vfs_mount.c revision 100363
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 100363 2002-07-19 16:05:31Z mux $ 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/malloc.h> 73#include <sys/mount.h> 74#include <sys/mutex.h> 75#include <sys/namei.h> 76#include <sys/proc.h> 77#include <sys/reboot.h> 78#include <sys/sysproto.h> 79#include <sys/sx.h> 80#include <sys/sysctl.h> 81#include <sys/sysent.h> 82#include <sys/systm.h> 83#include <sys/vnode.h> 84 85#include <machine/stdarg.h> 86 87#include "opt_rootdevname.h" 88#include "opt_ddb.h" 89 90#ifdef DDB 91#include <ddb/ddb.h> 92#endif 93 94#define ROOTNAME "root_device" 95 96static void checkdirs(struct vnode *olddp, struct vnode *newdp); 97static int vfs_nmount(struct thread *td, int, struct uio *); 98static int vfs_mountroot_try(char *mountfrom); 99static int vfs_mountroot_ask(void); 100static void gets(char *cp); 101 102static int usermount = 0; /* if 1, non-root can mount fs. */ 103SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, ""); 104 105MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 106 107/* List of mounted filesystems. */ 108struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist); 109 110/* For any iteration/modification of mountlist */ 111struct mtx mountlist_mtx; 112 113/* For any iteration/modification of mnt_vnodelist */ 114struct mtx mntvnode_mtx; 115 116/* 117 * The vnode of the system's root (/ in the filesystem, without chroot 118 * active.) 119 */ 120struct vnode *rootvnode; 121 122/* 123 * The root filesystem is detailed in the kernel environment variable 124 * vfs.root.mountfrom, which is expected to be in the general format 125 * 126 * <vfsname>:[<path>] 127 * vfsname := the name of a VFS known to the kernel and capable 128 * of being mounted as root 129 * path := disk device name or other data used by the filesystem 130 * to locate its physical store 131 */ 132 133/* 134 * The root specifiers we will try if RB_CDROM is specified. 135 */ 136static char *cdrom_rootdevnames[] = { 137 "cd9660:cd0a", 138 "cd9660:acd0a", 139 "cd9660:wcd0a", 140 NULL 141}; 142 143/* legacy find-root code */ 144char *rootdevnames[2] = {NULL, NULL}; 145static int setrootbyname(char *name); 146dev_t rootdev = NODEV; 147 148/* Remove one mount option. */ 149static void 150vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt) 151{ 152 153 TAILQ_REMOVE(opts, opt, link); 154 free(opt->name, M_MOUNT); 155 if (opt->value != NULL) 156 free(opt->value, M_MOUNT); 157#ifdef INVARIANTS 158 else if (opt->len != 0) 159 panic("%s: mount option with NULL value but length != 0", 160 __func__); 161#endif 162 free(opt, M_MOUNT); 163} 164 165/* Release all resources related to the mount options. */ 166static void 167vfs_freeopts(struct vfsoptlist *opts) 168{ 169 struct vfsopt *opt; 170 171 while (!TAILQ_EMPTY(opts)) { 172 opt = TAILQ_FIRST(opts); 173 vfs_freeopt(opts, opt); 174 } 175 free(opts, M_MOUNT); 176} 177 178/* 179 * If a mount option is specified several times, 180 * (with or without the "no" prefix) only keep 181 * the last occurence of it. 182 */ 183static void 184vfs_sanitizeopts(struct vfsoptlist *opts) 185{ 186 struct vfsopt *opt, *opt2, *tmp; 187 int noopt; 188 189 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) { 190 if (strncmp(opt->name, "no", 2) == 0) 191 noopt = 1; 192 else 193 noopt = 0; 194 opt2 = TAILQ_PREV(opt, vfsoptlist, link); 195 while (opt2 != NULL) { 196 if (strcmp(opt2->name, opt->name) == 0 || 197 (noopt && strcmp(opt->name + 2, opt2->name) == 0) || 198 (!noopt && strncmp(opt2->name, "no", 2) == 0 && 199 strcmp(opt2->name + 2, opt->name) == 0)) { 200 tmp = TAILQ_PREV(opt2, vfsoptlist, link); 201 vfs_freeopt(opts, opt2); 202 opt2 = tmp; 203 } else { 204 opt2 = TAILQ_PREV(opt2, vfsoptlist, link); 205 } 206 } 207 } 208} 209 210/* 211 * Build a linked list of mount options from a struct uio. 212 */ 213static int 214vfs_buildopts(struct uio *auio, struct vfsoptlist **options) 215{ 216 struct vfsoptlist *opts; 217 struct vfsopt *opt; 218 unsigned int i, iovcnt; 219 int error, namelen, optlen; 220 221 iovcnt = auio->uio_iovcnt; 222 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 223 TAILQ_INIT(opts); 224 for (i = 0; i < iovcnt; i += 2) { 225 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 226 namelen = auio->uio_iov[i].iov_len; 227 optlen = auio->uio_iov[i + 1].iov_len; 228 opt->name = malloc(namelen, M_MOUNT, M_WAITOK); 229 opt->len = optlen; 230 if (optlen == 0) { 231 opt->value = NULL; 232 } else { 233 opt->value = malloc(optlen, M_MOUNT, M_WAITOK); 234 if (auio->uio_segflg == UIO_SYSSPACE) { 235 bcopy(auio->uio_iov[i].iov_base, opt->name, 236 namelen); 237 bcopy(auio->uio_iov[i + 1].iov_base, opt->value, 238 optlen); 239 } else { 240 error = copyin(auio->uio_iov[i].iov_base, 241 opt->name, namelen); 242 if (error) 243 goto bad; 244 error = copyin(auio->uio_iov[i + 1].iov_base, 245 opt->value, optlen); 246 if (error) 247 goto bad; 248 } 249 } 250 TAILQ_INSERT_TAIL(opts, opt, link); 251 } 252 vfs_sanitizeopts(opts); 253 *options = opts; 254 return (0); 255bad: 256 vfs_freeopts(opts); 257 return (error); 258} 259 260/* 261 * Merge the old mount options with the new ones passed 262 * in the MNT_UPDATE case. 263 */ 264static void 265vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts) 266{ 267 struct vfsopt *opt, *opt2, *new; 268 269 TAILQ_FOREACH(opt, opts, link) { 270 /* 271 * Check that this option hasn't been redefined 272 * nor cancelled with a "no" mount option. 