1 /*- 2 * Copyright (c) 1999-2004 Poul-Henning Kamp 3 * Copyright (c) 1999 Michael Smith 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 */ 36 37 #include <sys/cdefs.h> 38 __FBSDID("$FreeBSD$"); 39 40 #include <sys/param.h> 41 #include <sys/conf.h> 42 #include <sys/cons.h> 43 #include <sys/jail.h> 44 #include <sys/kernel.h> 45 #include <sys/mac.h> 46 #include <sys/malloc.h> 47 #include <sys/mount.h> 48 #include <sys/mutex.h> 49 #include <sys/namei.h> 50 #include <sys/proc.h> 51 #include <sys/filedesc.h> 52 #include <sys/reboot.h> 53 #include <sys/syscallsubr.h> 54 #include <sys/sysproto.h> 55 #include <sys/sx.h> 56 #include <sys/sysctl.h> 57 #include <sys/sysent.h> 58 #include <sys/systm.h> 59 #include <sys/vnode.h> 60 61 #include <geom/geom.h> 62 63 #include <machine/stdarg.h> 64 65 #include "opt_rootdevname.h" 66 #include "opt_ddb.h" 67 #include "opt_mac.h" 68 69 #ifdef DDB 70 #include <ddb/ddb.h> 71 #endif 72 73 #define ROOTNAME "root_device" 74 #define VFS_MOUNTARG_SIZE_MAX (1024 * 64) 75 76 static void gets(char *cp); 77 static int vfs_domount(struct thread *td, const char *fstype, 78 char *fspath, int fsflags, void *fsdata); 79 static int vfs_mount_alloc(struct vnode *dvp, struct vfsconf *vfsp, 80 const char *fspath, struct thread *td, struct mount **mpp); 81 static int vfs_mountroot_ask(void); 82 static int vfs_mountroot_try(const char *mountfrom); 83 static int vfs_donmount(struct thread *td, int fsflags, 84 struct uio *fsoptions); 85 86 static int usermount = 0; 87 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, 88 "Unprivileged users may mount and unmount file systems"); 89 90 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 91 92 /* List of mounted filesystems. */ 93 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist); 94 95 /* For any iteration/modification of mountlist */ 96 struct mtx mountlist_mtx; 97 98 TAILQ_HEAD(vfsoptlist, vfsopt); 99 struct vfsopt { 100 TAILQ_ENTRY(vfsopt) link; 101 char *name; 102 void *value; 103 int len; 104 }; 105 106 /* 107 * The vnode of the system's root (/ in the filesystem, without chroot 108 * active.) 109 */ 110 struct vnode *rootvnode; 111 112 /* 113 * The root filesystem is detailed in the kernel environment variable 114 * vfs.root.mountfrom, which is expected to be in the general format 115 * 116 * <vfsname>:[<path>] 117 * vfsname := the name of a VFS known to the kernel and capable 118 * of being mounted as root 119 * path := disk device name or other data used by the filesystem 120 * to locate its physical store 121 */ 122 123 /* 124 * Global opts, taken by all filesystems 125 */ 126 static const char *global_opts[] = { 127 "fstype", 128 "fspath", 129 "ro", 130 "suid", 131 "exec", 132 NULL 133 }; 134 135 /* 136 * The root specifiers we will try if RB_CDROM is specified. 137 */ 138 static char *cdrom_rootdevnames[] = { 139 "cd9660:cd0", 140 "cd9660:acd0", 141 NULL 142 }; 143 144 /* legacy find-root code */ 145 char *rootdevnames[2] = {NULL, NULL}; 146 #ifndef ROOTDEVNAME 147 # define ROOTDEVNAME NULL 148 #endif 149 const char *ctrootdevname = ROOTDEVNAME; 150 151 /* 152 * --------------------------------------------------------------------- 153 * Functions for building and sanitizing the mount options 154 */ 155 156 /* Remove one mount option. */ 157 static void 158 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt) 159 { 160 161 TAILQ_REMOVE(opts, opt, link); 162 free(opt->name, M_MOUNT); 163 if (opt->value != NULL) 164 free(opt->value, M_MOUNT); 165 #ifdef INVARIANTS 166 else if (opt->len != 0) 167 panic("%s: mount option with NULL value but length != 0", 168 __func__); 169 #endif 170 free(opt, M_MOUNT); 171 } 172 173 /* Release all resources related to the mount options. */ 174 static void 175 vfs_freeopts(struct vfsoptlist *opts) 176 { 177 struct vfsopt *opt; 178 179 while (!TAILQ_EMPTY(opts)) { 180 opt = TAILQ_FIRST(opts); 181 vfs_freeopt(opts, opt); 182 } 183 free(opts, M_MOUNT); 184 } 185 186 /* 187 * Check if options are equal (with or without the "no" prefix). 188 */ 189 static int 190 vfs_equalopts(const char *opt1, const char *opt2) 191 { 192 193 /* "opt" vs. "opt" or "noopt" vs. "noopt" */ 194 if (strcmp(opt1, opt2) == 0) 195 return (1); 196 /* "noopt" vs. "opt" */ 197 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0) 198 return (1); 199 /* "opt" vs. "noopt" */ 200 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0) 201 return (1); 202 return (0); 203 } 204 205 /* 206 * If a mount option is specified several times, 207 * (with or without the "no" prefix) only keep 208 * the last occurence of it. 209 */ 210 static void 211 vfs_sanitizeopts(struct vfsoptlist *opts) 212 { 213 struct vfsopt *opt, *opt2, *tmp; 214 215 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) { 216 opt2 = TAILQ_PREV(opt, vfsoptlist, link); 217 while (opt2 != NULL) { 218 if (vfs_equalopts(opt->name, opt2->name)) { 219 tmp = TAILQ_PREV(opt2, vfsoptlist, link); 220 vfs_freeopt(opts, opt2); 221 opt2 = tmp; 222 } else { 223 opt2 = TAILQ_PREV(opt2, vfsoptlist, link); 224 } 225 } 226 } 227 } 228 229 /* 230 * Build a linked list of mount options from a struct uio. 231 */ 232 static int 233 vfs_buildopts(struct uio *auio, struct vfsoptlist **options) 234 { 235 struct vfsoptlist *opts; 236 struct vfsopt *opt; 237 size_t memused; 238 unsigned int i, iovcnt; 239 int error, namelen, optlen; 240 241 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 242 TAILQ_INIT(opts); 243 memused = 0; 244 iovcnt = auio->uio_iovcnt; 245 for (i = 0; i < iovcnt; i += 2) { 246 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 247 namelen = auio->uio_iov[i].iov_len; 248 optlen = auio->uio_iov[i + 1].iov_len; 249 opt->name = malloc(namelen, M_MOUNT, M_WAITOK); 250 opt->value = NULL; 251 opt->len = 0; 252 253 /* 254 * Do this early, so jumps to "bad" will free the current 255 * option. 