blob: 548f8a057b87feb9a45cb56e9bfb25711ef38cf6 [file] [log] [blame]
Stefan Weil9aebd982009-07-31 21:45:56 +02001/*
2 * Block driver for the Virtual Disk Image (VDI) format
3 *
Stefan Weil641543b2012-01-21 13:54:24 +01004 * Copyright (c) 2009, 2012 Stefan Weil
Stefan Weil9aebd982009-07-31 21:45:56 +02005 *
6 * This program is free software: you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation, either version 2 of the License, or
9 * (at your option) version 3 or any later version.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program. If not, see <http://www.gnu.org/licenses/>.
18 *
19 * Reference:
20 * http://forums.virtualbox.org/viewtopic.php?t=8046
21 *
22 * This driver supports create / read / write operations on VDI images.
23 *
24 * Todo (see also TODO in code):
25 *
26 * Some features like snapshots are still missing.
27 *
28 * Deallocation of zero-filled blocks and shrinking images are missing, too
29 * (might be added to common block layer).
30 *
31 * Allocation of blocks could be optimized (less writes to block map and
32 * header).
33 *
Deepak Kathayatdc6fb732014-03-24 16:30:17 +080034 * Read and write of adjacent blocks could be done in one operation
Stefan Weil9aebd982009-07-31 21:45:56 +020035 * (current code uses one operation per block (1 MiB).
36 *
37 * The code is not thread safe (missing locks for changes in header and
38 * block table, no problem with current QEMU).
39 *
40 * Hints:
41 *
42 * Blocks (VDI documentation) correspond to clusters (QEMU).
43 * QEMU's backing files could be implemented using VDI snapshot files (TODO).
44 * VDI snapshot files may also contain the complete machine state.
45 * Maybe this machine state can be converted to QEMU PC machine snapshot data.
46 *
47 * The driver keeps a block cache (little endian entries) in memory.
48 * For the standard block size (1 MiB), a 1 TiB disk will use 4 MiB RAM,
49 * so this seems to be reasonable.
50 */
51
Peter Maydell80c71a22016-01-18 18:01:42 +000052#include "qemu/osdep.h"
Philippe Mathieu-Daudéf0435682018-06-25 09:41:54 -030053#include "qemu/units.h"
Markus Armbrusterda34e652016-03-14 09:01:28 +010054#include "qapi/error.h"
Max Reitz49858b52018-03-12 17:55:26 +010055#include "qapi/qobject-input-visitor.h"
56#include "qapi/qapi-visit-block-core.h"
Paolo Bonzini737e1502012-12-17 18:19:44 +010057#include "block/block_int.h"
Markus Armbrusterf8534652018-06-14 21:14:34 +020058#include "block/qdict.h"
Kevin Wolfa08f0c32016-03-08 15:57:05 +010059#include "sysemu/block-backend.h"
Paolo Bonzini1de7afc2012-12-17 18:20:00 +010060#include "qemu/module.h"
Markus Armbruster922a01a2018-02-01 12:18:46 +010061#include "qemu/option.h"
Paolo Bonzini58369e22016-03-15 17:22:36 +010062#include "qemu/bswap.h"
Juan Quintela795c40b2017-04-06 12:00:28 +020063#include "migration/blocker.h"
Daniel P. Berrange10817bf2015-09-01 14:48:02 +010064#include "qemu/coroutine.h"
Veronia Bahaaf348b6d2016-03-20 19:16:19 +020065#include "qemu/cutils.h"
Fam Zheng7c6f55b2016-09-21 12:27:17 +080066#include "qemu/uuid.h"
Stefan Weil9aebd982009-07-31 21:45:56 +020067
68/* Code configuration options. */
69
70/* Enable debug messages. */
71//~ #define CONFIG_VDI_DEBUG
72
73/* Support write operations on VDI images. */
74#define CONFIG_VDI_WRITE
75
76/* Support non-standard block (cluster) size. This is untested.
77 * Maybe it will be needed for very large images.
78 */
79//~ #define CONFIG_VDI_BLOCK_SIZE
80
81/* Support static (fixed, pre-allocated) images. */
82#define CONFIG_VDI_STATIC_IMAGE
83
84/* Command line option for static images. */
85#define BLOCK_OPT_STATIC "static"
86
Stefan Weil9aebd982009-07-31 21:45:56 +020087#define SECTOR_SIZE 512
Markus Armbruster14632122019-01-11 20:14:01 +010088#define DEFAULT_CLUSTER_SIZE 1048576
89/* Note: can't use 1 * MiB, because it's passed to stringify() */
Stefan Weil9aebd982009-07-31 21:45:56 +020090
91#if defined(CONFIG_VDI_DEBUG)
Eric Blakeb80666b2018-02-13 14:26:56 -060092#define VDI_DEBUG 1
Stefan Weil9aebd982009-07-31 21:45:56 +020093#else
Eric Blakeb80666b2018-02-13 14:26:56 -060094#define VDI_DEBUG 0
Stefan Weil9aebd982009-07-31 21:45:56 +020095#endif
96
Eric Blakeb80666b2018-02-13 14:26:56 -060097#define logout(fmt, ...) \
98 do { \
99 if (VDI_DEBUG) { \
100 fprintf(stderr, "vdi\t%-24s" fmt, __func__, ##__VA_ARGS__); \
101 } \
102 } while (0)
103
Stefan Weil9aebd982009-07-31 21:45:56 +0200104/* Image signature. */
105#define VDI_SIGNATURE 0xbeda107f
106
107/* Image version. */
108#define VDI_VERSION_1_1 0x00010001
109
110/* Image type. */
111#define VDI_TYPE_DYNAMIC 1
112#define VDI_TYPE_STATIC 2
113
114/* Innotek / SUN images use these strings in header.text:
115 * "<<< innotek VirtualBox Disk Image >>>\n"
116 * "<<< Sun xVM VirtualBox Disk Image >>>\n"
117 * "<<< Sun VirtualBox Disk Image >>>\n"
118 * The value does not matter, so QEMU created images use a different text.
119 */
120#define VDI_TEXT "<<< QEMU VM Virtual Disk Image >>>\n"
121
Eric Sunshinec794b4e2011-10-26 15:51:18 -0400122/* A never-allocated block; semantically arbitrary content. */
123#define VDI_UNALLOCATED 0xffffffffU
124
125/* A discarded (no longer allocated) block; semantically zero-filled. */
126#define VDI_DISCARDED 0xfffffffeU
127
128#define VDI_IS_ALLOCATED(X) ((X) < VDI_DISCARDED)
Stefan Weil9aebd982009-07-31 21:45:56 +0200129
Max Reitzd20418e2014-10-28 11:12:32 +0100130/* The bmap will take up VDI_BLOCKS_IN_IMAGE_MAX * sizeof(uint32_t) bytes; since
131 * the bmap is read and written in a single operation, its size needs to be
132 * limited to INT_MAX; furthermore, when opening an image, the bmap size is
133 * rounded up to be aligned on BDRV_SECTOR_SIZE.
