storage_backend.c 56.7 KB
Newer Older
1
/*
2
 * storage_backend.c: internal storage driver backend contract
3
 *
4
 * Copyright (C) 2007-2013 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16 17
 * Copyright (C) 2007-2008 Daniel P. Berrange
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with this library.  If not, see
O
Osier Yang 已提交
19
 * <http://www.gnu.org/licenses/>.
20 21 22 23 24 25 26
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

#include <config.h>

#include <string.h>
27
#include <stdio.h>
28
#include <regex.h>
29
#include <sys/types.h>
E
Eric Blake 已提交
30
#include <sys/wait.h>
31
#include <unistd.h>
32 33
#include <fcntl.h>
#include <sys/stat.h>
34
#include <sys/param.h>
35
#include <dirent.h>
36
#include "dirname.h"
37 38 39 40
#ifdef __linux__
# include <sys/ioctl.h>
# include <linux/fs.h>
#endif
41

42
#if WITH_SELINUX
43
# include <selinux/selinux.h>
44
#endif
45

46
#include "datatypes.h"
47
#include "virerror.h"
48
#include "viralloc.h"
49
#include "internal.h"
50
#include "secret_conf.h"
51
#include "viruuid.h"
52
#include "virstoragefile.h"
D
Daniel P. Berrange 已提交
53
#include "storage_backend.h"
54
#include "virlog.h"
E
Eric Blake 已提交
55
#include "virfile.h"
56
#include "stat-time.h"
57
#include "virstring.h"
58

59
#if WITH_STORAGE_LVM
60
# include "storage_backend_logical.h"
61 62
#endif
#if WITH_STORAGE_ISCSI
63
# include "storage_backend_iscsi.h"
64
#endif
65
#if WITH_STORAGE_SCSI
66
# include "storage_backend_scsi.h"
67
#endif
D
Dave Allan 已提交
68
#if WITH_STORAGE_MPATH
69
# include "storage_backend_mpath.h"
D
Dave Allan 已提交
70
#endif
71
#if WITH_STORAGE_DISK
72
# include "storage_backend_disk.h"
73 74
#endif
#if WITH_STORAGE_DIR
75
# include "storage_backend_fs.h"
76
#endif
77 78 79
#if WITH_STORAGE_RBD
# include "storage_backend_rbd.h"
#endif
80 81 82
#if WITH_STORAGE_SHEEPDOG
# include "storage_backend_sheepdog.h"
#endif
83 84 85
#if WITH_STORAGE_GLUSTER
# include "storage_backend_gluster.h"
#endif
86

87 88
#define VIR_FROM_THIS VIR_FROM_STORAGE

89 90 91 92 93 94 95 96 97 98 99 100 101 102
static virStorageBackendPtr backends[] = {
#if WITH_STORAGE_DIR
    &virStorageBackendDirectory,
#endif
#if WITH_STORAGE_FS
    &virStorageBackendFileSystem,
    &virStorageBackendNetFileSystem,
#endif
#if WITH_STORAGE_LVM
    &virStorageBackendLogical,
#endif
#if WITH_STORAGE_ISCSI
    &virStorageBackendISCSI,
#endif
103 104 105
#if WITH_STORAGE_SCSI
    &virStorageBackendSCSI,
#endif
D
Dave Allan 已提交
106 107 108
#if WITH_STORAGE_MPATH
    &virStorageBackendMpath,
#endif
109 110
#if WITH_STORAGE_DISK
    &virStorageBackendDisk,
111 112 113
#endif
#if WITH_STORAGE_RBD
    &virStorageBackendRBD,
114 115 116
#endif
#if WITH_STORAGE_SHEEPDOG
    &virStorageBackendSheepdog,
117 118 119
#endif
#if WITH_STORAGE_GLUSTER
    &virStorageBackendGluster,
120 121 122
#endif
    NULL
};
123

124 125

static virStorageFileBackendPtr fileBackends[] = {
126 127 128
#if WITH_STORAGE_FS
    &virStorageFileBackendFile,
    &virStorageFileBackendBlock,
129 130 131
#endif
#if WITH_STORAGE_GLUSTER
    &virStorageFileBackendGluster,
132
#endif
133 134 135 136
    NULL
};


137 138 139 140 141 142
enum {
    TOOL_QEMU_IMG,
    TOOL_KVM_IMG,
    TOOL_QCOW_CREATE,
};

143 144 145
#define READ_BLOCK_SIZE_DEFAULT  (1024 * 1024)
#define WRITE_BLOCK_SIZE_DEFAULT (4 * 1024)

146
static int ATTRIBUTE_NONNULL(2)
147
virStorageBackendCopyToFD(virStorageVolDefPtr vol,
148 149
                          virStorageVolDefPtr inputvol,
                          int fd,
150
                          unsigned long long *total,
J
Ján Tomko 已提交
151
                          bool want_sparse)
152 153 154
{
    int inputfd = -1;
    int amtread = -1;
155
    int ret = 0;
156
    size_t rbytes = READ_BLOCK_SIZE_DEFAULT;
L
Li Zhang 已提交
157
    int wbytes = 0;
158
    int interval;
159
    char *zerobuf = NULL;
160
    char *buf = NULL;
161
    struct stat st;
162

163
    if ((inputfd = open(inputvol->target.path, O_RDONLY)) < 0) {
164
        ret = -errno;
165
        virReportSystemError(errno,
166 167 168
                             _("could not open input path '%s'"),
                             inputvol->target.path);
        goto cleanup;
169 170
    }

171 172 173 174 175 176 177 178 179 180 181 182 183 184
#ifdef __linux__
    if (ioctl(fd, BLKBSZGET, &wbytes) < 0) {
        wbytes = 0;
    }
#endif
    if ((wbytes == 0) && fstat(fd, &st) == 0)
        wbytes = st.st_blksize;
    if (wbytes < WRITE_BLOCK_SIZE_DEFAULT)
        wbytes = WRITE_BLOCK_SIZE_DEFAULT;

    if (VIR_ALLOC_N(zerobuf, wbytes) < 0) {
        ret = -errno;
        goto cleanup;
    }
185

186
    if (VIR_ALLOC_N(buf, rbytes) < 0) {
187
        ret = -errno;
188 189 190 191 192 193
        goto cleanup;
    }

    while (amtread != 0) {
        int amtleft;

194 195
        if (*total < rbytes)
            rbytes = *total;
196

197
        if ((amtread = saferead(inputfd, buf, rbytes)) < 0) {
198
            ret = -errno;
199
            virReportSystemError(errno,
200 201 202 203
                                 _("failed reading from file '%s'"),
                                 inputvol->target.path);
            goto cleanup;
        }
204
        *total -= amtread;
205 206 207 208 209

        /* Loop over amt read in 512 byte increments, looking for sparse
         * blocks */
        amtleft = amtread;
        do {
210
            interval = ((wbytes > amtleft) ? amtleft : wbytes);
211 212
            int offset = amtread - amtleft;

213
            if (want_sparse && memcmp(buf+offset, zerobuf, interval) == 0) {
214
                if (lseek(fd, interval, SEEK_CUR) < 0) {
215
                    ret = -errno;
216
                    virReportSystemError(errno,
217 218 219 220 221
                                         _("cannot extend file '%s'"),
                                         vol->target.path);
                    goto cleanup;
                }
            } else if (safewrite(fd, buf+offset, interval) < 0) {
222
                ret = -errno;
223
                virReportSystemError(errno,
224 225 226 227 228
                                     _("failed writing to file '%s'"),
                                     vol->target.path);
                goto cleanup;

            }
229
        } while ((amtleft -= interval) > 0);
230 231
    }

232 233 234 235 236 237 238 239
    if (fdatasync(fd) < 0) {
        ret = -errno;
        virReportSystemError(errno, _("cannot sync data to file '%s'"),
                             vol->target.path);
        goto cleanup;
    }


240
    if (VIR_CLOSE(inputfd) < 0) {
241
        ret = -errno;
242
        virReportSystemError(errno,
243 244 245 246 247 248 249
                             _("cannot close file '%s'"),
                             inputvol->target.path);
        goto cleanup;
    }
    inputfd = -1;

cleanup:
250
    VIR_FORCE_CLOSE(inputfd);
251

252
    VIR_FREE(zerobuf);
253 254
    VIR_FREE(buf);

255 256 257
    return ret;
}

258
static int
259
virStorageBackendCreateBlockFrom(virConnectPtr conn ATTRIBUTE_UNUSED,
260
                                 virStoragePoolObjPtr pool ATTRIBUTE_UNUSED,
261 262
                                 virStorageVolDefPtr vol,
                                 virStorageVolDefPtr inputvol,
E
Eric Blake 已提交
263
                                 unsigned int flags)
264 265 266 267
{
    int fd = -1;
    int ret = -1;
    unsigned long long remain;
268 269 270
    struct stat st;
    gid_t gid;
    uid_t uid;
271

