vfs_mount.c revision 138412
1/* 2 * Copyright (c) 1999-2004 Poul-Henning Kamp 3 * Copyright (c) 1989, 1993 4 * The Regents of the University of California. All rights reserved. 5 * (c) UNIX System Laboratories, Inc. 6 * All or some portions of this file are derived from material licensed 7 * to the University of California by American Telephone and Telegraph 8 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 9 * the permission of UNIX System Laboratories, Inc. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * Copyright (c) 1999 Michael Smith 36 * All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 47 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 48 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 49 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 50 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 51 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 52 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 53 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 54 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 55 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 56 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 57 * SUCH DAMAGE. 58 */ 59 60#include <sys/cdefs.h> 61__FBSDID("$FreeBSD: head/sys/kern/vfs_mount.c 138412 2004-12-05 22:41:02Z phk $"); 62 63#include <sys/param.h> 64#include <sys/conf.h> 65#include <sys/cons.h> 66#include <sys/jail.h> 67#include <sys/kernel.h> 68#include <sys/mac.h> 69#include <sys/malloc.h> 70#include <sys/mount.h> 71#include <sys/mutex.h> 72#include <sys/namei.h> 73#include <sys/proc.h> 74#include <sys/filedesc.h> 75#include <sys/reboot.h> 76#include <sys/sysproto.h> 77#include <sys/sx.h> 78#include <sys/sysctl.h> 79#include <sys/sysent.h> 80#include <sys/systm.h> 81#include <sys/vnode.h> 82 83#include <geom/geom.h> 84 85#include <machine/stdarg.h> 86 87#include "opt_rootdevname.h" 88#include "opt_ddb.h" 89#include "opt_mac.h" 90 91#ifdef DDB 92#include <ddb/ddb.h> 93#endif 94 95#define ROOTNAME "root_device" 96#define VFS_MOUNTARG_SIZE_MAX (1024 * 64) 97 98static void checkdirs(struct vnode *olddp, struct vnode *newdp); 99static struct cdev *getdiskbyname(char *_name); 100static void gets(char *cp); 101static int vfs_domount(struct thread *td, const char *fstype, 102 char *fspath, int fsflags, void *fsdata, int compat); 103static int vfs_mount_alloc(struct vnode *dvp, struct vfsconf *vfsp, 104 const char *fspath, struct thread *td, struct mount **mpp); 105static int vfs_mountroot_ask(void); 106static int vfs_mountroot_try(const char *mountfrom); 107static int vfs_donmount(struct thread *td, int fsflags, 108 struct uio *fsoptions); 109 110static int usermount = 0; 111SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, 112 "Unprivileged users may mount and unmount file systems"); 113 114MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 115 116/* List of mounted filesystems. */ 117struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist); 118 119/* For any iteration/modification of mountlist */ 120struct mtx mountlist_mtx; 121 122/* 123 * The vnode of the system's root (/ in the filesystem, without chroot 124 * active.) 125 */ 126struct vnode *rootvnode; 127 128/* 129 * The root filesystem is detailed in the kernel environment variable 130 * vfs.root.mountfrom, which is expected to be in the general format 131 * 132 * <vfsname>:[<path>] 133 * vfsname := the name of a VFS known to the kernel and capable 134 * of being mounted as root 135 * path := disk device name or other data used by the filesystem 136 * to locate its physical store 137 */ 138 139/* 140 * The root specifiers we will try if RB_CDROM is specified. 141 */ 142static char *cdrom_rootdevnames[] = { 143 "cd9660:cd0", 144 "cd9660:acd0", 145 NULL 146}; 147 148/* legacy find-root code */ 149char *rootdevnames[2] = {NULL, NULL}; 150struct cdev *rootdev = NULL; 151#ifdef ROOTDEVNAME 152const char *ctrootdevname = ROOTDEVNAME; 153#else 154const char *ctrootdevname = NULL; 155#endif 156 157/* 158 * Has to be dynamic as the value of rootdev can change; however, it can't 159 * change after the root is mounted, so a user process can't access this 160 * sysctl until after the value is unchangeable. 161 */ 162static int 163sysctl_rootdev(SYSCTL_HANDLER_ARGS) 164{ 165 int error; 166 167 /* _RD prevents this from happening. */ 168 KASSERT(req->newptr == NULL, ("Attempt to change root device name")); 169 170 if (rootdev != NULL) 171 error = sysctl_handle_string(oidp, rootdev->si_name, 0, req); 172 else 173 error = sysctl_handle_string(oidp, "", 0, req); 174 175 return (error); 176} 177 178SYSCTL_PROC(_kern, OID_AUTO, rootdev, CTLTYPE_STRING | CTLFLAG_RD, 179 0, 0, sysctl_rootdev, "A", "Root file system device"); 180 181/* Remove one mount option. */ 182static void 183vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt) 184{ 185 186 TAILQ_REMOVE(opts, opt, link); 187 free(opt->name, M_MOUNT); 188 if (opt->value != NULL) 189 free(opt->value, M_MOUNT); 190#ifdef INVARIANTS 191 else if (opt->len != 0) 192 panic("%s: mount option with NULL value but length != 0", 193 __func__); 194#endif 195 free(opt, M_MOUNT); 196} 197 198/* Release all resources related to the mount options. */ 199static void 200vfs_freeopts(struct vfsoptlist *opts) 201{ 202 struct vfsopt *opt; 203 204 while (!TAILQ_EMPTY(opts)) { 205 opt = TAILQ_FIRST(opts); 206 vfs_freeopt(opts, opt); 207 } 208 free(opts, M_MOUNT); 209} 210 211/* 212 * Check if options are equal (with or without the "no" prefix). 