134 * Therefore this should satisfy the following:
135 * VDI_BLOCKS_IN_IMAGE_MAX * sizeof(uint32_t) + BDRV_SECTOR_SIZE == INT_MAX + 1
136 * (INT_MAX + 1 is the first value not representable as an int)
137 * This guarantees that any value below or equal to the constant will, when
138 * multiplied by sizeof(uint32_t) and rounded up to a BDRV_SECTOR_SIZE boundary,
139 * still be below or equal to INT_MAX. */
140#define VDI_BLOCKS_IN_IMAGE_MAX \
141 ((unsigned)((INT_MAX + 1u - BDRV_SECTOR_SIZE) / sizeof(uint32_t)))
Jeff Cody63fa06d2014-03-28 11:42:24 -0400142#define VDI_DISK_SIZE_MAX ((uint64_t)VDI_BLOCKS_IN_IMAGE_MAX * \
143 (uint64_t)DEFAULT_CLUSTER_SIZE)
144
Max Reitz49858b52018-03-12 17:55:26 +0100145static QemuOptsList vdi_create_opts;
146
Stefan Weil9aebd982009-07-31 21:45:56 +0200147typedef struct {
Stefan Weil9aebd982009-07-31 21:45:56 +0200148 char text[0x40];
149 uint32_t signature;
150 uint32_t version;
151 uint32_t header_size;
152 uint32_t image_type;
153 uint32_t image_flags;
154 char description[256];
155 uint32_t offset_bmap;
156 uint32_t offset_data;
157 uint32_t cylinders; /* disk geometry, unused here */
158 uint32_t heads; /* disk geometry, unused here */
159 uint32_t sectors; /* disk geometry, unused here */
160 uint32_t sector_size;
161 uint32_t unused1;
162 uint64_t disk_size;
163 uint32_t block_size;
164 uint32_t block_extra; /* unused here */
165 uint32_t blocks_in_image;
166 uint32_t blocks_allocated;
Fam Zheng7c6f55b2016-09-21 12:27:17 +0800167 QemuUUID uuid_image;
168 QemuUUID uuid_last_snap;
169 QemuUUID uuid_link;
170 QemuUUID uuid_parent;
Stefan Weil9aebd982009-07-31 21:45:56 +0200171 uint64_t unused2[7];
Jeff Cody8368feb2013-09-25 12:08:48 -0400172} QEMU_PACKED VdiHeader;
Stefan Weil9aebd982009-07-31 21:45:56 +0200173
Alberto Garciad4f18972019-05-01 21:13:56 +0300174QEMU_BUILD_BUG_ON(sizeof(VdiHeader) != 512);
175
Stefan Weil9aebd982009-07-31 21:45:56 +0200176typedef struct {
Stefan Weil9aebd982009-07-31 21:45:56 +0200177 /* The block map entries are little endian (even in memory). */
178 uint32_t *bmap;
179 /* Size of block (bytes). */
180 uint32_t block_size;
Stefan Weil9aebd982009-07-31 21:45:56 +0200181 /* First sector of block map. */
182 uint32_t bmap_sector;
Stefan Weil4ff97862011-03-13 15:44:02 +0100183 /* VDI header (converted to host endianness). */
Stefan Weil9aebd982009-07-31 21:45:56 +0200184 VdiHeader header;
Kevin Wolffc9d1062011-11-22 16:46:26 +0100185
Paolo Bonzini1e886632017-06-29 15:27:41 +0200186 CoRwlock bmap_lock;
Max Reitzf0ab6f12015-02-27 14:54:39 -0500187
Kevin Wolffc9d1062011-11-22 16:46:26 +0100188 Error *migration_blocker;
Stefan Weil9aebd982009-07-31 21:45:56 +0200189} BDRVVdiState;
190
Stefan Weil9aebd982009-07-31 21:45:56 +0200191static void vdi_header_to_cpu(VdiHeader *header)
192{
Peter Maydell09190182018-10-16 18:25:03 +0100193 header->signature = le32_to_cpu(header->signature);
194 header->version = le32_to_cpu(header->version);
195 header->header_size = le32_to_cpu(header->header_size);
196 header->image_type = le32_to_cpu(header->image_type);
197 header->image_flags = le32_to_cpu(header->image_flags);
198 header->offset_bmap = le32_to_cpu(header->offset_bmap);
199 header->offset_data = le32_to_cpu(header->offset_data);
200 header->cylinders = le32_to_cpu(header->cylinders);
201 header->heads = le32_to_cpu(header->heads);
202 header->sectors = le32_to_cpu(header->sectors);
203 header->sector_size = le32_to_cpu(header->sector_size);
204 header->disk_size = le64_to_cpu(header->disk_size);
205 header->block_size = le32_to_cpu(header->block_size);
206 header->block_extra = le32_to_cpu(header->block_extra);
207 header->blocks_in_image = le32_to_cpu(header->blocks_in_image);
208 header->blocks_allocated = le32_to_cpu(header->blocks_allocated);
Peter Maydell1324f062018-12-10 11:26:49 +0000209 header->uuid_image = qemu_uuid_bswap(header->uuid_image);
210 header->uuid_last_snap = qemu_uuid_bswap(header->uuid_last_snap);
211 header->uuid_link = qemu_uuid_bswap(header->uuid_link);
212 header->uuid_parent = qemu_uuid_bswap(header->uuid_parent);
Stefan Weil9aebd982009-07-31 21:45:56 +0200213}
214
215static void vdi_header_to_le(VdiHeader *header)
216{
Peter Maydell09190182018-10-16 18:25:03 +0100217 header->signature = cpu_to_le32(header->signature);
218 header->version = cpu_to_le32(header->version);
219 header->header_size = cpu_to_le32(header->header_size);
220 header->image_type = cpu_to_le32(header->image_type);
221 header->image_flags = cpu_to_le32(header->image_flags);
222 header->offset_bmap = cpu_to_le32(header->offset_bmap);
223 header->offset_data = cpu_to_le32(header->offset_data);
224 header->cylinders = cpu_to_le32(header->cylinders);
225 header->heads = cpu_to_le32(header->heads);
226 header->sectors = cpu_to_le32(header->sectors);
227 header->sector_size = cpu_to_le32(header->sector_size);
228 header->disk_size = cpu_to_le64(header->disk_size);
229 header->block_size = cpu_to_le32(header->block_size);
230 header->block_extra = cpu_to_le32(header->block_extra);
231 header->blocks_in_image = cpu_to_le32(header->blocks_in_image);
232 header->blocks_allocated = cpu_to_le32(header->blocks_allocated);
Peter Maydell1324f062018-12-10 11:26:49 +0000233 header->uuid_image = qemu_uuid_bswap(header->uuid_image);
234 header->uuid_last_snap = qemu_uuid_bswap(header->uuid_last_snap);
235 header->uuid_link = qemu_uuid_bswap(header->uuid_link);
236 header->uuid_parent = qemu_uuid_bswap(header->uuid_parent);
Stefan Weil9aebd982009-07-31 21:45:56 +0200237}
238
Stefan Weil9aebd982009-07-31 21:45:56 +0200239static void vdi_header_print(VdiHeader *header)
240{
Peter Maydellac928b82018-12-10 11:26:48 +0000241 char uuidstr[37];
242 QemuUUID uuid;
Stefan Weil9aebd982009-07-31 21:45:56 +0200243 logout("text %s", header->text);
Stefan Weil9f0470b2013-01-17 21:45:26 +0100244 logout("signature 0x%08x\n", header->signature);
Stefan Weil9aebd982009-07-31 21:45:56 +0200245 logout("header size 0x%04x\n", header->header_size);
246 logout("image type 0x%04x\n", header->image_type);
247 logout("image flags 0x%04x\n", header->image_flags);
248 logout("description %s\n", header->description);
249 logout("offset bmap 0x%04x\n", header->offset_bmap);
250 logout("offset data 0x%04x\n", header->offset_data);
251 logout("cylinders 0x%04x\n", header->cylinders);
252 logout("heads 0x%04x\n", header->heads);
253 logout("sectors 0x%04x\n", header->sectors);
254 logout("sector size 0x%04x\n", header->sector_size);
255 logout("image size 0x%" PRIx64 " B (%" PRIu64 " MiB)\n",
