 585f8587 | bellard | |
2317 | 585f8587 | bellard | static int get_refcount(BlockDriverState *bs, int64_t cluster_index) |
2318 | 585f8587 | bellard | { |
2319 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2320 | 585f8587 | bellard | int refcount_table_index, block_index;
|
2321 | 585f8587 | bellard | int64_t refcount_block_offset; |
2322 | 585f8587 | bellard | |
2323 | 585f8587 | bellard | refcount_table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT); |
2324 | 585f8587 | bellard | if (refcount_table_index >= s->refcount_table_size)
|
2325 | 585f8587 | bellard | return 0; |
2326 | 585f8587 | bellard | refcount_block_offset = s->refcount_table[refcount_table_index]; |
2327 | 585f8587 | bellard | if (!refcount_block_offset)
|
2328 | 585f8587 | bellard | return 0; |
2329 | 585f8587 | bellard | if (refcount_block_offset != s->refcount_block_cache_offset) {
|
2330 | 585f8587 | bellard | /* better than nothing: return allocated if read error */
|
2331 | 585f8587 | bellard | if (load_refcount_block(bs, refcount_block_offset) < 0) |
2332 | 585f8587 | bellard | return 1; |
2333 | 585f8587 | bellard | } |
2334 | 5fafdf24 | ths | block_index = cluster_index & |
2335 | 585f8587 | bellard | ((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1); |
2336 | 585f8587 | bellard | return be16_to_cpu(s->refcount_block_cache[block_index]);
|
2337 | 585f8587 | bellard | } |
2338 | 585f8587 | bellard | |
2339 | 585f8587 | bellard | /* return < 0 if error */
|
2340 | 585f8587 | bellard | static int64_t alloc_clusters_noref(BlockDriverState *bs, int64_t size)
|
2341 | 585f8587 | bellard | { |
2342 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2343 | 585f8587 | bellard | int i, nb_clusters;
|
2344 | 585f8587 | bellard | |
2345 | 6db6c638 | aliguori | nb_clusters = size_to_clusters(s, size); |
2346 | 6db6c638 | aliguori | retry:
|
2347 | 6db6c638 | aliguori | for(i = 0; i < nb_clusters; i++) { |
2348 | 6db6c638 | aliguori | int64_t i = s->free_cluster_index++; |
2349 | 6db6c638 | aliguori | if (get_refcount(bs, i) != 0) |
2350 | 6db6c638 | aliguori | goto retry;
|
2351 | 6db6c638 | aliguori | } |
2352 | 585f8587 | bellard | #ifdef DEBUG_ALLOC2
|
2353 | 6db6c638 | aliguori | printf("alloc_clusters: size=%lld -> %lld\n",
|
2354 | 6db6c638 | aliguori | size, |
2355 | 6db6c638 | aliguori | (s->free_cluster_index - nb_clusters) << s->cluster_bits); |
2356 | 585f8587 | bellard | #endif
|
2357 | 6db6c638 | aliguori | return (s->free_cluster_index - nb_clusters) << s->cluster_bits;
|
2358 | 585f8587 | bellard | } |
2359 | 585f8587 | bellard | |
2360 | 585f8587 | bellard | static int64_t alloc_clusters(BlockDriverState *bs, int64_t size)
|
2361 | 585f8587 | bellard | { |
2362 | 585f8587 | bellard | int64_t offset; |
2363 | 585f8587 | bellard | |
2364 | 585f8587 | bellard | offset = alloc_clusters_noref(bs, size); |
2365 | 585f8587 | bellard | update_refcount(bs, offset, size, 1);
|
2366 | 585f8587 | bellard | return offset;
|
2367 | 585f8587 | bellard | } |
2368 | 585f8587 | bellard | |
2369 | 585f8587 | bellard | /* only used to allocate compressed sectors. We try to allocate
|
2370 | 585f8587 | bellard | contiguous sectors. size must be <= cluster_size */
|
2371 | 585f8587 | bellard | static int64_t alloc_bytes(BlockDriverState *bs, int size) |
2372 | 585f8587 | bellard | { |
2373 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2374 | 585f8587 | bellard | int64_t offset, cluster_offset; |
2375 | 585f8587 | bellard | int free_in_cluster;
|
2376 | 3b46e624 | ths | |
2377 | 585f8587 | bellard | assert(size > 0 && size <= s->cluster_size);
|
2378 | 585f8587 | bellard | if (s->free_byte_offset == 0) { |
2379 | 585f8587 | bellard | s->free_byte_offset = alloc_clusters(bs, s->cluster_size); |
2380 | 585f8587 | bellard | } |
2381 | 585f8587 | bellard | redo:
|
2382 | 5fafdf24 | ths | free_in_cluster = s->cluster_size - |
2383 | 585f8587 | bellard | (s->free_byte_offset & (s->cluster_size - 1));
|
2384 | 585f8587 | bellard | if (size <= free_in_cluster) {
|
2385 | 585f8587 | bellard | /* enough space in current cluster */
|
2386 | 585f8587 | bellard | offset = s->free_byte_offset; |
2387 | 585f8587 | bellard | s->free_byte_offset += size; |
2388 | 585f8587 | bellard | free_in_cluster -= size; |
2389 | 585f8587 | bellard | if (free_in_cluster == 0) |
2390 | 585f8587 | bellard | s->free_byte_offset = 0;
|