272 273 274 275 276 277 278 279
    virCheckFlags(VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA, -1);

    if (flags & VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("metadata preallocation is not supported for block "
                         "volumes"));
        goto cleanup;
    }
E
Eric Blake 已提交
280

281
    if ((fd = open(vol->target.path, O_RDWR)) < 0) {
282
        virReportSystemError(errno,
283 284 285 286 287 288 289 290
                             _("cannot create path '%s'"),
                             vol->target.path);
        goto cleanup;
    }

    remain = vol->allocation;

    if (inputvol) {
291
        int res = virStorageBackendCopyToFD(vol, inputvol,
J
Ján Tomko 已提交
292
                                            fd, &remain, false);
293 294 295 296
        if (res < 0)
            goto cleanup;
    }

297
    if (fstat(fd, &st) == -1) {
298
        virReportSystemError(errno, _("stat of '%s' failed"),
299 300 301
                             vol->target.path);
        goto cleanup;
    }
P
Philipp Hahn 已提交
302 303 304
    uid = (vol->target.perms.uid != st.st_uid) ? vol->target.perms.uid : (uid_t) -1;
    gid = (vol->target.perms.gid != st.st_gid) ? vol->target.perms.gid : (gid_t) -1;
    if (((uid != (uid_t) -1) || (gid != (gid_t) -1))
305
        && (fchown(fd, uid, gid) < 0)) {
306
        virReportSystemError(errno,
307
                             _("cannot chown '%s' to (%u, %u)"),
308 309
                             vol->target.path, (unsigned int) uid,
                             (unsigned int) gid);
310 311 312
        goto cleanup;
    }
    if (fchmod(fd, vol->target.perms.mode) < 0) {
313
        virReportSystemError(errno,
314 315 316 317
                             _("cannot set mode of '%s' to %04o"),
                             vol->target.path, vol->target.perms.mode);
        goto cleanup;
    }
318
    if (VIR_CLOSE(fd) < 0) {
319
        virReportSystemError(errno,
320 321 322 323 324 325 326 327
                             _("cannot close file '%s'"),
                             vol->target.path);
        goto cleanup;
    }
    fd = -1;

    ret = 0;
cleanup:
328
    VIR_FORCE_CLOSE(fd);
329 330 331 332

    return ret;
}

E
Eric Blake 已提交
333 334 335 336
static int
createRawFile(int fd, virStorageVolDefPtr vol,
              virStorageVolDefPtr inputvol)
{
J
Ján Tomko 已提交
337
    bool need_alloc = true;
338 339
    int ret = 0;
    unsigned long long remain;
340

341 342
    /* Seek to the final size, so the capacity is available upfront
     * for progress reporting */
E
Eric Blake 已提交
343
    if (ftruncate(fd, vol->capacity) < 0) {
344
        ret = -errno;
345
        virReportSystemError(errno,
346
                             _("cannot extend file '%s'"),
E
Eric Blake 已提交
347
                             vol->target.path);
348 349 350
        goto cleanup;
    }

E
Eric Blake 已提交
351 352
/* Avoid issues with older kernel's <linux/fs.h> namespace pollution. */
#if HAVE_FALLOCATE - 0
353
    /* Try to preallocate all requested disk space, but fall back to
354 355
     * other methods if this fails with ENOSYS or EOPNOTSUPP. If allocation
     * is 0 (or less than 0), then fallocate will fail with EINVAL.
356 357 358 359
     * NOTE: do not use posix_fallocate; posix_fallocate falls back
     * to writing zeroes block by block in case fallocate isn't
     * available, and since we're going to copy data from another
     * file it doesn't make sense to write the file twice. */
360 361 362 363 364 365 366 367 368 369
    if (vol->allocation) {
        if (fallocate(fd, 0, 0, vol->allocation) == 0) {
            need_alloc = false;
        } else if (errno != ENOSYS && errno != EOPNOTSUPP) {
            ret = -errno;
            virReportSystemError(errno,
                                 _("cannot allocate %llu bytes in file '%s'"),
                                 vol->allocation, vol->target.path);
            goto cleanup;
        }
370 371 372
    }
#endif

E
Eric Blake 已提交
373
    remain = vol->allocation;
374

E
Eric Blake 已提交
375
    if (inputvol) {
376 377 378
        /* allow zero blocks to be skipped if we've requested sparse
         * allocation (allocation < capacity) or we have already
         * been able to allocate the required space. */
J
Ján Tomko 已提交
379 380
        bool want_sparse = !need_alloc ||
                           (vol->allocation < inputvol->capacity);
381 382

        ret = virStorageBackendCopyToFD(vol, inputvol, fd, &remain, want_sparse);
383
        if (ret < 0) {
384
            goto cleanup;
385
        }
386 387
    }

388 389
    if (remain && need_alloc) {
        if (safezero(fd, vol->allocation - remain, remain) < 0) {
390 391 392 393
            ret = -errno;
            virReportSystemError(errno, _("cannot fill file '%s'"),
                                 vol->target.path);
            goto cleanup;
394
        }
395 396 397
    }

    if (fsync(fd) < 0) {
398
        ret = -errno;
399
        virReportSystemError(errno, _("cannot sync data to file '%s'"),
E
Eric Blake 已提交
400
                             vol->target.path);
401
        goto cleanup;
402 403
    }

404 405 406 407 408 409 410 411 412
cleanup:
    return ret;
}

int
virStorageBackendCreateRaw(virConnectPtr conn ATTRIBUTE_UNUSED,
                           virStoragePoolObjPtr pool,
                           virStorageVolDefPtr vol,
                           virStorageVolDefPtr inputvol,
E
Eric Blake 已提交
413
                           unsigned int flags)
414 415
{
    int ret = -1;
E
Eric Blake 已提交
416 417 418
    int fd = -1;
    int operation_flags;

419 420 421 422 423 424 425 426
    virCheckFlags(VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA, -1);

    if (flags & VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("metadata preallocation is not supported for raw "
                         "volumes"));
        goto cleanup;
    }
427 428

    if (vol->target.encryption != NULL) {
429 430
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("storage pool does not support encrypted volumes"));
431 432
        goto cleanup;
    }
433

L
Laine Stump 已提交
434
    operation_flags = VIR_FILE_OPEN_FORCE_MODE | VIR_FILE_OPEN_FORCE_OWNER;
E
Eric Blake 已提交
435
    if (pool->def->type == VIR_STORAGE_POOL_NETFS)
L
Laine Stump 已提交
436
        operation_flags |= VIR_FILE_OPEN_FORK;
E
Eric Blake 已提交
437

438 439
    if ((fd = virFileOpenAs(vol->target.path,
                            O_RDWR | O_CREAT | O_EXCL,
L
Laine Stump 已提交
440 441 442
                            vol->target.perms.mode,
                            vol->target.perms.uid,
                            vol->target.perms.gid,
443
                            operation_flags)) < 0) {
E
Eric Blake 已提交
444
        virReportSystemError(-fd,
445
                             _("Failed to create file '%s'"),
E
Eric Blake 已提交
446 447 448 449
                             vol->target.path);
        goto cleanup;
    }

450 451
    if ((ret = createRawFile(fd, vol, inputvol)) < 0)
        /* createRawFile already reported the exact error. */
E
Eric Blake 已提交
452
        ret = -1;
453 454

cleanup:
E
Eric Blake 已提交
455
    VIR_FORCE_CLOSE(fd);
456 457 458
    return ret;
}

459 460 461 462 463 464 465 466 467
static int
virStorageGenerateSecretUUID(virConnectPtr conn,
                             unsigned char *uuid)
{
    unsigned attempt;

    for (attempt = 0; attempt < 65536; attempt++) {
        virSecretPtr tmp;
        if (virUUIDGenerate(uuid) < 0) {
468 469
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("unable to generate uuid"));
470 471
            return -1;
        }
472
        tmp = conn->secretDriver->secretLookupByUUID(conn, uuid);
473 474 475 476 477 478
        if (tmp == NULL)
            return 0;

        virSecretFree(tmp);
    }

479 480
    virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                   _("too many conflicts when generating an uuid"));
481 482 483 484

    return -1;
}

485 486 487 488 489 490 491 492 493
static int
virStorageGenerateQcowEncryption(virConnectPtr conn,
                                 virStorageVolDefPtr vol)
{
    virSecretDefPtr def = NULL;
    virBuffer buf = VIR_BUFFER_INITIALIZER;
    virStorageEncryptionPtr enc;
    virStorageEncryptionSecretPtr enc_secret = NULL;
    virSecretPtr secret = NULL;
494
    char *xml;
495 496 497
    unsigned char value[VIR_STORAGE_QCOW_PASSPHRASE_SIZE];
    int ret = -1;