273 */ 274 opt2 = TAILQ_FIRST(toopts); 275 while (opt2 != NULL) { 276 if (strcmp(opt2->name, opt->name) == 0) 277 goto next; 278 if (strncmp(opt2->name, "no", 2) == 0 && 279 strcmp(opt2->name + 2, opt->name) == 0) { 280 vfs_freeopt(toopts, opt2); 281 goto next; 282 } 283 opt2 = TAILQ_NEXT(opt2, link); 284 } 285 /* We want this option, duplicate it. */ 286 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 287 new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK); 288 strcpy(new->name, opt->name); 289 if (opt->len != 0) { 290 new->value = malloc(opt->len, M_MOUNT, M_WAITOK); 291 bcopy(opt->value, new->value, opt->len); 292 } else { 293 new->value = NULL; 294 } 295 new->len = opt->len; 296 TAILQ_INSERT_TAIL(toopts, new, link); 297next: 298 continue; 299 } 300} 301 302/* 303 * New mount API. 304 */ 305int 306nmount(td, uap) 307 struct thread *td; 308 struct nmount_args /* { 309 syscallarg(struct iovec *) iovp; 310 syscallarg(unsigned int) iovcnt; 311 syscallarg(int) flags; 312 } */ *uap; 313{ 314 struct uio auio; 315 struct iovec *iov, *needfree; 316 struct iovec aiov[UIO_SMALLIOV]; 317 unsigned int i; 318 int error; 319 u_int iovlen, iovcnt; 320 321 iovcnt = SCARG(uap, iovcnt); 322 iovlen = iovcnt * sizeof (struct iovec); 323 /* 324 * Check that we have an even number of iovec's 325 * and that we have at least two options. 326 */ 327 if ((iovcnt & 1) || (iovcnt < 4) || (iovcnt > UIO_MAXIOV)) 328 return (EINVAL); 329 330 if (iovcnt > UIO_SMALLIOV) { 331 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 332 needfree = iov; 333 } else { 334 iov = aiov; 335 needfree = NULL; 336 } 337 auio.uio_iov = iov; 338 auio.uio_iovcnt = iovcnt; 339 auio.uio_segflg = UIO_USERSPACE; 340 if ((error = copyin(uap->iovp, iov, iovlen))) 341 goto finish; 342 343 for (i = 0; i < iovcnt; i++) { 344 if (iov->iov_len > MMAXOPTIONLEN) { 345 error = EINVAL; 346 goto finish; 347 } 348 iov++; 349 } 350 error = vfs_nmount(td, SCARG(uap, flags), &auio); 351finish: 352 if (needfree != NULL) 353 free(needfree, M_TEMP); 354 return (error); 355} 356 357int 358kernel_mount(iovp, iovcnt, flags) 359 struct iovec *iovp; 360 unsigned int iovcnt; 361 int flags; 362{ 363 struct uio auio; 364 int error; 365 366 /* 367 * Check that we have an even number of iovec's 368 * and that we have at least two options. 369 */ 370 if ((iovcnt & 1) || (iovcnt < 4)) 371 return (EINVAL); 372 373 auio.uio_iov = iovp; 374 auio.uio_iovcnt = iovcnt; 375 auio.uio_segflg = UIO_SYSSPACE; 376 377 error = vfs_nmount(curthread, flags, &auio); 378 return (error); 379} 380 381int 382kernel_vmount(int flags, ...) 383{ 384 struct iovec *iovp; 385 struct uio auio; 386 va_list ap; 387 unsigned int iovcnt, iovlen, len; 388 const char *cp; 389 char *buf, *pos; 390 size_t n; 391 int error, i; 392 393 len = 0; 394 va_start(ap, flags); 395 for (iovcnt = 0; (cp = va_arg(ap, const char *)) != NULL; iovcnt++) 396 len += strlen(cp) + 1; 397 va_end(ap); 398 399 if (iovcnt < 4 || iovcnt & 1) 400 return (EINVAL); 401 402 iovlen = iovcnt * sizeof (struct iovec); 403 MALLOC(iovp, struct iovec *, iovlen, M_MOUNT, M_WAITOK); 404 MALLOC(buf, char *, len, M_MOUNT, M_WAITOK); 405 pos = buf; 406 va_start(ap, flags); 407 for (i = 0; i < iovcnt; i++) { 408 cp = va_arg(ap, const char *); 409 copystr(cp, pos, len - (pos - buf), &n); 410 iovp[i].iov_base = pos; 411 iovp[i].iov_len = n; 412 pos += n; 413 } 414 va_end(ap); 415 416 auio.uio_iov = iovp; 417 auio.uio_iovcnt = iovcnt; 418 auio.uio_segflg = UIO_SYSSPACE; 419 420 error = vfs_nmount(curthread, flags, &auio); 421 FREE(iovp, M_MOUNT); 422 FREE(buf, M_MOUNT); 423 return (error); 424} 425 426/* 427 * vfs_nmount(): actually attempt a filesystem mount. 428 */ 429static int 430vfs_nmount(td, fsflags, fsoptions) 431 struct thread *td; 432 int fsflags; /* Flags common to all filesystems. */ 433 struct uio *fsoptions; /* Options local to the filesystem. */ 434{ 435 linker_file_t lf; 436 struct vnode *vp; 437 struct mount *mp; 438 struct vfsconf *vfsp; 439 struct vfsoptlist *optlist; 440 char *fstype, *fspath; 441 int error, flag = 0, kern_flag = 0; 442 int fstypelen, fspathlen; 443 struct vattr va; 444 struct nameidata nd; 445 446 error = vfs_buildopts(fsoptions, &optlist); 447 if (error) 448 return (error); 449 450 /* 451 * We need these two options before the others, 452 * and they are mandatory for any filesystem. 453 * Ensure they are NUL terminated as well. 454 */ 455 fstypelen = 0; 456 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen); 457 if (error || fstype[fstypelen - 1] != '\0') { 458 error = EINVAL; 459 goto bad; 460 } 461 fspathlen = 0; 462 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen); 463 if (error || fspath[fspathlen - 1] != '\0') { 464 error = EINVAL; 465 goto bad; 466 } 467 468 /* 469 * Be ultra-paranoid about making sure the type and fspath 470 * variables will fit in our mp buffers, including the 471 * terminating NUL. 472 */ 473 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) { 474 error = ENAMETOOLONG; 475 goto bad; 476 } 477 478 if (usermount == 0) { 479 error = suser(td); 480 if (error) 481 goto bad; 482 } 483 /* 484 * Do not allow NFS export by non-root users. 