2391 | 585f8587 | bellard | if ((offset & (s->cluster_size - 1)) != 0) |
2392 | 585f8587 | bellard | update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
|
2393 | 585f8587 | bellard | } else {
|
2394 | 585f8587 | bellard | offset = alloc_clusters(bs, s->cluster_size); |
2395 | 585f8587 | bellard | cluster_offset = s->free_byte_offset & ~(s->cluster_size - 1);
|
2396 | 585f8587 | bellard | if ((cluster_offset + s->cluster_size) == offset) {
|
2397 | 585f8587 | bellard | /* we are lucky: contiguous data */
|
2398 | 585f8587 | bellard | offset = s->free_byte_offset; |
2399 | 585f8587 | bellard | update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
|
2400 | 585f8587 | bellard | s->free_byte_offset += size; |
2401 | 585f8587 | bellard | } else {
|
2402 | 585f8587 | bellard | s->free_byte_offset = offset; |
2403 | 585f8587 | bellard | goto redo;
|
2404 | 585f8587 | bellard | } |
2405 | 585f8587 | bellard | } |
2406 | 585f8587 | bellard | return offset;
|
2407 | 585f8587 | bellard | } |
2408 | 585f8587 | bellard | |
2409 | 5fafdf24 | ths | static void free_clusters(BlockDriverState *bs, |
2410 | 585f8587 | bellard | int64_t offset, int64_t size) |
2411 | 585f8587 | bellard | { |
2412 | 585f8587 | bellard | update_refcount(bs, offset, size, -1);
|
2413 | 585f8587 | bellard | } |
2414 | 585f8587 | bellard | |
2415 | 585f8587 | bellard | static int grow_refcount_table(BlockDriverState *bs, int min_size) |
2416 | 585f8587 | bellard | { |
2417 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2418 | 585f8587 | bellard | int new_table_size, new_table_size2, refcount_table_clusters, i, ret;
|
2419 | 585f8587 | bellard | uint64_t *new_table; |
2420 | 585f8587 | bellard | int64_t table_offset; |
2421 | 643e5399 | aliguori | uint8_t data[12];
|
2422 | 23be50f1 | ths | int old_table_size;
|
2423 | 23be50f1 | ths | int64_t old_table_offset; |
2424 | 585f8587 | bellard | |
2425 | 585f8587 | bellard | if (min_size <= s->refcount_table_size)
|
2426 | 585f8587 | bellard | return 0; |
2427 | 585f8587 | bellard | /* compute new table size */
|
2428 | 585f8587 | bellard | refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3);
|
2429 | 585f8587 | bellard | for(;;) {
|
2430 | 585f8587 | bellard | if (refcount_table_clusters == 0) { |
2431 | 585f8587 | bellard | refcount_table_clusters = 1;
|
2432 | 585f8587 | bellard | } else {
|
2433 | 585f8587 | bellard | refcount_table_clusters = (refcount_table_clusters * 3 + 1) / 2; |
2434 | 585f8587 | bellard | } |
2435 | 585f8587 | bellard | new_table_size = refcount_table_clusters << (s->cluster_bits - 3);
|
2436 | 585f8587 | bellard | if (min_size <= new_table_size)
|
2437 | 585f8587 | bellard | break;
|
2438 | 585f8587 | bellard | } |
2439 | 15e6690a | bellard | #ifdef DEBUG_ALLOC2
|
2440 | 15e6690a | bellard | printf("grow_refcount_table from %d to %d\n",
|
2441 | 15e6690a | bellard | s->refcount_table_size, |
2442 | 15e6690a | bellard | new_table_size); |
2443 | 15e6690a | bellard | #endif
|
2444 | 585f8587 | bellard | new_table_size2 = new_table_size * sizeof(uint64_t);
|
2445 | 585f8587 | bellard | new_table = qemu_mallocz(new_table_size2); |
2446 | 5fafdf24 | ths | memcpy(new_table, s->refcount_table, |
2447 | 585f8587 | bellard | s->refcount_table_size * sizeof(uint64_t));
|
2448 | 585f8587 | bellard | for(i = 0; i < s->refcount_table_size; i++) |
2449 | 585f8587 | bellard | cpu_to_be64s(&new_table[i]); |
2450 | 585f8587 | bellard | /* Note: we cannot update the refcount now to avoid recursion */
|
2451 | 585f8587 | bellard | table_offset = alloc_clusters_noref(bs, new_table_size2); |
2452 | 585f8587 | bellard | ret = bdrv_pwrite(s->hd, table_offset, new_table, new_table_size2); |
2453 | 5fafdf24 | ths | if (ret != new_table_size2)
|
2454 | 585f8587 | bellard | goto fail;
|
2455 | 585f8587 | bellard | for(i = 0; i < s->refcount_table_size; i++) |
2456 | 585f8587 | bellard | be64_to_cpus(&new_table[i]); |
2457 | 585f8587 | bellard | |
2458 | 643e5399 | aliguori | cpu_to_be64w((uint64_t*)data, table_offset); |
2459 | 643e5399 | aliguori | cpu_to_be32w((uint32_t*)(data + 8), refcount_table_clusters);
|
2460 | 585f8587 | bellard | if (bdrv_pwrite(s->hd, offsetof(QCowHeader, refcount_table_offset),
|
2461 | 643e5399 | aliguori | data, sizeof(data)) != sizeof(data)) |
2462 | 585f8587 | bellard | goto fail;
|
2463 | 585f8587 | bellard | qemu_free(s->refcount_table); |