498
    if (conn->secretDriver == NULL ||
499 500 501
        conn->secretDriver->secretLookupByUUID == NULL ||
        conn->secretDriver->secretDefineXML == NULL ||
        conn->secretDriver->secretSetValue == NULL) {
502 503
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("secret storage not supported"));
504 505 506 507 508
        goto cleanup;
    }

    enc = vol->target.encryption;
    if (enc->nsecrets != 0) {
509 510
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("secrets already defined"));
511 512 513 514
        goto cleanup;
    }

    if (VIR_ALLOC(enc_secret) < 0 || VIR_REALLOC_N(enc->secrets, 1) < 0 ||
515
        VIR_ALLOC(def) < 0)
516 517
        goto cleanup;

518
    def->ephemeral = false;
519
    def->private = false;
520
    if (virStorageGenerateSecretUUID(conn, def->uuid) < 0)
521
        goto cleanup;
522

523
    def->usage_type = VIR_SECRET_USAGE_TYPE_VOLUME;
524
    if (VIR_STRDUP(def->usage.volume, vol->target.path) < 0)
525
        goto cleanup;
526
    xml = virSecretDefFormat(def);
527 528 529 530 531
    virSecretDefFree(def);
    def = NULL;
    if (xml == NULL)
        goto cleanup;

532
    secret = conn->secretDriver->secretDefineXML(conn, xml, 0);
533 534 535 536 537 538
    if (secret == NULL) {
        VIR_FREE(xml);
        goto cleanup;
    }
    VIR_FREE(xml);

539
    if (virStorageGenerateQcowPassphrase(value) < 0)
540 541
        goto cleanup;

542
    if (conn->secretDriver->secretSetValue(secret, value, sizeof(value), 0) < 0)
543 544 545
        goto cleanup;

    enc_secret->type = VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE;
546
    memcpy(enc_secret->uuid, secret->uuid, VIR_UUID_BUFLEN);
547 548 549 550 551 552 553 554 555
    enc->format = VIR_STORAGE_ENCRYPTION_FORMAT_QCOW;
    enc->secrets[0] = enc_secret; /* Space for secrets[0] allocated above */
    enc_secret = NULL;
    enc->nsecrets = 1;

    ret = 0;

cleanup:
    if (secret != NULL) {
556
        if (ret != 0 &&
557 558
            conn->secretDriver->secretUndefine != NULL)
            conn->secretDriver->secretUndefine(secret);
559 560
        virSecretFree(secret);
    }
561
    virBufferFreeAndReset(&buf);
562 563 564 565 566
    virSecretDefFree(def);
    VIR_FREE(enc_secret);
    return ret;
}

567 568 569 570 571
static int
virStorageBackendCreateExecCommand(virStoragePoolObjPtr pool,
                                   virStorageVolDefPtr vol,
                                   virCommandPtr cmd)
{
572 573 574
    struct stat st;
    gid_t gid;
    uid_t uid;
575
    bool filecreated = false;
576 577

    if ((pool->def->type == VIR_STORAGE_POOL_NETFS)
578
        && (((geteuid() == 0)
P
Philipp Hahn 已提交
579
             && (vol->target.perms.uid != (uid_t) -1)
580
             && (vol->target.perms.uid != 0))
P
Philipp Hahn 已提交
581
            || ((vol->target.perms.gid != (gid_t) -1)
582
                && (vol->target.perms.gid != getegid())))) {
583

584 585
        virCommandSetUID(cmd, vol->target.perms.uid);
        virCommandSetGID(cmd, vol->target.perms.gid);
586 587

        if (virCommandRun(cmd, NULL) == 0) {
588 589
            /* command was successfully run, check if the file was created */
            if (stat(vol->target.path, &st) >=0)
590
                filecreated = true;
591 592
        }
    }
593

594 595 596
    /* don't change uid/gid if we retry */
    virCommandSetUID(cmd, -1);
    virCommandSetGID(cmd, -1);
597

598
    if (!filecreated) {
599
        if (virCommandRun(cmd, NULL) < 0) {
600 601 602
            return -1;
        }
        if (stat(vol->target.path, &st) < 0) {
603
            virReportSystemError(errno,
604
                                 _("failed to create %s"), vol->target.path);
605 606 607 608
            return -1;
        }
    }

P
Philipp Hahn 已提交
609 610 611
    uid = (vol->target.perms.uid != st.st_uid) ? vol->target.perms.uid : (uid_t) -1;
    gid = (vol->target.perms.gid != st.st_gid) ? vol->target.perms.gid : (gid_t) -1;
    if (((uid != (uid_t) -1) || (gid != (gid_t) -1))
612
        && (chown(vol->target.path, uid, gid) < 0)) {
613
        virReportSystemError(errno,
614
                             _("cannot chown %s to (%u, %u)"),
615 616
                             vol->target.path, (unsigned int) uid,
                             (unsigned int) gid);
617 618 619
        return -1;
    }
    if (chmod(vol->target.path, vol->target.perms.mode) < 0) {
620
        virReportSystemError(errno,
621 622 623 624 625 626 627
                             _("cannot set mode of '%s' to %04o"),
                             vol->target.path, vol->target.perms.mode);
        return -1;
    }
    return 0;
}

628 629 630 631
enum {
    QEMU_IMG_BACKING_FORMAT_NONE = 0,
    QEMU_IMG_BACKING_FORMAT_FLAG,
    QEMU_IMG_BACKING_FORMAT_OPTIONS,
632
    QEMU_IMG_BACKING_FORMAT_OPTIONS_COMPAT,
633 634
};

635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659
static bool
virStorageBackendQemuImgSupportsCompat(const char *qemuimg)
{
    bool ret = false;
    char *output;
    virCommandPtr cmd = NULL;

    cmd = virCommandNewArgList(qemuimg, "create", "-o", "?", "-f", "qcow2",
                               "/dev/null", NULL);

    virCommandAddEnvString(cmd, "LC_ALL=C");
    virCommandSetOutputBuffer(cmd, &output);

    if (virCommandRun(cmd, NULL) < 0)
        goto cleanup;

    if (strstr(output, "\ncompat "))
        ret = true;

cleanup:
    virCommandFree(cmd);
    VIR_FREE(output);
    return ret;
}

660 661
static int
virStorageBackendQEMUImgBackingFormat(const char *qemuimg)
662 663 664 665 666 667
{
    char *help = NULL;
    char *start;
    char *end;
    char *tmp;
    int ret = -1;
668
    int exitstatus;
669
    virCommandPtr cmd = virCommandNewArgList(qemuimg, "-h", NULL);
670

671 672 673
    virCommandAddEnvString(cmd, "LC_ALL=C");
    virCommandSetOutputBuffer(cmd, &help);
    virCommandClearCaps(cmd);
674

675 676 677
    /* qemuimg doesn't return zero exit status on -h,
     * therefore we need to provide pointer for storing
     * exit status, although we don't parse it any later */
678
    if (virCommandRun(cmd, &exitstatus) < 0)
679 680
        goto cleanup;

681 682
    if ((start = strstr(help, " create ")) == NULL ||
        (end = strstr(start, "\n")) == NULL) {
683 684 685
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unable to parse qemu-img output '%s'"),
                       help);
686 687
        goto cleanup;
    }
688
    if (((tmp = strstr(start, "-F fmt")) && tmp < end) ||
689
        ((tmp = strstr(start, "-F backing_fmt")) && tmp < end)) {
690
        ret = QEMU_IMG_BACKING_FORMAT_FLAG;
691 692 693 694 695 696
    } else if ((tmp = strstr(start, "[-o options]")) && tmp < end) {
        if (virStorageBackendQemuImgSupportsCompat(qemuimg))
            ret = QEMU_IMG_BACKING_FORMAT_OPTIONS_COMPAT;
        else
            ret = QEMU_IMG_BACKING_FORMAT_OPTIONS;
    } else {
697
        ret = QEMU_IMG_BACKING_FORMAT_NONE;
698
    }
699 700

cleanup:
701
    virCommandFree(cmd);
702 703 704 705
    VIR_FREE(help);
    return ret;
}