485 */ 486 if (fsflags & MNT_EXPORTED) { 487 error = suser(td); 488 if (error) 489 goto bad; 490 } 491 /* 492 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users. 493 */ 494 if (suser(td)) 495 fsflags |= MNT_NOSUID | MNT_NODEV; 496 /* 497 * Get vnode to be covered 498 */ 499 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 500 if ((error = namei(&nd)) != 0) 501 goto bad; 502 NDFREE(&nd, NDF_ONLY_PNBUF); 503 vp = nd.ni_vp; 504 if (fsflags & MNT_UPDATE) { 505 if ((vp->v_flag & VROOT) == 0) { 506 vput(vp); 507 error = EINVAL; 508 goto bad; 509 } 510 mp = vp->v_mount; 511 flag = mp->mnt_flag; 512 kern_flag = mp->mnt_kern_flag; 513 /* 514 * We only allow the filesystem to be reloaded if it 515 * is currently mounted read-only. 516 */ 517 if ((fsflags & MNT_RELOAD) && 518 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 519 vput(vp); 520 error = EOPNOTSUPP; /* Needs translation */ 521 goto bad; 522 } 523 /* 524 * Only root, or the user that did the original mount is 525 * permitted to update it. 526 */ 527 if (mp->mnt_stat.f_owner != td->td_ucred->cr_uid) { 528 error = suser(td); 529 if (error) { 530 vput(vp); 531 goto bad; 532 } 533 } 534 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 535 vput(vp); 536 error = EBUSY; 537 goto bad; 538 } 539 mtx_lock(&vp->v_interlock); 540 if ((vp->v_flag & VMOUNT) != 0 || vp->v_mountedhere != NULL) { 541 mtx_unlock(&vp->v_interlock); 542 vfs_unbusy(mp, td); 543 vput(vp); 544 error = EBUSY; 545 goto bad; 546 } 547 vp->v_flag |= VMOUNT; 548 mtx_unlock(&vp->v_interlock); 549 mp->mnt_flag |= fsflags & 550 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 551 VOP_UNLOCK(vp, 0, td); 552 mp->mnt_optnew = optlist; 553 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 554 goto update; 555 } 556 /* 557 * If the user is not root, ensure that they own the directory 558 * onto which we are attempting to mount. 559 */ 560 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 561 if (error) { 562 vput(vp); 563 goto bad; 564 } 565 if (va.va_uid != td->td_ucred->cr_uid) { 566 error = suser(td); 567 if (error) { 568 vput(vp); 569 goto bad; 570 } 571 } 572 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 573 vput(vp); 574 goto bad; 575 } 576 if (vp->v_type != VDIR) { 577 vput(vp); 578 error = ENOTDIR; 579 goto bad; 580 } 581 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 582 if (!strcmp(vfsp->vfc_name, fstype)) 583 break; 584 if (vfsp == NULL) { 585 /* Only load modules for root (very important!). */ 586 error = suser(td); 587 if (error) { 588 vput(vp); 589 goto bad; 590 } 591 error = securelevel_gt(td->td_ucred, 0); 592 if (error) { 593 vput(vp); 594 goto bad; 595 } 596 error = linker_load_file(fstype, &lf); 597 if (error || lf == NULL) { 598 vput(vp); 599 if (lf == NULL) 600 error = ENODEV; 601 goto bad; 602 } 603 lf->userrefs++; 604 /* Look up again to see if the VFS was loaded. */ 605 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 606 if (!strcmp(vfsp->vfc_name, fstype)) 607 break; 608 if (vfsp == NULL) { 609 lf->userrefs--; 610 linker_file_unload(lf); 611 vput(vp); 612 error = ENODEV; 613 goto bad; 614 } 615 } 616 mtx_lock(&vp->v_interlock); 617 if ((vp->v_flag & VMOUNT) != 0 || 618 vp->v_mountedhere != NULL) { 619 mtx_unlock(&vp->v_interlock); 620 vput(vp); 621 error = EBUSY; 622 goto bad; 623 } 624 vp->v_flag |= VMOUNT; 625 mtx_unlock(&vp->v_interlock); 626 627 /* 628 * Allocate and initialize the filesystem. 629 */ 630 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 631 TAILQ_INIT(&mp->mnt_nvnodelist); 632 TAILQ_INIT(&mp->mnt_reservedvnlist); 633 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 634 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 635 mp->mnt_op = vfsp->vfc_vfsops; 636 mp->mnt_vfc = vfsp; 637 vfsp->vfc_refcount++; 638 mp->mnt_stat.f_type = vfsp->vfc_typenum; 639 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 640 strncpy(mp->mnt_stat.f_fstypename, fstype, MFSNAMELEN); 641 mp->mnt_vnodecovered = vp; 642 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 643 strncpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 644 mp->mnt_iosize_max = DFLTPHYS; 645 VOP_UNLOCK(vp, 0, td); 646 mp->mnt_optnew = optlist; 647 648update: 649 /* 650 * Check if the fs implements the new VFS_NMOUNT() 651 * function, since the new system call was used. 652 */ 653 if (mp->mnt_op->vfs_mount != NULL) { 654 printf("%s doesn't support the new mount syscall\n", 655 mp->mnt_vfc->vfc_name); 656 mtx_lock(&vp->v_interlock); 657 vp->v_flag &= ~VMOUNT; 658 mtx_unlock(&vp->v_interlock); 659 if (mp->mnt_flag & MNT_UPDATE) 660 vfs_unbusy(mp, td); 661 else { 662 mp->mnt_vfc->vfc_refcount--; 663 vfs_unbusy(mp, td); 664 free(mp, M_MOUNT); 665 } 666 vrele(vp); 667 error = EOPNOTSUPP; 668 goto bad; 669 } 670 671 /* 672 * Set the mount level flags. 673 */ 674 if (fsflags & MNT_RDONLY) 675 mp->mnt_flag |= MNT_RDONLY; 676 else if (mp->mnt_flag & MNT_RDONLY) 677 mp->mnt_kern_flag |= MNTK_WANTRDWR; 678 mp->mnt_flag &=~ MNT_UPDATEMASK; 679 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 680 /* 681 * Mount the filesystem. 682 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 683 * get. No freeing of cn_pnbuf. 