2464 | 23be50f1 | ths | old_table_offset = s->refcount_table_offset; |
2465 | 23be50f1 | ths | old_table_size = s->refcount_table_size; |
2466 | 585f8587 | bellard | s->refcount_table = new_table; |
2467 | 585f8587 | bellard | s->refcount_table_size = new_table_size; |
2468 | a4080ece | ths | s->refcount_table_offset = table_offset; |
2469 | 585f8587 | bellard | |
2470 | 585f8587 | bellard | update_refcount(bs, table_offset, new_table_size2, 1);
|
2471 | 23be50f1 | ths | free_clusters(bs, old_table_offset, old_table_size * sizeof(uint64_t));
|
2472 | 585f8587 | bellard | return 0; |
2473 | 585f8587 | bellard | fail:
|
2474 | 585f8587 | bellard | free_clusters(bs, table_offset, new_table_size2); |
2475 | 585f8587 | bellard | qemu_free(new_table); |
2476 | 585f8587 | bellard | return -EIO;
|
2477 | 585f8587 | bellard | } |
2478 | 585f8587 | bellard | |
2479 | 585f8587 | bellard | /* addend must be 1 or -1 */
|
2480 | 585f8587 | bellard | /* XXX: cache several refcount block clusters ? */
|
2481 | 5fafdf24 | ths | static int update_cluster_refcount(BlockDriverState *bs, |
2482 | 585f8587 | bellard | int64_t cluster_index, |
2483 | 585f8587 | bellard | int addend)
|
2484 | 585f8587 | bellard | { |
2485 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2486 | 585f8587 | bellard | int64_t offset, refcount_block_offset; |
2487 | 585f8587 | bellard | int ret, refcount_table_index, block_index, refcount;
|
2488 | 585f8587 | bellard | uint64_t data64; |
2489 | 585f8587 | bellard | |
2490 | 585f8587 | bellard | refcount_table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT); |
2491 | 585f8587 | bellard | if (refcount_table_index >= s->refcount_table_size) {
|
2492 | 585f8587 | bellard | if (addend < 0) |
2493 | 585f8587 | bellard | return -EINVAL;
|
2494 | 585f8587 | bellard | ret = grow_refcount_table(bs, refcount_table_index + 1);
|
2495 | 585f8587 | bellard | if (ret < 0) |
2496 | 585f8587 | bellard | return ret;
|
2497 | 585f8587 | bellard | } |
2498 | 585f8587 | bellard | refcount_block_offset = s->refcount_table[refcount_table_index]; |
2499 | 585f8587 | bellard | if (!refcount_block_offset) {
|
2500 | 585f8587 | bellard | if (addend < 0) |
2501 | 585f8587 | bellard | return -EINVAL;
|
2502 | 585f8587 | bellard | /* create a new refcount block */
|
2503 | 585f8587 | bellard | /* Note: we cannot update the refcount now to avoid recursion */
|
2504 | 585f8587 | bellard | offset = alloc_clusters_noref(bs, s->cluster_size); |
2505 | 585f8587 | bellard | memset(s->refcount_block_cache, 0, s->cluster_size);
|
2506 | 585f8587 | bellard | ret = bdrv_pwrite(s->hd, offset, s->refcount_block_cache, s->cluster_size); |
2507 | 585f8587 | bellard | if (ret != s->cluster_size)
|
2508 | 585f8587 | bellard | return -EINVAL;
|
2509 | 585f8587 | bellard | s->refcount_table[refcount_table_index] = offset; |
2510 | 585f8587 | bellard | data64 = cpu_to_be64(offset); |
2511 | 5fafdf24 | ths | ret = bdrv_pwrite(s->hd, s->refcount_table_offset + |
2512 | 5fafdf24 | ths | refcount_table_index * sizeof(uint64_t),
|
2513 | 585f8587 | bellard | &data64, sizeof(data64));
|
2514 | 585f8587 | bellard | if (ret != sizeof(data64)) |
2515 | 585f8587 | bellard | return -EINVAL;
|
2516 | 585f8587 | bellard | |
2517 | 585f8587 | bellard | refcount_block_offset = offset; |
2518 | 585f8587 | bellard | s->refcount_block_cache_offset = offset; |
2519 | 585f8587 | bellard | update_refcount(bs, offset, s->cluster_size, 1);
|
2520 | 585f8587 | bellard | } else {
|
2521 | 585f8587 | bellard | if (refcount_block_offset != s->refcount_block_cache_offset) {
|
2522 | 585f8587 | bellard | if (load_refcount_block(bs, refcount_block_offset) < 0) |
2523 | 585f8587 | bellard | return -EIO;
|
2524 | 585f8587 | bellard | } |
2525 | 585f8587 | bellard | } |
2526 | 585f8587 | bellard | /* we can update the count and save it */
|
2527 | 5fafdf24 | ths | block_index = cluster_index & |
2528 | 585f8587 | bellard | ((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1); |
2529 | 585f8587 | bellard | refcount = be16_to_cpu(s->refcount_block_cache[block_index]); |
2530 | 585f8587 | bellard | refcount += addend; |
2531 | 585f8587 | bellard | if (refcount < 0 || refcount > 0xffff) |
2532 | 585f8587 | bellard | return -EINVAL;
|
2533 | 585f8587 | bellard | if (refcount == 0 && cluster_index < s->free_cluster_index) { |
2534 | 585f8587 | bellard | s->free_cluster_index = cluster_index; |
2535 | 585f8587 | bellard | } |
