vfs_mount.c revision 107855
1/* 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * Copyright (c) 1999 Michael Smith 39 * All rights reserved. 40 * Copyright (c) 1999 Poul-Henning Kamp 41 * All rights reserved. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 52 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 55 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 62 * SUCH DAMAGE. 63 * 64 * $FreeBSD: head/sys/kern/vfs_mount.c 107855 2002-12-14 08:18:06Z alfred $ 65 */ 66 67#include <sys/param.h> 68#include <sys/conf.h> 69#include <sys/cons.h> 70#include <sys/kernel.h> 71#include <sys/linker.h> 72#include <sys/mac.h> 73#include <sys/malloc.h> 74#include <sys/mount.h> 75#include <sys/mutex.h> 76#include <sys/namei.h> 77#include <sys/proc.h> 78#include <sys/reboot.h> 79#include <sys/sysproto.h> 80#include <sys/sx.h> 81#include <sys/sysctl.h> 82#include <sys/sysent.h> 83#include <sys/systm.h> 84#include <sys/vnode.h> 85 86#include <geom/geom.h> 87 88#include <machine/stdarg.h> 89 90#include "opt_rootdevname.h" 91#include "opt_ddb.h" 92#include "opt_mac.h" 93 94#ifdef DDB 95#include <ddb/ddb.h> 96#endif 97 98#define ROOTNAME "root_device" 99 100static void checkdirs(struct vnode *olddp, struct vnode *newdp); 101static int vfs_nmount(struct thread *td, int, struct uio *); 102static int vfs_mountroot_try(char *mountfrom); 103static int vfs_mountroot_ask(void); 104static void gets(char *cp); 105 106static int usermount = 0; /* if 1, non-root can mount fs. */ 107SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, ""); 108 109MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 110 111/* List of mounted filesystems. */ 112struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist); 113 114/* For any iteration/modification of mountlist */ 115struct mtx mountlist_mtx; 116 117/* For any iteration/modification of mnt_vnodelist */ 118struct mtx mntvnode_mtx; 119 120/* 121 * The vnode of the system's root (/ in the filesystem, without chroot 122 * active.) 123 */ 124struct vnode *rootvnode; 125 126/* 127 * The root filesystem is detailed in the kernel environment variable 128 * vfs.root.mountfrom, which is expected to be in the general format 129 * 130 * <vfsname>:[<path>] 131 * vfsname := the name of a VFS known to the kernel and capable 132 * of being mounted as root 133 * path := disk device name or other data used by the filesystem 134 * to locate its physical store 135 */ 136 137/* 138 * The root specifiers we will try if RB_CDROM is specified. 139 */ 140static char *cdrom_rootdevnames[] = { 141 "cd9660:cd0a", 142 "cd9660:acd0a", 143 "cd9660:wcd0a", 144 NULL 145}; 146 147/* legacy find-root code */ 148char *rootdevnames[2] = {NULL, NULL}; 149static int setrootbyname(char *name); 150dev_t rootdev = NODEV; 151 152/* Remove one mount option. */ 153static void 154vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt) 155{ 156 157 TAILQ_REMOVE(opts, opt, link); 158 free(opt->name, M_MOUNT); 159 if (opt->value != NULL) 160 free(opt->value, M_MOUNT); 161#ifdef INVARIANTS 162 else if (opt->len != 0) 163 panic("%s: mount option with NULL value but length != 0", 164 __func__); 165#endif 166 free(opt, M_MOUNT); 167} 168 169/* Release all resources related to the mount options. */ 170static void 171vfs_freeopts(struct vfsoptlist *opts) 172{ 173 struct vfsopt *opt; 174 175 while (!TAILQ_EMPTY(opts)) { 176 opt = TAILQ_FIRST(opts); 177 vfs_freeopt(opts, opt); 178 } 179 free(opts, M_MOUNT); 180} 181 182/* 183 * If a mount option is specified several times, 184 * (with or without the "no" prefix) only keep 185 * the last occurence of it. 186 */ 187static void 188vfs_sanitizeopts(struct vfsoptlist *opts) 189{ 190 struct vfsopt *opt, *opt2, *tmp; 191 int noopt; 192 193 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) { 194 if (strncmp(opt->name, "no", 2) == 0) 195 noopt = 1; 196 else 197 noopt = 0; 198 opt2 = TAILQ_PREV(opt, vfsoptlist, link); 199 while (opt2 != NULL) { 200 if (strcmp(opt2->name, opt->name) == 0 || 201 (noopt && strcmp(opt->name + 2, opt2->name) == 0) || 202 (!noopt && strncmp(opt2->name, "no", 2) == 0 && 203 strcmp(opt2->name + 2, opt->name) == 0)) { 204 tmp = TAILQ_PREV(opt2, vfsoptlist, link); 205 vfs_freeopt(opts, opt2); 206 opt2 = tmp; 207 } else { 208 opt2 = TAILQ_PREV(opt2, vfsoptlist, link); 209 } 210 } 211 } 212} 213 214/* 215 * Build a linked list of mount options from a struct uio. 216 */ 217static int 218vfs_buildopts(struct uio *auio, struct vfsoptlist **options) 219{ 220 struct vfsoptlist *opts; 221 struct vfsopt *opt; 222 unsigned int i, iovcnt; 223 int error, namelen, optlen; 224 225 iovcnt = auio->uio_iovcnt; 226 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 227 TAILQ_INIT(opts); 228 for (i = 0; i < iovcnt; i += 2) { 229 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 230 namelen = auio->uio_iov[i].iov_len; 231 optlen = auio->uio_iov[i + 1].iov_len; 232 opt->name = malloc(namelen, M_MOUNT, M_WAITOK); 233 opt->value = NULL; 234 if (auio->uio_segflg == UIO_SYSSPACE) { 235 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen); 236 } else { 237 error = copyin(auio->uio_iov[i].iov_base, opt->name, 238 namelen); 239 if (error) 240 goto bad; 241 } 242 opt->len = optlen; 243 if (optlen != 0) { 244 opt->value = malloc(optlen, M_MOUNT, M_WAITOK); 245 if (auio->uio_segflg == UIO_SYSSPACE) { 246 bcopy(auio->uio_iov[i + 1].iov_base, opt->value, 247 optlen); 248 } else { 249 error = copyin(auio->uio_iov[i + 1].iov_base, 250 opt->value, optlen); 251 if (error) 252 goto bad; 253 } 254 } 255 TAILQ_INSERT_TAIL(opts, opt, link); 256 } 257 vfs_sanitizeopts(opts); 258 *options = opts; 259 return (0); 260bad: 261 vfs_freeopts(opts); 262 return (error); 263} 264 265/* 266 * Merge the old mount options with the new ones passed 267 * in the MNT_UPDATE case. 268 */ 269static void 270vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts) 271{ 272 struct vfsopt *opt, *opt2, *new; 273 274 TAILQ_FOREACH(opt, opts, link) { 275 /* 276 * Check that this option hasn't been redefined 277 * nor cancelled with a "no" mount option. 