684 */ 685 error = VFS_NMOUNT(mp, &nd, td); 686 if (!error) { 687 if (mp->mnt_opt != NULL) 688 vfs_freeopts(mp->mnt_opt); 689 mp->mnt_opt = mp->mnt_optnew; 690 } 691 /* 692 * Prevent external consumers of mount 693 * options to read mnt_optnew. 694 */ 695 mp->mnt_optnew = NULL; 696 if (mp->mnt_flag & MNT_UPDATE) { 697 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 698 mp->mnt_flag &= ~MNT_RDONLY; 699 mp->mnt_flag &=~ 700 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 701 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 702 if (error) { 703 mp->mnt_flag = flag; 704 mp->mnt_kern_flag = kern_flag; 705 } 706 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 707 if (mp->mnt_syncer == NULL) 708 error = vfs_allocate_syncvnode(mp); 709 } else { 710 if (mp->mnt_syncer != NULL) 711 vput(mp->mnt_syncer); 712 mp->mnt_syncer = NULL; 713 } 714 vfs_unbusy(mp, td); 715 mtx_lock(&vp->v_interlock); 716 vp->v_flag &= ~VMOUNT; 717 mtx_unlock(&vp->v_interlock); 718 vrele(vp); 719 return (error); 720 } 721 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 722 /* 723 * Put the new filesystem on the mount list after root. 724 */ 725 cache_purge(vp); 726 if (!error) { 727 struct vnode *newdp; 728 729 mtx_lock(&vp->v_interlock); 730 vp->v_flag &= ~VMOUNT; 731 vp->v_mountedhere = mp; 732 mtx_unlock(&vp->v_interlock); 733 mtx_lock(&mountlist_mtx); 734 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 735 mtx_unlock(&mountlist_mtx); 736 if (VFS_ROOT(mp, &newdp)) 737 panic("mount: lost mount"); 738 checkdirs(vp, newdp); 739 vput(newdp); 740 VOP_UNLOCK(vp, 0, td); 741 if ((mp->mnt_flag & MNT_RDONLY) == 0) 742 error = vfs_allocate_syncvnode(mp); 743 vfs_unbusy(mp, td); 744 if ((error = VFS_START(mp, 0, td)) != 0) { 745 vrele(vp); 746 goto bad; 747 } 748 } else { 749 mtx_lock(&vp->v_interlock); 750 vp->v_flag &= ~VMOUNT; 751 mtx_unlock(&vp->v_interlock); 752 mp->mnt_vfc->vfc_refcount--; 753 vfs_unbusy(mp, td); 754 free(mp, M_MOUNT); 755 vput(vp); 756 goto bad; 757 } 758 return (0); 759bad: 760 vfs_freeopts(optlist); 761 return (error); 762} 763 764/* 765 * Old mount API. 766 */ 767#ifndef _SYS_SYSPROTO_H_ 768struct mount_args { 769 char *type; 770 char *path; 771 int flags; 772 caddr_t data; 773}; 774#endif 775/* ARGSUSED */ 776int 777mount(td, uap) 778 struct thread *td; 779 struct mount_args /* { 780 syscallarg(char *) type; 781 syscallarg(char *) path; 782 syscallarg(int) flags; 783 syscallarg(caddr_t) data; 784 } */ *uap; 785{ 786 char *fstype; 787 char *fspath; 788 int error; 789 790 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK); 791 fspath = malloc(MNAMELEN, M_TEMP, M_WAITOK); 792 793 /* 794 * vfs_mount() actually takes a kernel string for `type' and 795 * `path' now, so extract them. 796 */ 797 error = copyinstr(SCARG(uap, type), fstype, MFSNAMELEN, NULL); 798 if (error) 799 goto finish; 800 error = copyinstr(SCARG(uap, path), fspath, MNAMELEN, NULL); 801 if (error) 802 goto finish; 803 error = vfs_mount(td, fstype, fspath, SCARG(uap, flags), 804 SCARG(uap, data)); 805finish: 806 free(fstype, M_TEMP); 807 free(fspath, M_TEMP); 808 return (error); 809} 810 811/* 812 * vfs_mount(): actually attempt a filesystem mount. 813 * 814 * This routine is designed to be a "generic" entry point for routines 815 * that wish to mount a filesystem. All parameters except `fsdata' are 816 * pointers into kernel space. `fsdata' is currently still a pointer 817 * into userspace. 818 */ 819int 820vfs_mount(td, fstype, fspath, fsflags, fsdata) 821 struct thread *td; 822 const char *fstype; 823 char *fspath; 824 int fsflags; 825 void *fsdata; 826{ 827 linker_file_t lf; 828 struct vnode *vp; 829 struct mount *mp; 830 struct vfsconf *vfsp; 831 int error, flag = 0, kern_flag = 0; 832 struct vattr va; 833 struct nameidata nd; 834 835 /* 836 * Be ultra-paranoid about making sure the type and fspath 837 * variables will fit in our mp buffers, including the 838 * terminating NUL. 839 */ 840 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 841 return (ENAMETOOLONG); 842 843 if (usermount == 0) { 844 error = suser(td); 845 if (error) 846 return (error); 847 } 848 /* 849 * Do not allow NFS export by non-root users. 850 */ 851 if (fsflags & MNT_EXPORTED) { 852 error = suser(td); 853 if (error) 854 return (error); 855 } 856 /* 857 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users. 858 */ 859 if (suser(td)) 860 fsflags |= MNT_NOSUID | MNT_NODEV; 861 /* 862 * Get vnode to be covered 863 */ 864 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 865 if ((error = namei(&nd)) != 0) 866 return (error); 867 NDFREE(&nd, NDF_ONLY_PNBUF); 868 vp = nd.ni_vp; 869 if (fsflags & MNT_UPDATE) { 870 if ((vp->v_flag & VROOT) == 0) { 871 vput(vp); 872 return (EINVAL); 873 } 874 mp = vp->v_mount; 875 flag = mp->mnt_flag; 876 kern_flag = mp->mnt_kern_flag; 877 /* 878 * We only allow the filesystem to be reloaded if it 879 * is currently mounted read-only. 880 */ 881 if ((fsflags & MNT_RELOAD) && 882 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 883 vput(vp); 884 return (EOPNOTSUPP); /* Needs translation */ 885 } 886 /* 887 * Only root, or the user that did the original mount is 888 * permitted to update it. 889 */ 890 if (mp->mnt_stat.f_owner != td->td_ucred->cr_uid) { 891 error = suser(td); 892 if (error) { 893 vput(vp); 894 return (error); 895 } 896 } 897 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 898 vput(vp); 899 return (EBUSY); 900 } 901 mtx_lock(&vp->v_interlock); 902 if ((vp->v_flag & VMOUNT) != 0 || vp->v_mountedhere != NULL) { 903 mtx_unlock(&vp->v_interlock); 904 vfs_unbusy(mp, td); 905 vput(vp); 906 return (EBUSY); 907 } 908 vp->v_flag |= VMOUNT; 909 mtx_unlock(&vp->v_interlock); 910 mp->mnt_flag |= fsflags & 911 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 912 VOP_UNLOCK(vp, 0, td); 913 goto update; 914 } 915 /* 916 * If the user is not root, ensure that they own the directory 917 * onto which we are attempting to mount. 