256 header->disk_size, header->disk_size / MiB);
257 logout("block size 0x%04x\n", header->block_size);
258 logout("block extra 0x%04x\n", header->block_extra);
259 logout("blocks tot. 0x%04x\n", header->blocks_in_image);
260 logout("blocks all. 0x%04x\n", header->blocks_allocated);
Peter Maydellac928b82018-12-10 11:26:48 +0000261 uuid = header->uuid_image;
262 qemu_uuid_unparse(&uuid, uuidstr);
263 logout("uuid image %s\n", uuidstr);
264 uuid = header->uuid_last_snap;
265 qemu_uuid_unparse(&uuid, uuidstr);
266 logout("uuid snap %s\n", uuidstr);
267 uuid = header->uuid_link;
268 qemu_uuid_unparse(&uuid, uuidstr);
269 logout("uuid link %s\n", uuidstr);
270 uuid = header->uuid_parent;
271 qemu_uuid_unparse(&uuid, uuidstr);
272 logout("uuid parent %s\n", uuidstr);
Stefan Weil9aebd982009-07-31 21:45:56 +0200273}
Stefan Weil9aebd982009-07-31 21:45:56 +0200274
Paolo Bonzini2fd61632018-03-01 17:36:19 +0100275static int coroutine_fn vdi_co_check(BlockDriverState *bs, BdrvCheckResult *res,
276 BdrvCheckMode fix)
Stefan Weil9aebd982009-07-31 21:45:56 +0200277{
278 /* TODO: additional checks possible. */
279 BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
Stefan Weil9aebd982009-07-31 21:45:56 +0200280 uint32_t blocks_allocated = 0;
281 uint32_t block;
282 uint32_t *bmap;
283 logout("\n");
284
Kevin Wolf4534ff52012-05-11 16:07:02 +0200285 if (fix) {
286 return -ENOTSUP;
287 }
288
Markus Armbruster5839e532014-08-19 10:31:08 +0200289 bmap = g_try_new(uint32_t, s->header.blocks_in_image);
Kevin Wolf17cce732014-05-20 13:25:43 +0200290 if (s->header.blocks_in_image && bmap == NULL) {
291 res->check_errors++;
292 return -ENOMEM;
293 }
294
Stefan Weil9aebd982009-07-31 21:45:56 +0200295 memset(bmap, 0xff, s->header.blocks_in_image * sizeof(uint32_t));
296
297 /* Check block map and value of blocks_allocated. */
298 for (block = 0; block < s->header.blocks_in_image; block++) {
299 uint32_t bmap_entry = le32_to_cpu(s->bmap[block]);
Eric Sunshinec794b4e2011-10-26 15:51:18 -0400300 if (VDI_IS_ALLOCATED(bmap_entry)) {
Stefan Weil9aebd982009-07-31 21:45:56 +0200301 if (bmap_entry < s->header.blocks_in_image) {
302 blocks_allocated++;
Eric Sunshinec794b4e2011-10-26 15:51:18 -0400303 if (!VDI_IS_ALLOCATED(bmap[bmap_entry])) {
Stefan Weil9aebd982009-07-31 21:45:56 +0200304 bmap[bmap_entry] = bmap_entry;
305 } else {
306 fprintf(stderr, "ERROR: block index %" PRIu32
307 " also used by %" PRIu32 "\n", bmap[bmap_entry], bmap_entry);
Kevin Wolf9ac228e2010-06-29 12:37:54 +0200308 res->corruptions++;
Stefan Weil9aebd982009-07-31 21:45:56 +0200309 }
310 } else {
311 fprintf(stderr, "ERROR: block index %" PRIu32
312 " too large, is %" PRIu32 "\n", block, bmap_entry);
Kevin Wolf9ac228e2010-06-29 12:37:54 +0200313 res->corruptions++;
Stefan Weil9aebd982009-07-31 21:45:56 +0200314 }
315 }
316 }
317 if (blocks_allocated != s->header.blocks_allocated) {
318 fprintf(stderr, "ERROR: allocated blocks mismatch, is %" PRIu32
319 ", should be %" PRIu32 "\n",
320 blocks_allocated, s->header.blocks_allocated);
Kevin Wolf9ac228e2010-06-29 12:37:54 +0200321 res->corruptions++;
Stefan Weil9aebd982009-07-31 21:45:56 +0200322 }
323
Anthony Liguori7267c092011-08-20 22:09:37 -0500324 g_free(bmap);
Stefan Weil9aebd982009-07-31 21:45:56 +0200325
Kevin Wolf9ac228e2010-06-29 12:37:54 +0200326 return 0;
Stefan Weil9aebd982009-07-31 21:45:56 +0200327}
328
329static int vdi_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
330{
331 /* TODO: vdi_get_info would be needed for machine snapshots.
332 vm_state_offset is still missing. */
333 BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
334 logout("\n");
335 bdi->cluster_size = s->block_size;
336 bdi->vm_state_offset = 0;
337 return 0;
338}
339
340static int vdi_make_empty(BlockDriverState *bs)
341{
342 /* TODO: missing code. */
343 logout("\n");
344 /* The return value for missing code must be 0, see block.c. */
345 return 0;
346}
347
348static int vdi_probe(const uint8_t *buf, int buf_size, const char *filename)
349{
350 const VdiHeader *header = (const VdiHeader *)buf;
Jeff Codydddc7752014-07-23 17:22:59 -0400351 int ret = 0;
Stefan Weil9aebd982009-07-31 21:45:56 +0200352
353 logout("\n");
354
355 if (buf_size < sizeof(*header)) {
356 /* Header too small, no VDI. */
357 } else if (le32_to_cpu(header->signature) == VDI_SIGNATURE) {
Jeff Codydddc7752014-07-23 17:22:59 -0400358 ret = 100;
Stefan Weil9aebd982009-07-31 21:45:56 +0200359 }
360
Jeff Codydddc7752014-07-23 17:22:59 -0400361 if (ret == 0) {
Stefan Weil9aebd982009-07-31 21:45:56 +0200362 logout("no vdi image\n");
363 } else {
364 logout("%s", header->text);
365 }
366
Jeff Codydddc7752014-07-23 17:22:59 -0400367 return ret;
Stefan Weil9aebd982009-07-31 21:45:56 +0200368}
369
Max Reitz015a1032013-09-05 14:22:29 +0200370static int vdi_open(BlockDriverState *bs, QDict *options, int flags,
371 Error **errp)
Stefan Weil9aebd982009-07-31 21:45:56 +0200372{
373 BDRVVdiState *s = bs->opaque;
374 VdiHeader header;
375 size_t bmap_size;
Stefan Weil8937f822013-01-17 21:45:27 +0100376 int ret;
Peter Maydellac928b82018-12-10 11:26:48 +0000377 QemuUUID uuid_link, uuid_parent;
Stefan Weil9aebd982009-07-31 21:45:56 +0200378
Max Reitz8b1869d2020-05-13 13:05:35 +0200379 bs->file = bdrv_open_child(NULL, options, "file", bs, &child_of_bds,
380 BDRV_CHILD_IMAGE, false, errp);
Kevin Wolf4e4bf5c2016-12-16 18:52:37 +0100381 if (!bs->file) {
382 return -EINVAL;
383 }
384
Stefan Weil9aebd982009-07-31 21:45:56 +0200385 logout("\n");
386
Alberto Garciad4f18972019-05-01 21:13:56 +0300387 ret = bdrv_pread(bs->file, 0, &header, sizeof(header));
Stefan Weil8937f822013-01-17 21:45:27 +0100388 if (ret < 0) {
Stefan Weil9aebd982009-07-31 21:45:56 +0200389 goto fail;
390 }
391
392 vdi_header_to_cpu(&header);
Kevin Wolf95a14d52018-03-20 14:41:53 +0100393 if (VDI_DEBUG) {
394 vdi_header_print(&header);
395 }
Stefan Weil9aebd982009-07-31 21:45:56 +0200396
Jeff Cody63fa06d2014-03-28 11:42:24 -0400397 if (header.disk_size > VDI_DISK_SIZE_MAX) {
398 error_setg(errp, "Unsupported VDI image size (size is 0x%" PRIx64
399 ", max supported is 0x%" PRIx64 ")",
400 header.disk_size, VDI_DISK_SIZE_MAX);
401 ret = -ENOTSUP;
402 goto fail;
403 }
404
Peter Maydellac928b82018-12-10 11:26:48 +0000405 uuid_link = header.uuid_link;
406 uuid_parent = header.uuid_parent;
407
Stefan Weilf21dc3a2010-05-12 20:25:45 +0200408 if (header.disk_size % SECTOR_SIZE != 0) {
409 /* 'VBoxManage convertfromraw' can create images with odd disk sizes.