706 707 708 709
static int
virStorageBackendCreateQemuImgOpts(char **opts,
                                   const char *backingType,
                                   bool encryption,
710 711 712 713
                                   bool preallocate,
                                   int format,
                                   const char *compat,
                                   virBitmapPtr features)
714 715
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;
716
    bool b;
717
    size_t i;
718

719 720 721 722 723 724 725
    if (backingType)
        virBufferAsprintf(&buf, "backing_fmt=%s,", backingType);
    if (encryption)
        virBufferAddLit(&buf, "encryption=on,");
    if (preallocate)
        virBufferAddLit(&buf, "preallocation=metadata,");

726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742
    if (compat)
        virBufferAsprintf(&buf, "compat=%s,", compat);
    if (features && format == VIR_STORAGE_FILE_QCOW2) {
        for (i = 0; i < VIR_STORAGE_FILE_FEATURE_LAST; i++) {
            ignore_value(virBitmapGetBit(features, i, &b));
            if (b) {
                switch ((enum virStorageFileFeature) i) {
                case VIR_STORAGE_FILE_FEATURE_LAZY_REFCOUNTS:
                    if (STREQ_NULLABLE(compat, "0.10")) {
                        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                                       _("Feature %s not supported with compat"
                                         " level %s"),
                                       virStorageFileFeatureTypeToString(i),
                                       compat);
                        goto error;
                    }
                    break;
743

744
                /* coverity[dead_error_begin] */
745 746 747 748 749 750 751 752 753
                case VIR_STORAGE_FILE_FEATURE_LAST:
                    ;
                }
                virBufferAsprintf(&buf, "%s,",
                                  virStorageFileFeatureTypeToString(i));
            }
        }
    }

754 755
    virBufferTrim(&buf, ",", -1);

756 757
    if (virBufferError(&buf))
        goto no_memory;
758 759 760

    *opts = virBufferContentAndReset(&buf);
    return 0;
761 762 763 764 765 766

no_memory:
    virReportOOMError();
error:
    virBufferFreeAndReset(&buf);
    return -1;
767 768
}

769 770 771 772 773 774 775 776
virCommandPtr
virStorageBackendCreateQemuImgCmd(virConnectPtr conn,
                                  virStoragePoolObjPtr pool,
                                  virStorageVolDefPtr vol,
                                  virStorageVolDefPtr inputvol,
                                  unsigned int flags,
                                  const char *create_tool,
                                  int imgformat)
777
{
778 779
    virCommandPtr cmd = NULL;
    bool do_encryption = (vol->target.encryption != NULL);
780
    unsigned long long int size_arg;
781 782 783 784 785
    bool preallocate = !!(flags & VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA);
    const char *type;
    const char *backingType = NULL;
    const char *inputPath = NULL;
    const char *inputType = NULL;
786
    const char *compat = vol->target.compat;
787 788 789
    char *opts = NULL;
    bool convert = false;
    bool backing = false;
790

791 792
    virCheckFlags(VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA, NULL);

793 794 795 796
    /* Treat output block devices as 'raw' format */
    type = virStorageFileFormatTypeToString(vol->type == VIR_STORAGE_VOL_BLOCK ?
                                            VIR_STORAGE_FILE_RAW :
                                            vol->target.format);
797

798
    if (!type) {
799 800 801
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unknown storage vol type %d"),
                       vol->target.format);
802
        return NULL;
803
    }
804

805 806 807
    if (preallocate && vol->target.format != VIR_STORAGE_FILE_QCOW2) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("metadata preallocation only available with qcow2"));
808
        return NULL;
809
    }
810 811 812 813 814 815 816 817 818 819
    if (vol->target.compat && vol->target.format != VIR_STORAGE_FILE_QCOW2) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("compatibility option only available with qcow2"));
        return NULL;
    }
    if (vol->target.features && vol->target.format != VIR_STORAGE_FILE_QCOW2) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("format features only available with qcow2"));
        return NULL;
    }
820

821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840
    if (inputvol) {
        if (!(inputPath = inputvol->target.path)) {
            virReportError(VIR_ERR_INVALID_ARG, "%s",
                           _("missing input volume target path"));
            return NULL;
        }

        inputType = virStorageFileFormatTypeToString(inputvol->type == VIR_STORAGE_VOL_BLOCK ?
                                                     VIR_STORAGE_FILE_RAW :
                                                     inputvol->target.format);

        if (!inputType) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unknown storage vol type %d"),
                           inputvol->target.format);
            return NULL;
        }

    }

841
    if (vol->backingStore.path) {
842 843
        int accessRetCode = -1;
        char *absolutePath = NULL;
844

845 846
        backingType = virStorageFileFormatTypeToString(vol->backingStore.format);

847 848 849 850
        if (preallocate) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("metadata preallocation conflicts with backing"
                             " store"));
851
            return NULL;
852 853
        }

854 855 856 857 858
        /* XXX: Not strictly required: qemu-img has an option a different
         * backing store, not really sure what use it serves though, and it
         * may cause issues with lvm. Untested essentially.
         */
        if (inputvol &&
859 860 861
            STRNEQ_NULLABLE(inputvol->backingStore.path, vol->backingStore.path)) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("a different backing store cannot be specified."));
862
            return NULL;
863 864 865
        }

        if (backingType == NULL) {
866 867 868
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unknown storage vol backing store type %d"),
                           vol->backingStore.format);
869
            return NULL;
870
        }
871 872 873 874 875 876

        /* Convert relative backing store paths to absolute paths for access
         * validation.
         */
        if ('/' != *(vol->backingStore.path) &&
            virAsprintf(&absolutePath, "%s/%s", pool->def->target.path,
877
                        vol->backingStore.path) < 0)
878
            return NULL;
879 880 881 882
        accessRetCode = access(absolutePath ? absolutePath
                               : vol->backingStore.path, R_OK);
        VIR_FREE(absolutePath);
        if (accessRetCode != 0) {
883
            virReportSystemError(errno,
884 885
                                 _("inaccessible backing store volume %s"),
                                 vol->backingStore.path);
886
            return NULL;
887 888 889
        }
    }

890
    if (do_encryption) {
891 892
        virStorageEncryptionPtr enc;

893 894
        if (vol->target.format != VIR_STORAGE_FILE_QCOW &&
            vol->target.format != VIR_STORAGE_FILE_QCOW2) {
895 896 897
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("qcow volume encryption unsupported with "
                             "volume format %s"), type);
898
            return NULL;
899
        }
900 901 902
        enc = vol->target.encryption;
        if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW &&
            enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_DEFAULT) {
903 904 905
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("unsupported volume encryption format %d"),
                           vol->target.encryption->format);
906
            return NULL;
907
        }
908
        if (enc->nsecrets > 1) {
909 910
            virReportError(VIR_ERR_XML_ERROR, "%s",
                           _("too many secrets for qcow encryption"));
911
            return NULL;
912
        }
913 914 915
        if (enc->format == VIR_STORAGE_ENCRYPTION_FORMAT_DEFAULT ||
            enc->nsecrets == 0) {
            if (virStorageGenerateQcowEncryption(conn, vol) < 0)
916
                return NULL;
917
        }
918 919
    }

920
    /* Size in KB */
921
    size_arg = VIR_DIV_UP(vol->capacity, 1024);
922

923 924
    cmd = virCommandNew(create_tool);

925 926
    convert = !!inputvol;
    backing = !inputvol && vol->backingStore.path;
927

928 929 930
    if (convert)
        virCommandAddArgList(cmd, "convert", "-f", inputType, "-O", type, NULL);
    else
931
        virCommandAddArgList(cmd, "create", "-f", type, NULL);
932

933 934 935
    if (backing)
        virCommandAddArgList(cmd, "-b", vol->backingStore.path, NULL);

936 937 938 939 940
    if (imgformat >= QEMU_IMG_BACKING_FORMAT_OPTIONS) {
        if (vol->target.format == VIR_STORAGE_FILE_QCOW2 && !compat &&
            imgformat == QEMU_IMG_BACKING_FORMAT_OPTIONS_COMPAT)
            compat = "0.10";

941 942
        if (virStorageBackendCreateQemuImgOpts(&opts,
                                               backing ? backingType : NULL,
943 944
                                               do_encryption, preallocate,
                                               vol->target.format,
945
                                               compat,
946 947
                                               vol->target.features) < 0) {
            virCommandFree(cmd);
948
            return NULL;
949
        }
950 951 952 953 954 955 956 957 958 959
        if (opts)
            virCommandAddArgList(cmd, "-o", opts, NULL);
        VIR_FREE(opts);
    } else {
        if (backing) {
            if (imgformat == QEMU_IMG_BACKING_FORMAT_FLAG)
                virCommandAddArgList(cmd, "-F", backingType, NULL);
            else
                VIR_DEBUG("Unable to set backing store format for %s with %s",
                          vol->target.path, create_tool);
960
        }
961 962
        if (do_encryption)
            virCommandAddArg(cmd, "-e");
963
    }
964

