vfs_mount.c revision 158924
1/*- 2 * Copyright (c) 1999-2004 Poul-Henning Kamp 3 * Copyright (c) 1999 Michael Smith 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 */ 36 37#include <sys/cdefs.h> 38__FBSDID("$FreeBSD: head/sys/kern/vfs_mount.c 158924 2006-05-26 00:32:21Z rodrigc $"); 39 40#include <sys/param.h> 41#include <sys/conf.h> 42#include <sys/jail.h> 43#include <sys/kernel.h> 44#include <sys/libkern.h> 45#include <sys/mac.h> 46#include <sys/malloc.h> 47#include <sys/mount.h> 48#include <sys/mutex.h> 49#include <sys/namei.h> 50#include <sys/proc.h> 51#include <sys/filedesc.h> 52#include <sys/reboot.h> 53#include <sys/syscallsubr.h> 54#include <sys/sysproto.h> 55#include <sys/sx.h> 56#include <sys/sysctl.h> 57#include <sys/sysent.h> 58#include <sys/systm.h> 59#include <sys/vnode.h> 60#include <vm/uma.h> 61 62#include <geom/geom.h> 63 64#include <machine/stdarg.h> 65 66#include "opt_rootdevname.h" 67#include "opt_ddb.h" 68#include "opt_mac.h" 69 70#ifdef DDB 71#include <ddb/ddb.h> 72#endif 73 74#define ROOTNAME "root_device" 75#define VFS_MOUNTARG_SIZE_MAX (1024 * 64) 76 77static int vfs_domount(struct thread *td, const char *fstype, 78 char *fspath, int fsflags, void *fsdata); 79static struct mount *vfs_mount_alloc(struct vnode *dvp, struct vfsconf *vfsp, 80 const char *fspath, struct thread *td); 81static int vfs_mountroot_ask(void); 82static int vfs_mountroot_try(const char *mountfrom); 83static int vfs_donmount(struct thread *td, int fsflags, 84 struct uio *fsoptions); 85static void free_mntarg(struct mntarg *ma); 86static void vfs_mount_destroy(struct mount *, struct thread *); 87static int vfs_getopt_pos(struct vfsoptlist *opts, const char *name); 88 89static int usermount = 0; 90SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, 91 "Unprivileged users may mount and unmount file systems"); 92 93MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure"); 94MALLOC_DEFINE(M_VNODE_MARKER, "vnodemarker", "vnode marker"); 95static uma_zone_t mount_zone; 96 97/* List of mounted filesystems. */ 98struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist); 99 100/* For any iteration/modification of mountlist */ 101struct mtx mountlist_mtx; 102MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF); 103 104TAILQ_HEAD(vfsoptlist, vfsopt); 105struct vfsopt { 106 TAILQ_ENTRY(vfsopt) link; 107 char *name; 108 void *value; 109 int len; 110}; 111 112/* 113 * The vnode of the system's root (/ in the filesystem, without chroot 114 * active.) 115 */ 116struct vnode *rootvnode; 117 118/* 119 * The root filesystem is detailed in the kernel environment variable 120 * vfs.root.mountfrom, which is expected to be in the general format 121 * 122 * <vfsname>:[<path>] 123 * vfsname := the name of a VFS known to the kernel and capable 124 * of being mounted as root 125 * path := disk device name or other data used by the filesystem 126 * to locate its physical store 127 */ 128 129/* 130 * Global opts, taken by all filesystems 131 */ 132static const char *global_opts[] = { 133 "errmsg", 134 "fstype", 135 "fspath", 136 "rdonly", 137 "ro", 138 "rw", 139 "suid", 140 "exec", 141 NULL 142}; 143 144/* 145 * The root specifiers we will try if RB_CDROM is specified. 146 */ 147static char *cdrom_rootdevnames[] = { 148 "cd9660:cd0", 149 "cd9660:acd0", 150 NULL 151}; 152 153/* legacy find-root code */ 154char *rootdevnames[2] = {NULL, NULL}; 155#ifndef ROOTDEVNAME 156# define ROOTDEVNAME NULL 157#endif 158static const char *ctrootdevname = ROOTDEVNAME; 159 160/* 161 * --------------------------------------------------------------------- 162 * Functions for building and sanitizing the mount options 163 */ 164 165/* Remove one mount option. */ 166static void 167vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt) 168{ 169 170 TAILQ_REMOVE(opts, opt, link); 171 free(opt->name, M_MOUNT); 172 if (opt->value != NULL) 173 free(opt->value, M_MOUNT); 174#ifdef INVARIANTS 175 else if (opt->len != 0) 176 panic("%s: mount option with NULL value but length != 0", 177 __func__); 178#endif 179 free(opt, M_MOUNT); 180} 181 182/* Release all resources related to the mount options. */ 183static void 184vfs_freeopts(struct vfsoptlist *opts) 185{ 186 struct vfsopt *opt; 187 188 while (!TAILQ_EMPTY(opts)) { 189 opt = TAILQ_FIRST(opts); 190 vfs_freeopt(opts, opt); 191 } 192 free(opts, M_MOUNT); 193} 194 195/* 196 * Check if options are equal (with or without the "no" prefix). 197 */ 198static int 199vfs_equalopts(const char *opt1, const char *opt2) 200{ 201 202 /* "opt" vs. "opt" or "noopt" vs. "noopt" */ 203 if (strcmp(opt1, opt2) == 0) 204 return (1); 205 /* "noopt" vs. "opt" */ 206 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0) 207 return (1); 208 /* "opt" vs. "noopt" */ 209 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0) 210 return (1); 211 return (0); 212} 213 214/* 215 * If a mount option is specified several times, 216 * (with or without the "no" prefix) only keep 217 * the last occurence of it. 218 */ 219static void 220vfs_sanitizeopts(struct vfsoptlist *opts) 221{ 222 struct vfsopt *opt, *opt2, *tmp; 223 224 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) { 225 opt2 = TAILQ_PREV(opt, vfsoptlist, link); 226 while (opt2 != NULL) { 227 if (vfs_equalopts(opt->name, opt2->name)) { 228 tmp = TAILQ_PREV(opt2, vfsoptlist, link); 229 vfs_freeopt(opts, opt2); 230 opt2 = tmp; 231 } else { 232 opt2 = TAILQ_PREV(opt2, vfsoptlist, link); 233 } 234 } 235 } 236} 237 238/* 239 * Build a linked list of mount options from a struct uio. 240 */ 241static int 242vfs_buildopts(struct uio *auio, struct vfsoptlist **options) 243{ 244 struct vfsoptlist *opts; 245 struct vfsopt *opt; 246 size_t memused; 247 unsigned int i, iovcnt; 248 int error, namelen, optlen; 249 250 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 251 TAILQ_INIT(opts); 252 memused = 0; 253 iovcnt = auio->uio_iovcnt; 254 for (i = 0; i < iovcnt; i += 2) { 255 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 256 namelen = auio->uio_iov[i].iov_len; 257 optlen = auio->uio_iov[i + 1].iov_len; 258 opt->name = malloc(namelen, M_MOUNT, M_WAITOK); 259 opt->value = NULL; 260 opt->len = 0; 261 262 /* 263 * Do this early, so jumps to "bad" will free the current 264 * option. 265 */ 266 TAILQ_INSERT_TAIL(opts, opt, link); 267 memused += sizeof(struct vfsopt) + optlen + namelen; 268 269 /* 270 * Avoid consuming too much memory, and attempts to overflow 271 * memused. 