2536 | 585f8587 | bellard | s->refcount_block_cache[block_index] = cpu_to_be16(refcount); |
2537 | 5fafdf24 | ths | if (bdrv_pwrite(s->hd,
|
2538 | 5fafdf24 | ths | refcount_block_offset + (block_index << REFCOUNT_SHIFT), |
2539 | 585f8587 | bellard | &s->refcount_block_cache[block_index], 2) != 2) |
2540 | 585f8587 | bellard | return -EIO;
|
2541 | 585f8587 | bellard | return refcount;
|
2542 | 585f8587 | bellard | } |
2543 | 585f8587 | bellard | |
2544 | 5fafdf24 | ths | static void update_refcount(BlockDriverState *bs, |
2545 | 5fafdf24 | ths | int64_t offset, int64_t length, |
2546 | 585f8587 | bellard | int addend)
|
2547 | 585f8587 | bellard | { |
2548 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2549 | 585f8587 | bellard | int64_t start, last, cluster_offset; |
2550 | 585f8587 | bellard | |
2551 | 585f8587 | bellard | #ifdef DEBUG_ALLOC2
|
2552 | 5fafdf24 | ths | printf("update_refcount: offset=%lld size=%lld addend=%d\n",
|
2553 | 585f8587 | bellard | offset, length, addend); |
2554 | 585f8587 | bellard | #endif
|
2555 | 585f8587 | bellard | if (length <= 0) |
2556 | 585f8587 | bellard | return;
|
2557 | 585f8587 | bellard | start = offset & ~(s->cluster_size - 1);
|
2558 | 585f8587 | bellard | last = (offset + length - 1) & ~(s->cluster_size - 1); |
2559 | 5fafdf24 | ths | for(cluster_offset = start; cluster_offset <= last;
|
2560 | 585f8587 | bellard | cluster_offset += s->cluster_size) { |
2561 | 585f8587 | bellard | update_cluster_refcount(bs, cluster_offset >> s->cluster_bits, addend); |
2562 | 585f8587 | bellard | } |
2563 | 585f8587 | bellard | } |
2564 | 585f8587 | bellard | |
2565 | e97fc193 | aliguori | /*
|
2566 | e97fc193 | aliguori | * Increases the refcount for a range of clusters in a given refcount table.
|
2567 | e97fc193 | aliguori | * This is used to construct a temporary refcount table out of L1 and L2 tables
|
2568 | e97fc193 | aliguori | * which can be compared the the refcount table saved in the image.
|
2569 | e97fc193 | aliguori | *
|
2570 | e97fc193 | aliguori | * Returns the number of errors in the image that were found
|
2571 | e97fc193 | aliguori | */
|
2572 | e97fc193 | aliguori | static int inc_refcounts(BlockDriverState *bs, |
2573 | 5fafdf24 | ths | uint16_t *refcount_table, |
2574 | 585f8587 | bellard | int refcount_table_size,
|
2575 | 585f8587 | bellard | int64_t offset, int64_t size) |
2576 | 585f8587 | bellard | { |
2577 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2578 | 585f8587 | bellard | int64_t start, last, cluster_offset; |
2579 | 585f8587 | bellard | int k;
|
2580 | e97fc193 | aliguori | int errors = 0; |
2581 | 3b46e624 | ths | |
2582 | 585f8587 | bellard | if (size <= 0) |
2583 | e97fc193 | aliguori | return 0; |
2584 | 585f8587 | bellard | |
2585 | 585f8587 | bellard | start = offset & ~(s->cluster_size - 1);
|
2586 | 585f8587 | bellard | last = (offset + size - 1) & ~(s->cluster_size - 1); |
2587 | 5fafdf24 | ths | for(cluster_offset = start; cluster_offset <= last;
|
2588 | 585f8587 | bellard | cluster_offset += s->cluster_size) { |
2589 | 585f8587 | bellard | k = cluster_offset >> s->cluster_bits; |
2590 | 585f8587 | bellard | if (k < 0 || k >= refcount_table_size) { |
2591 | 8ddbc04f | aliguori | fprintf(stderr, "ERROR: invalid cluster offset=0x%" PRIx64 "\n", |
2592 | 8ddbc04f | aliguori | cluster_offset); |
2593 | e97fc193 | aliguori | errors++; |
2594 | 585f8587 | bellard | } else {
|
2595 | 585f8587 | bellard | if (++refcount_table[k] == 0) { |
2596 | 8ddbc04f | aliguori | fprintf(stderr, "ERROR: overflow cluster offset=0x%" PRIx64
|
2597 | 8ddbc04f | aliguori | "\n", cluster_offset);
|
2598 | e97fc193 | aliguori | errors++; |
2599 | 585f8587 | bellard | } |
2600 | 585f8587 | bellard | } |
2601 | 585f8587 | bellard | } |
2602 | e97fc193 | aliguori | |
2603 | e97fc193 | aliguori | return errors;
|
2604 | 585f8587 | bellard | } |
2605 | 585f8587 | bellard | |
2606 | ae2f14af | aliguori | /*
|
2607 | ae2f14af | aliguori | * Increases the refcount in the given refcount table for the all clusters
|
2608 | ae2f14af | aliguori | * referenced in the L2 table. While doing so, performs some checks on L2
|
2609 | ae2f14af | aliguori | * entries.
|
2610 | ae2f14af | aliguori | *
|
2611 | ae2f14af | aliguori | * Returns the number of errors found by the checks or -errno if an internal
|
2612 | ae2f14af | aliguori | * error occurred.