256 */ 257 TAILQ_INSERT_TAIL(opts, opt, link); 258 memused += sizeof(struct vfsopt) + optlen + namelen; 259 260 /* 261 * Avoid consuming too much memory, and attempts to overflow 262 * memused. 263 */ 264 if (memused > VFS_MOUNTARG_SIZE_MAX || 265 optlen > VFS_MOUNTARG_SIZE_MAX || 266 namelen > VFS_MOUNTARG_SIZE_MAX) { 267 error = EINVAL; 268 goto bad; 269 } 270 271 if (auio->uio_segflg == UIO_SYSSPACE) { 272 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen); 273 } else { 274 error = copyin(auio->uio_iov[i].iov_base, opt->name, 275 namelen); 276 if (error) 277 goto bad; 278 } 279 /* Ensure names are null-terminated strings. */ 280 if (opt->name[namelen - 1] != '\0') { 281 error = EINVAL; 282 goto bad; 283 } 284 if (optlen != 0) { 285 opt->len = optlen; 286 opt->value = malloc(optlen, M_MOUNT, M_WAITOK); 287 if (auio->uio_segflg == UIO_SYSSPACE) { 288 bcopy(auio->uio_iov[i + 1].iov_base, opt->value, 289 optlen); 290 } else { 291 error = copyin(auio->uio_iov[i + 1].iov_base, 292 opt->value, optlen); 293 if (error) 294 goto bad; 295 } 296 } 297 } 298 vfs_sanitizeopts(opts); 299 *options = opts; 300 return (0); 301 bad: 302 vfs_freeopts(opts); 303 return (error); 304 } 305 306 /* 307 * Merge the old mount options with the new ones passed 308 * in the MNT_UPDATE case. 309 */ 310 static void 311 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts) 312 { 313 struct vfsopt *opt, *opt2, *new; 314 315 TAILQ_FOREACH(opt, opts, link) { 316 /* 317 * Check that this option hasn't been redefined 318 * nor cancelled with a "no" mount option. 319 */ 320 opt2 = TAILQ_FIRST(toopts); 321 while (opt2 != NULL) { 322 if (strcmp(opt2->name, opt->name) == 0) 323 goto next; 324 if (strncmp(opt2->name, "no", 2) == 0 && 325 strcmp(opt2->name + 2, opt->name) == 0) { 326 vfs_freeopt(toopts, opt2); 327 goto next; 328 } 329 opt2 = TAILQ_NEXT(opt2, link); 330 } 331 /* We want this option, duplicate it. */ 332 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 333 new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK); 334 strcpy(new->name, opt->name); 335 if (opt->len != 0) { 336 new->value = malloc(opt->len, M_MOUNT, M_WAITOK); 337 bcopy(opt->value, new->value, opt->len); 338 } else { 339 new->value = NULL; 340 } 341 new->len = opt->len; 342 TAILQ_INSERT_TAIL(toopts, new, link); 343 next: 344 continue; 345 } 346 } 347 348 /* 349 * --------------------------------------------------------------------- 350 * Mount a filesystem 351 */ 352 int 353 nmount(td, uap) 354 struct thread *td; 355 struct nmount_args /* { 356 struct iovec *iovp; 357 unsigned int iovcnt; 358 int flags; 359 } */ *uap; 360 { 361 struct uio *auio; 362 struct iovec *iov; 363 unsigned int i; 364 int error; 365 u_int iovcnt; 366 367 /* Kick out MNT_ROOTFS early as it is legal internally */ 368 if (uap->flags & MNT_ROOTFS) 369 return (EINVAL); 370 371 iovcnt = uap->iovcnt; 372 /* 373 * Check that we have an even number of iovec's 374 * and that we have at least two options. 375 */ 376 if ((iovcnt & 1) || (iovcnt < 4)) 377 return (EINVAL); 378 379 error = copyinuio(uap->iovp, iovcnt, &auio); 380 if (error) 381 return (error); 382 iov = auio->uio_iov; 383 for (i = 0; i < iovcnt; i++) { 384 if (iov->iov_len > MMAXOPTIONLEN) { 385 free(auio, M_IOV); 386 return (EINVAL); 387 } 388 iov++; 389 } 390 error = vfs_donmount(td, uap->flags, auio); 391 free(auio, M_IOV); 392 return (error); 393 } 394 395 /* 396 * --------------------------------------------------------------------- 397 * Various utility functions 398 */ 399 400 /* 401 * Allocate and initialize the mount point struct. 402 */ 403 static int 404 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, 405 const char *fspath, struct thread *td, struct mount **mpp) 406 { 407 struct mount *mp; 408 409 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 410 TAILQ_INIT(&mp->mnt_nvnodelist); 411 mp->mnt_nvnodelistsize = 0; 412 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF); 413 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 414 vfs_busy(mp, LK_NOWAIT, 0, td); 415 mp->mnt_op = vfsp->vfc_vfsops; 416 mp->mnt_vfc = vfsp; 417 vfsp->vfc_refcount++; 418 mp->mnt_stat.f_type = vfsp->vfc_typenum; 419 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 420 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 421 mp->mnt_vnodecovered = vp; 422 mp->mnt_cred = crdup(td->td_ucred); 423 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 424 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 425 mp->mnt_iosize_max = DFLTPHYS; 426 #ifdef MAC 427 mac_init_mount(mp); 428 mac_create_mount(td->td_ucred, mp); 429 #endif 430 *mpp = mp; 431 return (0); 432 } 433 434 /* 435 * Destroy the mount struct previously allocated by vfs_mount_alloc(). 