sys/kern/vfs_syscalls.c

   1 /*
   2  * Copyright (c) 1989, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  * (c) UNIX System Laboratories, Inc.
   5  * All or some portions of this file are derived from material licensed
   6  * to the University of California by American Telephone and Telegraph
   7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
   8  * the permission of UNIX System Laboratories, Inc.
   9  *
  10  * Redistribution and use in source and binary forms, with or without
  11  * modification, are permitted provided that the following conditions
  12  * are met:
  13  * 1. Redistributions of source code must retain the above copyright
  14  *    notice, this list of conditions and the following disclaimer.
  15  * 2. Redistributions in binary form must reproduce the above copyright
  16  *    notice, this list of conditions and the following disclaimer in the
  17  *    documentation and/or other materials provided with the distribution.
  18  * 3. All advertising materials mentioning features or use of this software
  19  *    must display the following acknowledgement:
  20  *      This product includes software developed by the University of
  21  *      California, Berkeley and its contributors.
  22  * 4. Neither the name of the University nor the names of its contributors
  23  *    may be used to endorse or promote products derived from this software
  24  *    without specific prior written permission.
  25  *
  26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  36  * SUCH DAMAGE.
  37  *
  38  *      @(#)vfs_syscalls.c      8.13 (Berkeley) 4/15/94
  39  * $FreeBSD: src/sys/kern/vfs_syscalls.c,v 1.151.2.18 2003/04/04 20:35:58 tegge Exp $
  40  * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.135 2008/11/11 00:55:49 pavalos Exp $
  41  */
  42
  43 #include <sys/param.h>
  44 #include <sys/systm.h>
  45 #include <sys/buf.h>
  46 #include <sys/conf.h>
  47 #include <sys/sysent.h>
  48 #include <sys/malloc.h>
  49 #include <sys/mount.h>
  50 #include <sys/mountctl.h>
  51 #include <sys/sysproto.h>
  52 #include <sys/filedesc.h>
  53 #include <sys/kernel.h>
  54 #include <sys/fcntl.h>
  55 #include <sys/file.h>
  56 #include <sys/linker.h>
  57 #include <sys/stat.h>
  58 #include <sys/unistd.h>
  59 #include <sys/vnode.h>
  60 #include <sys/proc.h>
  61 #include <sys/priv.h>
  62 #include <sys/namei.h>
  63 #include <sys/nlookup.h>
  64 #include <sys/dirent.h>
  65 #include <sys/extattr.h>
  66 #include <sys/spinlock.h>
  67 #include <sys/kern_syscall.h>
  68 #include <sys/objcache.h>
  69 #include <sys/sysctl.h>
  70
  71 #include <sys/buf2.h>
  72 #include <sys/file2.h>
  73 #include <sys/spinlock2.h>
  74
  75 #include <vm/vm.h>
  76 #include <vm/vm_object.h>
  77 #include <vm/vm_page.h>
  78
  79 #include <machine/limits.h>
  80 #include <machine/stdarg.h>
  81
  82 #include <vfs/union/union.h>
  83
  84 static void mount_warning(struct mount *mp, const char *ctl, ...);
  85 static int mount_path(struct proc *p, struct mount *mp, char **rb, char **fb);
  86 static int checkvp_chdir (struct vnode *vn, struct thread *td);
  87 static void checkdirs (struct nchandle *old_nch, struct nchandle *new_nch);
  88 static int chroot_refuse_vdir_fds (struct filedesc *fdp);
  89 static int chroot_visible_mnt(struct mount *mp, struct proc *p);
  90 static int getutimes (const struct timeval *, struct timespec *);
  91 static int setfown (struct vnode *, uid_t, gid_t);
  92 static int setfmode (struct vnode *, int);
  93 static int setfflags (struct vnode *, int);
  94 static int setutimes (struct vnode *, const struct timespec *, int);
  95 static int      usermount = 0;  /* if 1, non-root can mount fs. */
  96
  97 int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
  98
  99 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
 100
 101 /*
 102  * Virtual File System System Calls
 103  */
 104
 105 /*
 106  * Mount a file system.
 107  */
 108 /*
 109  * mount_args(char *type, char *path, int flags, caddr_t data)
 110  */
 111 /* ARGSUSED */
 112 int
 113 sys_mount(struct mount_args *uap)
 114 {
 115         struct thread *td = curthread;
 116         struct proc *p = td->td_proc;
 117         struct vnode *vp;
 118         struct nchandle nch;
 119         struct mount *mp;
 120         struct vfsconf *vfsp;
 121         int error, flag = 0, flag2 = 0;
 122         int hasmount;
 123         struct vattr va;
 124         struct nlookupdata nd;
 125         char fstypename[MFSNAMELEN];
 126         struct ucred *cred = p->p_ucred;
 127
 128         KKASSERT(p);
 129         if (jailed(cred))
 130                 return (EPERM);
 131         if (usermount == 0 && (error = priv_check(td, PRIV_ROOT)))
 132                 return (error);
 133         /*
 134          * Do not allow NFS export by non-root users.
 135          */
 136         if (uap->flags & MNT_EXPORTED) {
 137                 error = priv_check(td, PRIV_ROOT);
 138                 if (error)
 139                         return (error);
 140         }
 141         /*
 142          * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
 143          */
 144         if (priv_check(td, PRIV_ROOT))
 145                 uap->flags |= MNT_NOSUID | MNT_NODEV;
 146
 147         /*
 148          * Lookup the requested path and extract the nch and vnode.
 149          */
 150         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
 151         if (error == 0) {
 152                 if ((error = nlookup(&nd)) == 0) {
 153                         if (nd.nl_nch.ncp->nc_vp == NULL)
 154                                 error = ENOENT;
 155                 }
 156         }
 157         if (error) {
 158                 nlookup_done(&nd);
 159                 return (error);
 160         }
 161
 162         /*
 163          * Extract the locked+refd ncp and cleanup the nd structure
 164          */
 165         nch = nd.nl_nch;
 166         cache_zero(&nd.nl_nch);
 167         nlookup_done(&nd);
 168
 169         if ((nch.ncp->nc_flag & NCF_ISMOUNTPT) && cache_findmount(&nch))
 170                 hasmount = 1;
 171         else
 172                 hasmount = 0;
 173
 174
 175         /*
 176          * now we have the locked ref'd nch and unreferenced vnode.
 177          */
 178         vp = nch.ncp->nc_vp;
 179         if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
 180                 cache_put(&nch);
 181                 return (error);
 182         }
 183         cache_unlock(&nch);
 184
 185         /*
 186          * Now we have an unlocked ref'd nch and a locked ref'd vp
 187          */
 188         if (uap->flags & MNT_UPDATE) {
 189                 if ((vp->v_flag & (VROOT|VPFSROOT)) == 0) {
 190                         cache_drop(&nch);
 191                         vput(vp);
 192                         return (EINVAL);
 193                 }
 194                 mp = vp->v_mount;
 195                 flag = mp->mnt_flag;
 196                 flag2 = mp->mnt_kern_flag;
 197                 /*
 198                  * We only allow the filesystem to be reloaded if it
 199                  * is currently mounted read-only.
 200                  */
 201                 if ((uap->flags & MNT_RELOAD) &&
 202                     ((mp->mnt_flag & MNT_RDONLY) == 0)) {
 203                         cache_drop(&nch);
 204                         vput(vp);
 205                         return (EOPNOTSUPP);    /* Needs translation */
 206                 }
 207                 /*
 208                  * Only root, or the user that did the original mount is
 209                  * permitted to update it.
 210                  */
 211                 if (mp->mnt_stat.f_owner != cred->cr_uid &&
 212                     (error = priv_check(td, PRIV_ROOT))) {
 213                         cache_drop(&nch);
 214                         vput(vp);
 215                         return (error);
 216                 }
 217                 if (vfs_busy(mp, LK_NOWAIT)) {
 218                         cache_drop(&nch);
 219                         vput(vp);
 220                         return (EBUSY);
 221                 }
 222                 if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
 223                         cache_drop(&nch);
 224                         vfs_unbusy(mp);
 225                         vput(vp);
 226                         return (EBUSY);
 227                 }
 228                 vp->v_flag |= VMOUNT;
 229                 mp->mnt_flag |=
 230                     uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
 231                 vn_unlock(vp);
 232                 goto update;
 233         }
 234         /*
 235          * If the user is not root, ensure that they own the directory
 236          * onto which we are attempting to mount.
 237          */
 238         if ((error = VOP_GETATTR(vp, &va)) ||
 239             (va.va_uid != cred->cr_uid && (error = priv_check(td, PRIV_ROOT)))) {
 240                 cache_drop(&nch);
 241                 vput(vp);
 242                 return (error);
 243         }
 244         if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
 245                 cache_drop(&nch);
 246                 vput(vp);
 247                 return (error);
 248         }
 249         if (vp->v_type != VDIR) {
 250                 cache_drop(&nch);
 251                 vput(vp);
 252                 return (ENOTDIR);
 253         }
 254         if (vp->v_mount->mnt_kern_flag & MNTK_NOSTKMNT) {
 255                 cache_drop(&nch);
 256                 vput(vp);
 257                 return (EPERM);
 258         }
 259         if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
 260                 cache_drop(&nch);
 261                 vput(vp);
 262                 return (error);
 263         }
 264         vfsp = vfsconf_find_by_name(fstypename);
 265         if (vfsp == NULL) {
 266                 linker_file_t lf;
 267
 268                 /* Only load modules for root (very important!) */
 269                 if ((error = priv_check(td, PRIV_ROOT)) != 0) {
 270                         cache_drop(&nch);
 271                         vput(vp);
 272                         return error;
 273                 }
 274                 error = linker_load_file(fstypename, &lf);
 275                 if (error || lf == NULL) {
 276                         cache_drop(&nch);
 277                         vput(vp);
 278                         if (lf == NULL)
 279                                 error = ENODEV;
 280                         return error;
 281                 }
 282                 lf->userrefs++;
 283                 /* lookup again, see if the VFS was loaded */
 284                 vfsp = vfsconf_find_by_name(fstypename);
 285                 if (vfsp == NULL) {
 286                         lf->userrefs--;
 287                         linker_file_unload(lf);
 288                         cache_drop(&nch);
 289                         vput(vp);
 290                         return (ENODEV);
 291                 }
 292         }
 293         if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
 294                 cache_drop(&nch);
 295                 vput(vp);
 296                 return (EBUSY);
 297         }
 298         vp->v_flag |= VMOUNT;
 299
 300         /*
 301          * Allocate and initialize the filesystem.
 302          */
 303         mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
 304         TAILQ_INIT(&mp->mnt_nvnodelist);
 305         TAILQ_INIT(&mp->mnt_reservedvnlist);
 306         TAILQ_INIT(&mp->mnt_jlist);
 307         mp->mnt_nvnodelistsize = 0;
 308         lockinit(&mp->mnt_lock, "vfslock", 0, 0);
 309         vfs_busy(mp, LK_NOWAIT);
 310         mp->mnt_op = vfsp->vfc_vfsops;
 311         mp->mnt_vfc = vfsp;
 312         vfsp->vfc_refcount++;
 313         mp->mnt_stat.f_type = vfsp->vfc_typenum;
 314         mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
 315         strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
 316         mp->mnt_stat.f_owner = cred->cr_uid;
 317         mp->mnt_iosize_max = DFLTPHYS;
 318         vn_unlock(vp);
 319 update:
 320         /*
 321          * Set the mount level flags.