965 966 967 968 969 970
    if (convert)
        virCommandAddArg(cmd, inputPath);
    virCommandAddArg(cmd, vol->target.path);
    if (!convert)
        virCommandAddArgFormat(cmd, "%lluK", size_arg);

971 972 973 974 975 976 977 978 979 980 981
    return cmd;
}

static int
virStorageBackendCreateQemuImg(virConnectPtr conn,
                               virStoragePoolObjPtr pool,
                               virStorageVolDefPtr vol,
                               virStorageVolDefPtr inputvol,
                               unsigned int flags)
{
    int ret = -1;
982
    char *create_tool;
983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007
    int imgformat;
    virCommandPtr cmd;

    virCheckFlags(VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA, -1);

    /* KVM is usually ahead of qemu on features, so try that first */
    create_tool = virFindFileInPath("kvm-img");
    if (!create_tool)
        create_tool = virFindFileInPath("qemu-img");

    if (!create_tool) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("unable to find kvm-img or qemu-img"));
        return -1;
    }

    imgformat = virStorageBackendQEMUImgBackingFormat(create_tool);
    if (imgformat < 0)
        goto cleanup;

    cmd = virStorageBackendCreateQemuImgCmd(conn, pool, vol, inputvol, flags,
                                            create_tool, imgformat);
    if (!cmd)
        goto cleanup;

1008
    ret = virStorageBackendCreateExecCommand(pool, vol, cmd);
1009 1010

    virCommandFree(cmd);
1011
cleanup:
1012
    VIR_FREE(create_tool);
1013
    return ret;
1014 1015 1016 1017 1018 1019 1020
}

/*
 * Xen removed the fully-functional qemu-img, and replaced it
 * with a partially functional qcow-create. Go figure ??!?
 */
static int
1021
virStorageBackendCreateQcowCreate(virConnectPtr conn ATTRIBUTE_UNUSED,
1022
                                  virStoragePoolObjPtr pool,
1023
                                  virStorageVolDefPtr vol,
1024
                                  virStorageVolDefPtr inputvol,
E
Eric Blake 已提交
1025
                                  unsigned int flags)
1026
{
1027
    int ret;
E
Eric Blake 已提交
1028
    char *size;
1029
    virCommandPtr cmd;
1030

1031 1032 1033 1034 1035 1036 1037 1038
    virCheckFlags(VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA, -1);

    if (flags & VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("metadata preallocation is not supported with "
                         "qcow-create"));
        return -1;
    }
E
Eric Blake 已提交
1039

1040
    if (inputvol) {
1041 1042
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("cannot copy from volume with qcow-create"));
1043 1044 1045
        return -1;
    }

1046
    if (vol->target.format != VIR_STORAGE_FILE_QCOW2) {
1047 1048 1049
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unsupported storage vol type %d"),
                       vol->target.format);
1050 1051 1052
        return -1;
    }
    if (vol->backingStore.path != NULL) {
1053 1054 1055
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("copy-on-write image not supported with "
                         "qcow-create"));
1056 1057
        return -1;
    }
1058
    if (vol->target.encryption != NULL) {
1059 1060 1061
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       "%s", _("encrypted volumes not supported with "
                               "qcow-create"));
1062 1063
        return -1;
    }
1064 1065

    /* Size in MB - yes different units to qemu-img :-( */
1066
    if (virAsprintf(&size, "%llu",
1067
                    VIR_DIV_UP(vol->capacity, (1024 * 1024))) < 0)
E
Eric Blake 已提交
1068
        return -1;
1069

1070
    cmd = virCommandNewArgList("qcow-create", size, vol->target.path, NULL);
1071

1072 1073
    ret = virStorageBackendCreateExecCommand(pool, vol, cmd);
    virCommandFree(cmd);
E
Eric Blake 已提交
1074
    VIR_FREE(size);
1075

1076
    return ret;
1077 1078
}

1079
virStorageBackendBuildVolFrom
1080
virStorageBackendFSImageToolTypeToFunc(int tool_type)
1081 1082 1083 1084 1085 1086 1087 1088
{
    switch (tool_type) {
    case TOOL_KVM_IMG:
    case TOOL_QEMU_IMG:
        return virStorageBackendCreateQemuImg;
    case TOOL_QCOW_CREATE:
        return virStorageBackendCreateQcowCreate;
    default:
1089 1090 1091
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Unknown file create tool type '%d'."),
                       tool_type);
1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
    }

    return NULL;
}

int
virStorageBackendFindFSImageTool(char **tool)
{
    int tool_type = -1;
    char *tmp = NULL;

    if ((tmp = virFindFileInPath("kvm-img")) != NULL) {
        tool_type = TOOL_KVM_IMG;
    } else if ((tmp = virFindFileInPath("qemu-img")) != NULL) {
        tool_type = TOOL_QEMU_IMG;
    } else if ((tmp = virFindFileInPath("qcow-create")) != NULL) {
        tool_type = TOOL_QCOW_CREATE;
    }

    if (tool)
        *tool = tmp;
    else
        VIR_FREE(tmp);

    return tool_type;
}

1119
virStorageBackendBuildVolFrom
1120
virStorageBackendGetBuildVolFromFunction(virStorageVolDefPtr vol,
1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131
                                         virStorageVolDefPtr inputvol)
{
    int tool_type;

    if (!inputvol)
        return NULL;

    /* If either volume is a non-raw file vol, we need to use an external
     * tool for converting
     */
    if ((vol->type == VIR_STORAGE_VOL_FILE &&
1132
         vol->target.format != VIR_STORAGE_FILE_RAW) ||
1133
        (inputvol->type == VIR_STORAGE_VOL_FILE &&
1134
         inputvol->target.format != VIR_STORAGE_FILE_RAW)) {
1135

D
Daniel P. Berrange 已提交
1136
        if ((tool_type = virStorageBackendFindFSImageTool(NULL)) < 0) {
1137 1138 1139
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("creation of non-raw file images is "
                             "not supported without qemu-img."));
1140 1141 1142
            return NULL;
        }

1143
        return virStorageBackendFSImageToolTypeToFunc(tool_type);
1144 1145
    }

1146 1147 1148 1149
    if (vol->type == VIR_STORAGE_VOL_BLOCK)
        return virStorageBackendCreateBlockFrom;
    else
        return virStorageBackendCreateRaw;
1150
}
1151

1152

1153
virStorageBackendPtr
1154 1155
virStorageBackendForType(int type)
{
1156
    size_t i;
1157
    for (i = 0; backends[i]; i++)
1158 1159 1160
        if (backends[i]->type == type)
            return backends[i];

1161
    virReportError(VIR_ERR_INTERNAL_ERROR,
1162 1163
                   _("missing backend for pool type %d (%s)"),
                   type, NULLSTR(virStoragePoolTypeToString(type)));
1164 1165 1166 1167
    return NULL;
}


1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
virStorageFileBackendPtr
virStorageFileBackendForType(int type,
                             int protocol)
{
    size_t i;

    for (i = 0; fileBackends[i]; i++) {
        if (fileBackends[i]->type == type) {
            if (type == VIR_DOMAIN_DISK_TYPE_NETWORK &&
                fileBackends[i]->protocol != protocol)
                continue;

            return fileBackends[i];
        }
    }

    if (type == VIR_DOMAIN_DISK_TYPE_NETWORK) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("missing storage backend for network files "
                         "using %s protocol"),
                       virDomainDiskProtocolTypeToString(protocol));
    } else {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("missing storage backend for '%s' storage"),
                       virDomainDiskTypeToString(type));
    }

    return NULL;
}


1199 1200 1201
/*
 * Allows caller to silently ignore files with improper mode
 *
1202 1203
 * Returns -1 on error, -2 if file mode is unexpected or the
 * volume is a dangling symbolic link.
1204
 */
1205
int
E
Eric Blake 已提交
1206 1207
virStorageBackendVolOpenCheckMode(const char *path, struct stat *sb,
                                  unsigned int flags)
1208
{
1209
    int fd, mode = 0;
1210
    char *base = last_component(path);
1211

E
Eric Blake 已提交
1212
    if (lstat(path, sb) < 0) {
1213 1214 1215 1216 1217 1218
        virReportSystemError(errno,
                             _("cannot stat file '%s'"),
                             path);
        return -1;
    }