213 */ 214static int 215vfs_equalopts(const char *opt1, const char *opt2) 216{ 217 218 /* "opt" vs. "opt" or "noopt" vs. "noopt" */ 219 if (strcmp(opt1, opt2) == 0) 220 return (1); 221 /* "noopt" vs. "opt" */ 222 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0) 223 return (1); 224 /* "opt" vs. "noopt" */ 225 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0) 226 return (1); 227 return (0); 228} 229 230/* 231 * If a mount option is specified several times, 232 * (with or without the "no" prefix) only keep 233 * the last occurence of it. 234 */ 235static void 236vfs_sanitizeopts(struct vfsoptlist *opts) 237{ 238 struct vfsopt *opt, *opt2, *tmp; 239 240 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) { 241 opt2 = TAILQ_PREV(opt, vfsoptlist, link); 242 while (opt2 != NULL) { 243 if (vfs_equalopts(opt->name, opt2->name)) { 244 tmp = TAILQ_PREV(opt2, vfsoptlist, link); 245 vfs_freeopt(opts, opt2); 246 opt2 = tmp; 247 } else { 248 opt2 = TAILQ_PREV(opt2, vfsoptlist, link); 249 } 250 } 251 } 252} 253 254/* 255 * Build a linked list of mount options from a struct uio. 256 */ 257static int 258vfs_buildopts(struct uio *auio, struct vfsoptlist **options) 259{ 260 struct vfsoptlist *opts; 261 struct vfsopt *opt; 262 size_t memused; 263 unsigned int i, iovcnt; 264 int error, namelen, optlen; 265 266 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 267 TAILQ_INIT(opts); 268 memused = 0; 269 iovcnt = auio->uio_iovcnt; 270 for (i = 0; i < iovcnt; i += 2) { 271 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 272 namelen = auio->uio_iov[i].iov_len; 273 optlen = auio->uio_iov[i + 1].iov_len; 274 opt->name = malloc(namelen, M_MOUNT, M_WAITOK); 275 opt->value = NULL; 276 opt->len = 0; 277 278 /* 279 * Do this early, so jumps to "bad" will free the current 280 * option. 281 */ 282 TAILQ_INSERT_TAIL(opts, opt, link); 283 memused += sizeof(struct vfsopt) + optlen + namelen; 284 285 /* 286 * Avoid consuming too much memory, and attempts to overflow 287 * memused. 288 */ 289 if (memused > VFS_MOUNTARG_SIZE_MAX || 290 optlen > VFS_MOUNTARG_SIZE_MAX || 291 namelen > VFS_MOUNTARG_SIZE_MAX) { 292 error = EINVAL; 293 goto bad; 294 } 295 296 if (auio->uio_segflg == UIO_SYSSPACE) { 297 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen); 298 } else { 299 error = copyin(auio->uio_iov[i].iov_base, opt->name, 300 namelen); 301 if (error) 302 goto bad; 303 } 304 /* Ensure names are null-terminated strings. */ 305 if (opt->name[namelen - 1] != '\0') { 306 error = EINVAL; 307 goto bad; 308 } 309 if (optlen != 0) { 310 opt->len = optlen; 311 opt->value = malloc(optlen, M_MOUNT, M_WAITOK); 312 if (auio->uio_segflg == UIO_SYSSPACE) { 313 bcopy(auio->uio_iov[i + 1].iov_base, opt->value, 314 optlen); 315 } else { 316 error = copyin(auio->uio_iov[i + 1].iov_base, 317 opt->value, optlen); 318 if (error) 319 goto bad; 320 } 321 } 322 } 323 vfs_sanitizeopts(opts); 324 *options = opts; 325 return (0); 326bad: 327 vfs_freeopts(opts); 328 return (error); 329} 330 331/* 332 * Merge the old mount options with the new ones passed 333 * in the MNT_UPDATE case. 334 */ 335static void 336vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts) 337{ 338 struct vfsopt *opt, *opt2, *new; 339 340 TAILQ_FOREACH(opt, opts, link) { 341 /* 342 * Check that this option hasn't been redefined 343 * nor cancelled with a "no" mount option. 344 */ 345 opt2 = TAILQ_FIRST(toopts); 346 while (opt2 != NULL) { 347 if (strcmp(opt2->name, opt->name) == 0) 348 goto next; 349 if (strncmp(opt2->name, "no", 2) == 0 && 350 strcmp(opt2->name + 2, opt->name) == 0) { 351 vfs_freeopt(toopts, opt2); 352 goto next; 353 } 354 opt2 = TAILQ_NEXT(opt2, link); 355 } 356 /* We want this option, duplicate it. */ 357 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 358 new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK); 359 strcpy(new->name, opt->name); 360 if (opt->len != 0) { 361 new->value = malloc(opt->len, M_MOUNT, M_WAITOK); 362 bcopy(opt->value, new->value, opt->len); 363 } else { 364 new->value = NULL; 365 } 366 new->len = opt->len; 367 TAILQ_INSERT_TAIL(toopts, new, link); 368next: 369 continue; 370 } 371} 372 373/* 374 * New mount API. 375 */ 376int 377nmount(td, uap) 378 struct thread *td; 379 struct nmount_args /* { 380 struct iovec *iovp; 381 unsigned int iovcnt; 382 int flags; 383 } */ *uap; 384{ 385 struct uio *auio; 386 struct iovec *iov; 387 unsigned int i; 388 int error; 389 u_int iovcnt; 390 391 /* Kick out MNT_ROOTFS early as it is legal internally */ 392 if (uap->flags & MNT_ROOTFS) 393 return (EINVAL); 394 395 iovcnt = uap->iovcnt; 396 /* 397 * Check that we have an even number of iovec's 398 * and that we have at least two options. 