272 */ 273 if (memused > VFS_MOUNTARG_SIZE_MAX || 274 optlen > VFS_MOUNTARG_SIZE_MAX || 275 namelen > VFS_MOUNTARG_SIZE_MAX) { 276 error = EINVAL; 277 goto bad; 278 } 279 280 if (auio->uio_segflg == UIO_SYSSPACE) { 281 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen); 282 } else { 283 error = copyin(auio->uio_iov[i].iov_base, opt->name, 284 namelen); 285 if (error) 286 goto bad; 287 } 288 /* Ensure names are null-terminated strings. */ 289 if (opt->name[namelen - 1] != '\0') { 290 error = EINVAL; 291 goto bad; 292 } 293 if (optlen != 0) { 294 opt->len = optlen; 295 opt->value = malloc(optlen, M_MOUNT, M_WAITOK); 296 if (auio->uio_segflg == UIO_SYSSPACE) { 297 bcopy(auio->uio_iov[i + 1].iov_base, opt->value, 298 optlen); 299 } else { 300 error = copyin(auio->uio_iov[i + 1].iov_base, 301 opt->value, optlen); 302 if (error) 303 goto bad; 304 } 305 } 306 } 307 vfs_sanitizeopts(opts); 308 *options = opts; 309 return (0); 310bad: 311 vfs_freeopts(opts); 312 return (error); 313} 314 315/* 316 * Merge the old mount options with the new ones passed 317 * in the MNT_UPDATE case. 318 */ 319static void 320vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts) 321{ 322 struct vfsopt *opt, *opt2, *new; 323 324 TAILQ_FOREACH(opt, opts, link) { 325 /* 326 * Check that this option hasn't been redefined 327 * nor cancelled with a "no" mount option. 328 */ 329 opt2 = TAILQ_FIRST(toopts); 330 while (opt2 != NULL) { 331 if (strcmp(opt2->name, opt->name) == 0) 332 goto next; 333 if (strncmp(opt2->name, "no", 2) == 0 && 334 strcmp(opt2->name + 2, opt->name) == 0) { 335 vfs_freeopt(toopts, opt2); 336 goto next; 337 } 338 opt2 = TAILQ_NEXT(opt2, link); 339 } 340 /* We want this option, duplicate it. */ 341 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 342 new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK); 343 strcpy(new->name, opt->name); 344 if (opt->len != 0) { 345 new->value = malloc(opt->len, M_MOUNT, M_WAITOK); 346 bcopy(opt->value, new->value, opt->len); 347 } else { 348 new->value = NULL; 349 } 350 new->len = opt->len; 351 TAILQ_INSERT_TAIL(toopts, new, link); 352next: 353 continue; 354 } 355} 356 357/* 358 * --------------------------------------------------------------------- 359 * Mount a filesystem 360 */ 361int 362nmount(td, uap) 363 struct thread *td; 364 struct nmount_args /* { 365 struct iovec *iovp; 366 unsigned int iovcnt; 367 int flags; 368 } */ *uap; 369{ 370 struct uio *auio; 371 struct iovec *iov; 372 unsigned int i; 373 int error; 374 u_int iovcnt; 375 376 /* Kick out MNT_ROOTFS early as it is legal internally */ 377 if (uap->flags & MNT_ROOTFS) 378 return (EINVAL); 379 380 iovcnt = uap->iovcnt; 381 /* 382 * Check that we have an even number of iovec's 383 * and that we have at least two options. 384 */ 385 if ((iovcnt & 1) || (iovcnt < 4)) 386 return (EINVAL); 387 388 error = copyinuio(uap->iovp, iovcnt, &auio); 389 if (error) 390 return (error); 391 iov = auio->uio_iov; 392 for (i = 0; i < iovcnt; i++) { 393 if (iov->iov_len > MMAXOPTIONLEN) { 394 free(auio, M_IOV); 395 return (EINVAL); 396 } 397 iov++; 398 } 399 error = vfs_donmount(td, uap->flags, auio); 400 401 free(auio, M_IOV); 402 return (error); 403} 404 405/* 406 * --------------------------------------------------------------------- 407 * Various utility functions 408 */ 409 410void 411vfs_ref(struct mount *mp) 412{ 413 414 MNT_ILOCK(mp); 415 MNT_REF(mp); 416 MNT_IUNLOCK(mp); 417} 418 419void 420vfs_rel(struct mount *mp) 421{ 422 423 MNT_ILOCK(mp); 424 MNT_REL(mp); 425 MNT_IUNLOCK(mp); 426} 427 428static int 429mount_init(void *mem, int size, int flags) 430{ 431 struct mount *mp; 432 433 mp = (struct mount *)mem; 434 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF); 435 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 436 return (0); 437} 438 439static void 440mount_fini(void *mem, int size) 441{ 442 struct mount *mp; 443 444 mp = (struct mount *)mem; 445 lockdestroy(&mp->mnt_lock); 446 mtx_destroy(&mp->mnt_mtx); 447} 448 449/* 450 * Allocate and initialize the mount point struct. 451 */ 452static struct mount * 453vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, 454 const char *fspath, struct thread *td) 455{ 456 struct mount *mp; 457 458 mp = uma_zalloc(mount_zone, M_WAITOK); 459 bzero(&mp->mnt_startzero, 460 __rangeof(struct mount, mnt_startzero, mnt_endzero)); 461 TAILQ_INIT(&mp->mnt_nvnodelist); 462 mp->mnt_nvnodelistsize = 0; 463 mp->mnt_ref = 0; 464 (void) vfs_busy(mp, LK_NOWAIT, 0, td); 465 mp->mnt_op = vfsp->vfc_vfsops; 466 mp->mnt_vfc = vfsp; 467 vfsp->vfc_refcount++; /* XXX Unlocked */ 468 mp->mnt_stat.f_type = vfsp->vfc_typenum; 469 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 470 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 471 mp->mnt_vnodecovered = vp; 472 mp->mnt_cred = crdup(td->td_ucred); 473 mp->mnt_stat.f_owner = td->td_ucred->cr_uid; 474 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN); 475 mp->mnt_iosize_max = DFLTPHYS; 476#ifdef MAC 477 mac_init_mount(mp); 478 mac_create_mount(td->td_ucred, mp); 479#endif 480 arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0); 481 return (mp); 482} 483 484/* 485 * Destroy the mount struct previously allocated by vfs_mount_alloc(). 486 */ 487static void 488vfs_mount_destroy(struct mount *mp, struct thread *td) 489{ 490 int i; 491 492 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, td); 493 MNT_ILOCK(mp); 494 for (i = 0; mp->mnt_ref && i < 3; i++) 495 msleep(mp, MNT_MTX(mp), PVFS, "mntref", hz); 496 /* 497 * This will always cause a 3 second delay in rebooting due to 498 * refs on the root mountpoint that never go away. Most of these 499 * are held by init which never exits. 500 */ 501 if (i == 3 && (!rebooting || bootverbose)) 502 printf("Mount point %s had %d dangling refs\n", 503 mp->mnt_stat.f_mntonname, mp->mnt_ref); 504 if (mp->mnt_holdcnt != 0) { 505 printf("Waiting for mount point to be unheld\n"); 506 while (mp->mnt_holdcnt != 0) { 507 mp->mnt_holdcntwaiters++; 508 msleep(&mp->mnt_holdcnt, MNT_MTX(mp), 509 PZERO, "mntdestroy", 0); 510 mp->mnt_holdcntwaiters--; 511 } 512 printf("mount point unheld\n"); 513 } 514 if (mp->mnt_writeopcount > 0) { 515 printf("Waiting for mount point write ops\n"); 516 while (mp->mnt_writeopcount > 0) { 517 mp->mnt_kern_flag |= MNTK_SUSPEND; 518 msleep(&mp->mnt_writeopcount, 519 MNT_MTX(mp), 520 PZERO, "mntdestroy2", 0); 521 } 522 printf("mount point write ops completed\n"); 523 } 524 if (mp->mnt_secondary_writes > 0) { 525 printf("Waiting for mount point secondary write ops\n"); 526 while (mp->mnt_secondary_writes > 0) { 527 mp->mnt_kern_flag |= MNTK_SUSPEND; 528 msleep(&mp->mnt_secondary_writes, 529 MNT_MTX(mp), 530 PZERO, "mntdestroy3", 0); 531 } 532 printf("mount point secondary write ops completed\n"); 533 } 534 MNT_IUNLOCK(mp); 535 mp->mnt_vfc->vfc_refcount--; 536 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) { 537 struct vnode *vp; 538 539 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes) 540 vprint("", vp); 541 panic("unmount: dangling vnode"); 542 } 543 MNT_ILOCK(mp); 544 if (mp->mnt_kern_flag & MNTK_MWAIT) 545 wakeup(mp); 546 if (mp->mnt_writeopcount != 0) 547 panic("vfs_mount_destroy: nonzero writeopcount"); 548 if (mp->mnt_secondary_writes != 0) 549 panic("vfs_mount_destroy: nonzero secondary_writes"); 550 if (mp->mnt_nvnodelistsize != 0) 551 panic("vfs_mount_destroy: nonzero nvnodelistsize"); 552 mp->mnt_writeopcount = -1000; 553 mp->mnt_nvnodelistsize = -1000; 554 mp->mnt_secondary_writes = -1000; 555 MNT_IUNLOCK(mp); 556#ifdef MAC 557 mac_destroy_mount(mp); 558#endif 559 if (mp->mnt_opt != NULL) 560 vfs_freeopts(mp->mnt_opt); 561 crfree(mp->mnt_cred); 562 uma_zfree(mount_zone, mp); 563} 564 565static int 566vfs_donmount(struct thread *td, int fsflags, struct uio *fsoptions) 567{ 568 struct vfsoptlist *optlist; 569 struct vfsopt *opt, *noro_opt; 570 char *fstype, *fspath, *errmsg; 571 int error, fstypelen, fspathlen, errmsg_len, errmsg_pos; 572 int has_rw, has_noro; 573 574 errmsg_len = 0; 575 errmsg_pos = -1; 576 has_rw = 0; 577 has_noro = 0; 578 579 error = vfs_buildopts(fsoptions, &optlist); 580 if (error) 581 return (error); 582 583 if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0) 584 errmsg_pos = vfs_getopt_pos(optlist, "errmsg"); 585 else 586 errmsg_len = 0; 587 588 /* 589 * We need these two options before the others, 590 * and they are mandatory for any filesystem. 591 * Ensure they are NUL terminated as well. 592 */ 593 fstypelen = 0; 594 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen); 595 if (error || fstype[fstypelen - 1] != '\0') { 596 error = EINVAL; 597 if (errmsg != NULL) 598 strncpy(errmsg, "Invalid fstype", errmsg_len); 599 goto bail; 600 } 601 fspathlen = 0; 602 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen); 603 if (error || fspath[fspathlen - 1] != '\0') { 604 error = EINVAL; 605 if (errmsg != NULL) 606 strncpy(errmsg, "Invalid fspath", errmsg_len); 607 goto bail; 608 } 609 610 /* 611 * We need to see if we have the "update" option 612 * before we call vfs_domount(), since vfs_domount() has special 613 * logic based on MNT_UPDATE. This is very important 614 * when we want to update the root filesystem. 615 */ 616 TAILQ_FOREACH(opt, optlist, link) { 617 if (strcmp(opt->name, "update") == 0) 618 fsflags |= MNT_UPDATE; 619 else if (strcmp(opt->name, "async") == 0) 620 fsflags |= MNT_ASYNC; 621 else if (strcmp(opt->name, "force") == 0) 622 fsflags |= MNT_FORCE; 623 else if (strcmp(opt->name, "multilabel") == 0) 624 fsflags |= MNT_MULTILABEL; 625 else if (strcmp(opt->name, "noasync") == 0) 626 fsflags &= ~MNT_ASYNC; 627 else if (strcmp(opt->name, "noatime") == 0) 628 fsflags |= MNT_NOATIME; 629 else if (strcmp(opt->name, "noclusterr") == 0) 630 fsflags |= MNT_NOCLUSTERR; 631 else if (strcmp(opt->name, "noclusterw") == 0) 632 fsflags |= MNT_NOCLUSTERW; 633 else if (strcmp(opt->name, "noexec") == 0) 634 fsflags |= MNT_NOEXEC; 635 else if (strcmp(opt->name, "nosuid") == 0) 636 fsflags |= MNT_NOSUID; 637 else if (strcmp(opt->name, "nosymfollow") == 0) 638 fsflags |= MNT_NOSYMFOLLOW; 639 else if (strcmp(opt->name, "noro") == 0) { 640 fsflags &= ~MNT_RDONLY; 641 has_noro = 1; 642 } 643 else if (strcmp(opt->name, "rw") == 0) { 644 fsflags &= ~MNT_RDONLY; 645 has_rw = 1; 646 } 647 else if (strcmp(opt->name, "ro") == 0 || 648 strcmp(opt->name, "rdonly") == 0) 649 fsflags |= MNT_RDONLY; 650 else if (strcmp(opt->name, "snapshot") == 0) 651 fsflags |= MNT_SNAPSHOT; 652 else if (strcmp(opt->name, "suiddir") == 0) 653 fsflags |= MNT_SUIDDIR; 654 else if (strcmp(opt->name, "sync") == 0) 655 fsflags |= MNT_SYNCHRONOUS; 656 else if (strcmp(opt->name, "union") == 0) 657 fsflags |= MNT_UNION; 658 } 659 660 /* 661 * If "rw" was specified as a mount option, and we 662 * are trying to update a mount-point from "ro" to "rw", 663 * we need a mount option "noro", since in vfs_mergeopts(), 664 * "noro" will cancel "ro", but "rw" will not do anything. 665 */ 666 if (has_rw && !has_noro) { 667 noro_opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK); 668 noro_opt->name = strdup("noro", M_MOUNT); 669 noro_opt->value = NULL; 670 noro_opt->len = 0; 671 TAILQ_INSERT_TAIL(optlist, noro_opt, link); 672 } 673 674 /* 675 * Be ultra-paranoid about making sure the type and fspath 676 * variables will fit in our mp buffers, including the 677 * terminating NUL. 678 */ 679 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) { 680 error = ENAMETOOLONG; 681 goto bail; 682 } 683 684 mtx_lock(&Giant); 685 error = vfs_domount(td, fstype, fspath, fsflags, optlist); 686 mtx_unlock(&Giant); 687bail: 688 /* copyout the errmsg */ 689 if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt) 690 && errmsg_len > 0 && errmsg != NULL) { 691 if (fsoptions->uio_segflg == UIO_SYSSPACE) { 692 strncpy(fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base, 693 errmsg, 694 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len); 695 } else { 696 copystr(errmsg, 697 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base, 698 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len, 699 NULL); 700 } 701 } 702 703 if (error != 0) 704 vfs_freeopts(optlist); 705 return (error); 706} 707 708/* 709 * --------------------------------------------------------------------- 710 * Old mount API. 711 */ 712#ifndef _SYS_SYSPROTO_H_ 713struct mount_args { 714 char *type; 715 char *path; 716 int flags; 717 caddr_t data; 718}; 719#endif 720/* ARGSUSED */ 721int 722mount(td, uap) 723 struct thread *td; 724 struct mount_args /* { 725 char *type; 726 char *path; 727 int flags; 728 caddr_t data; 729 } */ *uap; 730{ 731 char *fstype; 732 struct vfsconf *vfsp = NULL; 733 struct mntarg *ma = NULL; 734 int error; 735 736 /* Kick out MNT_ROOTFS early as it is legal internally */ 737 uap->flags &= ~MNT_ROOTFS; 738 739 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK); 740 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL); 741 if (!error) { 742 mtx_lock(&Giant); /* XXX ? */ 743 vfsp = vfs_byname_kld(fstype, td, &error); 744 mtx_unlock(&Giant); 745 } 746 free(fstype, M_TEMP); 747 if (error) 748 return (error); 749 if (vfsp == NULL) 750 return (ENOENT); 751 if (vfsp->vfc_vfsops->vfs_cmount == NULL) 752 return (EOPNOTSUPP); 753 754 ma = mount_argsu(ma, "fstype", uap->type, MNAMELEN); 755 ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN); 756 ma = mount_argb(ma, uap->flags & MNT_RDONLY, "noro"); 757 ma = mount_argb(ma, !(uap->flags & MNT_NOSUID), "nosuid"); 758 ma = mount_argb(ma, !(uap->flags & MNT_NOEXEC), "noexec"); 759 760 error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, uap->flags, td); 761 return (error); 762} 763 764 765/* 766 * vfs_domount(): actually attempt a filesystem mount. 