|
2613 | ae2f14af | aliguori | */
|
2614 | ae2f14af | aliguori | static int check_refcounts_l2(BlockDriverState *bs, |
2615 | ae2f14af | aliguori | uint16_t *refcount_table, int refcount_table_size, int64_t l2_offset,
|
2616 | ae2f14af | aliguori | int check_copied)
|
2617 | ae2f14af | aliguori | { |
2618 | ae2f14af | aliguori | BDRVQcowState *s = bs->opaque; |
2619 | ae2f14af | aliguori | uint64_t *l2_table, offset; |
2620 | ae2f14af | aliguori | int i, l2_size, nb_csectors, refcount;
|
2621 | ae2f14af | aliguori | int errors = 0; |
2622 | ae2f14af | aliguori | |
2623 | ae2f14af | aliguori | /* Read L2 table from disk */
|
2624 | ae2f14af | aliguori | l2_size = s->l2_size * sizeof(uint64_t);
|
2625 | ae2f14af | aliguori | l2_table = qemu_malloc(l2_size); |
2626 | ae2f14af | aliguori | |
2627 | ae2f14af | aliguori | if (bdrv_pread(s->hd, l2_offset, l2_table, l2_size) != l2_size)
|
2628 | ae2f14af | aliguori | goto fail;
|
2629 | ae2f14af | aliguori | |
2630 | ae2f14af | aliguori | /* Do the actual checks */
|
2631 | ae2f14af | aliguori | for(i = 0; i < s->l2_size; i++) { |
2632 | ae2f14af | aliguori | offset = be64_to_cpu(l2_table[i]); |
2633 | ae2f14af | aliguori | if (offset != 0) { |
2634 | ae2f14af | aliguori | if (offset & QCOW_OFLAG_COMPRESSED) {
|
2635 | ae2f14af | aliguori | /* Compressed clusters don't have QCOW_OFLAG_COPIED */
|
2636 | ae2f14af | aliguori | if (offset & QCOW_OFLAG_COPIED) {
|
2637 | ae2f14af | aliguori | fprintf(stderr, "ERROR: cluster %" PRId64 ": " |
2638 | ae2f14af | aliguori | "copied flag must never be set for compressed "
|
2639 | ae2f14af | aliguori | "clusters\n", offset >> s->cluster_bits);
|
2640 | ae2f14af | aliguori | offset &= ~QCOW_OFLAG_COPIED; |
2641 | ae2f14af | aliguori | errors++; |
2642 | ae2f14af | aliguori | } |
2643 | ae2f14af | aliguori | |
2644 | ae2f14af | aliguori | /* Mark cluster as used */
|
2645 | ae2f14af | aliguori | nb_csectors = ((offset >> s->csize_shift) & |
2646 | ae2f14af | aliguori | s->csize_mask) + 1;
|
2647 | ae2f14af | aliguori | offset &= s->cluster_offset_mask; |
2648 | ae2f14af | aliguori | errors += inc_refcounts(bs, refcount_table, |
2649 | ae2f14af | aliguori | refcount_table_size, |
2650 | ae2f14af | aliguori | offset & ~511, nb_csectors * 512); |
2651 | ae2f14af | aliguori | } else {
|
2652 | ae2f14af | aliguori | /* QCOW_OFLAG_COPIED must be set iff refcount == 1 */
|
2653 | ae2f14af | aliguori | if (check_copied) {
|
2654 | ae2f14af | aliguori | uint64_t entry = offset; |
2655 | ae2f14af | aliguori | offset &= ~QCOW_OFLAG_COPIED; |
2656 | ae2f14af | aliguori | refcount = get_refcount(bs, offset >> s->cluster_bits); |
2657 | ae2f14af | aliguori | if ((refcount == 1) != ((entry & QCOW_OFLAG_COPIED) != 0)) { |
2658 | ae2f14af | aliguori | fprintf(stderr, "ERROR OFLAG_COPIED: offset=%"
|
2659 | ae2f14af | aliguori | PRIx64 " refcount=%d\n", entry, refcount);
|
2660 | ae2f14af | aliguori | errors++; |
2661 | ae2f14af | aliguori | } |
2662 | ae2f14af | aliguori | } |
2663 | ae2f14af | aliguori | |
2664 | ae2f14af | aliguori | /* Mark cluster as used */
|
2665 | ae2f14af | aliguori | offset &= ~QCOW_OFLAG_COPIED; |
2666 | ae2f14af | aliguori | errors += inc_refcounts(bs, refcount_table, |
2667 | ae2f14af | aliguori | refcount_table_size, |
2668 | ae2f14af | aliguori | offset, s->cluster_size); |
2669 | 54c42368 | aliguori | |
2670 | 54c42368 | aliguori | /* Correct offsets are cluster aligned */
|
2671 | 54c42368 | aliguori | if (offset & (s->cluster_size - 1)) { |
2672 | 54c42368 | aliguori | fprintf(stderr, "ERROR offset=%" PRIx64 ": Cluster is not " |
2673 | 54c42368 | aliguori | "properly aligned; L2 entry corrupted.\n", offset);
|
2674 | 54c42368 | aliguori | errors++; |
2675 | 54c42368 | aliguori | } |
2676 | ae2f14af | aliguori | } |
2677 | ae2f14af | aliguori | } |
2678 | ae2f14af | aliguori | } |
2679 | ae2f14af | aliguori | |
2680 | ae2f14af | aliguori | qemu_free(l2_table); |
2681 | ae2f14af | aliguori | return errors;
|
2682 | ae2f14af | aliguori | |
2683 | ae2f14af | aliguori | fail:
|
2684 | ae2f14af | aliguori | fprintf(stderr, "ERROR: I/O error in check_refcounts_l1\n");
|
2685 | ae2f14af | aliguori | qemu_free(l2_table); |
2686 | ae2f14af | aliguori | return -EIO;
|
2687 | ae2f14af | aliguori | } |
2688 | ae2f14af | aliguori | |
2689 | ae2f14af | aliguori | /*
|
2690 | ae2f14af | aliguori | * Increases the refcount for the L1 table, its L2 tables and all referenced
|
2691 | ae2f14af | aliguori | * clusters in the given refcount table. While doing so, performs some checks
|
2692 | ae2f14af | aliguori | * on L1 and L2 entries.