278 */ 279 opt2 = TAILQ_FIRST(toopts); 280 while (opt2 != NULL) { 281 if (strcmp(opt2->name, opt->name) == 0) 282 goto next; 283 if (strncmp(opt2->name, "no", 2) == 0 && 284 strcmp(opt2->name + 2, opt->name) == 0) { 285 vfs_freeopt(toopts, opt2); 286 goto next; 287 } 288 opt2 = TAILQ_NEXT(opt2, link); 289 } 290 /* We want this option, duplicate it. */ 291 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 292 new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK); 293 strcpy(new->name, opt->name); 294 if (opt->len != 0) { 295 new->value = malloc(opt->len, M_MOUNT, M_WAITOK); 296 bcopy(opt->value, new->value, opt->len); 297 } else { 298 new->value = NULL; 299 } 300 new->len = opt->len; 301 TAILQ_INSERT_TAIL(toopts, new, link); 302next: 303 continue; 304 } 305} 306 307/* 308 * New mount API. 309 */ 310int 311nmount(td, uap) 312 struct thread *td; 313 struct nmount_args /* { 314 struct iovec *iovp; 315 unsigned int iovcnt; 316 int flags; 317 } */ *uap; 318{ 319 struct uio auio; 320 struct iovec *iov, *needfree; 321 struct iovec aiov[UIO_SMALLIOV]; 322 unsigned int i; 323 int error; 324 u_int iovlen, iovcnt; 325 326 iovcnt = uap->iovcnt; 327 iovlen = iovcnt * sizeof (struct iovec); 328 /* 329 * Check that we have an even number of iovec's 330 * and that we have at least two options. 331 */ 332 if ((iovcnt & 1) || (iovcnt < 4) || (iovcnt > UIO_MAXIOV)) 333 return (EINVAL); 334 335 if (iovcnt > UIO_SMALLIOV) { 336 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 337 needfree = iov; 338 } else { 339 iov = aiov; 340 needfree = NULL; 341 } 342 auio.uio_iov = iov; 343 auio.uio_iovcnt = iovcnt; 344 auio.uio_segflg = UIO_USERSPACE; 345 if ((error = copyin(uap->iovp, iov, iovlen))) 346 goto finish; 347 348 for (i = 0; i < iovcnt; i++) { 349 if (iov->iov_len > MMAXOPTIONLEN) { 350 error = EINVAL; 351 goto finish; 352 } 353 iov++; 354 } 355 error = vfs_nmount(td, uap->flags, &auio); 356finish: 357 if (needfree != NULL) 358 free(needfree, M_TEMP); 359 return (error); 360} 361 362int 363kernel_mount(iovp, iovcnt, flags) 364 struct iovec *iovp; 365 unsigned int iovcnt; 366 int flags; 367{ 368 struct uio auio; 369 int error; 370 371 /* 372 * Check that we have an even number of iovec's 373 * and that we have at least two options. 374 */ 375 if ((iovcnt & 1) || (iovcnt < 4)) 376 return (EINVAL); 377 378 auio.uio_iov = iovp; 379 auio.uio_iovcnt = iovcnt; 380 auio.uio_segflg = UIO_SYSSPACE; 381 382 error = vfs_nmount(curthread, flags, &auio); 383 return (error); 384} 385 386int 387kernel_vmount(int flags, ...) 388{ 389 struct iovec *iovp; 390 struct uio auio; 391 va_list ap; 392 unsigned int iovcnt, iovlen, len; 393 const char *cp; 394 char *buf, *pos; 395 size_t n; 396 int error, i; 397 398 len = 0; 399 va_start(ap, flags); 400 for (iovcnt = 0; (cp = va_arg(ap, const char *)) != NULL; iovcnt++) 401 len += strlen(cp) + 1; 402 va_end(ap); 403 404 if (iovcnt < 4 || iovcnt & 1) 405 return (EINVAL); 406 407 iovlen = iovcnt * sizeof (struct iovec); 408 MALLOC(iovp, struct iovec *, iovlen, M_MOUNT, M_WAITOK); 409 MALLOC(buf, char *, len, M_MOUNT, M_WAITOK); 410 pos = buf; 411 va_start(ap, flags); 412 for (i = 0; i < iovcnt; i++) { 413 cp = va_arg(ap, const char *); 414 copystr(cp, pos, len - (pos - buf), &n); 415 iovp[i].iov_base = pos; 416 iovp[i].iov_len = n; 417 pos += n; 418 } 419 va_end(ap); 420 421 auio.uio_iov = iovp; 422 auio.uio_iovcnt = iovcnt; 423 auio.uio_segflg = UIO_SYSSPACE; 424 425 error = vfs_nmount(curthread, flags, &auio); 426 FREE(iovp, M_MOUNT); 427 FREE(buf, M_MOUNT); 428 return (error); 429} 430 431/* 432 * vfs_nmount(): actually attempt a filesystem mount. 433 */ 434static int 435vfs_nmount(td, fsflags, fsoptions) 436 struct thread *td; 437 int fsflags; /* Flags common to all filesystems. */ 438 struct uio *fsoptions; /* Options local to the filesystem. */ 439{ 440 linker_file_t lf; 441 struct vnode *vp; 442 struct mount *mp; 443 struct vfsconf *vfsp; 444 struct vfsoptlist *optlist; 445 char *fstype, *fspath; 446 int error, flag = 0, kern_flag = 0; 447 int fstypelen, fspathlen; 448 struct vattr va; 449 struct nameidata nd; 450 451 error = vfs_buildopts(fsoptions, &optlist); 452 if (error) 453 return (error); 454 455 /* 456 * We need these two options before the others, 457 * and they are mandatory for any filesystem. 458 * Ensure they are NUL terminated as well. 459 */ 460 fstypelen = 0; 461 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen); 462 if (error || fstype[fstypelen - 1] != '\0') { 463 error = EINVAL; 464 goto bad; 465 } 466 fspathlen = 0; 467 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen); 468 if (error || fspath[fspathlen - 1] != '\0') { 469 error = EINVAL; 470 goto bad; 471 } 472 473 /* 474 * Be ultra-paranoid about making sure the type and fspath 475 * variables will fit in our mp buffers, including the 476 * terminating NUL. 477 */ 478 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) { 479 error = ENAMETOOLONG; 480 goto bad; 481 } 482 483 if (usermount == 0) { 484 error = suser(td); 485 if (error) 486 goto bad; 487 } 488 /* 489 * Do not allow NFS export by non-root users. 