root / block / qcow2.c @ f67503e5
History | View | Annotate | Download (70.4 kB)
1 |
/*
|
---|---|
2 |
* Block driver for the QCOW version 2 format
|
3 |
*
|
4 |
* Copyright (c) 2004-2006 Fabrice Bellard
|
5 |
*
|
6 |
* Permission is hereby granted, free of charge, to any person obtaining a copy
|
7 |
* of this software and associated documentation files (the "Software"), to deal
|
8 |
* in the Software without restriction, including without limitation the rights
|
9 |
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
10 |
* copies of the Software, and to permit persons to whom the Software is
|
11 |
* furnished to do so, subject to the following conditions:
|
12 |
*
|
13 |
* The above copyright notice and this permission notice shall be included in
|
14 |
* all copies or substantial portions of the Software.
|
15 |
*
|
16 |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
17 |
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
18 |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
19 |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
20 |
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
21 |
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
22 |
* THE SOFTWARE.
|
23 |
*/
|
24 |
#include "qemu-common.h" |
25 |
#include "block/block_int.h" |
26 |
#include "qemu/module.h" |
27 |
#include <zlib.h> |
28 |
#include "qemu/aes.h" |
29 |
#include "block/qcow2.h" |
30 |
#include "qemu/error-report.h" |
31 |
#include "qapi/qmp/qerror.h" |
32 |
#include "qapi/qmp/qbool.h" |
33 |
#include "trace.h" |
34 |
|
35 |
/*
|
36 |
Differences with QCOW:
|
37 |
|
38 |
- Support for multiple incremental snapshots.
|
39 |
- Memory management by reference counts.
|
40 |
- Clusters which have a reference count of one have the bit
|
41 |
QCOW_OFLAG_COPIED to optimize write performance.
|
42 |
- Size of compressed clusters is stored in sectors to reduce bit usage
|
43 |
in the cluster offsets.
|
44 |
- Support for storing additional data (such as the VM state) in the
|
45 |
snapshots.
|
46 |
- If a backing store is used, the cluster size is not constrained
|
47 |
(could be backported to QCOW).
|
48 |
- L2 tables have always a size of one cluster.
|
49 |
*/
|
50 |
|
51 |
|
52 |
typedef struct { |
53 |
uint32_t magic; |
54 |
uint32_t len; |
55 |
} QEMU_PACKED QCowExtension; |
56 |
|
57 |
#define QCOW2_EXT_MAGIC_END 0 |
58 |
#define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA |
59 |
#define QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857 |
60 |
|
61 |
static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename) |
62 |
{ |
63 |
const QCowHeader *cow_header = (const void *)buf; |
64 |
|
65 |
if (buf_size >= sizeof(QCowHeader) && |
66 |
be32_to_cpu(cow_header->magic) == QCOW_MAGIC && |
67 |
be32_to_cpu(cow_header->version) >= 2)
|
68 |
return 100; |
69 |
else
|
70 |
return 0; |
71 |
} |
72 |
|
73 |
|
74 |
/*
|
75 |
* read qcow2 extension and fill bs
|
76 |
* start reading from start_offset
|
77 |
* finish reading upon magic of value 0 or when end_offset reached
|
78 |
* unknown magic is skipped (future extension this version knows nothing about)
|
79 |
* return 0 upon success, non-0 otherwise
|
80 |
*/
|
81 |
static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset, |
82 |
uint64_t end_offset, void **p_feature_table,
|
83 |
Error **errp) |
84 |
{ |
85 |
BDRVQcowState *s = bs->opaque; |
86 |
QCowExtension ext; |
87 |
uint64_t offset; |
88 |
int ret;
|
89 |
|
90 |
#ifdef DEBUG_EXT
|
91 |
printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset);
|
92 |
#endif
|
93 |
offset = start_offset; |
94 |
while (offset < end_offset) {
|
95 |
|
96 |
#ifdef DEBUG_EXT
|
97 |
/* Sanity check */
|
98 |
if (offset > s->cluster_size)
|
99 |
printf("qcow2_read_extension: suspicious offset %lu\n", offset);
|
100 |
|
101 |
printf("attempting to read extended header in offset %lu\n", offset);
|
102 |
#endif
|
103 |
|
104 |
ret = bdrv_pread(bs->file, offset, &ext, sizeof(ext));
|
105 |
if (ret < 0) { |
106 |
error_setg_errno(errp, -ret, "qcow2_read_extension: ERROR: "
|
107 |
"pread fail from offset %" PRIu64, offset);
|
108 |
return 1; |
109 |
} |
110 |
be32_to_cpus(&ext.magic); |
111 |
be32_to_cpus(&ext.len); |
112 |
offset += sizeof(ext);
|
113 |
#ifdef DEBUG_EXT
|
114 |
printf("ext.magic = 0x%x\n", ext.magic);
|
115 |
#endif
|
116 |
if (ext.len > end_offset - offset) {
|
117 |
error_setg(errp, "Header extension too large");
|
118 |
return -EINVAL;
|
119 |
} |
120 |
|
121 |
switch (ext.magic) {
|
122 |
case QCOW2_EXT_MAGIC_END:
|
123 |
return 0; |
124 |
|
125 |
case QCOW2_EXT_MAGIC_BACKING_FORMAT:
|
126 |
if (ext.len >= sizeof(bs->backing_format)) { |
127 |
error_setg(errp, "ERROR: ext_backing_format: len=%u too large"
|
128 |
" (>=%zu)", ext.len, sizeof(bs->backing_format)); |
129 |
return 2; |
130 |
} |
131 |
ret = bdrv_pread(bs->file, offset, bs->backing_format, ext.len); |
132 |
if (ret < 0) { |
133 |
error_setg_errno(errp, -ret, "ERROR: ext_backing_format: "
|
134 |
"Could not read format name");
|
135 |
return 3; |
136 |
} |
137 |
bs->backing_format[ext.len] = '\0';
|
138 |
#ifdef DEBUG_EXT
|
139 |
printf("Qcow2: Got format extension %s\n", bs->backing_format);
|
140 |
#endif
|
141 |
break;
|
142 |
|
143 |
case QCOW2_EXT_MAGIC_FEATURE_TABLE:
|
144 |
if (p_feature_table != NULL) { |
145 |
void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature)); |
146 |
ret = bdrv_pread(bs->file, offset , feature_table, ext.len); |
147 |
if (ret < 0) { |
148 |
error_setg_errno(errp, -ret, "ERROR: ext_feature_table: "
|
149 |
"Could not read table");
|
150 |
return ret;
|
151 |
} |
152 |
|
153 |
*p_feature_table = feature_table; |
154 |
} |
155 |
break;
|
156 |
|
157 |
default:
|
158 |
/* unknown magic - save it in case we need to rewrite the header */
|
159 |
{ |
160 |
Qcow2UnknownHeaderExtension *uext; |
161 |
|
162 |
uext = g_malloc0(sizeof(*uext) + ext.len);
|
163 |
uext->magic = ext.magic; |
164 |
uext->len = ext.len; |
165 |
QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next); |
166 |
|
167 |
ret = bdrv_pread(bs->file, offset , uext->data, uext->len); |
168 |
if (ret < 0) { |
169 |
error_setg_errno(errp, -ret, "ERROR: unknown extension: "
|
170 |
"Could not read data");
|
171 |
return ret;
|
172 |
} |
173 |
} |
174 |
break;
|
175 |
} |
176 |
|
177 |
offset += ((ext.len + 7) & ~7); |
178 |
} |
179 |
|
180 |
return 0; |
181 |
} |
182 |
|
183 |
static void cleanup_unknown_header_ext(BlockDriverState *bs) |
184 |
{ |
185 |
BDRVQcowState *s = bs->opaque; |
186 |
Qcow2UnknownHeaderExtension *uext, *next; |
187 |
|
188 |
QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) { |
189 |
QLIST_REMOVE(uext, next); |
190 |
g_free(uext); |
191 |
} |
192 |
} |
193 |
|
194 |
static void GCC_FMT_ATTR(3, 4) report_unsupported(BlockDriverState *bs, |
195 |
Error **errp, const char *fmt, ...) |
196 |
{ |
197 |
char msg[64]; |
198 |
va_list ap; |
199 |
|
200 |
va_start(ap, fmt); |
201 |
vsnprintf(msg, sizeof(msg), fmt, ap);
|
202 |
va_end(ap); |
203 |
|
204 |
error_set(errp, QERR_UNKNOWN_BLOCK_FORMAT_FEATURE, bs->device_name, "qcow2",
|
205 |
msg); |
206 |
} |
207 |
|
208 |
static void report_unsupported_feature(BlockDriverState *bs, |
209 |
Error **errp, Qcow2Feature *table, uint64_t mask) |
210 |
{ |
211 |
while (table && table->name[0] != '\0') { |
212 |
if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) {
|
213 |
if (mask & (1 << table->bit)) { |
214 |
report_unsupported(bs, errp, "%.46s", table->name);
|
215 |
mask &= ~(1 << table->bit);
|
216 |
} |
217 |
} |
218 |
table++; |
219 |
} |
220 |
|
221 |
if (mask) {
|
222 |
report_unsupported(bs, errp, "Unknown incompatible feature: %" PRIx64,
|
223 |
mask); |
224 |
} |
225 |
} |
226 |
|
227 |
/*
|
228 |
* Sets the dirty bit and flushes afterwards if necessary.
|
229 |
*
|
230 |
* The incompatible_features bit is only set if the image file header was
|
231 |
* updated successfully. Therefore it is not required to check the return
|
232 |
* value of this function.
|
233 |
*/
|
234 |
int qcow2_mark_dirty(BlockDriverState *bs)
|
235 |
{ |
236 |
BDRVQcowState *s = bs->opaque; |
237 |
uint64_t val; |
238 |
int ret;
|
239 |
|
240 |
assert(s->qcow_version >= 3);
|
241 |
|
242 |
if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) {
|
243 |
return 0; /* already dirty */ |
244 |
} |
245 |
|
246 |
val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY); |
247 |
ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features), |
248 |
&val, sizeof(val));
|
249 |
if (ret < 0) { |
250 |
return ret;
|
251 |
} |
252 |
ret = bdrv_flush(bs->file); |
253 |
if (ret < 0) { |
254 |
return ret;
|
255 |
} |
256 |
|
257 |
/* Only treat image as dirty if the header was updated successfully */
|
258 |
s->incompatible_features |= QCOW2_INCOMPAT_DIRTY; |
259 |
return 0; |
260 |
} |
261 |
|
262 |
/*
|
263 |
* Clears the dirty bit and flushes before if necessary. Only call this
|
264 |
* function when there are no pending requests, it does not guard against
|
265 |
* concurrent requests dirtying the image.
|
266 |
*/
|
267 |
static int qcow2_mark_clean(BlockDriverState *bs) |
268 |
{ |
269 |
BDRVQcowState *s = bs->opaque; |
270 |
|
271 |
if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) {
|
272 |
int ret = bdrv_flush(bs);
|
273 |
if (ret < 0) { |
274 |
return ret;
|
275 |
} |
276 |
|
277 |
s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY; |
278 |
return qcow2_update_header(bs);
|
279 |
} |
280 |
return 0; |
281 |
} |
282 |
|
283 |
/*
|
284 |
* Marks the image as corrupt.
|
285 |
*/
|
286 |
int qcow2_mark_corrupt(BlockDriverState *bs)
|
287 |
{ |
288 |
BDRVQcowState *s = bs->opaque; |
289 |
|
290 |
s->incompatible_features |= QCOW2_INCOMPAT_CORRUPT; |
291 |
return qcow2_update_header(bs);
|
292 |
} |
293 |
|
294 |
/*
|
295 |
* Marks the image as consistent, i.e., unsets the corrupt bit, and flushes
|
296 |
* before if necessary.
