blob: 052a753159d6e649cc314da6bd7ccc588b7b4920 [file] [log] [blame]
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001/*
2 * Block driver for Hyper-V VHDX Images
3 *
4 * Copyright (c) 2013 Red Hat, Inc.,
5 *
6 * Authors:
7 * Jeff Cody <jcody@redhat.com>
8 *
Jeff Cody6e9d2902013-10-30 10:44:38 -04009 * This is based on the "VHDX Format Specification v1.00", published 8/25/2012
Jeff Codye8d4e5f2013-04-29 14:48:18 -040010 * by Microsoft:
Jeff Cody6e9d2902013-10-30 10:44:38 -040011 * https://www.microsoft.com/en-us/download/details.aspx?id=34750
Jeff Codye8d4e5f2013-04-29 14:48:18 -040012 *
13 * This work is licensed under the terms of the GNU LGPL, version 2 or later.
14 * See the COPYING.LIB file in the top-level directory.
15 *
16 */
17
Peter Maydell80c71a22016-01-18 18:01:42 +000018#include "qemu/osdep.h"
Markus Armbrusterda34e652016-03-14 09:01:28 +010019#include "qapi/error.h"
Jeff Codye8d4e5f2013-04-29 14:48:18 -040020#include "qemu-common.h"
21#include "block/block_int.h"
Kevin Wolf10bf03a2016-03-08 15:57:05 +010022#include "sysemu/block-backend.h"
Jeff Codye8d4e5f2013-04-29 14:48:18 -040023#include "qemu/module.h"
24#include "qemu/crc32c.h"
Paolo Bonzini58369e22016-03-15 17:22:36 +010025#include "qemu/bswap.h"
Jeff Codye8d4e5f2013-04-29 14:48:18 -040026#include "block/vhdx.h"
Jeff Cody5641bf42013-10-01 11:59:20 -040027#include "migration/migration.h"
Fam Zhengcb6414d2016-09-21 12:27:16 +080028#include "qemu/uuid.h"
Jeff Cody3412f7b2013-10-30 10:44:53 -040029
30/* Options for VHDX creation */
31
32#define VHDX_BLOCK_OPT_LOG_SIZE "log_size"
33#define VHDX_BLOCK_OPT_BLOCK_SIZE "block_size"
34#define VHDX_BLOCK_OPT_ZERO "block_state_zero"
35
36typedef enum VHDXImageType {
37 VHDX_TYPE_DYNAMIC = 0,
38 VHDX_TYPE_FIXED,
39 VHDX_TYPE_DIFFERENCING, /* Currently unsupported */
40} VHDXImageType;
Jeff Codye8d4e5f2013-04-29 14:48:18 -040041
42/* Several metadata and region table data entries are identified by
43 * guids in a MS-specific GUID format. */
44
45
46/* ------- Known Region Table GUIDs ---------------------- */
47static const MSGUID bat_guid = { .data1 = 0x2dc27766,
48 .data2 = 0xf623,
49 .data3 = 0x4200,
50 .data4 = { 0x9d, 0x64, 0x11, 0x5e,
51 0x9b, 0xfd, 0x4a, 0x08} };
52
53static const MSGUID metadata_guid = { .data1 = 0x8b7ca206,
54 .data2 = 0x4790,
55 .data3 = 0x4b9a,
56 .data4 = { 0xb8, 0xfe, 0x57, 0x5f,
57 0x05, 0x0f, 0x88, 0x6e} };
58
59
60
61/* ------- Known Metadata Entry GUIDs ---------------------- */
62static const MSGUID file_param_guid = { .data1 = 0xcaa16737,
63 .data2 = 0xfa36,
64 .data3 = 0x4d43,
65 .data4 = { 0xb3, 0xb6, 0x33, 0xf0,
66 0xaa, 0x44, 0xe7, 0x6b} };
67
68static const MSGUID virtual_size_guid = { .data1 = 0x2FA54224,
69 .data2 = 0xcd1b,
70 .data3 = 0x4876,
71 .data4 = { 0xb2, 0x11, 0x5d, 0xbe,
72 0xd8, 0x3b, 0xf4, 0xb8} };
73
74static const MSGUID page83_guid = { .data1 = 0xbeca12ab,
75 .data2 = 0xb2e6,
76 .data3 = 0x4523,
77 .data4 = { 0x93, 0xef, 0xc3, 0x09,
78 0xe0, 0x00, 0xc7, 0x46} };
79
80
81static const MSGUID phys_sector_guid = { .data1 = 0xcda348c7,
82 .data2 = 0x445d,
83 .data3 = 0x4471,
84 .data4 = { 0x9c, 0xc9, 0xe9, 0x88,
85 0x52, 0x51, 0xc5, 0x56} };
86
87static const MSGUID parent_locator_guid = { .data1 = 0xa8d35f2d,
88 .data2 = 0xb30b,
89 .data3 = 0x454d,
90 .data4 = { 0xab, 0xf7, 0xd3,
91 0xd8, 0x48, 0x34,
92 0xab, 0x0c} };
93
94static const MSGUID logical_sector_guid = { .data1 = 0x8141bf1d,
95 .data2 = 0xa96f,
96 .data3 = 0x4709,
97 .data4 = { 0xba, 0x47, 0xf2,
98 0x33, 0xa8, 0xfa,
99 0xab, 0x5f} };
100
101/* Each parent type must have a valid GUID; this is for parent images
102 * of type 'VHDX'. If we were to allow e.g. a QCOW2 parent, we would
103 * need to make up our own QCOW2 GUID type */
Peter Maydellc2ebb052014-09-14 20:29:59 +0100104static const MSGUID parent_vhdx_guid __attribute__((unused))
105 = { .data1 = 0xb04aefb7,
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400106 .data2 = 0xd19e,
107 .data3 = 0x4a81,
108 .data4 = { 0xb7, 0x89, 0x25, 0xb8,
109 0xe9, 0x44, 0x59, 0x13} };
110
111
112#define META_FILE_PARAMETER_PRESENT 0x01
113#define META_VIRTUAL_DISK_SIZE_PRESENT 0x02
114#define META_PAGE_83_PRESENT 0x04
115#define META_LOGICAL_SECTOR_SIZE_PRESENT 0x08
116#define META_PHYS_SECTOR_SIZE_PRESENT 0x10
117#define META_PARENT_LOCATOR_PRESENT 0x20
118
119#define META_ALL_PRESENT \
120 (META_FILE_PARAMETER_PRESENT | META_VIRTUAL_DISK_SIZE_PRESENT | \
121 META_PAGE_83_PRESENT | META_LOGICAL_SECTOR_SIZE_PRESENT | \
122 META_PHYS_SECTOR_SIZE_PRESENT)
123
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400124
Jeff Cody059e2fb2013-04-29 14:48:19 -0400125typedef struct VHDXSectorInfo {
126 uint32_t bat_idx; /* BAT entry index */
127 uint32_t sectors_avail; /* sectors available in payload block */
128 uint32_t bytes_left; /* bytes left in the block after data to r/w */
129 uint32_t bytes_avail; /* bytes available in payload block */
130 uint64_t file_offset; /* absolute offset in bytes, in file */
131 uint64_t block_offset; /* block offset, in bytes */
132} VHDXSectorInfo;
133
Jeff Cody4f18b782013-10-30 10:44:39 -0400134/* Calculates new checksum.
135 *
136 * Zero is substituted during crc calculation for the original crc field
137 * crc_offset: byte offset in buf of the buffer crc
138 * buf: buffer pointer
139 * size: size of buffer (must be > crc_offset+4)
140 *
Jeff Cody4f75b522014-08-06 15:54:58 -0400141 * Note: The buffer should have all multi-byte data in little-endian format,
142 * and the resulting checksum is in little endian format.
Jeff Cody4f18b782013-10-30 10:44:39 -0400143 */
144uint32_t vhdx_update_checksum(uint8_t *buf, size_t size, int crc_offset)
145{
146 uint32_t crc;
147
148 assert(buf != NULL);
149 assert(size > (crc_offset + sizeof(crc)));
150
151 memset(buf + crc_offset, 0, sizeof(crc));
152 crc = crc32c(0xffffffff, buf, size);
Jeff Cody4f75b522014-08-06 15:54:58 -0400153 cpu_to_le32s(&crc);
Jeff Cody4f18b782013-10-30 10:44:39 -0400154 memcpy(buf + crc_offset, &crc, sizeof(crc));
155
156 return crc;
157}
158
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400159uint32_t vhdx_checksum_calc(uint32_t crc, uint8_t *buf, size_t size,
160 int crc_offset)
161{
162 uint32_t crc_new;
163 uint32_t crc_orig;
164 assert(buf != NULL);
165
166 if (crc_offset > 0) {
167 memcpy(&crc_orig, buf + crc_offset, sizeof(crc_orig));
168 memset(buf + crc_offset, 0, sizeof(crc_orig));
169 }
170
171 crc_new = crc32c(crc, buf, size);
172 if (crc_offset > 0) {
173 memcpy(buf + crc_offset, &crc_orig, sizeof(crc_orig));
174 }
175
176 return crc_new;
177}
178
179/* Validates the checksum of the buffer, with an in-place CRC.
180 *
181 * Zero is substituted during crc calculation for the original crc field,
182 * and the crc field is restored afterwards. But the buffer will be modifed
183 * during the calculation, so this may not be not suitable for multi-threaded
184 * use.
185 *
186 * crc_offset: byte offset in buf of the buffer crc
187 * buf: buffer pointer
188 * size: size of buffer (must be > crc_offset+4)
189 *
190 * returns true if checksum is valid, false otherwise
191 */
192bool vhdx_checksum_is_valid(uint8_t *buf, size_t size, int crc_offset)
193{
194 uint32_t crc_orig;
195 uint32_t crc;
196
197 assert(buf != NULL);
198 assert(size > (crc_offset + 4));
199
200 memcpy(&crc_orig, buf + crc_offset, sizeof(crc_orig));
201 crc_orig = le32_to_cpu(crc_orig);
202
203 crc = vhdx_checksum_calc(0xffffffff, buf, size, crc_offset);
204
205 return crc == crc_orig;
206}
207
208
209/*
Jeff Cody4f18b782013-10-30 10:44:39 -0400210 * This generates a UUID that is compliant with the MS GUIDs used
211 * in the VHDX spec (and elsewhere).