767 */ 768static int 769vfs_domount( 770 struct thread *td, /* Flags common to all filesystems. */ 771 const char *fstype, /* Filesystem type. */ 772 char *fspath, /* Mount path. */ 773 int fsflags, /* Flags common to all filesystems. */ 774 void *fsdata /* Options local to the filesystem. */ 775 ) 776{ 777 struct vnode *vp; 778 struct mount *mp; 779 struct vfsconf *vfsp; 780 struct export_args export; 781 int error, flag = 0, kern_flag = 0; 782 struct vattr va; 783 struct nameidata nd; 784 785 mtx_assert(&Giant, MA_OWNED); 786 /* 787 * Be ultra-paranoid about making sure the type and fspath 788 * variables will fit in our mp buffers, including the 789 * terminating NUL. 790 */ 791 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN) 792 return (ENAMETOOLONG); 793 794 if (jailed(td->td_ucred)) 795 return (EPERM); 796 if (usermount == 0) { 797 if ((error = suser(td)) != 0) 798 return (error); 799 } 800 801 /* 802 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users. 803 */ 804 if (fsflags & (MNT_EXPORTED | MNT_SUIDDIR)) { 805 if ((error = suser(td)) != 0) 806 return (error); 807 } 808 /* 809 * Silently enforce MNT_NOSUID and MNT_USER for 810 * unprivileged users. 811 */ 812 if (suser(td) != 0) 813 fsflags |= MNT_NOSUID | MNT_USER; 814 815 /* Load KLDs before we lock the covered vnode to avoid reversals. */ 816 vfsp = NULL; 817 if ((fsflags & MNT_UPDATE) == 0) { 818 /* Don't try to load KLDs if we're mounting the root. */ 819 if (fsflags & MNT_ROOTFS) 820 vfsp = vfs_byname(fstype); 821 else 822 vfsp = vfs_byname_kld(fstype, td, &error); 823 if (vfsp == NULL) 824 return (ENODEV); 825 } 826 /* 827 * Get vnode to be covered 828 */ 829 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td); 830 if ((error = namei(&nd)) != 0) 831 return (error); 832 NDFREE(&nd, NDF_ONLY_PNBUF); 833 vp = nd.ni_vp; 834 if (fsflags & MNT_UPDATE) { 835 if ((vp->v_vflag & VV_ROOT) == 0) { 836 vput(vp); 837 return (EINVAL); 838 } 839 mp = vp->v_mount; 840 flag = mp->mnt_flag; 841 kern_flag = mp->mnt_kern_flag; 842 /* 843 * We only allow the filesystem to be reloaded if it 844 * is currently mounted read-only. 845 */ 846 if ((fsflags & MNT_RELOAD) && 847 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 848 vput(vp); 849 return (EOPNOTSUPP); /* Needs translation */ 850 } 851 /* 852 * Only privileged root, or (if MNT_USER is set) the user that 853 * did the original mount is permitted to update it. 854 */ 855 error = vfs_suser(mp, td); 856 if (error) { 857 vput(vp); 858 return (error); 859 } 860 if (vfs_busy(mp, LK_NOWAIT, 0, td)) { 861 vput(vp); 862 return (EBUSY); 863 } 864 VI_LOCK(vp); 865 if ((vp->v_iflag & VI_MOUNT) != 0 || 866 vp->v_mountedhere != NULL) { 867 VI_UNLOCK(vp); 868 vfs_unbusy(mp, td); 869 vput(vp); 870 return (EBUSY); 871 } 872 vp->v_iflag |= VI_MOUNT; 873 VI_UNLOCK(vp); 874 mp->mnt_flag |= fsflags & 875 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT | MNT_ROOTFS); 876 VOP_UNLOCK(vp, 0, td); 877 mp->mnt_optnew = fsdata; 878 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt); 879 } else { 880 /* 881 * If the user is not root, ensure that they own the directory 882 * onto which we are attempting to mount. 883 */ 884 error = VOP_GETATTR(vp, &va, td->td_ucred, td); 885 if (error) { 886 vput(vp); 887 return (error); 888 } 889 if (va.va_uid != td->td_ucred->cr_uid) { 890 if ((error = suser(td)) != 0) { 891 vput(vp); 892 return (error); 893 } 894 } 895 error = vinvalbuf(vp, V_SAVE, td, 0, 0); 896 if (error != 0) { 897 vput(vp); 898 return (error); 899 } 900 if (vp->v_type != VDIR) { 901 vput(vp); 902 return (ENOTDIR); 903 } 904 VI_LOCK(vp); 905 if ((vp->v_iflag & VI_MOUNT) != 0 || 906 vp->v_mountedhere != NULL) { 907 VI_UNLOCK(vp); 908 vput(vp); 909 return (EBUSY); 910 } 911 vp->v_iflag |= VI_MOUNT; 912 VI_UNLOCK(vp); 913 914 /* 915 * Allocate and initialize the filesystem. 916 */ 917 mp = vfs_mount_alloc(vp, vfsp, fspath, td); 918 VOP_UNLOCK(vp, 0, td); 919 920 /* XXXMAC: pass to vfs_mount_alloc? */ 921 mp->mnt_optnew = fsdata; 922 } 923 924 /* 925 * Set the mount level flags. 926 */ 927 if (fsflags & MNT_RDONLY) 928 mp->mnt_flag |= MNT_RDONLY; 929 mp->mnt_flag &=~ MNT_UPDATEMASK; 930 mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE | MNT_ROOTFS); 931 /* 932 * Mount the filesystem. 933 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 934 * get. No freeing of cn_pnbuf. 935 */ 936 error = VFS_MOUNT(mp, td); 937 938 /* 939 * Process the export option only if we are 940 * updating mount options. 941 */ 942 if (!error && (fsflags & MNT_UPDATE)) { 943 if (vfs_copyopt(mp->mnt_optnew, "export", &export, 944 sizeof(export)) == 0) 945 error = vfs_export(mp, &export); 946 } 947 948 if (!error) { 949 if (mp->mnt_opt != NULL) 950 vfs_freeopts(mp->mnt_opt); 951 mp->mnt_opt = mp->mnt_optnew; 952 (void)VFS_STATFS(mp, &mp->mnt_stat, td); 953 } 954 /* 955 * Prevent external consumers of mount options from reading 956 * mnt_optnew. 957 */ 958 mp->mnt_optnew = NULL; 959 if (mp->mnt_flag & MNT_UPDATE) { 960 mp->mnt_flag &= 961 ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT); 962 if (error) { 963 mp->mnt_flag = flag; 964 mp->mnt_kern_flag = kern_flag; 965 } 966 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 967 if (mp->mnt_syncer == NULL) 968 error = vfs_allocate_syncvnode(mp); 969 } else { 970 if (mp->mnt_syncer != NULL) 971 vrele(mp->mnt_syncer); 972 mp->mnt_syncer = NULL; 973 } 974 vfs_unbusy(mp, td); 975 VI_LOCK(vp); 976 vp->v_iflag &= ~VI_MOUNT; 977 VI_UNLOCK(vp); 978 vrele(vp); 979 return (error); 980 } 981 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 982 /* 983 * Put the new filesystem on the mount list after root. 984 */ 985 cache_purge(vp); 986 if (!error) { 987 struct vnode *newdp; 988 989 VI_LOCK(vp); 990 vp->v_iflag &= ~VI_MOUNT; 991 VI_UNLOCK(vp); 992 vp->v_mountedhere = mp; 993 mtx_lock(&mountlist_mtx); 994 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 995 mtx_unlock(&mountlist_mtx); 996 vfs_event_signal(NULL, VQ_MOUNT, 0); 997 if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp, td)) 998 panic("mount: lost mount"); 999 mountcheckdirs(vp, newdp); 1000 vput(newdp); 1001 VOP_UNLOCK(vp, 0, td); 1002 if ((mp->mnt_flag & MNT_RDONLY) == 0) 1003 error = vfs_allocate_syncvnode(mp); 1004 vfs_unbusy(mp, td); 1005 if (error) 1006 vrele(vp); 1007 } else { 1008 VI_LOCK(vp); 1009 vp->v_iflag &= ~VI_MOUNT; 1010 VI_UNLOCK(vp); 1011 vfs_mount_destroy(mp, td); 1012 vput(vp); 1013 } 1014 return (error); 1015} 1016 1017/* 1018 * --------------------------------------------------------------------- 1019 * Unmount a filesystem. 1020 * 1021 * Note: unmount takes a path to the vnode mounted on as argument, 1022 * not special file (as before). 