block/raw_bsd.c

   1 /* BlockDriver implementation for "raw"
   2  *
   3  * Copyright (C) 2010-2016 Red Hat, Inc.
   4  * Copyright (C) 2010, Blue Swirl <blauwirbel@gmail.com>
   5  * Copyright (C) 2009, Anthony Liguori <aliguori@us.ibm.com>
   6  *
   7  * Author:
   8  *   Laszlo Ersek <lersek@redhat.com>
   9  *
  10  * Permission is hereby granted, free of charge, to any person obtaining a copy
  11  * of this software and associated documentation files (the "Software"), to
  12  * deal in the Software without restriction, including without limitation the
  13  * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
  14  * sell copies of the Software, and to permit persons to whom the Software is
  15  * furnished to do so, subject to the following conditions:
  16  *
  17  * The above copyright notice and this permission notice shall be included in
  18  * all copies or substantial portions of the Software.
  19  *
  20  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  21  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  23  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  24  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  25  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  26  * IN THE SOFTWARE.
  27  */
  28
  29 #include "qemu/osdep.h"
  30 #include "block/block_int.h"
  31 #include "qapi/error.h"
  32 #include "qemu/option.h"
  33
  34 typedef struct BDRVRawState {
  35     uint64_t offset;
  36     uint64_t size;
  37     bool has_size;
  38 } BDRVRawState;
  39
  40 static QemuOptsList raw_runtime_opts = {
  41     .name = "raw",
  42     .head = QTAILQ_HEAD_INITIALIZER(raw_runtime_opts.head),
  43     .desc = {
  44         {
  45             .name = "offset",
  46             .type = QEMU_OPT_SIZE,
  47             .help = "offset in the disk where the image starts",
  48         },
  49         {
  50             .name = "size",
  51             .type = QEMU_OPT_SIZE,
  52             .help = "virtual disk size",
  53         },
  54         { /* end of list */ }
  55     },
  56 };
  57
  58 static QemuOptsList raw_create_opts = {
  59     .name = "raw-create-opts",
  60     .head = QTAILQ_HEAD_INITIALIZER(raw_create_opts.head),
  61     .desc = {
  62         {
  63             .name = BLOCK_OPT_SIZE,
  64             .type = QEMU_OPT_SIZE,
  65             .help = "Virtual disk size"
  66         },
  67         { /* end of list */ }
  68     }
  69 };
  70
  71 static int raw_read_options(QDict *options, BlockDriverState *bs,
  72     BDRVRawState *s, Error **errp)
  73 {
  74     Error *local_err = NULL;
  75     QemuOpts *opts = NULL;
  76     int64_t real_size = 0;
  77     int ret;
  78
  79     real_size = bdrv_getlength(bs->file->bs);
  80     if (real_size < 0) {
  81         error_setg_errno(errp, -real_size, "Could not get image size");
  82         return real_size;
  83     }
  84
  85     opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort);
  86     qemu_opts_absorb_qdict(opts, options, &local_err);
  87     if (local_err) {
  88         error_propagate(errp, local_err);
  89         ret = -EINVAL;
  90         goto end;
  91     }
  92
  93     s->offset = qemu_opt_get_size(opts, "offset", 0);
  94     if (qemu_opt_find(opts, "size") != NULL) {
  95         s->size = qemu_opt_get_size(opts, "size", 0);
  96         s->has_size = true;
  97     } else {
  98         s->has_size = false;
  99         s->size = real_size - s->offset;
 100     }
 101
 102     /* Check size and offset */
 103     if (real_size < s->offset || (real_size - s->offset) < s->size) {
 104         error_setg(errp, "The sum of offset (%" PRIu64 ") and size "
 105             "(%" PRIu64 ") has to be smaller or equal to the "
 106             " actual size of the containing file (%" PRId64 ")",
 107             s->offset, s->size, real_size);
 108         ret = -EINVAL;
 109         goto end;
 110     }
 111
 112     /* Make sure size is multiple of BDRV_SECTOR_SIZE to prevent rounding