212 */
213void vhdx_guid_generate(MSGUID *guid)
214{
Fam Zhengcb6414d2016-09-21 12:27:16 +0800215 QemuUUID uuid;
Jeff Cody4f18b782013-10-30 10:44:39 -0400216 assert(guid != NULL);
217
Fam Zhengcb6414d2016-09-21 12:27:16 +0800218 qemu_uuid_generate(&uuid);
219 memcpy(guid, &uuid, sizeof(MSGUID));
Jeff Cody4f18b782013-10-30 10:44:39 -0400220}
221
Jeff Cody1a848fd2013-10-30 10:44:46 -0400222/* Check for region overlaps inside the VHDX image */
223static int vhdx_region_check(BDRVVHDXState *s, uint64_t start, uint64_t length)
224{
225 int ret = 0;
226 uint64_t end;
227 VHDXRegionEntry *r;
228
229 end = start + length;
230 QLIST_FOREACH(r, &s->regions, entries) {
231 if (!((start >= r->end) || (end <= r->start))) {
232 ret = -EINVAL;
233 goto exit;
234 }
235 }
236
237exit:
238 return ret;
239}
240
241/* Register a region for future checks */
242static void vhdx_region_register(BDRVVHDXState *s,
243 uint64_t start, uint64_t length)
244{
245 VHDXRegionEntry *r;
246
247 r = g_malloc0(sizeof(*r));
248
249 r->start = start;
250 r->end = start + length;
251
252 QLIST_INSERT_HEAD(&s->regions, r, entries);
253}
254
255/* Free all registered regions */
256static void vhdx_region_unregister_all(BDRVVHDXState *s)
257{
258 VHDXRegionEntry *r, *r_next;
259
260 QLIST_FOREACH_SAFE(r, &s->regions, entries, r_next) {
261 QLIST_REMOVE(r, entries);
262 g_free(r);
263 }
264}
265
Jeff Cody1e74a972013-10-30 10:44:51 -0400266static void vhdx_set_shift_bits(BDRVVHDXState *s)
267{
Max Reitz04a36152015-12-18 16:04:26 +0100268 s->logical_sector_size_bits = ctz32(s->logical_sector_size);
269 s->sectors_per_block_bits = ctz32(s->sectors_per_block);
270 s->chunk_ratio_bits = ctz64(s->chunk_ratio);
271 s->block_size_bits = ctz32(s->block_size);
Jeff Cody1e74a972013-10-30 10:44:51 -0400272}
273
Jeff Cody4f18b782013-10-30 10:44:39 -0400274/*
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400275 * Per the MS VHDX Specification, for every VHDX file:
276 * - The header section is fixed size - 1 MB
277 * - The header section is always the first "object"
278 * - The first 64KB of the header is the File Identifier
279 * - The first uint64 (8 bytes) is the VHDX Signature ("vhdxfile")
280 * - The following 512 bytes constitute a UTF-16 string identifiying the
281 * software that created the file, and is optional and diagnostic only.
282 *
283 * Therefore, we probe by looking for the vhdxfile signature "vhdxfile"
284 */
285static int vhdx_probe(const uint8_t *buf, int buf_size, const char *filename)
286{
287 if (buf_size >= 8 && !memcmp(buf, "vhdxfile", 8)) {
288 return 100;
289 }
290 return 0;
291}
292
Jeff Cody1e74a972013-10-30 10:44:51 -0400293/*
294 * Writes the header to the specified offset.
295 *
296 * This will optionally read in buffer data from disk (otherwise zero-fill),
297 * and then update the header checksum. Header is converted to proper
298 * endianness before being written to the specified file offset
299 */
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200300static int vhdx_write_header(BdrvChild *file, VHDXHeader *hdr,
Jeff Cody1e74a972013-10-30 10:44:51 -0400301 uint64_t offset, bool read)
302{
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200303 BlockDriverState *bs_file = file->bs;
Jeff Cody1e74a972013-10-30 10:44:51 -0400304 uint8_t *buffer = NULL;
305 int ret;
Jeff Cody4f75b522014-08-06 15:54:58 -0400306 VHDXHeader *header_le;
Jeff Cody1e74a972013-10-30 10:44:51 -0400307
308 assert(bs_file != NULL);
309 assert(hdr != NULL);
310
311 /* the header checksum is not over just the packed size of VHDXHeader,
312 * but rather over the entire 'reserved' range for the header, which is
313 * 4KB (VHDX_HEADER_SIZE). */
314
315 buffer = qemu_blockalign(bs_file, VHDX_HEADER_SIZE);
316 if (read) {
317 /* if true, we can't assume the extra reserved bytes are 0 */
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200318 ret = bdrv_pread(file, offset, buffer, VHDX_HEADER_SIZE);
Jeff Cody1e74a972013-10-30 10:44:51 -0400319 if (ret < 0) {
320 goto exit;
321 }
322 } else {
323 memset(buffer, 0, VHDX_HEADER_SIZE);
324 }
325
326 /* overwrite the actual VHDXHeader portion */
Jeff Cody4f75b522014-08-06 15:54:58 -0400327 header_le = (VHDXHeader *)buffer;
328 memcpy(header_le, hdr, sizeof(VHDXHeader));
329 vhdx_header_le_export(hdr, header_le);
330 vhdx_update_checksum(buffer, VHDX_HEADER_SIZE,
331 offsetof(VHDXHeader, checksum));
Kevin Wolfd9ca2ea2016-06-20 20:09:15 +0200332 ret = bdrv_pwrite_sync(file, offset, header_le, sizeof(VHDXHeader));
Jeff Cody1e74a972013-10-30 10:44:51 -0400333
334exit:
335 qemu_vfree(buffer);
336 return ret;
337}
338
Jeff Cody4f18b782013-10-30 10:44:39 -0400339/* Update the VHDX headers
340 *
341 * This follows the VHDX spec procedures for header updates.
342 *
343 * - non-current header is updated with largest sequence number
344 */
345static int vhdx_update_header(BlockDriverState *bs, BDRVVHDXState *s,
Jeff Codyc3906c52013-10-30 10:44:43 -0400346 bool generate_data_write_guid, MSGUID *log_guid)
Jeff Cody4f18b782013-10-30 10:44:39 -0400347{
348 int ret = 0;
349 int hdr_idx = 0;
350 uint64_t header_offset = VHDX_HEADER1_OFFSET;
351
352 VHDXHeader *active_header;
353 VHDXHeader *inactive_header;
Jeff Cody4f18b782013-10-30 10:44:39 -0400354
355 /* operate on the non-current header */
356 if (s->curr_header == 0) {
357 hdr_idx = 1;
358 header_offset = VHDX_HEADER2_OFFSET;
359 }
360
361 active_header = s->headers[s->curr_header];
362 inactive_header = s->headers[hdr_idx];
363
364 inactive_header->sequence_number = active_header->sequence_number + 1;
365
366 /* a new file guid must be generated before any file write, including
367 * headers */
368 inactive_header->file_write_guid = s->session_guid;
369
370 /* a new data guid only needs to be generated before any guest-visible
371 * writes (i.e. something observable via virtual disk read) */
372 if (generate_data_write_guid) {
373 vhdx_guid_generate(&inactive_header->data_write_guid);
374 }
375
Jeff Codyc3906c52013-10-30 10:44:43 -0400376 /* update the log guid if present */
377 if (log_guid) {
378 inactive_header->log_guid = *log_guid;
379 }
380
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200381 ret = vhdx_write_header(bs->file, inactive_header, header_offset, true);
Jeff Cody4f18b782013-10-30 10:44:39 -0400382 if (ret < 0) {
383 goto exit;
384 }
385 s->curr_header = hdr_idx;
386
387exit:
Jeff Cody4f18b782013-10-30 10:44:39 -0400388 return ret;
389}
390
391/*
392 * The VHDX spec calls for header updates to be performed twice, so that both
393 * the current and non-current header have valid info
394 */
Jeff Codyc3906c52013-10-30 10:44:43 -0400395int vhdx_update_headers(BlockDriverState *bs, BDRVVHDXState *s,
396 bool generate_data_write_guid, MSGUID *log_guid)
Jeff Cody4f18b782013-10-30 10:44:39 -0400397{
398 int ret;
399
Jeff Codyc3906c52013-10-30 10:44:43 -0400400 ret = vhdx_update_header(bs, s, generate_data_write_guid, log_guid);
Jeff Cody4f18b782013-10-30 10:44:39 -0400401 if (ret < 0) {
402 return ret;
403 }
Jeff Codyc3906c52013-10-30 10:44:43 -0400404 ret = vhdx_update_header(bs, s, generate_data_write_guid, log_guid);
Jeff Cody4f18b782013-10-30 10:44:39 -0400405 return ret;
406}
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400407
408/* opens the specified header block from the VHDX file header section */
Paolo Bonzini6890aad2014-02-17 14:44:00 +0100409static void vhdx_parse_header(BlockDriverState *bs, BDRVVHDXState *s,
410 Error **errp)
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400411{
Paolo Bonzini6890aad2014-02-17 14:44:00 +0100412 int ret;
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400413 VHDXHeader *header1;
414 VHDXHeader *header2;
415 bool h1_valid = false;
416 bool h2_valid = false;
417 uint64_t h1_seq = 0;
418 uint64_t h2_seq = 0;
419 uint8_t *buffer;
420
Jeff Cody6e9d2902013-10-30 10:44:38 -0400421 /* header1 & header2 are freed in vhdx_close() */
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400422 header1 = qemu_blockalign(bs, sizeof(VHDXHeader));
423 header2 = qemu_blockalign(bs, sizeof(VHDXHeader));
424
425 buffer = qemu_blockalign(bs, VHDX_HEADER_SIZE);
426
427 s->headers[0] = header1;
428 s->headers[1] = header2;
429
430 /* We have to read the whole VHDX_HEADER_SIZE instead of
431 * sizeof(VHDXHeader), because the checksum is over the whole
432 * region */
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200433 ret = bdrv_pread(bs->file, VHDX_HEADER1_OFFSET, buffer,
Kevin Wolf9a4f4c32015-06-16 14:19:22 +0200434 VHDX_HEADER_SIZE);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400435 if (ret < 0) {
436 goto fail;
437 }
438 /* copy over just the relevant portion that we need */
439 memcpy(header1, buffer, sizeof(VHDXHeader));
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400440
Jeff Cody4f75b522014-08-06 15:54:58 -0400441 if (vhdx_checksum_is_valid(buffer, VHDX_HEADER_SIZE, 4)) {
442 vhdx_header_le_import(header1);
443 if (header1->signature == VHDX_HEADER_SIGNATURE &&
444 header1->version == 1) {
445 h1_seq = header1->sequence_number;
446 h1_valid = true;
447 }
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400448 }
449
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200450 ret = bdrv_pread(bs->file, VHDX_HEADER2_OFFSET, buffer,
Kevin Wolf9a4f4c32015-06-16 14:19:22 +0200451 VHDX_HEADER_SIZE);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400452 if (ret < 0) {
453 goto fail;
454 }
455 /* copy over just the relevant portion that we need */
456 memcpy(header2, buffer, sizeof(VHDXHeader));
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400457
Jeff Cody4f75b522014-08-06 15:54:58 -0400458 if (vhdx_checksum_is_valid(buffer, VHDX_HEADER_SIZE, 4)) {
459 vhdx_header_le_import(header2);
460 if (header2->signature == VHDX_HEADER_SIGNATURE &&
461 header2->version == 1) {
462 h2_seq = header2->sequence_number;
463 h2_valid = true;
464 }
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400465 }
466
467 /* If there is only 1 valid header (or no valid headers), we
468 * don't care what the sequence numbers are */
469 if (h1_valid && !h2_valid) {
470 s->curr_header = 0;
471 } else if (!h1_valid && h2_valid) {
472 s->curr_header = 1;
473 } else if (!h1_valid && !h2_valid) {
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400474 goto fail;
475 } else {
476 /* If both headers are valid, then we choose the active one by the
477 * highest sequence number. If the sequence numbers are equal, that is
478 * invalid */
479 if (h1_seq > h2_seq) {
480 s->curr_header = 0;
481 } else if (h2_seq > h1_seq) {
482 s->curr_header = 1;
483 } else {
Jeff Cody69060462014-05-13 10:00:52 -0400484 /* The Microsoft Disk2VHD tool will create 2 identical
485 * headers, with identical sequence numbers. If the headers are
486 * identical, don't consider the file corrupt */
487 if (!memcmp(header1, header2, sizeof(VHDXHeader))) {
488 s->curr_header = 0;
489 } else {
490 goto fail;
491 }
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400492 }