|
297 |
*/
|
298 |
int qcow2_mark_consistent(BlockDriverState *bs)
|
299 |
{ |
300 |
BDRVQcowState *s = bs->opaque; |
301 |
|
302 |
if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) {
|
303 |
int ret = bdrv_flush(bs);
|
304 |
if (ret < 0) { |
305 |
return ret;
|
306 |
} |
307 |
|
308 |
s->incompatible_features &= ~QCOW2_INCOMPAT_CORRUPT; |
309 |
return qcow2_update_header(bs);
|
310 |
} |
311 |
return 0; |
312 |
} |
313 |
|
314 |
static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result, |
315 |
BdrvCheckMode fix) |
316 |
{ |
317 |
int ret = qcow2_check_refcounts(bs, result, fix);
|
318 |
if (ret < 0) { |
319 |
return ret;
|
320 |
} |
321 |
|
322 |
if (fix && result->check_errors == 0 && result->corruptions == 0) { |
323 |
ret = qcow2_mark_clean(bs); |
324 |
if (ret < 0) { |
325 |
return ret;
|
326 |
} |
327 |
return qcow2_mark_consistent(bs);
|
328 |
} |
329 |
return ret;
|
330 |
} |
331 |
|
332 |
static QemuOptsList qcow2_runtime_opts = {
|
333 |
.name = "qcow2",
|
334 |
.head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head), |
335 |
.desc = { |
336 |
{ |
337 |
.name = QCOW2_OPT_LAZY_REFCOUNTS, |
338 |
.type = QEMU_OPT_BOOL, |
339 |
.help = "Postpone refcount updates",
|
340 |
}, |
341 |
{ |
342 |
.name = QCOW2_OPT_DISCARD_REQUEST, |
343 |
.type = QEMU_OPT_BOOL, |
344 |
.help = "Pass guest discard requests to the layer below",
|
345 |
}, |
346 |
{ |
347 |
.name = QCOW2_OPT_DISCARD_SNAPSHOT, |
348 |
.type = QEMU_OPT_BOOL, |
349 |
.help = "Generate discard requests when snapshot related space "
|
350 |
"is freed",
|
351 |
}, |
352 |
{ |
353 |
.name = QCOW2_OPT_DISCARD_OTHER, |
354 |
.type = QEMU_OPT_BOOL, |
355 |
.help = "Generate discard requests when other clusters are freed",
|
356 |
}, |
357 |
{ |
358 |
.name = QCOW2_OPT_OVERLAP, |
359 |
.type = QEMU_OPT_STRING, |
360 |
.help = "Selects which overlap checks to perform from a range of "
|
361 |
"templates (none, constant, cached, all)",
|
362 |
}, |
363 |
{ |
364 |
.name = QCOW2_OPT_OVERLAP_MAIN_HEADER, |
365 |
.type = QEMU_OPT_BOOL, |
366 |
.help = "Check for unintended writes into the main qcow2 header",
|
367 |
}, |
368 |
{ |
369 |
.name = QCOW2_OPT_OVERLAP_ACTIVE_L1, |
370 |
.type = QEMU_OPT_BOOL, |
371 |
.help = "Check for unintended writes into the active L1 table",
|
372 |
}, |
373 |
{ |
374 |
.name = QCOW2_OPT_OVERLAP_ACTIVE_L2, |
375 |
.type = QEMU_OPT_BOOL, |
376 |
.help = "Check for unintended writes into an active L2 table",
|
377 |
}, |
378 |
{ |
379 |
.name = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, |
380 |
.type = QEMU_OPT_BOOL, |
381 |
.help = "Check for unintended writes into the refcount table",
|
382 |
}, |
383 |
{ |
384 |
.name = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, |
385 |
.type = QEMU_OPT_BOOL, |
386 |
.help = "Check for unintended writes into a refcount block",
|
387 |
}, |
388 |
{ |
389 |
.name = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, |
390 |
.type = QEMU_OPT_BOOL, |
391 |
.help = "Check for unintended writes into the snapshot table",
|
392 |
}, |
393 |
{ |
394 |
.name = QCOW2_OPT_OVERLAP_INACTIVE_L1, |
395 |
.type = QEMU_OPT_BOOL, |
396 |
.help = "Check for unintended writes into an inactive L1 table",
|
397 |
}, |
398 |
{ |
399 |
.name = QCOW2_OPT_OVERLAP_INACTIVE_L2, |
400 |
.type = QEMU_OPT_BOOL, |
401 |
.help = "Check for unintended writes into an inactive L2 table",
|
402 |
}, |
403 |
{ /* end of list */ }
|
404 |
}, |
405 |
}; |
406 |
|
407 |
static const char *overlap_bool_option_names[QCOW2_OL_MAX_BITNR] = { |
408 |
[QCOW2_OL_MAIN_HEADER_BITNR] = QCOW2_OPT_OVERLAP_MAIN_HEADER, |
409 |
[QCOW2_OL_ACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L1, |
410 |
[QCOW2_OL_ACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L2, |
411 |
[QCOW2_OL_REFCOUNT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, |
412 |
[QCOW2_OL_REFCOUNT_BLOCK_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, |
413 |
[QCOW2_OL_SNAPSHOT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, |
414 |
[QCOW2_OL_INACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L1, |
415 |
[QCOW2_OL_INACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L2, |
416 |
}; |
417 |
|
418 |
static int qcow2_open(BlockDriverState *bs, QDict *options, int flags, |
419 |
Error **errp) |
420 |
{ |
421 |
BDRVQcowState *s = bs->opaque; |
422 |
int len, i, ret = 0; |
423 |
QCowHeader header; |
424 |
QemuOpts *opts; |
425 |
Error *local_err = NULL;
|
426 |
uint64_t ext_end; |
427 |
uint64_t l1_vm_state_index; |
428 |
const char *opt_overlap_check; |
429 |
int overlap_check_template = 0; |
430 |
|
431 |
ret = bdrv_pread(bs->file, 0, &header, sizeof(header)); |
432 |
if (ret < 0) { |
433 |
error_setg_errno(errp, -ret, "Could not read qcow2 header");
|
434 |
goto fail;
|
435 |
} |
436 |
be32_to_cpus(&header.magic); |
437 |
be32_to_cpus(&header.version); |
438 |
be64_to_cpus(&header.backing_file_offset); |
439 |
be32_to_cpus(&header.backing_file_size); |
440 |
be64_to_cpus(&header.size); |
441 |
be32_to_cpus(&header.cluster_bits); |
442 |
be32_to_cpus(&header.crypt_method); |
443 |
be64_to_cpus(&header.l1_table_offset); |
444 |
be32_to_cpus(&header.l1_size); |
445 |
be64_to_cpus(&header.refcount_table_offset); |
446 |
be32_to_cpus(&header.refcount_table_clusters); |
447 |
be64_to_cpus(&header.snapshots_offset); |
448 |
be32_to_cpus(&header.nb_snapshots); |
449 |
|
450 |
if (header.magic != QCOW_MAGIC) {
|
451 |
error_setg(errp, "Image is not in qcow2 format");
|
452 |
ret = -EMEDIUMTYPE; |
453 |
goto fail;
|
454 |
} |
455 |
if (header.version < 2 || header.version > 3) { |
456 |
report_unsupported(bs, errp, "QCOW version %d", header.version);
|
457 |
ret = -ENOTSUP; |
458 |
goto fail;
|
459 |
} |
460 |
|
461 |
s->qcow_version = header.version; |
462 |
|
463 |
/* Initialise version 3 header fields */
|
464 |
if (header.version == 2) { |
465 |
header.incompatible_features = 0;
|
466 |
header.compatible_features = 0;
|
467 |
header.autoclear_features = 0;
|
468 |
header.refcount_order = 4;
|
469 |
header.header_length = 72;
|
470 |
} else {
|
471 |
be64_to_cpus(&header.incompatible_features); |
472 |
be64_to_cpus(&header.compatible_features); |
473 |
be64_to_cpus(&header.autoclear_features); |
474 |
be32_to_cpus(&header.refcount_order); |
475 |
be32_to_cpus(&header.header_length); |
476 |
} |
477 |
|
478 |
if (header.header_length > sizeof(header)) { |
479 |
s->unknown_header_fields_size = header.header_length - sizeof(header);
|
480 |
s->unknown_header_fields = g_malloc(s->unknown_header_fields_size); |
481 |
ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields,
|
482 |
s->unknown_header_fields_size); |
483 |
if (ret < 0) { |
484 |
error_setg_errno(errp, -ret, "Could not read unknown qcow2 header "
|
485 |
"fields");
|
486 |
goto fail;
|
487 |
} |
488 |
} |
489 |
|
490 |
if (header.backing_file_offset) {
|
491 |
ext_end = header.backing_file_offset; |
492 |
} else {
|
493 |
ext_end = 1 << header.cluster_bits;
|
494 |
} |
495 |
|
496 |
/* Handle feature bits */
|
497 |
s->incompatible_features = header.incompatible_features; |
498 |
s->compatible_features = header.compatible_features; |
499 |
s->autoclear_features = header.autoclear_features; |
500 |
|
501 |
if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) {
|
502 |
void *feature_table = NULL; |
503 |
qcow2_read_extensions(bs, header.header_length, ext_end, |
504 |
&feature_table, NULL);
|
505 |
report_unsupported_feature(bs, errp, feature_table, |
506 |
s->incompatible_features & |
507 |
~QCOW2_INCOMPAT_MASK); |
508 |
ret = -ENOTSUP; |
509 |
goto fail;
|
510 |
} |
511 |
|
512 |
if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) {
|
513 |
/* Corrupt images may not be written to unless they are being repaired
|
514 |
*/
|
515 |
if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_CHECK)) {
|
516 |
error_setg(errp, "qcow2: Image is corrupt; cannot be opened "
|
517 |
"read/write");
|
518 |
ret = -EACCES; |
519 |
goto fail;
|
520 |
} |
521 |
} |
522 |
|
523 |
/* Check support for various header values */
|
524 |
if (header.refcount_order != 4) { |
525 |
report_unsupported(bs, errp, "%d bit reference counts",
|
526 |
1 << header.refcount_order);
|
527 |
ret = -ENOTSUP; |
528 |
goto fail;
|
529 |
} |
530 |
s->refcount_order = header.refcount_order; |
531 |
|
532 |
if (header.cluster_bits < MIN_CLUSTER_BITS ||
|
533 |
header.cluster_bits > MAX_CLUSTER_BITS) { |
534 |
error_setg(errp, "Unsupported cluster size: 2^%i", header.cluster_bits);
|
535 |
ret = -EINVAL; |
536 |
goto fail;
|
537 |
} |
538 |
if (header.crypt_method > QCOW_CRYPT_AES) {
|
539 |
error_setg(errp, "Unsupported encryption method: %i",
|
540 |
header.crypt_method); |
541 |
ret = -EINVAL; |
542 |
goto fail;
|
543 |
} |
544 |
s->crypt_method_header = header.crypt_method; |
545 |
if (s->crypt_method_header) {
|
546 |
bs->encrypted = 1;
|
547 |
} |
548 |
s->cluster_bits = header.cluster_bits; |
549 |
s->cluster_size = 1 << s->cluster_bits;
|
550 |
s->cluster_sectors = 1 << (s->cluster_bits - 9); |
551 |
s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ |
552 |
s->l2_size = 1 << s->l2_bits;
|
553 |
bs->total_sectors = header.size / 512;
|
554 |
s->csize_shift = (62 - (s->cluster_bits - 8)); |
555 |
s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; |
556 |
s->cluster_offset_mask = (1LL << s->csize_shift) - 1; |
557 |
s->refcount_table_offset = header.refcount_table_offset; |
558 |
s->refcount_table_size = |
559 |
header.refcount_table_clusters << (s->cluster_bits - 3);
|
560 |
|
561 |
s->snapshots_offset = header.snapshots_offset; |
562 |
s->nb_snapshots = header.nb_snapshots; |
563 |
|
564 |
/* read the level 1 table */
|
565 |
s->l1_size = header.l1_size; |
566 |
|
567 |
l1_vm_state_index = size_to_l1(s, header.size); |
568 |
if (l1_vm_state_index > INT_MAX) {
|
569 |
error_setg(errp, "Image is too big");
|
570 |
ret = -EFBIG; |
571 |
goto fail;
|
572 |
} |
573 |
s->l1_vm_state_index = l1_vm_state_index; |
574 |
|
575 |
/* the L1 table must contain at least enough entries to put
|
576 |
header.size bytes */
|
577 |
if (s->l1_size < s->l1_vm_state_index) {
|
578 |
error_setg(errp, "L1 table is too small");
|
579 |
ret = -EINVAL; |
580 |
goto fail;
|
581 |
} |
582 |
s->l1_table_offset = header.l1_table_offset; |
583 |
if (s->l1_size > 0) { |
584 |
s->l1_table = g_malloc0( |
585 |
align_offset(s->l1_size * sizeof(uint64_t), 512)); |
586 |
ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table, |
587 |
s->l1_size * sizeof(uint64_t));
|
588 |
if (ret < 0) { |
589 |
error_setg_errno(errp, -ret, "Could not read L1 table");
|
590 |
goto fail;
|
591 |
} |
592 |
for(i = 0;i < s->l1_size; i++) { |
593 |
be64_to_cpus(&s->l1_table[i]); |
594 |
} |
595 |
} |
596 |
|
597 |
/* alloc L2 table/refcount block cache */
|
598 |
s->l2_table_cache = qcow2_cache_create(bs, L2_CACHE_SIZE); |
599 |
s->refcount_block_cache = qcow2_cache_create(bs, REFCOUNT_CACHE_SIZE); |
600 |
|
601 |
s->cluster_cache = g_malloc(s->cluster_size); |
602 |
/* one more sector for decompressed data alignment */
|
603 |
s->cluster_data = qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size |