410 We accept them but round the disk size to the next multiple of
411 SECTOR_SIZE. */
412 logout("odd disk size %" PRIu64 " B, round up\n", header.disk_size);
Max Reitze9082e42014-10-21 10:51:25 +0200413 header.disk_size = ROUND_UP(header.disk_size, SECTOR_SIZE);
Stefan Weilf21dc3a2010-05-12 20:25:45 +0200414 }
415
Stefan Weil0e87ba22013-01-17 21:45:28 +0100416 if (header.signature != VDI_SIGNATURE) {
Max Reitz521b2b52014-04-29 19:03:12 +0200417 error_setg(errp, "Image not in VDI format (bad signature %08" PRIx32
418 ")", header.signature);
Paolo Bonzini76abe402014-02-17 14:44:06 +0100419 ret = -EINVAL;
Stefan Weil0e87ba22013-01-17 21:45:28 +0100420 goto fail;
421 } else if (header.version != VDI_VERSION_1_1) {
Max Reitz521b2b52014-04-29 19:03:12 +0200422 error_setg(errp, "unsupported VDI image (version %" PRIu32 ".%" PRIu32
423 ")", header.version >> 16, header.version & 0xffff);
Stefan Weil8937f822013-01-17 21:45:27 +0100424 ret = -ENOTSUP;
Stefan Weil9aebd982009-07-31 21:45:56 +0200425 goto fail;
426 } else if (header.offset_bmap % SECTOR_SIZE != 0) {
427 /* We only support block maps which start on a sector boundary. */
Paolo Bonzini5b7aa9b2014-02-17 14:44:07 +0100428 error_setg(errp, "unsupported VDI image (unaligned block map offset "
Max Reitz521b2b52014-04-29 19:03:12 +0200429 "0x%" PRIx32 ")", header.offset_bmap);
Stefan Weil8937f822013-01-17 21:45:27 +0100430 ret = -ENOTSUP;
Stefan Weil9aebd982009-07-31 21:45:56 +0200431 goto fail;
432 } else if (header.offset_data % SECTOR_SIZE != 0) {
433 /* We only support data blocks which start on a sector boundary. */
Max Reitz521b2b52014-04-29 19:03:12 +0200434 error_setg(errp, "unsupported VDI image (unaligned data offset 0x%"
435 PRIx32 ")", header.offset_data);
Stefan Weil8937f822013-01-17 21:45:27 +0100436 ret = -ENOTSUP;
Stefan Weil9aebd982009-07-31 21:45:56 +0200437 goto fail;
438 } else if (header.sector_size != SECTOR_SIZE) {
Max Reitz521b2b52014-04-29 19:03:12 +0200439 error_setg(errp, "unsupported VDI image (sector size %" PRIu32
440 " is not %u)", header.sector_size, SECTOR_SIZE);
Stefan Weil8937f822013-01-17 21:45:27 +0100441 ret = -ENOTSUP;
Stefan Weil9aebd982009-07-31 21:45:56 +0200442 goto fail;
Jeff Cody63fa06d2014-03-28 11:42:24 -0400443 } else if (header.block_size != DEFAULT_CLUSTER_SIZE) {
Max Reitz521b2b52014-04-29 19:03:12 +0200444 error_setg(errp, "unsupported VDI image (block size %" PRIu32
Leonid Bloch3dd5b8f2018-11-04 18:09:28 +0000445 " is not %" PRIu32 ")",
Philippe Mathieu-Daudéf0435682018-06-25 09:41:54 -0300446 header.block_size, DEFAULT_CLUSTER_SIZE);
Stefan Weil8937f822013-01-17 21:45:27 +0100447 ret = -ENOTSUP;
Stefan Weil9aebd982009-07-31 21:45:56 +0200448 goto fail;
Stefan Weilf21dc3a2010-05-12 20:25:45 +0200449 } else if (header.disk_size >
450 (uint64_t)header.blocks_in_image * header.block_size) {
Paolo Bonzini5b7aa9b2014-02-17 14:44:07 +0100451 error_setg(errp, "unsupported VDI image (disk size %" PRIu64 ", "
452 "image bitmap has room for %" PRIu64 ")",
453 header.disk_size,
454 (uint64_t)header.blocks_in_image * header.block_size);
Stefan Weil8937f822013-01-17 21:45:27 +0100455 ret = -ENOTSUP;
Stefan Weil9aebd982009-07-31 21:45:56 +0200456 goto fail;
Peter Maydellac928b82018-12-10 11:26:48 +0000457 } else if (!qemu_uuid_is_null(&uuid_link)) {
Paolo Bonzini5b7aa9b2014-02-17 14:44:07 +0100458 error_setg(errp, "unsupported VDI image (non-NULL link UUID)");
Stefan Weil8937f822013-01-17 21:45:27 +0100459 ret = -ENOTSUP;
Stefan Weil9aebd982009-07-31 21:45:56 +0200460 goto fail;
Peter Maydellac928b82018-12-10 11:26:48 +0000461 } else if (!qemu_uuid_is_null(&uuid_parent)) {
Paolo Bonzini5b7aa9b2014-02-17 14:44:07 +0100462 error_setg(errp, "unsupported VDI image (non-NULL parent UUID)");
Stefan Weil8937f822013-01-17 21:45:27 +0100463 ret = -ENOTSUP;
Stefan Weil9aebd982009-07-31 21:45:56 +0200464 goto fail;
Jeff Cody63fa06d2014-03-28 11:42:24 -0400465 } else if (header.blocks_in_image > VDI_BLOCKS_IN_IMAGE_MAX) {
466 error_setg(errp, "unsupported VDI image "
467 "(too many blocks %u, max is %u)",
468 header.blocks_in_image, VDI_BLOCKS_IN_IMAGE_MAX);
469 ret = -ENOTSUP;
470 goto fail;
Stefan Weil9aebd982009-07-31 21:45:56 +0200471 }
472
473 bs->total_sectors = header.disk_size / SECTOR_SIZE;
474
475 s->block_size = header.block_size;
Stefan Weil9aebd982009-07-31 21:45:56 +0200476 s->bmap_sector = header.offset_bmap / SECTOR_SIZE;
477 s->header = header;
478
479 bmap_size = header.blocks_in_image * sizeof(uint32_t);
Max Reitze9082e42014-10-21 10:51:25 +0200480 bmap_size = DIV_ROUND_UP(bmap_size, SECTOR_SIZE);
Kevin Wolf9a4f4c32015-06-16 14:19:22 +0200481 s->bmap = qemu_try_blockalign(bs->file->bs, bmap_size * SECTOR_SIZE);
Kevin Wolf17cce732014-05-20 13:25:43 +0200482 if (s->bmap == NULL) {
483 ret = -ENOMEM;
484 goto fail;
485 }
486
Alberto Garciad4f18972019-05-01 21:13:56 +0300487 ret = bdrv_pread(bs->file, header.offset_bmap, s->bmap,
488 bmap_size * SECTOR_SIZE);
Stefan Weil8937f822013-01-17 21:45:27 +0100489 if (ret < 0) {
Stefan Weil9aebd982009-07-31 21:45:56 +0200490 goto fail_free_bmap;
491 }
492
Kevin Wolffc9d1062011-11-22 16:46:26 +0100493 /* Disable migration when vdi images are used */
Alberto Garcia81e5f782015-04-08 12:29:19 +0300494 error_setg(&s->migration_blocker, "The vdi format used by node '%s' "
495 "does not support live migration",
496 bdrv_get_device_or_node_name(bs));
Markus Armbruster386f6c02020-07-07 18:06:10 +0200497 ret = migrate_add_blocker(s->migration_blocker, errp);
498 if (ret < 0) {
Ashijeet Acharyafe44dc92017-01-16 17:01:53 +0530499 error_free(s->migration_blocker);
500 goto fail_free_bmap;
501 }
Kevin Wolffc9d1062011-11-22 16:46:26 +0100502
Paolo Bonzini1e886632017-06-29 15:27:41 +0200503 qemu_co_rwlock_init(&s->bmap_lock);
Max Reitzf0ab6f12015-02-27 14:54:39 -0500504
Stefan Weil9aebd982009-07-31 21:45:56 +0200505 return 0;
506
507 fail_free_bmap:
Kevin Wolf17cce732014-05-20 13:25:43 +0200508 qemu_vfree(s->bmap);
Stefan Weil9aebd982009-07-31 21:45:56 +0200509
510 fail:
Stefan Weil8937f822013-01-17 21:45:27 +0100511 return ret;
Stefan Weil9aebd982009-07-31 21:45:56 +0200512}
513
Jeff Codyecfe2bb2012-09-20 15:13:32 -0400514static int vdi_reopen_prepare(BDRVReopenState *state,
515 BlockReopenQueue *queue, Error **errp)
516{
517 return 0;
518}
519
Eric Blake67635f62018-02-13 14:26:57 -0600520static int coroutine_fn vdi_co_block_status(BlockDriverState *bs,
521 bool want_zero,
522 int64_t offset, int64_t bytes,
523 int64_t *pnum, int64_t *map,
524 BlockDriverState **file)
Stefan Weil9aebd982009-07-31 21:45:56 +0200525{
Stefan Weil9aebd982009-07-31 21:45:56 +0200526 BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
Eric Blake67635f62018-02-13 14:26:57 -0600527 size_t bmap_index = offset / s->block_size;
528 size_t index_in_block = offset % s->block_size;
Stefan Weil9aebd982009-07-31 21:45:56 +0200529 uint32_t bmap_entry = le32_to_cpu(s->bmap[bmap_index]);
Paolo Bonzini4bc74be2013-09-04 19:00:30 +0200530 int result;
531
Eric Blake67635f62018-02-13 14:26:57 -0600532 logout("%p, %" PRId64 ", %" PRId64 ", %p\n", bs, offset, bytes, pnum);