493 }
494
Jeff Cody1a848fd2013-10-30 10:44:46 -0400495 vhdx_region_register(s, s->headers[s->curr_header]->log_offset,
496 s->headers[s->curr_header]->log_length);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400497 goto exit;
498
499fail:
Paolo Bonzini6890aad2014-02-17 14:44:00 +0100500 error_setg_errno(errp, -ret, "No valid VHDX header found");
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400501 qemu_vfree(header1);
502 qemu_vfree(header2);
503 s->headers[0] = NULL;
504 s->headers[1] = NULL;
505exit:
506 qemu_vfree(buffer);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400507}
508
509
510static int vhdx_open_region_tables(BlockDriverState *bs, BDRVVHDXState *s)
511{
512 int ret = 0;
513 uint8_t *buffer;
514 int offset = 0;
515 VHDXRegionTableEntry rt_entry;
516 uint32_t i;
517 bool bat_rt_found = false;
518 bool metadata_rt_found = false;
519
520 /* We have to read the whole 64KB block, because the crc32 is over the
521 * whole block */
522 buffer = qemu_blockalign(bs, VHDX_HEADER_BLOCK_SIZE);
523
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200524 ret = bdrv_pread(bs->file, VHDX_REGION_TABLE_OFFSET, buffer,
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400525 VHDX_HEADER_BLOCK_SIZE);
526 if (ret < 0) {
527 goto fail;
528 }
529 memcpy(&s->rt, buffer, sizeof(s->rt));
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400530 offset += sizeof(s->rt);
531
Jeff Cody4f75b522014-08-06 15:54:58 -0400532 if (!vhdx_checksum_is_valid(buffer, VHDX_HEADER_BLOCK_SIZE, 4)) {
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400533 ret = -EINVAL;
534 goto fail;
535 }
536
Jeff Cody4f75b522014-08-06 15:54:58 -0400537 vhdx_region_header_le_import(&s->rt);
538
539 if (s->rt.signature != VHDX_REGION_SIGNATURE) {
540 ret = -EINVAL;
541 goto fail;
542 }
543
544
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400545 /* Per spec, maximum region table entry count is 2047 */
546 if (s->rt.entry_count > 2047) {
547 ret = -EINVAL;
548 goto fail;
549 }
550
551 for (i = 0; i < s->rt.entry_count; i++) {
552 memcpy(&rt_entry, buffer + offset, sizeof(rt_entry));
553 offset += sizeof(rt_entry);
554
Jeff Codyc325ee12013-10-30 10:44:50 -0400555 vhdx_region_entry_le_import(&rt_entry);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400556
Jeff Cody1a848fd2013-10-30 10:44:46 -0400557 /* check for region overlap between these entries, and any
558 * other memory regions in the file */
559 ret = vhdx_region_check(s, rt_entry.file_offset, rt_entry.length);
560 if (ret < 0) {
561 goto fail;
562 }
563
564 vhdx_region_register(s, rt_entry.file_offset, rt_entry.length);
565
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400566 /* see if we recognize the entry */
567 if (guid_eq(rt_entry.guid, bat_guid)) {
568 /* must be unique; if we have already found it this is invalid */
569 if (bat_rt_found) {
570 ret = -EINVAL;
571 goto fail;
572 }
573 bat_rt_found = true;
574 s->bat_rt = rt_entry;
575 continue;
576 }
577
578 if (guid_eq(rt_entry.guid, metadata_guid)) {
579 /* must be unique; if we have already found it this is invalid */
580 if (metadata_rt_found) {
581 ret = -EINVAL;
582 goto fail;
583 }
584 metadata_rt_found = true;
585 s->metadata_rt = rt_entry;
586 continue;
587 }
588
589 if (rt_entry.data_bits & VHDX_REGION_ENTRY_REQUIRED) {
590 /* cannot read vhdx file - required region table entry that
591 * we do not understand. per spec, we must fail to open */
592 ret = -ENOTSUP;
593 goto fail;
594 }
595 }
Jeff Cody1a848fd2013-10-30 10:44:46 -0400596
597 if (!bat_rt_found || !metadata_rt_found) {
598 ret = -EINVAL;
599 goto fail;
600 }
601
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400602 ret = 0;
603
604fail:
605 qemu_vfree(buffer);
606 return ret;
607}
608
609
610
611/* Metadata initial parser
612 *
613 * This loads all the metadata entry fields. This may cause additional
614 * fields to be processed (e.g. parent locator, etc..).
615 *
616 * There are 5 Metadata items that are always required:
617 * - File Parameters (block size, has a parent)
618 * - Virtual Disk Size (size, in bytes, of the virtual drive)
619 * - Page 83 Data (scsi page 83 guid)
620 * - Logical Sector Size (logical sector size in bytes, either 512 or
621 * 4096. We only support 512 currently)
622 * - Physical Sector Size (512 or 4096)
623 *
624 * Also, if the File Parameters indicate this is a differencing file,
625 * we must also look for the Parent Locator metadata item.
626 */
627static int vhdx_parse_metadata(BlockDriverState *bs, BDRVVHDXState *s)
628{
629 int ret = 0;
630 uint8_t *buffer;
631 int offset = 0;
632 uint32_t i = 0;
633 VHDXMetadataTableEntry md_entry;
634
635 buffer = qemu_blockalign(bs, VHDX_METADATA_TABLE_MAX_SIZE);
636
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200637 ret = bdrv_pread(bs->file, s->metadata_rt.file_offset, buffer,
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400638 VHDX_METADATA_TABLE_MAX_SIZE);
639 if (ret < 0) {
640 goto exit;
641 }
642 memcpy(&s->metadata_hdr, buffer, sizeof(s->metadata_hdr));
643 offset += sizeof(s->metadata_hdr);
644
Jeff Codyc325ee12013-10-30 10:44:50 -0400645 vhdx_metadata_header_le_import(&s->metadata_hdr);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400646
Jeff Cody4f75b522014-08-06 15:54:58 -0400647 if (s->metadata_hdr.signature != VHDX_METADATA_SIGNATURE) {
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400648 ret = -EINVAL;
649 goto exit;
650 }
651
652 s->metadata_entries.present = 0;
653
654 if ((s->metadata_hdr.entry_count * sizeof(md_entry)) >
655 (VHDX_METADATA_TABLE_MAX_SIZE - offset)) {
656 ret = -EINVAL;
657 goto exit;
658 }
659
660 for (i = 0; i < s->metadata_hdr.entry_count; i++) {
661 memcpy(&md_entry, buffer + offset, sizeof(md_entry));
662 offset += sizeof(md_entry);
663
Jeff Codyc325ee12013-10-30 10:44:50 -0400664 vhdx_metadata_entry_le_import(&md_entry);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400665
666 if (guid_eq(md_entry.item_id, file_param_guid)) {
667 if (s->metadata_entries.present & META_FILE_PARAMETER_PRESENT) {
668 ret = -EINVAL;
669 goto exit;
670 }
671 s->metadata_entries.file_parameters_entry = md_entry;
672 s->metadata_entries.present |= META_FILE_PARAMETER_PRESENT;
673 continue;
674 }
675
676 if (guid_eq(md_entry.item_id, virtual_size_guid)) {
677 if (s->metadata_entries.present & META_VIRTUAL_DISK_SIZE_PRESENT) {
678 ret = -EINVAL;
679 goto exit;
680 }
681 s->metadata_entries.virtual_disk_size_entry = md_entry;
682 s->metadata_entries.present |= META_VIRTUAL_DISK_SIZE_PRESENT;
683 continue;
684 }
685
686 if (guid_eq(md_entry.item_id, page83_guid)) {
687 if (s->metadata_entries.present & META_PAGE_83_PRESENT) {
688 ret = -EINVAL;
689 goto exit;
690 }
691 s->metadata_entries.page83_data_entry = md_entry;
692 s->metadata_entries.present |= META_PAGE_83_PRESENT;
693 continue;
694 }
695
696 if (guid_eq(md_entry.item_id, logical_sector_guid)) {
697 if (s->metadata_entries.present &
698 META_LOGICAL_SECTOR_SIZE_PRESENT) {
699 ret = -EINVAL;
700 goto exit;
701 }
702 s->metadata_entries.logical_sector_size_entry = md_entry;
703 s->metadata_entries.present |= META_LOGICAL_SECTOR_SIZE_PRESENT;
704 continue;
705 }
706
707 if (guid_eq(md_entry.item_id, phys_sector_guid)) {
708 if (s->metadata_entries.present & META_PHYS_SECTOR_SIZE_PRESENT) {
709 ret = -EINVAL;
710 goto exit;
711 }
712 s->metadata_entries.phys_sector_size_entry = md_entry;
713 s->metadata_entries.present |= META_PHYS_SECTOR_SIZE_PRESENT;
714 continue;
715 }
716
717 if (guid_eq(md_entry.item_id, parent_locator_guid)) {
718 if (s->metadata_entries.present & META_PARENT_LOCATOR_PRESENT) {
719 ret = -EINVAL;
720 goto exit;
721 }
722 s->metadata_entries.parent_locator_entry = md_entry;
723 s->metadata_entries.present |= META_PARENT_LOCATOR_PRESENT;
724 continue;
725 }
726
727 if (md_entry.data_bits & VHDX_META_FLAGS_IS_REQUIRED) {
728 /* cannot read vhdx file - required region table entry that
729 * we do not understand. per spec, we must fail to open */
730 ret = -ENOTSUP;
731 goto exit;
732 }
733 }
734
735 if (s->metadata_entries.present != META_ALL_PRESENT) {
736 ret = -ENOTSUP;
737 goto exit;
738 }
739
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200740 ret = bdrv_pread(bs->file,
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400741 s->metadata_entries.file_parameters_entry.offset
742 + s->metadata_rt.file_offset,
743 &s->params,
744 sizeof(s->params));
745
746 if (ret < 0) {
747 goto exit;
748 }
749
750 le32_to_cpus(&s->params.block_size);
751 le32_to_cpus(&s->params.data_bits);
752
753
754 /* We now have the file parameters, so we can tell if this is a
755 * differencing file (i.e.. has_parent), is dynamic or fixed
756 * sized (leave_blocks_allocated), and the block size */
757
758 /* The parent locator required iff the file parameters has_parent set */
759 if (s->params.data_bits & VHDX_PARAMS_HAS_PARENT) {
760 if (s->metadata_entries.present & META_PARENT_LOCATOR_PRESENT) {
761 /* TODO: parse parent locator fields */
762 ret = -ENOTSUP; /* temp, until differencing files are supported */
763 goto exit;
764 } else {
765 /* if has_parent is set, but there is not parent locator present,
766 * then that is an invalid combination */
767 ret = -EINVAL;
768 goto exit;
769 }
770 }
771
772 /* determine virtual disk size, logical sector size,
773 * and phys sector size */
774
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200775 ret = bdrv_pread(bs->file,
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400776 s->metadata_entries.virtual_disk_size_entry.offset
777 + s->metadata_rt.file_offset,
778 &s->virtual_disk_size,
779 sizeof(uint64_t));
780 if (ret < 0) {
781 goto exit;
782 }
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200783 ret = bdrv_pread(bs->file,
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400784 s->metadata_entries.logical_sector_size_entry.offset
785 + s->metadata_rt.file_offset,
786 &s->logical_sector_size,
787 sizeof(uint32_t));
788 if (ret < 0) {
789 goto exit;
790 }
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200791 ret = bdrv_pread(bs->file,
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400792 s->metadata_entries.phys_sector_size_entry.offset
793 + s->metadata_rt.file_offset,
794 &s->physical_sector_size,
795 sizeof(uint32_t));
796 if (ret < 0) {
797 goto exit;
798 }
799
800 le64_to_cpus(&s->virtual_disk_size);
801 le32_to_cpus(&s->logical_sector_size);
802 le32_to_cpus(&s->physical_sector_size);
803
Jeff Cody1d7678d2014-03-26 13:05:39 +0100804 if (s->params.block_size < VHDX_BLOCK_SIZE_MIN ||
805 s->params.block_size > VHDX_BLOCK_SIZE_MAX) {
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400806 ret = -EINVAL;
807 goto exit;
808 }
809
Jeff Cody1d7678d2014-03-26 13:05:39 +0100810 /* only 2 supported sector sizes */
811 if (s->logical_sector_size != 512 && s->logical_sector_size != 4096) {
812 ret = -EINVAL;
813 goto exit;
814 }
815
816 /* Both block_size and sector_size are guaranteed powers of 2, below.