490 */ 491 if (fsflags & MNT_EXPORTED) { 492 error = suser(td); 493 if (error) 494 goto bad; 495 } 496 /* 497 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users. 498 */ 499 if (suser(td)) 500 fsflags |= MNT_NOSUID | MNT_NODEV; 501 /* 502 * Get vnode to be covered 503 */ 504 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 505 if ((error = namei(&nd)) != 0) 506 goto bad; 507 NDFREE(&nd, NDF_ONLY_PNBUF); 508 vp = nd.ni_vp; 509 if (fsflags & MNT_UPDATE) { 510 if ((vp->v_vflag & VV_ROOT) == 0) { 511 vput(vp); 512 error = EINVAL; 513 goto bad; 514 } 515 mp = vp->v_mount; 516 flag = mp->mnt_flag; 517 kern_flag = mp->mnt_kern_flag; 518 /* 519 * We only allow the filesystem to be reloaded if it 520 * is currently mounted read-only. 521 */ 522 if ((fsflags & MNT_RELOAD) && 523 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 524 vput(vp); 525 error = EOPNOTSUPP; /* Needs translation */ 526 goto bad; 527 } 528 /* 529 * Only root, or the user that did the original mount is 530 * permitted to update it. 531 */ 532 if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) { 533 error = suser(td); 534 if (error) { 535 vput(vp); 536 goto bad; 537 } 538 } 539 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 540 vput(vp); 541 error = EBUSY; 542 goto bad; 543 } 544 VI_LOCK(vp); 545 if ((vp->v_iflag & VI_MOUNT) != 0 || 546 vp->v_mountedhere != NULL) { 547 VI_UNLOCK(vp); 548 vfs_unbusy(mp, td); 549 vput(vp); 550 error = EBUSY; 551 goto bad; 552 } 553 vp->v_iflag |= VI_MOUNT; 554 VI_UNLOCK(vp); 555 mp->mnt_flag |= fsflags & 556 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 557 VOP_UNLOCK(vp, 0, td); 558 mp->mnt_optnew = optlist; 559 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 560 goto update; 561 } 562 /* 563 * If the user is not root, ensure that they own the directory 564 * onto which we are attempting to mount. 565 */ 566 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 567 if (error) { 568 vput(vp); 569 goto bad; 570 } 571 if (va.va_uid != td->td_ucred->cr_uid) { 572 error = suser(td); 573 if (error) { 574 vput(vp); 575 goto bad; 576 } 577 } 578 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 579 vput(vp); 580 goto bad; 581 } 582 if (vp->v_type != VDIR) { 583 vput(vp); 584 error = ENOTDIR; 585 goto bad; 586 } 587 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 588 if (!strcmp(vfsp->vfc_name, fstype)) 589 break; 590 if (vfsp == NULL) { 591 /* Only load modules for root (very important!). */ 592 error = suser(td); 593 if (error) { 594 vput(vp); 595 goto bad; 596 } 597 error = securelevel_gt(td->td_ucred, 0); 598 if (error) { 599 vput(vp); 600 goto bad; 601 } 602 error = linker_load_module(NULL, fstype, NULL, NULL, &lf); 603 if (error || lf == NULL) { 604 vput(vp); 605 if (lf == NULL) 606 error = ENODEV; 607 goto bad; 608 } 609 lf->userrefs++; 610 /* Look up again to see if the VFS was loaded. */ 611 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 612 if (!strcmp(vfsp->vfc_name, fstype)) 613 break; 614 if (vfsp == NULL) { 615 lf->userrefs--; 616 linker_file_unload(lf); 617 vput(vp); 618 error = ENODEV; 619 goto bad; 620 } 621 } 622 VI_LOCK(vp); 623 if ((vp->v_iflag & VI_MOUNT) != 0 || 624 vp->v_mountedhere != NULL) { 625 VI_UNLOCK(vp); 626 vput(vp); 627 error = EBUSY; 628 goto bad; 629 } 630 vp->v_iflag |= VI_MOUNT; 631 VI_UNLOCK(vp); 632 633 /* 634 * Allocate and initialize the filesystem. 635 */ 636 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 637 TAILQ_INIT(&mp->mnt_nvnodelist); 638 TAILQ_INIT(&mp->mnt_reservedvnlist); 639 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 640 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 641 mp->mnt_op = vfsp->vfc_vfsops; 642 mp->mnt_vfc = vfsp; 643 vfsp->vfc_refcount++; 644 mp->mnt_stat.f_type = vfsp->vfc_typenum; 645 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 646 strlcpy(mp->mnt_stat.f_fstypename, fstype, MFSNAMELEN); 647 mp->mnt_vnodecovered = vp; 648 mp->mnt_cred = crdup(td->td_ucred); 649 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 650 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 651 mp->mnt_iosize_max = DFLTPHYS; 652#ifdef MAC 653 mac_init_mount(mp); 654 mac_create_mount(td->td_ucred, mp); 655#endif 656 VOP_UNLOCK(vp, 0, td); 657 mp->mnt_optnew = optlist; /* XXXMAC: should this be above? */ 658 659update: 660 /* 661 * Check if the fs implements the new VFS_NMOUNT() 662 * function, since the new system call was used. 663 */ 664 if (mp->mnt_op->vfs_mount != NULL) { 665 printf("%s doesn't support the new mount syscall\n", 666 mp->mnt_vfc->vfc_name); 667 VI_LOCK(vp); 668 vp->v_iflag &= ~VI_MOUNT; 669 VI_UNLOCK(vp); 670 if (mp->mnt_flag & MNT_UPDATE) 671 vfs_unbusy(mp, td); 672 else { 673 mp->mnt_vfc->vfc_refcount--; 674 vfs_unbusy(mp, td); 675#ifdef MAC 676 mac_destroy_mount(mp); 677#endif 678 free(mp, M_MOUNT); 679 } 680 vrele(vp); 681 error = EOPNOTSUPP; 682 goto bad; 683 } 684 685 /* 686 * Set the mount level flags. 687 */ 688 if (fsflags & MNT_RDONLY) 689 mp->mnt_flag |= MNT_RDONLY; 690 else if (mp->mnt_flag & MNT_RDONLY) 691 mp->mnt_kern_flag |= MNTK_WANTRDWR; 692 mp->mnt_flag &=~ MNT_UPDATEMASK; 693 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 694 /* 695 * Mount the filesystem. 