1023 */ 1024#ifndef _SYS_SYSPROTO_H_ 1025struct unmount_args { 1026 char *path; 1027 int flags; 1028}; 1029#endif 1030/* ARGSUSED */ 1031int 1032unmount(td, uap) 1033 struct thread *td; 1034 register struct unmount_args /* { 1035 char *path; 1036 int flags; 1037 } */ *uap; 1038{ 1039 struct mount *mp; 1040 char *pathbuf; 1041 int error, id0, id1; 1042 1043 if (jailed(td->td_ucred)) 1044 return (EPERM); 1045 if (usermount == 0) { 1046 if ((error = suser(td)) != 0) 1047 return (error); 1048 } 1049 1050 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK); 1051 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL); 1052 if (error) { 1053 free(pathbuf, M_TEMP); 1054 return (error); 1055 } 1056 if (uap->flags & MNT_BYFSID) { 1057 /* Decode the filesystem ID. */ 1058 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) { 1059 free(pathbuf, M_TEMP); 1060 return (EINVAL); 1061 } 1062 1063 mtx_lock(&mountlist_mtx); 1064 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1065 if (mp->mnt_stat.f_fsid.val[0] == id0 && 1066 mp->mnt_stat.f_fsid.val[1] == id1) 1067 break; 1068 } 1069 mtx_unlock(&mountlist_mtx); 1070 } else { 1071 mtx_lock(&mountlist_mtx); 1072 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) { 1073 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) 1074 break; 1075 } 1076 mtx_unlock(&mountlist_mtx); 1077 } 1078 free(pathbuf, M_TEMP); 1079 if (mp == NULL) { 1080 /* 1081 * Previously we returned ENOENT for a nonexistent path and 1082 * EINVAL for a non-mountpoint. We cannot tell these apart 1083 * now, so in the !MNT_BYFSID case return the more likely 1084 * EINVAL for compatibility. 1085 */ 1086 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL); 1087 } 1088 1089 /* 1090 * Only privileged root, or (if MNT_USER is set) the user that did the 1091 * original mount is permitted to unmount this filesystem. 1092 */ 1093 error = vfs_suser(mp, td); 1094 if (error) 1095 return (error); 1096 1097 /* 1098 * Don't allow unmounting the root filesystem. 1099 */ 1100 if (mp->mnt_flag & MNT_ROOTFS) 1101 return (EINVAL); 1102 mtx_lock(&Giant); 1103 error = dounmount(mp, uap->flags, td); 1104 mtx_unlock(&Giant); 1105 return (error); 1106} 1107 1108/* 1109 * Do the actual filesystem unmount. 1110 */ 1111int 1112dounmount(mp, flags, td) 1113 struct mount *mp; 1114 int flags; 1115 struct thread *td; 1116{ 1117 struct vnode *coveredvp, *fsrootvp; 1118 int error; 1119 int async_flag; 1120 1121 mtx_assert(&Giant, MA_OWNED); 1122 1123 if ((coveredvp = mp->mnt_vnodecovered) != NULL) 1124 vn_lock(coveredvp, LK_EXCLUSIVE | LK_RETRY, td); 1125 MNT_ILOCK(mp); 1126 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 1127 MNT_IUNLOCK(mp); 1128 if (coveredvp) 1129 VOP_UNLOCK(coveredvp, 0, td); 1130 return (EBUSY); 1131 } 1132 mp->mnt_kern_flag |= MNTK_UNMOUNT; 1133 /* Allow filesystems to detect that a forced unmount is in progress. */ 1134 if (flags & MNT_FORCE) 1135 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 1136 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 1137 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), MNT_MTX(mp), td); 1138 if (error) { 1139 MNT_ILOCK(mp); 1140 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1141 if (mp->mnt_kern_flag & MNTK_MWAIT) 1142 wakeup(mp); 1143 MNT_IUNLOCK(mp); 1144 if (coveredvp) 1145 VOP_UNLOCK(coveredvp, 0, td); 1146 return (error); 1147 } 1148 vn_start_write(NULL, &mp, V_WAIT); 1149 1150 if (mp->mnt_flag & MNT_EXPUBLIC) 1151 vfs_setpublicfs(NULL, NULL, NULL); 1152 1153 vfs_msync(mp, MNT_WAIT); 1154 async_flag = mp->mnt_flag & MNT_ASYNC; 1155 mp->mnt_flag &= ~MNT_ASYNC; 1156 cache_purgevfs(mp); /* remove cache entries for this file sys */ 1157 if (mp->mnt_syncer != NULL) 1158 vrele(mp->mnt_syncer); 1159 /* 1160 * For forced unmounts, move process cdir/rdir refs on the fs root 1161 * vnode to the covered vnode. For non-forced unmounts we want 1162 * such references to cause an EBUSY error. 1163 */ 1164 if ((flags & MNT_FORCE) && 1165 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp, td) == 0) { 1166 if (mp->mnt_vnodecovered != NULL) 1167 mountcheckdirs(fsrootvp, mp->mnt_vnodecovered); 1168 if (fsrootvp == rootvnode) { 1169 vrele(rootvnode); 1170 rootvnode = NULL; 1171 } 1172 vput(fsrootvp); 1173 } 1174 if (((mp->mnt_flag & MNT_RDONLY) || 1175 (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) || 1176 (flags & MNT_FORCE)) { 1177 error = VFS_UNMOUNT(mp, flags, td); 1178 } 1179 vn_finished_write(mp); 1180 if (error) { 1181 /* Undo cdir/rdir and rootvnode changes made above. */ 1182 if ((flags & MNT_FORCE) && 1183 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp, td) == 0) { 1184 if (mp->mnt_vnodecovered != NULL) 1185 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp); 1186 if (rootvnode == NULL) { 1187 rootvnode = fsrootvp; 1188 vref(rootvnode); 1189 } 1190 vput(fsrootvp); 1191 } 1192 if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL) 1193 (void) vfs_allocate_syncvnode(mp); 1194 MNT_ILOCK(mp); 1195 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 1196 mp->mnt_flag |= async_flag; 1197 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, td); 1198 if (mp->mnt_kern_flag & MNTK_MWAIT) 1199 wakeup(mp); 1200 MNT_IUNLOCK(mp); 1201 if (coveredvp) 1202 VOP_UNLOCK(coveredvp, 0, td); 1203 return (error); 1204 } 1205 mtx_lock(&mountlist_mtx); 1206 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1207 mtx_unlock(&mountlist_mtx); 1208 if (coveredvp != NULL) { 1209 coveredvp->v_mountedhere = NULL; 1210 vput(coveredvp); 1211 } 1212 vfs_event_signal(NULL, VQ_UNMOUNT, 0); 1213 vfs_mount_destroy(mp, td); 1214 return (0); 1215} 1216 1217/* 1218 * --------------------------------------------------------------------- 1219 * Mounting of root filesystem 1220 * 1221 */ 1222 1223struct root_hold_token { 1224 const char *who; 1225 LIST_ENTRY(root_hold_token) list; 1226}; 1227 1228static LIST_HEAD(, root_hold_token) root_holds = 1229 LIST_HEAD_INITIALIZER(&root_holds); 1230 1231struct root_hold_token * 1232root_mount_hold(const char *identifier) 1233{ 1234 struct root_hold_token *h; 1235 1236 h = malloc(sizeof *h, M_DEVBUF, M_ZERO | M_WAITOK); 1237 h->who = identifier; 1238 mtx_lock(&mountlist_mtx); 1239 LIST_INSERT_HEAD(&root_holds, h, list); 1240 mtx_unlock(&mountlist_mtx); 1241 return (h); 1242} 1243 1244void 1245root_mount_rel(struct root_hold_token *h) 1246{ 1247 1248 mtx_lock(&mountlist_mtx); 1249 LIST_REMOVE(h, list); 1250 wakeup(&root_holds); 1251 mtx_unlock(&mountlist_mtx); 1252 free(h, M_DEVBUF); 1253} 1254 1255static void 1256root_mount_wait(void) 1257{ 1258 struct root_hold_token *h; 1259 1260 for (;;) { 1261 DROP_GIANT(); 1262 g_waitidle(); 1263 PICKUP_GIANT(); 1264 mtx_lock(&mountlist_mtx); 1265 if (LIST_EMPTY(&root_holds)) { 1266 mtx_unlock(&mountlist_mtx); 1267 break; 1268 } 1269 printf("Root mount waiting for:"); 1270 LIST_FOREACH(h, &root_holds, list) 1271 printf(" %s", h->who); 1272 printf("\n"); 1273 msleep(&root_holds, &mountlist_mtx, PZERO | PDROP, "roothold", 1274 hz); 1275 } 1276} 1277 1278static void 1279set_rootvnode(struct thread *td) 1280{ 1281 struct proc *p; 1282 1283 if (VFS_ROOT(TAILQ_FIRST(&mountlist), LK_EXCLUSIVE, &rootvnode, td)) 1284 panic("Cannot find root vnode"); 1285 1286 p = td->td_proc; 1287 FILEDESC_LOCK(p->p_fd); 1288 1289 if (p->p_fd->fd_cdir != NULL) 1290 vrele(p->p_fd->fd_cdir); 1291 p->p_fd->fd_cdir = rootvnode; 1292 VREF(rootvnode); 1293 1294 if (p->p_fd->fd_rdir != NULL) 1295 vrele(p->p_fd->fd_rdir); 1296 p->p_fd->fd_rdir = rootvnode; 1297 VREF(rootvnode); 1298 1299 FILEDESC_UNLOCK(p->p_fd); 1300 1301 VOP_UNLOCK(rootvnode, 0, td); 1302} 1303 1304/* 1305 * Mount /devfs as our root filesystem, but do not put it on the mountlist 1306 * yet. Create a /dev -> / symlink so that absolute pathnames will lookup. 