399 */ 400 if ((iovcnt & 1) || (iovcnt < 4)) 401 return (EINVAL); 402 403 error = copyinuio(uap->iovp, iovcnt, &auio); 404 if (error) 405 return (error); 406 iov = auio->uio_iov; 407 for (i = 0; i < iovcnt; i++) { 408 if (iov->iov_len > MMAXOPTIONLEN) { 409 free(auio, M_IOV); 410 return (EINVAL); 411 } 412 iov++; 413 } 414 error = vfs_donmount(td, uap->flags, auio); 415 free(auio, M_IOV); 416 return (error); 417} 418 419struct mntarg { 420 struct iovec *v; 421 int len; 422}; 423 424struct mntarg * 425mount_arg(struct mntarg *ma, const char *name, const void *val, int len) 426{ 427 428 if (ma == NULL) 429 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 430 431 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 432 M_MOUNT, M_WAITOK); 433 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 434 ma->v[ma->len].iov_len = strlen(name) + 1; 435 ma->len++; 436 437 ma->v[ma->len].iov_base = (void *)(uintptr_t)val; 438 if (len < 0) 439 ma->v[ma->len].iov_len = strlen(val) + 1; 440 else 441 ma->v[ma->len].iov_len = len; 442 ma->len++; 443 return (ma); 444} 445 446int 447kernel_mount(struct mntarg *ma, int flags) 448{ 449 struct uio auio; 450 int error; 451 452 KASSERT(ma != NULL, ("kernel_mount NULL ma")); 453 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v")); 454 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len)); 455 456 auio.uio_iov = ma->v; 457 auio.uio_iovcnt = ma->len; 458 auio.uio_segflg = UIO_SYSSPACE; 459 460 error = vfs_donmount(curthread, flags, &auio); 461 free(ma->v, M_MOUNT); 462 free(ma, M_MOUNT); 463 return (error); 464} 465 466int 467kernel_vmount(int flags, ...) 468{ 469 struct mntarg *ma = NULL; 470 va_list ap; 471 const char *cp; 472 const void *vp; 473 int error; 474 475 va_start(ap, flags); 476 for (;;) { 477 cp = va_arg(ap, const char *); 478 if (cp == NULL) 479 break; 480 vp = va_arg(ap, const void *); 481 ma = mount_arg(ma, cp, vp, -1); 482 } 483 va_end(ap); 484 485 error = kernel_mount(ma, flags); 486 return (error); 487} 488 489/* 490 * Allocate and initialize the mount point struct. 491 */ 492static int 493vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, 494 const char *fspath, struct thread *td, struct mount **mpp) 495{ 496 struct mount *mp; 497 498 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO); 499 TAILQ_INIT(&mp->mnt_nvnodelist); 500 mp->mnt_nvnodelistsize = 0; 501 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF); 502 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE); 503 vfs_busy(mp, LK_NOWAIT, 0, td); 504 mp->mnt_op = vfsp->vfc_vfsops; 505 mp->mnt_vfc = vfsp; 506 vfsp->vfc_refcount++; 507 mp->mnt_stat.f_type = vfsp->vfc_typenum; 508 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 509 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 510 mp->mnt_vnodecovered = vp; 511 mp->mnt_cred = crdup(td->td_ucred); 512 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 513 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 514 mp->mnt_iosize_max = DFLTPHYS; 515#ifdef MAC 516 mac_init_mount(mp); 517 mac_create_mount(td->td_ucred, mp); 518#endif 519 *mpp = mp; 520 return (0); 521} 522 523/* 524 * Destroy the mount struct previously allocated by vfs_mount_alloc(). 525 */ 526void 527vfs_mount_destroy(struct mount *mp, struct thread *td) 528{ 529 530 mp->mnt_vfc->vfc_refcount--; 531 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) 532 panic("unmount: dangling vnode"); 533 vfs_unbusy(mp,td); 534 lockdestroy(&mp->mnt_lock); 535 mtx_destroy(&mp->mnt_mtx); 536 if (mp->mnt_kern_flag & MNTK_MWAIT) 537 wakeup(mp); 538#ifdef MAC 539 mac_destroy_mount(mp); 540#endif 541 if (mp->mnt_opt != NULL) 542 vfs_freeopts(mp->mnt_opt); 543 crfree(mp->mnt_cred); 544 free(mp, M_MOUNT); 545} 546 547static int 548vfs_donmount(struct thread *td, int fsflags, struct uio *fsoptions) 549{ 550 struct vfsoptlist *optlist; 551 char *fstype, *fspath; 552 int error, fstypelen, fspathlen; 553 554 error = vfs_buildopts(fsoptions, &optlist); 555 if (error) 556 return (error); 557 558 /* 559 * We need these two options before the others, 560 * and they are mandatory for any filesystem. 561 * Ensure they are NUL terminated as well. 562 */ 563 fstypelen = 0; 564 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen); 565 if (error || fstype[fstypelen - 1] != '\0') { 566 error = EINVAL; 567 goto bail; 568 } 569 fspathlen = 0; 570 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen); 571 if (error || fspath[fspathlen - 1] != '\0') { 572 error = EINVAL; 573 goto bail; 574 } 575 576 /* 577 * Be ultra-paranoid about making sure the type and fspath 578 * variables will fit in our mp buffers, including the 579 * terminating NUL. 580 */ 581 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) { 582 error = ENAMETOOLONG; 583 goto bail; 584 } 585 586 mtx_lock(&Giant); 587 error = vfs_domount(td, fstype, fspath, fsflags, optlist, 0); 588 mtx_unlock(&Giant); 