696 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 697 * get. No freeing of cn_pnbuf. 698 */ 699 error = VFS_NMOUNT(mp, &nd, td); 700 if (!error) { 701 if (mp->mnt_opt != NULL) 702 vfs_freeopts(mp->mnt_opt); 703 mp->mnt_opt = mp->mnt_optnew; 704 } 705 /* 706 * Prevent external consumers of mount 707 * options to read mnt_optnew. 708 */ 709 mp->mnt_optnew = NULL; 710 if (mp->mnt_flag & MNT_UPDATE) { 711 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 712 mp->mnt_flag &= ~MNT_RDONLY; 713 mp->mnt_flag &=~ 714 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 715 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 716 if (error) { 717 mp->mnt_flag = flag; 718 mp->mnt_kern_flag = kern_flag; 719 } 720 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 721 if (mp->mnt_syncer == NULL) 722 error = vfs_allocate_syncvnode(mp); 723 } else { 724 if (mp->mnt_syncer != NULL) 725 vrele(mp->mnt_syncer); 726 mp->mnt_syncer = NULL; 727 } 728 vfs_unbusy(mp, td); 729 VI_LOCK(vp); 730 vp->v_iflag &= ~VI_MOUNT; 731 VI_UNLOCK(vp); 732 vrele(vp); 733 return (error); 734 } 735 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 736 /* 737 * Put the new filesystem on the mount list after root. 738 */ 739 cache_purge(vp); 740 if (!error) { 741 struct vnode *newdp; 742 743 VI_LOCK(vp); 744 vp->v_iflag &= ~VI_MOUNT; 745 VI_UNLOCK(vp); 746 vp->v_mountedhere = mp; 747 mtx_lock(&mountlist_mtx); 748 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 749 mtx_unlock(&mountlist_mtx); 750 if (VFS_ROOT(mp, &newdp)) 751 panic("mount: lost mount"); 752 checkdirs(vp, newdp); 753 vput(newdp); 754 VOP_UNLOCK(vp, 0, td); 755 if ((mp->mnt_flag & MNT_RDONLY) == 0) 756 error = vfs_allocate_syncvnode(mp); 757 vfs_unbusy(mp, td); 758 if ((error = VFS_START(mp, 0, td)) != 0) { 759 vrele(vp); 760 goto bad; 761 } 762 } else { 763 VI_LOCK(vp); 764 vp->v_iflag &= ~VI_MOUNT; 765 VI_UNLOCK(vp); 766 mp->mnt_vfc->vfc_refcount--; 767 vfs_unbusy(mp, td); 768#ifdef MAC 769 mac_destroy_mount(mp); 770#endif 771 free(mp, M_MOUNT); 772 vput(vp); 773 goto bad; 774 } 775 return (0); 776bad: 777 vfs_freeopts(optlist); 778 return (error); 779} 780 781/* 782 * Old mount API. 783 */ 784#ifndef _SYS_SYSPROTO_H_ 785struct mount_args { 786 char *type; 787 char *path; 788 int flags; 789 caddr_t data; 790}; 791#endif 792/* ARGSUSED */ 793int 794mount(td, uap) 795 struct thread *td; 796 struct mount_args /* { 797 char *type; 798 char *path; 799 int flags; 800 caddr_t data; 801 } */ *uap; 802{ 803 char *fstype; 804 char *fspath; 805 int error; 806 807 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK); 808 fspath = malloc(MNAMELEN, M_TEMP, M_WAITOK); 809 810 /* 811 * vfs_mount() actually takes a kernel string for `type' and 812 * `path' now, so extract them. 813 */ 814 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL); 815 if (error == 0) 816 error = copyinstr(uap->path, fspath, MNAMELEN, NULL); 817 if (error == 0) 818 error = vfs_mount(td, fstype, fspath, uap->flags, uap->data); 819 free(fstype, M_TEMP); 820 free(fspath, M_TEMP); 821 return (error); 822} 823 824/* 825 * vfs_mount(): actually attempt a filesystem mount. 826 * 827 * This routine is designed to be a "generic" entry point for routines 828 * that wish to mount a filesystem. All parameters except `fsdata' are 829 * pointers into kernel space. `fsdata' is currently still a pointer 830 * into userspace. 831 */ 832int 833vfs_mount(td, fstype, fspath, fsflags, fsdata) 834 struct thread *td; 835 const char *fstype; 836 char *fspath; 837 int fsflags; 838 void *fsdata; 839{ 840 linker_file_t lf; 841 struct vnode *vp; 842 struct mount *mp; 843 struct vfsconf *vfsp; 844 int error, flag = 0, kern_flag = 0; 845 struct vattr va; 846 struct nameidata nd; 847 848 /* 849 * Be ultra-paranoid about making sure the type and fspath 850 * variables will fit in our mp buffers, including the 851 * terminating NUL. 852 */ 853 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 854 return (ENAMETOOLONG); 855 856 if (usermount == 0) { 857 error = suser(td); 858 if (error) 859 return (error); 860 } 861 /* 862 * Do not allow NFS export by non-root users. 863 */ 864 if (fsflags & MNT_EXPORTED) { 865 error = suser(td); 866 if (error) 867 return (error); 868 } 869 /* 870 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users. 871 */ 872 if (suser(td)) 873 fsflags |= MNT_NOSUID | MNT_NODEV; 874 /* 875 * Get vnode to be covered 876 */ 877 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 878 if ((error = namei(&nd)) != 0) 879 return (error); 880 NDFREE(&nd, NDF_ONLY_PNBUF); 881 vp = nd.ni_vp; 882 if (fsflags & MNT_UPDATE) { 883 if ((vp->v_vflag & VV_ROOT) == 0) { 884 vput(vp); 885 return (EINVAL); 886 } 887 mp = vp->v_mount; 888 flag = mp->mnt_flag; 889 kern_flag = mp->mnt_kern_flag; 890 /* 891 * We only allow the filesystem to be reloaded if it 892 * is currently mounted read-only. 893 */ 894 if ((fsflags & MNT_RELOAD) && 895 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 896 vput(vp); 897 return (EOPNOTSUPP); /* Needs translation */ 898 } 899 /* 900 * Only root, or the user that did the original mount is 901 * permitted to update it. 902 */ 903 if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) { 904 error = suser(td); 905 if (error) { 906 vput(vp); 907 return (error); 908 } 909 } 910 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 911 vput(vp); 912 return (EBUSY); 913 } 914 VI_LOCK(vp); 915 if ((vp->v_iflag & VI_MOUNT) != 0 || 916 vp->v_mountedhere != NULL) { 917 VI_UNLOCK(vp); 918 vfs_unbusy(mp, td); 919 vput(vp); 920 return (EBUSY); 921 } 922 vp->v_iflag |= VI_MOUNT; 923 VI_UNLOCK(vp); 924 mp->mnt_flag |= fsflags & 925 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 926 VOP_UNLOCK(vp, 0, td); 927 goto update; 928 } 929 /* 930 * If the user is not root, ensure that they own the directory 931 * onto which we are attempting to mount. 