436 */ 437 void 438 vfs_mount_destroy(struct mount *mp, struct thread *td) 439 { 440 441 mp->mnt_vfc->vfc_refcount--; 442 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) 443 panic("unmount: dangling vnode"); 444 vfs_unbusy(mp,td); 445 lockdestroy(&mp->mnt_lock); 446 mtx_destroy(&mp->mnt_mtx); 447 if (mp->mnt_kern_flag & MNTK_MWAIT) 448 wakeup(mp); 449 #ifdef MAC 450 mac_destroy_mount(mp); 451 #endif 452 if (mp->mnt_opt != NULL) 453 vfs_freeopts(mp->mnt_opt); 454 crfree(mp->mnt_cred); 455 free(mp, M_MOUNT); 456 } 457 458 static int 459 vfs_donmount(struct thread *td, int fsflags, struct uio *fsoptions) 460 { 461 struct vfsoptlist *optlist; 462 char *fstype, *fspath; 463 int error, fstypelen, fspathlen; 464 465 error = vfs_buildopts(fsoptions, &optlist); 466 if (error) 467 return (error); 468 469 /* 470 * We need these two options before the others, 471 * and they are mandatory for any filesystem. 472 * Ensure they are NUL terminated as well. 473 */ 474 fstypelen = 0; 475 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen); 476 if (error || fstype[fstypelen - 1] != '\0') { 477 error = EINVAL; 478 goto bail; 479 } 480 fspathlen = 0; 481 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen); 482 if (error || fspath[fspathlen - 1] != '\0') { 483 error = EINVAL; 484 goto bail; 485 } 486 487 /* 488 * Be ultra-paranoid about making sure the type and fspath 489 * variables will fit in our mp buffers, including the 490 * terminating NUL. 491 */ 492 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) { 493 error = ENAMETOOLONG; 494 goto bail; 495 } 496 497 mtx_lock(&Giant); 498 error = vfs_domount(td, fstype, fspath, fsflags, optlist); 499 mtx_unlock(&Giant); 500 bail: 501 if (error) 502 vfs_freeopts(optlist); 503 return (error); 504 } 505 506 /* 507 * --------------------------------------------------------------------- 508 * Old mount API. 509 */ 510 #ifndef _SYS_SYSPROTO_H_ 511 struct mount_args { 512 char *type; 513 char *path; 514 int flags; 515 caddr_t data; 516 }; 517 #endif 518 /* ARGSUSED */ 519 int 520 mount(td, uap) 521 struct thread *td; 522 struct mount_args /* { 523 char *type; 524 char *path; 525 int flags; 526 caddr_t data; 527 } */ *uap; 528 { 529 char *fstype; 530 struct vfsconf *vfsp = NULL; 531 struct mntarg *ma = NULL; 532 int error; 533 534 /* Kick out MNT_ROOTFS early as it is legal internally */ 535 uap->flags &= ~MNT_ROOTFS; 536 537 if (uap->data == NULL) 538 return (EINVAL); 539 540 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK); 541 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL); 542 if (!error) { 543 mtx_lock(&Giant); /* XXX ? */ 544 vfsp = vfs_byname_kld(fstype, td, &error); 545 mtx_unlock(&Giant); 546 } 547 free(fstype, M_TEMP); 548 if (error) 549 return (error); 550 if (vfsp == NULL) 551 return (ENOENT); 552 if (vfsp->vfc_vfsops->vfs_cmount == NULL) 553 return (EOPNOTSUPP); 554 555 ma = mount_argsu(ma, "fstype", uap->type, MNAMELEN); 556 ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN); 557 ma = mount_argb(ma, uap->flags & MNT_RDONLY, "noro"); 558 ma = mount_argb(ma, !(uap->flags & MNT_NOSUID), "nosuid"); 559 ma = mount_argb(ma, !(uap->flags & MNT_NOEXEC), "noexec"); 560 561 error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, uap->flags, td); 562 return (error); 563 } 564 565 566 /* 567 * vfs_domount(): actually attempt a filesystem mount. 568 */ 569 static int 570 vfs_domount( 571 struct thread *td, /* Flags common to all filesystems. */ 572 const char *fstype, /* Filesystem type. */ 573 char *fspath, /* Mount path. */ 574 int fsflags, /* Flags common to all filesystems. */ 575 void *fsdata /* Options local to the filesystem. */ 576 ) 577 { 578 struct vnode *vp; 579 struct mount *mp; 580 struct vfsconf *vfsp; 581 int error, flag = 0, kern_flag = 0; 582 struct vattr va; 583 struct nameidata nd; 584 585 mtx_assert(&Giant, MA_OWNED); 586 587 /* 588 * Be ultra-paranoid about making sure the type and fspath 589 * variables will fit in our mp buffers, including the 590 * terminating NUL. 591 */ 592 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 593 return (ENAMETOOLONG); 594 595 if (jailed(td->td_ucred)) 596 return (EPERM); 597 if (usermount == 0) { 598 if ((error = suser(td)) != 0) 599 return (error); 600 } 601 602 /* 603 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users. 604 */ 605 if (fsflags & (MNT_EXPORTED | MNT_SUIDDIR)) { 606 if ((error = suser(td)) != 0) 607 return (error); 608 } 609 /* 610 * Silently enforce MNT_NOSUID and MNT_USER for 611 * unprivileged users. 612 */ 613 if (suser(td) != 0) 614 fsflags |= MNT_NOSUID | MNT_USER; 615 /* 616 * Get vnode to be covered 617 */ 618 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 619 if ((error = namei(&nd)) != 0) 620 return (error); 621 NDFREE(&nd, NDF_ONLY_PNBUF); 622 vp = nd.ni_vp; 623 if (fsflags & MNT_UPDATE) { 624 if ((vp->v_vflag & VV_ROOT) == 0) { 625 vput(vp); 626 return (EINVAL); 627 } 628 mp = vp->v_mount; 629 flag = mp->mnt_flag; 630 kern_flag = mp->mnt_kern_flag; 631 /* 632 * We only allow the filesystem to be reloaded if it 633 * is currently mounted read-only. 634 */ 635 if ((fsflags & MNT_RELOAD) && 636 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 637 vput(vp); 638 return (EOPNOTSUPP); /* Needs translation */ 639 } 640 /* 641 * Only privileged root, or (if MNT_USER is set) the user that 642 * did the original mount is permitted to update it. 643 */ 644 error = vfs_suser(mp, td); 645 if (error) { 646 vput(vp); 647 return (error); 648 } 649 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 650 vput(vp); 651 return (EBUSY); 652 } 653 VI_LOCK(vp); 654 if ((vp->v_iflag & VI_MOUNT) != 0 || 655 vp->v_mountedhere != NULL) { 656 VI_UNLOCK(vp); 657 vfs_unbusy(mp, td); 658 vput(vp); 659 return (EBUSY); 660 } 661 vp->v_iflag |= VI_MOUNT; 662 VI_UNLOCK(vp); 663 mp->mnt_flag |= fsflags & 664 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT | MNT_ROOTFS); 665 VOP_UNLOCK(vp, 0, td); 666 mp->mnt_optnew = fsdata; 667 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 668 } else { 669 /* 670 * If the user is not root, ensure that they own the directory 671 * onto which we are attempting to mount. 