589bail: 590 if (error) 591 vfs_freeopts(optlist); 592 return (error); 593} 594 595/* 596 * Old mount API. 597 */ 598#ifndef _SYS_SYSPROTO_H_ 599struct mount_args { 600 char *type; 601 char *path; 602 int flags; 603 caddr_t data; 604}; 605#endif 606/* ARGSUSED */ 607int 608mount(td, uap) 609 struct thread *td; 610 struct mount_args /* { 611 char *type; 612 char *path; 613 int flags; 614 caddr_t data; 615 } */ *uap; 616{ 617 char *fstype; 618 char *fspath; 619 struct vfsconf *vfsp; 620 int error; 621 622 /* Kick out MNT_ROOTFS early as it is legal internally */ 623 if (uap->flags & MNT_ROOTFS) 624 return (EINVAL); 625 626 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK); 627 fspath = malloc(MNAMELEN, M_TEMP, M_WAITOK); 628 629 /* 630 * vfs_mount() actually takes a kernel string for `type' and 631 * `path' now, so extract them. 632 */ 633 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL); 634 if (error == 0) 635 error = copyinstr(uap->path, fspath, MNAMELEN, NULL); 636 if (error == 0) { 637 mtx_lock(&Giant); 638 vfsp = vfs_byname_kld(fstype, td, &error); 639 if (vfsp != NULL) { 640 if (vfsp->vfc_vfsops->vfs_cmount != NULL) { 641 error = vfsp->vfc_vfsops->vfs_cmount( 642 fspath, uap->data, uap->flags, td); 643 } else { 644 error = vfs_domount(td, fstype, fspath, 645 uap->flags, uap->data, 1); 646 } 647 } 648 mtx_unlock(&Giant); 649 } 650 free(fstype, M_TEMP); 651 free(fspath, M_TEMP); 652 return (error); 653} 654 655/* 656 * vfs_domount(): actually attempt a filesystem mount. 657 */ 658static int 659vfs_domount( 660 struct thread *td, /* Flags common to all filesystems. */ 661 const char *fstype, /* Filesystem type. */ 662 char *fspath, /* Mount path. */ 663 int fsflags, /* Flags common to all filesystems. */ 664 void *fsdata, /* Options local to the filesystem. */ 665 int compat /* Invocation from compat syscall. */ 666 ) 667{ 668 struct vnode *vp; 669 struct mount *mp; 670 struct vfsconf *vfsp; 671 int error, flag = 0, kern_flag = 0; 672 struct vattr va; 673 struct nameidata nd; 674 675 mtx_assert(&Giant, MA_OWNED); 676 677 /* 678 * Be ultra-paranoid about making sure the type and fspath 679 * variables will fit in our mp buffers, including the 680 * terminating NUL. 681 */ 682 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 683 return (ENAMETOOLONG); 684 685 if (jailed(td->td_ucred)) 686 return (EPERM); 687 if (usermount == 0) { 688 if ((error = suser(td)) != 0) 689 return (error); 690 } 691 692 /* 693 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users. 694 */ 695 if (fsflags & (MNT_EXPORTED | MNT_SUIDDIR)) { 696 if ((error = suser(td)) != 0) 697 return (error); 698 } 699 /* 700 * Silently enforce MNT_NOSUID and MNT_USER for 701 * unprivileged users. 702 */ 703 if (suser(td) != 0) 704 fsflags |= MNT_NOSUID | MNT_USER; 705 /* 706 * Get vnode to be covered 707 */ 708 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 709 if ((error = namei(&nd)) != 0) 710 return (error); 711 NDFREE(&nd, NDF_ONLY_PNBUF); 712 vp = nd.ni_vp; 713 if (fsflags & MNT_UPDATE) { 714 if ((vp->v_vflag & VV_ROOT) == 0) { 715 vput(vp); 716 return (EINVAL); 717 } 718 mp = vp->v_mount; 719 flag = mp->mnt_flag; 720 kern_flag = mp->mnt_kern_flag; 721 /* 722 * We only allow the filesystem to be reloaded if it 723 * is currently mounted read-only. 724 */ 725 if ((fsflags & MNT_RELOAD) && 726 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 727 vput(vp); 728 return (EOPNOTSUPP); /* Needs translation */ 729 } 730 /* 731 * Only privileged root, or (if MNT_USER is set) the user that 732 * did the original mount is permitted to update it. 733 */ 734 error = vfs_suser(mp, td); 735 if (error) { 736 vput(vp); 737 return (error); 738 } 739 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 740 vput(vp); 741 return (EBUSY); 742 } 743 VI_LOCK(vp); 744 if ((vp->v_iflag & VI_MOUNT) != 0 || 745 vp->v_mountedhere != NULL) { 746 VI_UNLOCK(vp); 747 vfs_unbusy(mp, td); 748 vput(vp); 749 return (EBUSY); 750 } 751 vp->v_iflag |= VI_MOUNT; 752 VI_UNLOCK(vp); 753 mp->mnt_flag |= fsflags & 754 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT); 755 VOP_UNLOCK(vp, 0, td); 756 if (compat == 0) { 757 mp->mnt_optnew = fsdata; 758 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 759 } 760 } else { 761 /* 762 * If the user is not root, ensure that they own the directory 763 * onto which we are attempting to mount. 764 */ 765 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 766 if (error) { 767 vput(vp); 768 return (error); 769 } 770 if (va.va_uid != td->td_ucred->cr_uid) { 771 if ((error = suser(td)) != 0) { 772 vput(vp); 773 return (error); 774 } 775 } 776 if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) { 777 vput(vp); 778 return (error); 779 } 780 if (vp->v_type != VDIR) { 781 vput(vp); 782 return (ENOTDIR); 783 } 784 vfsp = vfs_byname_kld(fstype, td, &error); 785 if (vfsp == NULL) { 786 vput(vp); 787 return (error); 788 } 789 VI_LOCK(vp); 790 if ((vp->v_iflag & VI_MOUNT) != 0 || 791 vp->v_mountedhere != NULL) { 792 VI_UNLOCK(vp); 793 vput(vp); 794 return (EBUSY); 795 } 796 vp->v_iflag |= VI_MOUNT; 797 VI_UNLOCK(vp); 798 799 /* 800 * Allocate and initialize the filesystem. 