932 */ 933 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 934 if (error) { 935 vput(vp); 936 return (error); 937 } 938 if (va.va_uid != td->td_ucred->cr_uid) { 939 error = suser(td); 940 if (error) { 941 vput(vp); 942 return (error); 943 } 944 } 945 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 946 vput(vp); 947 return (error); 948 } 949 if (vp->v_type != VDIR) { 950 vput(vp); 951 return (ENOTDIR); 952 } 953 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 954 if (!strcmp(vfsp->vfc_name, fstype)) 955 break; 956 if (vfsp == NULL) { 957 /* Only load modules for root (very important!). */ 958 error = suser(td); 959 if (error) { 960 vput(vp); 961 return (error); 962 } 963 error = securelevel_gt(td->td_ucred, 0); 964 if (error) { 965 vput(vp); 966 return (error); 967 } 968 error = linker_load_module(NULL, fstype, NULL, NULL, &lf); 969 if (error || lf == NULL) { 970 vput(vp); 971 if (lf == NULL) 972 error = ENODEV; 973 return (error); 974 } 975 lf->userrefs++; 976 /* Look up again to see if the VFS was loaded. */ 977 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 978 if (!strcmp(vfsp->vfc_name, fstype)) 979 break; 980 if (vfsp == NULL) { 981 lf->userrefs--; 982 linker_file_unload(lf); 983 vput(vp); 984 return (ENODEV); 985 } 986 } 987 VI_LOCK(vp); 988 if ((vp->v_iflag & VI_MOUNT) != 0 || 989 vp->v_mountedhere != NULL) { 990 VI_UNLOCK(vp); 991 vput(vp); 992 return (EBUSY); 993 } 994 vp->v_iflag |= VI_MOUNT; 995 VI_UNLOCK(vp); 996 997 /* 998 * Allocate and initialize the filesystem. 999 */ 1000 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 1001 TAILQ_INIT(&mp->mnt_nvnodelist); 1002 TAILQ_INIT(&mp->mnt_reservedvnlist); 1003 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 1004 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 1005 mp->mnt_op = vfsp->vfc_vfsops; 1006 mp->mnt_vfc = vfsp; 1007 vfsp->vfc_refcount++; 1008 mp->mnt_stat.f_type = vfsp->vfc_typenum; 1009 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 1010 strlcpy(mp->mnt_stat.f_fstypename, fstype, MFSNAMELEN); 1011 mp->mnt_vnodecovered = vp; 1012 mp->mnt_cred = crdup(td->td_ucred); 1013 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 1014 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 1015 mp->mnt_iosize_max = DFLTPHYS; 1016#ifdef MAC 1017 mac_init_mount(mp); 1018 mac_create_mount(td->td_ucred, mp); 1019#endif 1020 VOP_UNLOCK(vp, 0, td); 1021update: 1022 /* 1023 * Check if the fs implements the old VFS_MOUNT() 1024 * function, since the old system call was used. 1025 */ 1026 if (mp->mnt_op->vfs_mount == NULL) { 1027 printf("%s doesn't support the old mount syscall\n", 1028 mp->mnt_vfc->vfc_name); 1029 VI_LOCK(vp); 1030 vp->v_iflag &= ~VI_MOUNT; 1031 VI_UNLOCK(vp); 1032 if (mp->mnt_flag & MNT_UPDATE) 1033 vfs_unbusy(mp, td); 1034 else { 1035 mp->mnt_vfc->vfc_refcount--; 1036 vfs_unbusy(mp, td); 1037#ifdef MAC 1038 mac_destroy_mount(mp); 1039#endif 1040 free(mp, M_MOUNT); 1041 } 1042 vrele(vp); 1043 return (EOPNOTSUPP); 1044 } 1045 1046 /* 1047 * Set the mount level flags. 1048 */ 1049 if (fsflags & MNT_RDONLY) 1050 mp->mnt_flag |= MNT_RDONLY; 1051 else if (mp->mnt_flag & MNT_RDONLY) 1052 mp->mnt_kern_flag |= MNTK_WANTRDWR; 1053 mp->mnt_flag &=~ MNT_UPDATEMASK; 1054 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 1055 /* 1056 * Mount the filesystem. 1057 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 1058 * get. No freeing of cn_pnbuf. 1059 */ 1060 error = VFS_MOUNT(mp, fspath, fsdata, &nd, td); 1061 if (mp->mnt_flag & MNT_UPDATE) { 1062 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 1063 mp->mnt_flag &= ~MNT_RDONLY; 1064 mp->mnt_flag &=~ 1065 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 1066 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 1067 if (error) { 1068 mp->mnt_flag = flag; 1069 mp->mnt_kern_flag = kern_flag; 1070 } 1071 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 1072 if (mp->mnt_syncer == NULL) 1073 error = vfs_allocate_syncvnode(mp); 1074 } else { 1075 if (mp->mnt_syncer != NULL) 1076 vrele(mp->mnt_syncer); 1077 mp->mnt_syncer = NULL; 1078 } 1079 vfs_unbusy(mp, td); 1080 VI_LOCK(vp); 1081 vp->v_iflag &= ~VI_MOUNT; 1082 VI_UNLOCK(vp); 1083 vrele(vp); 1084 return (error); 1085 } 1086 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 1087 /* 1088 * Put the new filesystem on the mount list after root. 1089 */ 1090 cache_purge(vp); 1091 if (!error) { 1092 struct vnode *newdp; 1093 1094 VI_LOCK(vp); 1095 vp->v_iflag &= ~VI_MOUNT; 1096 VI_UNLOCK(vp); 1097 vp->v_mountedhere = mp; 1098 mtx_lock(&mountlist_mtx); 1099 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 1100 mtx_unlock(&mountlist_mtx); 1101 if (VFS_ROOT(mp, &newdp)) 1102 panic("mount: lost mount"); 1103 checkdirs(vp, newdp); 1104 vput(newdp); 1105 VOP_UNLOCK(vp, 0, td); 1106 if ((mp->mnt_flag & MNT_RDONLY) == 0) 1107 error = vfs_allocate_syncvnode(mp); 1108 vfs_unbusy(mp, td); 1109 if ((error = VFS_START(mp, 0, td)) != 0) 1110 vrele(vp); 1111 } else { 1112 VI_LOCK(vp); 1113 vp->v_iflag &= ~VI_MOUNT; 1114 VI_UNLOCK(vp); 1115 mp->mnt_vfc->vfc_refcount--; 1116 vfs_unbusy(mp, td); 1117#ifdef MAC 1118 mac_destroy_mount(mp); 1119#endif 1120 free(mp, M_MOUNT); 1121 vput(vp); 1122 } 1123 return (error); 1124} 1125 1126/* 1127 * Scan all active processes to see if any of them have a current 1128 * or root directory of `olddp'. If so, replace them with the new 1129 * mount point. 