 113      * up and leaking out of the specified area. */
 114     if (!QEMU_IS_ALIGNED(s->size, BDRV_SECTOR_SIZE)) {
 115         error_setg(errp, "Specified size is not multiple of %llu",
 116             BDRV_SECTOR_SIZE);
 117         ret = -EINVAL;
 118         goto end;
 119     }
 120
 121     ret = 0;
 122
 123 end:
 124
 125     qemu_opts_del(opts);
 126
 127     return ret;
 128 }
 129
 130 static int raw_reopen_prepare(BDRVReopenState *reopen_state,
 131                               BlockReopenQueue *queue, Error **errp)
 132 {
 133     assert(reopen_state != NULL);
 134     assert(reopen_state->bs != NULL);
 135
 136     reopen_state->opaque = g_new0(BDRVRawState, 1);
 137
 138     return raw_read_options(
 139         reopen_state->options,
 140         reopen_state->bs,
 141         reopen_state->opaque,
 142         errp);
 143 }
 144
 145 static void raw_reopen_commit(BDRVReopenState *state)
 146 {
 147     BDRVRawState *new_s = state->opaque;
 148     BDRVRawState *s = state->bs->opaque;
 149
 150     memcpy(s, new_s, sizeof(BDRVRawState));
 151
 152     g_free(state->opaque);
 153     state->opaque = NULL;
 154 }
 155
 156 static void raw_reopen_abort(BDRVReopenState *state)
 157 {
 158     g_free(state->opaque);
 159     state->opaque = NULL;
 160 }
 161
 162 static int coroutine_fn raw_co_preadv(BlockDriverState *bs, uint64_t offset,
 163                                       uint64_t bytes, QEMUIOVector *qiov,
 164                                       int flags)
 165 {
 166     BDRVRawState *s = bs->opaque;
 167
 168     if (offset > UINT64_MAX - s->offset) {
 169         return -EINVAL;
 170     }
 171     offset += s->offset;
 172
 173     BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO);
 174     return bdrv_co_preadv(bs->file, offset, bytes, qiov, flags);
 175 }
 176
 177 static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, uint64_t offset,
 178                                        uint64_t bytes, QEMUIOVector *qiov,
 179                                        int flags)
 180 {
 181     BDRVRawState *s = bs->opaque;
 182     void *buf = NULL;
 183     BlockDriver *drv;
 184     QEMUIOVector local_qiov;
 185     int ret;
 186
 187     if (s->has_size && (offset > s->size || bytes > (s->size - offset))) {
 188         /* There's not enough space for the data. Don't write anything and just
 189          * fail to prevent leaking out of the size specified in options. */
 190         return -ENOSPC;
 191     }
 192
 193     if (offset > UINT64_MAX - s->offset) {
 194         ret = -EINVAL;
 195         goto fail;
 196     }
 197
 198     if (bs->probed && offset < BLOCK_PROBE_BUF_SIZE && bytes) {
 199         /* Handling partial writes would be a pain - so we just
 200          * require that guests have 512-byte request alignment if
 201          * probing occurred */
 202         QEMU_BUILD_BUG_ON(BLOCK_PROBE_BUF_SIZE != 512);
 203         QEMU_BUILD_BUG_ON(BDRV_SECTOR_SIZE != 512);
 204         assert(offset == 0 && bytes >= BLOCK_PROBE_BUF_SIZE);
 205
 206         buf = qemu_try_blockalign(bs->file->bs, 512);
 207         if (!buf) {
 208             ret = -ENOMEM;
 209             goto fail;
 210         }
 211
 212         ret = qemu_iovec_to_buf(qiov, 0, buf, 512);
 213         if (ret != 512) {
 214             ret = -EINVAL;
 215             goto fail;
 216         }
 217
 218         drv = bdrv_probe_all(buf, 512, NULL);
 219         if (drv != bs->drv) {
 220             ret = -EPERM;
 221             goto fail;
 222         }
 223
 224         /* Use the checked buffer, a malicious guest might be overwriting its
 225          * original buffer in the background. */
 226         qemu_iovec_init(&local_qiov, qiov->niov + 1);
 227         qemu_iovec_add(&local_qiov, buf, 512);
 228         qemu_iovec_concat(&local_qiov, qiov, 512, qiov->size - 512);
 229         qiov = &local_qiov;
 230     }
 231
 232     offset += s->offset;
 233
 234     BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO);