817 Due to range checks above, s->sectors_per_block can never be < 256 */
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400818 s->sectors_per_block = s->params.block_size / s->logical_sector_size;
819 s->chunk_ratio = (VHDX_MAX_SECTORS_PER_BLOCK) *
820 (uint64_t)s->logical_sector_size /
821 (uint64_t)s->params.block_size;
822
823 /* These values are ones we will want to use for division / multiplication
824 * later on, and they are all guaranteed (per the spec) to be powers of 2,
825 * so we can take advantage of that for shift operations during
826 * reads/writes */
827 if (s->logical_sector_size & (s->logical_sector_size - 1)) {
828 ret = -EINVAL;
829 goto exit;
830 }
831 if (s->sectors_per_block & (s->sectors_per_block - 1)) {
832 ret = -EINVAL;
833 goto exit;
834 }
835 if (s->chunk_ratio & (s->chunk_ratio - 1)) {
836 ret = -EINVAL;
837 goto exit;
838 }
839 s->block_size = s->params.block_size;
840 if (s->block_size & (s->block_size - 1)) {
841 ret = -EINVAL;
842 goto exit;
843 }
844
Jeff Cody1e74a972013-10-30 10:44:51 -0400845 vhdx_set_shift_bits(s);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400846
847 ret = 0;
848
849exit:
850 qemu_vfree(buffer);
851 return ret;
852}
853
Jeff Cody1e74a972013-10-30 10:44:51 -0400854/*
855 * Calculate the number of BAT entries, including sector
856 * bitmap entries.
857 */
858static void vhdx_calc_bat_entries(BDRVVHDXState *s)
859{
860 uint32_t data_blocks_cnt, bitmap_blocks_cnt;
861
Max Reitz939901d2015-12-18 16:04:25 +0100862 data_blocks_cnt = DIV_ROUND_UP(s->virtual_disk_size, s->block_size);
863 bitmap_blocks_cnt = DIV_ROUND_UP(data_blocks_cnt, s->chunk_ratio);
Jeff Cody1e74a972013-10-30 10:44:51 -0400864
865 if (s->parent_entries) {
866 s->bat_entries = bitmap_blocks_cnt * (s->chunk_ratio + 1);
867 } else {
868 s->bat_entries = data_blocks_cnt +
869 ((data_blocks_cnt - 1) >> s->chunk_ratio_bits);
870 }
871
872}
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400873
Jeff Codyc46415a2013-10-30 10:44:44 -0400874static void vhdx_close(BlockDriverState *bs)
875{
876 BDRVVHDXState *s = bs->opaque;
877 qemu_vfree(s->headers[0]);
Jeff Cody0a43a1b2013-10-30 10:44:45 -0400878 s->headers[0] = NULL;
Jeff Codyc46415a2013-10-30 10:44:44 -0400879 qemu_vfree(s->headers[1]);
Jeff Cody0a43a1b2013-10-30 10:44:45 -0400880 s->headers[1] = NULL;
Jeff Codyc46415a2013-10-30 10:44:44 -0400881 qemu_vfree(s->bat);
Jeff Cody0a43a1b2013-10-30 10:44:45 -0400882 s->bat = NULL;
Jeff Codyc46415a2013-10-30 10:44:44 -0400883 qemu_vfree(s->parent_entries);
Jeff Cody0a43a1b2013-10-30 10:44:45 -0400884 s->parent_entries = NULL;
Jeff Codyc46415a2013-10-30 10:44:44 -0400885 migrate_del_blocker(s->migration_blocker);
886 error_free(s->migration_blocker);
Jeff Cody0a43a1b2013-10-30 10:44:45 -0400887 qemu_vfree(s->log.hdr);
888 s->log.hdr = NULL;
Jeff Cody1a848fd2013-10-30 10:44:46 -0400889 vhdx_region_unregister_all(s);
Jeff Codyc46415a2013-10-30 10:44:44 -0400890}
891
Max Reitz015a1032013-09-05 14:22:29 +0200892static int vhdx_open(BlockDriverState *bs, QDict *options, int flags,
893 Error **errp)
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400894{
895 BDRVVHDXState *s = bs->opaque;
896 int ret = 0;
897 uint32_t i;
898 uint64_t signature;
Paolo Bonzini6890aad2014-02-17 14:44:00 +0100899 Error *local_err = NULL;
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400900
Kevin Wolf4e4bf5c2016-12-16 18:52:37 +0100901 bs->file = bdrv_open_child(NULL, options, "file", bs, &child_file,
902 false, errp);
903 if (!bs->file) {
904 return -EINVAL;
905 }
906
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400907 s->bat = NULL;
Jeff Codyc3906c52013-10-30 10:44:43 -0400908 s->first_visible_write = true;
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400909
910 qemu_co_mutex_init(&s->lock);
Jeff Cody1a848fd2013-10-30 10:44:46 -0400911 QLIST_INIT(&s->regions);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400912
913 /* validate the file signature */
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200914 ret = bdrv_pread(bs->file, 0, &signature, sizeof(uint64_t));
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400915 if (ret < 0) {
916 goto fail;
917 }
918 if (memcmp(&signature, "vhdxfile", 8)) {
919 ret = -EINVAL;
920 goto fail;
921 }
922
Jeff Cody4f18b782013-10-30 10:44:39 -0400923 /* This is used for any header updates, for the file_write_guid.
924 * The spec dictates that a new value should be used for the first
925 * header update */
926 vhdx_guid_generate(&s->session_guid);
927
Paolo Bonzini6890aad2014-02-17 14:44:00 +0100928 vhdx_parse_header(bs, s, &local_err);
929 if (local_err != NULL) {
930 error_propagate(errp, local_err);
931 ret = -EINVAL;
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400932 goto fail;
933 }
934
Jeff Cody7e30e6a2013-12-17 05:33:37 -0500935 ret = vhdx_parse_log(bs, s, &s->log_replayed_on_open, errp);
Jeff Cody0a43a1b2013-10-30 10:44:45 -0400936 if (ret < 0) {
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400937 goto fail;
938 }
939
940 ret = vhdx_open_region_tables(bs, s);
Jeff Cody0a43a1b2013-10-30 10:44:45 -0400941 if (ret < 0) {
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400942 goto fail;
943 }
944
945 ret = vhdx_parse_metadata(bs, s);
Jeff Cody0a43a1b2013-10-30 10:44:45 -0400946 if (ret < 0) {
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400947 goto fail;
948 }
Jeff Cody0a43a1b2013-10-30 10:44:45 -0400949
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400950 s->block_size = s->params.block_size;
951
952 /* the VHDX spec dictates that virtual_disk_size is always a multiple of
953 * logical_sector_size */
954 bs->total_sectors = s->virtual_disk_size >> s->logical_sector_size_bits;
955
Jeff Cody1e74a972013-10-30 10:44:51 -0400956 vhdx_calc_bat_entries(s);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400957
958 s->bat_offset = s->bat_rt.file_offset;
959
960 if (s->bat_entries > s->bat_rt.length / sizeof(VHDXBatEntry)) {
961 /* BAT allocation is not large enough for all entries */
962 ret = -EINVAL;
963 goto fail;
964 }
965
Jeff Cody6e9d2902013-10-30 10:44:38 -0400966 /* s->bat is freed in vhdx_close() */
Kevin Wolf9a4f4c32015-06-16 14:19:22 +0200967 s->bat = qemu_try_blockalign(bs->file->bs, s->bat_rt.length);
Kevin Wolfa67e1282014-05-20 13:55:50 +0200968 if (s->bat == NULL) {
969 ret = -ENOMEM;
970 goto fail;
971 }
972
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +0200973 ret = bdrv_pread(bs->file, s->bat_offset, s->bat, s->bat_rt.length);
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400974 if (ret < 0) {
975 goto fail;
976 }
977
Jeff Cody1a848fd2013-10-30 10:44:46 -0400978 uint64_t payblocks = s->chunk_ratio;
979 /* endian convert, and verify populated BAT field file offsets against
980 * region table and log entries */
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400981 for (i = 0; i < s->bat_entries; i++) {
982 le64_to_cpus(&s->bat[i]);
Jeff Cody1a848fd2013-10-30 10:44:46 -0400983 if (payblocks--) {
984 /* payload bat entries */
985 if ((s->bat[i] & VHDX_BAT_STATE_BIT_MASK) ==
Jeff Codyd92aa882013-10-30 10:44:48 -0400986 PAYLOAD_BLOCK_FULLY_PRESENT) {
Jeff Cody1a848fd2013-10-30 10:44:46 -0400987 ret = vhdx_region_check(s, s->bat[i] & VHDX_BAT_FILE_OFF_MASK,
988 s->block_size);
989 if (ret < 0) {
990 goto fail;
991 }
992 }
993 } else {
994 payblocks = s->chunk_ratio;
995 /* Once differencing files are supported, verify sector bitmap
996 * blocks here */
997 }
Jeff Codye8d4e5f2013-04-29 14:48:18 -0400998 }
999
Ashijeet Acharyafe44dc92017-01-16 17:01:53 +05301000 /* Disable migration when VHDX images are used */
1001 error_setg(&s->migration_blocker, "The vhdx format used by node '%s' "
1002 "does not support live migration",
1003 bdrv_get_device_or_node_name(bs));
1004 ret = migrate_add_blocker(s->migration_blocker, &local_err);
1005 if (local_err) {
1006 error_propagate(errp, local_err);
1007 error_free(s->migration_blocker);
1008 goto fail;
1009 }
1010
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001011 if (flags & BDRV_O_RDWR) {
Jeff Codyc3906c52013-10-30 10:44:43 -04001012 ret = vhdx_update_headers(bs, s, false, NULL);
Jeff Cody4f18b782013-10-30 10:44:39 -04001013 if (ret < 0) {
1014 goto fail;
1015 }
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001016 }
1017
Jeff Codyd92aa882013-10-30 10:44:48 -04001018 /* TODO: differencing files */
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001019
1020 return 0;
1021fail:
Jeff Cody0a43a1b2013-10-30 10:44:45 -04001022 vhdx_close(bs);
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001023 return ret;
1024}
1025
1026static int vhdx_reopen_prepare(BDRVReopenState *state,
1027 BlockReopenQueue *queue, Error **errp)
1028{
1029 return 0;
1030}
1031
1032
Jeff Cody059e2fb2013-04-29 14:48:19 -04001033/*
1034 * Perform sector to block offset translations, to get various
1035 * sector and file offsets into the image. See VHDXSectorInfo
1036 */
1037static void vhdx_block_translate(BDRVVHDXState *s, int64_t sector_num,
1038 int nb_sectors, VHDXSectorInfo *sinfo)
1039{
1040 uint32_t block_offset;
1041
1042 sinfo->bat_idx = sector_num >> s->sectors_per_block_bits;
1043 /* effectively a modulo - this gives us the offset into the block
1044 * (in sector sizes) for our sector number */
1045 block_offset = sector_num - (sinfo->bat_idx << s->sectors_per_block_bits);
1046 /* the chunk ratio gives us the interleaving of the sector
1047 * bitmaps, so we need to advance our page block index by the
1048 * sector bitmaps entry number */
1049 sinfo->bat_idx += sinfo->bat_idx >> s->chunk_ratio_bits;
1050
1051 /* the number of sectors we can read/write in this cycle */
1052 sinfo->sectors_avail = s->sectors_per_block - block_offset;
1053
1054 sinfo->bytes_left = sinfo->sectors_avail << s->logical_sector_size_bits;
1055
1056 if (sinfo->sectors_avail > nb_sectors) {
1057 sinfo->sectors_avail = nb_sectors;
1058 }
1059