1307 */ 1308 1309static void 1310devfs_first(void) 1311{ 1312 struct thread *td = curthread; 1313 struct vfsoptlist *opts; 1314 struct vfsconf *vfsp; 1315 struct mount *mp = NULL; 1316 int error; 1317 1318 vfsp = vfs_byname("devfs"); 1319 KASSERT(vfsp != NULL, ("Could not find devfs by name")); 1320 if (vfsp == NULL) 1321 return; 1322 1323 mp = vfs_mount_alloc(NULLVP, vfsp, "/dev", td); 1324 1325 error = VFS_MOUNT(mp, td); 1326 KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error)); 1327 if (error) 1328 return; 1329 1330 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK); 1331 TAILQ_INIT(opts); 1332 mp->mnt_opt = opts; 1333 1334 mtx_lock(&mountlist_mtx); 1335 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 1336 mtx_unlock(&mountlist_mtx); 1337 1338 set_rootvnode(td); 1339 1340 error = kern_symlink(td, "/", "dev", UIO_SYSSPACE); 1341 if (error) 1342 printf("kern_symlink /dev -> / returns %d\n", error); 1343} 1344 1345/* 1346 * Surgically move our devfs to be mounted on /dev. 1347 */ 1348 1349static void 1350devfs_fixup(struct thread *td) 1351{ 1352 struct nameidata nd; 1353 int error; 1354 struct vnode *vp, *dvp; 1355 struct mount *mp; 1356 1357 /* Remove our devfs mount from the mountlist and purge the cache */ 1358 mtx_lock(&mountlist_mtx); 1359 mp = TAILQ_FIRST(&mountlist); 1360 TAILQ_REMOVE(&mountlist, mp, mnt_list); 1361 mtx_unlock(&mountlist_mtx); 1362 cache_purgevfs(mp); 1363 1364 VFS_ROOT(mp, LK_EXCLUSIVE, &dvp, td); 1365 VI_LOCK(dvp); 1366 dvp->v_iflag &= ~VI_MOUNT; 1367 dvp->v_mountedhere = NULL; 1368 VI_UNLOCK(dvp); 1369 1370 /* Set up the real rootvnode, and purge the cache */ 1371 TAILQ_FIRST(&mountlist)->mnt_vnodecovered = NULL; 1372 set_rootvnode(td); 1373 cache_purgevfs(rootvnode->v_mount); 1374 1375 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td); 1376 error = namei(&nd); 1377 if (error) { 1378 printf("Lookup of /dev for devfs, error: %d\n", error); 1379 return; 1380 } 1381 NDFREE(&nd, NDF_ONLY_PNBUF); 1382 vp = nd.ni_vp; 1383 if (vp->v_type != VDIR) { 1384 vput(vp); 1385 } 1386 error = vinvalbuf(vp, V_SAVE, td, 0, 0); 1387 if (error) { 1388 vput(vp); 1389 } 1390 cache_purge(vp); 1391 mp->mnt_vnodecovered = vp; 1392 vp->v_mountedhere = mp; 1393 mtx_lock(&mountlist_mtx); 1394 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 1395 mtx_unlock(&mountlist_mtx); 1396 VOP_UNLOCK(vp, 0, td); 1397 vput(dvp); 1398 vfs_unbusy(mp, td); 1399 1400 /* Unlink the no longer needed /dev/dev -> / symlink */ 1401 kern_unlink(td, "/dev/dev", UIO_SYSSPACE); 1402} 1403 1404/* 1405 * Report errors during filesystem mounting. 1406 */ 1407void 1408vfs_mount_error(struct mount *mp, const char *fmt, ...) 1409{ 1410 struct vfsoptlist *moptlist = mp->mnt_optnew; 1411 va_list ap; 1412 int error, len; 1413 char *errmsg; 1414 1415 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len); 1416 if (error || errmsg == NULL || len <= 0) 1417 return; 1418 1419 va_start(ap, fmt); 1420 vsnprintf(errmsg, (size_t)len, fmt, ap); 1421 va_end(ap); 1422} 1423 1424/* 1425 * Find and mount the root filesystem 1426 */ 1427void 1428vfs_mountroot(void) 1429{ 1430 char *cp; 1431 int error, i, asked = 0; 1432 1433 root_mount_wait(); 1434 1435 mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), 1436 NULL, NULL, mount_init, mount_fini, 1437 UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 1438 devfs_first(); 1439 1440 /* 1441 * We are booted with instructions to prompt for the root filesystem. 1442 */ 1443 if (boothowto & RB_ASKNAME) { 1444 if (!vfs_mountroot_ask()) 1445 return; 1446 asked = 1; 1447 } 1448 1449 /* 1450 * The root filesystem information is compiled in, and we are 1451 * booted with instructions to use it. 1452 */ 1453 if (ctrootdevname != NULL && (boothowto & RB_DFLTROOT)) { 1454 if (!vfs_mountroot_try(ctrootdevname)) 1455 return; 1456 ctrootdevname = NULL; 1457 } 1458 1459 /* 1460 * We've been given the generic "use CDROM as root" flag. This is 1461 * necessary because one media may be used in many different 1462 * devices, so we need to search for them. 1463 */ 1464 if (boothowto & RB_CDROM) { 1465 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) { 1466 if (!vfs_mountroot_try(cdrom_rootdevnames[i])) 1467 return; 1468 } 1469 } 1470 1471 /* 1472 * Try to use the value read by the loader from /etc/fstab, or 1473 * supplied via some other means. This is the preferred 1474 * mechanism. 1475 */ 1476 cp = getenv("vfs.root.mountfrom"); 1477 if (cp != NULL) { 1478 error = vfs_mountroot_try(cp); 1479 freeenv(cp); 1480 if (!error) 1481 return; 1482 } 1483 1484 /* 1485 * Try values that may have been computed by code during boot 1486 */ 1487 if (!vfs_mountroot_try(rootdevnames[0])) 1488 return; 1489 if (!vfs_mountroot_try(rootdevnames[1])) 1490 return; 1491 1492 /* 1493 * If we (still) have a compiled-in default, try it. 1494 */ 1495 if (ctrootdevname != NULL) 1496 if (!vfs_mountroot_try(ctrootdevname)) 1497 return; 1498 /* 1499 * Everything so far has failed, prompt on the console if we haven't 1500 * already tried that. 1501 */ 1502 if (!asked) 1503 if (!vfs_mountroot_ask()) 1504 return; 1505 1506 panic("Root mount failed, startup aborted."); 1507} 1508 1509/* 1510 * Mount (mountfrom) as the root filesystem. 1511 */ 1512static int 1513vfs_mountroot_try(const char *mountfrom) 1514{ 1515 struct mount *mp; 1516 char *vfsname, *path; 1517 time_t timebase; 1518 int error; 1519 char patt[32]; 1520 1521 vfsname = NULL; 1522 path = NULL; 1523 mp = NULL; 1524 error = EINVAL; 1525 1526 if (mountfrom == NULL) 1527 return (error); /* don't complain */ 1528 printf("Trying to mount root from %s\n", mountfrom); 1529 1530 /* parse vfs name and path */ 1531 vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK); 1532 path = malloc(MNAMELEN, M_MOUNT, M_WAITOK); 1533 vfsname[0] = path[0] = 0; 1534 sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN); 1535 if (sscanf(mountfrom, patt, vfsname, path) < 1) 1536 goto out; 1537 1538 if (path[0] == '\0') 1539 strcpy(path, ROOTNAME); 1540 1541 error = kernel_vmount( 1542 MNT_RDONLY | MNT_ROOTFS, 1543 "fstype", vfsname, 1544 "fspath", "/", 1545 "from", path, 1546 NULL); 1547 if (error == 0) { 1548 /* 1549 * We mount devfs prior to mounting the / FS, so the first 1550 * entry will typically be devfs. 