 322          */
 323         if (uap->flags & MNT_RDONLY)
 324                 mp->mnt_flag |= MNT_RDONLY;
 325         else if (mp->mnt_flag & MNT_RDONLY)
 326                 mp->mnt_kern_flag |= MNTK_WANTRDWR;
 327         mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
 328             MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
 329             MNT_NOSYMFOLLOW | MNT_IGNORE |
 330             MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
 331         mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
 332             MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
 333             MNT_NOSYMFOLLOW | MNT_IGNORE |
 334             MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
 335         /*
 336          * Mount the filesystem.
 337          * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
 338          * get.
 339          */
 340         error = VFS_MOUNT(mp, uap->path, uap->data, cred);
 341         if (mp->mnt_flag & MNT_UPDATE) {
 342                 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
 343                         mp->mnt_flag &= ~MNT_RDONLY;
 344                 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
 345                 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
 346                 if (error) {
 347                         mp->mnt_flag = flag;
 348                         mp->mnt_kern_flag = flag2;
 349                 }
 350                 vfs_unbusy(mp);
 351                 vp->v_flag &= ~VMOUNT;
 352                 vrele(vp);
 353                 cache_drop(&nch);
 354                 return (error);
 355         }
 356         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 357         /*
 358          * Put the new filesystem on the mount list after root.  The mount
 359          * point gets its own mnt_ncmountpt (unless the VFS already set one
 360          * up) which represents the root of the mount.  The lookup code
 361          * detects the mount point going forward and checks the root of
 362          * the mount going backwards.
 363          *
 364          * It is not necessary to invalidate or purge the vnode underneath
 365          * because elements under the mount will be given their own glue
 366          * namecache record.
 367          */
 368         if (!error) {
 369                 if (mp->mnt_ncmountpt.ncp == NULL) {
 370                         /*
 371                          * allocate, then unlock, but leave the ref intact
 372                          */
 373                         cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
 374                         cache_unlock(&mp->mnt_ncmountpt);
 375                 }
 376                 mp->mnt_ncmounton = nch;                /* inherits ref */
 377                 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
 378
 379                 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
 380                 vp->v_flag &= ~VMOUNT;
 381                 mountlist_insert(mp, MNTINS_LAST);
 382                 checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
 383                 vn_unlock(vp);
 384                 error = vfs_allocate_syncvnode(mp);
 385                 vfs_unbusy(mp);
 386                 error = VFS_START(mp, 0);
 387                 vrele(vp);
 388         } else {
 389                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
 390                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
 391                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
 392                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
 393                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
 394                 vp->v_flag &= ~VMOUNT;
 395                 mp->mnt_vfc->vfc_refcount--;
 396                 vfs_unbusy(mp);
 397                 kfree(mp, M_MOUNT);
 398                 cache_drop(&nch);
 399                 vput(vp);
 400         }
 401         return (error);
 402 }
 403
 404 /*
 405  * Scan all active processes to see if any of them have a current
 406  * or root directory onto which the new filesystem has just been
 407  * mounted. If so, replace them with the new mount point.
 408  *
 409  * The passed ncp is ref'd and locked (from the mount code) and
 410  * must be associated with the vnode representing the root of the
 411  * mount point.
 412  */
 413 struct checkdirs_info {
 414         struct nchandle old_nch;
 415         struct nchandle new_nch;
 416         struct vnode *old_vp;
 417         struct vnode *new_vp;
 418 };
 419
 420 static int checkdirs_callback(struct proc *p, void *data);
 421
 422 static void
 423 checkdirs(struct nchandle *old_nch, struct nchandle *new_nch)
 424 {
 425         struct checkdirs_info info;
 426         struct vnode *olddp;
 427         struct vnode *newdp;
 428         struct mount *mp;
 429
 430         /*
 431          * If the old mount point's vnode has a usecount of 1, it is not
 432          * being held as a descriptor anywhere.
 433          */
 434         olddp = old_nch->ncp->nc_vp;
 435         if (olddp == NULL || olddp->v_sysref.refcnt == 1)
 436                 return;
 437
 438         /*
 439          * Force the root vnode of the new mount point to be resolved
 440          * so we can update any matching processes.
 441          */
 442         mp = new_nch->mount;
 443         if (VFS_ROOT(mp, &newdp))
 444                 panic("mount: lost mount");
 445         cache_setunresolved(new_nch);
 446         cache_setvp(new_nch, newdp);
 447
 448         /*
 449          * Special handling of the root node
 450          */
 451         if (rootvnode == olddp) {
 452                 vref(newdp);
 453                 vfs_cache_setroot(newdp, cache_hold(new_nch));
 454         }
 455
 456         /*
 457          * Pass newdp separately so the callback does not have to access
 458          * it via new_nch->ncp->nc_vp.
 459          */
 460         info.old_nch = *old_nch;
 461         info.new_nch = *new_nch;
 462         info.new_vp = newdp;
 463         allproc_scan(checkdirs_callback, &info);
 464         vput(newdp);
 465 }
 466
 467 /*
 468  * NOTE: callback is not MP safe because the scanned process's filedesc
 469  * structure can be ripped out from under us, amoung other things.
 470  */
 471 static int
 472 checkdirs_callback(struct proc *p, void *data)
 473 {
 474         struct checkdirs_info *info = data;
 475         struct filedesc *fdp;
 476         struct nchandle ncdrop1;
 477         struct nchandle ncdrop2;
 478         struct vnode *vprele1;
 479         struct vnode *vprele2;
 480
 481         if ((fdp = p->p_fd) != NULL) {
 482                 cache_zero(&ncdrop1);
 483                 cache_zero(&ncdrop2);
 484                 vprele1 = NULL;
 485                 vprele2 = NULL;
 486
 487                 /*
 488                  * MPUNSAFE - XXX fdp can be pulled out from under a
 489                  * foreign process.
 490                  *
 491                  * A shared filedesc is ok, we don't have to copy it
 492                  * because we are making this change globally.
 493                  */
 494                 spin_lock_wr(&fdp->fd_spin);
 495                 if (fdp->fd_ncdir.mount == info->old_nch.mount &&
 496                     fdp->fd_ncdir.ncp == info->old_nch.ncp) {
 497                         vprele1 = fdp->fd_cdir;
 498                         vref(info->new_vp);
 499                         fdp->fd_cdir = info->new_vp;
 500                         ncdrop1 = fdp->fd_ncdir;
 501                         cache_copy(&info->new_nch, &fdp->fd_ncdir);
 502                 }
 503                 if (fdp->fd_nrdir.mount == info->old_nch.mount &&
 504                     fdp->fd_nrdir.ncp == info->old_nch.ncp) {
 505                         vprele2 = fdp->fd_rdir;
 506                         vref(info->new_vp);
 507                         fdp->fd_rdir = info->new_vp;
 508                         ncdrop2 = fdp->fd_nrdir;
 509                         cache_copy(&info->new_nch, &fdp->fd_nrdir);
 510                 }
 511                 spin_unlock_wr(&fdp->fd_spin);
 512                 if (ncdrop1.ncp)
 513                         cache_drop(&ncdrop1);
 514                 if (ncdrop2.ncp)
 515                         cache_drop(&ncdrop2);
 516                 if (vprele1)
 517                         vrele(vprele1);
 518                 if (vprele2)
 519                         vrele(vprele2);
 520         }
 521         return(0);
 522 }
 523
 524 /*
 525  * Unmount a file system.
 526  *
 527  * Note: unmount takes a path to the vnode mounted on as argument,
 528  * not special file (as before).
 529  */
 530 /*
 531  * umount_args(char *path, int flags)
 532  */
 533 /* ARGSUSED */
 534 int
 535 sys_unmount(struct unmount_args *uap)
 536 {
 537         struct thread *td = curthread;
 538         struct proc *p = td->td_proc;
 539         struct mount *mp = NULL;
 540         int error;
 541         struct nlookupdata nd;
 542
 543         KKASSERT(p);
 544         if (p->p_ucred->cr_prison != NULL)
 545                 return (EPERM);
 546         if (usermount == 0 && (error = priv_check(td, PRIV_ROOT)))
 547                 return (error);
 548
 549         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
 550         if (error == 0)
 551                 error = nlookup(&nd);
 552         if (error)
 553                 goto out;
 554
 555         mp = nd.nl_nch.mount;
 556
 557         /*
 558          * Only root, or the user that did the original mount is
 559          * permitted to unmount this filesystem.
 560          */
 561         if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
 562             (error = priv_check(td, PRIV_ROOT)))
 563                 goto out;
 564
 565         /*
 566          * Don't allow unmounting the root file system.
 567          */
 568         if (mp->mnt_flag & MNT_ROOTFS) {
 569                 error = EINVAL;
 570                 goto out;
 571         }
 572
 573         /*
 574          * Must be the root of the filesystem
 575          */
 576         if (nd.nl_nch.ncp != mp->mnt_ncmountpt.ncp) {
 577                 error = EINVAL;
 578                 goto out;
 579         }
 580
 581 out:
 582         nlookup_done(&nd);
 583         if (error)
 584                 return (error);
 585         return (dounmount(mp, uap->flags));
 586 }
 587
 588 /*
 589  * Do the actual file system unmount.
 590  */
 591 static int
 592 dounmount_interlock(struct mount *mp)
 593 {
 594         if (mp->mnt_kern_flag & MNTK_UNMOUNT)
 595                 return (EBUSY);
 596         mp->mnt_kern_flag |= MNTK_UNMOUNT;
 597         return(0);
 598 }
 599
 600 int
 601 dounmount(struct mount *mp, int flags)
 602 {
 603         struct namecache *ncp;
 604         struct nchandle nch;
 605         struct vnode *vp;
 606         int error;
 607         int async_flag;
 608         int lflags;
 609         int freeok = 1;
 610
 611         /*
 612          * Exclusive access for unmounting purposes
 613          */
 614         if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
 615                 return (error);
 616
 617         /*
 618          * Allow filesystems to detect that a forced unmount is in progress.
 619          */
 620         if (flags & MNT_FORCE)
 621                 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
 622         lflags = LK_EXCLUSIVE | ((flags & MNT_FORCE) ? 0 : LK_NOWAIT);
 623         error = lockmgr(&mp->mnt_lock, lflags);
 624         if (error) {
 625                 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
 626                 if (mp->mnt_kern_flag & MNTK_MWAIT)
 627                         wakeup(mp);
 628                 return (error);
 629         }
 630
 631         if (mp->mnt_flag & MNT_EXPUBLIC)
 632                 vfs_setpublicfs(NULL, NULL, NULL);
 633
 634         vfs_msync(mp, MNT_WAIT);
 635         async_flag = mp->mnt_flag & MNT_ASYNC;
 636         mp->mnt_flag &=~ MNT_ASYNC;
 637
 638         /*
 639          * If this filesystem isn't aliasing other filesystems,
 640          * try to invalidate any remaining namecache entries and
 641          * check the count afterwords.