672 */ 673 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 674 if (error) { 675 vput(vp); 676 return (error); 677 } 678 if (va.va_uid != td->td_ucred->cr_uid) { 679 if ((error = suser(td)) != 0) { 680 vput(vp); 681 return (error); 682 } 683 } 684 error = vinvalbuf(vp, V_SAVE, td, 0, 0); 685 if (error != 0) { 686 vput(vp); 687 return (error); 688 } 689 if (vp->v_type != VDIR) { 690 vput(vp); 691 return (ENOTDIR); 692 } 693 vfsp = vfs_byname_kld(fstype, td, &error); 694 if (vfsp == NULL) { 695 vput(vp); 696 return (error); 697 } 698 VI_LOCK(vp); 699 if ((vp->v_iflag & VI_MOUNT) != 0 || 700 vp->v_mountedhere != NULL) { 701 VI_UNLOCK(vp); 702 vput(vp); 703 return (EBUSY); 704 } 705 vp->v_iflag |= VI_MOUNT; 706 VI_UNLOCK(vp); 707 708 /* 709 * Allocate and initialize the filesystem. 710 */ 711 error = vfs_mount_alloc(vp, vfsp, fspath, td, &mp); 712 if (error) { 713 vput(vp); 714 return (error); 715 } 716 VOP_UNLOCK(vp, 0, td); 717 718 /* XXXMAC: pass to vfs_mount_alloc? */ 719 mp->mnt_optnew = fsdata; 720 } 721 722 /* 723 * Set the mount level flags. 724 */ 725 if (fsflags & MNT_RDONLY) 726 mp->mnt_flag |= MNT_RDONLY; 727 mp->mnt_flag &=~ MNT_UPDATEMASK; 728 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE | MNT_ROOTFS); 729 /* 730 * Mount the filesystem. 731 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 732 * get. No freeing of cn_pnbuf. 733 */ 734 error = VFS_MOUNT(mp, td); 735 if (!error) { 736 if (mp->mnt_opt != NULL) 737 vfs_freeopts(mp->mnt_opt); 738 mp->mnt_opt = mp->mnt_optnew; 739 VFS_STATFS(mp, &mp->mnt_stat, td); 740 } 741 /* 742 * Prevent external consumers of mount options from reading 743 * mnt_optnew. 744 */ 745 mp->mnt_optnew = NULL; 746 if (mp->mnt_flag & MNT_UPDATE) { 747 mp->mnt_flag &= 748 ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 749 if (error) { 750 mp->mnt_flag = flag; 751 mp->mnt_kern_flag = kern_flag; 752 } 753 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 754 if (mp->mnt_syncer == NULL) 755 error = vfs_allocate_syncvnode(mp); 756 } else { 757 if (mp->mnt_syncer != NULL) 758 vrele(mp->mnt_syncer); 759 mp->mnt_syncer = NULL; 760 } 761 vfs_unbusy(mp, td); 762 VI_LOCK(vp); 763 vp->v_iflag &= ~VI_MOUNT; 764 VI_UNLOCK(vp); 765 vrele(vp); 766 return (error); 767 } 768 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 769 /* 770 * Put the new filesystem on the mount list after root. 771 */ 772 cache_purge(vp); 773 if (!error) { 774 struct vnode *newdp; 775 776 VI_LOCK(vp); 777 vp->v_iflag &= ~VI_MOUNT; 778 VI_UNLOCK(vp); 779 vp->v_mountedhere = mp; 780 mtx_lock(&mountlist_mtx); 781 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 782 mtx_unlock(&mountlist_mtx); 783 vfs_event_signal(NULL, VQ_MOUNT, 0); 784 if (VFS_ROOT(mp, &newdp, td)) 785 panic("mount: lost mount"); 786 mountcheckdirs(vp, newdp); 787 vput(newdp); 788 VOP_UNLOCK(vp, 0, td); 789 if ((mp->mnt_flag & MNT_RDONLY) == 0) 790 error = vfs_allocate_syncvnode(mp); 791 vfs_unbusy(mp, td); 792 if (error || (error = VFS_START(mp, 0, td)) != 0) 793 vrele(vp); 794 } else { 795 VI_LOCK(vp); 796 vp->v_iflag &= ~VI_MOUNT; 797 VI_UNLOCK(vp); 798 vfs_mount_destroy(mp, td); 799 vput(vp); 800 } 801 return (error); 802 } 803 804 /* 805 * --------------------------------------------------------------------- 806 * Unmount a filesystem. 807 * 808 * Note: unmount takes a path to the vnode mounted on as argument, 809 * not special file (as before). 810 */ 811 #ifndef _SYS_SYSPROTO_H_ 812 struct unmount_args { 813 char *path; 814 int flags; 815 }; 816 #endif 817 /* ARGSUSED */ 818 int 819 unmount(td, uap) 820 struct thread *td; 821 register struct unmount_args /* { 822 char *path; 823 int flags; 824 } */ *uap; 825 { 826 struct mount *mp; 827 char *pathbuf; 828 int error, id0, id1; 829 830 if (jailed(td->td_ucred)) 831 return (EPERM); 832 if (usermount == 0) { 833 if ((error = suser(td)) != 0) 834 return (error); 835 } 836 837 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 838 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL); 839 if (error) { 840 free(pathbuf, M_TEMP); 841 return (error); 842 } 843 if (uap->flags & MNT_BYFSID) { 844 /* Decode the filesystem ID. */ 845 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) { 846 free(pathbuf, M_TEMP); 847 return (EINVAL); 848 } 849 850 mtx_lock(&mountlist_mtx); 851 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 852 if (mp->mnt_stat.f_fsid.val[0] == id0 && 853 mp->mnt_stat.f_fsid.val[1] == id1) 854 break; 855 } 856 mtx_unlock(&mountlist_mtx); 857 } else { 858 mtx_lock(&mountlist_mtx); 859 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 860 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) 861 break; 862 } 863 mtx_unlock(&mountlist_mtx); 864 } 865 free(pathbuf, M_TEMP); 866 if (mp == NULL) { 867 /* 868 * Previously we returned ENOENT for a nonexistent path and 869 * EINVAL for a non-mountpoint. We cannot tell these apart 870 * now, so in the !MNT_BYFSID case return the more likely 871 * EINVAL for compatibility. 