E
Eric Blake 已提交
1219
    if (S_ISFIFO(sb->st_mode)) {
1220 1221
        VIR_WARN("ignoring FIFO '%s'", path);
        return -2;
E
Eric Blake 已提交
1222
    } else if (S_ISSOCK(sb->st_mode)) {
1223 1224 1225 1226
        VIR_WARN("ignoring socket '%s'", path);
        return -2;
    }

1227 1228 1229 1230 1231 1232
    /* O_NONBLOCK should only matter during open() for fifos and
     * sockets, which we already filtered; but using it prevents a
     * TOCTTOU race.  However, later on we will want to read() the
     * header from this fd, and virFileRead* routines require a
     * blocking fd, so fix it up after verifying we avoided a
     * race.  */
1233
    if ((fd = open(path, O_RDONLY|O_NONBLOCK|O_NOCTTY)) < 0) {
1234
        if ((errno == ENOENT || errno == ELOOP) &&
E
Eric Blake 已提交
1235
            S_ISLNK(sb->st_mode)) {
1236 1237 1238 1239
            VIR_WARN("ignoring dangling symlink '%s'", path);
            return -2;
        }

1240
        virReportSystemError(errno,
1241
                             _("cannot open volume '%s'"),
1242
                             path);
1243 1244 1245
        return -1;
    }

E
Eric Blake 已提交
1246
    if (fstat(fd, sb) < 0) {
1247 1248 1249 1250 1251 1252 1253
        virReportSystemError(errno,
                             _("cannot stat file '%s'"),
                             path);
        VIR_FORCE_CLOSE(fd);
        return -1;
    }

1254
    if (S_ISREG(sb->st_mode)) {
1255
        mode = VIR_STORAGE_VOL_OPEN_REG;
1256
    } else if (S_ISCHR(sb->st_mode)) {
1257
        mode = VIR_STORAGE_VOL_OPEN_CHAR;
1258
    } else if (S_ISBLK(sb->st_mode)) {
1259
        mode = VIR_STORAGE_VOL_OPEN_BLOCK;
1260
    } else if (S_ISDIR(sb->st_mode)) {
1261 1262 1263 1264 1265 1266 1267 1268
        mode = VIR_STORAGE_VOL_OPEN_DIR;

        if (STREQ(base, ".") ||
            STREQ(base, "..")) {
            VIR_FORCE_CLOSE(fd);
            VIR_INFO("Skipping special dir '%s'", base);
            return -2;
        }
1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279
    } else {
        VIR_WARN("ignoring unexpected type for file '%s'", path);
        VIR_FORCE_CLOSE(fd);
        return -2;
    }

    if (virSetBlocking(fd, true) < 0) {
        virReportSystemError(errno, _("unable to set blocking mode for '%s'"),
                             path);
        VIR_FORCE_CLOSE(fd);
        return -2;
1280
    }
1281 1282

    if (!(mode & flags)) {
1283
        VIR_FORCE_CLOSE(fd);
1284
        VIR_INFO("Skipping volume '%s'", path);
1285 1286

        if (mode & VIR_STORAGE_VOL_OPEN_ERROR) {
1287 1288
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unexpected storage mode for '%s'"), path);
1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299
            return -1;
        }

        return -2;
    }

    return fd;
}

int virStorageBackendVolOpen(const char *path)
{
E
Eric Blake 已提交
1300 1301
    struct stat sb;
    return virStorageBackendVolOpenCheckMode(path, &sb,
1302 1303 1304 1305 1306 1307
                                             VIR_STORAGE_VOL_OPEN_DEFAULT);
}

int
virStorageBackendUpdateVolTargetInfo(virStorageVolTargetPtr target,
                                     unsigned long long *allocation,
1308 1309
                                     unsigned long long *capacity,
                                     unsigned int openflags)
1310 1311
{
    int ret, fd;
E
Eric Blake 已提交
1312
    struct stat sb;
1313

E
Eric Blake 已提交
1314
    if ((ret = virStorageBackendVolOpenCheckMode(target->path, &sb,
1315
                                                 openflags)) < 0)
1316 1317 1318
        return ret;

    fd = ret;
1319
    ret = virStorageBackendUpdateVolTargetInfoFD(target,
1320
                                                 fd,
E
Eric Blake 已提交
1321
                                                 &sb,
1322 1323
                                                 allocation,
                                                 capacity);
1324

1325
    VIR_FORCE_CLOSE(fd);
1326 1327 1328 1329

    return ret;
}

1330
int
1331 1332 1333
virStorageBackendUpdateVolInfoFlags(virStorageVolDefPtr vol,
                                    int withCapacity,
                                    unsigned int openflags)
1334 1335 1336
{
    int ret;

1337
    if ((ret = virStorageBackendUpdateVolTargetInfo(&vol->target,
1338 1339 1340
                                    &vol->allocation,
                                    withCapacity ? &vol->capacity : NULL,
                                    openflags)) < 0)
1341 1342 1343
        return ret;

    if (vol->backingStore.path &&
1344
        (ret = virStorageBackendUpdateVolTargetInfo(&vol->backingStore,
1345 1346
                                            NULL, NULL,
                                            VIR_STORAGE_VOL_OPEN_DEFAULT)) < 0)
1347 1348 1349 1350 1351
        return ret;

    return 0;
}

1352 1353 1354 1355 1356 1357 1358
int virStorageBackendUpdateVolInfo(virStorageVolDefPtr vol,
                                   int withCapacity)
{
    return virStorageBackendUpdateVolInfoFlags(vol, withCapacity,
                                               VIR_STORAGE_VOL_OPEN_DEFAULT);
}

1359 1360 1361
/*
 * virStorageBackendUpdateVolTargetInfoFD:
 * @target: target definition ptr of volume to update
1362 1363
 * @fd: fd of storage volume to update, via virStorageBackendOpenVol*, or -1
 * @sb: details about file (must match @fd, if that is provided)
1364 1365 1366
 * @allocation: If not NULL, updated allocation information will be stored
 * @capacity: If not NULL, updated capacity info will be stored
 *
1367
 * Returns 0 for success, -1 on a legitimate error condition.
1368
 */
1369
int
1370
virStorageBackendUpdateVolTargetInfoFD(virStorageVolTargetPtr target,
1371
                                       int fd,
E
Eric Blake 已提交
1372
                                       struct stat *sb,
1373 1374
                                       unsigned long long *allocation,
                                       unsigned long long *capacity)
1375
{
1376
#if WITH_SELINUX
1377 1378 1379
    security_context_t filecon = NULL;
#endif

1380
    if (allocation) {
E
Eric Blake 已提交
1381
        if (S_ISREG(sb->st_mode)) {
1382
#ifndef WIN32
E
Eric Blake 已提交
1383
            *allocation = (unsigned long long)sb->st_blocks *
1384
                          (unsigned long long)DEV_BSIZE;
D
Daniel P. Berrange 已提交
1385
#else
E
Eric Blake 已提交
1386
            *allocation = sb->st_size;
D
Daniel P. Berrange 已提交
1387
#endif
1388 1389 1390 1391
            /* Regular files may be sparse, so logical size (capacity) is not same
             * as actual allocation above
             */
            if (capacity)
E
Eric Blake 已提交
1392 1393
                *capacity = sb->st_size;
        } else if (S_ISDIR(sb->st_mode)) {
1394 1395 1396 1397
            *allocation = 0;
            if (capacity)
                *capacity = 0;

1398
        } else if (fd >= 0) {
1399
            off_t end;
E
Eric Blake 已提交
1400
            /* XXX this is POSIX compliant, but doesn't work for CHAR files,
1401 1402 1403 1404 1405 1406
             * only BLOCK. There is a Linux specific ioctl() for getting
             * size of both CHAR / BLOCK devices we should check for in
             * configure
             */
            end = lseek(fd, 0, SEEK_END);
            if (end == (off_t)-1) {
1407
                virReportSystemError(errno,
1408 1409 1410 1411 1412 1413 1414
                                     _("cannot seek to end of file '%s'"),
                                     target->path);
                return -1;
            }
            *allocation = end;
            if (capacity)
                *capacity = end;
1415 1416 1417
        }
    }

E
Eric Blake 已提交
1418 1419 1420
    target->perms.mode = sb->st_mode & S_IRWXUGO;
    target->perms.uid = sb->st_uid;
    target->perms.gid = sb->st_gid;
1421