 642          */
 643         if ((mp->mnt_kern_flag & MNTK_NCALIASED) == 0) {
 644                 cache_lock(&mp->mnt_ncmountpt);
 645                 cache_inval(&mp->mnt_ncmountpt, CINV_DESTROY|CINV_CHILDREN);
 646                 cache_unlock(&mp->mnt_ncmountpt);
 647
 648                 if ((ncp = mp->mnt_ncmountpt.ncp) != NULL &&
 649                     (ncp->nc_refs != 1 || TAILQ_FIRST(&ncp->nc_list))) {
 650
 651                         if ((flags & MNT_FORCE) == 0) {
 652                                 error = EBUSY;
 653                                 mount_warning(mp, "Cannot unmount: "
 654                                                   "%d namecache "
 655                                                   "references still "
 656                                                   "present",
 657                                                   ncp->nc_refs - 1);
 658                         } else {
 659                                 mount_warning(mp, "Forced unmount: "
 660                                                   "%d namecache "
 661                                                   "references still "
 662                                                   "present",
 663                                                   ncp->nc_refs - 1);
 664                                 freeok = 0;
 665                         }
 666                 }
 667         }
 668
 669         /*
 670          * nchandle records ref the mount structure.  Expect a count of 1
 671          * (our mount->mnt_ncmountpt).
 672          */
 673         if (mp->mnt_refs != 1) {
 674                 if ((flags & MNT_FORCE) == 0) {
 675                         mount_warning(mp, "Cannot unmount: "
 676                                           "%d process references still "
 677                                           "present", mp->mnt_refs);
 678                         error = EBUSY;
 679                 } else {
 680                         mount_warning(mp, "Forced unmount: "
 681                                           "%d process references still "
 682                                           "present", mp->mnt_refs);
 683                         freeok = 0;
 684                 }
 685         }
 686
 687         /*
 688          * Decomission our special mnt_syncer vnode.  This also stops
 689          * the vnlru code.  If we are unable to unmount we recommission
 690          * the vnode.
 691          */
 692         if (error == 0) {
 693                 if ((vp = mp->mnt_syncer) != NULL) {
 694                         mp->mnt_syncer = NULL;
 695                         vrele(vp);
 696                 }
 697                 if (((mp->mnt_flag & MNT_RDONLY) ||
 698                      (error = VFS_SYNC(mp, MNT_WAIT)) == 0) ||
 699                     (flags & MNT_FORCE)) {
 700                         error = VFS_UNMOUNT(mp, flags);
 701                 }
 702         }
 703         if (error) {
 704                 if (mp->mnt_syncer == NULL)
 705                         vfs_allocate_syncvnode(mp);
 706                 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
 707                 mp->mnt_flag |= async_flag;
 708                 lockmgr(&mp->mnt_lock, LK_RELEASE);
 709                 if (mp->mnt_kern_flag & MNTK_MWAIT)
 710                         wakeup(mp);
 711                 return (error);
 712         }
 713         /*
 714          * Clean up any journals still associated with the mount after
 715          * filesystem activity has ceased.
 716          */
 717         journal_remove_all_journals(mp,
 718             ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));
 719
 720         mountlist_remove(mp);
 721
 722         /*
 723          * Remove any installed vnode ops here so the individual VFSs don't
 724          * have to.
 725          */
 726         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
 727         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
 728         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
 729         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
 730         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
 731
 732         if (mp->mnt_ncmountpt.ncp != NULL) {
 733                 nch = mp->mnt_ncmountpt;
 734                 cache_zero(&mp->mnt_ncmountpt);
 735                 cache_clrmountpt(&nch);
 736                 cache_drop(&nch);
 737         }
 738         if (mp->mnt_ncmounton.ncp != NULL) {
 739                 nch = mp->mnt_ncmounton;
 740                 cache_zero(&mp->mnt_ncmounton);
 741                 cache_clrmountpt(&nch);
 742                 cache_drop(&nch);
 743         }
 744
 745         mp->mnt_vfc->vfc_refcount--;
 746         if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
 747                 panic("unmount: dangling vnode");
 748         lockmgr(&mp->mnt_lock, LK_RELEASE);
 749         if (mp->mnt_kern_flag & MNTK_MWAIT)
 750                 wakeup(mp);
 751         if (freeok)
 752                 kfree(mp, M_MOUNT);
 753         return (0);
 754 }
 755
 756 static
 757 void
 758 mount_warning(struct mount *mp, const char *ctl, ...)
 759 {
 760         char *ptr;
 761         char *buf;
 762         __va_list va;
 763
 764         __va_start(va, ctl);
 765         if (cache_fullpath(NULL, &mp->mnt_ncmounton, &ptr, &buf) == 0) {
 766                 kprintf("unmount(%s): ", ptr);
 767                 kvprintf(ctl, va);
 768                 kprintf("\n");
 769                 kfree(buf, M_TEMP);
 770         } else {
 771                 kprintf("unmount(%p", mp);
 772                 if (mp->mnt_ncmounton.ncp && mp->mnt_ncmounton.ncp->nc_name)
 773                         kprintf(",%s", mp->mnt_ncmounton.ncp->nc_name);
 774                 kprintf("): ");
 775                 kvprintf(ctl, va);
 776                 kprintf("\n");
 777         }
 778         __va_end(va);
 779 }
 780
 781 /*
 782  * Shim cache_fullpath() to handle the case where a process is chrooted into
 783  * a subdirectory of a mount.  In this case if the root mount matches the
 784  * process root directory's mount we have to specify the process's root
 785  * directory instead of the mount point, because the mount point might
 786  * be above the root directory.
 787  */
 788 static
 789 int
 790 mount_path(struct proc *p, struct mount *mp, char **rb, char **fb)
 791 {
 792         struct nchandle *nch;
 793
 794         if (p && p->p_fd->fd_nrdir.mount == mp)
 795                 nch = &p->p_fd->fd_nrdir;
 796         else
 797                 nch = &mp->mnt_ncmountpt;
 798         return(cache_fullpath(p, nch, rb, fb));
 799 }
 800
 801 /*
 802  * Sync each mounted filesystem.
 803  */
 804
 805 #ifdef DEBUG
 806 static int syncprt = 0;
 807 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
 808 #endif /* DEBUG */
 809
 810 static int sync_callback(struct mount *mp, void *data);
 811
 812 /* ARGSUSED */
 813 int
 814 sys_sync(struct sync_args *uap)
 815 {
 816         mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
 817 #ifdef DEBUG
 818         /*
 819          * print out buffer pool stat information on each sync() call.
 820          */
 821         if (syncprt)
 822                 vfs_bufstats();
 823 #endif /* DEBUG */
 824         return (0);
 825 }
 826
 827 static
 828 int
 829 sync_callback(struct mount *mp, void *data __unused)
 830 {
 831         int asyncflag;
 832
 833         if ((mp->mnt_flag & MNT_RDONLY) == 0) {
 834                 asyncflag = mp->mnt_flag & MNT_ASYNC;
 835                 mp->mnt_flag &= ~MNT_ASYNC;
 836                 vfs_msync(mp, MNT_NOWAIT);
 837                 VFS_SYNC(mp, MNT_NOWAIT);
 838                 mp->mnt_flag |= asyncflag;
 839         }
 840         return(0);
 841 }
 842
 843 /* XXX PRISON: could be per prison flag */
 844 static int prison_quotas;
 845 #if 0
 846 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
 847 #endif
 848
 849 /*
 850  *  quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
 851  *
 852  * Change filesystem quotas.
 853  */
 854 /* ARGSUSED */
 855 int
 856 sys_quotactl(struct quotactl_args *uap)
 857 {
 858         struct nlookupdata nd;
 859         struct thread *td;
 860         struct proc *p;
 861         struct mount *mp;
 862         int error;
 863
 864         td = curthread;
 865         p = td->td_proc;
 866         if (p->p_ucred->cr_prison && !prison_quotas)
 867                 return (EPERM);
 868
 869         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
 870         if (error == 0)
 871                 error = nlookup(&nd);
 872         if (error == 0) {
 873                 mp = nd.nl_nch.mount;
 874                 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
 875                                     uap->arg, nd.nl_cred);
 876         }
 877         nlookup_done(&nd);
 878         return (error);
 879 }
 880
 881 /*
 882  * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
 883  *              void *buf, int buflen)
 884  *
 885  * This function operates on a mount point and executes the specified
 886  * operation using the specified control data, and possibly returns data.
 887  *
 888  * The actual number of bytes stored in the result buffer is returned, 0
 889  * if none, otherwise an error is returned.
 890  */
 891 /* ARGSUSED */
 892 int
 893 sys_mountctl(struct mountctl_args *uap)
 894 {
 895         struct thread *td = curthread;
 896         struct proc *p = td->td_proc;
 897         struct file *fp;
 898         void *ctl = NULL;
 899         void *buf = NULL;
 900         char *path = NULL;
 901         int error;
 902
 903         /*
 904          * Sanity and permissions checks.  We must be root.
 905          */
 906         KKASSERT(p);
 907         if (p->p_ucred->cr_prison != NULL)
 908                 return (EPERM);
 909         if ((error = priv_check(td, PRIV_ROOT)) != 0)
 910                 return (error);
 911
 912         /*
 913          * Argument length checks
 914          */
 915         if (uap->ctllen < 0 || uap->ctllen > 1024)
 916                 return (EINVAL);
 917         if (uap->buflen < 0 || uap->buflen > 16 * 1024)
 918                 return (EINVAL);
 919         if (uap->path == NULL)
 920                 return (EINVAL);
 921
 922         /*
 923          * Allocate the necessary buffers and copyin data
 924          */
 925         path = objcache_get(namei_oc, M_WAITOK);
 926         error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
 927         if (error)
 928                 goto done;
 929
 930         if (uap->ctllen) {
 931                 ctl = kmalloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
 932                 error = copyin(uap->ctl, ctl, uap->ctllen);
 933                 if (error)
 934                         goto done;
 935         }
 936         if (uap->buflen)
 937                 buf = kmalloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
 938
 939         /*
 940          * Validate the descriptor
 941          */
 942         if (uap->fd >= 0) {
 943                 fp = holdfp(p->p_fd, uap->fd, -1);
 944                 if (fp == NULL) {
 945                         error = EBADF;
 946                         goto done;
 947                 }
 948         } else {
 949                 fp = NULL;
 950         }
 951
 952         /*
 953          * Execute the internal kernel function and clean up.
 954          */
 955         error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
 956         if (fp)
 957                 fdrop(fp);
 958         if (error == 0 && uap->sysmsg_result > 0)
 959                 error = copyout(buf, uap->buf, uap->sysmsg_result);
 960 done:
 961         if (path)
 962                 objcache_put(namei_oc, path);
 963         if (ctl)
 964                 kfree(ctl, M_TEMP);
 965         if (buf)
 966                 kfree(buf, M_TEMP);
 967         return (error);
 968 }
 969
 970 /*
 971  * Execute a mount control operation by resolving the path to a mount point
 972  * and calling vop_mountctl().