|
2693 | ae2f14af | aliguori | *
|
2694 | ae2f14af | aliguori | * Returns the number of errors found by the checks or -errno if an internal
|
2695 | ae2f14af | aliguori | * error occurred.
|
2696 | ae2f14af | aliguori | */
|
2697 | 5fafdf24 | ths | static int check_refcounts_l1(BlockDriverState *bs, |
2698 | 5fafdf24 | ths | uint16_t *refcount_table, |
2699 | 585f8587 | bellard | int refcount_table_size,
|
2700 | 585f8587 | bellard | int64_t l1_table_offset, int l1_size,
|
2701 | 585f8587 | bellard | int check_copied)
|
2702 | 585f8587 | bellard | { |
2703 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2704 | ae2f14af | aliguori | uint64_t *l1_table, l2_offset, l1_size2; |
2705 | ae2f14af | aliguori | int i, refcount, ret;
|
2706 | e97fc193 | aliguori | int errors = 0; |
2707 | 585f8587 | bellard | |
2708 | 585f8587 | bellard | l1_size2 = l1_size * sizeof(uint64_t);
|
2709 | 585f8587 | bellard | |
2710 | ae2f14af | aliguori | /* Mark L1 table as used */
|
2711 | e97fc193 | aliguori | errors += inc_refcounts(bs, refcount_table, refcount_table_size, |
2712 | 585f8587 | bellard | l1_table_offset, l1_size2); |
2713 | 585f8587 | bellard | |
2714 | ae2f14af | aliguori | /* Read L1 table entries from disk */
|
2715 | 585f8587 | bellard | l1_table = qemu_malloc(l1_size2); |
2716 | 5fafdf24 | ths | if (bdrv_pread(s->hd, l1_table_offset,
|
2717 | 585f8587 | bellard | l1_table, l1_size2) != l1_size2) |
2718 | 585f8587 | bellard | goto fail;
|
2719 | 585f8587 | bellard | for(i = 0;i < l1_size; i++) |
2720 | 585f8587 | bellard | be64_to_cpus(&l1_table[i]); |
2721 | 3b46e624 | ths | |
2722 | ae2f14af | aliguori | /* Do the actual checks */
|
2723 | 585f8587 | bellard | for(i = 0; i < l1_size; i++) { |
2724 | 585f8587 | bellard | l2_offset = l1_table[i]; |
2725 | 585f8587 | bellard | if (l2_offset) {
|
2726 | ae2f14af | aliguori | /* QCOW_OFLAG_COPIED must be set iff refcount == 1 */
|
2727 | 585f8587 | bellard | if (check_copied) {
|
2728 | ae2f14af | aliguori | refcount = get_refcount(bs, (l2_offset & ~QCOW_OFLAG_COPIED) |
2729 | ae2f14af | aliguori | >> s->cluster_bits); |
2730 | 585f8587 | bellard | if ((refcount == 1) != ((l2_offset & QCOW_OFLAG_COPIED) != 0)) { |
2731 | 8ddbc04f | aliguori | fprintf(stderr, "ERROR OFLAG_COPIED: l2_offset=%" PRIx64
|
2732 | 8ddbc04f | aliguori | " refcount=%d\n", l2_offset, refcount);
|
2733 | e97fc193 | aliguori | errors++; |
2734 | 585f8587 | bellard | } |
2735 | 585f8587 | bellard | } |
2736 | ae2f14af | aliguori | |
2737 | ae2f14af | aliguori | /* Mark L2 table as used */
|
2738 | 585f8587 | bellard | l2_offset &= ~QCOW_OFLAG_COPIED; |
2739 | e97fc193 | aliguori | errors += inc_refcounts(bs, refcount_table, |
2740 | 585f8587 | bellard | refcount_table_size, |
2741 | 585f8587 | bellard | l2_offset, |
2742 | 585f8587 | bellard | s->cluster_size); |
2743 | ae2f14af | aliguori | |
2744 | 54c42368 | aliguori | /* L2 tables are cluster aligned */
|
2745 | 54c42368 | aliguori | if (l2_offset & (s->cluster_size - 1)) { |
2746 | 54c42368 | aliguori | fprintf(stderr, "ERROR l2_offset=%" PRIx64 ": Table is not " |
2747 | 54c42368 | aliguori | "cluster aligned; L1 entry corrupted\n", l2_offset);
|
2748 | 54c42368 | aliguori | errors++; |
2749 | 54c42368 | aliguori | } |
2750 | 54c42368 | aliguori | |
2751 | ae2f14af | aliguori | /* Process and check L2 entries */
|
2752 | ae2f14af | aliguori | ret = check_refcounts_l2(bs, refcount_table, refcount_table_size, |
2753 | ae2f14af | aliguori | l2_offset, check_copied); |
2754 | ae2f14af | aliguori | if (ret < 0) { |
2755 | ae2f14af | aliguori | goto fail;
|
2756 | ae2f14af | aliguori | } |
2757 | ae2f14af | aliguori | errors += ret; |
2758 | 585f8587 | bellard | } |
2759 | 585f8587 | bellard | } |
2760 | 585f8587 | bellard | qemu_free(l1_table); |
2761 | e97fc193 | aliguori | return errors;
|
2762 | ae2f14af | aliguori | |
2763 | ae2f14af | aliguori | fail:
|
2764 | 8ddbc04f | aliguori | fprintf(stderr, "ERROR: I/O error in check_refcounts_l1\n");
|
2765 | 585f8587 | bellard | qemu_free(l1_table); |
2766 | 585f8587 | bellard | return -EIO;
|
2767 | 585f8587 | bellard | } |
2768 | 585f8587 | bellard | |
2769 | e97fc193 | aliguori | /*
|
2770 | e97fc193 | aliguori | * Checks an image for refcount consistency.
