]> git.proxmox.com Git - qemu.git/blame - block-vmdk.c
new disk image layer
[qemu.git] / block-vmdk.c
CommitLineData
ea2384d3
FB
1/*
2 * Block driver for the VMDK format
3 *
4 * Copyright (c) 2004 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24#include "vl.h"
25#include "block_int.h"
26
27/* XXX: this code is untested */
28/* XXX: add write support */
29
30#define VMDK3_MAGIC (('C' << 24) | ('O' << 16) | ('W' << 8) | 'D')
31#define VMDK4_MAGIC (('K' << 24) | ('D' << 16) | ('M' << 8) | 'V')
32
33typedef struct {
34 uint32_t version;
35 uint32_t flags;
36 uint32_t disk_sectors;
37 uint32_t granularity;
38 uint32_t l1dir_offset;
39 uint32_t l1dir_size;
40 uint32_t file_sectors;
41 uint32_t cylinders;
42 uint32_t heads;
43 uint32_t sectors_per_track;
44} VMDK3Header;
45
46typedef struct {
47 uint32_t version;
48 uint32_t flags;
49 int64_t capacity;
50 int64_t granularity;
51 int64_t desc_offset;
52 int64_t desc_size;
53 int32_t num_gtes_per_gte;
54 int64_t rgd_offset;
55 int64_t gd_offset;
56 int64_t grain_offset;
57 char filler[1];
58 char check_bytes[4];
59} VMDK4Header;
60
61#define L2_CACHE_SIZE 16
62
63typedef struct BDRVVmdkState {
64 int fd;
65 int64_t l1_table_offset;
66 uint32_t *l1_table;
67 unsigned int l1_size;
68 uint32_t l1_entry_sectors;
69
70 unsigned int l2_size;
71 uint32_t *l2_cache;
72 uint32_t l2_cache_offsets[L2_CACHE_SIZE];
73 uint32_t l2_cache_counts[L2_CACHE_SIZE];
74
75 unsigned int cluster_sectors;
76} BDRVVmdkState;
77
78static int vmdk_probe(const uint8_t *buf, int buf_size, const char *filename)
79{
80 uint32_t magic;
81
82 if (buf_size < 4)
83 return 0;
84 magic = be32_to_cpu(*(uint32_t *)buf);
85 if (magic == VMDK3_MAGIC ||
86 magic == VMDK4_MAGIC)
87 return 100;
88 else
89 return 0;
90}
91
92static int vmdk_open(BlockDriverState *bs, const char *filename)
93{
94 BDRVVmdkState *s = bs->opaque;
95 int fd, i;
96 uint32_t magic;
97 int l1_size;
98
99 fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
100 if (fd < 0)
101 return -1;
102 if (read(fd, &magic, sizeof(magic)) != sizeof(magic))
103 goto fail;
104 magic = le32_to_cpu(magic);
105
106 if (magic == VMDK3_MAGIC) {
107 VMDK3Header header;
108 if (read(fd, &header, sizeof(header)) !=
109 sizeof(header))
110 goto fail;
111 s->cluster_sectors = le32_to_cpu(header.granularity);
112 s->l2_size = 1 << 9;
113 s->l1_size = 1 << 6;
114 bs->total_sectors = le32_to_cpu(header.disk_sectors);
115 s->l1_table_offset = le32_to_cpu(header.l1dir_offset) * 512;
116 s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
117 } else if (magic == VMDK4_MAGIC) {
118 VMDK4Header header;
119
120 if (read(fd, &header, sizeof(header)) != sizeof(header))
121 goto fail;
122 bs->total_sectors = le32_to_cpu(header.capacity);
123 s->cluster_sectors = le32_to_cpu(header.granularity);
124 s->l2_size = le32_to_cpu(header.num_gtes_per_gte);
125 s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
126 if (s->l1_entry_sectors <= 0)
127 goto fail;
128 s->l1_size = (bs->total_sectors + s->l1_entry_sectors - 1)
129 / s->l1_entry_sectors;
130 s->l1_table_offset = le64_to_cpu(header.rgd_offset) * 512;
131 } else {
132 goto fail;
133 }
134 /* read the L1 table */
135 l1_size = s->l1_size * sizeof(uint32_t);
136 s->l1_table = qemu_malloc(l1_size);
137 if (!s->l1_table)
138 goto fail;
139 if (read(s->fd, s->l1_table, l1_size) != l1_size)
140 goto fail;
141 for(i = 0; i < s->l1_size; i++) {
142 le32_to_cpus(&s->l1_table[i]);
143 }
144
145 s->l2_cache = qemu_malloc(s->l2_size * L2_CACHE_SIZE * sizeof(uint32_t));