604 |
+ 512);
|
605 |
s->cluster_cache_offset = -1;
|
606 |
s->flags = flags; |
607 |
|
608 |
ret = qcow2_refcount_init(bs); |
609 |
if (ret != 0) { |
610 |
error_setg_errno(errp, -ret, "Could not initialize refcount handling");
|
611 |
goto fail;
|
612 |
} |
613 |
|
614 |
QLIST_INIT(&s->cluster_allocs); |
615 |
QTAILQ_INIT(&s->discards); |
616 |
|
617 |
/* read qcow2 extensions */
|
618 |
if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL, |
619 |
&local_err)) { |
620 |
error_propagate(errp, local_err); |
621 |
ret = -EINVAL; |
622 |
goto fail;
|
623 |
} |
624 |
|
625 |
/* read the backing file name */
|
626 |
if (header.backing_file_offset != 0) { |
627 |
len = header.backing_file_size; |
628 |
if (len > 1023) { |
629 |
len = 1023;
|
630 |
} |
631 |
ret = bdrv_pread(bs->file, header.backing_file_offset, |
632 |
bs->backing_file, len); |
633 |
if (ret < 0) { |
634 |
error_setg_errno(errp, -ret, "Could not read backing file name");
|
635 |
goto fail;
|
636 |
} |
637 |
bs->backing_file[len] = '\0';
|
638 |
} |
639 |
|
640 |
ret = qcow2_read_snapshots(bs); |
641 |
if (ret < 0) { |
642 |
error_setg_errno(errp, -ret, "Could not read snapshots");
|
643 |
goto fail;
|
644 |
} |
645 |
|
646 |
/* Clear unknown autoclear feature bits */
|
647 |
if (!bs->read_only && s->autoclear_features != 0) { |
648 |
s->autoclear_features = 0;
|
649 |
ret = qcow2_update_header(bs); |
650 |
if (ret < 0) { |
651 |
error_setg_errno(errp, -ret, "Could not update qcow2 header");
|
652 |
goto fail;
|
653 |
} |
654 |
} |
655 |
|
656 |
/* Initialise locks */
|
657 |
qemu_co_mutex_init(&s->lock); |
658 |
|
659 |
/* Repair image if dirty */
|
660 |
if (!(flags & BDRV_O_CHECK) && !bs->read_only &&
|
661 |
(s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) { |
662 |
BdrvCheckResult result = {0};
|
663 |
|
664 |
ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS); |
665 |
if (ret < 0) { |
666 |
error_setg_errno(errp, -ret, "Could not repair dirty image");
|
667 |
goto fail;
|
668 |
} |
669 |
} |
670 |
|
671 |
/* Enable lazy_refcounts according to image and command line options */
|
672 |
opts = qemu_opts_create(&qcow2_runtime_opts, NULL, 0, &error_abort); |
673 |
qemu_opts_absorb_qdict(opts, options, &local_err); |
674 |
if (local_err) {
|
675 |
error_propagate(errp, local_err); |
676 |
ret = -EINVAL; |
677 |
goto fail;
|
678 |
} |
679 |
|
680 |
s->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS, |
681 |
(s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS)); |
682 |
|
683 |
s->discard_passthrough[QCOW2_DISCARD_NEVER] = false;
|
684 |
s->discard_passthrough[QCOW2_DISCARD_ALWAYS] = true;
|
685 |
s->discard_passthrough[QCOW2_DISCARD_REQUEST] = |
686 |
qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_REQUEST, |
687 |
flags & BDRV_O_UNMAP); |
688 |
s->discard_passthrough[QCOW2_DISCARD_SNAPSHOT] = |
689 |
qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_SNAPSHOT, true);
|
690 |
s->discard_passthrough[QCOW2_DISCARD_OTHER] = |
691 |
qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_OTHER, false);
|
692 |
|
693 |
opt_overlap_check = qemu_opt_get(opts, "overlap-check") ?: "cached"; |
694 |
if (!strcmp(opt_overlap_check, "none")) { |
695 |
overlap_check_template = 0;
|
696 |
} else if (!strcmp(opt_overlap_check, "constant")) { |
697 |
overlap_check_template = QCOW2_OL_CONSTANT; |
698 |
} else if (!strcmp(opt_overlap_check, "cached")) { |
699 |
overlap_check_template = QCOW2_OL_CACHED; |
700 |
} else if (!strcmp(opt_overlap_check, "all")) { |
701 |
overlap_check_template = QCOW2_OL_ALL; |
702 |
} else {
|
703 |
error_setg(errp, "Unsupported value '%s' for qcow2 option "
|
704 |
"'overlap-check'. Allowed are either of the following: "
|
705 |
"none, constant, cached, all", opt_overlap_check);
|
706 |
qemu_opts_del(opts); |
707 |
ret = -EINVAL; |
708 |
goto fail;
|
709 |
} |
710 |
|
711 |
s->overlap_check = 0;
|
712 |
for (i = 0; i < QCOW2_OL_MAX_BITNR; i++) { |
713 |
/* overlap-check defines a template bitmask, but every flag may be
|
714 |
* overwritten through the associated boolean option */
|
715 |
s->overlap_check |= |
716 |
qemu_opt_get_bool(opts, overlap_bool_option_names[i], |
717 |
overlap_check_template & (1 << i)) << i;
|
718 |
} |
719 |
|
720 |
qemu_opts_del(opts); |
721 |
|
722 |
if (s->use_lazy_refcounts && s->qcow_version < 3) { |
723 |
error_setg(errp, "Lazy refcounts require a qcow2 image with at least "
|
724 |
"qemu 1.1 compatibility level");
|
725 |
ret = -EINVAL; |
726 |
goto fail;
|
727 |
} |
728 |
|
729 |
#ifdef DEBUG_ALLOC
|
730 |
{ |
731 |
BdrvCheckResult result = {0};
|
732 |
qcow2_check_refcounts(bs, &result, 0);
|
733 |
} |
734 |
#endif
|
735 |
return ret;
|
736 |
|
737 |
fail:
|
738 |
g_free(s->unknown_header_fields); |
739 |
cleanup_unknown_header_ext(bs); |
740 |
qcow2_free_snapshots(bs); |
741 |
qcow2_refcount_close(bs); |
742 |
g_free(s->l1_table); |
743 |
/* else pre-write overlap checks in cache_destroy may crash */
|
744 |
s->l1_table = NULL;
|
745 |
if (s->l2_table_cache) {
|
746 |
qcow2_cache_destroy(bs, s->l2_table_cache); |
747 |
} |
748 |
g_free(s->cluster_cache); |
749 |
qemu_vfree(s->cluster_data); |
750 |
return ret;
|
751 |
} |
752 |
|
753 |
static int qcow2_refresh_limits(BlockDriverState *bs) |
754 |
{ |
755 |
BDRVQcowState *s = bs->opaque; |
756 |
|
757 |
bs->bl.write_zeroes_alignment = s->cluster_sectors; |
758 |
|
759 |
return 0; |
760 |
} |
761 |
|
762 |
static int qcow2_set_key(BlockDriverState *bs, const char *key) |
763 |
{ |
764 |
BDRVQcowState *s = bs->opaque; |
765 |
uint8_t keybuf[16];
|
766 |
int len, i;
|
767 |
|
768 |
memset(keybuf, 0, 16); |
769 |
len = strlen(key); |
770 |
if (len > 16) |
771 |
len = 16;
|
772 |
/* XXX: we could compress the chars to 7 bits to increase
|
773 |
entropy */
|
774 |
for(i = 0;i < len;i++) { |
775 |
keybuf[i] = key[i]; |
776 |
} |
777 |
s->crypt_method = s->crypt_method_header; |
778 |
|
779 |
if (AES_set_encrypt_key(keybuf, 128, &s->aes_encrypt_key) != 0) |
780 |
return -1; |
781 |
if (AES_set_decrypt_key(keybuf, 128, &s->aes_decrypt_key) != 0) |
782 |
return -1; |
783 |
#if 0
|
784 |
/* test */
|
785 |
{
|
786 |
uint8_t in[16];
|
787 |
uint8_t out[16];
|
788 |
uint8_t tmp[16];
|
789 |
for(i=0;i<16;i++)
|
790 |
in[i] = i;
|
791 |
AES_encrypt(in, tmp, &s->aes_encrypt_key);
|
792 |
AES_decrypt(tmp, out, &s->aes_decrypt_key);
|
793 |
for(i = 0; i < 16; i++)
|
794 |
printf(" %02x", tmp[i]);
|
795 |
printf("\n");
|
796 |
for(i = 0; i < 16; i++)
|
797 |
printf(" %02x", out[i]);
|
798 |
printf("\n");
|
799 |
}
|
800 |
#endif
|
801 |
return 0; |
802 |
} |
803 |
|
804 |
/* We have nothing to do for QCOW2 reopen, stubs just return
|
805 |
* success */
|
806 |
static int qcow2_reopen_prepare(BDRVReopenState *state, |
807 |
BlockReopenQueue *queue, Error **errp) |
808 |
{ |
809 |
return 0; |
810 |
} |
811 |
|
812 |
static int64_t coroutine_fn qcow2_co_get_block_status(BlockDriverState *bs,
|
813 |
int64_t sector_num, int nb_sectors, int *pnum) |
814 |
{ |
815 |
BDRVQcowState *s = bs->opaque; |
816 |
uint64_t cluster_offset; |
817 |
int index_in_cluster, ret;
|
818 |
int64_t status = 0;
|
819 |
|
820 |
*pnum = nb_sectors; |
821 |
qemu_co_mutex_lock(&s->lock); |
822 |
ret = qcow2_get_cluster_offset(bs, sector_num << 9, pnum, &cluster_offset);
|
823 |
qemu_co_mutex_unlock(&s->lock); |
824 |
if (ret < 0) { |
825 |
return ret;
|
826 |
} |
827 |
|
828 |
if (cluster_offset != 0 && ret != QCOW2_CLUSTER_COMPRESSED && |
829 |
!s->crypt_method) { |
830 |
index_in_cluster = sector_num & (s->cluster_sectors - 1);
|
831 |
cluster_offset |= (index_in_cluster << BDRV_SECTOR_BITS); |
832 |
status |= BDRV_BLOCK_OFFSET_VALID | cluster_offset; |
833 |
} |
834 |
if (ret == QCOW2_CLUSTER_ZERO) {
|
835 |
status |= BDRV_BLOCK_ZERO; |
836 |
} else if (ret != QCOW2_CLUSTER_UNALLOCATED) { |
837 |
status |= BDRV_BLOCK_DATA; |
838 |
} |
839 |
return status;
|
840 |
} |
841 |
|
842 |
/* handle reading after the end of the backing file */
|
843 |
int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov,
|
844 |
int64_t sector_num, int nb_sectors)
|
845 |
{ |
846 |
int n1;
|
847 |
if ((sector_num + nb_sectors) <= bs->total_sectors)
|
848 |
return nb_sectors;
|
849 |
if (sector_num >= bs->total_sectors)
|
850 |
n1 = 0;
|
851 |
else
|
852 |
n1 = bs->total_sectors - sector_num; |
853 |
|
854 |
qemu_iovec_memset(qiov, 512 * n1, 0, 512 * (nb_sectors - n1)); |
855 |
|
856 |
return n1;
|
857 |
} |
858 |
|
859 |
static coroutine_fn int qcow2_co_readv(BlockDriverState *bs, int64_t sector_num, |
860 |
int remaining_sectors, QEMUIOVector *qiov)
|
861 |
{ |
862 |
BDRVQcowState *s = bs->opaque; |
863 |
int index_in_cluster, n1;
|
864 |
int ret;
|
865 |
int cur_nr_sectors; /* number of sectors in current iteration */ |
866 |
uint64_t cluster_offset = 0;
|
867 |
uint64_t bytes_done = 0;
|
868 |
QEMUIOVector hd_qiov; |
869 |
uint8_t *cluster_data = NULL;
|
870 |
|
871 |
qemu_iovec_init(&hd_qiov, qiov->niov); |
872 |
|
873 |
qemu_co_mutex_lock(&s->lock); |
874 |
|
875 |
while (remaining_sectors != 0) { |
876 |
|
877 |
/* prepare next request */
|
878 |
cur_nr_sectors = remaining_sectors; |
879 |
if (s->crypt_method) {
|
880 |
cur_nr_sectors = MIN(cur_nr_sectors, |
881 |
QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); |
882 |
} |
883 |
|
884 |
ret = qcow2_get_cluster_offset(bs, sector_num << 9,
|
885 |
&cur_nr_sectors, &cluster_offset); |
886 |
if (ret < 0) { |
887 |
goto fail;
|
888 |
} |
889 |
|
890 |
index_in_cluster = sector_num & (s->cluster_sectors - 1);
|
891 |
|
892 |
qemu_iovec_reset(&hd_qiov); |
893 |
qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
894 |
cur_nr_sectors * 512);
|
895 |
|
896 |
switch (ret) {
|
897 |
case QCOW2_CLUSTER_UNALLOCATED:
|
898 |
|
899 |
if (bs->backing_hd) {
|
900 |
/* read from the base image */
|
901 |
n1 = qcow2_backing_read1(bs->backing_hd, &hd_qiov, |
902 |
sector_num, cur_nr_sectors); |
903 |
if (n1 > 0) { |
904 |
BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO); |
905 |
qemu_co_mutex_unlock(&s->lock); |
906 |
ret = bdrv_co_readv(bs->backing_hd, sector_num, |
907 |
n1, &hd_qiov); |
908 |
qemu_co_mutex_lock(&s->lock); |
909 |
if (ret < 0) { |
910 |
goto fail;
|
911 |
} |
912 |
} |
913 |
} else {
|
914 |
/* Note: in this case, no need to wait */
|
915 |
qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
916 |
} |
917 |
break;
|
918 |
|
919 |
case QCOW2_CLUSTER_ZERO:
|
920 |
qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
921 |
break;
|
922 |
|
923 |
case QCOW2_CLUSTER_COMPRESSED:
|
924 |
/* add AIO support for compressed blocks ? */
|
925 |
ret = qcow2_decompress_cluster(bs, cluster_offset); |
926 |
if (ret < 0) { |
927 |
goto fail;
|
928 |
} |
929 |
|
930 |
qemu_iovec_from_buf(&hd_qiov, 0,
|
931 |
s->cluster_cache + index_in_cluster * 512,
|
932 |
512 * cur_nr_sectors);
|
933 |
break;
|
934 |
|
935 |
case QCOW2_CLUSTER_NORMAL:
|
936 |
if ((cluster_offset & 511) != 0) { |
937 |
ret = -EIO; |
938 |
goto fail;
|
939 |
} |
940 |
|
941 |
if (s->crypt_method) {
|
942 |
/*
|
943 |
* For encrypted images, read everything into a temporary
|
944 |
* contiguous buffer on which the AES functions can work.