872 */ 873 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL); 874 } 875 876 /* 877 * Only privileged root, or (if MNT_USER is set) the user that did the 878 * original mount is permitted to unmount this filesystem. 879 */ 880 error = vfs_suser(mp, td); 881 if (error) 882 return (error); 883 884 /* 885 * Don't allow unmounting the root filesystem. 886 */ 887 if (mp->mnt_flag & MNT_ROOTFS) 888 return (EINVAL); 889 mtx_lock(&Giant); 890 error = dounmount(mp, uap->flags, td); 891 mtx_unlock(&Giant); 892 return (error); 893 } 894 895 /* 896 * Do the actual filesystem unmount. 897 */ 898 int 899 dounmount(mp, flags, td) 900 struct mount *mp; 901 int flags; 902 struct thread *td; 903 { 904 struct vnode *coveredvp, *fsrootvp; 905 int error; 906 int async_flag; 907 908 mtx_assert(&Giant, MA_OWNED); 909 910 mtx_lock(&mountlist_mtx); 911 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 912 mtx_unlock(&mountlist_mtx); 913 return (EBUSY); 914 } 915 mp->mnt_kern_flag |= MNTK_UNMOUNT; 916 /* Allow filesystems to detect that a forced unmount is in progress. */ 917 if (flags & MNT_FORCE) 918 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 919 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 920 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_mtx, td); 921 if (error) { 922 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 923 if (mp->mnt_kern_flag & MNTK_MWAIT) 924 wakeup(mp); 925 return (error); 926 } 927 vn_start_write(NULL, &mp, V_WAIT); 928 929 if (mp->mnt_flag & MNT_EXPUBLIC) 930 vfs_setpublicfs(NULL, NULL, NULL); 931 932 vfs_msync(mp, MNT_WAIT); 933 async_flag = mp->mnt_flag & MNT_ASYNC; 934 mp->mnt_flag &= ~MNT_ASYNC; 935 cache_purgevfs(mp); /* remove cache entries for this file sys */ 936 if (mp->mnt_syncer != NULL) 937 vrele(mp->mnt_syncer); 938 /* 939 * For forced unmounts, move process cdir/rdir refs on the fs root 940 * vnode to the covered vnode. For non-forced unmounts we want 941 * such references to cause an EBUSY error. 942 */ 943 if ((flags & MNT_FORCE) && VFS_ROOT(mp, &fsrootvp, td) == 0) { 944 if (mp->mnt_vnodecovered != NULL) 945 mountcheckdirs(fsrootvp, mp->mnt_vnodecovered); 946 if (fsrootvp == rootvnode) { 947 vrele(rootvnode); 948 rootvnode = NULL; 949 } 950 vput(fsrootvp); 951 } 952 if (((mp->mnt_flag & MNT_RDONLY) || 953 (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) || 954 (flags & MNT_FORCE)) { 955 error = VFS_UNMOUNT(mp, flags, td); 956 } 957 vn_finished_write(mp); 958 if (error) { 959 /* Undo cdir/rdir and rootvnode changes made above. */ 960 if ((flags & MNT_FORCE) && VFS_ROOT(mp, &fsrootvp, td) == 0) { 961 if (mp->mnt_vnodecovered != NULL) 962 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp); 963 if (rootvnode == NULL) { 964 rootvnode = fsrootvp; 965 vref(rootvnode); 966 } 967 vput(fsrootvp); 968 } 969 if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL) 970 (void) vfs_allocate_syncvnode(mp); 971 mtx_lock(&mountlist_mtx); 972 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 973 mp->mnt_flag |= async_flag; 974 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, 975 &mountlist_mtx, td); 976 if (mp->mnt_kern_flag & MNTK_MWAIT) 977 wakeup(mp); 978 return (error); 979 } 980 mtx_lock(&mountlist_mtx); 981 TAILQ_REMOVE(&mountlist, mp, mnt_list); 982 if ((coveredvp = mp->mnt_vnodecovered) != NULL) 983 coveredvp->v_mountedhere = NULL; 984 mtx_unlock(&mountlist_mtx); 985 vfs_event_signal(NULL, VQ_UNMOUNT, 0); 986 vfs_mount_destroy(mp, td); 987 if (coveredvp != NULL) 988 vrele(coveredvp); 989 return (0); 990 } 991 992 /* 993 * --------------------------------------------------------------------- 994 * Mounting of root filesystem 995 * 996 */ 997 998 static void 999 set_rootvnode(struct thread *td) 1000 { 1001 struct proc *p; 1002 1003 if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode, td)) 1004 panic("Cannot find root vnode"); 1005 1006 p = td->td_proc; 1007 FILEDESC_LOCK(p->p_fd); 1008 1009 if (p->p_fd->fd_cdir != NULL) 1010 vrele(p->p_fd->fd_cdir); 1011 p->p_fd->fd_cdir = rootvnode; 1012 VREF(rootvnode); 1013 1014 if (p->p_fd->fd_rdir != NULL) 1015 vrele(p->p_fd->fd_rdir); 1016 p->p_fd->fd_rdir = rootvnode; 1017 VREF(rootvnode); 1018 1019 FILEDESC_UNLOCK(p->p_fd); 1020 1021 VOP_UNLOCK(rootvnode, 0, td); 1022 } 1023 1024 /* 1025 * Mount /devfs as our root filesystem, but do not put it on the mountlist 1026 * yet. Create a /dev -> / symlink so that absolute pathnames will lookup. 1027 */ 1028 1029 static struct mount * 1030 devfs_first(void) 1031 { 1032 struct thread *td = curthread; 1033 struct vfsconf *vfsp; 1034 struct mount *mp = NULL; 1035 int error; 1036 1037 vfsp = vfs_byname("devfs"); 1038 KASSERT(vfsp != NULL, ("Could not find devfs by name")); 1039 if (vfsp == NULL) 1040 return(NULL); 1041 1042 error = vfs_mount_alloc(NULLVP, vfsp, "/dev", td, &mp); 1043 KASSERT(error == 0, ("vfs_mount_alloc failed %d", error)); 1044 if (error) 1045 return (NULL); 1046 1047 error = VFS_MOUNT(mp, curthread); 1048 KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error)); 1049 if (error) 1050 return (NULL); 1051 1052 VFS_START(mp, 0, td); 1053 1054 mtx_lock(&mountlist_mtx); 1055 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 1056 mtx_unlock(&mountlist_mtx); 1057 1058 set_rootvnode(td); 1059 1060 error = kern_symlink(td, "/", "dev", UIO_SYSSPACE); 1061 if (error) 1062 printf("kern_symlink /dev -> / returns %d\n", error); 1063 1064 return (mp); 1065 } 1066 1067 /* 1068 * Surgically move our devfs to be mounted on /dev. 1069 */ 1070 1071 static void 1072 devfs_fixup(struct thread *td) 1073 { 1074 struct nameidata nd; 1075 int error; 1076 struct vnode *vp, *dvp; 1077 struct mount *mp; 1078 1079 /* Remove our devfs mount from the mountlist and purge the cache */ 1080 mtx_lock(&mountlist_mtx); 1081 mp = TAILQ_FIRST(&mountlist); 1082 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1083 mtx_unlock(&mountlist_mtx); 1084 cache_purgevfs(mp); 1085 1086 VFS_ROOT(mp, &dvp, td); 1087 VI_LOCK(dvp); 1088 dvp->v_iflag &= ~VI_MOUNT; 1089 dvp->v_mountedhere = NULL; 1090 VI_UNLOCK(dvp); 1091 1092 /* Set up the real rootvnode, and purge the cache */ 1093 TAILQ_FIRST(&mountlist)->mnt_vnodecovered = NULL; 1094 set_rootvnode(td); 1095 cache_purgevfs(rootvnode->v_mount); 1096 1097 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td); 1098 error = namei(&nd); 1099 if (error) { 1100 printf("Lookup /dev -> %d\n", error); 1101 return; 1102 } 1103 NDFREE(&nd, NDF_ONLY_PNBUF); 1104 vp = nd.ni_vp; 1105 if (vp->v_type != VDIR) { 1106 vput(vp); 1107 } 1108 error = vinvalbuf(vp, V_SAVE, td, 0, 0); 1109 if (error) { 1110 vput(vp); 1111 } 1112 cache_purge(vp); 1113 mp->mnt_vnodecovered = vp; 1114 vp->v_mountedhere = mp; 1115 mtx_lock(&mountlist_mtx); 1116 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 1117 mtx_unlock(&mountlist_mtx); 1118 VOP_UNLOCK(vp, 0, td); 1119 vfs_unbusy(mp, td); 1120 vput(dvp); 1121 1122 /* Unlink the no longer needed /dev/dev -> / symlink */ 1123 kern_unlink(td, "/dev/dev", UIO_SYSSPACE); 1124 } 1125 1126 /* 1127 * Find and mount the root filesystem 1128 */ 1129 void 1130 vfs_mountroot(void) 1131 { 1132 char *cp; 1133 int error, i, asked = 0; 1134 struct mount *mp; 1135 1136 /* 1137 * Wait for GEOM to settle down 1138 */ 1139 DROP_GIANT(); 1140 g_waitidle(); 1141 PICKUP_GIANT(); 1142 1143 mp = devfs_first(); 1144 1145 /* 1146 * We are booted with instructions to prompt for the root filesystem. 1147 */ 1148 if (boothowto & RB_ASKNAME) { 1149 if (!vfs_mountroot_ask()) 1150 return; 1151 asked = 1; 1152 } 1153 1154 /* 1155 * The root filesystem information is compiled in, and we are 1156 * booted with instructions to use it. 1157 */ 1158 if (ctrootdevname != NULL && (boothowto & RB_DFLTROOT)) { 1159 if (!vfs_mountroot_try(ctrootdevname)) 1160 return; 1161 ctrootdevname = NULL; 1162 } 1163 1164 /* 1165 * We've been given the generic "use CDROM as root" flag. This is 1166 * necessary because one media may be used in many different 1167 * devices, so we need to search for them. 1168 */ 1169 if (boothowto & RB_CDROM) { 1170 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 1171 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 1172 return; 1173 } 1174 } 1175 1176 /* 1177 * Try to use the value read by the loader from /etc/fstab, or 1178 * supplied via some other means. This is the preferred 1179 * mechanism. 1180 */ 1181 cp = getenv("vfs.root.mountfrom"); 1182 if (cp != NULL) { 1183 error = vfs_mountroot_try(cp); 1184 freeenv(cp); 1185 if (!error) 1186 return; 1187 } 1188 1189 /* 1190 * Try values that may have been computed by code during boot 1191 */ 1192 if (!vfs_mountroot_try(rootdevnames[0])) 1193 return; 1194 if (!vfs_mountroot_try(rootdevnames[1])) 1195 return; 1196 1197 /* 1198 * If we (still) have a compiled-in default, try it. 1199 */ 1200 if (ctrootdevname != NULL) 1201 if (!vfs_mountroot_try(ctrootdevname)) 1202 return; 1203 /* 1204 * Everything so far has failed, prompt on the console if we haven't 1205 * already tried that. 1206 */ 1207 if (!asked) 1208 if (!vfs_mountroot_ask()) 1209 return; 1210 1211 panic("Root mount failed, startup aborted."); 1212 } 1213 1214 /* 1215 * Mount (mountfrom) as the root filesystem. 1216 */ 1217 static int 1218 vfs_mountroot_try(const char *mountfrom) 1219 { 1220 struct mount *mp; 1221 char *vfsname, *path; 1222 int error; 1223 char patt[32]; 1224 int s; 1225 1226 vfsname = NULL; 1227 path = NULL; 1228 mp = NULL; 1229 error = EINVAL; 1230 1231 if (mountfrom == NULL) 1232 return (error); /* don't complain */ 1233 1234 s = splcam(); /* Overkill, but annoying without it */ 1235 printf("Trying to mount root from %s\n", mountfrom); 1236 splx(s); 1237 1238 /* parse vfs name and path */ 1239 vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK); 1240 path = malloc(MNAMELEN, M_MOUNT, M_WAITOK); 1241 vfsname[0] = path[0] = 0; 1242 sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN); 1243 if (sscanf(mountfrom, patt, vfsname, path) < 1) 1244 return (error); 1245 1246 if (path[0] == '\0') 1247 strcpy(path, ROOTNAME); 1248 1249 error = kernel_vmount( 1250 MNT_RDONLY | MNT_ROOTFS, 1251 "fstype", vfsname, 1252 "fspath", "/", 1253 "from", path, 1254 NULL); 1255 if (error == 0) { 1256 mp = TAILQ_FIRST(&mountlist); 1257 1258 /* sanity check system clock against root fs timestamp */ 1259 inittodr(mp->mnt_time); 1260 vfs_unbusy(mp, curthread); 1261 error = VFS_START(mp, 0, curthread); 1262 1263 devfs_fixup(curthread); 1264 } 1265 return (error); 1266 } 1267 1268 /* 1269 * --------------------------------------------------------------------- 1270 * Interactive root filesystem selection code. 1271 */ 1272 1273 static int 1274 vfs_mountroot_ask(void) 1275 { 1276 char name[128]; 1277 1278 for(;;) { 1279 printf("\nManual root filesystem specification:\n"); 1280 printf(" <fstype>:<device> Mount <device> using filesystem <fstype>\n"); 1281 #if