918 */ 919 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 920 if (error) { 921 vput(vp); 922 return (error); 923 } 924 if (va.va_uid != td->td_ucred->cr_uid) { 925 error = suser(td); 926 if (error) { 927 vput(vp); 928 return (error); 929 } 930 } 931 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 932 vput(vp); 933 return (error); 934 } 935 if (vp->v_type != VDIR) { 936 vput(vp); 937 return (ENOTDIR); 938 } 939 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 940 if (!strcmp(vfsp->vfc_name, fstype)) 941 break; 942 if (vfsp == NULL) { 943 /* Only load modules for root (very important!). */ 944 error = suser(td); 945 if (error) { 946 vput(vp); 947 return (error); 948 } 949 error = securelevel_gt(td->td_ucred, 0); 950 if (error) { 951 vput(vp); 952 return (error); 953 } 954 error = linker_load_file(fstype, &lf); 955 if (error || lf == NULL) { 956 vput(vp); 957 if (lf == NULL) 958 error = ENODEV; 959 return (error); 960 } 961 lf->userrefs++; 962 /* Look up again to see if the VFS was loaded. */ 963 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 964 if (!strcmp(vfsp->vfc_name, fstype)) 965 break; 966 if (vfsp == NULL) { 967 lf->userrefs--; 968 linker_file_unload(lf); 969 vput(vp); 970 return (ENODEV); 971 } 972 } 973 mtx_lock(&vp->v_interlock); 974 if ((vp->v_flag & VMOUNT) != 0 || 975 vp->v_mountedhere != NULL) { 976 mtx_unlock(&vp->v_interlock); 977 vput(vp); 978 return (EBUSY); 979 } 980 vp->v_flag |= VMOUNT; 981 mtx_unlock(&vp->v_interlock); 982 983 /* 984 * Allocate and initialize the filesystem. 985 */ 986 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 987 TAILQ_INIT(&mp->mnt_nvnodelist); 988 TAILQ_INIT(&mp->mnt_reservedvnlist); 989 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 990 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 991 mp->mnt_op = vfsp->vfc_vfsops; 992 mp->mnt_vfc = vfsp; 993 vfsp->vfc_refcount++; 994 mp->mnt_stat.f_type = vfsp->vfc_typenum; 995 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 996 strncpy(mp->mnt_stat.f_fstypename, fstype, MFSNAMELEN); 997 mp->mnt_vnodecovered = vp; 998 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 999 strncpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 1000 mp->mnt_iosize_max = DFLTPHYS; 1001 VOP_UNLOCK(vp, 0, td); 1002update: 1003 /* 1004 * Check if the fs implements the old VFS_MOUNT() 1005 * function, since the old system call was used. 1006 */ 1007 if (mp->mnt_op->vfs_mount == NULL) { 1008 printf("%s doesn't support the old mount syscall\n", 1009 mp->mnt_vfc->vfc_name); 1010 mtx_lock(&vp->v_interlock); 1011 vp->v_flag &= ~VMOUNT; 1012 mtx_unlock(&vp->v_interlock); 1013 if (mp->mnt_flag & MNT_UPDATE) 1014 vfs_unbusy(mp, td); 1015 else { 1016 mp->mnt_vfc->vfc_refcount--; 1017 vfs_unbusy(mp, td); 1018 free(mp, M_MOUNT); 1019 } 1020 vrele(vp); 1021 return (EOPNOTSUPP); 1022 } 1023 1024 /* 1025 * Set the mount level flags. 1026 */ 1027 if (fsflags & MNT_RDONLY) 1028 mp->mnt_flag |= MNT_RDONLY; 1029 else if (mp->mnt_flag & MNT_RDONLY) 1030 mp->mnt_kern_flag |= MNTK_WANTRDWR; 1031 mp->mnt_flag &=~ MNT_UPDATEMASK; 1032 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 1033 /* 1034 * Mount the filesystem. 1035 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 1036 * get. No freeing of cn_pnbuf. 1037 */ 1038 error = VFS_MOUNT(mp, fspath, fsdata, &nd, td); 1039 if (mp->mnt_flag & MNT_UPDATE) { 1040 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 1041 mp->mnt_flag &= ~MNT_RDONLY; 1042 mp->mnt_flag &=~ 1043 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 1044 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 1045 if (error) { 1046 mp->mnt_flag = flag; 1047 mp->mnt_kern_flag = kern_flag; 1048 } 1049 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 1050 if (mp->mnt_syncer == NULL) 1051 error = vfs_allocate_syncvnode(mp); 1052 } else { 1053 if (mp->mnt_syncer != NULL) 1054 vput(mp->mnt_syncer); 1055 mp->mnt_syncer = NULL; 1056 } 1057 vfs_unbusy(mp, td); 1058 mtx_lock(&vp->v_interlock); 1059 vp->v_flag &= ~VMOUNT; 1060 mtx_unlock(&vp->v_interlock); 1061 vrele(vp); 1062 return (error); 1063 } 1064 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 1065 /* 1066 * Put the new filesystem on the mount list after root. 1067 */ 1068 cache_purge(vp); 1069 if (!error) { 1070 struct vnode *newdp; 1071 1072 mtx_lock(&vp->v_interlock); 1073 vp->v_flag &= ~VMOUNT; 1074 vp->v_mountedhere = mp; 1075 mtx_unlock(&vp->v_interlock); 1076 mtx_lock(&mountlist_mtx); 1077 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 1078 mtx_unlock(&mountlist_mtx); 1079 if (VFS_ROOT(mp, &newdp)) 1080 panic("mount: lost mount"); 1081 checkdirs(vp, newdp); 1082 vput(newdp); 1083 VOP_UNLOCK(vp, 0, td); 1084 if ((mp->mnt_flag & MNT_RDONLY) == 0) 1085 error = vfs_allocate_syncvnode(mp); 1086 vfs_unbusy(mp, td); 1087 if ((error = VFS_START(mp, 0, td)) != 0) 1088 vrele(vp); 1089 } else { 1090 mtx_lock(&vp->v_interlock); 1091 vp->v_flag &= ~VMOUNT; 1092 mtx_unlock(&vp->v_interlock); 1093 mp->mnt_vfc->vfc_refcount--; 1094 vfs_unbusy(mp, td); 1095 free(mp, M_MOUNT); 1096 vput(vp); 1097 } 1098 return (error); 1099} 1100 1101/* 1102 * Scan all active processes to see if any of them have a current 1103 * or root directory of `olddp'. If so, replace them with the new 1104 * mount point. 