|
945 |
*/
|
946 |
if (!cluster_data) {
|
947 |
cluster_data = |
948 |
qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
949 |
} |
950 |
|
951 |
assert(cur_nr_sectors <= |
952 |
QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); |
953 |
qemu_iovec_reset(&hd_qiov); |
954 |
qemu_iovec_add(&hd_qiov, cluster_data, |
955 |
512 * cur_nr_sectors);
|
956 |
} |
957 |
|
958 |
BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); |
959 |
qemu_co_mutex_unlock(&s->lock); |
960 |
ret = bdrv_co_readv(bs->file, |
961 |
(cluster_offset >> 9) + index_in_cluster,
|
962 |
cur_nr_sectors, &hd_qiov); |
963 |
qemu_co_mutex_lock(&s->lock); |
964 |
if (ret < 0) { |
965 |
goto fail;
|
966 |
} |
967 |
if (s->crypt_method) {
|
968 |
qcow2_encrypt_sectors(s, sector_num, cluster_data, |
969 |
cluster_data, cur_nr_sectors, 0, &s->aes_decrypt_key);
|
970 |
qemu_iovec_from_buf(qiov, bytes_done, |
971 |
cluster_data, 512 * cur_nr_sectors);
|
972 |
} |
973 |
break;
|
974 |
|
975 |
default:
|
976 |
g_assert_not_reached(); |
977 |
ret = -EIO; |
978 |
goto fail;
|
979 |
} |
980 |
|
981 |
remaining_sectors -= cur_nr_sectors; |
982 |
sector_num += cur_nr_sectors; |
983 |
bytes_done += cur_nr_sectors * 512;
|
984 |
} |
985 |
ret = 0;
|
986 |
|
987 |
fail:
|
988 |
qemu_co_mutex_unlock(&s->lock); |
989 |
|
990 |
qemu_iovec_destroy(&hd_qiov); |
991 |
qemu_vfree(cluster_data); |
992 |
|
993 |
return ret;
|
994 |
} |
995 |
|
996 |
static coroutine_fn int qcow2_co_writev(BlockDriverState *bs, |
997 |
int64_t sector_num, |
998 |
int remaining_sectors,
|
999 |
QEMUIOVector *qiov) |
1000 |
{ |
1001 |
BDRVQcowState *s = bs->opaque; |
1002 |
int index_in_cluster;
|
1003 |
int ret;
|
1004 |
int cur_nr_sectors; /* number of sectors in current iteration */ |
1005 |
uint64_t cluster_offset; |
1006 |
QEMUIOVector hd_qiov; |
1007 |
uint64_t bytes_done = 0;
|
1008 |
uint8_t *cluster_data = NULL;
|
1009 |
QCowL2Meta *l2meta = NULL;
|
1010 |
|
1011 |
trace_qcow2_writev_start_req(qemu_coroutine_self(), sector_num, |
1012 |
remaining_sectors); |
1013 |
|
1014 |
qemu_iovec_init(&hd_qiov, qiov->niov); |
1015 |
|
1016 |
s->cluster_cache_offset = -1; /* disable compressed cache */ |
1017 |
|
1018 |
qemu_co_mutex_lock(&s->lock); |
1019 |
|
1020 |
while (remaining_sectors != 0) { |
1021 |
|
1022 |
l2meta = NULL;
|
1023 |
|
1024 |
trace_qcow2_writev_start_part(qemu_coroutine_self()); |
1025 |
index_in_cluster = sector_num & (s->cluster_sectors - 1);
|
1026 |
cur_nr_sectors = remaining_sectors; |
1027 |
if (s->crypt_method &&
|
1028 |
cur_nr_sectors > |
1029 |
QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors - index_in_cluster) { |
1030 |
cur_nr_sectors = |
1031 |
QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors - index_in_cluster; |
1032 |
} |
1033 |
|
1034 |
ret = qcow2_alloc_cluster_offset(bs, sector_num << 9,
|
1035 |
&cur_nr_sectors, &cluster_offset, &l2meta); |
1036 |
if (ret < 0) { |
1037 |
goto fail;
|
1038 |
} |
1039 |
|
1040 |
assert((cluster_offset & 511) == 0); |
1041 |
|
1042 |
qemu_iovec_reset(&hd_qiov); |
1043 |
qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
1044 |
cur_nr_sectors * 512);
|
1045 |
|
1046 |
if (s->crypt_method) {
|
1047 |
if (!cluster_data) {
|
1048 |
cluster_data = qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * |
1049 |
s->cluster_size); |
1050 |
} |
1051 |
|
1052 |
assert(hd_qiov.size <= |
1053 |
QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
1054 |
qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size);
|
1055 |
|
1056 |
qcow2_encrypt_sectors(s, sector_num, cluster_data, |
1057 |
cluster_data, cur_nr_sectors, 1, &s->aes_encrypt_key);
|
1058 |
|
1059 |
qemu_iovec_reset(&hd_qiov); |
1060 |
qemu_iovec_add(&hd_qiov, cluster_data, |
1061 |
cur_nr_sectors * 512);
|
1062 |
} |
1063 |
|
1064 |
ret = qcow2_pre_write_overlap_check(bs, 0,
|
1065 |
cluster_offset + index_in_cluster * BDRV_SECTOR_SIZE, |
1066 |
cur_nr_sectors * BDRV_SECTOR_SIZE); |
1067 |
if (ret < 0) { |
1068 |
goto fail;
|
1069 |
} |
1070 |
|
1071 |
qemu_co_mutex_unlock(&s->lock); |
1072 |
BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); |
1073 |
trace_qcow2_writev_data(qemu_coroutine_self(), |
1074 |
(cluster_offset >> 9) + index_in_cluster);
|
1075 |
ret = bdrv_co_writev(bs->file, |
1076 |
(cluster_offset >> 9) + index_in_cluster,
|
1077 |
cur_nr_sectors, &hd_qiov); |
1078 |
qemu_co_mutex_lock(&s->lock); |
1079 |
if (ret < 0) { |
1080 |
goto fail;
|
1081 |
} |
1082 |
|
1083 |
while (l2meta != NULL) { |
1084 |
QCowL2Meta *next; |
1085 |
|
1086 |
ret = qcow2_alloc_cluster_link_l2(bs, l2meta); |
1087 |
if (ret < 0) { |
1088 |
goto fail;
|
1089 |
} |
1090 |
|
1091 |
/* Take the request off the list of running requests */
|
1092 |
if (l2meta->nb_clusters != 0) { |
1093 |
QLIST_REMOVE(l2meta, next_in_flight); |
1094 |
} |
1095 |
|
1096 |
qemu_co_queue_restart_all(&l2meta->dependent_requests); |
1097 |
|
1098 |
next = l2meta->next; |
1099 |
g_free(l2meta); |
1100 |
l2meta = next; |
1101 |
} |
1102 |
|
1103 |
remaining_sectors -= cur_nr_sectors; |
1104 |
sector_num += cur_nr_sectors; |
1105 |
bytes_done += cur_nr_sectors * 512;
|
1106 |
trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_nr_sectors); |
1107 |
} |
1108 |
ret = 0;
|
1109 |
|
1110 |
fail:
|
1111 |
qemu_co_mutex_unlock(&s->lock); |
1112 |
|
1113 |
while (l2meta != NULL) { |
1114 |
QCowL2Meta *next; |
1115 |
|
1116 |
if (l2meta->nb_clusters != 0) { |
1117 |
QLIST_REMOVE(l2meta, next_in_flight); |
1118 |
} |
1119 |
qemu_co_queue_restart_all(&l2meta->dependent_requests); |
1120 |
|
1121 |
next = l2meta->next; |
1122 |
g_free(l2meta); |
1123 |
l2meta = next; |
1124 |
} |
1125 |
|
1126 |
qemu_iovec_destroy(&hd_qiov); |
1127 |
qemu_vfree(cluster_data); |
1128 |
trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); |
1129 |
|
1130 |
return ret;
|
1131 |
} |
1132 |
|
1133 |
static void qcow2_close(BlockDriverState *bs) |
1134 |
{ |
1135 |
BDRVQcowState *s = bs->opaque; |
1136 |
g_free(s->l1_table); |
1137 |
/* else pre-write overlap checks in cache_destroy may crash */
|
1138 |
s->l1_table = NULL;
|
1139 |
|
1140 |
qcow2_cache_flush(bs, s->l2_table_cache); |
1141 |
qcow2_cache_flush(bs, s->refcount_block_cache); |
1142 |
|
1143 |
qcow2_mark_clean(bs); |
1144 |
|
1145 |
qcow2_cache_destroy(bs, s->l2_table_cache); |
1146 |
qcow2_cache_destroy(bs, s->refcount_block_cache); |
1147 |
|
1148 |
g_free(s->unknown_header_fields); |
1149 |
cleanup_unknown_header_ext(bs); |
1150 |
|
1151 |
g_free(s->cluster_cache); |
1152 |
qemu_vfree(s->cluster_data); |
1153 |
qcow2_refcount_close(bs); |
1154 |
qcow2_free_snapshots(bs); |
1155 |
} |
1156 |
|
1157 |
static void qcow2_invalidate_cache(BlockDriverState *bs) |
1158 |
{ |
1159 |
BDRVQcowState *s = bs->opaque; |
1160 |
int flags = s->flags;
|
1161 |
AES_KEY aes_encrypt_key; |
1162 |
AES_KEY aes_decrypt_key; |
1163 |
uint32_t crypt_method = 0;
|
1164 |
QDict *options; |
1165 |
|
1166 |
/*
|
1167 |
* Backing files are read-only which makes all of their metadata immutable,
|
1168 |
* that means we don't have to worry about reopening them here.