|
2771 | e97fc193 | aliguori | *
|
2772 | e97fc193 | aliguori | * Returns 0 if no errors are found, the number of errors in case the image is
|
2773 | e97fc193 | aliguori | * detected as corrupted, and -errno when an internal error occured.
|
2774 | e97fc193 | aliguori | */
|
2775 | e97fc193 | aliguori | static int check_refcounts(BlockDriverState *bs) |
2776 | 585f8587 | bellard | { |
2777 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque; |
2778 | 585f8587 | bellard | int64_t size; |
2779 | 585f8587 | bellard | int nb_clusters, refcount1, refcount2, i;
|
2780 | 585f8587 | bellard | QCowSnapshot *sn; |
2781 | 585f8587 | bellard | uint16_t *refcount_table; |
2782 | e97fc193 | aliguori | int ret, errors = 0; |
2783 | 585f8587 | bellard | |
2784 | 585f8587 | bellard | size = bdrv_getlength(s->hd); |
2785 | 6db6c638 | aliguori | nb_clusters = size_to_clusters(s, size); |
2786 | 585f8587 | bellard | refcount_table = qemu_mallocz(nb_clusters * sizeof(uint16_t));
|
2787 | 15e6690a | bellard | |
2788 | 585f8587 | bellard | /* header */
|
2789 | e97fc193 | aliguori | errors += inc_refcounts(bs, refcount_table, nb_clusters, |
2790 | 585f8587 | bellard | 0, s->cluster_size);
|
2791 | 3b46e624 | ths | |
2792 | ae2f14af | aliguori | /* current L1 table */
|
2793 | e97fc193 | aliguori | ret = check_refcounts_l1(bs, refcount_table, nb_clusters, |
2794 | 585f8587 | bellard | s->l1_table_offset, s->l1_size, 1);
|
2795 | e97fc193 | aliguori | if (ret < 0) { |
2796 | e97fc193 | aliguori | return ret;
|
2797 | e97fc193 | aliguori | } |
2798 | e97fc193 | aliguori | errors += ret; |
2799 | 585f8587 | bellard | |
2800 | 585f8587 | bellard | /* snapshots */
|
2801 | 585f8587 | bellard | for(i = 0; i < s->nb_snapshots; i++) { |
2802 | 585f8587 | bellard | sn = s->snapshots + i; |
2803 | 585f8587 | bellard | check_refcounts_l1(bs, refcount_table, nb_clusters, |
2804 | 585f8587 | bellard | sn->l1_table_offset, sn->l1_size, 0);
|
2805 | 585f8587 | bellard | } |
2806 | e97fc193 | aliguori | errors += inc_refcounts(bs, refcount_table, nb_clusters, |
2807 | 585f8587 | bellard | s->snapshots_offset, s->snapshots_size); |
2808 | 585f8587 | bellard | |
2809 | 585f8587 | bellard | /* refcount data */
|
2810 | e97fc193 | aliguori | errors += inc_refcounts(bs, refcount_table, nb_clusters, |
2811 | 5fafdf24 | ths | s->refcount_table_offset, |
2812 | 585f8587 | bellard | s->refcount_table_size * sizeof(uint64_t));
|
2813 | 585f8587 | bellard | for(i = 0; i < s->refcount_table_size; i++) { |
2814 | 585f8587 | bellard | int64_t offset; |
2815 | 585f8587 | bellard | offset = s->refcount_table[i]; |
2816 | 585f8587 | bellard | if (offset != 0) { |
2817 | e97fc193 | aliguori | errors += inc_refcounts(bs, refcount_table, nb_clusters, |
2818 | 585f8587 | bellard | offset, s->cluster_size); |
2819 | 585f8587 | bellard | } |
2820 | 585f8587 | bellard | } |
2821 | 585f8587 | bellard | |
2822 | 585f8587 | bellard | /* compare ref counts */
|
2823 | 585f8587 | bellard | for(i = 0; i < nb_clusters; i++) { |
2824 | 585f8587 | bellard | refcount1 = get_refcount(bs, i); |
2825 | 585f8587 | bellard | refcount2 = refcount_table[i]; |
2826 | e97fc193 | aliguori | if (refcount1 != refcount2) {
|
2827 | 8ddbc04f | aliguori | fprintf(stderr, "ERROR cluster %d refcount=%d reference=%d\n",
|
2828 | 585f8587 | bellard | i, refcount1, refcount2); |
2829 | e97fc193 | aliguori | errors++; |
2830 | e97fc193 | aliguori | } |
2831 | 585f8587 | bellard | } |
2832 | 585f8587 | bellard | |
2833 | 585f8587 | bellard | qemu_free(refcount_table); |
2834 | e97fc193 | aliguori | |
2835 | e97fc193 | aliguori | return errors;
|
2836 | e97fc193 | aliguori | } |
2837 | e97fc193 | aliguori | |
2838 | e97fc193 | aliguori | static int qcow_check(BlockDriverState *bs) |
2839 | e97fc193 | aliguori | { |
2840 | e97fc193 | aliguori | return check_refcounts(bs);
|
2841 | 585f8587 | bellard | } |
2842 | 585f8587 | bellard | |
2843 | 585f8587 | bellard | #if 0
|