1105 */ 1106static void 1107checkdirs(olddp, newdp) 1108 struct vnode *olddp, *newdp; 1109{ 1110 struct filedesc *fdp; 1111 struct proc *p; 1112 int nrele; 1113 1114 if (olddp->v_usecount == 1) 1115 return; 1116 sx_slock(&allproc_lock); 1117 LIST_FOREACH(p, &allproc, p_list) { 1118 PROC_LOCK(p); 1119 fdp = p->p_fd; 1120 if (fdp == NULL) { 1121 PROC_UNLOCK(p); 1122 continue; 1123 } 1124 nrele = 0; 1125 FILEDESC_LOCK(fdp); 1126 if (fdp->fd_cdir == olddp) { 1127 VREF(newdp); 1128 fdp->fd_cdir = newdp; 1129 nrele++; 1130 } 1131 if (fdp->fd_rdir == olddp) { 1132 VREF(newdp); 1133 fdp->fd_rdir = newdp; 1134 nrele++; 1135 } 1136 FILEDESC_UNLOCK(fdp); 1137 PROC_UNLOCK(p); 1138 while (nrele--) 1139 vrele(olddp); 1140 } 1141 sx_sunlock(&allproc_lock); 1142 if (rootvnode == olddp) { 1143 vrele(rootvnode); 1144 VREF(newdp); 1145 rootvnode = newdp; 1146 } 1147} 1148 1149/* 1150 * Unmount a filesystem. 1151 * 1152 * Note: unmount takes a path to the vnode mounted on as argument, 1153 * not special file (as before). 1154 */ 1155#ifndef _SYS_SYSPROTO_H_ 1156struct unmount_args { 1157 char *path; 1158 int flags; 1159}; 1160#endif 1161/* ARGSUSED */ 1162int 1163unmount(td, uap) 1164 struct thread *td; 1165 register struct unmount_args /* { 1166 syscallarg(char *) path; 1167 syscallarg(int) flags; 1168 } */ *uap; 1169{ 1170 register struct vnode *vp; 1171 struct mount *mp; 1172 int error; 1173 struct nameidata nd; 1174 1175 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1176 SCARG(uap, path), td); 1177 if ((error = namei(&nd)) != 0) 1178 return (error); 1179 vp = nd.ni_vp; 1180 NDFREE(&nd, NDF_ONLY_PNBUF); 1181 mp = vp->v_mount; 1182 1183 /* 1184 * Only root, or the user that did the original mount is 1185 * permitted to unmount this filesystem. 1186 */ 1187 if (mp->mnt_stat.f_owner != td->td_ucred->cr_uid) { 1188 error = suser(td); 1189 if (error) { 1190 vput(vp); 1191 return (error); 1192 } 1193 } 1194 1195 /* 1196 * Don't allow unmounting the root filesystem. 1197 */ 1198 if (mp->mnt_flag & MNT_ROOTFS) { 1199 vput(vp); 1200 return (EINVAL); 1201 } 1202 1203 /* 1204 * Must be the root of the filesystem 1205 */ 1206 if ((vp->v_flag & VROOT) == 0) { 1207 vput(vp); 1208 return (EINVAL); 1209 } 1210 vput(vp); 1211 return (dounmount(mp, SCARG(uap, flags), td)); 1212} 1213 1214/* 1215 * Do the actual filesystem unmount. 1216 */ 1217int 1218dounmount(mp, flags, td) 1219 struct mount *mp; 1220 int flags; 1221 struct thread *td; 1222{ 1223 struct vnode *coveredvp, *fsrootvp; 1224 int error; 1225 int async_flag; 1226 1227 mtx_lock(&mountlist_mtx); 1228 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 1229 mtx_unlock(&mountlist_mtx); 1230 return (EBUSY); 1231 } 1232 mp->mnt_kern_flag |= MNTK_UNMOUNT; 1233 /* Allow filesystems to detect that a forced unmount is in progress. */ 1234 if (flags & MNT_FORCE) 1235 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1236 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 1237 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_mtx, td); 1238 if (error) { 1239 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1240 if (mp->mnt_kern_flag & MNTK_MWAIT) 1241 wakeup(mp); 1242 return (error); 1243 } 1244 vn_start_write(NULL, &mp, V_WAIT); 1245 1246 if (mp->mnt_flag & MNT_EXPUBLIC) 1247 vfs_setpublicfs(NULL, NULL, NULL); 1248 1249 vfs_msync(mp, MNT_WAIT); 1250 async_flag = mp->mnt_flag & MNT_ASYNC; 1251 mp->mnt_flag &=~ MNT_ASYNC; 1252 cache_purgevfs(mp); /* remove cache entries for this file sys */ 1253 if (mp->mnt_syncer != NULL) 1254 vput(mp->mnt_syncer); 1255 /* Move process cdir/rdir refs on fs root to underlying vnode. */ 1256 if (VFS_ROOT(mp, &fsrootvp) == 0) { 1257 if (mp->mnt_vnodecovered != NULL) 1258 checkdirs(fsrootvp, mp->mnt_vnodecovered); 1259 if (fsrootvp == rootvnode) { 1260 vrele(rootvnode); 1261 rootvnode = NULL; 1262 } 1263 vput(fsrootvp); 1264 } 1265 if (((mp->mnt_flag & MNT_RDONLY) || 1266 (error = VFS_SYNC(mp, MNT_WAIT, td->td_ucred, td)) == 0) || 1267 (flags & MNT_FORCE)) { 1268 error = VFS_UNMOUNT(mp, flags, td); 1269 } 1270 vn_finished_write(mp); 1271 if (error) { 1272 /* Undo cdir/rdir and rootvnode changes made above. */ 1273 if (VFS_ROOT(mp, &fsrootvp) == 0) { 1274 if (mp->mnt_vnodecovered != NULL) 1275 checkdirs(mp->mnt_vnodecovered, fsrootvp); 1276 if (rootvnode == NULL) { 1277 rootvnode = fsrootvp; 1278 vref(rootvnode); 1279 } 1280 vput(fsrootvp); 1281 } 1282 if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL) 1283 (void) vfs_allocate_syncvnode(mp); 1284 mtx_lock(&mountlist_mtx); 1285 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1286 mp->mnt_flag |= async_flag; 1287 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, 1288 &mountlist_mtx, td); 1289 if (mp->mnt_kern_flag & MNTK_MWAIT) 1290 wakeup(mp); 1291 return (error); 1292 } 1293 mtx_lock(&mountlist_mtx); 1294 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1295 if ((coveredvp = mp->mnt_vnodecovered) != NULL) 1296 coveredvp->v_mountedhere = NULL; 1297 mp->mnt_vfc->vfc_refcount--; 1298 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) 1299 panic("unmount: dangling vnode"); 1300 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_mtx, td); 1301 lockdestroy(&mp->mnt_lock); 1302 if (coveredvp != NULL) 1303 vrele(coveredvp); 1304 if (mp->mnt_kern_flag & MNTK_MWAIT) 1305 wakeup(mp); 1306 if (mp->mnt_op->vfs_mount == NULL) 1307 vfs_freeopts(mp->mnt_opt); 1308 free(mp, M_MOUNT); 1309 return (0); 1310} 1311 1312/* 1313 * Lookup a filesystem type, and if found allocate and initialize 1314 * a mount structure for it. 1315 * 1316 * Devname is usually updated by mount(8) after booting. 