|
1169 |
*/
|
1170 |
|
1171 |
if (s->crypt_method) {
|
1172 |
crypt_method = s->crypt_method; |
1173 |
memcpy(&aes_encrypt_key, &s->aes_encrypt_key, sizeof(aes_encrypt_key));
|
1174 |
memcpy(&aes_decrypt_key, &s->aes_decrypt_key, sizeof(aes_decrypt_key));
|
1175 |
} |
1176 |
|
1177 |
qcow2_close(bs); |
1178 |
|
1179 |
options = qdict_new(); |
1180 |
qdict_put(options, QCOW2_OPT_LAZY_REFCOUNTS, |
1181 |
qbool_from_int(s->use_lazy_refcounts)); |
1182 |
|
1183 |
memset(s, 0, sizeof(BDRVQcowState)); |
1184 |
qcow2_open(bs, options, flags, NULL);
|
1185 |
|
1186 |
QDECREF(options); |
1187 |
|
1188 |
if (crypt_method) {
|
1189 |
s->crypt_method = crypt_method; |
1190 |
memcpy(&s->aes_encrypt_key, &aes_encrypt_key, sizeof(aes_encrypt_key));
|
1191 |
memcpy(&s->aes_decrypt_key, &aes_decrypt_key, sizeof(aes_decrypt_key));
|
1192 |
} |
1193 |
} |
1194 |
|
1195 |
static size_t header_ext_add(char *buf, uint32_t magic, const void *s, |
1196 |
size_t len, size_t buflen) |
1197 |
{ |
1198 |
QCowExtension *ext_backing_fmt = (QCowExtension*) buf; |
1199 |
size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7); |
1200 |
|
1201 |
if (buflen < ext_len) {
|
1202 |
return -ENOSPC;
|
1203 |
} |
1204 |
|
1205 |
*ext_backing_fmt = (QCowExtension) { |
1206 |
.magic = cpu_to_be32(magic), |
1207 |
.len = cpu_to_be32(len), |
1208 |
}; |
1209 |
memcpy(buf + sizeof(QCowExtension), s, len);
|
1210 |
|
1211 |
return ext_len;
|
1212 |
} |
1213 |
|
1214 |
/*
|
1215 |
* Updates the qcow2 header, including the variable length parts of it, i.e.
|
1216 |
* the backing file name and all extensions. qcow2 was not designed to allow
|
1217 |
* such changes, so if we run out of space (we can only use the first cluster)
|
1218 |
* this function may fail.
|
1219 |
*
|
1220 |
* Returns 0 on success, -errno in error cases.
|
1221 |
*/
|
1222 |
int qcow2_update_header(BlockDriverState *bs)
|
1223 |
{ |
1224 |
BDRVQcowState *s = bs->opaque; |
1225 |
QCowHeader *header; |
1226 |
char *buf;
|
1227 |
size_t buflen = s->cluster_size; |
1228 |
int ret;
|
1229 |
uint64_t total_size; |
1230 |
uint32_t refcount_table_clusters; |
1231 |
size_t header_length; |
1232 |
Qcow2UnknownHeaderExtension *uext; |
1233 |
|
1234 |
buf = qemu_blockalign(bs, buflen); |
1235 |
|
1236 |
/* Header structure */
|
1237 |
header = (QCowHeader*) buf; |
1238 |
|
1239 |
if (buflen < sizeof(*header)) { |
1240 |
ret = -ENOSPC; |
1241 |
goto fail;
|
1242 |
} |
1243 |
|
1244 |
header_length = sizeof(*header) + s->unknown_header_fields_size;
|
1245 |
total_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
1246 |
refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3);
|
1247 |
|
1248 |
*header = (QCowHeader) { |
1249 |
/* Version 2 fields */
|
1250 |
.magic = cpu_to_be32(QCOW_MAGIC), |
1251 |
.version = cpu_to_be32(s->qcow_version), |
1252 |
.backing_file_offset = 0,
|
1253 |
.backing_file_size = 0,
|
1254 |
.cluster_bits = cpu_to_be32(s->cluster_bits), |
1255 |
.size = cpu_to_be64(total_size), |
1256 |
.crypt_method = cpu_to_be32(s->crypt_method_header), |
1257 |
.l1_size = cpu_to_be32(s->l1_size), |
1258 |
.l1_table_offset = cpu_to_be64(s->l1_table_offset), |
1259 |
.refcount_table_offset = cpu_to_be64(s->refcount_table_offset), |
1260 |
.refcount_table_clusters = cpu_to_be32(refcount_table_clusters), |
1261 |
.nb_snapshots = cpu_to_be32(s->nb_snapshots), |
1262 |
.snapshots_offset = cpu_to_be64(s->snapshots_offset), |
1263 |
|
1264 |
/* Version 3 fields */
|
1265 |
.incompatible_features = cpu_to_be64(s->incompatible_features), |
1266 |
.compatible_features = cpu_to_be64(s->compatible_features), |
1267 |
.autoclear_features = cpu_to_be64(s->autoclear_features), |
1268 |
.refcount_order = cpu_to_be32(s->refcount_order), |
1269 |
.header_length = cpu_to_be32(header_length), |
1270 |
}; |
1271 |
|
1272 |
/* For older versions, write a shorter header */
|
1273 |
switch (s->qcow_version) {
|
1274 |
case 2: |
1275 |
ret = offsetof(QCowHeader, incompatible_features); |
1276 |
break;
|
1277 |
case 3: |
1278 |
ret = sizeof(*header);
|
1279 |
break;
|
1280 |
default:
|
1281 |
ret = -EINVAL; |
1282 |
goto fail;
|
1283 |
} |
1284 |
|
1285 |
buf += ret; |
1286 |
buflen -= ret; |
1287 |
memset(buf, 0, buflen);
|
1288 |
|
1289 |
/* Preserve any unknown field in the header */
|
1290 |
if (s->unknown_header_fields_size) {
|
1291 |
if (buflen < s->unknown_header_fields_size) {
|
1292 |
ret = -ENOSPC; |
1293 |
goto fail;
|
1294 |
} |
1295 |
|
1296 |
memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size); |
1297 |
buf += s->unknown_header_fields_size; |
1298 |
buflen -= s->unknown_header_fields_size; |
1299 |
} |
1300 |
|
1301 |
/* Backing file format header extension */
|
1302 |
if (*bs->backing_format) {
|
1303 |
ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT, |
1304 |
bs->backing_format, strlen(bs->backing_format), |
1305 |
buflen); |
1306 |
if (ret < 0) { |
1307 |
goto fail;
|
1308 |
} |
1309 |
|
1310 |
buf += ret; |
1311 |
buflen -= ret; |
1312 |
} |
1313 |
|
1314 |
/* Feature table */
|
1315 |
Qcow2Feature features[] = { |
1316 |
{ |
1317 |
.type = QCOW2_FEAT_TYPE_INCOMPATIBLE, |
1318 |
.bit = QCOW2_INCOMPAT_DIRTY_BITNR, |
1319 |
.name = "dirty bit",
|
1320 |
}, |
1321 |
{ |
1322 |
.type = QCOW2_FEAT_TYPE_INCOMPATIBLE, |
1323 |
.bit = QCOW2_INCOMPAT_CORRUPT_BITNR, |
1324 |
.name = "corrupt bit",
|
1325 |
}, |
1326 |
{ |
1327 |
.type = QCOW2_FEAT_TYPE_COMPATIBLE, |
1328 |
.bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, |
1329 |
.name = "lazy refcounts",
|
1330 |
}, |
1331 |
}; |
1332 |
|
1333 |
ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE, |
1334 |
features, sizeof(features), buflen);
|
1335 |
if (ret < 0) { |
1336 |
goto fail;
|
1337 |
} |
1338 |
buf += ret; |
1339 |
buflen -= ret; |
1340 |
|
1341 |
/* Keep unknown header extensions */
|
1342 |
QLIST_FOREACH(uext, &s->unknown_header_ext, next) { |
1343 |
ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen); |
1344 |
if (ret < 0) { |
1345 |
goto fail;
|
1346 |
} |
1347 |
|
1348 |
buf += ret; |
1349 |
buflen -= ret; |
1350 |
} |
1351 |
|
1352 |
/* End of header extensions */
|
1353 |
ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen); |
1354 |
if (ret < 0) { |
1355 |
goto fail;
|
1356 |
} |
1357 |
|
1358 |
buf += ret; |
1359 |
buflen -= ret; |
1360 |
|
1361 |
/* Backing file name */
|
1362 |
if (*bs->backing_file) {
|
1363 |
size_t backing_file_len = strlen(bs->backing_file); |
1364 |
|
1365 |
if (buflen < backing_file_len) {
|
1366 |
ret = -ENOSPC; |
1367 |
goto fail;
|
1368 |
} |
1369 |
|
1370 |
/* Using strncpy is ok here, since buf is not NUL-terminated. */
|
1371 |
strncpy(buf, bs->backing_file, buflen); |
1372 |
|
1373 |
header->backing_file_offset = cpu_to_be64(buf - ((char*) header));
|
1374 |
header->backing_file_size = cpu_to_be32(backing_file_len); |
1375 |
} |
1376 |
|
1377 |
/* Write the new header */
|
1378 |
ret = bdrv_pwrite(bs->file, 0, header, s->cluster_size);
|
1379 |
if (ret < 0) { |
1380 |
goto fail;
|
1381 |
} |
1382 |
|
1383 |
ret = 0;
|
1384 |
fail:
|
1385 |
qemu_vfree(header); |
1386 |
return ret;
|
1387 |
} |
1388 |
|
1389 |
static int qcow2_change_backing_file(BlockDriverState *bs, |
1390 |
const char *backing_file, const char *backing_fmt) |
1391 |
{ |
1392 |
pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); |
1393 |
pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); |
1394 |
|
1395 |
return qcow2_update_header(bs);
|
1396 |
} |
1397 |
|
1398 |
static int preallocate(BlockDriverState *bs) |
1399 |
{ |
1400 |
uint64_t nb_sectors; |
1401 |
uint64_t offset; |
1402 |
uint64_t host_offset = 0;
|
1403 |
int num;
|
1404 |
int ret;
|
1405 |
QCowL2Meta *meta; |
1406 |
|
1407 |
nb_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
1408 |
offset = 0;
|
1409 |
|
1410 |
while (nb_sectors) {
|
1411 |
num = MIN(nb_sectors, INT_MAX >> BDRV_SECTOR_BITS); |
1412 |
ret = qcow2_alloc_cluster_offset(bs, offset, &num, |
1413 |
&host_offset, &meta); |
1414 |
if (ret < 0) { |
1415 |
return ret;
|
1416 |
} |
1417 |
|
1418 |
if (meta != NULL) { |
1419 |
ret = qcow2_alloc_cluster_link_l2(bs, meta); |
1420 |
if (ret < 0) { |
1421 |
qcow2_free_any_clusters(bs, meta->alloc_offset, |
1422 |
meta->nb_clusters, QCOW2_DISCARD_NEVER); |
1423 |
return ret;
|
1424 |
} |
1425 |
|
1426 |
/* There are no dependent requests, but we need to remove our
|
1427 |
* request from the list of in-flight requests */
|
1428 |
QLIST_REMOVE(meta, next_in_flight); |
1429 |
} |
1430 |
|
1431 |
/* TODO Preallocate data if requested */
|
1432 |
|
1433 |
nb_sectors -= num; |
1434 |
offset += num << BDRV_SECTOR_BITS; |
1435 |
} |
1436 |
|
1437 |
/*
|
1438 |
* It is expected that the image file is large enough to actually contain
|
1439 |
* all of the allocated clusters (otherwise we get failing reads after
|
1440 |
* EOF). Extend the image to the last allocated sector.
|
1441 |
*/
|
1442 |
if (host_offset != 0) { |
1443 |
uint8_t buf[BDRV_SECTOR_SIZE]; |
1444 |
memset(buf, 0, BDRV_SECTOR_SIZE);
|
1445 |
ret = bdrv_write(bs->file, (host_offset >> BDRV_SECTOR_BITS) + num - 1,
|
1446 |
buf, 1);
|
1447 |
if (ret < 0) { |
1448 |
return ret;
|
1449 |
} |
1450 |
} |
1451 |
|
1452 |
return 0; |
1453 |
} |
1454 |
|
1455 |
static int qcow2_create2(const char *filename, int64_t total_size, |
1456 |
const char *backing_file, const char *backing_format, |
1457 |
int flags, size_t cluster_size, int prealloc, |
1458 |
QEMUOptionParameter *options, int version,
|
1459 |
Error **errp) |
1460 |
{ |
1461 |
/* Calculate cluster_bits */
|
1462 |
int cluster_bits;
|
1463 |
cluster_bits = ffs(cluster_size) - 1;
|
1464 |
if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS ||
|
1465 |
(1 << cluster_bits) != cluster_size)
|
1466 |
{ |
1467 |
error_setg(errp, "Cluster size must be a power of two between %d and "
|
1468 |
"%dk", 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10)); |
1469 |
return -EINVAL;
|
1470 |
} |
1471 |
|
1472 |
/*
|
1473 |
* Open the image file and write a minimal qcow2 header.