 235     ret = bdrv_co_pwritev(bs->file, offset, bytes, qiov, flags);
 236
 237 fail:
 238     if (qiov == &local_qiov) {
 239         qemu_iovec_destroy(&local_qiov);
 240     }
 241     qemu_vfree(buf);
 242     return ret;
 243 }
 244
 245 static int64_t coroutine_fn raw_co_get_block_status(BlockDriverState *bs,
 246                                             int64_t sector_num,
 247                                             int nb_sectors, int *pnum,
 248                                             BlockDriverState **file)
 249 {
 250     BDRVRawState *s = bs->opaque;
 251     *pnum = nb_sectors;
 252     *file = bs->file->bs;
 253     sector_num += s->offset / BDRV_SECTOR_SIZE;
 254     return BDRV_BLOCK_RAW | BDRV_BLOCK_OFFSET_VALID | BDRV_BLOCK_DATA |
 255            (sector_num << BDRV_SECTOR_BITS);
 256 }
 257
 258 static int coroutine_fn raw_co_pwrite_zeroes(BlockDriverState *bs,
 259                                              int64_t offset, int count,
 260                                              BdrvRequestFlags flags)
 261 {
 262     BDRVRawState *s = bs->opaque;
 263     if (offset > UINT64_MAX - s->offset) {
 264         return -EINVAL;
 265     }
 266     offset += s->offset;
 267     return bdrv_co_pwrite_zeroes(bs->file, offset, count, flags);
 268 }
 269
 270 static int coroutine_fn raw_co_pdiscard(BlockDriverState *bs,
 271                                         int64_t offset, int count)
 272 {
 273     BDRVRawState *s = bs->opaque;
 274     if (offset > UINT64_MAX - s->offset) {
 275         return -EINVAL;
 276     }
 277     offset += s->offset;
 278     return bdrv_co_pdiscard(bs->file->bs, offset, count);
 279 }
 280
 281 static int64_t raw_getlength(BlockDriverState *bs)
 282 {
 283     int64_t len;
 284     BDRVRawState *s = bs->opaque;
 285
 286     /* Update size. It should not change unless the file was externally
 287      * modified. */
 288     len = bdrv_getlength(bs->file->bs);
 289     if (len < 0) {
 290         return len;
 291     }
 292
 293     if (len < s->offset) {
 294         s->size = 0;
 295     } else {
 296         if (s->has_size) {
 297             /* Try to honour the size */
 298             s->size = MIN(s->size, len - s->offset);
 299         } else {
 300             s->size = len - s->offset;
 301         }
 302     }
 303
 304     return s->size;
 305 }
 306
 307 static int raw_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
 308 {
 309     return bdrv_get_info(bs->file->bs, bdi);
 310 }
 311
 312 static void raw_refresh_limits(BlockDriverState *bs, Error **errp)
 313 {
 314     if (bs->probed) {
 315         /* To make it easier to protect the first sector, any probed
 316          * image is restricted to read-modify-write on sub-sector
 317          * operations. */
 318         bs->bl.request_alignment = BDRV_SECTOR_SIZE;
 319     }
 320 }
 321
 322 static int raw_truncate(BlockDriverState *bs, int64_t offset)
 323 {
 324     BDRVRawState *s = bs->opaque;
 325
 326     if (s->has_size) {
 327         return -ENOTSUP;
 328     }
 329
 330     if (INT64_MAX - offset < s->offset) {
 331         return -EINVAL;
 332     }
 333
 334     s->size = offset;
 335     offset += s->offset;
 336     return bdrv_truncate(bs->file->bs, offset);
 337 }
 338
 339 static int raw_media_changed(BlockDriverState *bs)
 340 {
 341     return bdrv_media_changed(bs->file->bs);
 342 }
 343
 344 static void raw_eject(BlockDriverState *bs, bool eject_flag)
 345 {
 346     bdrv_eject(bs->file->bs, eject_flag);
 347 }
 348
 349 static void raw_lock_medium(BlockDriverState *bs, bool locked)
 350 {
 351     bdrv_lock_medium(bs->file->bs, locked);
 352 }
 353
 354 static int raw_co_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
 355 {
 356     BDRVRawState *s = bs->opaque;
 357     if (s->offset || s->has_size) {
 358         return -ENOTSUP;
 359     }
 360     return bdrv_co_ioctl(bs->file->bs, req, buf);
 361 }
 362
 363 static int raw_has_zero_init(BlockDriverState *bs)
 364 {
 365     return bdrv_has_zero_init(bs->file->bs);