1060 sinfo->bytes_avail = sinfo->sectors_avail << s->logical_sector_size_bits;
1061
Jeff Cody0b7da092013-10-30 10:44:49 -04001062 sinfo->file_offset = s->bat[sinfo->bat_idx] & VHDX_BAT_FILE_OFF_MASK;
Jeff Cody059e2fb2013-04-29 14:48:19 -04001063
1064 sinfo->block_offset = block_offset << s->logical_sector_size_bits;
1065
1066 /* The file offset must be past the header section, so must be > 0 */
1067 if (sinfo->file_offset == 0) {
1068 return;
1069 }
1070
1071 /* block offset is the offset in vhdx logical sectors, in
1072 * the payload data block. Convert that to a byte offset
1073 * in the block, and add in the payload data block offset
1074 * in the file, in bytes, to get the final read address */
1075
Jeff Cody059e2fb2013-04-29 14:48:19 -04001076 sinfo->file_offset += sinfo->block_offset;
1077}
1078
1079
Paolo Bonzini97b00e22013-11-22 13:39:49 +01001080static int vhdx_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1081{
1082 BDRVVHDXState *s = bs->opaque;
1083
1084 bdi->cluster_size = s->block_size;
1085
Paolo Bonzini95de6d72013-11-22 13:39:50 +01001086 bdi->unallocated_blocks_are_zero =
1087 (s->params.data_bits & VHDX_PARAMS_HAS_PARENT) == 0;
1088
Paolo Bonzini97b00e22013-11-22 13:39:49 +01001089 return 0;
1090}
1091
Jeff Cody059e2fb2013-04-29 14:48:19 -04001092
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001093static coroutine_fn int vhdx_co_readv(BlockDriverState *bs, int64_t sector_num,
1094 int nb_sectors, QEMUIOVector *qiov)
1095{
Jeff Cody059e2fb2013-04-29 14:48:19 -04001096 BDRVVHDXState *s = bs->opaque;
1097 int ret = 0;
1098 VHDXSectorInfo sinfo;
1099 uint64_t bytes_done = 0;
1100 QEMUIOVector hd_qiov;
1101
1102 qemu_iovec_init(&hd_qiov, qiov->niov);
1103
1104 qemu_co_mutex_lock(&s->lock);
1105
1106 while (nb_sectors > 0) {
1107 /* We are a differencing file, so we need to inspect the sector bitmap
1108 * to see if we have the data or not */
1109 if (s->params.data_bits & VHDX_PARAMS_HAS_PARENT) {
1110 /* not supported yet */
1111 ret = -ENOTSUP;
1112 goto exit;
1113 } else {
1114 vhdx_block_translate(s, sector_num, nb_sectors, &sinfo);
1115
1116 qemu_iovec_reset(&hd_qiov);
1117 qemu_iovec_concat(&hd_qiov, qiov, bytes_done, sinfo.bytes_avail);
1118
1119 /* check the payload block state */
1120 switch (s->bat[sinfo.bat_idx] & VHDX_BAT_STATE_BIT_MASK) {
1121 case PAYLOAD_BLOCK_NOT_PRESENT: /* fall through */
Jeff Cody0571df42014-12-08 01:07:42 -05001122 case PAYLOAD_BLOCK_UNDEFINED:
1123 case PAYLOAD_BLOCK_UNMAPPED:
Jeff Codya9d1e9d2014-12-08 01:07:43 -05001124 case PAYLOAD_BLOCK_UNMAPPED_v095:
Jeff Cody059e2fb2013-04-29 14:48:19 -04001125 case PAYLOAD_BLOCK_ZERO:
1126 /* return zero */
1127 qemu_iovec_memset(&hd_qiov, 0, 0, sinfo.bytes_avail);
1128 break;
Jeff Codyd92aa882013-10-30 10:44:48 -04001129 case PAYLOAD_BLOCK_FULLY_PRESENT:
Jeff Cody059e2fb2013-04-29 14:48:19 -04001130 qemu_co_mutex_unlock(&s->lock);
Kevin Wolf28b04a82016-05-24 17:21:22 +02001131 ret = bdrv_co_readv(bs->file,
Jeff Cody059e2fb2013-04-29 14:48:19 -04001132 sinfo.file_offset >> BDRV_SECTOR_BITS,
1133 sinfo.sectors_avail, &hd_qiov);
1134 qemu_co_mutex_lock(&s->lock);
1135 if (ret < 0) {
1136 goto exit;
1137 }
1138 break;
1139 case PAYLOAD_BLOCK_PARTIALLY_PRESENT:
1140 /* we don't yet support difference files, fall through
1141 * to error */
1142 default:
1143 ret = -EIO;
1144 goto exit;
1145 break;
1146 }
1147 nb_sectors -= sinfo.sectors_avail;
1148 sector_num += sinfo.sectors_avail;
1149 bytes_done += sinfo.bytes_avail;
1150 }
1151 }
1152 ret = 0;
1153exit:
1154 qemu_co_mutex_unlock(&s->lock);
1155 qemu_iovec_destroy(&hd_qiov);
1156 return ret;
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001157}
1158
Jeff Codyd92aa882013-10-30 10:44:48 -04001159/*
1160 * Allocate a new payload block at the end of the file.
1161 *
1162 * Allocation will happen at 1MB alignment inside the file
1163 *
1164 * Returns the file offset start of the new payload block
1165 */
1166static int vhdx_allocate_block(BlockDriverState *bs, BDRVVHDXState *s,
1167 uint64_t *new_offset)
1168{
Kevin Wolf9a4f4c32015-06-16 14:19:22 +02001169 *new_offset = bdrv_getlength(bs->file->bs);
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001170
Jeff Codyd92aa882013-10-30 10:44:48 -04001171 /* per the spec, the address for a block is in units of 1MB */
1172 *new_offset = ROUND_UP(*new_offset, 1024 * 1024);
1173
Kevin Wolf52cdbc52017-02-16 18:39:03 +01001174 return bdrv_truncate(bs->file, *new_offset + s->block_size);
Jeff Codyd92aa882013-10-30 10:44:48 -04001175}
1176
1177/*
1178 * Update the BAT table entry with the new file offset, and the new entry
1179 * state */
1180static void vhdx_update_bat_table_entry(BlockDriverState *bs, BDRVVHDXState *s,
1181 VHDXSectorInfo *sinfo,
1182 uint64_t *bat_entry_le,
1183 uint64_t *bat_offset, int state)
1184{
1185 /* The BAT entry is a uint64, with 44 bits for the file offset in units of
1186 * 1MB, and 3 bits for the block state. */
Jeff Codycdf96342015-01-20 16:01:43 -05001187 if ((state == PAYLOAD_BLOCK_ZERO) ||
1188 (state == PAYLOAD_BLOCK_UNDEFINED) ||
1189 (state == PAYLOAD_BLOCK_NOT_PRESENT) ||
1190 (state == PAYLOAD_BLOCK_UNMAPPED)) {
1191 s->bat[sinfo->bat_idx] = 0; /* For PAYLOAD_BLOCK_ZERO, the
1192 FileOffsetMB field is denoted as
1193 'reserved' in the v1.0 spec. If it is
1194 non-zero, MS Hyper-V will fail to read
1195 the disk image */
1196 } else {
1197 s->bat[sinfo->bat_idx] = sinfo->file_offset;
1198 }
Jeff Codyd92aa882013-10-30 10:44:48 -04001199
1200 s->bat[sinfo->bat_idx] |= state & VHDX_BAT_STATE_BIT_MASK;
1201
1202 *bat_entry_le = cpu_to_le64(s->bat[sinfo->bat_idx]);
1203 *bat_offset = s->bat_offset + sinfo->bat_idx * sizeof(VHDXBatEntry);
1204
1205}
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001206
Jeff Codyc3906c52013-10-30 10:44:43 -04001207/* Per the spec, on the first write of guest-visible data to the file the
1208 * data write guid must be updated in the header */
1209int vhdx_user_visible_write(BlockDriverState *bs, BDRVVHDXState *s)
1210{
1211 int ret = 0;
1212 if (s->first_visible_write) {
1213 s->first_visible_write = false;
1214 ret = vhdx_update_headers(bs, s, true, NULL);
1215 }
1216 return ret;
1217}
1218
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001219static coroutine_fn int vhdx_co_writev(BlockDriverState *bs, int64_t sector_num,
1220 int nb_sectors, QEMUIOVector *qiov)
1221{
Jeff Codyd92aa882013-10-30 10:44:48 -04001222 int ret = -ENOTSUP;
1223 BDRVVHDXState *s = bs->opaque;
1224 VHDXSectorInfo sinfo;
1225 uint64_t bytes_done = 0;
1226 uint64_t bat_entry = 0;
1227 uint64_t bat_entry_offset = 0;
1228 QEMUIOVector hd_qiov;
1229 struct iovec iov1 = { 0 };
1230 struct iovec iov2 = { 0 };
1231 int sectors_to_write;
1232 int bat_state;
1233 uint64_t bat_prior_offset = 0;
1234 bool bat_update = false;
1235
1236 qemu_iovec_init(&hd_qiov, qiov->niov);
1237
1238 qemu_co_mutex_lock(&s->lock);
1239
1240 ret = vhdx_user_visible_write(bs, s);
1241 if (ret < 0) {
1242 goto exit;
1243 }
1244
1245 while (nb_sectors > 0) {
1246 bool use_zero_buffers = false;
1247 bat_update = false;
1248 if (s->params.data_bits & VHDX_PARAMS_HAS_PARENT) {
1249 /* not supported yet */
1250 ret = -ENOTSUP;
1251 goto exit;
1252 } else {
1253 vhdx_block_translate(s, sector_num, nb_sectors, &sinfo);
1254 sectors_to_write = sinfo.sectors_avail;
1255
1256 qemu_iovec_reset(&hd_qiov);
1257 /* check the payload block state */
1258 bat_state = s->bat[sinfo.bat_idx] & VHDX_BAT_STATE_BIT_MASK;
1259 switch (bat_state) {
1260 case PAYLOAD_BLOCK_ZERO:
1261 /* in this case, we need to preserve zero writes for
1262 * data that is not part of this write, so we must pad
1263 * the rest of the buffer to zeroes */
1264
1265 /* if we are on a posix system with ftruncate() that extends
1266 * a file, then it is zero-filled for us. On Win32, the raw
1267 * layer uses SetFilePointer and SetFileEnd, which does not
1268 * zero fill AFAIK */
1269
1270 /* Queue another write of zero buffers if the underlying file
1271 * does not zero-fill on file extension */
1272
Kevin Wolf9a4f4c32015-06-16 14:19:22 +02001273 if (bdrv_has_zero_init(bs->file->bs) == 0) {
Jeff Codyd92aa882013-10-30 10:44:48 -04001274 use_zero_buffers = true;
1275
1276 /* zero fill the front, if any */
1277 if (sinfo.block_offset) {
1278 iov1.iov_len = sinfo.block_offset;
1279 iov1.iov_base = qemu_blockalign(bs, iov1.iov_len);
1280 memset(iov1.iov_base, 0, iov1.iov_len);
1281 qemu_iovec_concat_iov(&hd_qiov, &iov1, 1, 0,
Kevin Wolfd1a126c2015-04-14 16:36:16 +02001282 iov1.iov_len);
Jeff Codyd92aa882013-10-30 10:44:48 -04001283 sectors_to_write += iov1.iov_len >> BDRV_SECTOR_BITS;
1284 }
1285
1286 /* our actual data */
1287 qemu_iovec_concat(&hd_qiov, qiov, bytes_done,
1288 sinfo.bytes_avail);
1289
1290 /* zero fill the back, if any */
1291 if ((sinfo.bytes_avail - sinfo.block_offset) <
1292 s->block_size) {
1293 iov2.iov_len = s->block_size -
1294 (sinfo.bytes_avail + sinfo.block_offset);
1295 iov2.iov_base = qemu_blockalign(bs, iov2.iov_len);
1296 memset(iov2.iov_base, 0, iov2.iov_len);
1297 qemu_iovec_concat_iov(&hd_qiov, &iov2, 1, 0,
Kevin Wolfd1a126c2015-04-14 16:36:16 +02001298 iov2.iov_len);
Jeff Codyd92aa882013-10-30 10:44:48 -04001299 sectors_to_write += iov2.iov_len >> BDRV_SECTOR_BITS;
1300 }
1301 }
Jeff Codyd92aa882013-10-30 10:44:48 -04001302 /* fall through */
1303 case PAYLOAD_BLOCK_NOT_PRESENT: /* fall through */
Jeff Cody0571df42014-12-08 01:07:42 -05001304 case PAYLOAD_BLOCK_UNMAPPED:
Jeff Codya9d1e9d2014-12-08 01:07:43 -05001305 case PAYLOAD_BLOCK_UNMAPPED_v095:
Jeff Cody0571df42014-12-08 01:07:42 -05001306 case PAYLOAD_BLOCK_UNDEFINED:
Jeff Codyd92aa882013-10-30 10:44:48 -04001307 bat_prior_offset = sinfo.file_offset;
1308 ret = vhdx_allocate_block(bs, s, &sinfo.file_offset);
1309 if (ret < 0) {
1310 goto exit;
1311 }
1312 /* once we support differencing files, this may also be
1313 * partially present */
1314 /* update block state to the newly specified state */