533 *pnum = MIN(s->block_size - index_in_block, bytes);
Paolo Bonzini4bc74be2013-09-04 19:00:30 +0200534 result = VDI_IS_ALLOCATED(bmap_entry);
535 if (!result) {
Vladimir Sementsov-Ogievskiy2ea03322020-05-28 12:43:58 +0300536 return BDRV_BLOCK_ZERO;
Paolo Bonzini4bc74be2013-09-04 19:00:30 +0200537 }
538
Eric Blake67635f62018-02-13 14:26:57 -0600539 *map = s->header.offset_data + (uint64_t)bmap_entry * s->block_size +
540 index_in_block;
Fam Zheng8bfb1372016-01-26 11:58:56 +0800541 *file = bs->file->bs;
Max Reitzad6434d2019-07-25 17:55:10 +0200542 return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID |
543 (s->header.image_type == VDI_TYPE_STATIC ? BDRV_BLOCK_RECURSE : 0);
Stefan Weil9aebd982009-07-31 21:45:56 +0200544}
545
Kevin Wolf0865bb62016-04-25 16:22:39 +0200546static int coroutine_fn
547vdi_co_preadv(BlockDriverState *bs, uint64_t offset, uint64_t bytes,
548 QEMUIOVector *qiov, int flags)
Stefan Weil9aebd982009-07-31 21:45:56 +0200549{
Stefan Weil9aebd982009-07-31 21:45:56 +0200550 BDRVVdiState *s = bs->opaque;
Kevin Wolf0865bb62016-04-25 16:22:39 +0200551 QEMUIOVector local_qiov;
Stefan Weil9aebd982009-07-31 21:45:56 +0200552 uint32_t bmap_entry;
553 uint32_t block_index;
Kevin Wolf0865bb62016-04-25 16:22:39 +0200554 uint32_t offset_in_block;
555 uint32_t n_bytes;
556 uint64_t bytes_done = 0;
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100557 int ret = 0;
Paolo Bonzini4de659e2012-03-19 18:07:47 +0100558
559 logout("\n");
Stefan Weil9aebd982009-07-31 21:45:56 +0200560
Kevin Wolf0865bb62016-04-25 16:22:39 +0200561 qemu_iovec_init(&local_qiov, qiov->niov);
Stefan Weil9aebd982009-07-31 21:45:56 +0200562
Kevin Wolf0865bb62016-04-25 16:22:39 +0200563 while (ret >= 0 && bytes > 0) {
564 block_index = offset / s->block_size;
565 offset_in_block = offset % s->block_size;
566 n_bytes = MIN(bytes, s->block_size - offset_in_block);
567
568 logout("will read %u bytes starting at offset %" PRIu64 "\n",
569 n_bytes, offset);
Stefan Weil9aebd982009-07-31 21:45:56 +0200570
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100571 /* prepare next AIO request */
Paolo Bonzini1e886632017-06-29 15:27:41 +0200572 qemu_co_rwlock_rdlock(&s->bmap_lock);
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100573 bmap_entry = le32_to_cpu(s->bmap[block_index]);
Paolo Bonzini1e886632017-06-29 15:27:41 +0200574 qemu_co_rwlock_unlock(&s->bmap_lock);
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100575 if (!VDI_IS_ALLOCATED(bmap_entry)) {
576 /* Block not allocated, return zeros, no need to wait. */
Kevin Wolf0865bb62016-04-25 16:22:39 +0200577 qemu_iovec_memset(qiov, bytes_done, 0, n_bytes);
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100578 ret = 0;
579 } else {
Kevin Wolf0865bb62016-04-25 16:22:39 +0200580 uint64_t data_offset = s->header.offset_data +
581 (uint64_t)bmap_entry * s->block_size +
582 offset_in_block;
Paolo Bonzini0c7bfc32012-03-19 18:07:46 +0100583
Kevin Wolf0865bb62016-04-25 16:22:39 +0200584 qemu_iovec_reset(&local_qiov);
585 qemu_iovec_concat(&local_qiov, qiov, bytes_done, n_bytes);
586
Kevin Wolfa03ef882016-06-20 21:31:46 +0200587 ret = bdrv_co_preadv(bs->file, data_offset, n_bytes,
Kevin Wolf0865bb62016-04-25 16:22:39 +0200588 &local_qiov, 0);
589 }
590 logout("%u bytes read\n", n_bytes);
591
592 bytes -= n_bytes;
593 offset += n_bytes;
594 bytes_done += n_bytes;
Paolo Bonzini3d46a752012-03-19 18:07:45 +0100595 }
596
Kevin Wolf0865bb62016-04-25 16:22:39 +0200597 qemu_iovec_destroy(&local_qiov);
598
Paolo Bonzini3d46a752012-03-19 18:07:45 +0100599 return ret;
Stefan Weil9aebd982009-07-31 21:45:56 +0200600}
601
Kevin Wolffde9d562016-04-25 16:22:39 +0200602static int coroutine_fn
603vdi_co_pwritev(BlockDriverState *bs, uint64_t offset, uint64_t bytes,
604 QEMUIOVector *qiov, int flags)
Stefan Weil9aebd982009-07-31 21:45:56 +0200605{
Stefan Weil9aebd982009-07-31 21:45:56 +0200606 BDRVVdiState *s = bs->opaque;
Kevin Wolffde9d562016-04-25 16:22:39 +0200607 QEMUIOVector local_qiov;
Stefan Weil9aebd982009-07-31 21:45:56 +0200608 uint32_t bmap_entry;
609 uint32_t block_index;
Kevin Wolffde9d562016-04-25 16:22:39 +0200610 uint32_t offset_in_block;
611 uint32_t n_bytes;
Paolo Bonzini1e886632017-06-29 15:27:41 +0200612 uint64_t data_offset;
Paolo Bonzinibfc45fc2012-03-19 18:07:48 +0100613 uint32_t bmap_first = VDI_UNALLOCATED;
614 uint32_t bmap_last = VDI_UNALLOCATED;
Paolo Bonzinibfc45fc2012-03-19 18:07:48 +0100615 uint8_t *block = NULL;
Kevin Wolffde9d562016-04-25 16:22:39 +0200616 uint64_t bytes_done = 0;
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100617 int ret = 0;
Paolo Bonzini4de659e2012-03-19 18:07:47 +0100618
619 logout("\n");
Stefan Weil9aebd982009-07-31 21:45:56 +0200620
Kevin Wolffde9d562016-04-25 16:22:39 +0200621 qemu_iovec_init(&local_qiov, qiov->niov);
Paolo Bonzini0c7bfc32012-03-19 18:07:46 +0100622
Kevin Wolffde9d562016-04-25 16:22:39 +0200623 while (ret >= 0 && bytes > 0) {
624 block_index = offset / s->block_size;
625 offset_in_block = offset % s->block_size;
626 n_bytes = MIN(bytes, s->block_size - offset_in_block);
627
628 logout("will write %u bytes starting at offset %" PRIu64 "\n",
629 n_bytes, offset);
Paolo Bonzini0c7bfc32012-03-19 18:07:46 +0100630
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100631 /* prepare next AIO request */
Paolo Bonzini1e886632017-06-29 15:27:41 +0200632 qemu_co_rwlock_rdlock(&s->bmap_lock);
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100633 bmap_entry = le32_to_cpu(s->bmap[block_index]);
634 if (!VDI_IS_ALLOCATED(bmap_entry)) {
635 /* Allocate new block and write to it. */
Kevin Wolffde9d562016-04-25 16:22:39 +0200636 uint64_t data_offset;
Paolo Bonzini1e886632017-06-29 15:27:41 +0200637 qemu_co_rwlock_upgrade(&s->bmap_lock);
638 bmap_entry = le32_to_cpu(s->bmap[block_index]);
639 if (VDI_IS_ALLOCATED(bmap_entry)) {
640 /* A concurrent allocation did the work for us. */
641 qemu_co_rwlock_downgrade(&s->bmap_lock);
642 goto nonallocating_write;
643 }
644
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100645 bmap_entry = s->header.blocks_allocated;
646 s->bmap[block_index] = cpu_to_le32(bmap_entry);
647 s->header.blocks_allocated++;
Kevin Wolffde9d562016-04-25 16:22:39 +0200648 data_offset = s->header.offset_data +
649 (uint64_t)bmap_entry * s->block_size;
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100650 if (block == NULL) {
651 block = g_malloc(s->block_size);
652 bmap_first = block_index;
653 }
654 bmap_last = block_index;
655 /* Copy data to be written to new block and zero unused parts. */
Kevin Wolffde9d562016-04-25 16:22:39 +0200656 memset(block, 0, offset_in_block);
657 qemu_iovec_to_buf(qiov, bytes_done, block + offset_in_block,
658 n_bytes);
659 memset(block + offset_in_block + n_bytes, 0,
660 s->block_size - n_bytes - offset_in_block);
Max Reitzf0ab6f12015-02-27 14:54:39 -0500661
Paolo Bonzini1e886632017-06-29 15:27:41 +0200662 /* Write the new block under CoRwLock write-side protection,
663 * so this full-cluster write does not overlap a partial write
664 * of the same cluster, issued from the "else" branch.