 973  *
 974  * Use the mount point from the nch instead of the vnode so nullfs mounts
 975  * can properly spike the VOP.
 976  */
 977 int
 978 kern_mountctl(const char *path, int op, struct file *fp,
 979                 const void *ctl, int ctllen,
 980                 void *buf, int buflen, int *res)
 981 {
 982         struct vnode *vp;
 983         struct mount *mp;
 984         struct nlookupdata nd;
 985         int error;
 986
 987         *res = 0;
 988         vp = NULL;
 989         error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
 990         if (error == 0)
 991                 error = nlookup(&nd);
 992         if (error == 0)
 993                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
 994         mp = nd.nl_nch.mount;
 995         nlookup_done(&nd);
 996         if (error)
 997                 return (error);
 998
 999         /*
1000          * Must be the root of the filesystem
1001          */
1002         if ((vp->v_flag & (VROOT|VPFSROOT)) == 0) {
1003                 vput(vp);
1004                 return (EINVAL);
1005         }
1006         error = vop_mountctl(mp->mnt_vn_use_ops, op, fp, ctl, ctllen,
1007                                 buf, buflen, res);
1008         vput(vp);
1009         return (error);
1010 }
1011
1012 int
1013 kern_statfs(struct nlookupdata *nd, struct statfs *buf)
1014 {
1015         struct thread *td = curthread;
1016         struct proc *p = td->td_proc;
1017         struct mount *mp;
1018         struct statfs *sp;
1019         char *fullpath, *freepath;
1020         int error;
1021
1022         if ((error = nlookup(nd)) != 0)
1023                 return (error);
1024         mp = nd->nl_nch.mount;
1025         sp = &mp->mnt_stat;
1026         if ((error = VFS_STATFS(mp, sp, nd->nl_cred)) != 0)
1027                 return (error);
1028
1029         error = mount_path(p, mp, &fullpath, &freepath);
1030         if (error)
1031                 return(error);
1032         bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1033         strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1034         kfree(freepath, M_TEMP);
1035
1036         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1037         bcopy(sp, buf, sizeof(*buf));
1038         /* Only root should have access to the fsid's. */
1039         if (priv_check(td, PRIV_ROOT))
1040                 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
1041         return (0);
1042 }
1043
1044 /*
1045  * statfs_args(char *path, struct statfs *buf)
1046  *
1047  * Get filesystem statistics.
1048  */
1049 int
1050 sys_statfs(struct statfs_args *uap)
1051 {
1052         struct nlookupdata nd;
1053         struct statfs buf;
1054         int error;
1055
1056         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1057         if (error == 0)
1058                 error = kern_statfs(&nd, &buf);
1059         nlookup_done(&nd);
1060         if (error == 0)
1061                 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1062         return (error);
1063 }
1064
1065 int
1066 kern_fstatfs(int fd, struct statfs *buf)
1067 {
1068         struct thread *td = curthread;
1069         struct proc *p = td->td_proc;
1070         struct file *fp;
1071         struct mount *mp;
1072         struct statfs *sp;
1073         char *fullpath, *freepath;
1074         int error;
1075
1076         KKASSERT(p);
1077         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
1078                 return (error);
1079         mp = ((struct vnode *)fp->f_data)->v_mount;
1080         if (mp == NULL) {
1081                 error = EBADF;
1082                 goto done;
1083         }
1084         if (fp->f_cred == NULL) {
1085                 error = EINVAL;
1086                 goto done;
1087         }
1088         sp = &mp->mnt_stat;
1089         if ((error = VFS_STATFS(mp, sp, fp->f_cred)) != 0)
1090                 goto done;
1091
1092         if ((error = mount_path(p, mp, &fullpath, &freepath)) != 0)
1093                 goto done;
1094         bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1095         strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1096         kfree(freepath, M_TEMP);
1097
1098         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1099         bcopy(sp, buf, sizeof(*buf));
1100
1101         /* Only root should have access to the fsid's. */
1102         if (priv_check(td, PRIV_ROOT))
1103                 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
1104         error = 0;
1105 done:
1106         fdrop(fp);
1107         return (error);
1108 }
1109
1110 /*
1111  * fstatfs_args(int fd, struct statfs *buf)
1112  *
1113  * Get filesystem statistics.
1114  */
1115 int
1116 sys_fstatfs(struct fstatfs_args *uap)
1117 {
1118         struct statfs buf;
1119         int error;
1120
1121         error = kern_fstatfs(uap->fd, &buf);
1122
1123         if (error == 0)
1124                 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1125         return (error);
1126 }
1127
1128 int
1129 kern_statvfs(struct nlookupdata *nd, struct statvfs *buf)
1130 {
1131         struct mount *mp;
1132         struct statvfs *sp;
1133         int error;
1134
1135         if ((error = nlookup(nd)) != 0)
1136                 return (error);
1137         mp = nd->nl_nch.mount;
1138         sp = &mp->mnt_vstat;
1139         if ((error = VFS_STATVFS(mp, sp, nd->nl_cred)) != 0)
1140                 return (error);
1141
1142         sp->f_flag = 0;
1143         if (mp->mnt_flag & MNT_RDONLY)
1144                 sp->f_flag |= ST_RDONLY;
1145         if (mp->mnt_flag & MNT_NOSUID)
1146                 sp->f_flag |= ST_NOSUID;
1147         bcopy(sp, buf, sizeof(*buf));
1148         return (0);
1149 }
1150
1151 /*
1152  * statfs_args(char *path, struct statfs *buf)
1153  *
1154  * Get filesystem statistics.
1155  */
1156 int
1157 sys_statvfs(struct statvfs_args *uap)
1158 {
1159         struct nlookupdata nd;
1160         struct statvfs buf;
1161         int error;
1162
1163         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1164         if (error == 0)
1165                 error = kern_statvfs(&nd, &buf);
1166         nlookup_done(&nd);
1167         if (error == 0)
1168                 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1169         return (error);
1170 }
1171
1172 int
1173 kern_fstatvfs(int fd, struct statvfs *buf)
1174 {
1175         struct thread *td = curthread;
1176         struct proc *p = td->td_proc;
1177         struct file *fp;
1178         struct mount *mp;
1179         struct statvfs *sp;
1180         int error;
1181
1182         KKASSERT(p);
1183         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
1184                 return (error);
1185         mp = ((struct vnode *)fp->f_data)->v_mount;
1186         if (mp == NULL) {
1187                 error = EBADF;
1188                 goto done;
1189         }
1190         if (fp->f_cred == NULL) {
1191                 error = EINVAL;
1192                 goto done;
1193         }
1194         sp = &mp->mnt_vstat;
1195         if ((error = VFS_STATVFS(mp, sp, fp->f_cred)) != 0)
1196                 goto done;
1197
1198         sp->f_flag = 0;
1199         if (mp->mnt_flag & MNT_RDONLY)
1200                 sp->f_flag |= ST_RDONLY;
1201         if (mp->mnt_flag & MNT_NOSUID)
1202                 sp->f_flag |= ST_NOSUID;
1203
1204         bcopy(sp, buf, sizeof(*buf));
1205         error = 0;
1206 done:
1207         fdrop(fp);
1208         return (error);
1209 }
1210
1211 /*
1212  * fstatfs_args(int fd, struct statfs *buf)
1213  *
1214  * Get filesystem statistics.
1215  */
1216 int
1217 sys_fstatvfs(struct fstatvfs_args *uap)
1218 {
1219         struct statvfs buf;
1220         int error;
1221
1222         error = kern_fstatvfs(uap->fd, &buf);
1223
1224         if (error == 0)
1225                 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1226         return (error);
1227 }
1228
1229 /*
1230  * getfsstat_args(struct statfs *buf, long bufsize, int flags)
1231  *
1232  * Get statistics on all filesystems.
1233  */
1234
1235 struct getfsstat_info {
1236         struct statfs *sfsp;
1237         long count;
1238         long maxcount;
1239         int error;
1240         int flags;
1241         struct proc *p;
1242 };
1243
1244 static int getfsstat_callback(struct mount *, void *);
1245
1246 /* ARGSUSED */
1247 int
1248 sys_getfsstat(struct getfsstat_args *uap)
1249 {
1250         struct thread *td = curthread;
1251         struct proc *p = td->td_proc;
1252         struct getfsstat_info info;
1253
1254         bzero(&info, sizeof(info));
1255
1256         info.maxcount = uap->bufsize / sizeof(struct statfs);
1257         info.sfsp = uap->buf;
1258         info.count = 0;
1259         info.flags = uap->flags;
1260         info.p = p;
1261
1262         mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
1263         if (info.sfsp && info.count > info.maxcount)
1264                 uap->sysmsg_result = info.maxcount;
1265         else
1266                 uap->sysmsg_result = info.count;
1267         return (info.error);
1268 }
1269
1270 static int
1271 getfsstat_callback(struct mount *mp, void *data)
1272 {
1273         struct getfsstat_info *info = data;
1274         struct statfs *sp;
1275         char *freepath;
1276         char *fullpath;
1277         int error;
1278
1279         if (info->sfsp && info->count < info->maxcount) {
1280                 if (info->p && !chroot_visible_mnt(mp, info->p))
1281                         return(0);
1282                 sp = &mp->mnt_stat;
1283
1284                 /*
1285                  * If MNT_NOWAIT or MNT_LAZY is specified, do not
1286                  * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1287                  * overrides MNT_WAIT.
1288                  */
1289                 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1290                     (info->flags & MNT_WAIT)) &&
1291                     (error = VFS_STATFS(mp, sp, info->p->p_ucred))) {
1292                         return(0);
1293                 }
1294                 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1295
1296                 error = mount_path(info->p, mp, &fullpath, &freepath);
1297                 if (error) {
1298                         info->error = error;
1299                         return(-1);
1300                 }
1301                 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1302                 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1303                 kfree(freepath, M_TEMP);
1304
1305                 error = copyout(sp, info->sfsp, sizeof(*sp));
1306                 if (error) {
1307                         info->error = error;
1308                         return (-1);
1309                 }
1310                 ++info->sfsp;
1311         }
1312         info->count++;
1313         return(0);
1314 }
1315
1316 /*
1317  * getvfsstat_args(struct statfs *buf, struct statvfs *vbuf,
1318                    long bufsize, int flags)
1319  *
1320  * Get statistics on all filesystems.