1315 vhdx_update_bat_table_entry(bs, s, &sinfo, &bat_entry,
1316 &bat_entry_offset,
1317 PAYLOAD_BLOCK_FULLY_PRESENT);
1318 bat_update = true;
1319 /* since we just allocated a block, file_offset is the
1320 * beginning of the payload block. It needs to be the
1321 * write address, which includes the offset into the block */
1322 if (!use_zero_buffers) {
1323 sinfo.file_offset += sinfo.block_offset;
1324 }
1325 /* fall through */
1326 case PAYLOAD_BLOCK_FULLY_PRESENT:
1327 /* if the file offset address is in the header zone,
1328 * there is a problem */
1329 if (sinfo.file_offset < (1024 * 1024)) {
1330 ret = -EFAULT;
1331 goto error_bat_restore;
1332 }
1333
1334 if (!use_zero_buffers) {
1335 qemu_iovec_concat(&hd_qiov, qiov, bytes_done,
1336 sinfo.bytes_avail);
1337 }
1338 /* block exists, so we can just overwrite it */
1339 qemu_co_mutex_unlock(&s->lock);
Kevin Wolf25ec1772016-05-24 17:21:22 +02001340 ret = bdrv_co_writev(bs->file,
Jeff Codyd92aa882013-10-30 10:44:48 -04001341 sinfo.file_offset >> BDRV_SECTOR_BITS,
1342 sectors_to_write, &hd_qiov);
1343 qemu_co_mutex_lock(&s->lock);
1344 if (ret < 0) {
1345 goto error_bat_restore;
1346 }
1347 break;
1348 case PAYLOAD_BLOCK_PARTIALLY_PRESENT:
1349 /* we don't yet support difference files, fall through
1350 * to error */
1351 default:
1352 ret = -EIO;
1353 goto exit;
1354 break;
1355 }
1356
1357 if (bat_update) {
1358 /* this will update the BAT entry into the log journal, and
1359 * then flush the log journal out to disk */
1360 ret = vhdx_log_write_and_flush(bs, s, &bat_entry,
1361 sizeof(VHDXBatEntry),
1362 bat_entry_offset);
1363 if (ret < 0) {
1364 goto exit;
1365 }
1366 }
1367
1368 nb_sectors -= sinfo.sectors_avail;
1369 sector_num += sinfo.sectors_avail;
1370 bytes_done += sinfo.bytes_avail;
1371
1372 }
1373 }
1374
1375 goto exit;
1376
1377error_bat_restore:
1378 if (bat_update) {
1379 /* keep metadata in sync, and restore the bat entry state
1380 * if error. */
1381 sinfo.file_offset = bat_prior_offset;
1382 vhdx_update_bat_table_entry(bs, s, &sinfo, &bat_entry,
1383 &bat_entry_offset, bat_state);
1384 }
1385exit:
1386 qemu_vfree(iov1.iov_base);
1387 qemu_vfree(iov2.iov_base);
1388 qemu_co_mutex_unlock(&s->lock);
1389 qemu_iovec_destroy(&hd_qiov);
1390 return ret;
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001391}
1392
1393
Jeff Cody3412f7b2013-10-30 10:44:53 -04001394
1395/*
1396 * Create VHDX Headers
1397 *
1398 * There are 2 headers, and the highest sequence number will represent
1399 * the active header
1400 */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001401static int vhdx_create_new_headers(BlockBackend *blk, uint64_t image_size,
Jeff Cody3412f7b2013-10-30 10:44:53 -04001402 uint32_t log_size)
1403{
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001404 BlockDriverState *bs = blk_bs(blk);
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +02001405 BdrvChild *child;
Jeff Cody3412f7b2013-10-30 10:44:53 -04001406 int ret = 0;
1407 VHDXHeader *hdr = NULL;
1408
Markus Armbruster5839e532014-08-19 10:31:08 +02001409 hdr = g_new0(VHDXHeader, 1);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001410
1411 hdr->signature = VHDX_HEADER_SIGNATURE;
1412 hdr->sequence_number = g_random_int();
1413 hdr->log_version = 0;
1414 hdr->version = 1;
1415 hdr->log_length = log_size;
1416 hdr->log_offset = VHDX_HEADER_SECTION_END;
1417 vhdx_guid_generate(&hdr->file_write_guid);
1418 vhdx_guid_generate(&hdr->data_write_guid);
1419
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +02001420 /* XXX Ugly way to get blk->root, but that's a feature, not a bug. This
1421 * hack makes it obvious that vhdx_write_header() bypasses the BlockBackend
1422 * here, which it really shouldn't be doing. */
1423 child = QLIST_FIRST(&bs->parents);
1424 assert(!QLIST_NEXT(child, next_parent));
1425
1426 ret = vhdx_write_header(child, hdr, VHDX_HEADER1_OFFSET, false);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001427 if (ret < 0) {
1428 goto exit;
1429 }
1430 hdr->sequence_number++;
Kevin Wolfcf2ab8f2016-06-20 18:24:02 +02001431 ret = vhdx_write_header(child, hdr, VHDX_HEADER2_OFFSET, false);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001432 if (ret < 0) {
1433 goto exit;
1434 }
1435
1436exit:
1437 g_free(hdr);
1438 return ret;
1439}
1440
Jeff Codye91a8b22014-09-16 15:12:06 -04001441#define VHDX_METADATA_ENTRY_BUFFER_SIZE \
1442 (sizeof(VHDXFileParameters) +\
1443 sizeof(VHDXVirtualDiskSize) +\
1444 sizeof(VHDXPage83Data) +\
1445 sizeof(VHDXVirtualDiskLogicalSectorSize) +\
1446 sizeof(VHDXVirtualDiskPhysicalSectorSize))
Jeff Cody3412f7b2013-10-30 10:44:53 -04001447
1448/*
1449 * Create the Metadata entries.
1450 *
1451 * For more details on the entries, see section 3.5 (pg 29) in the
1452 * VHDX 1.00 specification.
1453 *
1454 * We support 5 metadata entries (all required by spec):
1455 * File Parameters,
1456 * Virtual Disk Size,
1457 * Page 83 Data,
1458 * Logical Sector Size,
1459 * Physical Sector Size
1460 *
1461 * The first 64KB of the Metadata section is reserved for the metadata
1462 * header and entries; beyond that, the metadata items themselves reside.
1463 */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001464static int vhdx_create_new_metadata(BlockBackend *blk,
Jeff Cody3412f7b2013-10-30 10:44:53 -04001465 uint64_t image_size,
1466 uint32_t block_size,
1467 uint32_t sector_size,
1468 uint64_t metadata_offset,
1469 VHDXImageType type)
1470{
1471 int ret = 0;
1472 uint32_t offset = 0;
1473 void *buffer = NULL;
1474 void *entry_buffer;
Daniel P. Berrangea8f15a22015-08-26 12:17:12 +01001475 VHDXMetadataTableHeader *md_table;
Jeff Cody3412f7b2013-10-30 10:44:53 -04001476 VHDXMetadataTableEntry *md_table_entry;
1477
1478 /* Metadata entries */
1479 VHDXFileParameters *mt_file_params;
1480 VHDXVirtualDiskSize *mt_virtual_size;
1481 VHDXPage83Data *mt_page83;
1482 VHDXVirtualDiskLogicalSectorSize *mt_log_sector_size;
1483 VHDXVirtualDiskPhysicalSectorSize *mt_phys_sector_size;
1484
Jeff Codye91a8b22014-09-16 15:12:06 -04001485 entry_buffer = g_malloc0(VHDX_METADATA_ENTRY_BUFFER_SIZE);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001486
1487 mt_file_params = entry_buffer;
1488 offset += sizeof(VHDXFileParameters);
1489 mt_virtual_size = entry_buffer + offset;
1490 offset += sizeof(VHDXVirtualDiskSize);
1491 mt_page83 = entry_buffer + offset;
1492 offset += sizeof(VHDXPage83Data);
1493 mt_log_sector_size = entry_buffer + offset;
1494 offset += sizeof(VHDXVirtualDiskLogicalSectorSize);
1495 mt_phys_sector_size = entry_buffer + offset;
1496
1497 mt_file_params->block_size = cpu_to_le32(block_size);
1498 if (type == VHDX_TYPE_FIXED) {
1499 mt_file_params->data_bits |= VHDX_PARAMS_LEAVE_BLOCKS_ALLOCED;
1500 cpu_to_le32s(&mt_file_params->data_bits);
1501 }
1502
1503 vhdx_guid_generate(&mt_page83->page_83_data);
1504 cpu_to_leguids(&mt_page83->page_83_data);
1505 mt_virtual_size->virtual_disk_size = cpu_to_le64(image_size);
1506 mt_log_sector_size->logical_sector_size = cpu_to_le32(sector_size);
1507 mt_phys_sector_size->physical_sector_size = cpu_to_le32(sector_size);
1508
1509 buffer = g_malloc0(VHDX_HEADER_BLOCK_SIZE);
1510 md_table = buffer;
1511
1512 md_table->signature = VHDX_METADATA_SIGNATURE;
1513 md_table->entry_count = 5;
1514 vhdx_metadata_header_le_export(md_table);
1515
1516
1517 /* This will reference beyond the reserved table portion */
1518 offset = 64 * KiB;
1519
1520 md_table_entry = buffer + sizeof(VHDXMetadataTableHeader);
1521
1522 md_table_entry[0].item_id = file_param_guid;
1523 md_table_entry[0].offset = offset;
1524 md_table_entry[0].length = sizeof(VHDXFileParameters);
1525 md_table_entry[0].data_bits |= VHDX_META_FLAGS_IS_REQUIRED;
1526 offset += md_table_entry[0].length;
1527 vhdx_metadata_entry_le_export(&md_table_entry[0]);
1528
1529 md_table_entry[1].item_id = virtual_size_guid;
1530 md_table_entry[1].offset = offset;
1531 md_table_entry[1].length = sizeof(VHDXVirtualDiskSize);
1532 md_table_entry[1].data_bits |= VHDX_META_FLAGS_IS_REQUIRED |
1533 VHDX_META_FLAGS_IS_VIRTUAL_DISK;
1534 offset += md_table_entry[1].length;
1535 vhdx_metadata_entry_le_export(&md_table_entry[1]);
1536
1537 md_table_entry[2].item_id = page83_guid;
1538 md_table_entry[2].offset = offset;
1539 md_table_entry[2].length = sizeof(VHDXPage83Data);
1540 md_table_entry[2].data_bits |= VHDX_META_FLAGS_IS_REQUIRED |
1541 VHDX_META_FLAGS_IS_VIRTUAL_DISK;
1542 offset += md_table_entry[2].length;
1543 vhdx_metadata_entry_le_export(&md_table_entry[2]);
1544
1545 md_table_entry[3].item_id = logical_sector_guid;
1546 md_table_entry[3].offset = offset;
1547 md_table_entry[3].length = sizeof(VHDXVirtualDiskLogicalSectorSize);
1548 md_table_entry[3].data_bits |= VHDX_META_FLAGS_IS_REQUIRED |
1549 VHDX_META_FLAGS_IS_VIRTUAL_DISK;
1550 offset += md_table_entry[3].length;
1551 vhdx_metadata_entry_le_export(&md_table_entry[3]);
1552
1553 md_table_entry[4].item_id = phys_sector_guid;
1554 md_table_entry[4].offset = offset;
1555 md_table_entry[4].length = sizeof(VHDXVirtualDiskPhysicalSectorSize);
1556 md_table_entry[4].data_bits |= VHDX_META_FLAGS_IS_REQUIRED |
1557 VHDX_META_FLAGS_IS_VIRTUAL_DISK;
1558 vhdx_metadata_entry_le_export(&md_table_entry[4]);
1559
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001560 ret = blk_pwrite(blk, metadata_offset, buffer, VHDX_HEADER_BLOCK_SIZE, 0);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001561 if (ret < 0) {
1562 goto exit;
1563 }
1564
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001565 ret = blk_pwrite(blk, metadata_offset + (64 * KiB), entry_buffer,
1566 VHDX_METADATA_ENTRY_BUFFER_SIZE, 0);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001567 if (ret < 0) {
1568 goto exit;
1569 }
1570
1571
1572exit:
1573 g_free(buffer);
1574 g_free(entry_buffer);
1575 return ret;
1576}
1577
1578/* This create the actual BAT itself. We currently only support
1579 * 'Dynamic' and 'Fixed' image types.