665 */
Kevin Wolfd9ca2ea2016-06-20 20:09:15 +0200666 ret = bdrv_pwrite(bs->file, data_offset, block, s->block_size);
Paolo Bonzini1e886632017-06-29 15:27:41 +0200667 qemu_co_rwlock_unlock(&s->bmap_lock);
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100668 } else {
Paolo Bonzini1e886632017-06-29 15:27:41 +0200669nonallocating_write:
670 data_offset = s->header.offset_data +
671 (uint64_t)bmap_entry * s->block_size +
672 offset_in_block;
673 qemu_co_rwlock_unlock(&s->bmap_lock);
Kevin Wolffde9d562016-04-25 16:22:39 +0200674
675 qemu_iovec_reset(&local_qiov);
676 qemu_iovec_concat(&local_qiov, qiov, bytes_done, n_bytes);
677
Kevin Wolfa03ef882016-06-20 21:31:46 +0200678 ret = bdrv_co_pwritev(bs->file, data_offset, n_bytes,
Kevin Wolffde9d562016-04-25 16:22:39 +0200679 &local_qiov, 0);
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100680 }
681
Kevin Wolffde9d562016-04-25 16:22:39 +0200682 bytes -= n_bytes;
683 offset += n_bytes;
684 bytes_done += n_bytes;
Paolo Bonzinieb9566d2012-03-19 18:07:51 +0100685
Kevin Wolffde9d562016-04-25 16:22:39 +0200686 logout("%u bytes written\n", n_bytes);
Paolo Bonzini3d46a752012-03-19 18:07:45 +0100687 }
Stefan Weil9aebd982009-07-31 21:45:56 +0200688
Kevin Wolffde9d562016-04-25 16:22:39 +0200689 qemu_iovec_destroy(&local_qiov);
690
Paolo Bonzini0c7bfc32012-03-19 18:07:46 +0100691 logout("finished data write\n");
Paolo Bonzini4eea78e2012-03-19 18:07:50 +0100692 if (ret < 0) {
David Edmondson574b8302021-03-25 12:29:36 +0100693 g_free(block);
Paolo Bonzini4eea78e2012-03-19 18:07:50 +0100694 return ret;
695 }
696
697 if (block) {
698 /* One or more new blocks were allocated. */
David Edmondson07ee2ab2021-03-25 12:29:37 +0100699 VdiHeader *header;
Paolo Bonzini4eea78e2012-03-19 18:07:50 +0100700 uint8_t *base;
701 uint64_t offset;
Kevin Wolffde9d562016-04-25 16:22:39 +0200702 uint32_t n_sectors;
Paolo Bonzini4eea78e2012-03-19 18:07:50 +0100703
David Edmondson07ee2ab2021-03-25 12:29:37 +0100704 g_free(block);
705 header = g_malloc(sizeof(*header));
706
Paolo Bonzini4eea78e2012-03-19 18:07:50 +0100707 logout("now writing modified header\n");
708 assert(VDI_IS_ALLOCATED(bmap_first));
709 *header = s->header;
710 vdi_header_to_le(header);
David Edmondson07ee2ab2021-03-25 12:29:37 +0100711 ret = bdrv_pwrite(bs->file, 0, header, sizeof(*header));
712 g_free(header);
Paolo Bonzini4eea78e2012-03-19 18:07:50 +0100713
714 if (ret < 0) {
715 return ret;
Paolo Bonzini0c7bfc32012-03-19 18:07:46 +0100716 }
Paolo Bonzini4eea78e2012-03-19 18:07:50 +0100717
718 logout("now writing modified block map entry %u...%u\n",
719 bmap_first, bmap_last);
720 /* Write modified sectors from block map. */
721 bmap_first /= (SECTOR_SIZE / sizeof(uint32_t));
722 bmap_last /= (SECTOR_SIZE / sizeof(uint32_t));
723 n_sectors = bmap_last - bmap_first + 1;
724 offset = s->bmap_sector + bmap_first;
725 base = ((uint8_t *)&s->bmap[0]) + bmap_first * SECTOR_SIZE;
726 logout("will write %u block map sectors starting from entry %u\n",
727 n_sectors, bmap_first);
Alberto Garciad4f18972019-05-01 21:13:56 +0300728 ret = bdrv_pwrite(bs->file, offset * SECTOR_SIZE, base,
729 n_sectors * SECTOR_SIZE);
Paolo Bonzini0c7bfc32012-03-19 18:07:46 +0100730 }
731
Alberto Garciad4f18972019-05-01 21:13:56 +0300732 return ret < 0 ? ret : 0;
Stefan Weil9aebd982009-07-31 21:45:56 +0200733}
734
Max Reitze3810572018-03-12 17:55:28 +0100735static int coroutine_fn vdi_co_do_create(BlockdevCreateOptions *create_options,
Max Reitz49858b52018-03-12 17:55:26 +0100736 size_t block_size, Error **errp)
Stefan Weil9aebd982009-07-31 21:45:56 +0200737{
Max Reitze3810572018-03-12 17:55:28 +0100738 BlockdevCreateOptionsVdi *vdi_opts;
Jeff Codydddc7752014-07-23 17:22:59 -0400739 int ret = 0;
Stefan Weil9aebd982009-07-31 21:45:56 +0200740 uint64_t bytes = 0;
741 uint32_t blocks;
Kevin Wolf61fa6482018-03-20 15:08:00 +0100742 uint32_t image_type;
Stefan Weil9aebd982009-07-31 21:45:56 +0200743 VdiHeader header;
744 size_t i;
745 size_t bmap_size;
Jeff Cody70747862014-07-23 17:22:58 -0400746 int64_t offset = 0;
Max Reitzec73f062018-03-12 17:55:27 +0100747 BlockDriverState *bs_file = NULL;
Kevin Wolfa08f0c32016-03-08 15:57:05 +0100748 BlockBackend *blk = NULL;
Jeff Cody70747862014-07-23 17:22:58 -0400749 uint32_t *bmap = NULL;
Peter Maydellac928b82018-12-10 11:26:48 +0000750 QemuUUID uuid;
Stefan Weil9aebd982009-07-31 21:45:56 +0200751
Max Reitze3810572018-03-12 17:55:28 +0100752 assert(create_options->driver == BLOCKDEV_DRIVER_VDI);
753 vdi_opts = &create_options->u.vdi;
754
Stefan Weil9aebd982009-07-31 21:45:56 +0200755 logout("\n");
756
Kevin Wolfda232482018-03-09 19:53:19 +0100757 /* Validate options and set default values */
Max Reitz49858b52018-03-12 17:55:26 +0100758 bytes = vdi_opts->size;
Kevin Wolf61fa6482018-03-20 15:08:00 +0100759
760 if (!vdi_opts->has_preallocation) {
761 vdi_opts->preallocation = PREALLOC_MODE_OFF;
Stefan Weil9aebd982009-07-31 21:45:56 +0200762 }
Kevin Wolf61fa6482018-03-20 15:08:00 +0100763 switch (vdi_opts->preallocation) {
764 case PREALLOC_MODE_OFF:
765 image_type = VDI_TYPE_DYNAMIC;
766 break;
767 case PREALLOC_MODE_METADATA:
768 image_type = VDI_TYPE_STATIC;
769 break;
770 default:
771 error_setg(errp, "Preallocation mode not supported for vdi");
772 return -EINVAL;
773 }
774
Max Reitz49858b52018-03-12 17:55:26 +0100775#ifndef CONFIG_VDI_STATIC_IMAGE
776 if (image_type == VDI_TYPE_STATIC) {
777 ret = -ENOTSUP;
778 error_setg(errp, "Statically allocated images cannot be created in "
779 "this build");
780 goto exit;
781 }
782#endif
783#ifndef CONFIG_VDI_BLOCK_SIZE
784 if (block_size != DEFAULT_CLUSTER_SIZE) {
785 ret = -ENOTSUP;
786 error_setg(errp,
787 "A non-default cluster size is not supported in this build");
788 goto exit;
789 }
Chunyan Liu004b7f22014-06-05 17:21:07 +0800790#endif
Stefan Weil9aebd982009-07-31 21:45:56 +0200791
Jeff Cody63fa06d2014-03-28 11:42:24 -0400792 if (bytes > VDI_DISK_SIZE_MAX) {
Jeff Codydddc7752014-07-23 17:22:59 -0400793 ret = -ENOTSUP;