1422
    if (!target->timestamps && VIR_ALLOC(target->timestamps) < 0)
1423
        return -1;
E
Eric Blake 已提交
1424 1425 1426 1427
    target->timestamps->atime = get_stat_atime(sb);
    target->timestamps->btime = get_stat_birthtime(sb);
    target->timestamps->ctime = get_stat_ctime(sb);
    target->timestamps->mtime = get_stat_mtime(sb);
1428

1429
    VIR_FREE(target->perms.label);
1430

1431
#if WITH_SELINUX
1432
    /* XXX: make this a security driver call */
1433 1434 1435 1436 1437 1438 1439 1440
    if (fd >= 0) {
        if (fgetfilecon_raw(fd, &filecon) == -1) {
            if (errno != ENODATA && errno != ENOTSUP) {
                virReportSystemError(errno,
                                     _("cannot get file context of '%s'"),
                                     target->path);
                return -1;
            }
1441
        } else {
1442 1443 1444 1445
            if (VIR_STRDUP(target->perms.label, filecon) < 0) {
                freecon(filecon);
                return -1;
            }
1446
            freecon(filecon);
1447
        }
1448 1449 1450 1451 1452 1453
    }
#endif

    return 0;
}

D
Dave Allan 已提交
1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488

struct diskType {
    int part_table_type;
    unsigned short offset;
    unsigned short length;
    unsigned long long magic;
};


static struct diskType const disk_types[] = {
    { VIR_STORAGE_POOL_DISK_LVM2, 0x218, 8, 0x31303020324D564CULL },
    { VIR_STORAGE_POOL_DISK_GPT,  0x200, 8, 0x5452415020494645ULL },
    { VIR_STORAGE_POOL_DISK_DVH,  0x0,   4, 0x41A9E50BULL },
    { VIR_STORAGE_POOL_DISK_MAC,  0x0,   2, 0x5245ULL },
    { VIR_STORAGE_POOL_DISK_BSD,  0x40,  4, 0x82564557ULL },
    { VIR_STORAGE_POOL_DISK_SUN,  0x1fc, 2, 0xBEDAULL },
    /*
     * NOTE: pc98 is funky; the actual signature is 0x55AA (just like dos), so
     * we can't use that.  At the moment I'm relying on the "dummy" IPL
     * bootloader data that comes from parted.  Luckily, the chances of running
     * into a pc98 machine running libvirt are approximately nil.
     */
    /*{ 0x1fe, 2, 0xAA55UL },*/
    { VIR_STORAGE_POOL_DISK_PC98, 0x0,   8, 0x314C5049000000CBULL },
    /*
     * NOTE: the order is important here; some other disk types (like GPT and
     * and PC98) also have 0x55AA at this offset.  For that reason, the DOS
     * one must be the last one.
     */
    { VIR_STORAGE_POOL_DISK_DOS,  0x1fe, 2, 0xAA55ULL },
    { -1,                         0x0,   0, 0x0ULL },
};


int
1489 1490
virStorageBackendDetectBlockVolFormatFD(virStorageVolTargetPtr target,
                                        int fd)
D
Dave Allan 已提交
1491
{
1492
    size_t i;
D
Dave Allan 已提交
1493 1494 1495 1496 1497 1498 1499 1500 1501
    off_t start;
    unsigned char buffer[1024];
    ssize_t bytes;

    /* make sure to set the target format "unknown" to begin with */
    target->format = VIR_STORAGE_POOL_DISK_UNKNOWN;

    start = lseek(fd, 0, SEEK_SET);
    if (start < 0) {
1502
        virReportSystemError(errno,
D
Dave Allan 已提交
1503 1504 1505 1506 1507 1508
                             _("cannot seek to beginning of file '%s'"),
                             target->path);
        return -1;
    }
    bytes = saferead(fd, buffer, sizeof(buffer));
    if (bytes < 0) {
1509
        virReportSystemError(errno,
D
Dave Allan 已提交
1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528
                             _("cannot read beginning of file '%s'"),
                             target->path);
        return -1;
    }

    for (i = 0; disk_types[i].part_table_type != -1; i++) {
        if (disk_types[i].offset + disk_types[i].length > bytes)
            continue;
        if (memcmp(buffer+disk_types[i].offset, &disk_types[i].magic,
            disk_types[i].length) == 0) {
            target->format = disk_types[i].part_table_type;
            break;
        }
    }

    return 0;
}


1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540
/*
 * Given a volume path directly in /dev/XXX, iterate over the
 * entries in the directory pool->def->target.path and find the
 * first symlink pointing to the volume path.
 *
 * If, the target.path is /dev/, then return the original volume
 * path.
 *
 * If no symlink is found, then return the original volume path
 *
 * Typically target.path is one of the /dev/disk/by-XXX dirs
 * with stable paths.
1541
 *
1542
 * If 'loop' is true, we use a timeout loop to give dynamic paths
1543
 * a change to appear.
1544 1545
 */
char *
1546
virStorageBackendStablePath(virStoragePoolObjPtr pool,
1547
                            const char *devpath,
1548
                            bool loop)
1549 1550 1551
{
    DIR *dh;
    struct dirent *dent;
1552
    char *stablepath;
1553
    int opentries = 0;
1554
    int retry = 0;
1555 1556 1557 1558 1559

    /* Short circuit if pool has no target, or if its /dev */
    if (pool->def->target.path == NULL ||
        STREQ(pool->def->target.path, "/dev") ||
        STREQ(pool->def->target.path, "/dev/"))
1560
        goto ret_strdup;
1561

1562
    /* Skip whole thing for a pool which isn't in /dev
1563
     * so we don't mess filesystem/dir based pools
1564 1565 1566 1567
     */
    if (!STRPREFIX(pool->def->target.path, "/dev"))
        goto ret_strdup;

1568 1569 1570 1571
    /* Logical pools are under /dev but already have stable paths */
    if (pool->def->type == VIR_STORAGE_POOL_LOGICAL)
        goto ret_strdup;

1572 1573 1574
    /* We loop here because /dev/disk/by-{id,path} may not have existed
     * before we started this operation, so we have to give it some time to
     * get created.
1575
     */
1576
 reopen:
1577
    if ((dh = opendir(pool->def->target.path)) == NULL) {
1578
        opentries++;
1579
        if (loop && errno == ENOENT && opentries < 50) {
1580 1581 1582
            usleep(100 * 1000);
            goto reopen;
        }
1583
        virReportSystemError(errno,
1584 1585
                             _("cannot read dir '%s'"),
                             pool->def->target.path);
1586 1587 1588
        return NULL;
    }

1589 1590 1591
    /* The pool is pointing somewhere like /dev/disk/by-path
     * or /dev/disk/by-id, so we need to check all symlinks in
     * the target directory and figure out which one points
1592 1593
     * to this device node.
     *
1594
     * And it might need some time till the stable path shows
1595
     * up, so add timeout to retry here.
1596
     */
1597
 retry:
1598 1599 1600 1601
    while ((dent = readdir(dh)) != NULL) {
        if (dent->d_name[0] == '.')
            continue;

1602 1603 1604
        if (virAsprintf(&stablepath, "%s/%s",
                        pool->def->target.path,
                        dent->d_name) == -1) {
1605 1606 1607 1608 1609 1610 1611 1612 1613
            closedir(dh);
            return NULL;
        }

        if (virFileLinkPointsTo(stablepath, devpath)) {
            closedir(dh);
            return stablepath;
        }

1614
        VIR_FREE(stablepath);
1615 1616
    }

1617
    if (loop && ++retry < 100) {
1618 1619 1620 1621
        usleep(100 * 1000);
        goto retry;
    }

1622 1623
    closedir(dh);

1624
 ret_strdup:
1625 1626 1627
    /* Couldn't find any matching stable link so give back
     * the original non-stable dev path
     */
1628

1629
    ignore_value(VIR_STRDUP(stablepath, devpath));
1630 1631

    return stablepath;
1632 1633
}

D
Daniel P. Berrange 已提交
1634

1635
#ifndef WIN32
1636 1637 1638 1639
/*
 * Run an external program.
 *
 * Read its output and apply a series of regexes to each line
R
Richard W.M. Jones 已提交
1640
 * When the entire set of regexes has matched consecutively
1641 1642 1643
 * then run a callback passing in all the matches
 */
int
1644
virStorageBackendRunProgRegex(virStoragePoolObjPtr pool,
1645
                              virCommandPtr cmd,
1646 1647 1648 1649
                              int nregex,
                              const char **regex,
                              int *nvars,
                              virStorageBackendListVolRegexFunc func,
1650
                              void *data, const char *prefix)
1651
{
1652
    int fd = -1, err, ret = -1;
1653 1654 1655 1656
    FILE *list = NULL;
    regex_t *reg;
    regmatch_t *vars = NULL;
    char line[1024];
1657 1658
    int maxReg = 0;
    size_t i, j;
1659 1660 1661 1662
    int totgroups = 0, ngroup = 0, maxvars = 0;
    char **groups;