1580 *
1581 * Dynamic images: default state of the BAT is all zeroes.
1582 *
1583 * Fixed images: default state of the BAT is fully populated, with
1584 * file offsets and state PAYLOAD_BLOCK_FULLY_PRESENT.
1585 */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001586static int vhdx_create_bat(BlockBackend *blk, BDRVVHDXState *s,
Jeff Cody3412f7b2013-10-30 10:44:53 -04001587 uint64_t image_size, VHDXImageType type,
Jeff Cody4f75b522014-08-06 15:54:58 -04001588 bool use_zero_blocks, uint64_t file_offset,
1589 uint32_t length)
Jeff Cody3412f7b2013-10-30 10:44:53 -04001590{
1591 int ret = 0;
1592 uint64_t data_file_offset;
1593 uint64_t total_sectors = 0;
1594 uint64_t sector_num = 0;
1595 uint64_t unused;
1596 int block_state;
1597 VHDXSectorInfo sinfo;
1598
1599 assert(s->bat == NULL);
1600
1601 /* this gives a data start after BAT/bitmap entries, and well
1602 * past any metadata entries (with a 4 MB buffer for future
1603 * expansion */
Jeff Cody4f75b522014-08-06 15:54:58 -04001604 data_file_offset = file_offset + length + 5 * MiB;
Jeff Cody3412f7b2013-10-30 10:44:53 -04001605 total_sectors = image_size >> s->logical_sector_size_bits;
1606
1607 if (type == VHDX_TYPE_DYNAMIC) {
1608 /* All zeroes, so we can just extend the file - the end of the BAT
1609 * is the furthest thing we have written yet */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001610 ret = blk_truncate(blk, data_file_offset);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001611 if (ret < 0) {
1612 goto exit;
1613 }
1614 } else if (type == VHDX_TYPE_FIXED) {
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001615 ret = blk_truncate(blk, data_file_offset + image_size);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001616 if (ret < 0) {
1617 goto exit;
1618 }
1619 } else {
1620 ret = -ENOTSUP;
1621 goto exit;
1622 }
1623
1624 if (type == VHDX_TYPE_FIXED ||
1625 use_zero_blocks ||
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001626 bdrv_has_zero_init(blk_bs(blk)) == 0) {
Jeff Cody3412f7b2013-10-30 10:44:53 -04001627 /* for a fixed file, the default BAT entry is not zero */
Kevin Wolfa67e1282014-05-20 13:55:50 +02001628 s->bat = g_try_malloc0(length);
Adelina Tuveniea0118982014-09-18 18:17:44 +03001629 if (length && s->bat == NULL) {
Kevin Wolfa67e1282014-05-20 13:55:50 +02001630 ret = -ENOMEM;
1631 goto exit;
1632 }
Jeff Cody3412f7b2013-10-30 10:44:53 -04001633 block_state = type == VHDX_TYPE_FIXED ? PAYLOAD_BLOCK_FULLY_PRESENT :
1634 PAYLOAD_BLOCK_NOT_PRESENT;
1635 block_state = use_zero_blocks ? PAYLOAD_BLOCK_ZERO : block_state;
1636 /* fill the BAT by emulating sector writes of sectors_per_block size */
1637 while (sector_num < total_sectors) {
1638 vhdx_block_translate(s, sector_num, s->sectors_per_block, &sinfo);
1639 sinfo.file_offset = data_file_offset +
1640 (sector_num << s->logical_sector_size_bits);
1641 sinfo.file_offset = ROUND_UP(sinfo.file_offset, MiB);
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001642 vhdx_update_bat_table_entry(blk_bs(blk), s, &sinfo, &unused, &unused,
Jeff Cody3412f7b2013-10-30 10:44:53 -04001643 block_state);
1644 cpu_to_le64s(&s->bat[sinfo.bat_idx]);
1645 sector_num += s->sectors_per_block;
1646 }
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001647 ret = blk_pwrite(blk, file_offset, s->bat, length, 0);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001648 if (ret < 0) {
1649 goto exit;
1650 }
1651 }
1652
1653
1654
1655exit:
1656 g_free(s->bat);
1657 return ret;
1658}
1659
1660/* Creates the region table header, and region table entries.
1661 * There are 2 supported region table entries: BAT, and Metadata/
1662 *
1663 * As the calculations for the BAT region table are also needed
1664 * to create the BAT itself, we will also cause the BAT to be
1665 * created.
1666 */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001667static int vhdx_create_new_region_table(BlockBackend *blk,
Jeff Cody3412f7b2013-10-30 10:44:53 -04001668 uint64_t image_size,
1669 uint32_t block_size,
1670 uint32_t sector_size,
1671 uint32_t log_size,
1672 bool use_zero_blocks,
1673 VHDXImageType type,
1674 uint64_t *metadata_offset)
1675{
1676 int ret = 0;
1677 uint32_t offset = 0;
1678 void *buffer = NULL;
Jeff Cody4f75b522014-08-06 15:54:58 -04001679 uint64_t bat_file_offset;
1680 uint32_t bat_length;
Jeff Cody3412f7b2013-10-30 10:44:53 -04001681 BDRVVHDXState *s = NULL;
1682 VHDXRegionTableHeader *region_table;
1683 VHDXRegionTableEntry *rt_bat;
1684 VHDXRegionTableEntry *rt_metadata;
1685
1686 assert(metadata_offset != NULL);
1687
1688 /* Populate enough of the BDRVVHDXState to be able to use the
1689 * pre-existing BAT calculation, translation, and update functions */
Markus Armbruster5839e532014-08-19 10:31:08 +02001690 s = g_new0(BDRVVHDXState, 1);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001691
1692 s->chunk_ratio = (VHDX_MAX_SECTORS_PER_BLOCK) *
1693 (uint64_t) sector_size / (uint64_t) block_size;
1694
1695 s->sectors_per_block = block_size / sector_size;
1696 s->virtual_disk_size = image_size;
1697 s->block_size = block_size;
1698 s->logical_sector_size = sector_size;
1699
1700 vhdx_set_shift_bits(s);
1701
1702 vhdx_calc_bat_entries(s);
1703
1704 /* At this point the VHDX state is populated enough for creation */
1705
1706 /* a single buffer is used so we can calculate the checksum over the
1707 * entire 64KB block */
1708 buffer = g_malloc0(VHDX_HEADER_BLOCK_SIZE);
1709 region_table = buffer;
1710 offset += sizeof(VHDXRegionTableHeader);
1711 rt_bat = buffer + offset;
1712 offset += sizeof(VHDXRegionTableEntry);
1713 rt_metadata = buffer + offset;
1714
1715 region_table->signature = VHDX_REGION_SIGNATURE;
1716 region_table->entry_count = 2; /* BAT and Metadata */
1717
1718 rt_bat->guid = bat_guid;
1719 rt_bat->length = ROUND_UP(s->bat_entries * sizeof(VHDXBatEntry), MiB);
1720 rt_bat->file_offset = ROUND_UP(VHDX_HEADER_SECTION_END + log_size, MiB);
1721 s->bat_offset = rt_bat->file_offset;
1722
1723 rt_metadata->guid = metadata_guid;
1724 rt_metadata->file_offset = ROUND_UP(rt_bat->file_offset + rt_bat->length,
1725 MiB);
1726 rt_metadata->length = 1 * MiB; /* min size, and more than enough */
1727 *metadata_offset = rt_metadata->file_offset;
1728
Jeff Cody4f75b522014-08-06 15:54:58 -04001729 bat_file_offset = rt_bat->file_offset;
1730 bat_length = rt_bat->length;
1731
1732 vhdx_region_header_le_export(region_table);
1733 vhdx_region_entry_le_export(rt_bat);
1734 vhdx_region_entry_le_export(rt_metadata);
1735
Jeff Cody3412f7b2013-10-30 10:44:53 -04001736 vhdx_update_checksum(buffer, VHDX_HEADER_BLOCK_SIZE,
1737 offsetof(VHDXRegionTableHeader, checksum));
1738
1739
1740 /* The region table gives us the data we need to create the BAT,
1741 * so do that now */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001742 ret = vhdx_create_bat(blk, s, image_size, type, use_zero_blocks,
Jeff Cody4f75b522014-08-06 15:54:58 -04001743 bat_file_offset, bat_length);
1744 if (ret < 0) {
1745 goto exit;
1746 }
Jeff Cody3412f7b2013-10-30 10:44:53 -04001747
1748 /* Now write out the region headers to disk */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001749 ret = blk_pwrite(blk, VHDX_REGION_TABLE_OFFSET, buffer,
1750 VHDX_HEADER_BLOCK_SIZE, 0);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001751 if (ret < 0) {
1752 goto exit;
1753 }
1754
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001755 ret = blk_pwrite(blk, VHDX_REGION_TABLE2_OFFSET, buffer,
1756 VHDX_HEADER_BLOCK_SIZE, 0);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001757 if (ret < 0) {
1758 goto exit;
1759 }
1760
Jeff Cody3412f7b2013-10-30 10:44:53 -04001761exit:
1762 g_free(s);
1763 g_free(buffer);
1764 return ret;
1765}
1766
1767/* We need to create the following elements:
1768 *
1769 * .-----------------------------------------------------------------.
1770 * | (A) | (B) | (C) | (D) | (E) |
1771 * | File ID | Header1 | Header 2 | Region Tbl 1 | Region Tbl 2 |
1772 * | | | | | |
1773 * .-----------------------------------------------------------------.
1774 * 0 64KB 128KB 192KB 256KB 320KB
1775 *
1776 *
1777 * .---- ~ ----------- ~ ------------ ~ ---------------- ~ -----------.