1321  */
1322
1323 struct getvfsstat_info {
1324         struct statfs *sfsp;
1325         struct statvfs *vsfsp;
1326         long count;
1327         long maxcount;
1328         int error;
1329         int flags;
1330         struct proc *p;
1331 };
1332
1333 static int getvfsstat_callback(struct mount *, void *);
1334
1335 /* ARGSUSED */
1336 int
1337 sys_getvfsstat(struct getvfsstat_args *uap)
1338 {
1339         struct thread *td = curthread;
1340         struct proc *p = td->td_proc;
1341         struct getvfsstat_info info;
1342
1343         bzero(&info, sizeof(info));
1344
1345         info.maxcount = uap->vbufsize / sizeof(struct statvfs);
1346         info.sfsp = uap->buf;
1347         info.vsfsp = uap->vbuf;
1348         info.count = 0;
1349         info.flags = uap->flags;
1350         info.p = p;
1351
1352         mountlist_scan(getvfsstat_callback, &info, MNTSCAN_FORWARD);
1353         if (info.vsfsp && info.count > info.maxcount)
1354                 uap->sysmsg_result = info.maxcount;
1355         else
1356                 uap->sysmsg_result = info.count;
1357         return (info.error);
1358 }
1359
1360 static int
1361 getvfsstat_callback(struct mount *mp, void *data)
1362 {
1363         struct getvfsstat_info *info = data;
1364         struct statfs *sp;
1365         struct statvfs *vsp;
1366         char *freepath;
1367         char *fullpath;
1368         int error;
1369
1370         if (info->vsfsp && info->count < info->maxcount) {
1371                 if (info->p && !chroot_visible_mnt(mp, info->p))
1372                         return(0);
1373                 sp = &mp->mnt_stat;
1374                 vsp = &mp->mnt_vstat;
1375
1376                 /*
1377                  * If MNT_NOWAIT or MNT_LAZY is specified, do not
1378                  * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1379                  * overrides MNT_WAIT.
1380                  */
1381                 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1382                     (info->flags & MNT_WAIT)) &&
1383                     (error = VFS_STATFS(mp, sp, info->p->p_ucred))) {
1384                         return(0);
1385                 }
1386                 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1387
1388                 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1389                     (info->flags & MNT_WAIT)) &&
1390                     (error = VFS_STATVFS(mp, vsp, info->p->p_ucred))) {
1391                         return(0);
1392                 }
1393                 vsp->f_flag = 0;
1394                 if (mp->mnt_flag & MNT_RDONLY)
1395                         vsp->f_flag |= ST_RDONLY;
1396                 if (mp->mnt_flag & MNT_NOSUID)
1397                         vsp->f_flag |= ST_NOSUID;
1398
1399                 error = mount_path(info->p, mp, &fullpath, &freepath);
1400                 if (error) {
1401                         info->error = error;
1402                         return(-1);
1403                 }
1404                 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1405                 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1406                 kfree(freepath, M_TEMP);
1407
1408                 error = copyout(sp, info->sfsp, sizeof(*sp));
1409                 if (error == 0)
1410                         error = copyout(vsp, info->vsfsp, sizeof(*vsp));
1411                 if (error) {
1412                         info->error = error;
1413                         return (-1);
1414                 }
1415                 ++info->sfsp;
1416                 ++info->vsfsp;
1417         }
1418         info->count++;
1419         return(0);
1420 }
1421
1422
1423 /*
1424  * fchdir_args(int fd)
1425  *
1426  * Change current working directory to a given file descriptor.
1427  */
1428 /* ARGSUSED */
1429 int
1430 sys_fchdir(struct fchdir_args *uap)
1431 {
1432         struct thread *td = curthread;
1433         struct proc *p = td->td_proc;
1434         struct filedesc *fdp = p->p_fd;
1435         struct vnode *vp, *ovp;
1436         struct mount *mp;
1437         struct file *fp;
1438         struct nchandle nch, onch, tnch;
1439         int error;
1440
1441         if ((error = holdvnode(fdp, uap->fd, &fp)) != 0)
1442                 return (error);
1443         vp = (struct vnode *)fp->f_data;
1444         vref(vp);
1445         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1446         if (vp->v_type != VDIR || fp->f_nchandle.ncp == NULL)
1447                 error = ENOTDIR;
1448         else
1449                 error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
1450         if (error) {
1451                 vput(vp);
1452                 fdrop(fp);
1453                 return (error);
1454         }
1455         cache_copy(&fp->f_nchandle, &nch);
1456
1457         /*
1458          * If the ncp has become a mount point, traverse through
1459          * the mount point.
1460          */
1461
1462         while (!error && (nch.ncp->nc_flag & NCF_ISMOUNTPT) &&
1463                (mp = cache_findmount(&nch)) != NULL
1464         ) {
1465                 error = nlookup_mp(mp, &tnch);
1466                 if (error == 0) {
1467                         cache_unlock(&tnch);    /* leave ref intact */
1468                         vput(vp);
1469                         vp = tnch.ncp->nc_vp;
1470                         error = vget(vp, LK_SHARED);
1471                         KKASSERT(error == 0);
1472                         cache_drop(&nch);
1473                         nch = tnch;
1474                 }
1475         }
1476         if (error == 0) {
1477                 ovp = fdp->fd_cdir;
1478                 onch = fdp->fd_ncdir;
1479                 vn_unlock(vp);          /* leave ref intact */
1480                 fdp->fd_cdir = vp;
1481                 fdp->fd_ncdir = nch;
1482                 cache_drop(&onch);
1483                 vrele(ovp);
1484         } else {
1485                 cache_drop(&nch);
1486                 vput(vp);
1487         }
1488         fdrop(fp);
1489         return (error);
1490 }
1491
1492 int
1493 kern_chdir(struct nlookupdata *nd)
1494 {
1495         struct thread *td = curthread;
1496         struct proc *p = td->td_proc;
1497         struct filedesc *fdp = p->p_fd;
1498         struct vnode *vp, *ovp;
1499         struct nchandle onch;
1500         int error;
1501
1502         if ((error = nlookup(nd)) != 0)
1503                 return (error);
1504         if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
1505                 return (ENOENT);
1506         if ((error = vget(vp, LK_SHARED)) != 0)
1507                 return (error);
1508
1509         error = checkvp_chdir(vp, td);
1510         vn_unlock(vp);
1511         if (error == 0) {
1512                 ovp = fdp->fd_cdir;
1513                 onch = fdp->fd_ncdir;
1514                 cache_unlock(&nd->nl_nch);      /* leave reference intact */
1515                 fdp->fd_ncdir = nd->nl_nch;
1516                 fdp->fd_cdir = vp;
1517                 cache_drop(&onch);
1518                 vrele(ovp);
1519                 cache_zero(&nd->nl_nch);
1520         } else {
1521                 vrele(vp);
1522         }
1523         return (error);
1524 }
1525
1526 /*
1527  * chdir_args(char *path)
1528  *
1529  * Change current working directory (``.'').
1530  */
1531 int
1532 sys_chdir(struct chdir_args *uap)
1533 {
1534         struct nlookupdata nd;
1535         int error;
1536
1537         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1538         if (error == 0)
1539                 error = kern_chdir(&nd);
1540         nlookup_done(&nd);
1541         return (error);
1542 }
1543
1544 /*
1545  * Helper function for raised chroot(2) security function:  Refuse if
1546  * any filedescriptors are open directories.
1547  */
1548 static int
1549 chroot_refuse_vdir_fds(struct filedesc *fdp)
1550 {
1551         struct vnode *vp;
1552         struct file *fp;
1553         int error;
1554         int fd;
1555
1556         for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
1557                 if ((error = holdvnode(fdp, fd, &fp)) != 0)
1558                         continue;
1559                 vp = (struct vnode *)fp->f_data;
1560                 if (vp->v_type != VDIR) {
1561                         fdrop(fp);
1562                         continue;
1563                 }
1564                 fdrop(fp);
1565                 return(EPERM);
1566         }
1567         return (0);
1568 }
1569
1570 /*
1571  * This sysctl determines if we will allow a process to chroot(2) if it
1572  * has a directory open:
1573  *      0: disallowed for all processes.
1574  *      1: allowed for processes that were not already chroot(2)'ed.
1575  *      2: allowed for all processes.
1576  */
1577
1578 static int chroot_allow_open_directories = 1;
1579
1580 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1581      &chroot_allow_open_directories, 0, "");
1582
1583 /*
1584  * chroot to the specified namecache entry.  We obtain the vp from the
1585  * namecache data.  The passed ncp must be locked and referenced and will
1586  * remain locked and referenced on return.
1587  */
1588 int
1589 kern_chroot(struct nchandle *nch)
1590 {
1591         struct thread *td = curthread;
1592         struct proc *p = td->td_proc;
1593         struct filedesc *fdp = p->p_fd;
1594         struct vnode *vp;
1595         int error;
1596
1597         /*
1598          * Only root can chroot
1599          */
1600         if ((error = priv_check_cred(p->p_ucred, PRIV_ROOT, PRISON_ROOT)) != 0)
1601                 return (error);
1602
1603         /*
1604          * Disallow open directory descriptors (fchdir() breakouts).
1605          */
1606         if (chroot_allow_open_directories == 0 ||
1607            (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1608                 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1609                         return (error);
1610         }
1611         if ((vp = nch->ncp->nc_vp) == NULL)
1612                 return (ENOENT);
1613
1614         if ((error = vget(vp, LK_SHARED)) != 0)
1615                 return (error);
1616
1617         /*
1618          * Check the validity of vp as a directory to change to and
1619          * associate it with rdir/jdir.
1620          */
1621         error = checkvp_chdir(vp, td);
1622         vn_unlock(vp);                  /* leave reference intact */
1623         if (error == 0) {
1624                 vrele(fdp->fd_rdir);
1625                 fdp->fd_rdir = vp;      /* reference inherited by fd_rdir */
1626                 cache_drop(&fdp->fd_nrdir);
1627                 cache_copy(nch, &fdp->fd_nrdir);
1628                 if (fdp->fd_jdir == NULL) {
1629                         fdp->fd_jdir = vp;
1630                         vref(fdp->fd_jdir);
1631                         cache_copy(nch, &fdp->fd_njdir);
1632                 }
1633         } else {
1634                 vrele(vp);
1635         }
1636         return (error);
1637 }
1638
1639 /*
1640  * chroot_args(char *path)
1641  *
1642  * Change notion of root (``/'') directory.
1643  */
1644 /* ARGSUSED */
1645 int
1646 sys_chroot(struct chroot_args *uap)
1647 {
1648         struct thread *td = curthread;
1649         struct nlookupdata nd;
1650         int error;
1651
1652         KKASSERT(td->td_proc);
1653         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1654         if (error) {
1655                 nlookup_done(&nd);
1656                 return(error);
1657         }
1658         error = nlookup(&nd);
1659         if (error == 0)
1660                 error = kern_chroot(&nd.nl_nch);
1661         nlookup_done(&nd);
1662         return(error);
1663 }
1664
1665 /*
1666  * Common routine for chroot and chdir.  Given a locked, referenced vnode,
1667  * determine whether it is legal to chdir to the vnode.  The vnode's state
1668  * is not changed by this call.
1669  */
1670 int
1671 checkvp_chdir(struct vnode *vp, struct thread *td)
1672 {
1673         int error;
1674
1675         if (vp->v_type != VDIR)
1676                 error = ENOTDIR;
1677         else
1678                 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred);
1679         return (error);
1680 }
1681
1682 int
1683 kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
1684 {
1685         struct thread *td = curthread;
1686         struct proc *p = td->td_proc;
1687         struct lwp *lp = td->td_lwp;
1688         struct filedesc *fdp = p->p_fd;
1689         int cmode, flags;
1690         struct file *nfp;
1691         struct file *fp;
1692         struct vnode *vp;
1693         int type, indx, error;
1694         struct flock lf;
1695
1696         if ((oflags & O_ACCMODE) == O_ACCMODE)
1697                 return (EINVAL);
1698         flags = FFLAGS(oflags);
1699         error = falloc(p, &nfp, NULL);
1700         if (error)
1701                 return (error);
1702         fp = nfp;
1703         cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1704
1705         /*
1706          * XXX p_dupfd is a real mess.  It allows a device to return a
1707          * file descriptor to be duplicated rather then doing the open
1708          * itself.