|
1474 |
*
|
1475 |
* We keep things simple and start with a zero-sized image. We also
|
1476 |
* do without refcount blocks or a L1 table for now. We'll fix the
|
1477 |
* inconsistency later.
|
1478 |
*
|
1479 |
* We do need a refcount table because growing the refcount table means
|
1480 |
* allocating two new refcount blocks - the seconds of which would be at
|
1481 |
* 2 GB for 64k clusters, and we don't want to have a 2 GB initial file
|
1482 |
* size for any qcow2 image.
|
1483 |
*/
|
1484 |
BlockDriverState* bs; |
1485 |
QCowHeader *header; |
1486 |
uint8_t* refcount_table; |
1487 |
Error *local_err = NULL;
|
1488 |
int ret;
|
1489 |
|
1490 |
ret = bdrv_create_file(filename, options, &local_err); |
1491 |
if (ret < 0) { |
1492 |
error_propagate(errp, local_err); |
1493 |
return ret;
|
1494 |
} |
1495 |
|
1496 |
ret = bdrv_file_open(&bs, filename, NULL, NULL, BDRV_O_RDWR, &local_err); |
1497 |
if (ret < 0) { |
1498 |
error_propagate(errp, local_err); |
1499 |
return ret;
|
1500 |
} |
1501 |
|
1502 |
/* Write the header */
|
1503 |
QEMU_BUILD_BUG_ON((1 << MIN_CLUSTER_BITS) < sizeof(*header)); |
1504 |
header = g_malloc0(cluster_size); |
1505 |
*header = (QCowHeader) { |
1506 |
.magic = cpu_to_be32(QCOW_MAGIC), |
1507 |
.version = cpu_to_be32(version), |
1508 |
.cluster_bits = cpu_to_be32(cluster_bits), |
1509 |
.size = cpu_to_be64(0),
|
1510 |
.l1_table_offset = cpu_to_be64(0),
|
1511 |
.l1_size = cpu_to_be32(0),
|
1512 |
.refcount_table_offset = cpu_to_be64(cluster_size), |
1513 |
.refcount_table_clusters = cpu_to_be32(1),
|
1514 |
.refcount_order = cpu_to_be32(3 + REFCOUNT_SHIFT),
|
1515 |
.header_length = cpu_to_be32(sizeof(*header)),
|
1516 |
}; |
1517 |
|
1518 |
if (flags & BLOCK_FLAG_ENCRYPT) {
|
1519 |
header->crypt_method = cpu_to_be32(QCOW_CRYPT_AES); |
1520 |
} else {
|
1521 |
header->crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); |
1522 |
} |
1523 |
|
1524 |
if (flags & BLOCK_FLAG_LAZY_REFCOUNTS) {
|
1525 |
header->compatible_features |= |
1526 |
cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS); |
1527 |
} |
1528 |
|
1529 |
ret = bdrv_pwrite(bs, 0, header, cluster_size);
|
1530 |
g_free(header); |
1531 |
if (ret < 0) { |
1532 |
error_setg_errno(errp, -ret, "Could not write qcow2 header");
|
1533 |
goto out;
|
1534 |
} |
1535 |
|
1536 |
/* Write an empty refcount table */
|
1537 |
refcount_table = g_malloc0(cluster_size); |
1538 |
ret = bdrv_pwrite(bs, cluster_size, refcount_table, cluster_size); |
1539 |
g_free(refcount_table); |
1540 |
|
1541 |
if (ret < 0) { |
1542 |
error_setg_errno(errp, -ret, "Could not write refcount table");
|
1543 |
goto out;
|
1544 |
} |
1545 |
|
1546 |
bdrv_unref(bs); |
1547 |
bs = NULL;
|
1548 |
|
1549 |
/*
|
1550 |
* And now open the image and make it consistent first (i.e. increase the
|
1551 |
* refcount of the cluster that is occupied by the header and the refcount
|
1552 |
* table)
|
1553 |
*/
|
1554 |
BlockDriver* drv = bdrv_find_format("qcow2");
|
1555 |
assert(drv != NULL);
|
1556 |
ret = bdrv_open(&bs, filename, NULL,
|
1557 |
BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_FLUSH, drv, &local_err); |
1558 |
if (ret < 0) { |
1559 |
error_propagate(errp, local_err); |
1560 |
goto out;
|
1561 |
} |
1562 |
|
1563 |
ret = qcow2_alloc_clusters(bs, 2 * cluster_size);
|
1564 |
if (ret < 0) { |
1565 |
error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 "
|
1566 |
"header and refcount table");
|
1567 |
goto out;
|
1568 |
|
1569 |
} else if (ret != 0) { |
1570 |
error_report("Huh, first cluster in empty image is already in use?");
|
1571 |
abort(); |
1572 |
} |
1573 |
|
1574 |
/* Okay, now that we have a valid image, let's give it the right size */
|
1575 |
ret = bdrv_truncate(bs, total_size * BDRV_SECTOR_SIZE); |
1576 |
if (ret < 0) { |
1577 |
error_setg_errno(errp, -ret, "Could not resize image");
|
1578 |
goto out;
|
1579 |
} |
1580 |
|
1581 |
/* Want a backing file? There you go.*/
|
1582 |
if (backing_file) {
|
1583 |
ret = bdrv_change_backing_file(bs, backing_file, backing_format); |
1584 |
if (ret < 0) { |
1585 |
error_setg_errno(errp, -ret, "Could not assign backing file '%s' "
|
1586 |
"with format '%s'", backing_file, backing_format);
|
1587 |
goto out;
|
1588 |
} |
1589 |
} |
1590 |
|
1591 |
/* And if we're supposed to preallocate metadata, do that now */
|
1592 |
if (prealloc) {
|
1593 |
BDRVQcowState *s = bs->opaque; |
1594 |
qemu_co_mutex_lock(&s->lock); |
1595 |
ret = preallocate(bs); |
1596 |
qemu_co_mutex_unlock(&s->lock); |
1597 |
if (ret < 0) { |
1598 |
error_setg_errno(errp, -ret, "Could not preallocate metadata");
|
1599 |
goto out;
|
1600 |
} |
1601 |
} |
1602 |
|
1603 |
bdrv_unref(bs); |
1604 |
bs = NULL;
|
1605 |
|
1606 |
/* Reopen the image without BDRV_O_NO_FLUSH to flush it before returning */
|
1607 |
ret = bdrv_open(&bs, filename, NULL,
|
1608 |
BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_BACKING, |
1609 |
drv, &local_err); |
1610 |
if (local_err) {
|
1611 |
error_propagate(errp, local_err); |
1612 |
goto out;
|
1613 |
} |
1614 |
|
1615 |
ret = 0;
|
1616 |
out:
|
1617 |
if (bs) {
|
1618 |
bdrv_unref(bs); |
1619 |
} |
1620 |
return ret;
|
1621 |
} |
1622 |
|
1623 |
static int qcow2_create(const char *filename, QEMUOptionParameter *options, |
1624 |
Error **errp) |
1625 |
{ |
1626 |
const char *backing_file = NULL; |
1627 |
const char *backing_fmt = NULL; |
1628 |
uint64_t sectors = 0;
|
1629 |
int flags = 0; |
1630 |
size_t cluster_size = DEFAULT_CLUSTER_SIZE; |
1631 |
int prealloc = 0; |
1632 |
int version = 3; |
1633 |
Error *local_err = NULL;
|
1634 |
int ret;
|
1635 |
|
1636 |
/* Read out options */
|
1637 |
while (options && options->name) {
|
1638 |
if (!strcmp(options->name, BLOCK_OPT_SIZE)) {
|
1639 |
sectors = options->value.n / 512;
|
1640 |
} else if (!strcmp(options->name, BLOCK_OPT_BACKING_FILE)) { |
1641 |
backing_file = options->value.s; |
1642 |
} else if (!strcmp(options->name, BLOCK_OPT_BACKING_FMT)) { |
1643 |
backing_fmt = options->value.s; |
1644 |
} else if (!strcmp(options->name, BLOCK_OPT_ENCRYPT)) { |
1645 |
flags |= options->value.n ? BLOCK_FLAG_ENCRYPT : 0;
|
1646 |
} else if (!strcmp(options->name, BLOCK_OPT_CLUSTER_SIZE)) { |
1647 |
if (options->value.n) {
|
1648 |
cluster_size = options->value.n; |
1649 |
} |
1650 |
} else if (!strcmp(options->name, BLOCK_OPT_PREALLOC)) { |
1651 |
if (!options->value.s || !strcmp(options->value.s, "off")) { |
1652 |
prealloc = 0;
|
1653 |
} else if (!strcmp(options->value.s, "metadata")) { |
1654 |
prealloc = 1;
|
1655 |
} else {
|
1656 |
error_setg(errp, "Invalid preallocation mode: '%s'",
|
1657 |
options->value.s); |
1658 |
return -EINVAL;
|
1659 |
} |
1660 |
} else if (!strcmp(options->name, BLOCK_OPT_COMPAT_LEVEL)) { |
1661 |
if (!options->value.s) {
|
1662 |
/* keep the default */
|
1663 |
} else if (!strcmp(options->value.s, "0.10")) { |
1664 |
version = 2;
|
1665 |
} else if (!strcmp(options->value.s, "1.1")) { |
1666 |
version = 3;
|
1667 |
} else {
|
1668 |
error_setg(errp, "Invalid compatibility level: '%s'",
|
1669 |
options->value.s); |
1670 |
return -EINVAL;
|
1671 |
} |
1672 |
} else if (!strcmp(options->name, BLOCK_OPT_LAZY_REFCOUNTS)) { |
1673 |
flags |= options->value.n ? BLOCK_FLAG_LAZY_REFCOUNTS : 0;
|
1674 |
} |
1675 |
options++; |
1676 |
} |
1677 |
|
1678 |
if (backing_file && prealloc) {
|
1679 |
error_setg(errp, "Backing file and preallocation cannot be used at "
|
1680 |
"the same time");
|
1681 |
return -EINVAL;
|
1682 |
} |
1683 |
|
1684 |
if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) { |
1685 |
error_setg(errp, "Lazy refcounts only supported with compatibility "
|
1686 |
"level 1.1 and above (use compat=1.1 or greater)");
|
1687 |
return -EINVAL;
|
1688 |
} |
1689 |
|
1690 |
ret = qcow2_create2(filename, sectors, backing_file, backing_fmt, flags, |
1691 |
cluster_size, prealloc, options, version, &local_err); |
1692 |
if (local_err) {
|
1693 |
error_propagate(errp, local_err); |
1694 |
} |
1695 |
return ret;
|
1696 |
} |
1697 |
|
1698 |
static coroutine_fn int qcow2_co_write_zeroes(BlockDriverState *bs, |
1699 |
int64_t sector_num, int nb_sectors, BdrvRequestFlags flags)
|
1700 |
{ |
1701 |
int ret;
|
1702 |
BDRVQcowState *s = bs->opaque; |
1703 |
|
1704 |
/* Emulate misaligned zero writes */
|
1705 |
if (sector_num % s->cluster_sectors || nb_sectors % s->cluster_sectors) {
|
1706 |
return -ENOTSUP;
|
1707 |
} |
1708 |
|
1709 |
/* Whatever is left can use real zero clusters */
|
1710 |
qemu_co_mutex_lock(&s->lock); |
1711 |
ret = qcow2_zero_clusters(bs, sector_num << BDRV_SECTOR_BITS, |
1712 |
nb_sectors); |
1713 |
qemu_co_mutex_unlock(&s->lock); |
1714 |
|
1715 |
return ret;
|
1716 |
} |
1717 |
|
1718 |
static coroutine_fn int qcow2_co_discard(BlockDriverState *bs, |
1719 |
int64_t sector_num, int nb_sectors)
|
1720 |
{ |
1721 |
int ret;
|
1722 |
BDRVQcowState *s = bs->opaque; |
1723 |
|