1130 */ 1131static void 1132checkdirs(olddp, newdp) 1133 struct vnode *olddp, *newdp; 1134{ 1135 struct filedesc *fdp; 1136 struct proc *p; 1137 int nrele; 1138 1139 if (vrefcnt(olddp) == 1) 1140 return; 1141 sx_slock(&allproc_lock); 1142 LIST_FOREACH(p, &allproc, p_list) { 1143 PROC_LOCK(p); 1144 fdp = p->p_fd; 1145 if (fdp == NULL) { 1146 PROC_UNLOCK(p); 1147 continue; 1148 } 1149 nrele = 0; 1150 FILEDESC_LOCK(fdp); 1151 if (fdp->fd_cdir == olddp) { 1152 VREF(newdp); 1153 fdp->fd_cdir = newdp; 1154 nrele++; 1155 } 1156 if (fdp->fd_rdir == olddp) { 1157 VREF(newdp); 1158 fdp->fd_rdir = newdp; 1159 nrele++; 1160 } 1161 FILEDESC_UNLOCK(fdp); 1162 PROC_UNLOCK(p); 1163 while (nrele--) 1164 vrele(olddp); 1165 } 1166 sx_sunlock(&allproc_lock); 1167 if (rootvnode == olddp) { 1168 vrele(rootvnode); 1169 VREF(newdp); 1170 rootvnode = newdp; 1171 } 1172} 1173 1174/* 1175 * Unmount a filesystem. 1176 * 1177 * Note: unmount takes a path to the vnode mounted on as argument, 1178 * not special file (as before). 1179 */ 1180#ifndef _SYS_SYSPROTO_H_ 1181struct unmount_args { 1182 char *path; 1183 int flags; 1184}; 1185#endif 1186/* ARGSUSED */ 1187int 1188unmount(td, uap) 1189 struct thread *td; 1190 register struct unmount_args /* { 1191 char *path; 1192 int flags; 1193 } */ *uap; 1194{ 1195 register struct vnode *vp; 1196 struct mount *mp; 1197 int error; 1198 struct nameidata nd; 1199 1200 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->path, td); 1201 if ((error = namei(&nd)) != 0) 1202 return (error); 1203 vp = nd.ni_vp; 1204 NDFREE(&nd, NDF_ONLY_PNBUF); 1205 mp = vp->v_mount; 1206 1207 /* 1208 * Only root, or the user that did the original mount is 1209 * permitted to unmount this filesystem. 1210 */ 1211 if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) { 1212 error = suser(td); 1213 if (error) { 1214 vput(vp); 1215 return (error); 1216 } 1217 } 1218 1219 /* 1220 * Don't allow unmounting the root filesystem. 1221 */ 1222 if (mp->mnt_flag & MNT_ROOTFS) { 1223 vput(vp); 1224 return (EINVAL); 1225 } 1226 1227 /* 1228 * Must be the root of the filesystem 1229 */ 1230 if ((vp->v_vflag & VV_ROOT) == 0) { 1231 vput(vp); 1232 return (EINVAL); 1233 } 1234 vput(vp); 1235 return (dounmount(mp, uap->flags, td)); 1236} 1237 1238/* 1239 * Do the actual filesystem unmount. 1240 */ 1241int 1242dounmount(mp, flags, td) 1243 struct mount *mp; 1244 int flags; 1245 struct thread *td; 1246{ 1247 struct vnode *coveredvp, *fsrootvp; 1248 int error; 1249 int async_flag; 1250 1251 mtx_lock(&mountlist_mtx); 1252 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 1253 mtx_unlock(&mountlist_mtx); 1254 return (EBUSY); 1255 } 1256 mp->mnt_kern_flag |= MNTK_UNMOUNT; 1257 /* Allow filesystems to detect that a forced unmount is in progress. */ 1258 if (flags & MNT_FORCE) 1259 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1260 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 1261 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_mtx, td); 1262 if (error) { 1263 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1264 if (mp->mnt_kern_flag & MNTK_MWAIT) 1265 wakeup(mp); 1266 return (error); 1267 } 1268 vn_start_write(NULL, &mp, V_WAIT); 1269 1270 if (mp->mnt_flag & MNT_EXPUBLIC) 1271 vfs_setpublicfs(NULL, NULL, NULL); 1272 1273 vfs_msync(mp, MNT_WAIT); 1274 async_flag = mp->mnt_flag & MNT_ASYNC; 1275 mp->mnt_flag &=~ MNT_ASYNC; 1276 cache_purgevfs(mp); /* remove cache entries for this file sys */ 1277 if (mp->mnt_syncer != NULL) 1278 vrele(mp->mnt_syncer); 1279 /* Move process cdir/rdir refs on fs root to underlying vnode. */ 1280 if (VFS_ROOT(mp, &fsrootvp) == 0) { 1281 if (mp->mnt_vnodecovered != NULL) 1282 checkdirs(fsrootvp, mp->mnt_vnodecovered); 1283 if (fsrootvp == rootvnode) { 1284 vrele(rootvnode); 1285 rootvnode = NULL; 1286 } 1287 vput(fsrootvp); 1288 } 1289 if (((mp->mnt_flag & MNT_RDONLY) || 1290 (error = VFS_SYNC(mp, MNT_WAIT, td->td_ucred, td)) == 0) || 1291 (flags & MNT_FORCE)) { 1292 error = VFS_UNMOUNT(mp, flags, td); 1293 } 1294 vn_finished_write(mp); 1295 if (error) { 1296 /* Undo cdir/rdir and rootvnode changes made above. */ 1297 if (VFS_ROOT(mp, &fsrootvp) == 0) { 1298 if (mp->mnt_vnodecovered != NULL) 1299 checkdirs(mp->mnt_vnodecovered, fsrootvp); 1300 if (rootvnode == NULL) { 1301 rootvnode = fsrootvp; 1302 vref(rootvnode); 1303 } 1304 vput(fsrootvp); 1305 } 1306 if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL) 1307 (void) vfs_allocate_syncvnode(mp); 1308 mtx_lock(&mountlist_mtx); 1309 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1310 mp->mnt_flag |= async_flag; 1311 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, 1312 &mountlist_mtx, td); 1313 if (mp->mnt_kern_flag & MNTK_MWAIT) 1314 wakeup(mp); 1315 return (error); 1316 } 1317 crfree(mp->mnt_cred); 1318 mtx_lock(&mountlist_mtx); 1319 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1320 if ((coveredvp = mp->mnt_vnodecovered) != NULL) 1321 coveredvp->v_mountedhere = NULL; 1322 mp->mnt_vfc->vfc_refcount--; 1323 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) 1324 panic("unmount: dangling vnode"); 1325 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_mtx, td); 1326 lockdestroy(&mp->mnt_lock); 1327 if (coveredvp != NULL) 1328 vrele(coveredvp); 1329 if (mp->mnt_kern_flag & MNTK_MWAIT) 1330 wakeup(mp); 1331#ifdef MAC 1332 mac_destroy_mount(mp); 1333#endif 1334 if (mp->mnt_op->vfs_mount == NULL) 1335 vfs_freeopts(mp->mnt_opt); 1336 free(mp, M_MOUNT); 1337 return (0); 1338} 1339 1340/* 1341 * Lookup a filesystem type, and if found allocate and initialize 1342 * a mount structure for it. 1343 * 1344 * Devname is usually updated by mount(8) after booting. 