1709          */
1710         lp->lwp_dupfd = -1;
1711
1712         /*
1713          * Call vn_open() to do the lookup and assign the vnode to the
1714          * file pointer.  vn_open() does not change the ref count on fp
1715          * and the vnode, on success, will be inherited by the file pointer
1716          * and unlocked.
1717          */
1718         nd->nl_flags |= NLC_LOCKVP;
1719         error = vn_open(nd, fp, flags, cmode);
1720         nlookup_done(nd);
1721         if (error) {
1722                 /*
1723                  * handle special fdopen() case.  bleh.  dupfdopen() is
1724                  * responsible for dropping the old contents of ofiles[indx]
1725                  * if it succeeds.
1726                  *
1727                  * Note that fsetfd() will add a ref to fp which represents
1728                  * the fd_files[] assignment.  We must still drop our
1729                  * reference.
1730                  */
1731                 if ((error == ENODEV || error == ENXIO) && lp->lwp_dupfd >= 0) {
1732                         if (fdalloc(p, 0, &indx) == 0) {
1733                                 error = dupfdopen(p, indx, lp->lwp_dupfd, flags, error);
1734                                 if (error == 0) {
1735                                         *res = indx;
1736                                         fdrop(fp);      /* our ref */
1737                                         return (0);
1738                                 }
1739                                 fsetfd(p, NULL, indx);
1740                         }
1741                 }
1742                 fdrop(fp);      /* our ref */
1743                 if (error == ERESTART)
1744                         error = EINTR;
1745                 return (error);
1746         }
1747
1748         /*
1749          * ref the vnode for ourselves so it can't be ripped out from under
1750          * is.  XXX need an ND flag to request that the vnode be returned
1751          * anyway.
1752          *
1753          * Reserve a file descriptor but do not assign it until the open
1754          * succeeds.
1755          */
1756         vp = (struct vnode *)fp->f_data;
1757         vref(vp);
1758         if ((error = fdalloc(p, 0, &indx)) != 0) {
1759                 fdrop(fp);
1760                 vrele(vp);
1761                 return (error);
1762         }
1763
1764         /*
1765          * If no error occurs the vp will have been assigned to the file
1766          * pointer.
1767          */
1768         lp->lwp_dupfd = 0;
1769
1770         if (flags & (O_EXLOCK | O_SHLOCK)) {
1771                 lf.l_whence = SEEK_SET;
1772                 lf.l_start = 0;
1773                 lf.l_len = 0;
1774                 if (flags & O_EXLOCK)
1775                         lf.l_type = F_WRLCK;
1776                 else
1777                         lf.l_type = F_RDLCK;
1778                 if (flags & FNONBLOCK)
1779                         type = 0;
1780                 else
1781                         type = F_WAIT;
1782
1783                 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1784                         /*
1785                          * lock request failed.  Clean up the reserved
1786                          * descriptor.
1787                          */
1788                         vrele(vp);
1789                         fsetfd(p, NULL, indx);
1790                         fdrop(fp);
1791                         return (error);
1792                 }
1793                 fp->f_flag |= FHASLOCK;
1794         }
1795 #if 0
1796         /*
1797          * Assert that all regular file vnodes were created with a object.
1798          */
1799         KASSERT(vp->v_type != VREG || vp->v_object != NULL,
1800                 ("open: regular file has no backing object after vn_open"));
1801 #endif
1802
1803         vrele(vp);
1804
1805         /*
1806          * release our private reference, leaving the one associated with the
1807          * descriptor table intact.
1808          */
1809         fsetfd(p, fp, indx);
1810         fdrop(fp);
1811         *res = indx;
1812         return (0);
1813 }
1814
1815 /*
1816  * open_args(char *path, int flags, int mode)
1817  *
1818  * Check permissions, allocate an open file structure,
1819  * and call the device open routine if any.
1820  */
1821 int
1822 sys_open(struct open_args *uap)
1823 {
1824         struct nlookupdata nd;
1825         int error;
1826
1827         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1828         if (error == 0) {
1829                 error = kern_open(&nd, uap->flags,
1830                                     uap->mode, &uap->sysmsg_result);
1831         }
1832         nlookup_done(&nd);
1833         return (error);
1834 }
1835
1836 int
1837 kern_mknod(struct nlookupdata *nd, int mode, int rmajor, int rminor)
1838 {
1839         struct thread *td = curthread;
1840         struct proc *p = td->td_proc;
1841         struct vnode *vp;
1842         struct vattr vattr;
1843         int error;
1844         int whiteout = 0;
1845
1846         KKASSERT(p);
1847
1848         switch (mode & S_IFMT) {
1849         case S_IFCHR:
1850         case S_IFBLK:
1851                 error = priv_check(td, PRIV_ROOT);
1852                 break;
1853         default:
1854                 error = priv_check_cred(p->p_ucred, PRIV_ROOT, PRISON_ROOT);
1855                 break;
1856         }
1857         if (error)
1858                 return (error);
1859
1860         bwillinode(1);
1861         nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
1862         if ((error = nlookup(nd)) != 0)
1863                 return (error);
1864         if (nd->nl_nch.ncp->nc_vp)
1865                 return (EEXIST);
1866         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
1867                 return (error);
1868
1869         VATTR_NULL(&vattr);
1870         vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1871         vattr.va_rmajor = rmajor;
1872         vattr.va_rminor = rminor;
1873         whiteout = 0;
1874
1875         switch (mode & S_IFMT) {
1876         case S_IFMT:    /* used by badsect to flag bad sectors */
1877                 vattr.va_type = VBAD;
1878                 break;
1879         case S_IFCHR:
1880                 vattr.va_type = VCHR;
1881                 break;
1882         case S_IFBLK:
1883                 vattr.va_type = VBLK;
1884                 break;
1885         case S_IFWHT:
1886                 whiteout = 1;
1887                 break;
1888         case S_IFDIR:
1889                 /* special directories support for HAMMER */
1890                 vattr.va_type = VDIR;
1891                 break;
1892         default:
1893                 error = EINVAL;
1894                 break;
1895         }
1896         if (error == 0) {
1897                 if (whiteout) {
1898                         error = VOP_NWHITEOUT(&nd->nl_nch, nd->nl_dvp,
1899                                               nd->nl_cred, NAMEI_CREATE);
1900                 } else {
1901                         vp = NULL;
1902                         error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp,
1903                                            &vp, nd->nl_cred, &vattr);
1904                         if (error == 0)
1905                                 vput(vp);
1906                 }
1907         }
1908         return (error);
1909 }
1910
1911 /*
1912  * mknod_args(char *path, int mode, int dev)
1913  *
1914  * Create a special file.
1915  */
1916 int
1917 sys_mknod(struct mknod_args *uap)
1918 {
1919         struct nlookupdata nd;
1920         int error;
1921
1922         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1923         if (error == 0) {
1924                 error = kern_mknod(&nd, uap->mode,
1925                                    umajor(uap->dev), uminor(uap->dev));
1926         }
1927         nlookup_done(&nd);
1928         return (error);
1929 }
1930
1931 int
1932 kern_mkfifo(struct nlookupdata *nd, int mode)
1933 {
1934         struct thread *td = curthread;
1935         struct proc *p = td->td_proc;
1936         struct vattr vattr;
1937         struct vnode *vp;
1938         int error;
1939
1940         bwillinode(1);
1941
1942         nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
1943         if ((error = nlookup(nd)) != 0)
1944                 return (error);
1945         if (nd->nl_nch.ncp->nc_vp)
1946                 return (EEXIST);
1947         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
1948                 return (error);
1949
1950         VATTR_NULL(&vattr);
1951         vattr.va_type = VFIFO;
1952         vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1953         vp = NULL;
1954         error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp, &vp, nd->nl_cred, &vattr);
1955         if (error == 0)
1956                 vput(vp);
1957         return (error);
1958 }
1959
1960 /*
1961  * mkfifo_args(char *path, int mode)
1962  *
1963  * Create a named pipe.
1964  */
1965 int
1966 sys_mkfifo(struct mkfifo_args *uap)
1967 {
1968         struct nlookupdata nd;
1969         int error;
1970
1971         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1972         if (error == 0)
1973                 error = kern_mkfifo(&nd, uap->mode);
1974         nlookup_done(&nd);
1975         return (error);
1976 }
1977
1978 static int hardlink_check_uid = 0;
1979 SYSCTL_INT(_security, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1980     &hardlink_check_uid, 0,
1981     "Unprivileged processes cannot create hard links to files owned by other "
1982     "users");
1983 static int hardlink_check_gid = 0;
1984 SYSCTL_INT(_security, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1985     &hardlink_check_gid, 0,
1986     "Unprivileged processes cannot create hard links to files owned by other "
1987     "groups");
1988
1989 static int
1990 can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
1991 {
1992         struct vattr va;
1993         int error;
1994
1995         /*
1996          * Shortcut if disabled
1997          */
1998         if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
1999                 return (0);
2000
2001         /*
2002          * root cred can always hardlink
2003          */
2004         if (priv_check_cred(cred, PRIV_ROOT, PRISON_ROOT) == 0)
2005                 return (0);
2006
2007         /*
2008          * Otherwise only if the originating file is owned by the
2009          * same user or group.  Note that any group is allowed if
2010          * the file is owned by the caller.
2011          */
2012         error = VOP_GETATTR(vp, &va);
2013         if (error != 0)
2014                 return (error);
2015
2016         if (hardlink_check_uid) {
2017                 if (cred->cr_uid != va.va_uid)
2018                         return (EPERM);
2019         }
2020
2021         if (hardlink_check_gid) {
2022                 if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
2023                         return (EPERM);
2024         }
2025
2026         return (0);
2027 }
2028
2029 int
2030 kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
2031 {
2032         struct thread *td = curthread;
2033         struct vnode *vp;
2034         int error;
2035
2036         /*
2037          * Lookup the source and obtained a locked vnode.
2038          *
2039          * XXX relookup on vget failure / race ?
2040          */
2041         bwillinode(1);
2042         if ((error = nlookup(nd)) != 0)
2043                 return (error);
2044         vp = nd->nl_nch.ncp->nc_vp;
2045         KKASSERT(vp != NULL);
2046         if (vp->v_type == VDIR)
2047                 return (EPERM);         /* POSIX */
2048         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2049                 return (error);
2050         if ((error = vget(vp, LK_EXCLUSIVE)) != 0)
2051                 return (error);
2052
2053         /*
2054          * Unlock the source so we can lookup the target without deadlocking
2055          * (XXX vp is locked already, possible other deadlock?).  The target
2056          * must not exist.
2057          */
2058         KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
2059         nd->nl_flags &= ~NLC_NCPISLOCKED;
2060         cache_unlock(&nd->nl_nch);
2061
2062         linknd->nl_flags |= NLC_CREATE | NLC_REFDVP;
2063         if ((error = nlookup(linknd)) != 0) {
2064                 vput(vp);
2065                 return (error);
2066         }
2067         if (linknd->nl_nch.ncp->nc_vp) {
2068                 vput(vp);
2069                 return (EEXIST);
2070         }
2071
2072         /*
2073          * Finally run the new API VOP.