1317 */ 1318int 1319vfs_rootmountalloc(fstypename, devname, mpp) 1320 char *fstypename; 1321 char *devname; 1322 struct mount **mpp; 1323{ 1324 struct thread *td = curthread; /* XXX */ 1325 struct vfsconf *vfsp; 1326 struct mount *mp; 1327 1328 if (fstypename == NULL) 1329 return (ENODEV); 1330 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1331 if (!strcmp(vfsp->vfc_name, fstypename)) 1332 break; 1333 if (vfsp == NULL) 1334 return (ENODEV); 1335 mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 1336 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 1337 (void)vfs_busy(mp, LK_NOWAIT, 0, td); 1338 TAILQ_INIT(&mp->mnt_nvnodelist); 1339 TAILQ_INIT(&mp->mnt_reservedvnlist); 1340 mp->mnt_vfc = vfsp; 1341 mp->mnt_op = vfsp->vfc_vfsops; 1342 mp->mnt_flag = MNT_RDONLY; 1343 mp->mnt_vnodecovered = NULLVP; 1344 vfsp->vfc_refcount++; 1345 mp->mnt_iosize_max = DFLTPHYS; 1346 mp->mnt_stat.f_type = vfsp->vfc_typenum; 1347 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 1348 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 1349 mp->mnt_stat.f_mntonname[0] = '/'; 1350 mp->mnt_stat.f_mntonname[1] = 0; 1351 (void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); 1352 *mpp = mp; 1353 return (0); 1354} 1355 1356/* 1357 * Find and mount the root filesystem 1358 */ 1359void 1360vfs_mountroot(void) 1361{ 1362 char *cp; 1363 int i, error; 1364 1365 /* 1366 * The root filesystem information is compiled in, and we are 1367 * booted with instructions to use it. 1368 */ 1369#ifdef ROOTDEVNAME 1370 if ((boothowto & RB_DFLTROOT) && 1371 !vfs_mountroot_try(ROOTDEVNAME)) 1372 return; 1373#endif 1374 /* 1375 * We are booted with instructions to prompt for the root filesystem, 1376 * or to use the compiled-in default when it doesn't exist. 1377 */ 1378 if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) { 1379 if (!vfs_mountroot_ask()) 1380 return; 1381 } 1382 1383 /* 1384 * We've been given the generic "use CDROM as root" flag. This is 1385 * necessary because one media may be used in many different 1386 * devices, so we need to search for them. 1387 */ 1388 if (boothowto & RB_CDROM) { 1389 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 1390 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 1391 return; 1392 } 1393 } 1394 1395 /* 1396 * Try to use the value read by the loader from /etc/fstab, or 1397 * supplied via some other means. This is the preferred 1398 * mechanism. 1399 */ 1400 if ((cp = getenv("vfs.root.mountfrom")) != NULL) { 1401 error = vfs_mountroot_try(cp); 1402 freeenv(cp); 1403 if (!error) 1404 return; 1405 } 1406 1407 /* 1408 * Try values that may have been computed by the machine-dependant 1409 * legacy code. 1410 */ 1411 if (!vfs_mountroot_try(rootdevnames[0])) 1412 return; 1413 if (!vfs_mountroot_try(rootdevnames[1])) 1414 return; 1415 1416 /* 1417 * If we have a compiled-in default, and haven't already tried it, try 1418 * it now. 1419 */ 1420#ifdef ROOTDEVNAME 1421 if (!(boothowto & RB_DFLTROOT)) 1422 if (!vfs_mountroot_try(ROOTDEVNAME)) 1423 return; 1424#endif 1425 1426 /* 1427 * Everything so far has failed, prompt on the console if we haven't 1428 * already tried that. 1429 */ 1430 if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask()) 1431 return; 1432 panic("Root mount failed, startup aborted."); 1433} 1434 1435/* 1436 * Mount (mountfrom) as the root filesystem. 1437 */ 1438static int 1439vfs_mountroot_try(char *mountfrom) 1440{ 1441 struct mount *mp; 1442 char *vfsname, *path; 1443 int error; 1444 char patt[32]; 1445 int s; 1446 1447 vfsname = NULL; 1448 path = NULL; 1449 mp = NULL; 1450 error = EINVAL; 1451 1452 if (mountfrom == NULL) 1453 return(error); /* don't complain */ 1454 1455 s = splcam(); /* Overkill, but annoying without it */ 1456 printf("Mounting root from %s\n", mountfrom); 1457 splx(s); 1458 1459 /* parse vfs name and path */ 1460 vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK); 1461 path = malloc(MNAMELEN, M_MOUNT, M_WAITOK); 1462 vfsname[0] = path[0] = 0; 1463 sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN); 1464 if (sscanf(mountfrom, patt, vfsname, path) < 1) 1465 goto done; 1466 1467 /* allocate a root mount */ 1468 error = vfs_rootmountalloc(vfsname, path[0] != 0 ? path : ROOTNAME, 1469 &mp); 1470 if (error != 0) { 1471 printf("Can't allocate root mount for filesystem '%s': %d\n", 1472 vfsname, error); 1473 goto done; 1474 } 1475 mp->mnt_flag |= MNT_ROOTFS; 1476 1477 /* do our best to set rootdev */ 1478 if ((path[0] != 0) && setrootbyname(path)) 1479 printf("setrootbyname failed\n"); 1480 1481 /* If the root device is a type "memory disk", mount RW */ 1482 if (rootdev != NODEV && devsw(rootdev) && 1483 (devsw(rootdev)->d_flags & D_MEMDISK)) 1484 mp->mnt_flag &= ~MNT_RDONLY; 1485 1486 /* 1487 * Set the mount path to be something useful, because the 1488 * filesystem code isn't responsible now for initialising 1489 * f_mntonname unless they want to override the default 1490 * (which is `path'.) 