146 if (!s->l2_cache)
147 goto fail;
148 s->fd = fd;
149 /* XXX: currently only read only */
150 bs->read_only = 1;
151 return 0;
152 fail:
153 qemu_free(s->l1_table);
154 qemu_free(s->l2_cache);
155 close(fd);
156 return -1;
157}
158
159static uint64_t get_cluster_offset(BlockDriverState *bs,
160 uint64_t offset)
161{
162 BDRVVmdkState *s = bs->opaque;
163 unsigned int l1_index, l2_offset, l2_index;
164 int min_index, i, j;
165 uint32_t min_count, *l2_table;
166 uint64_t cluster_offset;
167
168 l1_index = (offset >> 9) / s->l1_entry_sectors;
169 if (l1_index >= s->l1_size)
170 return 0;
171 l2_offset = s->l1_table[l1_index];
172 if (!l2_offset)
173 return 0;
174
175 for(i = 0; i < L2_CACHE_SIZE; i++) {
176 if (l2_offset == s->l2_cache_offsets[i]) {
177 /* increment the hit count */
178 if (++s->l2_cache_counts[i] == 0xffffffff) {
179 for(j = 0; j < L2_CACHE_SIZE; j++) {
180 s->l2_cache_counts[j] >>= 1;
181 }
182 }
183 l2_table = s->l2_cache + (i * s->l2_size);
184 goto found;
185 }
186 }
187 /* not found: load a new entry in the least used one */
188 min_index = 0;
189 min_count = 0xffffffff;
190 for(i = 0; i < L2_CACHE_SIZE; i++) {
191 if (s->l2_cache_counts[i] < min_count) {
192 min_count = s->l2_cache_counts[i];
193 min_index = i;
194 }
195 }
196 l2_table = s->l2_cache + (min_index * s->l2_size);
197 lseek(s->fd, (int64_t)l2_offset * 512, SEEK_SET);
198 if (read(s->fd, l2_table, s->l2_size * sizeof(uint32_t)) !=
199 s->l2_size * sizeof(uint32_t))
200 return 0;
201 s->l2_cache_offsets[min_index] = l2_offset;
202 s->l2_cache_counts[min_index] = 1;
203 found:
204 l2_index = ((offset >> 9) / s->cluster_sectors) % s->l2_size;
205 cluster_offset = le32_to_cpu(l2_table[l2_index]);
206 cluster_offset <<= 9;
207 return cluster_offset;
208}
209
210static int vmdk_is_allocated(BlockDriverState *bs, int64_t sector_num,
211 int nb_sectors, int *pnum)
212{
213 BDRVVmdkState *s = bs->opaque;
214 int index_in_cluster, n;
215 uint64_t cluster_offset;
216
217 cluster_offset = get_cluster_offset(bs, sector_num << 9);
218 index_in_cluster = sector_num % s->cluster_sectors;
219 n = s->cluster_sectors - index_in_cluster;
220 if (n > nb_sectors)
221 n = nb_sectors;
222 *pnum = n;
223 return (cluster_offset != 0);
224}
225
226static int vmdk_read(BlockDriverState *bs, int64_t sector_num,
227 uint8_t *buf, int nb_sectors)
228{
229 BDRVVmdkState *s = bs->opaque;
230 int ret, index_in_cluster, n;
231 uint64_t cluster_offset;
232
233 while (nb_sectors > 0) {
234 cluster_offset = get_cluster_offset(bs, sector_num << 9);
235 index_in_cluster = sector_num % s->cluster_sectors;
236 n = s->cluster_sectors - index_in_cluster;
237 if (n > nb_sectors)
238 n = nb_sectors;
239 if (!cluster_offset) {
240 memset(buf, 0, 512 * n);
241 } else {
242 lseek64(s->fd, cluster_offset + index_in_cluster * 512, SEEK_SET);
243 ret = read(s->fd, buf, n * 512);
244 if (ret != n * 512)
245 return -1;
246 }
247 nb_sectors -= n;
248 sector_num += n;
249 buf += n * 512;
250 }
251 return 0;
252}
253
254static int vmdk_write(BlockDriverState *bs, int64_t sector_num,
255 const uint8_t *buf, int nb_sectors)
256{
257 return -1;
258}
259
260static int vmdk_close(BlockDriverState *bs)
261{
262 BDRVVmdkState *s = bs->opaque;
263 qemu_free(s->l1_table);
264 qemu_free(s->l2_cache);
265 close(s->fd);
266}
267
268BlockDriver bdrv_vmdk = {
269 "vmdk",
270 sizeof(BDRVVmdkState),
271 vmdk_probe,
272 vmdk_open,
273 vmdk_read,
274 vmdk_write,
275 vmdk_close,
276 NULL, /* no create yet */
277 vmdk_is_allocated,
278};