defined(__i386__) || defined(__ia64__) 1282 printf(" eg. ufs:da0s1a\n"); 1283 #else 1284 printf(" eg. ufs:/dev/da0a\n"); 1285 #endif 1286 printf(" ? List valid disk boot devices\n"); 1287 printf(" <empty line> Abort manual input\n"); 1288 printf("\nmountroot> "); 1289 gets(name); 1290 if (name[0] == '\0') 1291 return (1); 1292 if (name[0] == '?') { 1293 printf("\nList of GEOM managed disk devices:\n "); 1294 g_dev_print(); 1295 continue; 1296 } 1297 if (!vfs_mountroot_try(name)) 1298 return (0); 1299 } 1300 } 1301 1302 /* 1303 * Local helper function for vfs_mountroot_ask. 1304 */ 1305 static void 1306 gets(char *cp) 1307 { 1308 char *lp; 1309 int c; 1310 1311 lp = cp; 1312 for (;;) { 1313 printf("%c", c = cngetc() & 0177); 1314 switch (c) { 1315 case -1: 1316 case '\n': 1317 case '\r': 1318 *lp++ = '\0'; 1319 return; 1320 case '\b': 1321 case '\177': 1322 if (lp > cp) { 1323 printf(" \b"); 1324 lp--; 1325 } 1326 continue; 1327 case '#': 1328 lp--; 1329 if (lp < cp) 1330 lp = cp; 1331 continue; 1332 case '@': 1333 case 'u' & 037: 1334 lp = cp; 1335 printf("%c", '\n'); 1336 continue; 1337 default: 1338 *lp++ = c; 1339 } 1340 } 1341 } 1342 1343 /* 1344 * --------------------------------------------------------------------- 1345 * Functions for querying mount options/arguments from filesystems. 1346 */ 1347 1348 /* 1349 * Check that no unknown options are given 1350 */ 1351 int 1352 vfs_filteropt(struct vfsoptlist *opts, const char **legal) 1353 { 1354 struct vfsopt *opt; 1355 const char **t, *p; 1356 1357 1358 TAILQ_FOREACH(opt, opts, link) { 1359 p = opt->name; 1360 if (p[0] == 'n' && p[1] == 'o') 1361 p += 2; 1362 for(t = global_opts; *t != NULL; t++) 1363 if (!strcmp(*t, p)) 1364 break; 1365 if (*t != NULL) 1366 continue; 1367 for(t = legal; *t != NULL; t++) 1368 if (!strcmp(*t, p)) 1369 break; 1370 if (*t != NULL) 1371 continue; 1372 printf("mount option <%s> is unknown\n", p); 1373 return (EINVAL); 1374 } 1375 return (0); 1376 } 1377 1378 /* 1379 * Get a mount option by its name. 1380 * 1381 * Return 0 if the option was found, ENOENT otherwise. 1382 * If len is non-NULL it will be filled with the length 1383 * of the option. If buf is non-NULL, it will be filled 1384 * with the address of the option. 1385 */ 1386 int 1387 vfs_getopt(opts, name, buf, len) 1388 struct vfsoptlist *opts; 1389 const char *name; 1390 void **buf; 1391 int *len; 1392 { 1393 struct vfsopt *opt; 1394 1395 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1396 1397 TAILQ_FOREACH(opt, opts, link) { 1398 if (strcmp(name, opt->name) == 0) { 1399 if (len != NULL) 1400 *len = opt->len; 1401 if (buf != NULL) 1402 *buf = opt->value; 1403 return (0); 1404 } 1405 } 1406 return (ENOENT); 1407 } 1408 1409 char * 1410 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error) 1411 { 1412 struct vfsopt *opt; 1413 1414 *error = 0; 1415 TAILQ_FOREACH(opt, opts, link) { 1416 if (strcmp(name, opt->name) != 0) 1417 continue; 1418 if (((char *)opt->value)[opt->len - 1] != '\0') { 1419 *error = EINVAL; 1420 return (NULL); 1421 } 1422 return (opt->value); 1423 } 1424 return (NULL); 1425 } 1426 1427 int 1428 vfs_flagopt(struct vfsoptlist *opts, const char *name, u_int *w, u_int val) 1429 { 1430 struct vfsopt *opt; 1431 1432 TAILQ_FOREACH(opt, opts, link) { 1433 if (strcmp(name, opt->name) == 0) { 1434 if (w != NULL) 1435 *w |= val; 1436 return (1); 1437 } 1438 } 1439 if (w != NULL) 1440 *w &= ~val; 1441 return (0); 1442 } 1443 1444 int 1445 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...) 1446 { 1447 va_list ap; 1448 struct vfsopt *opt; 1449 int ret; 1450 1451 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1452 1453 TAILQ_FOREACH(opt, opts, link) { 1454 if (strcmp(name, opt->name) != 0) 1455 continue; 1456 if (((char *)opt->value)[opt->len - 1] != '\0') 1457 return (0); 1458 va_start(ap, fmt); 1459 ret = vsscanf(opt->value, fmt, ap); 1460 va_end(ap); 1461 return (ret); 1462 } 1463 return (0); 1464 } 1465 1466 /* 1467 * Find and copy a mount option. 1468 * 1469 * The size of the buffer has to be specified 1470 * in len, if it is not the same length as the 1471 * mount option, EINVAL is returned. 1472 * Returns ENOENT if the option is not found. 1473 */ 1474 int 1475 vfs_copyopt(opts, name, dest, len) 1476 struct vfsoptlist *opts; 1477 const char *name; 1478 void *dest; 1479 int len; 1480 { 1481 struct vfsopt *opt; 1482 1483 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL")); 1484 1485 TAILQ_FOREACH(opt, opts, link) { 1486 if (strcmp(name, opt->name) == 0) { 1487 if (len != opt->len) 1488 return (EINVAL); 1489 bcopy(opt->value, dest, opt->len); 1490 return (0); 1491 } 1492 } 1493 return (ENOENT); 1494 } 1495 1496 /* 1497 * This is a helper function for filesystems to traverse their 1498 * vnodes. See MNT_VNODE_FOREACH() in sys/mount.h 1499 */ 1500 1501 struct vnode * 1502 __mnt_vnode_next(struct vnode **nvp, struct mount *mp) 1503 { 1504 struct vnode *vp; 1505 1506 mtx_assert(&mp->mnt_mtx, MA_OWNED); 1507 1508 vp = *nvp; 1509 /* Check if we are done */ 1510 if (vp == NULL) 1511 return (NULL); 1512 /* If our next vnode is no longer ours, start over */ 1513 if (vp->v_mount != mp) 1514 vp = TAILQ_FIRST(&mp->mnt_nvnodelist); 1515 /* Save pointer to next vnode in list */ 1516 if (vp != NULL) 1517 *nvp = TAILQ_NEXT(vp, v_nmntvnodes); 1518 else 1519 *nvp = NULL; 1520 return (vp); 1521 } 1522 1523 int 1524 __vfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td) 1525 { 1526 int error; 1527 1528 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat, td); 1529 if (sbp != &mp->mnt_stat) 1530 *sbp = mp->mnt_stat; 1531 return (error); 1532 } 1533 1534 void 1535 vfs_mountedfrom(struct mount *mp, const char *from) 1536 { 1537 1538 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname); 1539 strlcpy(mp->mnt_stat.f_mntfromname, from, 1540 sizeof mp->mnt_stat.f_mntfromname); 1541 } 1542 1543 /* 1544 * --------------------------------------------------------------------- 1545 * This is the api for building mount args and mounting filesystems from 1546 * inside the kernel. 