2074          */
2075         error = can_hardlink(vp, td, td->td_proc->p_ucred);
2076         if (error == 0) {
2077                 error = VOP_NLINK(&linknd->nl_nch, linknd->nl_dvp,
2078                                   vp, linknd->nl_cred);
2079         }
2080         vput(vp);
2081         return (error);
2082 }
2083
2084 /*
2085  * link_args(char *path, char *link)
2086  *
2087  * Make a hard file link.
2088  */
2089 int
2090 sys_link(struct link_args *uap)
2091 {
2092         struct nlookupdata nd, linknd;
2093         int error;
2094
2095         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2096         if (error == 0) {
2097                 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
2098                 if (error == 0)
2099                         error = kern_link(&nd, &linknd);
2100                 nlookup_done(&linknd);
2101         }
2102         nlookup_done(&nd);
2103         return (error);
2104 }
2105
2106 int
2107 kern_symlink(struct nlookupdata *nd, char *path, int mode)
2108 {
2109         struct vattr vattr;
2110         struct vnode *vp;
2111         struct vnode *dvp;
2112         int error;
2113
2114         bwillinode(1);
2115         nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
2116         if ((error = nlookup(nd)) != 0)
2117                 return (error);
2118         if (nd->nl_nch.ncp->nc_vp)
2119                 return (EEXIST);
2120         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2121                 return (error);
2122         dvp = nd->nl_dvp;
2123         VATTR_NULL(&vattr);
2124         vattr.va_mode = mode;
2125         error = VOP_NSYMLINK(&nd->nl_nch, dvp, &vp, nd->nl_cred, &vattr, path);
2126         if (error == 0)
2127                 vput(vp);
2128         return (error);
2129 }
2130
2131 /*
2132  * symlink(char *path, char *link)
2133  *
2134  * Make a symbolic link.
2135  */
2136 int
2137 sys_symlink(struct symlink_args *uap)
2138 {
2139         struct thread *td = curthread;
2140         struct nlookupdata nd;
2141         char *path;
2142         int error;
2143         int mode;
2144
2145         path = objcache_get(namei_oc, M_WAITOK);
2146         error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
2147         if (error == 0) {
2148                 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
2149                 if (error == 0) {
2150                         mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
2151                         error = kern_symlink(&nd, path, mode);
2152                 }
2153                 nlookup_done(&nd);
2154         }
2155         objcache_put(namei_oc, path);
2156         return (error);
2157 }
2158
2159 /*
2160  * undelete_args(char *path)
2161  *
2162  * Delete a whiteout from the filesystem.
2163  */
2164 /* ARGSUSED */
2165 int
2166 sys_undelete(struct undelete_args *uap)
2167 {
2168         struct nlookupdata nd;
2169         int error;
2170
2171         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2172         bwillinode(1);
2173         nd.nl_flags |= NLC_DELETE | NLC_REFDVP;
2174         if (error == 0)
2175                 error = nlookup(&nd);
2176         if (error == 0)
2177                 error = ncp_writechk(&nd.nl_nch);
2178         if (error == 0) {
2179                 error = VOP_NWHITEOUT(&nd.nl_nch, nd.nl_dvp, nd.nl_cred,
2180                                       NAMEI_DELETE);
2181         }
2182         nlookup_done(&nd);
2183         return (error);
2184 }
2185
2186 int
2187 kern_unlink(struct nlookupdata *nd)
2188 {
2189         int error;
2190
2191         bwillinode(1);
2192         nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
2193         if ((error = nlookup(nd)) != 0)
2194                 return (error);
2195         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2196                 return (error);
2197         error = VOP_NREMOVE(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
2198         return (error);
2199 }
2200
2201 /*
2202  * unlink_args(char *path)
2203  *
2204  * Delete a name from the filesystem.
2205  */
2206 int
2207 sys_unlink(struct unlink_args *uap)
2208 {
2209         struct nlookupdata nd;
2210         int error;
2211
2212         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2213         if (error == 0)
2214                 error = kern_unlink(&nd);
2215         nlookup_done(&nd);
2216         return (error);
2217 }
2218
2219 int
2220 kern_lseek(int fd, off_t offset, int whence, off_t *res)
2221 {
2222         struct thread *td = curthread;
2223         struct proc *p = td->td_proc;
2224         struct file *fp;
2225         struct vnode *vp;
2226         struct vattr vattr;
2227         off_t new_offset;
2228         int error;
2229
2230         fp = holdfp(p->p_fd, fd, -1);
2231         if (fp == NULL)
2232                 return (EBADF);
2233         if (fp->f_type != DTYPE_VNODE) {
2234                 error = ESPIPE;
2235                 goto done;
2236         }
2237         vp = (struct vnode *)fp->f_data;
2238
2239         switch (whence) {
2240         case L_INCR:
2241                 new_offset = fp->f_offset + offset;
2242                 error = 0;
2243                 break;
2244         case L_XTND:
2245                 error = VOP_GETATTR(vp, &vattr);
2246                 new_offset = offset + vattr.va_size;
2247                 break;
2248         case L_SET:
2249                 new_offset = offset;
2250                 error = 0;
2251                 break;
2252         default:
2253                 new_offset = 0;
2254                 error = EINVAL;
2255                 break;
2256         }
2257
2258         /*
2259          * Validate the seek position.  Negative offsets are not allowed
2260          * for regular files, block specials, or directories.
2261          */
2262         if (error == 0) {
2263                 if (new_offset < 0 &&
2264                     (vp->v_type == VREG || vp->v_type == VDIR ||
2265                      vp->v_type == VCHR || vp->v_type == VBLK)) {
2266                         error = EINVAL;
2267                 } else {
2268                         fp->f_offset = new_offset;
2269                 }
2270         }
2271         *res = fp->f_offset;
2272 done:
2273         fdrop(fp);
2274         return (error);
2275 }
2276
2277 /*
2278  * lseek_args(int fd, int pad, off_t offset, int whence)
2279  *
2280  * Reposition read/write file offset.
2281  */
2282 int
2283 sys_lseek(struct lseek_args *uap)
2284 {
2285         int error;
2286
2287         error = kern_lseek(uap->fd, uap->offset, uap->whence,
2288                            &uap->sysmsg_offset);
2289
2290         return (error);
2291 }
2292
2293 int
2294 kern_access(struct nlookupdata *nd, int aflags)
2295 {
2296         struct vnode *vp;
2297         int error, flags;
2298
2299         if ((error = nlookup(nd)) != 0)
2300                 return (error);
2301 retry:
2302         error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
2303         if (error)
2304                 return (error);
2305
2306         /* Flags == 0 means only check for existence. */
2307         if (aflags) {
2308                 flags = 0;
2309                 if (aflags & R_OK)
2310                         flags |= VREAD;
2311                 if (aflags & W_OK)
2312                         flags |= VWRITE;
2313                 if (aflags & X_OK)
2314                         flags |= VEXEC;
2315                 if ((flags & VWRITE) == 0 ||
2316                     (error = vn_writechk(vp, &nd->nl_nch)) == 0)
2317                         error = VOP_ACCESS(vp, flags, nd->nl_cred);
2318
2319                 /*
2320                  * If the file handle is stale we have to re-resolve the
2321                  * entry.  This is a hack at the moment.
2322                  */
2323                 if (error == ESTALE) {
2324                         vput(vp);
2325                         cache_setunresolved(&nd->nl_nch);
2326                         error = cache_resolve(&nd->nl_nch, nd->nl_cred);
2327                         if (error == 0) {
2328                                 vp = NULL;
2329                                 goto retry;
2330                         }
2331                         return(error);
2332                 }
2333         }
2334         vput(vp);
2335         return (error);
2336 }
2337
2338 /*
2339  * access_args(char *path, int flags)
2340  *
2341  * Check access permissions.
2342  */
2343 int
2344 sys_access(struct access_args *uap)
2345 {
2346         struct nlookupdata nd;
2347         int error;
2348
2349         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2350         if (error == 0)
2351                 error = kern_access(&nd, uap->flags);
2352         nlookup_done(&nd);
2353         return (error);
2354 }
2355
2356 int
2357 kern_stat(struct nlookupdata *nd, struct stat *st)
2358 {
2359         int error;
2360         struct vnode *vp;
2361         thread_t td;
2362
2363         if ((error = nlookup(nd)) != 0)
2364                 return (error);
2365 again:
2366         if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
2367                 return (ENOENT);
2368
2369         td = curthread;
2370         if ((error = vget(vp, LK_SHARED)) != 0)
2371                 return (error);
2372         error = vn_stat(vp, st, nd->nl_cred);
2373
2374         /*
2375          * If the file handle is stale we have to re-resolve the entry.  This
2376          * is a hack at the moment.
2377          */
2378         if (error == ESTALE) {
2379                 vput(vp);
2380                 cache_setunresolved(&nd->nl_nch);
2381                 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
2382                 if (error == 0)
2383                         goto again;
2384         } else {
2385                 vput(vp);
2386         }
2387         return (error);
2388 }
2389
2390 /*
2391  * stat_args(char *path, struct stat *ub)
2392  *
2393  * Get file status; this version follows links.
2394  */
2395 int
2396 sys_stat(struct stat_args *uap)
2397 {
2398         struct nlookupdata nd;
2399         struct stat st;
2400         int error;
2401
2402         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2403         if (error == 0) {
2404                 error = kern_stat(&nd, &st);
2405                 if (error == 0)
2406                         error = copyout(&st, uap->ub, sizeof(*uap->ub));
2407         }
2408         nlookup_done(&nd);
2409         return (error);
2410 }
2411
2412 /*
2413  * lstat_args(char *path, struct stat *ub)
2414  *
2415  * Get file status; this version does not follow links.
2416  */
2417 int
2418 sys_lstat(struct lstat_args *uap)
2419 {
2420         struct nlookupdata nd;
2421         struct stat st;
2422         int error;
2423
2424         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2425         if (error == 0) {
2426                 error = kern_stat(&nd, &st);
2427                 if (error == 0)
2428                         error = copyout(&st, uap->ub, sizeof(*uap->ub));
2429         }
2430         nlookup_done(&nd);
2431         return (error);
2432 }
2433
2434 /*
2435  * pathconf_Args(char *path, int name)
2436  *
2437  * Get configurable pathname variables.
2438  */
2439 /* ARGSUSED */
2440 int
2441 sys_pathconf(struct pathconf_args *uap)
2442 {
2443         struct nlookupdata nd;
2444         struct vnode *vp;
2445         int error;
2446
2447         vp = NULL;
2448         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2449         if (error == 0)
2450                 error = nlookup(&nd);
2451         if (error == 0)
2452                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
2453         nlookup_done(&nd);
2454         if (error == 0) {
2455                 error = VOP_PATHCONF(vp, uap->name, uap->sysmsg_fds);
2456                 vput(vp);
2457         }
2458         return (error);
2459 }
2460
2461 /*
2462  * XXX: daver
2463  * kern_readlink isn't properly split yet.  There is a copyin burried
2464  * in VOP_READLINK().