1778 * | (F) | (G) | (H) | |
1779 * | Journal Log | BAT / Bitmap | Metadata | .... data ...... |
1780 * | | | | |
1781 * .---- ~ ----------- ~ ------------ ~ ---------------- ~ -----------.
1782 * 1MB
1783 */
Chunyan Liu53660922014-06-05 17:21:08 +08001784static int vhdx_create(const char *filename, QemuOpts *opts, Error **errp)
Jeff Cody3412f7b2013-10-30 10:44:53 -04001785{
1786 int ret = 0;
1787 uint64_t image_size = (uint64_t) 2 * GiB;
1788 uint32_t log_size = 1 * MiB;
1789 uint32_t block_size = 0;
1790 uint64_t signature;
1791 uint64_t metadata_offset;
1792 bool use_zero_blocks = false;
1793
1794 gunichar2 *creator = NULL;
1795 glong creator_items;
Kevin Wolf10bf03a2016-03-08 15:57:05 +01001796 BlockBackend *blk;
Chunyan Liu53660922014-06-05 17:21:08 +08001797 char *type = NULL;
Jeff Cody3412f7b2013-10-30 10:44:53 -04001798 VHDXImageType image_type;
1799 Error *local_err = NULL;
1800
Hu Taoc2eb9182014-09-10 17:05:45 +08001801 image_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0),
1802 BDRV_SECTOR_SIZE);
Chunyan Liu53660922014-06-05 17:21:08 +08001803 log_size = qemu_opt_get_size_del(opts, VHDX_BLOCK_OPT_LOG_SIZE, 0);
1804 block_size = qemu_opt_get_size_del(opts, VHDX_BLOCK_OPT_BLOCK_SIZE, 0);
1805 type = qemu_opt_get_del(opts, BLOCK_OPT_SUBFMT);
Jeff Cody30af51c2014-12-08 01:07:44 -05001806 use_zero_blocks = qemu_opt_get_bool_del(opts, VHDX_BLOCK_OPT_ZERO, true);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001807
1808 if (image_size > VHDX_MAX_IMAGE_SIZE) {
1809 error_setg_errno(errp, EINVAL, "Image size too large; max of 64TB");
1810 ret = -EINVAL;
1811 goto exit;
1812 }
1813
1814 if (type == NULL) {
Chunyan Liu53660922014-06-05 17:21:08 +08001815 type = g_strdup("dynamic");
Jeff Cody3412f7b2013-10-30 10:44:53 -04001816 }
1817
1818 if (!strcmp(type, "dynamic")) {
1819 image_type = VHDX_TYPE_DYNAMIC;
1820 } else if (!strcmp(type, "fixed")) {
1821 image_type = VHDX_TYPE_FIXED;
1822 } else if (!strcmp(type, "differencing")) {
1823 error_setg_errno(errp, ENOTSUP,
1824 "Differencing files not yet supported");
1825 ret = -ENOTSUP;
1826 goto exit;
1827 } else {
1828 ret = -EINVAL;
1829 goto exit;
1830 }
1831
1832 /* These are pretty arbitrary, and mainly designed to keep the BAT
1833 * size reasonable to load into RAM */
1834 if (block_size == 0) {
1835 if (image_size > 32 * TiB) {
1836 block_size = 64 * MiB;
1837 } else if (image_size > (uint64_t) 100 * GiB) {
1838 block_size = 32 * MiB;
1839 } else if (image_size > 1 * GiB) {
1840 block_size = 16 * MiB;
1841 } else {
1842 block_size = 8 * MiB;
1843 }
1844 }
1845
1846
1847 /* make the log size close to what was specified, but must be
1848 * min 1MB, and multiple of 1MB */
1849 log_size = ROUND_UP(log_size, MiB);
1850
1851 block_size = ROUND_UP(block_size, MiB);
1852 block_size = block_size > VHDX_BLOCK_SIZE_MAX ? VHDX_BLOCK_SIZE_MAX :
1853 block_size;
1854
Chunyan Liuc282e1f2014-06-05 17:21:11 +08001855 ret = bdrv_create_file(filename, opts, &local_err);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001856 if (ret < 0) {
1857 error_propagate(errp, local_err);
1858 goto exit;
1859 }
1860
Max Reitzefaa7c42016-03-16 19:54:38 +01001861 blk = blk_new_open(filename, NULL, NULL,
Kevin Wolf55880602017-02-17 15:07:38 +01001862 BDRV_O_RDWR | BDRV_O_RESIZE | BDRV_O_PROTOCOL,
1863 &local_err);
Kevin Wolf10bf03a2016-03-08 15:57:05 +01001864 if (blk == NULL) {
Jeff Cody3412f7b2013-10-30 10:44:53 -04001865 error_propagate(errp, local_err);
Kevin Wolf10bf03a2016-03-08 15:57:05 +01001866 ret = -EIO;
Jeff Cody3412f7b2013-10-30 10:44:53 -04001867 goto exit;
1868 }
1869
Kevin Wolf10bf03a2016-03-08 15:57:05 +01001870 blk_set_allow_write_beyond_eof(blk, true);
1871
Jeff Cody3412f7b2013-10-30 10:44:53 -04001872 /* Create (A) */
1873
1874 /* The creator field is optional, but may be useful for
1875 * debugging / diagnostics */
1876 creator = g_utf8_to_utf16("QEMU v" QEMU_VERSION, -1, NULL,
1877 &creator_items, NULL);
1878 signature = cpu_to_le64(VHDX_FILE_SIGNATURE);
Eric Blake8341f002016-05-06 10:26:27 -06001879 ret = blk_pwrite(blk, VHDX_FILE_ID_OFFSET, &signature, sizeof(signature),
1880 0);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001881 if (ret < 0) {
1882 goto delete_and_exit;
1883 }
1884 if (creator) {
Kevin Wolf10bf03a2016-03-08 15:57:05 +01001885 ret = blk_pwrite(blk, VHDX_FILE_ID_OFFSET + sizeof(signature),
Eric Blake8341f002016-05-06 10:26:27 -06001886 creator, creator_items * sizeof(gunichar2), 0);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001887 if (ret < 0) {
1888 goto delete_and_exit;
1889 }
1890 }
1891
1892
1893 /* Creates (B),(C) */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001894 ret = vhdx_create_new_headers(blk, image_size, log_size);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001895 if (ret < 0) {
1896 goto delete_and_exit;
1897 }
1898
1899 /* Creates (D),(E),(G) explicitly. (F) created as by-product */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001900 ret = vhdx_create_new_region_table(blk, image_size, block_size, 512,
Jeff Cody3412f7b2013-10-30 10:44:53 -04001901 log_size, use_zero_blocks, image_type,
1902 &metadata_offset);
1903 if (ret < 0) {
1904 goto delete_and_exit;
1905 }
1906
1907 /* Creates (H) */
Kevin Wolfdb1e80e2016-06-20 22:05:39 +02001908 ret = vhdx_create_new_metadata(blk, image_size, block_size, 512,
Jeff Cody3412f7b2013-10-30 10:44:53 -04001909 metadata_offset, image_type);
1910 if (ret < 0) {
1911 goto delete_and_exit;
1912 }
1913
1914
Jeff Cody3412f7b2013-10-30 10:44:53 -04001915delete_and_exit:
Kevin Wolf10bf03a2016-03-08 15:57:05 +01001916 blk_unref(blk);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001917exit:
Chunyan Liu53660922014-06-05 17:21:08 +08001918 g_free(type);
Jeff Cody3412f7b2013-10-30 10:44:53 -04001919 g_free(creator);
1920 return ret;
1921}
1922
Jeff Cody7e30e6a2013-12-17 05:33:37 -05001923/* If opened r/w, the VHDX driver will automatically replay the log,
1924 * if one is present, inside the vhdx_open() call.
1925 *
1926 * If qemu-img check -r all is called, the image is automatically opened
1927 * r/w and any log has already been replayed, so there is nothing (currently)
1928 * for us to do here
1929 */
1930static int vhdx_check(BlockDriverState *bs, BdrvCheckResult *result,
1931 BdrvCheckMode fix)
1932{
1933 BDRVVHDXState *s = bs->opaque;
1934
1935 if (s->log_replayed_on_open) {
1936 result->corruptions_fixed++;
1937 }
1938 return 0;
1939}
1940
Chunyan Liu53660922014-06-05 17:21:08 +08001941static QemuOptsList vhdx_create_opts = {
1942 .name = "vhdx-create-opts",
1943 .head = QTAILQ_HEAD_INITIALIZER(vhdx_create_opts.head),
1944 .desc = {
1945 {
1946 .name = BLOCK_OPT_SIZE,
1947 .type = QEMU_OPT_SIZE,
1948 .help = "Virtual disk size; max of 64TB."
1949 },
1950 {
1951 .name = VHDX_BLOCK_OPT_LOG_SIZE,
1952 .type = QEMU_OPT_SIZE,
1953 .def_value_str = stringify(DEFAULT_LOG_SIZE),
1954 .help = "Log size; min 1MB."
1955 },
1956 {
1957 .name = VHDX_BLOCK_OPT_BLOCK_SIZE,
1958 .type = QEMU_OPT_SIZE,
1959 .def_value_str = stringify(0),
1960 .help = "Block Size; min 1MB, max 256MB. " \
1961 "0 means auto-calculate based on image size."
1962 },
1963 {
1964 .name = BLOCK_OPT_SUBFMT,
1965 .type = QEMU_OPT_STRING,
1966 .help = "VHDX format type, can be either 'dynamic' or 'fixed'. "\
1967 "Default is 'dynamic'."
1968 },
1969 {
1970 .name = VHDX_BLOCK_OPT_ZERO,
1971 .type = QEMU_OPT_BOOL,
Jeff Cody30af51c2014-12-08 01:07:44 -05001972 .help = "Force use of payload blocks of type 'ZERO'. "\
1973 "Non-standard, but default. Do not set to 'off' when "\
1974 "using 'qemu-img convert' with subformat=dynamic."
Chunyan Liu53660922014-06-05 17:21:08 +08001975 },
1976 { NULL }
1977 }
Jeff Cody3412f7b2013-10-30 10:44:53 -04001978};
1979
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001980static BlockDriver bdrv_vhdx = {
1981 .format_name = "vhdx",
1982 .instance_size = sizeof(BDRVVHDXState),
1983 .bdrv_probe = vhdx_probe,
1984 .bdrv_open = vhdx_open,
1985 .bdrv_close = vhdx_close,
1986 .bdrv_reopen_prepare = vhdx_reopen_prepare,
Kevin Wolf862f2152016-12-19 16:36:02 +01001987 .bdrv_child_perm = bdrv_format_default_perms,
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001988 .bdrv_co_readv = vhdx_co_readv,
1989 .bdrv_co_writev = vhdx_co_writev,
Chunyan Liuc282e1f2014-06-05 17:21:11 +08001990 .bdrv_create = vhdx_create,
Paolo Bonzini97b00e22013-11-22 13:39:49 +01001991 .bdrv_get_info = vhdx_get_info,
Jeff Cody7e30e6a2013-12-17 05:33:37 -05001992 .bdrv_check = vhdx_check,
Jeff Cody85b712c2014-12-08 01:07:45 -05001993 .bdrv_has_zero_init = bdrv_has_zero_init_1,
Jeff Cody3412f7b2013-10-30 10:44:53 -04001994
Chunyan Liu53660922014-06-05 17:21:08 +08001995 .create_opts = &vhdx_create_opts,
Jeff Codye8d4e5f2013-04-29 14:48:18 -04001996};
1997
1998static void bdrv_vhdx_init(void)
1999{
2000 bdrv_register(&bdrv_vhdx);
2001}
2002
2003block_init(bdrv_vhdx_init);