Jeff Cody63fa06d2014-03-28 11:42:24 -0400794 error_setg(errp, "Unsupported VDI image size (size is 0x%" PRIx64
795 ", max supported is 0x%" PRIx64 ")",
796 bytes, VDI_DISK_SIZE_MAX);
797 goto exit;
798 }
799
Kevin Wolfda232482018-03-09 19:53:19 +0100800 /* Create BlockBackend to write to the image */
Max Reitzec73f062018-03-12 17:55:27 +0100801 bs_file = bdrv_open_blockdev_ref(vdi_opts->file, errp);
802 if (!bs_file) {
803 ret = -EIO;
Jeff Cody63fa06d2014-03-28 11:42:24 -0400804 goto exit;
Stefan Weil9aebd982009-07-31 21:45:56 +0200805 }
Kevin Wolfa08f0c32016-03-08 15:57:05 +0100806
Eric Blakea3aeeab2020-04-28 14:26:46 -0500807 blk = blk_new_with_bs(bs_file, BLK_PERM_WRITE | BLK_PERM_RESIZE,
808 BLK_PERM_ALL, errp);
809 if (!blk) {
810 ret = -EPERM;
Jeff Cody70747862014-07-23 17:22:58 -0400811 goto exit;
Chunyan Liu4ab15592014-06-30 14:29:58 +0800812 }
813
Kevin Wolfa08f0c32016-03-08 15:57:05 +0100814 blk_set_allow_write_beyond_eof(blk, true);
815
Stefan Weilf21dc3a2010-05-12 20:25:45 +0200816 /* We need enough blocks to store the given disk size,
817 so always round up. */
Max Reitze9082e42014-10-21 10:51:25 +0200818 blocks = DIV_ROUND_UP(bytes, block_size);
Stefan Weilf21dc3a2010-05-12 20:25:45 +0200819
Stefan Weil9aebd982009-07-31 21:45:56 +0200820 bmap_size = blocks * sizeof(uint32_t);
Max Reitze9082e42014-10-21 10:51:25 +0200821 bmap_size = ROUND_UP(bmap_size, SECTOR_SIZE);
Stefan Weil9aebd982009-07-31 21:45:56 +0200822
823 memset(&header, 0, sizeof(header));
Blue Swirl1786dc12009-08-15 11:33:58 +0000824 pstrcpy(header.text, sizeof(header.text), VDI_TEXT);
Stefan Weil9aebd982009-07-31 21:45:56 +0200825 header.signature = VDI_SIGNATURE;
826 header.version = VDI_VERSION_1_1;
827 header.header_size = 0x180;
828 header.image_type = image_type;
829 header.offset_bmap = 0x200;
830 header.offset_data = 0x200 + bmap_size;
831 header.sector_size = SECTOR_SIZE;
832 header.disk_size = bytes;
833 header.block_size = block_size;
834 header.blocks_in_image = blocks;
Stefan Weil6eea90e2009-08-14 21:50:02 +0200835 if (image_type == VDI_TYPE_STATIC) {
836 header.blocks_allocated = blocks;
837 }
Peter Maydellac928b82018-12-10 11:26:48 +0000838 qemu_uuid_generate(&uuid);
839 header.uuid_image = uuid;
840 qemu_uuid_generate(&uuid);
841 header.uuid_last_snap = uuid;
Stefan Weil9aebd982009-07-31 21:45:56 +0200842 /* There is no need to set header.uuid_link or header.uuid_parent here. */
Kevin Wolf95a14d52018-03-20 14:41:53 +0100843 if (VDI_DEBUG) {
844 vdi_header_print(&header);
845 }
Stefan Weil9aebd982009-07-31 21:45:56 +0200846 vdi_header_to_le(&header);
Eric Blake8341f002016-05-06 10:26:27 -0600847 ret = blk_pwrite(blk, offset, &header, sizeof(header), 0);
Jeff Codydddc7752014-07-23 17:22:59 -0400848 if (ret < 0) {
Max Reitzec73f062018-03-12 17:55:27 +0100849 error_setg(errp, "Error writing header");
Jeff Cody70747862014-07-23 17:22:58 -0400850 goto exit;
Stefan Weil9aebd982009-07-31 21:45:56 +0200851 }
Jeff Cody70747862014-07-23 17:22:58 -0400852 offset += sizeof(header);
Stefan Weil9aebd982009-07-31 21:45:56 +0200853
Stefan Weilb76b6e92010-05-06 20:53:47 +0200854 if (bmap_size > 0) {
Kevin Wolf17cce732014-05-20 13:25:43 +0200855 bmap = g_try_malloc0(bmap_size);
856 if (bmap == NULL) {
857 ret = -ENOMEM;
858 error_setg(errp, "Could not allocate bmap");
859 goto exit;
860 }
Stefan Weil514f21a2012-08-17 15:23:24 +0200861 for (i = 0; i < blocks; i++) {
862 if (image_type == VDI_TYPE_STATIC) {
863 bmap[i] = i;
864 } else {
865 bmap[i] = VDI_UNALLOCATED;
866 }
Stefan Weil9aebd982009-07-31 21:45:56 +0200867 }
Eric Blake8341f002016-05-06 10:26:27 -0600868 ret = blk_pwrite(blk, offset, bmap, bmap_size, 0);
Jeff Codydddc7752014-07-23 17:22:59 -0400869 if (ret < 0) {
Max Reitzec73f062018-03-12 17:55:27 +0100870 error_setg(errp, "Error writing bmap");
Jeff Cody70747862014-07-23 17:22:58 -0400871 goto exit;
Stefan Weil514f21a2012-08-17 15:23:24 +0200872 }
Jeff Cody70747862014-07-23 17:22:58 -0400873 offset += bmap_size;
Stefan Weil9aebd982009-07-31 21:45:56 +0200874 }
Stefan Weil514f21a2012-08-17 15:23:24 +0200875
Stefan Weil9aebd982009-07-31 21:45:56 +0200876 if (image_type == VDI_TYPE_STATIC) {
Max Reitzc80d8b02019-09-18 11:51:40 +0200877 ret = blk_truncate(blk, offset + blocks * block_size, false,
Kevin Wolf8c6242b2020-04-24 14:54:41 +0200878 PREALLOC_MODE_OFF, 0, errp);
Jeff Codydddc7752014-07-23 17:22:59 -0400879 if (ret < 0) {
Max Reitzec73f062018-03-12 17:55:27 +0100880 error_prepend(errp, "Failed to statically allocate file");
Jeff Cody70747862014-07-23 17:22:58 -0400881 goto exit;
Stefan Weil9aebd982009-07-31 21:45:56 +0200882 }
883 }
884
Kevin Wolf53618dd2018-05-25 14:48:16 +0200885 ret = 0;
Jeff Cody63fa06d2014-03-28 11:42:24 -0400886exit:
Kevin Wolfa08f0c32016-03-08 15:57:05 +0100887 blk_unref(blk);
Max Reitzec73f062018-03-12 17:55:27 +0100888 bdrv_unref(bs_file);
Jeff Cody70747862014-07-23 17:22:58 -0400889 g_free(bmap);
Jeff Codydddc7752014-07-23 17:22:59 -0400890 return ret;
Stefan Weil9aebd982009-07-31 21:45:56 +0200891}
892
Max Reitze3810572018-03-12 17:55:28 +0100893static int coroutine_fn vdi_co_create(BlockdevCreateOptions *create_options,
894 Error **errp)
895{
896 return vdi_co_do_create(create_options, DEFAULT_CLUSTER_SIZE, errp);
897}
898
Maxim Levitskyb92902d2020-03-26 03:12:17 +0200899static int coroutine_fn vdi_co_create_opts(BlockDriver *drv,
900 const char *filename,
901 QemuOpts *opts,
Max Reitz49858b52018-03-12 17:55:26 +0100902 Error **errp)
903{
904 QDict *qdict = NULL;
Max Reitze3810572018-03-12 17:55:28 +0100905 BlockdevCreateOptions *create_options = NULL;
Max Reitzec73f062018-03-12 17:55:27 +0100906 BlockDriverState *bs_file = NULL;
Max Reitz49858b52018-03-12 17:55:26 +0100907 uint64_t block_size = DEFAULT_CLUSTER_SIZE;
Kevin Wolf61fa6482018-03-20 15:08:00 +0100908 bool is_static = false;
Max Reitz49858b52018-03-12 17:55:26 +0100909 Visitor *v;
Max Reitz49858b52018-03-12 17:55:26 +0100910 int ret;
911
Kevin Wolfda232482018-03-09 19:53:19 +0100912 /* Parse options and convert legacy syntax.