801 */ 802 error = vfs_mount_alloc(vp, vfsp, fspath, td, &mp); 803 if (error) { 804 vput(vp); 805 return (error); 806 } 807 VOP_UNLOCK(vp, 0, td); 808 809 /* XXXMAC: pass to vfs_mount_alloc? */ 810 if (compat == 0) 811 mp->mnt_optnew = fsdata; 812 } 813 /* 814 * Check if the fs implements the type VFS_[O]MOUNT() 815 * function we are looking for. 816 */ 817 if ((compat && (mp->mnt_op->vfs_omount == NULL)) || 818 (!compat && (mp->mnt_op->vfs_mount == NULL))) { 819 printf("%s doesn't support the %s mount syscall\n", 820 mp->mnt_vfc->vfc_name, compat ? "old" : "new"); 821 VI_LOCK(vp); 822 vp->v_iflag &= ~VI_MOUNT; 823 VI_UNLOCK(vp); 824 if (mp->mnt_flag & MNT_UPDATE) 825 vfs_unbusy(mp, td); 826 else 827 vfs_mount_destroy(mp, td); 828 vrele(vp); 829 return (EOPNOTSUPP); 830 } 831 832 /* 833 * Set the mount level flags. 834 */ 835 if (fsflags & MNT_RDONLY) 836 mp->mnt_flag |= MNT_RDONLY; 837 else if (mp->mnt_flag & MNT_RDONLY) 838 mp->mnt_kern_flag |= MNTK_WANTRDWR; 839 mp->mnt_flag &=~ MNT_UPDATEMASK; 840 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE); 841 /* 842 * Mount the filesystem. 843 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 844 * get. No freeing of cn_pnbuf. 845 */ 846 if (compat) 847 error = VFS_OMOUNT(mp, fspath, fsdata, td); 848 else 849 error = VFS_MOUNT(mp, td); 850 if (!error) { 851 if (mp->mnt_opt != NULL) 852 vfs_freeopts(mp->mnt_opt); 853 mp->mnt_opt = mp->mnt_optnew; 854 } 855 /* 856 * Prevent external consumers of mount options from reading 857 * mnt_optnew. 858 */ 859 mp->mnt_optnew = NULL; 860 if (mp->mnt_flag & MNT_UPDATE) { 861 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 862 mp->mnt_flag &= ~MNT_RDONLY; 863 mp->mnt_flag &= 864 ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 865 mp->mnt_kern_flag &= ~MNTK_WANTRDWR; 866 if (error) { 867 mp->mnt_flag = flag; 868 mp->mnt_kern_flag = kern_flag; 869 } 870 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 871 if (mp->mnt_syncer == NULL) 872 error = vfs_allocate_syncvnode(mp); 873 } else { 874 if (mp->mnt_syncer != NULL) 875 vrele(mp->mnt_syncer); 876 mp->mnt_syncer = NULL; 877 } 878 vfs_unbusy(mp, td); 879 VI_LOCK(vp); 880 vp->v_iflag &= ~VI_MOUNT; 881 VI_UNLOCK(vp); 882 vrele(vp); 883 return (error); 884 } 885 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 886 /* 887 * Put the new filesystem on the mount list after root. 888 */ 889 cache_purge(vp); 890 if (!error) { 891 struct vnode *newdp; 892 893 VI_LOCK(vp); 894 vp->v_iflag &= ~VI_MOUNT; 895 VI_UNLOCK(vp); 896 vp->v_mountedhere = mp; 897 mtx_lock(&mountlist_mtx); 898 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 899 mtx_unlock(&mountlist_mtx); 900 vfs_event_signal(NULL, VQ_MOUNT, 0); 901 if (VFS_ROOT(mp, &newdp, td)) 902 panic("mount: lost mount"); 903 checkdirs(vp, newdp); 904 vput(newdp); 905 VOP_UNLOCK(vp, 0, td); 906 if ((mp->mnt_flag & MNT_RDONLY) == 0) 907 error = vfs_allocate_syncvnode(mp); 908 vfs_unbusy(mp, td); 909 if (error || (error = VFS_START(mp, 0, td)) != 0) 910 vrele(vp); 911 } else { 912 VI_LOCK(vp); 913 vp->v_iflag &= ~VI_MOUNT; 914 VI_UNLOCK(vp); 915 vfs_mount_destroy(mp, td); 916 vput(vp); 917 } 918 return (error); 919} 920 921/* 922 * Scan all active processes to see if any of them have a current 923 * or root directory of `olddp'. If so, replace them with the new 924 * mount point. 925 */ 926static void 927checkdirs(olddp, newdp) 928 struct vnode *olddp, *newdp; 929{ 930 struct filedesc *fdp; 931 struct proc *p; 932 int nrele; 933 934 if (vrefcnt(olddp) == 1) 935 return; 936 sx_slock(&allproc_lock); 937 LIST_FOREACH(p, &allproc, p_list) { 938 mtx_lock(&fdesc_mtx); 939 fdp = p->p_fd; 940 if (fdp == NULL) { 941 mtx_unlock(&fdesc_mtx); 942 continue; 943 } 944 nrele = 0; 945 FILEDESC_LOCK_FAST(fdp); 946 if (fdp->fd_cdir == olddp) { 947 vref(newdp); 948 fdp->fd_cdir = newdp; 949 nrele++; 950 } 951 if (fdp->fd_rdir == olddp) { 952 vref(newdp); 953 fdp->fd_rdir = newdp; 954 nrele++; 955 } 956 FILEDESC_UNLOCK_FAST(fdp); 957 mtx_unlock(&fdesc_mtx); 958 while (nrele--) 959 vrele(olddp); 960 } 961 sx_sunlock(&allproc_lock); 962 if (rootvnode == olddp) { 963 vrele(rootvnode); 964 vref(newdp); 965 rootvnode = newdp; 966 } 967} 968 969/* 970 * Unmount a filesystem. 971 * 972 * Note: unmount takes a path to the vnode mounted on as argument, 973 * not special file (as before). 