1551 */ 1552 mp = TAILQ_FIRST(&mountlist); 1553 KASSERT(mp != NULL, ("%s: mountlist is empty", __func__)); 1554 1555 /* 1556 * Iterate over all currently mounted file systems and use 1557 * the time stamp found to check and/or initialize the RTC. 1558 * Typically devfs has no time stamp and the only other FS 1559 * is the actual / FS. 1560 * Call inittodr() only once and pass it the largest of the 1561 * timestamps we encounter. 1562 */ 1563 timebase = 0; 1564 do { 1565 if (mp->mnt_time > timebase) 1566 timebase = mp->mnt_time; 1567 mp = TAILQ_NEXT(mp, mnt_list); 1568 } while (mp != NULL); 1569 inittodr(timebase); 1570 1571 devfs_fixup(curthread); 1572 } 1573out: 1574 free(path, M_MOUNT); 1575 free(vfsname, M_MOUNT); 1576 return (error); 1577} 1578 1579/* 1580 * --------------------------------------------------------------------- 1581 * Interactive root filesystem selection code. 1582 */ 1583 1584static int 1585vfs_mountroot_ask(void) 1586{ 1587 char name[128]; 1588 1589 for(;;) { 1590 printf("\nManual root filesystem specification:\n"); 1591 printf(" <fstype>:<device> Mount <device> using filesystem <fstype>\n"); 1592#if defined(__amd64__) || defined(__i386__) || defined(__ia64__) 1593 printf(" eg. ufs:da0s1a\n"); 1594#else 1595 printf(" eg. ufs:/dev/da0a\n"); 1596#endif 1597 printf(" ? List valid disk boot devices\n"); 1598 printf(" <empty line> Abort manual input\n"); 1599 printf("\nmountroot> "); 1600 gets(name, sizeof(name), 1); 1601 if (name[0] == '\0') 1602 return (1); 1603 if (name[0] == '?') { 1604 printf("\nList of GEOM managed disk devices:\n "); 1605 g_dev_print(); 1606 continue; 1607 } 1608 if (!vfs_mountroot_try(name)) 1609 return (0); 1610 } 1611} 1612 1613/* 1614 * --------------------------------------------------------------------- 1615 * Functions for querying mount options/arguments from filesystems. 1616 */ 1617 1618/* 1619 * Check that no unknown options are given 1620 */ 1621int 1622vfs_filteropt(struct vfsoptlist *opts, const char **legal) 1623{ 1624 struct vfsopt *opt; 1625 const char **t, *p; 1626 1627 1628 TAILQ_FOREACH(opt, opts, link) { 1629 p = opt->name; 1630 if (p[0] == 'n' && p[1] == 'o') 1631 p += 2; 1632 for(t = global_opts; *t != NULL; t++) 1633 if (!strcmp(*t, p)) 1634 break; 1635 if (*t != NULL) 1636 continue; 1637 for(t = legal; *t != NULL; t++) 1638 if (!strcmp(*t, p)) 1639 break; 1640 if (*t != NULL) 1641 continue; 1642 printf("mount option <%s> is unknown\n", p); 1643 return (EINVAL); 1644 } 1645 return (0); 1646} 1647 1648/* 1649 * Get a mount option by its name. 1650 * 1651 * Return 0 if the option was found, ENOENT otherwise. 1652 * If len is non-NULL it will be filled with the length 1653 * of the option. If buf is non-NULL, it will be filled 1654 * with the address of the option. 1655 */ 1656int 1657vfs_getopt(opts, name, buf, len) 1658 struct vfsoptlist *opts; 1659 const char *name; 1660 void **buf; 1661 int *len; 1662{ 1663 struct vfsopt *opt; 1664 1665 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1666 1667 TAILQ_FOREACH(opt, opts, link) { 1668 if (strcmp(name, opt->name) == 0) { 1669 if (len != NULL) 1670 *len = opt->len; 1671 if (buf != NULL) 1672 *buf = opt->value; 1673 return (0); 1674 } 1675 } 1676 return (ENOENT); 1677} 1678 1679static int 1680vfs_getopt_pos(struct vfsoptlist *opts, const char *name) 1681{ 1682 struct vfsopt *opt; 1683 int i; 1684 1685 if (opts == NULL) 1686 return (-1); 1687 1688 i = 0; 1689 TAILQ_FOREACH(opt, opts, link) { 1690 if (strcmp(name, opt->name) == 0) 1691 return (i); 1692 ++i; 1693 } 1694 return (-1); 1695} 1696 1697char * 1698vfs_getopts(struct vfsoptlist *opts, const char *name, int *error) 1699{ 1700 struct vfsopt *opt; 1701 1702 *error = 0; 1703 TAILQ_FOREACH(opt, opts, link) { 1704 if (strcmp(name, opt->name) != 0) 1705 continue; 1706 if (((char *)opt->value)[opt->len - 1] != '\0') { 1707 *error = EINVAL; 1708 return (NULL); 1709 } 1710 return (opt->value); 1711 } 1712 return (NULL); 1713} 1714 1715int 1716vfs_flagopt(struct vfsoptlist *opts, const char *name, u_int *w, u_int val) 1717{ 1718 struct vfsopt *opt; 1719 1720 TAILQ_FOREACH(opt, opts, link) { 1721 if (strcmp(name, opt->name) == 0) { 1722 if (w != NULL) 1723 *w |= val; 1724 return (1); 1725 } 1726 } 1727 if (w != NULL) 1728 *w &= ~val; 1729 return (0); 1730} 1731 1732int 1733vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...) 1734{ 1735 va_list ap; 1736 struct vfsopt *opt; 1737 int ret; 1738 1739 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL")); 1740 1741 TAILQ_FOREACH(opt, opts, link) { 1742 if (strcmp(name, opt->name) != 0) 1743 continue; 1744 if (((char *)opt->value)[opt->len - 1] != '\0') 1745 return (0); 1746 va_start(ap, fmt); 1747 ret = vsscanf(opt->value, fmt, ap); 1748 va_end(ap); 1749 return (ret); 1750 } 1751 return (0); 1752} 1753 1754/* 1755 * Find and copy a mount option. 1756 * 1757 * The size of the buffer has to be specified 1758 * in len, if it is not the same length as the 1759 * mount option, EINVAL is returned. 1760 * Returns ENOENT if the option is not found. 1761 */ 1762int 1763vfs_copyopt(opts, name, dest, len) 1764 struct vfsoptlist *opts; 1765 const char *name; 1766 void *dest; 1767 int len; 1768{ 1769 struct vfsopt *opt; 1770 1771 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL")); 1772 1773 TAILQ_FOREACH(opt, opts, link) { 1774 if (strcmp(name, opt->name) == 0) { 1775 if (len != opt->len) 1776 return (EINVAL); 1777 bcopy(opt->value, dest, opt->len); 1778 return (0); 1779 } 1780 } 1781 return (ENOENT); 1782} 1783 1784/* 1785 * This is a helper function for filesystems to traverse their 1786 * vnodes. See MNT_VNODE_FOREACH() in sys/mount.h 1787 */ 1788 1789struct vnode * 1790__mnt_vnode_next(struct vnode **mvp, struct mount *mp) 1791{ 1792 struct vnode *vp; 1793 1794 mtx_assert(MNT_MTX(mp), MA_OWNED); 1795 1796 KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch")); 1797 vp = TAILQ_NEXT(*mvp, v_nmntvnodes); 1798 while (vp != NULL && vp->v_type == VMARKER) 1799 vp = TAILQ_NEXT(vp, v_nmntvnodes); 1800 1801 /* Check if we are done */ 1802 if (vp == NULL) { 1803 __mnt_vnode_markerfree(mvp, mp); 1804 return (NULL); 1805 } 1806 TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes); 1807 TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes); 1808 return (vp); 1809} 1810 1811struct vnode * 1812__mnt_vnode_first(struct vnode **mvp, struct mount *mp) 1813{ 1814 struct vnode *vp; 1815 1816 mtx_assert(MNT_MTX(mp), MA_OWNED); 1817 1818 vp = TAILQ_FIRST(&mp->mnt_nvnodelist); 1819 while (vp != NULL && vp->v_type == VMARKER) 1820 vp = TAILQ_NEXT(vp, v_nmntvnodes); 1821 1822 /* Check if we are done */ 1823 if (vp == NULL) { 1824 *mvp = NULL; 1825 return (NULL); 1826 } 1827 mp->mnt_holdcnt++; 1828 MNT_IUNLOCK(mp); 