2844 | 585f8587 | bellard | static void dump_refcounts(BlockDriverState *bs)
|
2845 | 585f8587 | bellard | {
|
2846 | 585f8587 | bellard | BDRVQcowState *s = bs->opaque;
|
2847 | 585f8587 | bellard | int64_t nb_clusters, k, k1, size;
|
2848 | 585f8587 | bellard | int refcount;
|
2849 | 585f8587 | bellard | |
2850 | 585f8587 | bellard | size = bdrv_getlength(s->hd);
|
2851 | 6db6c638 | aliguori | nb_clusters = size_to_clusters(s, size);
|
2852 | 585f8587 | bellard | for(k = 0; k < nb_clusters;) {
|
2853 | 585f8587 | bellard | k1 = k;
|
2854 | 585f8587 | bellard | refcount = get_refcount(bs, k);
|
2855 | 585f8587 | bellard | k++;
|
2856 | 585f8587 | bellard | while (k < nb_clusters && get_refcount(bs, k) == refcount)
|
2857 | 585f8587 | bellard | k++;
|
2858 | 585f8587 | bellard | printf("%lld: refcount=%d nb=%lld\n", k, refcount, k - k1);
|
2859 | 585f8587 | bellard | }
|
2860 | 585f8587 | bellard | }
|
2861 | 585f8587 | bellard | #endif
|
2862 | 585f8587 | bellard | |
2863 | 178e08a5 | aliguori | static int qcow_put_buffer(BlockDriverState *bs, const uint8_t *buf, |
2864 | 178e08a5 | aliguori | int64_t pos, int size)
|
2865 | 178e08a5 | aliguori | { |
2866 | 178e08a5 | aliguori | int growable = bs->growable;
|
2867 | 178e08a5 | aliguori | |
2868 | 178e08a5 | aliguori | bs->growable = 1;
|
2869 | 178e08a5 | aliguori | bdrv_pwrite(bs, pos, buf, size); |
2870 | 178e08a5 | aliguori | bs->growable = growable; |
2871 | 178e08a5 | aliguori | |
2872 | 178e08a5 | aliguori | return size;
|
2873 | 178e08a5 | aliguori | } |
2874 | 178e08a5 | aliguori | |
2875 | 178e08a5 | aliguori | static int qcow_get_buffer(BlockDriverState *bs, uint8_t *buf, |
2876 | 178e08a5 | aliguori | int64_t pos, int size)
|
2877 | 178e08a5 | aliguori | { |
2878 | 178e08a5 | aliguori | int growable = bs->growable;
|
2879 | 178e08a5 | aliguori | int ret;
|
2880 | 178e08a5 | aliguori | |
2881 | 178e08a5 | aliguori | bs->growable = 1;
|
2882 | 178e08a5 | aliguori | ret = bdrv_pread(bs, pos, buf, size); |
2883 | 178e08a5 | aliguori | bs->growable = growable; |
2884 | 178e08a5 | aliguori | |
2885 | 178e08a5 | aliguori | return ret;
|
2886 | 178e08a5 | aliguori | } |
2887 | 178e08a5 | aliguori | |
2888 | 585f8587 | bellard | BlockDriver bdrv_qcow2 = { |
2889 | e60f469c | aurel32 | .format_name = "qcow2",
|
2890 | e60f469c | aurel32 | .instance_size = sizeof(BDRVQcowState),
|
2891 | e60f469c | aurel32 | .bdrv_probe = qcow_probe, |
2892 | e60f469c | aurel32 | .bdrv_open = qcow_open, |
2893 | e60f469c | aurel32 | .bdrv_close = qcow_close, |
2894 | e60f469c | aurel32 | .bdrv_create = qcow_create, |
2895 | e60f469c | aurel32 | .bdrv_flush = qcow_flush, |
2896 | e60f469c | aurel32 | .bdrv_is_allocated = qcow_is_allocated, |
2897 | e60f469c | aurel32 | .bdrv_set_key = qcow_set_key, |
2898 | e60f469c | aurel32 | .bdrv_make_empty = qcow_make_empty, |
2899 | e60f469c | aurel32 | |
2900 | f141eafe | aliguori | .bdrv_aio_readv = qcow_aio_readv, |
2901 | f141eafe | aliguori | .bdrv_aio_writev = qcow_aio_writev, |
2902 | e60f469c | aurel32 | .bdrv_aio_cancel = qcow_aio_cancel, |
2903 | e60f469c | aurel32 | .aiocb_size = sizeof(QCowAIOCB),
|
2904 | 585f8587 | bellard | .bdrv_write_compressed = qcow_write_compressed, |
2905 | 585f8587 | bellard | |
2906 | 585f8587 | bellard | .bdrv_snapshot_create = qcow_snapshot_create, |
2907 | e60f469c | aurel32 | .bdrv_snapshot_goto = qcow_snapshot_goto, |
2908 | 585f8587 | bellard | .bdrv_snapshot_delete = qcow_snapshot_delete, |
2909 | e60f469c | aurel32 | .bdrv_snapshot_list = qcow_snapshot_list, |
2910 | e60f469c | aurel32 | .bdrv_get_info = qcow_get_info, |
2911 | f965509c | aliguori | |
2912 | 178e08a5 | aliguori | .bdrv_put_buffer = qcow_put_buffer, |
2913 | 178e08a5 | aliguori | .bdrv_get_buffer = qcow_get_buffer, |
2914 | 178e08a5 | aliguori | |
2915 | f965509c | aliguori | .bdrv_create2 = qcow_create2, |
2916 | e97fc193 | aliguori | .bdrv_check = qcow_check, |
2917 | 585f8587 | bellard | }; |