    /* Compile all regular expressions */
1663
    if (VIR_ALLOC_N(reg, nregex) < 0)
1664 1665
        return -1;

1666
    for (i = 0; i < nregex; i++) {
1667 1668 1669 1670
        err = regcomp(&reg[i], regex[i], REG_EXTENDED);
        if (err != 0) {
            char error[100];
            regerror(err, &reg[i], error, sizeof(error));
1671 1672
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Failed to compile regex %s"), error);
1673
            for (j = 0; j < i; j++)
1674
                regfree(&reg[j]);
1675
            VIR_FREE(reg);
1676 1677 1678 1679 1680 1681 1682 1683 1684 1685
            return -1;
        }

        totgroups += nvars[i];
        if (nvars[i] > maxvars)
            maxvars = nvars[i];

    }

    /* Storage for matched variables */
1686
    if (VIR_ALLOC_N(groups, totgroups) < 0)
1687
        goto cleanup;
1688
    if (VIR_ALLOC_N(vars, maxvars+1) < 0)
1689 1690
        goto cleanup;

1691 1692
    virCommandSetOutputFD(cmd, &fd);
    if (virCommandRunAsync(cmd, NULL) < 0) {
1693 1694 1695
        goto cleanup;
    }

1696
    if ((list = VIR_FDOPEN(fd, "r")) == NULL) {
1697 1698
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot read fd"));
1699 1700 1701 1702
        goto cleanup;
    }

    while (fgets(line, sizeof(line), list) != NULL) {
1703
        char *p = NULL;
1704 1705 1706 1707 1708
        /* Strip trailing newline */
        int len = strlen(line);
        if (len && line[len-1] == '\n')
            line[len-1] = '\0';

1709 1710 1711 1712 1713 1714
        /* ignore any command prefix */
        if (prefix)
            p = STRSKIP(line, prefix);
        if (!p)
            p = line;

1715
        for (i = 0; i <= maxReg && i < nregex; i++) {
1716
            if (regexec(&reg[i], p, nvars[i]+1, vars, 0) == 0) {
1717 1718 1719 1720 1721 1722
                maxReg++;

                if (i == 0)
                    ngroup = 0;

                /* NULL terminate each captured group in the line */
1723
                for (j = 0; j < nvars[i]; j++) {
1724
                    /* NB vars[0] is the full pattern, so we offset j by 1 */
1725
                    p[vars[j+1].rm_eo] = '\0';
1726
                    if (VIR_STRDUP(groups[ngroup++], p + vars[j+1].rm_so) < 0)
1727 1728 1729 1730 1731
                        goto cleanup;
                }

                /* We're matching on the last regex, so callback time */
                if (i == (nregex-1)) {
1732
                    if (((*func)(pool, groups, data)) < 0)
1733 1734 1735
                        goto cleanup;

                    /* Release matches & restart to matching the first regex */
1736
                    for (j = 0; j < totgroups; j++)
1737
                        VIR_FREE(groups[j]);
1738 1739 1740 1741 1742 1743 1744
                    maxReg = 0;
                    ngroup = 0;
                }
            }
        }
    }

1745
    ret = virCommandWait(cmd, NULL);
1746
cleanup:
1747
    if (groups) {
1748
        for (j = 0; j < totgroups; j++)
1749 1750
            VIR_FREE(groups[j]);
        VIR_FREE(groups);
1751
    }
1752
    VIR_FREE(vars);
1753

1754
    for (i = 0; i < nregex; i++)
1755 1756
        regfree(&reg[i]);

1757
    VIR_FREE(reg);
1758

1759 1760
    VIR_FORCE_FCLOSE(list);
    VIR_FORCE_CLOSE(fd);
1761

1762
    return ret;
1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776
}

/*
 * Run an external program and read from its standard output
 * a stream of tokens from IN_STREAM, applying FUNC to
 * each successive sequence of N_COLUMNS tokens.
 * If FUNC returns < 0, stop processing input and return -1.
 * Return -1 if N_COLUMNS == 0.
 * Return -1 upon memory allocation error.
 * If the number of input tokens is not a multiple of N_COLUMNS,
 * then the final FUNC call will specify a number smaller than N_COLUMNS.
 * If there are no input tokens (empty input), call FUNC with N_COLUMNS == 0.
 */
int
1777
virStorageBackendRunProgNul(virStoragePoolObjPtr pool,
1778
                            virCommandPtr cmd,
1779 1780 1781 1782 1783
                            size_t n_columns,
                            virStorageBackendListVolNulFunc func,
                            void *data)
{
    size_t n_tok = 0;
1784
    int fd = -1;
1785 1786
    FILE *fp = NULL;
    char **v;
1787
    int ret = -1;
1788
    size_t i;
1789 1790 1791 1792

    if (n_columns == 0)
        return -1;

1793
    if (VIR_ALLOC_N(v, n_columns) < 0)
1794 1795 1796 1797
        return -1;
    for (i = 0; i < n_columns; i++)
        v[i] = NULL;

1798 1799
    virCommandSetOutputFD(cmd, &fd);
    if (virCommandRunAsync(cmd, NULL) < 0) {
1800 1801 1802
        goto cleanup;
    }

1803
    if ((fp = VIR_FDOPEN(fd, "r")) == NULL) {
1804 1805
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot open file using fd"));
1806 1807 1808 1809 1810 1811 1812 1813
        goto cleanup;
    }

    while (1) {
        char *buf = NULL;
        size_t buf_len = 0;
        /* Be careful: even when it returns -1,
           this use of getdelim allocates memory.  */
1814
        ssize_t tok_len = getdelim(&buf, &buf_len, 0, fp);
1815 1816 1817 1818
        v[n_tok] = buf;
        if (tok_len < 0) {
            /* Maybe EOF, maybe an error.
               If n_tok > 0, then we know it's an error.  */
1819
            if (n_tok && func(pool, n_tok, v, data) < 0)
1820 1821 1822 1823 1824
                goto cleanup;
            break;
        }
        ++n_tok;
        if (n_tok == n_columns) {
1825
            if (func(pool, n_tok, v, data) < 0)
1826 1827 1828
                goto cleanup;
            n_tok = 0;
            for (i = 0; i < n_columns; i++) {
1829
                VIR_FREE(v[i]);
1830 1831 1832 1833
            }
        }
    }

1834 1835 1836
    if (feof(fp) < 0) {
        virReportSystemError(errno, "%s",
                             _("read error on pipe"));
1837 1838
        goto cleanup;
    }
1839

1840
    ret = virCommandWait(cmd, NULL);
1841 1842
 cleanup:
    for (i = 0; i < n_columns; i++)
1843 1844
        VIR_FREE(v[i]);
    VIR_FREE(v);
1845

1846 1847
    VIR_FORCE_FCLOSE(fp);
    VIR_FORCE_CLOSE(fd);
1848

1849
    return ret;
1850
}
D
Daniel P. Berrange 已提交
1851

1852
#else /* WIN32 */
D
Daniel P. Berrange 已提交
1853 1854 1855 1856 1857 1858 1859 1860 1861

int
virStorageBackendRunProgRegex(virConnectPtr conn,
                              virStoragePoolObjPtr pool ATTRIBUTE_UNUSED,
                              const char *const*prog ATTRIBUTE_UNUSED,
                              int nregex ATTRIBUTE_UNUSED,
                              const char **regex ATTRIBUTE_UNUSED,
                              int *nvars ATTRIBUTE_UNUSED,
                              virStorageBackendListVolRegexFunc func ATTRIBUTE_UNUSED,
1862
                              void *data ATTRIBUTE_UNUSED)
D
Daniel P. Berrange 已提交
1863
{
1864 1865
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("%s not implemented on Win32"), __FUNCTION__);
D
Daniel P. Berrange 已提交
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876
    return -1;
}

int
virStorageBackendRunProgNul(virConnectPtr conn,
                            virStoragePoolObjPtr pool ATTRIBUTE_UNUSED,
                            const char **prog ATTRIBUTE_UNUSED,
                            size_t n_columns ATTRIBUTE_UNUSED,
                            virStorageBackendListVolNulFunc func ATTRIBUTE_UNUSED,
                            void *data ATTRIBUTE_UNUSED)
{
1877 1878
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("%s not implemented on Win32"), __FUNCTION__);
D
Daniel P. Berrange 已提交
1879 1880
    return -1;
}
1881
#endif /* WIN32 */