1829 *mvp = (struct vnode *) malloc(sizeof(struct vnode), 1830 M_VNODE_MARKER, 1831 M_WAITOK | M_ZERO); 1832 MNT_ILOCK(mp); 1833 (*mvp)->v_type = VMARKER; 1834 1835 vp = TAILQ_FIRST(&mp->mnt_nvnodelist); 1836 while (vp != NULL && vp->v_type == VMARKER) 1837 vp = TAILQ_NEXT(vp, v_nmntvnodes); 1838 1839 /* Check if we are done */ 1840 if (vp == NULL) { 1841 MNT_IUNLOCK(mp); 1842 free(*mvp, M_VNODE_MARKER); 1843 MNT_ILOCK(mp); 1844 *mvp = NULL; 1845 mp->mnt_holdcnt--; 1846 if (mp->mnt_holdcnt == 0 && mp->mnt_holdcntwaiters != 0) 1847 wakeup(&mp->mnt_holdcnt); 1848 return (NULL); 1849 } 1850 mp->mnt_markercnt++; 1851 (*mvp)->v_mount = mp; 1852 TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes); 1853 return (vp); 1854} 1855 1856 1857void 1858__mnt_vnode_markerfree(struct vnode **mvp, struct mount *mp) 1859{ 1860 1861 if (*mvp == NULL) 1862 return; 1863 1864 mtx_assert(MNT_MTX(mp), MA_OWNED); 1865 1866 KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch")); 1867 TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes); 1868 MNT_IUNLOCK(mp); 1869 free(*mvp, M_VNODE_MARKER); 1870 MNT_ILOCK(mp); 1871 *mvp = NULL; 1872 1873 mp->mnt_markercnt--; 1874 mp->mnt_holdcnt--; 1875 if (mp->mnt_holdcnt == 0 && mp->mnt_holdcntwaiters != 0) 1876 wakeup(&mp->mnt_holdcnt); 1877} 1878 1879 1880int 1881__vfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td) 1882{ 1883 int error; 1884 1885 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat, td); 1886 if (sbp != &mp->mnt_stat) 1887 *sbp = mp->mnt_stat; 1888 return (error); 1889} 1890 1891void 1892vfs_mountedfrom(struct mount *mp, const char *from) 1893{ 1894 1895 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname); 1896 strlcpy(mp->mnt_stat.f_mntfromname, from, 1897 sizeof mp->mnt_stat.f_mntfromname); 1898} 1899 1900/* 1901 * --------------------------------------------------------------------- 1902 * This is the api for building mount args and mounting filesystems from 1903 * inside the kernel. 1904 * 1905 * The API works by accumulation of individual args. First error is 1906 * latched. 1907 * 1908 * XXX: should be documented in new manpage kernel_mount(9) 1909 */ 1910 1911/* A memory allocation which must be freed when we are done */ 1912struct mntaarg { 1913 SLIST_ENTRY(mntaarg) next; 1914}; 1915 1916/* The header for the mount arguments */ 1917struct mntarg { 1918 struct iovec *v; 1919 int len; 1920 int error; 1921 SLIST_HEAD(, mntaarg) list; 1922}; 1923 1924/* 1925 * Add a boolean argument. 1926 * 1927 * flag is the boolean value. 1928 * name must start with "no". 1929 */ 1930struct mntarg * 1931mount_argb(struct mntarg *ma, int flag, const char *name) 1932{ 1933 1934 KASSERT(name[0] == 'n' && name[1] == 'o', 1935 ("mount_argb(...,%s): name must start with 'no'", name)); 1936 1937 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0)); 1938} 1939 1940/* 1941 * Add an argument printf style 1942 */ 1943struct mntarg * 1944mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...) 1945{ 1946 va_list ap; 1947 struct mntaarg *maa; 1948 struct sbuf *sb; 1949 int len; 1950 1951 if (ma == NULL) { 1952 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1953 SLIST_INIT(&ma->list); 1954 } 1955 if (ma->error) 1956 return (ma); 1957 1958 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 1959 M_MOUNT, M_WAITOK); 1960 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 1961 ma->v[ma->len].iov_len = strlen(name) + 1; 1962 ma->len++; 1963 1964 sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND); 1965 va_start(ap, fmt); 1966 sbuf_vprintf(sb, fmt, ap); 1967 va_end(ap); 1968 sbuf_finish(sb); 1969 len = sbuf_len(sb) + 1; 1970 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 1971 SLIST_INSERT_HEAD(&ma->list, maa, next); 1972 bcopy(sbuf_data(sb), maa + 1, len); 1973 sbuf_delete(sb); 1974 1975 ma->v[ma->len].iov_base = maa + 1; 1976 ma->v[ma->len].iov_len = len; 1977 ma->len++; 1978 1979 return (ma); 1980} 1981 1982/* 1983 * Add an argument which is a userland string. 1984 */ 1985struct mntarg * 1986mount_argsu(struct mntarg *ma, const char *name, const void *val, int len) 1987{ 1988 struct mntaarg *maa; 1989 char *tbuf; 1990 1991 if (val == NULL) 1992 return (ma); 1993 if (ma == NULL) { 1994 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 1995 SLIST_INIT(&ma->list); 1996 } 1997 if (ma->error) 1998 return (ma); 1999 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO); 2000 SLIST_INSERT_HEAD(&ma->list, maa, next); 2001 tbuf = (void *)(maa + 1); 2002 ma->error = copyinstr(val, tbuf, len, NULL); 2003 return (mount_arg(ma, name, tbuf, -1)); 2004} 2005 2006/* 2007 * Plain argument. 2008 * 2009 * If length is -1, use printf. 2010 */ 2011struct mntarg * 2012mount_arg(struct mntarg *ma, const char *name, const void *val, int len) 2013{ 2014 2015 if (ma == NULL) { 2016 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO); 2017 SLIST_INIT(&ma->list); 2018 } 2019 if (ma->error) 2020 return (ma); 2021 2022 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2), 2023 M_MOUNT, M_WAITOK); 2024 ma->v[ma->len].iov_base = (void *)(uintptr_t)name; 2025 ma->v[ma->len].iov_len = strlen(name) + 1; 2026 ma->len++; 2027 2028 ma->v[ma->len].iov_base = (void *)(uintptr_t)val; 2029 if (len < 0) 2030 ma->v[ma->len].iov_len = strlen(val) + 1; 2031 else 2032 ma->v[ma->len].iov_len = len; 2033 ma->len++; 2034 return (ma); 2035} 2036 2037/* 2038 * Free a mntarg structure 2039 */ 2040static void 2041free_mntarg(struct mntarg *ma) 2042{ 2043 struct mntaarg *maa; 2044 2045 while (!SLIST_EMPTY(&ma->list)) { 2046 maa = SLIST_FIRST(&ma->list); 2047 SLIST_REMOVE_HEAD(&ma->list, next); 2048 free(maa, M_MOUNT); 2049 } 2050 free(ma->v, M_MOUNT); 2051 free(ma, M_MOUNT); 2052} 2053 2054/* 2055 * Mount a filesystem 2056 */ 2057int 2058kernel_mount(struct mntarg *ma, int flags) 2059{ 2060 struct uio auio; 2061 int error; 2062 2063 KASSERT(ma != NULL, ("kernel_mount NULL ma")); 2064 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v")); 2065 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len)); 2066 2067 auio.uio_iov = ma->v; 2068 auio.uio_iovcnt = ma->len; 2069 auio.uio_segflg = UIO_SYSSPACE; 2070 2071 error = ma->error; 2072 if (!error) 2073 error = vfs_donmount(curthread, flags, &auio); 2074 free_mntarg(ma); 2075 return (error); 2076} 2077 2078/* 2079 * A printflike function to mount a filesystem. 2080 */ 2081int 2082kernel_vmount(int flags, ...) 2083{ 2084 struct mntarg *ma = NULL; 2085 va_list ap; 2086 const char *cp; 2087 const void *vp; 2088 int error; 2089 2090 va_start(ap, flags); 2091 for (;;) { 2092 cp = va_arg(ap, const char *); 2093 if (cp == NULL) 2094 break; 2095 vp = va_arg(ap, const void *); 2096 ma = mount_arg(ma, cp, vp, -1); 2097 } 2098 va_end(ap); 2099 2100 error = kernel_mount(ma, flags); 2101 return (error); 2102} 2103