913 *
914 * Since CONFIG_VDI_BLOCK_SIZE is disabled by default,
Max Reitz49858b52018-03-12 17:55:26 +0100915 * cluster-size is not part of the QAPI schema; therefore we have
916 * to parse it before creating the QAPI object. */
917#if defined(CONFIG_VDI_BLOCK_SIZE)
918 block_size = qemu_opt_get_size_del(opts,
919 BLOCK_OPT_CLUSTER_SIZE,
920 DEFAULT_CLUSTER_SIZE);
921 if (block_size < BDRV_SECTOR_SIZE || block_size > UINT32_MAX ||
922 !is_power_of_2(block_size))
923 {
924 error_setg(errp, "Invalid cluster size");
925 ret = -EINVAL;
926 goto done;
927 }
928#endif
Kevin Wolf61fa6482018-03-20 15:08:00 +0100929 if (qemu_opt_get_bool_del(opts, BLOCK_OPT_STATIC, false)) {
930 is_static = true;
931 }
Max Reitz49858b52018-03-12 17:55:26 +0100932
933 qdict = qemu_opts_to_qdict_filtered(opts, NULL, &vdi_create_opts, true);
934
Kevin Wolfda232482018-03-09 19:53:19 +0100935 /* Create and open the file (protocol layer) */
Max Reitzec73f062018-03-12 17:55:27 +0100936 ret = bdrv_create_file(filename, opts, errp);
937 if (ret < 0) {
938 goto done;
939 }
940
941 bs_file = bdrv_open(filename, NULL, NULL,
942 BDRV_O_RDWR | BDRV_O_RESIZE | BDRV_O_PROTOCOL, errp);
943 if (!bs_file) {
944 ret = -EIO;
945 goto done;
946 }
947
Max Reitze3810572018-03-12 17:55:28 +0100948 qdict_put_str(qdict, "driver", "vdi");
Max Reitzec73f062018-03-12 17:55:27 +0100949 qdict_put_str(qdict, "file", bs_file->node_name);
Kevin Wolf61fa6482018-03-20 15:08:00 +0100950 if (is_static) {
951 qdict_put_str(qdict, "preallocation", "metadata");
952 }
Max Reitz49858b52018-03-12 17:55:26 +0100953
954 /* Get the QAPI object */
Markus Armbrusterf8534652018-06-14 21:14:34 +0200955 v = qobject_input_visitor_new_flat_confused(qdict, errp);
956 if (!v) {
957 ret = -EINVAL;
958 goto done;
959 }
Markus Armbrusterb11a0932020-07-07 18:06:07 +0200960 visit_type_BlockdevCreateOptions(v, NULL, &create_options, errp);
Max Reitz49858b52018-03-12 17:55:26 +0100961 visit_free(v);
Markus Armbrusterb11a0932020-07-07 18:06:07 +0200962 if (!create_options) {
Max Reitz49858b52018-03-12 17:55:26 +0100963 ret = -EINVAL;
964 goto done;
965 }
966
Kevin Wolfda232482018-03-09 19:53:19 +0100967 /* Silently round up size */
Max Reitze3810572018-03-12 17:55:28 +0100968 assert(create_options->driver == BLOCKDEV_DRIVER_VDI);
969 create_options->u.vdi.size = ROUND_UP(create_options->u.vdi.size,
970 BDRV_SECTOR_SIZE);
Max Reitz49858b52018-03-12 17:55:26 +0100971
Kevin Wolfda232482018-03-09 19:53:19 +0100972 /* Create the vdi image (format layer) */
Max Reitzec73f062018-03-12 17:55:27 +0100973 ret = vdi_co_do_create(create_options, block_size, errp);
Max Reitz49858b52018-03-12 17:55:26 +0100974done:
Marc-André Lureaucb3e7f02018-04-19 17:01:43 +0200975 qobject_unref(qdict);
Max Reitze3810572018-03-12 17:55:28 +0100976 qapi_free_BlockdevCreateOptions(create_options);
Max Reitzec73f062018-03-12 17:55:27 +0100977 bdrv_unref(bs_file);
Max Reitz49858b52018-03-12 17:55:26 +0100978 return ret;
979}
980
Stefan Weil9aebd982009-07-31 21:45:56 +0200981static void vdi_close(BlockDriverState *bs)
982{
Kevin Wolffc9d1062011-11-22 16:46:26 +0100983 BDRVVdiState *s = bs->opaque;
Kevin Wolf6ac5f382011-11-22 16:57:34 +0100984
Kevin Wolf17cce732014-05-20 13:25:43 +0200985 qemu_vfree(s->bmap);
Kevin Wolf6ac5f382011-11-22 16:57:34 +0100986
Kevin Wolffc9d1062011-11-22 16:46:26 +0100987 migrate_del_blocker(s->migration_blocker);
988 error_free(s->migration_blocker);
Stefan Weil9aebd982009-07-31 21:45:56 +0200989}
990
Max Reitz0a28bf22019-07-24 19:12:35 +0200991static int vdi_has_zero_init(BlockDriverState *bs)
992{
993 BDRVVdiState *s = bs->opaque;
994
995 if (s->header.image_type == VDI_TYPE_STATIC) {
996 return bdrv_has_zero_init(bs->file->bs);
997 } else {
998 return 1;
999 }
1000}
1001
Chunyan Liu004b7f22014-06-05 17:21:07 +08001002static QemuOptsList vdi_create_opts = {
1003 .name = "vdi-create-opts",
1004 .head = QTAILQ_HEAD_INITIALIZER(vdi_create_opts.head),
1005 .desc = {
1006 {
1007 .name = BLOCK_OPT_SIZE,
1008 .type = QEMU_OPT_SIZE,
1009 .help = "Virtual disk size"
1010 },
Stefan Weil9aebd982009-07-31 21:45:56 +02001011#if defined(CONFIG_VDI_BLOCK_SIZE)
Chunyan Liu004b7f22014-06-05 17:21:07 +08001012 {
1013 .name = BLOCK_OPT_CLUSTER_SIZE,
1014 .type = QEMU_OPT_SIZE,
1015 .help = "VDI cluster (block) size",
1016 .def_value_str = stringify(DEFAULT_CLUSTER_SIZE)
1017 },
Stefan Weil9aebd982009-07-31 21:45:56 +02001018#endif
1019#if defined(CONFIG_VDI_STATIC_IMAGE)
Chunyan Liu004b7f22014-06-05 17:21:07 +08001020 {
1021 .name = BLOCK_OPT_STATIC,
1022 .type = QEMU_OPT_BOOL,
1023 .help = "VDI static (pre-allocated) image",
1024 .def_value_str = "off"
1025 },
Stefan Weil9aebd982009-07-31 21:45:56 +02001026#endif
Chunyan Liu004b7f22014-06-05 17:21:07 +08001027 /* TODO: An additional option to set UUID values might be useful. */
1028 { /* end of list */ }
1029 }
Stefan Weil9aebd982009-07-31 21:45:56 +02001030};
1031
1032static BlockDriver bdrv_vdi = {
1033 .format_name = "vdi",
1034 .instance_size = sizeof(BDRVVdiState),
1035 .bdrv_probe = vdi_probe,
1036 .bdrv_open = vdi_open,
1037 .bdrv_close = vdi_close,
Jeff Codyecfe2bb2012-09-20 15:13:32 -04001038 .bdrv_reopen_prepare = vdi_reopen_prepare,
Max Reitz69dca432020-05-13 13:05:39 +02001039 .bdrv_child_perm = bdrv_default_perms,
Max Reitze3810572018-03-12 17:55:28 +01001040 .bdrv_co_create = vdi_co_create,
Kevin Wolfda232482018-03-09 19:53:19 +01001041 .bdrv_co_create_opts = vdi_co_create_opts,
Max Reitz0a28bf22019-07-24 19:12:35 +02001042 .bdrv_has_zero_init = vdi_has_zero_init,
Eric Blake67635f62018-02-13 14:26:57 -06001043 .bdrv_co_block_status = vdi_co_block_status,
Stefan Weil9aebd982009-07-31 21:45:56 +02001044 .bdrv_make_empty = vdi_make_empty,
1045
Kevin Wolf0865bb62016-04-25 16:22:39 +02001046 .bdrv_co_preadv = vdi_co_preadv,
Stefan Weil9aebd982009-07-31 21:45:56 +02001047#if defined(CONFIG_VDI_WRITE)
Kevin Wolffde9d562016-04-25 16:22:39 +02001048 .bdrv_co_pwritev = vdi_co_pwritev,
Stefan Weil9aebd982009-07-31 21:45:56 +02001049#endif
1050
1051 .bdrv_get_info = vdi_get_info,
1052
Max Reitzd67066d2020-05-13 13:05:12 +02001053 .is_format = true,
Chunyan Liu004b7f22014-06-05 17:21:07 +08001054 .create_opts = &vdi_create_opts,
Paolo Bonzini2fd61632018-03-01 17:36:19 +01001055 .bdrv_co_check = vdi_co_check,
Stefan Weil9aebd982009-07-31 21:45:56 +02001056};
1057
1058static void bdrv_vdi_init(void)
1059{
1060 logout("\n");
1061 bdrv_register(&bdrv_vdi);
1062}
1063
1064block_init(bdrv_vdi_init);