1547 * 1548 * The API works by accumulation of individual args. First error is 1549 * latched. 1550 * 1551 * XXX: should be documented in new manpage kernel_mount(9) 1552 */ 1553 1554 /* A memory allocation which must be freed when we are done */ 1555 struct mntaarg { 1556 SLIST_ENTRY(mntaarg) next; 1557 }; 1558 1559 /* The header for the mount arguments */ 1560 struct mntarg { 1561 struct iovec *v; 1562 int len; 1563 int error; 1564 SLIST_HEAD(, mntaarg) list; 1565 }; 1566 1567 /* 1568 * Add a boolean argument. 1569 * 1570 * flag is the boolean value. 1571 * name must start with "no". 1572 */ 1573 struct mntarg * 1574 mount_argb(struct mntarg *ma, int flag, const char *name) 1575 { 1576 1577 KASSERT(name[0] == 'n' && name[1] == 'o', 1578 ("mount_argb(...,%s): name must start with 'no'", name)); 1579 1580 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0)); 1581 } 1582 1583 /* 1584 * Add an argument printf style 1585 */ 1586 struct mntarg * 1587 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...) 1588 { 1589 va_list ap; 1590 struct mntaarg *maa; 1591 struct sbuf *sb; 1592 int len; 1593 1594 if (ma == NULL) { 1595 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1596 SLIST_INIT(&ma->list); 1597 } 1598 if (ma->error) 1599 return (ma); 1600 1601 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 1602 M_MOUNT, M_WAITOK); 1603 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 1604 ma->v[ma->len].iov_len = strlen(name) + 1; 1605 ma->len++; 1606 1607 sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND); 1608 va_start(ap, fmt); 1609 sbuf_vprintf(sb, fmt, ap); 1610 va_end(ap); 1611 sbuf_finish(sb); 1612 len = sbuf_len(sb) + 1; 1613 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 1614 SLIST_INSERT_HEAD(&ma->list, maa, next); 1615 bcopy(sbuf_data(sb), maa + 1, len); 1616 sbuf_delete(sb); 1617 1618 ma->v[ma->len].iov_base = maa + 1; 1619 ma->v[ma->len].iov_len = len; 1620 ma->len++; 1621 1622 return (ma); 1623 } 1624 1625 /* 1626 * Add an argument which is a userland string. 1627 */ 1628 struct mntarg * 1629 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len) 1630 { 1631 struct mntaarg *maa; 1632 char *tbuf; 1633 1634 if (val == NULL) 1635 return (ma); 1636 if (ma == NULL) { 1637 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1638 SLIST_INIT(&ma->list); 1639 } 1640 if (ma->error) 1641 return (ma); 1642 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 1643 SLIST_INSERT_HEAD(&ma->list, maa, next); 1644 tbuf = (void *)(maa + 1); 1645 ma->error = copyinstr(val, tbuf, len, NULL); 1646 return (mount_arg(ma, name, tbuf, -1)); 1647 } 1648 1649 /* 1650 * Plain argument. 1651 * 1652 * If length is -1, use printf. 1653 */ 1654 struct mntarg * 1655 mount_arg(struct mntarg *ma, const char *name, const void *val, int len) 1656 { 1657 1658 if (ma == NULL) { 1659 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1660 SLIST_INIT(&ma->list); 1661 } 1662 if (ma->error) 1663 return (ma); 1664 1665 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 1666 M_MOUNT, M_WAITOK); 1667 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 1668 ma->v[ma->len].iov_len = strlen(name) + 1; 1669 ma->len++; 1670 1671 ma->v[ma->len].iov_base = (void *)(uintptr_t)val; 1672 if (len < 0) 1673 ma->v[ma->len].iov_len = strlen(val) + 1; 1674 else 1675 ma->v[ma->len].iov_len = len; 1676 ma->len++; 1677 return (ma); 1678 } 1679 1680 /* 1681 * Free a mntarg structure 1682 */ 1683 void 1684 free_mntarg(struct mntarg *ma) 1685 { 1686 struct mntaarg *maa; 1687 1688 while (!SLIST_EMPTY(&ma->list)) { 1689 maa = SLIST_FIRST(&ma->list); 1690 SLIST_REMOVE_HEAD(&ma->list, next); 1691 free(maa, M_MOUNT); 1692 } 1693 free(ma->v, M_MOUNT); 1694 free(ma, M_MOUNT); 1695 } 1696 1697 /* 1698 * Mount a filesystem 1699 */ 1700 int 1701 kernel_mount(struct mntarg *ma, int flags) 1702 { 1703 struct uio auio; 1704 int error; 1705 1706 KASSERT(ma != NULL, ("kernel_mount NULL ma")); 1707 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v")); 1708 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len)); 1709 1710 auio.uio_iov = ma->v; 1711 auio.uio_iovcnt = ma->len; 1712 auio.uio_segflg = UIO_SYSSPACE; 1713 1714 error = ma->error; 1715 if (!error) 1716 error = vfs_donmount(curthread, flags, &auio); 1717 free_mntarg(ma); 1718 return (error); 1719 } 1720 1721 /* 1722 * A printflike function to mount a filesystem. 1723 */ 1724 int 1725 kernel_vmount(int flags, ...) 1726 { 1727 struct mntarg *ma = NULL; 1728 va_list ap; 1729 const char *cp; 1730 const void *vp; 1731 int error; 1732 1733 va_start(ap, flags); 1734 for (;;) { 1735 cp = va_arg(ap, const char *); 1736 if (cp == NULL) 1737 break; 1738 vp = va_arg(ap, const void *); 1739 ma = mount_arg(ma, cp, vp, -1); 1740 } 1741 va_end(ap); 1742 1743 error = kernel_mount(ma, flags); 1744 return (error); 1745 } 1746