 366 }
 367
 368 static int raw_create(const char *filename, QemuOpts *opts, Error **errp)
 369 {
 370     return bdrv_create_file(filename, opts, errp);
 371 }
 372
 373 static int raw_open(BlockDriverState *bs, QDict *options, int flags,
 374                     Error **errp)
 375 {
 376     BDRVRawState *s = bs->opaque;
 377     int ret;
 378
 379     bs->sg = bs->file->bs->sg;
 380     bs->supported_write_flags = BDRV_REQ_FUA &
 381         bs->file->bs->supported_write_flags;
 382     bs->supported_zero_flags = (BDRV_REQ_FUA | BDRV_REQ_MAY_UNMAP) &
 383         bs->file->bs->supported_zero_flags;
 384
 385     if (bs->probed && !bdrv_is_read_only(bs)) {
 386         fprintf(stderr,
 387                 "WARNING: Image format was not specified for '%s' and probing "
 388                 "guessed raw.\n"
 389                 "         Automatically detecting the format is dangerous for "
 390                 "raw images, write operations on block 0 will be restricted.\n"
 391                 "         Specify the 'raw' format explicitly to remove the "
 392                 "restrictions.\n",
 393                 bs->file->bs->filename);
 394     }
 395
 396     ret = raw_read_options(options, bs, s, errp);
 397     if (ret < 0) {
 398         return ret;
 399     }
 400
 401     if (bs->sg && (s->offset || s->has_size)) {
 402         error_setg(errp, "Cannot use offset/size with SCSI generic devices");
 403         return -EINVAL;
 404     }
 405
 406     return 0;
 407 }
 408
 409 static void raw_close(BlockDriverState *bs)
 410 {
 411 }
 412
 413 static int raw_probe(const uint8_t *buf, int buf_size, const char *filename)
 414 {
 415     /* smallest possible positive score so that raw is used if and only if no
 416      * other block driver works
 417      */
 418     return 1;
 419 }
 420
 421 static int raw_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
 422 {
 423     BDRVRawState *s = bs->opaque;
 424     int ret;
 425
 426     ret = bdrv_probe_blocksizes(bs->file->bs, bsz);
 427     if (ret < 0) {
 428         return ret;
 429     }
 430
 431     if (!QEMU_IS_ALIGNED(s->offset, MAX(bsz->log, bsz->phys))) {
 432         return -ENOTSUP;
 433     }
 434
 435     return 0;
 436 }
 437
 438 static int raw_probe_geometry(BlockDriverState *bs, HDGeometry *geo)
 439 {
 440     BDRVRawState *s = bs->opaque;
 441     if (s->offset || s->has_size) {
 442         return -ENOTSUP;
 443     }
 444     return bdrv_probe_geometry(bs->file->bs, geo);
 445 }
 446
 447 BlockDriver bdrv_raw = {
 448     .format_name          = "raw",
 449     .instance_size        = sizeof(BDRVRawState),
 450     .bdrv_probe           = &raw_probe,
 451     .bdrv_reopen_prepare  = &raw_reopen_prepare,
 452     .bdrv_reopen_commit   = &raw_reopen_commit,
 453     .bdrv_reopen_abort    = &raw_reopen_abort,
 454     .bdrv_open            = &raw_open,
 455     .bdrv_close           = &raw_close,
 456     .bdrv_create          = &raw_create,
 457     .bdrv_co_preadv       = &raw_co_preadv,
 458     .bdrv_co_pwritev      = &raw_co_pwritev,
 459     .bdrv_co_pwrite_zeroes = &raw_co_pwrite_zeroes,
 460     .bdrv_co_pdiscard     = &raw_co_pdiscard,
 461     .bdrv_co_get_block_status = &raw_co_get_block_status,
 462     .bdrv_truncate        = &raw_truncate,
 463     .bdrv_getlength       = &raw_getlength,
 464     .has_variable_length  = true,
 465     .bdrv_get_info        = &raw_get_info,
 466     .bdrv_refresh_limits  = &raw_refresh_limits,
 467     .bdrv_probe_blocksizes = &raw_probe_blocksizes,
 468     .bdrv_probe_geometry  = &raw_probe_geometry,
 469     .bdrv_media_changed   = &raw_media_changed,
 470     .bdrv_eject           = &raw_eject,
 471     .bdrv_lock_medium     = &raw_lock_medium,
 472     .bdrv_co_ioctl        = &raw_co_ioctl,
 473     .create_opts          = &raw_create_opts,
 474     .bdrv_has_zero_init   = &raw_has_zero_init
 475 };
 476
 477 static void bdrv_raw_init(void)
 478 {
 479     bdrv_register(&bdrv_raw);
 480 }
 481
 482 block_init(bdrv_raw_init);