974 */ 975#ifndef _SYS_SYSPROTO_H_ 976struct unmount_args { 977 char *path; 978 int flags; 979}; 980#endif 981/* ARGSUSED */ 982int 983unmount(td, uap) 984 struct thread *td; 985 register struct unmount_args /* { 986 char *path; 987 int flags; 988 } */ *uap; 989{ 990 struct mount *mp; 991 char *pathbuf; 992 int error, id0, id1; 993 994 if (jailed(td->td_ucred)) 995 return (EPERM); 996 if (usermount == 0) { 997 if ((error = suser(td)) != 0) 998 return (error); 999 } 1000 1001 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1002 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL); 1003 if (error) { 1004 free(pathbuf, M_TEMP); 1005 return (error); 1006 } 1007 if (uap->flags & MNT_BYFSID) { 1008 /* Decode the filesystem ID. */ 1009 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) { 1010 free(pathbuf, M_TEMP); 1011 return (EINVAL); 1012 } 1013 1014 mtx_lock(&mountlist_mtx); 1015 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1016 if (mp->mnt_stat.f_fsid.val[0] == id0 && 1017 mp->mnt_stat.f_fsid.val[1] == id1) 1018 break; 1019 } 1020 mtx_unlock(&mountlist_mtx); 1021 } else { 1022 mtx_lock(&mountlist_mtx); 1023 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1024 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) 1025 break; 1026 } 1027 mtx_unlock(&mountlist_mtx); 1028 } 1029 free(pathbuf, M_TEMP); 1030 if (mp == NULL) { 1031 /* 1032 * Previously we returned ENOENT for a nonexistent path and 1033 * EINVAL for a non-mountpoint. We cannot tell these apart 1034 * now, so in the !MNT_BYFSID case return the more likely 1035 * EINVAL for compatibility. 1036 */ 1037 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL); 1038 } 1039 1040 /* 1041 * Only privileged root, or (if MNT_USER is set) the user that did the 1042 * original mount is permitted to unmount this filesystem. 1043 */ 1044 error = vfs_suser(mp, td); 1045 if (error) 1046 return (error); 1047 1048 /* 1049 * Don't allow unmounting the root filesystem. 1050 */ 1051 if (mp->mnt_flag & MNT_ROOTFS) 1052 return (EINVAL); 1053 mtx_lock(&Giant); 1054 error = dounmount(mp, uap->flags, td); 1055 mtx_unlock(&Giant); 1056 return (error); 1057} 1058 1059/* 1060 * Do the actual filesystem unmount. 1061 */ 1062int 1063dounmount(mp, flags, td) 1064 struct mount *mp; 1065 int flags; 1066 struct thread *td; 1067{ 1068 struct vnode *coveredvp, *fsrootvp; 1069 int error; 1070 int async_flag; 1071 1072 mtx_assert(&Giant, MA_OWNED); 1073 1074 mtx_lock(&mountlist_mtx); 1075 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 1076 mtx_unlock(&mountlist_mtx); 1077 return (EBUSY); 1078 } 1079 mp->mnt_kern_flag |= MNTK_UNMOUNT; 1080 /* Allow filesystems to detect that a forced unmount is in progress. */ 1081 if (flags & MNT_FORCE) 1082 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1083 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 1084 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_mtx, td); 1085 if (error) { 1086 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1087 if (mp->mnt_kern_flag & MNTK_MWAIT) 1088 wakeup(mp); 1089 return (error); 1090 } 1091 vn_start_write(NULL, &mp, V_WAIT); 1092 1093 if (mp->mnt_flag & MNT_EXPUBLIC) 1094 vfs_setpublicfs(NULL, NULL, NULL); 1095 1096 vfs_msync(mp, MNT_WAIT); 1097 async_flag = mp->mnt_flag & MNT_ASYNC; 1098 mp->mnt_flag &= ~MNT_ASYNC; 1099 cache_purgevfs(mp); /* remove cache entries for this file sys */ 1100 if (mp->mnt_syncer != NULL) 1101 vrele(mp->mnt_syncer); 1102 /* 1103 * For forced unmounts, move process cdir/rdir refs on the fs root 1104 * vnode to the covered vnode. For non-forced unmounts we want 1105 * such references to cause an EBUSY error. 1106 */ 1107 if ((flags & MNT_FORCE) && VFS_ROOT(mp, &fsrootvp, td) == 0) { 1108 if (mp->mnt_vnodecovered != NULL) 1109 checkdirs(fsrootvp, mp->mnt_vnodecovered); 1110 if (fsrootvp == rootvnode) { 1111 vrele(rootvnode); 1112 rootvnode = NULL; 1113 } 1114 vput(fsrootvp); 1115 } 1116 if (((mp->mnt_flag & MNT_RDONLY) || 1117 (error = VFS_SYNC(mp, MNT_WAIT, td->td_ucred, td)) == 0) || 1118 (flags & MNT_FORCE)) { 1119 error = VFS_UNMOUNT(mp, flags, td); 1120 } 1121 vn_finished_write(mp); 1122 if (error) { 1123 /* Undo cdir/rdir and rootvnode changes made above. */ 1124 if ((flags & MNT_FORCE) && VFS_ROOT(mp, &fsrootvp, td) == 0) { 1125 if (mp->mnt_vnodecovered != NULL) 1126 checkdirs(mp->mnt_vnodecovered, fsrootvp); 1127 if (rootvnode == NULL) { 1128 rootvnode = fsrootvp; 1129 vref(rootvnode); 1130 } 1131 vput(fsrootvp); 1132 } 1133 if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL) 1134 (void) vfs_allocate_syncvnode(mp); 1135 mtx_lock(&mountlist_mtx); 1136 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1137 mp->mnt_flag |= async_flag; 1138 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, 1139 &mountlist_mtx, td); 1140 if (mp->mnt_kern_flag & MNTK_MWAIT) 1141 wakeup(mp); 1142 return (error); 1143 } 1144 mtx_lock(&mountlist_mtx); 1145 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1146 if ((coveredvp = mp->mnt_vnodecovered) != NULL) 1147 coveredvp->v_mountedhere = NULL; 1148 mtx_unlock(&mountlist_mtx); 1149 vfs_event_signal(NULL, VQ_UNMOUNT, 0); 1150 vfs_mount_destroy(mp, td); 1151 if (coveredvp != NULL) 1152 vrele(coveredvp); 1153 return (0); 1154} 1155 1156/* 1157 * Find and mount the root filesystem 1158 */ 1159void 1160vfs_mountroot(void) 1161{ 1162 char *cp; 1163 int error, i, asked = 0; 1164 1165 1166 /* 1167 * Wait for GEOM to settle down 1168 */ 1169 DROP_GIANT(); 1170 g_waitidle(); 1171 PICKUP_GIANT(); 1172 1173 /* 1174 * We are booted with instructions to prompt for the root filesystem. 