1345 */ 1346int 1347vfs_rootmountalloc(fstypename, devname, mpp) 1348 char *fstypename; 1349 char *devname; 1350 struct mount **mpp; 1351{ 1352 struct thread *td = curthread; /* XXX */ 1353 struct vfsconf *vfsp; 1354 struct mount *mp; 1355 1356 if (fstypename == NULL) 1357 return (ENODEV); 1358 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1359 if (!strcmp(vfsp->vfc_name, fstypename)) 1360 break; 1361 if (vfsp == NULL) 1362 return (ENODEV); 1363 mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 1364 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 1365 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 1366 TAILQ_INIT(&mp->mnt_nvnodelist); 1367 TAILQ_INIT(&mp->mnt_reservedvnlist); 1368 mp->mnt_vfc = vfsp; 1369 mp->mnt_op = vfsp->vfc_vfsops; 1370 mp->mnt_flag = MNT_RDONLY; 1371 mp->mnt_vnodecovered = NULLVP; 1372 mp->mnt_cred = crdup(td->td_ucred); 1373 vfsp->vfc_refcount++; 1374 mp->mnt_iosize_max = DFLTPHYS; 1375 mp->mnt_stat.f_type = vfsp->vfc_typenum; 1376 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 1377 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 1378 mp->mnt_stat.f_mntonname[0] = '/'; 1379 mp->mnt_stat.f_mntonname[1] = 0; 1380 strlcpy(mp->mnt_stat.f_mntfromname, devname, MNAMELEN); 1381#ifdef MAC 1382 mac_init_mount(mp); 1383 mac_create_mount(td->td_ucred, mp); 1384#endif 1385 *mpp = mp; 1386 return (0); 1387} 1388 1389/* 1390 * Find and mount the root filesystem 1391 */ 1392void 1393vfs_mountroot(void) 1394{ 1395 char *cp; 1396 int i, error; 1397 1398 g_waitidle(); 1399 1400 /* 1401 * The root filesystem information is compiled in, and we are 1402 * booted with instructions to use it. 1403 */ 1404#ifdef ROOTDEVNAME 1405 if ((boothowto & RB_DFLTROOT) && 1406 !vfs_mountroot_try(ROOTDEVNAME)) 1407 return; 1408#endif 1409 /* 1410 * We are booted with instructions to prompt for the root filesystem, 1411 * or to use the compiled-in default when it doesn't exist. 1412 */ 1413 if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) { 1414 if (!vfs_mountroot_ask()) 1415 return; 1416 } 1417 1418 /* 1419 * We've been given the generic "use CDROM as root" flag. This is 1420 * necessary because one media may be used in many different 1421 * devices, so we need to search for them. 1422 */ 1423 if (boothowto & RB_CDROM) { 1424 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 1425 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 1426 return; 1427 } 1428 } 1429 1430 /* 1431 * Try to use the value read by the loader from /etc/fstab, or 1432 * supplied via some other means. This is the preferred 1433 * mechanism. 1434 */ 1435 if ((cp = getenv("vfs.root.mountfrom")) != NULL) { 1436 error = vfs_mountroot_try(cp); 1437 freeenv(cp); 1438 if (!error) 1439 return; 1440 } 1441 1442 /* 1443 * Try values that may have been computed by the machine-dependant 1444 * legacy code. 1445 */ 1446 if (!vfs_mountroot_try(rootdevnames[0])) 1447 return; 1448 if (!vfs_mountroot_try(rootdevnames[1])) 1449 return; 1450 1451 /* 1452 * If we have a compiled-in default, and haven't already tried it, try 1453 * it now. 1454 */ 1455#ifdef ROOTDEVNAME 1456 if (!(boothowto & RB_DFLTROOT)) 1457 if (!vfs_mountroot_try(ROOTDEVNAME)) 1458 return; 1459#endif 1460 1461 /* 1462 * Everything so far has failed, prompt on the console if we haven't 1463 * already tried that. 1464 */ 1465 if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask()) 1466 return; 1467 panic("Root mount failed, startup aborted."); 1468} 1469 1470/* 1471 * Mount (mountfrom) as the root filesystem. 1472 */ 1473static int 1474vfs_mountroot_try(char *mountfrom) 1475{ 1476 struct mount *mp; 1477 char *vfsname, *path; 1478 const char *devname; 1479 int error; 1480 char patt[32]; 1481 int s; 1482 1483 vfsname = NULL; 1484 path = NULL; 1485 mp = NULL; 1486 error = EINVAL; 1487 1488 if (mountfrom == NULL) 1489 return(error); /* don't complain */ 1490 1491 s = splcam(); /* Overkill, but annoying without it */ 1492 printf("Mounting root from %s\n", mountfrom); 1493 splx(s); 1494 1495 /* parse vfs name and path */ 1496 vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK); 1497 path = malloc(MNAMELEN, M_MOUNT, M_WAITOK); 1498 vfsname[0] = path[0] = 0; 1499 sprintf(patt, "%%%d[a-z0-9]:%%%zds", MFSNAMELEN, MNAMELEN); 1500 if (sscanf(mountfrom, patt, vfsname, path) < 1) 1501 goto done; 1502 1503 /* allocate a root mount */ 1504 error = vfs_rootmountalloc(vfsname, path[0] != 0 ? path : ROOTNAME, 1505 &mp); 1506 if (error != 0) { 1507 printf("Can't allocate root mount for filesystem '%s': %d\n", 1508 vfsname, error); 1509 goto done; 1510 } 1511 mp->mnt_flag |= MNT_ROOTFS; 1512 1513 /* do our best to set rootdev */ 1514 if ((path[0] != 0) && setrootbyname(path)) 1515 printf("setrootbyname failed\n"); 1516 1517 /* If the root device is a type "memory disk", mount RW */ 1518 if (rootdev != NODEV && devsw(rootdev) != NULL) { 1519 devname = devtoname(rootdev); 1520 if (devname[0] == 'm' && devname[1] == 'd') 1521 mp->mnt_flag &= ~MNT_RDONLY; 1522 } 1523 1524 /* 1525 * Set the mount path to be something useful, because the 1526 * filesystem code isn't responsible now for initialising 1527 * f_mntonname unless they want to override the default 1528 * (which is `path'.) 