1724 |
qemu_co_mutex_lock(&s->lock); |
1725 |
ret = qcow2_discard_clusters(bs, sector_num << BDRV_SECTOR_BITS, |
1726 |
nb_sectors, QCOW2_DISCARD_REQUEST); |
1727 |
qemu_co_mutex_unlock(&s->lock); |
1728 |
return ret;
|
1729 |
} |
1730 |
|
1731 |
static int qcow2_truncate(BlockDriverState *bs, int64_t offset) |
1732 |
{ |
1733 |
BDRVQcowState *s = bs->opaque; |
1734 |
int64_t new_l1_size; |
1735 |
int ret;
|
1736 |
|
1737 |
if (offset & 511) { |
1738 |
error_report("The new size must be a multiple of 512");
|
1739 |
return -EINVAL;
|
1740 |
} |
1741 |
|
1742 |
/* cannot proceed if image has snapshots */
|
1743 |
if (s->nb_snapshots) {
|
1744 |
error_report("Can't resize an image which has snapshots");
|
1745 |
return -ENOTSUP;
|
1746 |
} |
1747 |
|
1748 |
/* shrinking is currently not supported */
|
1749 |
if (offset < bs->total_sectors * 512) { |
1750 |
error_report("qcow2 doesn't support shrinking images yet");
|
1751 |
return -ENOTSUP;
|
1752 |
} |
1753 |
|
1754 |
new_l1_size = size_to_l1(s, offset); |
1755 |
ret = qcow2_grow_l1_table(bs, new_l1_size, true);
|
1756 |
if (ret < 0) { |
1757 |
return ret;
|
1758 |
} |
1759 |
|
1760 |
/* write updated header.size */
|
1761 |
offset = cpu_to_be64(offset); |
1762 |
ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, size), |
1763 |
&offset, sizeof(uint64_t));
|
1764 |
if (ret < 0) { |
1765 |
return ret;
|
1766 |
} |
1767 |
|
1768 |
s->l1_vm_state_index = new_l1_size; |
1769 |
return 0; |
1770 |
} |
1771 |
|
1772 |
/* XXX: put compressed sectors first, then all the cluster aligned
|
1773 |
tables to avoid losing bytes in alignment */
|
1774 |
static int qcow2_write_compressed(BlockDriverState *bs, int64_t sector_num, |
1775 |
const uint8_t *buf, int nb_sectors) |
1776 |
{ |
1777 |
BDRVQcowState *s = bs->opaque; |
1778 |
z_stream strm; |
1779 |
int ret, out_len;
|
1780 |
uint8_t *out_buf; |
1781 |
uint64_t cluster_offset; |
1782 |
|
1783 |
if (nb_sectors == 0) { |
1784 |
/* align end of file to a sector boundary to ease reading with
|
1785 |
sector based I/Os */
|
1786 |
cluster_offset = bdrv_getlength(bs->file); |
1787 |
cluster_offset = (cluster_offset + 511) & ~511; |
1788 |
bdrv_truncate(bs->file, cluster_offset); |
1789 |
return 0; |
1790 |
} |
1791 |
|
1792 |
if (nb_sectors != s->cluster_sectors) {
|
1793 |
ret = -EINVAL; |
1794 |
|
1795 |
/* Zero-pad last write if image size is not cluster aligned */
|
1796 |
if (sector_num + nb_sectors == bs->total_sectors &&
|
1797 |
nb_sectors < s->cluster_sectors) { |
1798 |
uint8_t *pad_buf = qemu_blockalign(bs, s->cluster_size); |
1799 |
memset(pad_buf, 0, s->cluster_size);
|
1800 |
memcpy(pad_buf, buf, nb_sectors * BDRV_SECTOR_SIZE); |
1801 |
ret = qcow2_write_compressed(bs, sector_num, |
1802 |
pad_buf, s->cluster_sectors); |
1803 |
qemu_vfree(pad_buf); |
1804 |
} |
1805 |
return ret;
|
1806 |
} |
1807 |
|
1808 |
out_buf = g_malloc(s->cluster_size + (s->cluster_size / 1000) + 128); |
1809 |
|
1810 |
/* best compression, small window, no zlib header */
|
1811 |
memset(&strm, 0, sizeof(strm)); |
1812 |
ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, |
1813 |
Z_DEFLATED, -12,
|
1814 |
9, Z_DEFAULT_STRATEGY);
|
1815 |
if (ret != 0) { |
1816 |
ret = -EINVAL; |
1817 |
goto fail;
|
1818 |
} |
1819 |
|
1820 |
strm.avail_in = s->cluster_size; |
1821 |
strm.next_in = (uint8_t *)buf; |
1822 |
strm.avail_out = s->cluster_size; |
1823 |
strm.next_out = out_buf; |
1824 |
|
1825 |
ret = deflate(&strm, Z_FINISH); |
1826 |
if (ret != Z_STREAM_END && ret != Z_OK) {
|
1827 |
deflateEnd(&strm); |
1828 |
ret = -EINVAL; |
1829 |
goto fail;
|
1830 |
} |
1831 |
out_len = strm.next_out - out_buf; |
1832 |
|
1833 |
deflateEnd(&strm); |
1834 |
|
1835 |
if (ret != Z_STREAM_END || out_len >= s->cluster_size) {
|
1836 |
/* could not compress: write normal cluster */
|
1837 |
ret = bdrv_write(bs, sector_num, buf, s->cluster_sectors); |
1838 |
if (ret < 0) { |
1839 |
goto fail;
|
1840 |
} |
1841 |
} else {
|
1842 |
cluster_offset = qcow2_alloc_compressed_cluster_offset(bs, |
1843 |
sector_num << 9, out_len);
|
1844 |
if (!cluster_offset) {
|
1845 |
ret = -EIO; |
1846 |
goto fail;
|
1847 |
} |
1848 |
cluster_offset &= s->cluster_offset_mask; |
1849 |
|
1850 |
ret = qcow2_pre_write_overlap_check(bs, 0, cluster_offset, out_len);
|
1851 |
if (ret < 0) { |
1852 |
goto fail;
|
1853 |
} |
1854 |
|
1855 |
BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED); |
1856 |
ret = bdrv_pwrite(bs->file, cluster_offset, out_buf, out_len); |
1857 |
if (ret < 0) { |
1858 |
goto fail;
|
1859 |
} |
1860 |
} |
1861 |
|
1862 |
ret = 0;
|
1863 |
fail:
|
1864 |
g_free(out_buf); |
1865 |
return ret;
|
1866 |
} |
1867 |
|
1868 |
static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs) |
1869 |
{ |
1870 |
BDRVQcowState *s = bs->opaque; |
1871 |
int ret;
|
1872 |
|
1873 |
qemu_co_mutex_lock(&s->lock); |
1874 |
ret = qcow2_cache_flush(bs, s->l2_table_cache); |
1875 |
if (ret < 0) { |
1876 |
qemu_co_mutex_unlock(&s->lock); |
1877 |
return ret;
|
1878 |
} |
1879 |
|
1880 |
if (qcow2_need_accurate_refcounts(s)) {
|
1881 |
ret = qcow2_cache_flush(bs, s->refcount_block_cache); |
1882 |
if (ret < 0) { |
1883 |
qemu_co_mutex_unlock(&s->lock); |
1884 |
return ret;
|
1885 |
} |
1886 |
} |
1887 |
qemu_co_mutex_unlock(&s->lock); |
1888 |
|
1889 |
return 0; |
1890 |
} |
1891 |
|
1892 |
static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
1893 |
{ |
1894 |
BDRVQcowState *s = bs->opaque; |
1895 |
bdi->unallocated_blocks_are_zero = true;
|
1896 |
bdi->can_write_zeroes_with_unmap = (s->qcow_version >= 3);
|
1897 |
bdi->cluster_size = s->cluster_size; |
1898 |
bdi->vm_state_offset = qcow2_vm_state_offset(s); |
1899 |
return 0; |
1900 |
} |
1901 |
|
1902 |
static ImageInfoSpecific *qcow2_get_specific_info(BlockDriverState *bs)
|
1903 |
{ |
1904 |
BDRVQcowState *s = bs->opaque; |
1905 |
ImageInfoSpecific *spec_info = g_new(ImageInfoSpecific, 1);
|
1906 |
|
1907 |
*spec_info = (ImageInfoSpecific){ |
1908 |
.kind = IMAGE_INFO_SPECIFIC_KIND_QCOW2, |
1909 |
{ |
1910 |
.qcow2 = g_new(ImageInfoSpecificQCow2, 1),
|
1911 |
}, |
1912 |
}; |
1913 |
if (s->qcow_version == 2) { |
1914 |
*spec_info->qcow2 = (ImageInfoSpecificQCow2){ |
1915 |
.compat = g_strdup("0.10"),
|
1916 |
}; |
1917 |
} else if (s->qcow_version == 3) { |
1918 |
*spec_info->qcow2 = (ImageInfoSpecificQCow2){ |
1919 |
.compat = g_strdup("1.1"),
|
1920 |
.lazy_refcounts = s->compatible_features & |
1921 |
QCOW2_COMPAT_LAZY_REFCOUNTS, |
1922 |
.has_lazy_refcounts = true,
|
1923 |
}; |
1924 |
} |
1925 |
|
1926 |
return spec_info;
|
1927 |
} |
1928 |
|
1929 |
#if 0
|
1930 |
static void dump_refcounts(BlockDriverState *bs)
|
1931 |
{
|
1932 |
BDRVQcowState *s = bs->opaque;
|
1933 |
int64_t nb_clusters, k, k1, size;
|
1934 |
int refcount;
|
1935 |
|
1936 |
size = bdrv_getlength(bs->file);
|
1937 |
nb_clusters = size_to_clusters(s, size);
|
1938 |
for(k = 0; k < nb_clusters;) {
|
1939 |
k1 = k;
|
1940 |
refcount = get_refcount(bs, k);
|
1941 |
k++;
|
1942 |
while (k < nb_clusters && get_refcount(bs, k) == refcount)
|
1943 |
k++;
|
1944 |
printf("%" PRId64 ": refcount=%d nb=%" PRId64 "\n", k, refcount,
|
1945 |
k - k1);
|
1946 |
}
|
1947 |
}
|
1948 |
#endif
|
1949 |
|
1950 |
static int qcow2_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
1951 |
int64_t pos) |
1952 |
{ |
1953 |
BDRVQcowState *s = bs->opaque; |
1954 |
int64_t total_sectors = bs->total_sectors; |
1955 |
int growable = bs->growable;
|
1956 |
bool zero_beyond_eof = bs->zero_beyond_eof;
|
1957 |
int ret;
|
1958 |
|
1959 |
BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE); |
1960 |
bs->growable = 1;
|
1961 |
bs->zero_beyond_eof = false;
|
1962 |
ret = bdrv_pwritev(bs, qcow2_vm_state_offset(s) + pos, qiov); |
1963 |
bs->growable = growable; |
1964 |
bs->zero_beyond_eof = zero_beyond_eof; |
1965 |
|
1966 |
/* bdrv_co_do_writev will have increased the total_sectors value to include
|
1967 |
* the VM state - the VM state is however not an actual part of the block
|
1968 |
* device, therefore, we need to restore the old value. */
|
1969 |
bs->total_sectors = total_sectors; |
1970 |
|
1971 |
return ret;
|
1972 |
} |
1973 |
|
1974 |
static int qcow2_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
1975 |
int64_t pos, int size)
|
1976 |
{ |
1977 |
BDRVQcowState *s = bs->opaque; |
1978 |
int growable = bs->growable;
|
1979 |
bool zero_beyond_eof = bs->zero_beyond_eof;
|
1980 |
int ret;
|
1981 |
|
1982 |
BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD); |
1983 |
bs->growable = 1;
|
1984 |
bs->zero_beyond_eof = false;
|
1985 |
ret = bdrv_pread(bs, qcow2_vm_state_offset(s) + pos, buf, size); |
1986 |
bs->growable = growable; |
1987 |
bs->zero_beyond_eof = zero_beyond_eof; |
1988 |
|
1989 |
return ret;
|
1990 |
} |
1991 |
|
1992 |
/*
|
1993 |
* Downgrades an image's version. To achieve this, any incompatible features
|
1994 |
* have to be removed.