1175 */ 1176 if (boothowto & RB_ASKNAME) { 1177 if (!vfs_mountroot_ask()) 1178 return; 1179 asked = 1; 1180 } 1181 1182 /* 1183 * The root filesystem information is compiled in, and we are 1184 * booted with instructions to use it. 1185 */ 1186 if (ctrootdevname != NULL && (boothowto & RB_DFLTROOT)) { 1187 if (!vfs_mountroot_try(ctrootdevname)) 1188 return; 1189 ctrootdevname = NULL; 1190 } 1191 1192 /* 1193 * We've been given the generic "use CDROM as root" flag. This is 1194 * necessary because one media may be used in many different 1195 * devices, so we need to search for them. 1196 */ 1197 if (boothowto & RB_CDROM) { 1198 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 1199 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 1200 return; 1201 } 1202 } 1203 1204 /* 1205 * Try to use the value read by the loader from /etc/fstab, or 1206 * supplied via some other means. This is the preferred 1207 * mechanism. 1208 */ 1209 cp = getenv("vfs.root.mountfrom"); 1210 if (cp != NULL) { 1211 error = vfs_mountroot_try(cp); 1212 freeenv(cp); 1213 if (!error) 1214 return; 1215 } 1216 1217 /* 1218 * Try values that may have been computed by code during boot 1219 */ 1220 if (!vfs_mountroot_try(rootdevnames[0])) 1221 return; 1222 if (!vfs_mountroot_try(rootdevnames[1])) 1223 return; 1224 1225 /* 1226 * If we (still) have a compiled-in default, try it. 1227 */ 1228 if (ctrootdevname != NULL) 1229 if (!vfs_mountroot_try(ctrootdevname)) 1230 return; 1231 1232 /* 1233 * Everything so far has failed, prompt on the console if we haven't 1234 * already tried that. 1235 */ 1236 if (!asked) 1237 if (!vfs_mountroot_ask()) 1238 return; 1239 panic("Root mount failed, startup aborted."); 1240} 1241 1242/* 1243 * Mount (mountfrom) as the root filesystem. 1244 */ 1245static int 1246vfs_mountroot_try(const char *mountfrom) 1247{ 1248 struct mount *mp; 1249 struct thread *td = curthread; 1250 struct vfsconf *vfsp; 1251 char *vfsname, *path; 1252 int error; 1253 char patt[32]; 1254 int s; 1255 1256 vfsname = NULL; 1257 path = NULL; 1258 mp = NULL; 1259 error = EINVAL; 1260 1261 if (mountfrom == NULL) 1262 return (error); /* don't complain */ 1263 1264 s = splcam(); /* Overkill, but annoying without it */ 1265 printf("Trying to mount root from %s\n", mountfrom); 1266 splx(s); 1267 1268 /* parse vfs name and path */ 1269 vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK); 1270 path = malloc(MNAMELEN, M_MOUNT, M_WAITOK); 1271 vfsname[0] = path[0] = 0; 1272 sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN); 1273 if (sscanf(mountfrom, patt, vfsname, path) < 1) 1274 goto done; 1275 1276 if (path[0] == '\0') 1277 strcpy(path, ROOTNAME); 1278 1279 vfsp = vfs_byname(vfsname); 1280 if (vfsp == NULL) { 1281 printf("Can't find filesystem \"%s\"\n", vfsname); 1282 goto done; 1283 } 1284 error = vfs_mount_alloc(NULLVP, vfsp, "/", td, &mp); 1285 if (error) { 1286 printf("Could not alloc mountpoint\n"); 1287 goto done; 1288 } 1289 1290 mp->mnt_flag |= MNT_RDONLY | MNT_ROOTFS; 1291 1292 strlcpy(mp->mnt_stat.f_mntfromname, path, MNAMELEN); 1293 1294 /* 1295 * do our best to set rootdev 1296 * XXX: This does not belong here! 1297 */ 1298 if (path[0] != '\0') { 1299 struct cdev *diskdev; 1300 diskdev = getdiskbyname(path); 1301 if (diskdev != NULL) 1302 rootdev = diskdev; 1303 else 1304 printf("setrootbyname failed\n"); 1305 } 1306 1307 error = VFS_OMOUNT(mp, path, NULL, curthread); 1308 1309done: 1310 if (vfsname != NULL) 1311 free(vfsname, M_MOUNT); 1312 if (path != NULL) 1313 free(path, M_MOUNT); 1314 if (error != 0) { 1315 if (mp != NULL) 1316 vfs_mount_destroy(mp, curthread); 1317 printf("Root mount failed: %d\n", error); 1318 } else { 1319 1320 /* register with list of mounted filesystems */ 1321 mtx_lock(&mountlist_mtx); 1322 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 1323 mtx_unlock(&mountlist_mtx); 1324 1325 /* sanity check system clock against root fs timestamp */ 1326 inittodr(mp->mnt_time); 1327 vfs_unbusy(mp, curthread); 1328 error = VFS_START(mp, 0, curthread); 1329 } 1330 return (error); 1331} 1332 1333/* 1334 * Spin prompting on the console for a suitable root filesystem 1335 */ 1336static int 1337vfs_mountroot_ask(void) 1338{ 1339 char name[128]; 1340 1341 for(;;) { 1342 printf("\nManual root filesystem specification:\n"); 1343 printf(" <fstype>:<device> Mount <device> using filesystem <fstype>\n"); 1344#if defined(__i386__) || defined(__ia64__) 1345 printf(" eg. ufs:da0s1a\n"); 1346#else 1347 printf(" eg. ufs:/dev/da0a\n"); 1348#endif 1349 printf(" ? List valid disk boot devices\n"); 1350 printf(" <empty line> Abort manual input\n"); 1351 printf("\nmountroot> "); 1352 gets(name); 1353 if (name[0] == '\0') 1354 return (1); 1355 if (name[0] == '?') { 1356 printf("\nList of GEOM managed disk devices:\n "); 1357 g_dev_print(); 1358 continue; 1359 } 1360 if (!vfs_mountroot_try(name)) 1361 return (0); 1362 } 1363} 1364 1365/* 1366 * Local helper function for vfs_mountroot_ask. 1367 */ 1368static void 1369gets(char *cp) 