1491 */ 1492 strncpy(mp->mnt_stat.f_mntonname, "/", MNAMELEN); 1493 1494 error = VFS_MOUNT(mp, NULL, NULL, NULL, curthread); 1495 1496done: 1497 if (vfsname != NULL) 1498 free(vfsname, M_MOUNT); 1499 if (path != NULL) 1500 free(path, M_MOUNT); 1501 if (error != 0) { 1502 if (mp != NULL) { 1503 vfs_unbusy(mp, curthread); 1504 free(mp, M_MOUNT); 1505 } 1506 printf("Root mount failed: %d\n", error); 1507 } else { 1508 1509 /* register with list of mounted filesystems */ 1510 mtx_lock(&mountlist_mtx); 1511 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 1512 mtx_unlock(&mountlist_mtx); 1513 1514 /* sanity check system clock against root fs timestamp */ 1515 inittodr(mp->mnt_time); 1516 vfs_unbusy(mp, curthread); 1517 error = VFS_START(mp, 0, curthread); 1518 } 1519 return(error); 1520} 1521 1522/* 1523 * Spin prompting on the console for a suitable root filesystem 1524 */ 1525static int 1526vfs_mountroot_ask(void) 1527{ 1528 char name[128]; 1529 int i; 1530 dev_t dev; 1531 1532 for(;;) { 1533 printf("\nManual root filesystem specification:\n"); 1534 printf(" <fstype>:<device> Mount <device> using filesystem <fstype>\n"); 1535#if defined(__i386__) || defined(__ia64__) 1536 printf(" eg. ufs:da0s1a\n"); 1537#else 1538 printf(" eg. ufs:da0a\n"); 1539#endif 1540 printf(" ? List valid disk boot devices\n"); 1541 printf(" <empty line> Abort manual input\n"); 1542 printf("\nmountroot> "); 1543 gets(name); 1544 if (name[0] == 0) 1545 return(1); 1546 if (name[0] == '?') { 1547 printf("Possibly valid devices for 'ufs' root:\n"); 1548 for (i = 0; i < NUMCDEVSW; i++) { 1549 dev = makedev(i, 0); 1550 if (devsw(dev) != NULL) 1551 printf(" \"%s\"", devsw(dev)->d_name); 1552 } 1553 printf("\n"); 1554 continue; 1555 } 1556 if (!vfs_mountroot_try(name)) 1557 return(0); 1558 } 1559} 1560 1561/* 1562 * Local helper function for vfs_mountroot_ask. 1563 */ 1564static void 1565gets(char *cp) 1566{ 1567 char *lp; 1568 int c; 1569 1570 lp = cp; 1571 for (;;) { 1572 printf("%c", c = cngetc() & 0177); 1573 switch (c) { 1574 case -1: 1575 case '\n': 1576 case '\r': 1577 *lp++ = '\0'; 1578 return; 1579 case '\b': 1580 case '\177': 1581 if (lp > cp) { 1582 printf(" \b"); 1583 lp--; 1584 } 1585 continue; 1586 case '#': 1587 lp--; 1588 if (lp < cp) 1589 lp = cp; 1590 continue; 1591 case '@': 1592 case 'u' & 037: 1593 lp = cp; 1594 printf("%c", '\n'); 1595 continue; 1596 default: 1597 *lp++ = c; 1598 } 1599 } 1600} 1601 1602/* 1603 * Convert a given name to the dev_t of the disk-like device 1604 * it refers to. 1605 */ 1606dev_t 1607getdiskbyname(char *name) { 1608 char *cp; 1609 dev_t dev; 1610 1611 cp = name; 1612 if (!bcmp(cp, "/dev/", 5)) 1613 cp += 5; 1614 1615 dev = NODEV; 1616 EVENTHANDLER_INVOKE(dev_clone, cp, strlen(cp), &dev); 1617 return (dev); 1618} 1619 1620/* 1621 * Set rootdev to match (name), given that we expect it to 1622 * refer to a disk-like device. 1623 */ 1624static int 1625setrootbyname(char *name) 1626{ 1627 dev_t diskdev; 1628 1629 diskdev = getdiskbyname(name); 1630 if (diskdev != NODEV) { 1631 rootdev = diskdev; 1632 return (0); 1633 } 1634 1635 return (1); 1636} 1637 1638/* Show the dev_t for a disk specified by name */ 1639#ifdef DDB 1640DB_SHOW_COMMAND(disk, db_getdiskbyname) 1641{ 1642 dev_t dev; 1643 1644 if (modif[0] == '\0') { 1645 db_error("usage: show disk/devicename"); 1646 return; 1647 } 1648 dev = getdiskbyname(modif); 1649 if (dev != NODEV) 1650 db_printf("dev_t = %p\n", dev); 1651 else 1652 db_printf("No disk device matched.\n"); 1653} 1654#endif 1655 1656/* 1657 * Get a mount option by its name. 1658 * 1659 * Return 0 if the option was found, ENOENT otherwise. 1660 * If len is non-NULL it will be filled with the length 1661 * of the option. If buf is non-NULL, it will be filled 1662 * with the address of the option. 1663 */ 1664int 1665vfs_getopt(opts, name, buf, len) 1666 struct vfsoptlist *opts; 1667 const char *name; 1668 void **buf; 1669 int *len; 1670{ 1671 struct vfsopt *opt; 1672 1673 TAILQ_FOREACH(opt, opts, link) { 1674 if (strcmp(name, opt->name) == 0) { 1675 if (len != NULL) 1676 *len = opt->len; 1677 if (buf != NULL) 1678 *buf = opt->value; 1679 return (0); 1680 } 1681 } 1682 return (ENOENT); 1683} 1684 1685/* 1686 * Find and copy a mount option. 1687 * 1688 * The size of the buffer has to be specified 1689 * in len, if it is not the same length as the 1690 * mount option, EINVAL is returned. 1691 * Returns ENOENT if the option is not found. 1692 */ 1693int 1694vfs_copyopt(opts, name, dest, len) 1695 struct vfsoptlist *opts; 1696 const char *name; 1697 void *dest; 1698 int len; 1699{ 1700 struct vfsopt *opt; 1701 1702 TAILQ_FOREACH(opt, opts, link) { 1703 if (strcmp(name, opt->name) == 0) { 1704 if (len != opt->len) 1705 return (EINVAL); 1706 bcopy(opt->value, dest, opt->len); 1707 return (0); 1708 } 1709 } 1710 return (ENOENT); 1711} 1712