|
1995 |
*/
|
1996 |
static int qcow2_downgrade(BlockDriverState *bs, int target_version) |
1997 |
{ |
1998 |
BDRVQcowState *s = bs->opaque; |
1999 |
int current_version = s->qcow_version;
|
2000 |
int ret;
|
2001 |
|
2002 |
if (target_version == current_version) {
|
2003 |
return 0; |
2004 |
} else if (target_version > current_version) { |
2005 |
return -EINVAL;
|
2006 |
} else if (target_version != 2) { |
2007 |
return -EINVAL;
|
2008 |
} |
2009 |
|
2010 |
if (s->refcount_order != 4) { |
2011 |
/* we would have to convert the image to a refcount_order == 4 image
|
2012 |
* here; however, since qemu (at the time of writing this) does not
|
2013 |
* support anything different than 4 anyway, there is no point in doing
|
2014 |
* so right now; however, we should error out (if qemu supports this in
|
2015 |
* the future and this code has not been adapted) */
|
2016 |
error_report("qcow2_downgrade: Image refcount orders other than 4 are "
|
2017 |
"currently not supported.");
|
2018 |
return -ENOTSUP;
|
2019 |
} |
2020 |
|
2021 |
/* clear incompatible features */
|
2022 |
if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) {
|
2023 |
ret = qcow2_mark_clean(bs); |
2024 |
if (ret < 0) { |
2025 |
return ret;
|
2026 |
} |
2027 |
} |
2028 |
|
2029 |
/* with QCOW2_INCOMPAT_CORRUPT, it is pretty much impossible to get here in
|
2030 |
* the first place; if that happens nonetheless, returning -ENOTSUP is the
|
2031 |
* best thing to do anyway */
|
2032 |
|
2033 |
if (s->incompatible_features) {
|
2034 |
return -ENOTSUP;
|
2035 |
} |
2036 |
|
2037 |
/* since we can ignore compatible features, we can set them to 0 as well */
|
2038 |
s->compatible_features = 0;
|
2039 |
/* if lazy refcounts have been used, they have already been fixed through
|
2040 |
* clearing the dirty flag */
|
2041 |
|
2042 |
/* clearing autoclear features is trivial */
|
2043 |
s->autoclear_features = 0;
|
2044 |
|
2045 |
ret = qcow2_expand_zero_clusters(bs); |
2046 |
if (ret < 0) { |
2047 |
return ret;
|
2048 |
} |
2049 |
|
2050 |
s->qcow_version = target_version; |
2051 |
ret = qcow2_update_header(bs); |
2052 |
if (ret < 0) { |
2053 |
s->qcow_version = current_version; |
2054 |
return ret;
|
2055 |
} |
2056 |
return 0; |
2057 |
} |
2058 |
|
2059 |
static int qcow2_amend_options(BlockDriverState *bs, |
2060 |
QEMUOptionParameter *options) |
2061 |
{ |
2062 |
BDRVQcowState *s = bs->opaque; |
2063 |
int old_version = s->qcow_version, new_version = old_version;
|
2064 |
uint64_t new_size = 0;
|
2065 |
const char *backing_file = NULL, *backing_format = NULL; |
2066 |
bool lazy_refcounts = s->use_lazy_refcounts;
|
2067 |
int ret;
|
2068 |
int i;
|
2069 |
|
2070 |
for (i = 0; options[i].name; i++) |
2071 |
{ |
2072 |
if (!options[i].assigned) {
|
2073 |
/* only change explicitly defined options */
|
2074 |
continue;
|
2075 |
} |
2076 |
|
2077 |
if (!strcmp(options[i].name, "compat")) { |
2078 |
if (!options[i].value.s) {
|
2079 |
/* preserve default */
|
2080 |
} else if (!strcmp(options[i].value.s, "0.10")) { |
2081 |
new_version = 2;
|
2082 |
} else if (!strcmp(options[i].value.s, "1.1")) { |
2083 |
new_version = 3;
|
2084 |
} else {
|
2085 |
fprintf(stderr, "Unknown compatibility level %s.\n",
|
2086 |
options[i].value.s); |
2087 |
return -EINVAL;
|
2088 |
} |
2089 |
} else if (!strcmp(options[i].name, "preallocation")) { |
2090 |
fprintf(stderr, "Cannot change preallocation mode.\n");
|
2091 |
return -ENOTSUP;
|
2092 |
} else if (!strcmp(options[i].name, "size")) { |
2093 |
new_size = options[i].value.n; |
2094 |
} else if (!strcmp(options[i].name, "backing_file")) { |
2095 |
backing_file = options[i].value.s; |
2096 |
} else if (!strcmp(options[i].name, "backing_fmt")) { |
2097 |
backing_format = options[i].value.s; |
2098 |
} else if (!strcmp(options[i].name, "encryption")) { |
2099 |
if ((options[i].value.n != !!s->crypt_method)) {
|
2100 |
fprintf(stderr, "Changing the encryption flag is not "
|
2101 |
"supported.\n");
|
2102 |
return -ENOTSUP;
|
2103 |
} |
2104 |
} else if (!strcmp(options[i].name, "cluster_size")) { |
2105 |
if (options[i].value.n != s->cluster_size) {
|
2106 |
fprintf(stderr, "Changing the cluster size is not "
|
2107 |
"supported.\n");
|
2108 |
return -ENOTSUP;
|
2109 |
} |
2110 |
} else if (!strcmp(options[i].name, "lazy_refcounts")) { |
2111 |
lazy_refcounts = options[i].value.n; |
2112 |
} else {
|
2113 |
/* if this assertion fails, this probably means a new option was
|
2114 |
* added without having it covered here */
|
2115 |
assert(false);
|
2116 |
} |
2117 |
} |
2118 |
|
2119 |
if (new_version != old_version) {
|
2120 |
if (new_version > old_version) {
|
2121 |
/* Upgrade */
|
2122 |
s->qcow_version = new_version; |
2123 |
ret = qcow2_update_header(bs); |
2124 |
if (ret < 0) { |
2125 |
s->qcow_version = old_version; |
2126 |
return ret;
|
2127 |
} |
2128 |
} else {
|
2129 |
ret = qcow2_downgrade(bs, new_version); |
2130 |
if (ret < 0) { |
2131 |
return ret;
|
2132 |
} |
2133 |
} |
2134 |
} |
2135 |
|
2136 |
if (backing_file || backing_format) {
|
2137 |
ret = qcow2_change_backing_file(bs, backing_file ?: bs->backing_file, |
2138 |
backing_format ?: bs->backing_format); |
2139 |
if (ret < 0) { |
2140 |
return ret;
|
2141 |
} |
2142 |
} |
2143 |
|
2144 |
if (s->use_lazy_refcounts != lazy_refcounts) {
|
2145 |
if (lazy_refcounts) {
|
2146 |
if (s->qcow_version < 3) { |
2147 |
fprintf(stderr, "Lazy refcounts only supported with compatibility "
|
2148 |
"level 1.1 and above (use compat=1.1 or greater)\n");
|
2149 |
return -EINVAL;
|
2150 |
} |
2151 |
s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; |
2152 |
ret = qcow2_update_header(bs); |
2153 |
if (ret < 0) { |
2154 |
s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; |
2155 |
return ret;
|
2156 |
} |
2157 |
s->use_lazy_refcounts = true;
|
2158 |
} else {
|
2159 |
/* make image clean first */
|
2160 |
ret = qcow2_mark_clean(bs); |
2161 |
if (ret < 0) { |
2162 |
return ret;
|
2163 |
} |
2164 |
/* now disallow lazy refcounts */
|
2165 |
s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; |
2166 |
ret = qcow2_update_header(bs); |
2167 |
if (ret < 0) { |
2168 |
s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; |
2169 |
return ret;
|
2170 |
} |
2171 |
s->use_lazy_refcounts = false;
|
2172 |
} |
2173 |
} |
2174 |
|
2175 |
if (new_size) {
|
2176 |
ret = bdrv_truncate(bs, new_size); |
2177 |
if (ret < 0) { |
2178 |
return ret;
|
2179 |
} |
2180 |
} |
2181 |
|
2182 |
return 0; |
2183 |
} |
2184 |
|
2185 |
static QEMUOptionParameter qcow2_create_options[] = {
|
2186 |
{ |
2187 |
.name = BLOCK_OPT_SIZE, |
2188 |
.type = OPT_SIZE, |
2189 |
.help = "Virtual disk size"
|
2190 |
}, |
2191 |
{ |
2192 |
.name = BLOCK_OPT_COMPAT_LEVEL, |
2193 |
.type = OPT_STRING, |
2194 |
.help = "Compatibility level (0.10 or 1.1)"
|
2195 |
}, |
2196 |
{ |
2197 |
.name = BLOCK_OPT_BACKING_FILE, |
2198 |
.type = OPT_STRING, |
2199 |
.help = "File name of a base image"
|
2200 |
}, |
2201 |
{ |
2202 |
.name = BLOCK_OPT_BACKING_FMT, |
2203 |
.type = OPT_STRING, |
2204 |
.help = "Image format of the base image"
|
2205 |
}, |
2206 |
{ |
2207 |
.name = BLOCK_OPT_ENCRYPT, |
2208 |
.type = OPT_FLAG, |
2209 |
.help = "Encrypt the image"
|
2210 |
}, |
2211 |
{ |
2212 |
.name = BLOCK_OPT_CLUSTER_SIZE, |
2213 |
.type = OPT_SIZE, |
2214 |
.help = "qcow2 cluster size",
|
2215 |
.value = { .n = DEFAULT_CLUSTER_SIZE }, |
2216 |
}, |
2217 |
{ |
2218 |
.name = BLOCK_OPT_PREALLOC, |
2219 |
.type = OPT_STRING, |
2220 |
.help = "Preallocation mode (allowed values: off, metadata)"
|
2221 |
}, |
2222 |
{ |
2223 |
.name = BLOCK_OPT_LAZY_REFCOUNTS, |
2224 |
.type = OPT_FLAG, |
2225 |
.help = "Postpone refcount updates",
|
2226 |
}, |
2227 |
{ NULL }
|
2228 |
}; |
2229 |
|
2230 |
static BlockDriver bdrv_qcow2 = {
|
2231 |
.format_name = "qcow2",
|
2232 |
.instance_size = sizeof(BDRVQcowState),
|
2233 |
.bdrv_probe = qcow2_probe, |
2234 |
.bdrv_open = qcow2_open, |
2235 |
.bdrv_close = qcow2_close, |
2236 |
.bdrv_reopen_prepare = qcow2_reopen_prepare, |
2237 |
.bdrv_create = qcow2_create, |
2238 |
.bdrv_has_zero_init = bdrv_has_zero_init_1, |
2239 |
.bdrv_co_get_block_status = qcow2_co_get_block_status, |
2240 |
.bdrv_set_key = qcow2_set_key, |
2241 |
|
2242 |
.bdrv_co_readv = qcow2_co_readv, |
2243 |
.bdrv_co_writev = qcow2_co_writev, |
2244 |
.bdrv_co_flush_to_os = qcow2_co_flush_to_os, |
2245 |
|
2246 |
.bdrv_co_write_zeroes = qcow2_co_write_zeroes, |
2247 |
.bdrv_co_discard = qcow2_co_discard, |
2248 |
.bdrv_truncate = qcow2_truncate, |
2249 |
.bdrv_write_compressed = qcow2_write_compressed, |
2250 |
|
2251 |
.bdrv_snapshot_create = qcow2_snapshot_create, |
2252 |
.bdrv_snapshot_goto = qcow2_snapshot_goto, |
2253 |
.bdrv_snapshot_delete = qcow2_snapshot_delete, |
2254 |
.bdrv_snapshot_list = qcow2_snapshot_list, |
2255 |
.bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp, |
2256 |
.bdrv_get_info = qcow2_get_info, |
2257 |
.bdrv_get_specific_info = qcow2_get_specific_info, |
2258 |
|
2259 |
.bdrv_save_vmstate = qcow2_save_vmstate, |
2260 |
.bdrv_load_vmstate = qcow2_load_vmstate, |
2261 |
|
2262 |
.bdrv_change_backing_file = qcow2_change_backing_file, |
2263 |
|
2264 |
.bdrv_refresh_limits = qcow2_refresh_limits, |
2265 |
.bdrv_invalidate_cache = qcow2_invalidate_cache, |
2266 |
|
2267 |
.create_options = qcow2_create_options, |
2268 |
.bdrv_check = qcow2_check, |
2269 |
.bdrv_amend_options = qcow2_amend_options, |
2270 |
}; |
2271 |
|
2272 |
static void bdrv_qcow2_init(void) |
2273 |
{ |
2274 |
bdrv_register(&bdrv_qcow2); |
2275 |
} |
2276 |
|
2277 |
block_init(bdrv_qcow2_init); |