1370{ 1371 char *lp; 1372 int c; 1373 1374 lp = cp; 1375 for (;;) { 1376 printf("%c", c = cngetc() & 0177); 1377 switch (c) { 1378 case -1: 1379 case '\n': 1380 case '\r': 1381 *lp++ = '\0'; 1382 return; 1383 case '\b': 1384 case '\177': 1385 if (lp > cp) { 1386 printf(" \b"); 1387 lp--; 1388 } 1389 continue; 1390 case '#': 1391 lp--; 1392 if (lp < cp) 1393 lp = cp; 1394 continue; 1395 case '@': 1396 case 'u' & 037: 1397 lp = cp; 1398 printf("%c", '\n'); 1399 continue; 1400 default: 1401 *lp++ = c; 1402 } 1403 } 1404} 1405 1406/* 1407 * Convert a given name to the cdev pointer of the device, which is probably 1408 * but not by definition, a disk. Mount a DEVFS (on nothing), look the name 1409 * up, extract the cdev from the vnode and unmount it again. Unfortunately 1410 * we cannot use the vnode directly (because we unmount the DEVFS again) 1411 * so the filesystems still have to do the bdevvp() stunt. 1412 */ 1413static struct cdev * 1414getdiskbyname(char *name) 1415{ 1416 char *cp = name; 1417 struct cdev *dev = NULL; 1418 struct thread *td = curthread; 1419 struct vfsconf *vfsp; 1420 struct mount *mp = NULL; 1421 struct vnode *vroot = NULL; 1422 struct nameidata nid; 1423 int error; 1424 1425 if (!bcmp(cp, "/dev/", 5)) 1426 cp += 5; 1427 1428 do { 1429 vfsp = vfs_byname("devfs"); 1430 if (vfsp == NULL) 1431 break; 1432 error = vfs_mount_alloc(NULLVP, vfsp, "/dev", td, &mp); 1433 if (error) 1434 break; 1435 mp->mnt_flag |= MNT_RDONLY; 1436 1437 error = VFS_MOUNT(mp, curthread); 1438 if (error) 1439 break; 1440 VFS_START(mp, 0, td); 1441 VFS_ROOT(mp, &vroot, td); 1442 VOP_UNLOCK(vroot, 0, td); 1443 1444 NDINIT(&nid, LOOKUP, NOCACHE|FOLLOW, 1445 UIO_SYSSPACE, cp, curthread); 1446 nid.ni_startdir = vroot; 1447 nid.ni_pathlen = strlen(cp); 1448 nid.ni_cnd.cn_cred = curthread->td_ucred; 1449 nid.ni_cnd.cn_nameptr = cp; 1450 1451 error = lookup(&nid); 1452 if (error) 1453 break; 1454 if (nid.ni_vp->v_type != VCHR) 1455 dev = NULL; 1456 else 1457 dev = nid.ni_vp->v_rdev; 1458 NDFREE(&nid, 0); 1459 } while (0); 1460 1461 if (vroot != NULL) 1462 VFS_UNMOUNT(mp, 0, td); 1463 if (mp != NULL) 1464 vfs_mount_destroy(mp, td); 1465 return (dev); 1466} 1467 1468/* Show the struct cdev *for a disk specified by name */ 1469#ifdef DDB 1470DB_SHOW_COMMAND(disk, db_getdiskbyname) 1471{ 1472 struct cdev *dev; 1473 1474 if (modif[0] == '\0') { 1475 db_error("usage: show disk/devicename"); 1476 return; 1477 } 1478 dev = getdiskbyname(modif); 1479 if (dev != NULL) 1480 db_printf("struct cdev *= %p\n", dev); 1481 else 1482 db_printf("No disk device matched.\n"); 1483} 1484#endif 1485 1486/* 1487 * Get a mount option by its name. 1488 * 1489 * Return 0 if the option was found, ENOENT otherwise. 1490 * If len is non-NULL it will be filled with the length 1491 * of the option. If buf is non-NULL, it will be filled 1492 * with the address of the option. 1493 */ 1494int 1495vfs_getopt(opts, name, buf, len) 1496 struct vfsoptlist *opts; 1497 const char *name; 1498 void **buf; 1499 int *len; 1500{ 1501 struct vfsopt *opt; 1502 1503 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1504 1505 TAILQ_FOREACH(opt, opts, link) { 1506 if (strcmp(name, opt->name) == 0) { 1507 if (len != NULL) 1508 *len = opt->len; 1509 if (buf != NULL) 1510 *buf = opt->value; 1511 return (0); 1512 } 1513 } 1514 return (ENOENT); 1515} 1516 1517/* 1518 * Find and copy a mount option. 1519 * 1520 * The size of the buffer has to be specified 1521 * in len, if it is not the same length as the 1522 * mount option, EINVAL is returned. 1523 * Returns ENOENT if the option is not found. 1524 */ 1525int 1526vfs_copyopt(opts, name, dest, len) 1527 struct vfsoptlist *opts; 1528 const char *name; 1529 void *dest; 1530 int len; 1531{ 1532 struct vfsopt *opt; 1533 1534 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL")); 1535 1536 TAILQ_FOREACH(opt, opts, link) { 1537 if (strcmp(name, opt->name) == 0) { 1538 if (len != opt->len) 1539 return (EINVAL); 1540 bcopy(opt->value, dest, opt->len); 1541 return (0); 1542 } 1543 } 1544 return (ENOENT); 1545} 1546 1547 1548/* 1549 * This is a helper function for filesystems to traverse their 1550 * vnodes. See MNT_VNODE_FOREACH() in sys/mount.h 1551 */ 1552 1553struct vnode * 1554__mnt_vnode_next(struct vnode **nvp, struct mount *mp) 1555{ 1556 struct vnode *vp; 1557 1558 mtx_assert(&mp->mnt_mtx, MA_OWNED); 1559 1560 vp = *nvp; 1561 /* Check if we are done */ 1562 if (vp == NULL) 1563 return (NULL); 1564 /* If our next vnode is no longer ours, start over */ 1565 if (vp->v_mount != mp) 1566 vp = TAILQ_FIRST(&mp->mnt_nvnodelist); 1567 /* Save pointer to next vnode in list */ 1568 if (vp != NULL) 1569 *nvp = TAILQ_NEXT(vp, v_nmntvnodes); 1570 else 1571 *nvp = NULL; 1572 return (vp); 1573} 1574 1575int 1576__vfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td) 1577{ 1578 int error; 1579 1580 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat, td); 1581 if (sbp != &mp->mnt_stat) 1582 memcpy(sbp, &mp->mnt_stat, sizeof sbp); 1583 return (error); 1584} 1585