1529 */ 1530 strlcpy(mp->mnt_stat.f_mntonname, "/", MNAMELEN); 1531 1532 error = VFS_MOUNT(mp, NULL, NULL, NULL, curthread); 1533 1534done: 1535 if (vfsname != NULL) 1536 free(vfsname, M_MOUNT); 1537 if (path != NULL) 1538 free(path, M_MOUNT); 1539 if (error != 0) { 1540 if (mp != NULL) { 1541 vfs_unbusy(mp, curthread); 1542#ifdef MAC 1543 mac_destroy_mount(mp); 1544#endif 1545 free(mp, M_MOUNT); 1546 } 1547 printf("Root mount failed: %d\n", error); 1548 } else { 1549 1550 /* register with list of mounted filesystems */ 1551 mtx_lock(&mountlist_mtx); 1552 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 1553 mtx_unlock(&mountlist_mtx); 1554 1555 /* sanity check system clock against root fs timestamp */ 1556 inittodr(mp->mnt_time); 1557 vfs_unbusy(mp, curthread); 1558 error = VFS_START(mp, 0, curthread); 1559 } 1560 return(error); 1561} 1562 1563/* 1564 * Spin prompting on the console for a suitable root filesystem 1565 */ 1566static int 1567vfs_mountroot_ask(void) 1568{ 1569 char name[128]; 1570 int i; 1571 dev_t dev; 1572 1573 for(;;) { 1574 printf("\nManual root filesystem specification:\n"); 1575 printf(" <fstype>:<device> Mount <device> using filesystem <fstype>\n"); 1576#if defined(__i386__) || defined(__ia64__) 1577 printf(" eg. ufs:da0s1a\n"); 1578#else 1579 printf(" eg. ufs:da0a\n"); 1580#endif 1581 printf(" ? List valid disk boot devices\n"); 1582 printf(" <empty line> Abort manual input\n"); 1583 printf("\nmountroot> "); 1584 gets(name); 1585 if (name[0] == 0) 1586 return(1); 1587 if (name[0] == '?') { 1588 if (!g_dev_print()) { 1589 printf("Possibly valid devices for 'ufs' root:\n"); 1590 for (i = 0; i < NUMCDEVSW; i++) { 1591 dev = makedev(i, 0); 1592 if (devsw(dev) != NULL) 1593 printf(" \"%s\"", devsw(dev)->d_name); 1594 } 1595 } 1596 printf("\n"); 1597 continue; 1598 } 1599 if (!vfs_mountroot_try(name)) 1600 return(0); 1601 } 1602} 1603 1604/* 1605 * Local helper function for vfs_mountroot_ask. 1606 */ 1607static void 1608gets(char *cp) 1609{ 1610 char *lp; 1611 int c; 1612 1613 lp = cp; 1614 for (;;) { 1615 printf("%c", c = cngetc() & 0177); 1616 switch (c) { 1617 case -1: 1618 case '\n': 1619 case '\r': 1620 *lp++ = '\0'; 1621 return; 1622 case '\b': 1623 case '\177': 1624 if (lp > cp) { 1625 printf(" \b"); 1626 lp--; 1627 } 1628 continue; 1629 case '#': 1630 lp--; 1631 if (lp < cp) 1632 lp = cp; 1633 continue; 1634 case '@': 1635 case 'u' & 037: 1636 lp = cp; 1637 printf("%c", '\n'); 1638 continue; 1639 default: 1640 *lp++ = c; 1641 } 1642 } 1643} 1644 1645/* 1646 * Convert a given name to the dev_t of the disk-like device 1647 * it refers to. 1648 */ 1649dev_t 1650getdiskbyname(char *name) { 1651 char *cp; 1652 dev_t dev; 1653 1654 cp = name; 1655 if (!bcmp(cp, "/dev/", 5)) 1656 cp += 5; 1657 1658 dev = NODEV; 1659 EVENTHANDLER_INVOKE(dev_clone, cp, strlen(cp), &dev); 1660 return (dev); 1661} 1662 1663/* 1664 * Set rootdev to match (name), given that we expect it to 1665 * refer to a disk-like device. 1666 */ 1667static int 1668setrootbyname(char *name) 1669{ 1670 dev_t diskdev; 1671 1672 diskdev = getdiskbyname(name); 1673 if (diskdev != NODEV) { 1674 rootdev = diskdev; 1675 return (0); 1676 } 1677 1678 return (1); 1679} 1680 1681/* Show the dev_t for a disk specified by name */ 1682#ifdef DDB 1683DB_SHOW_COMMAND(disk, db_getdiskbyname) 1684{ 1685 dev_t dev; 1686 1687 if (modif[0] == '\0') { 1688 db_error("usage: show disk/devicename"); 1689 return; 1690 } 1691 dev = getdiskbyname(modif); 1692 if (dev != NODEV) 1693 db_printf("dev_t = %p\n", dev); 1694 else 1695 db_printf("No disk device matched.\n"); 1696} 1697#endif 1698 1699/* 1700 * Get a mount option by its name. 1701 * 1702 * Return 0 if the option was found, ENOENT otherwise. 1703 * If len is non-NULL it will be filled with the length 1704 * of the option. If buf is non-NULL, it will be filled 1705 * with the address of the option. 1706 */ 1707int 1708vfs_getopt(opts, name, buf, len) 1709 struct vfsoptlist *opts; 1710 const char *name; 1711 void **buf; 1712 int *len; 1713{ 1714 struct vfsopt *opt; 1715 1716 TAILQ_FOREACH(opt, opts, link) { 1717 if (strcmp(name, opt->name) == 0) { 1718 if (len != NULL) 1719 *len = opt->len; 1720 if (buf != NULL) 1721 *buf = opt->value; 1722 return (0); 1723 } 1724 } 1725 return (ENOENT); 1726} 1727 1728/* 1729 * Find and copy a mount option. 1730 * 1731 * The size of the buffer has to be specified 1732 * in len, if it is not the same length as the 1733 * mount option, EINVAL is returned. 1734 * Returns ENOENT if the option is not found. 1735 */ 1736int 1737vfs_copyopt(opts, name, dest, len) 1738 struct vfsoptlist *opts; 1739 const char *name; 1740 void *dest; 1741 int len; 1742{ 1743 struct vfsopt *opt; 1744 1745 TAILQ_FOREACH(opt, opts, link) { 1746 if (strcmp(name, opt->name) == 0) { 1747 if (len != opt->len) 1748 return (EINVAL); 1749 bcopy(opt->value, dest, opt->len); 1750 return (0); 1751 } 1752 } 1753 return (ENOENT); 1754} 1755