2465  */
2466 int
2467 kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
2468 {
2469         struct thread *td = curthread;
2470         struct proc *p = td->td_proc;
2471         struct vnode *vp;
2472         struct iovec aiov;
2473         struct uio auio;
2474         int error;
2475
2476         if ((error = nlookup(nd)) != 0)
2477                 return (error);
2478         error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
2479         if (error)
2480                 return (error);
2481         if (vp->v_type != VLNK) {
2482                 error = EINVAL;
2483         } else {
2484                 aiov.iov_base = buf;
2485                 aiov.iov_len = count;
2486                 auio.uio_iov = &aiov;
2487                 auio.uio_iovcnt = 1;
2488                 auio.uio_offset = 0;
2489                 auio.uio_rw = UIO_READ;
2490                 auio.uio_segflg = UIO_USERSPACE;
2491                 auio.uio_td = td;
2492                 auio.uio_resid = count;
2493                 error = VOP_READLINK(vp, &auio, p->p_ucred);
2494         }
2495         vput(vp);
2496         *res = count - auio.uio_resid;
2497         return (error);
2498 }
2499
2500 /*
2501  * readlink_args(char *path, char *buf, int count)
2502  *
2503  * Return target name of a symbolic link.
2504  */
2505 int
2506 sys_readlink(struct readlink_args *uap)
2507 {
2508         struct nlookupdata nd;
2509         int error;
2510
2511         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2512         if (error == 0) {
2513                 error = kern_readlink(&nd, uap->buf, uap->count,
2514                                         &uap->sysmsg_result);
2515         }
2516         nlookup_done(&nd);
2517         return (error);
2518 }
2519
2520 static int
2521 setfflags(struct vnode *vp, int flags)
2522 {
2523         struct thread *td = curthread;
2524         struct proc *p = td->td_proc;
2525         int error;
2526         struct vattr vattr;
2527
2528         /*
2529          * Prevent non-root users from setting flags on devices.  When
2530          * a device is reused, users can retain ownership of the device
2531          * if they are allowed to set flags and programs assume that
2532          * chown can't fail when done as root.
2533          */
2534         if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
2535             ((error = priv_check_cred(p->p_ucred, PRIV_ROOT, PRISON_ROOT)) != 0))
2536                 return (error);
2537
2538         /*
2539          * note: vget is required for any operation that might mod the vnode
2540          * so VINACTIVE is properly cleared.
2541          */
2542         if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2543                 VATTR_NULL(&vattr);
2544                 vattr.va_flags = flags;
2545                 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2546                 vput(vp);
2547         }
2548         return (error);
2549 }
2550
2551 /*
2552  * chflags(char *path, int flags)
2553  *
2554  * Change flags of a file given a path name.
2555  */
2556 /* ARGSUSED */
2557 int
2558 sys_chflags(struct chflags_args *uap)
2559 {
2560         struct nlookupdata nd;
2561         struct vnode *vp;
2562         int error;
2563
2564         vp = NULL;
2565         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2566         /* XXX Add NLC flag indicating modifying operation? */
2567         if (error == 0)
2568                 error = nlookup(&nd);
2569         if (error == 0)
2570                 error = ncp_writechk(&nd.nl_nch);
2571         if (error == 0)
2572                 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
2573         nlookup_done(&nd);
2574         if (error == 0) {
2575                 error = setfflags(vp, uap->flags);
2576                 vrele(vp);
2577         }
2578         return (error);
2579 }
2580
2581 /*
2582  * lchflags(char *path, int flags)
2583  *
2584  * Change flags of a file given a path name, but don't follow symlinks.
2585  */
2586 /* ARGSUSED */
2587 int
2588 sys_lchflags(struct lchflags_args *uap)
2589 {
2590         struct nlookupdata nd;
2591         struct vnode *vp;
2592         int error;
2593
2594         vp = NULL;
2595         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2596         /* XXX Add NLC flag indicating modifying operation? */
2597         if (error == 0)
2598                 error = nlookup(&nd);
2599         if (error == 0)
2600                 error = ncp_writechk(&nd.nl_nch);
2601         if (error == 0)
2602                 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
2603         nlookup_done(&nd);
2604         if (error == 0) {
2605                 error = setfflags(vp, uap->flags);
2606                 vrele(vp);
2607         }
2608         return (error);
2609 }
2610
2611 /*
2612  * fchflags_args(int fd, int flags)
2613  *
2614  * Change flags of a file given a file descriptor.
2615  */
2616 /* ARGSUSED */
2617 int
2618 sys_fchflags(struct fchflags_args *uap)
2619 {
2620         struct thread *td = curthread;
2621         struct proc *p = td->td_proc;
2622         struct file *fp;
2623         int error;
2624
2625         if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2626                 return (error);
2627         if (fp->f_nchandle.ncp)
2628                 error = ncp_writechk(&fp->f_nchandle);
2629         if (error == 0)
2630                 error = setfflags((struct vnode *) fp->f_data, uap->flags);
2631         fdrop(fp);
2632         return (error);
2633 }
2634
2635 static int
2636 setfmode(struct vnode *vp, int mode)
2637 {
2638         struct thread *td = curthread;
2639         struct proc *p = td->td_proc;
2640         int error;
2641         struct vattr vattr;
2642
2643         /*
2644          * note: vget is required for any operation that might mod the vnode
2645          * so VINACTIVE is properly cleared.
2646          */
2647         if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2648                 VATTR_NULL(&vattr);
2649                 vattr.va_mode = mode & ALLPERMS;
2650                 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2651                 vput(vp);
2652         }
2653         return error;
2654 }
2655
2656 int
2657 kern_chmod(struct nlookupdata *nd, int mode)
2658 {
2659         struct vnode *vp;
2660         int error;
2661
2662         /* XXX Add NLC flag indicating modifying operation? */
2663         if ((error = nlookup(nd)) != 0)
2664                 return (error);
2665         if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
2666                 return (error);
2667         if ((error = ncp_writechk(&nd->nl_nch)) == 0)
2668                 error = setfmode(vp, mode);
2669         vrele(vp);
2670         return (error);
2671 }
2672
2673 /*
2674  * chmod_args(char *path, int mode)
2675  *
2676  * Change mode of a file given path name.
2677  */
2678 /* ARGSUSED */
2679 int
2680 sys_chmod(struct chmod_args *uap)
2681 {
2682         struct nlookupdata nd;
2683         int error;
2684
2685         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2686         if (error == 0)
2687                 error = kern_chmod(&nd, uap->mode);
2688         nlookup_done(&nd);
2689         return (error);
2690 }
2691
2692 /*
2693  * lchmod_args(char *path, int mode)
2694  *
2695  * Change mode of a file given path name (don't follow links.)
2696  */
2697 /* ARGSUSED */
2698 int
2699 sys_lchmod(struct lchmod_args *uap)
2700 {
2701         struct nlookupdata nd;
2702         int error;
2703
2704         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2705         if (error == 0)
2706                 error = kern_chmod(&nd, uap->mode);
2707         nlookup_done(&nd);
2708         return (error);
2709 }
2710
2711 /*
2712  * fchmod_args(int fd, int mode)
2713  *
2714  * Change mode of a file given a file descriptor.
2715  */
2716 /* ARGSUSED */
2717 int
2718 sys_fchmod(struct fchmod_args *uap)
2719 {
2720         struct thread *td = curthread;
2721         struct proc *p = td->td_proc;
2722         struct file *fp;
2723         int error;
2724
2725         if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2726                 return (error);
2727         if (fp->f_nchandle.ncp)
2728                 error = ncp_writechk(&fp->f_nchandle);
2729         if (error == 0)
2730                 error = setfmode((struct vnode *)fp->f_data, uap->mode);
2731         fdrop(fp);
2732         return (error);
2733 }
2734
2735 static int
2736 setfown(struct vnode *vp, uid_t uid, gid_t gid)
2737 {
2738         struct thread *td = curthread;
2739         struct proc *p = td->td_proc;
2740         int error;
2741         struct vattr vattr;
2742
2743         /*
2744          * note: vget is required for any operation that might mod the vnode
2745          * so VINACTIVE is properly cleared.
2746          */
2747         if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2748                 VATTR_NULL(&vattr);
2749                 vattr.va_uid = uid;
2750                 vattr.va_gid = gid;
2751                 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2752                 vput(vp);
2753         }
2754         return error;
2755 }
2756
2757 int
2758 kern_chown(struct nlookupdata *nd, int uid, int gid)
2759 {
2760         struct vnode *vp;
2761         int error;
2762
2763         /* XXX Add NLC flag indicating modifying operation? */
2764         if ((error = nlookup(nd)) != 0)
2765                 return (error);
2766         if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
2767                 return (error);
2768         if ((error = ncp_writechk(&nd->nl_nch)) == 0)
2769                 error = setfown(vp, uid, gid);
2770         vrele(vp);
2771         return (error);
2772 }
2773
2774 /*
2775  * chown(char *path, int uid, int gid)
2776  *
2777  * Set ownership given a path name.
2778  */
2779 int
2780 sys_chown(struct chown_args *uap)
2781 {
2782         struct nlookupdata nd;
2783         int error;
2784
2785         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2786         if (error == 0)
2787                 error = kern_chown(&nd, uap->uid, uap->gid);
2788         nlookup_done(&nd);
2789         return (error);
2790 }
2791
2792 /*
2793  * lchown_args(char *path, int uid, int gid)
2794  *
2795  * Set ownership given a path name, do not cross symlinks.
2796  */
2797 int
2798 sys_lchown(struct lchown_args *uap)
2799 {
2800         struct nlookupdata nd;
2801         int error;
2802
2803         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2804         if (error == 0)
2805                 error = kern_chown(&nd, uap->uid, uap->gid);
2806         nlookup_done(&nd);
2807         return (error);
2808 }
2809
2810 /*
2811  * fchown_args(int fd, int uid, int gid)
2812  *
2813  * Set ownership given a file descriptor.
2814  */
2815 /* ARGSUSED */
2816 int
2817 sys_fchown(struct fchown_args *uap)
2818 {
2819         struct thread *td = curthread;
2820         struct proc *p = td->td_proc;
2821         struct file *fp;
2822         int error;
2823
2824         if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2825                 return (error);
2826         if (fp->f_nchandle.ncp)
2827                 error = ncp_writechk(&fp->f_nchandle);
2828         if (error == 0)
2829                 error = setfown((struct vnode *)fp->f_data, uap->uid, uap->gid);
2830         fdrop(fp);
2831         return (error);
2832 }
2833
2834 static int
2835 getutimes(const struct timeval *tvp, struct timespec *tsp)
2836 {
2837         struct timeval tv[2];
2838
2839         if (tvp == NULL) {
2840                 microtime(&tv[0]);
2841                 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2842                 tsp[1] = tsp[0];
2843         } else {
2844                 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2845                 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2846         }
2847         return 0;
2848 }
2849
2850 static int
2851 setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
2852 {
2853         struct thread *td = curthread;
2854         struct proc *p = td->td_proc;
2855         int error;
2856         struct vattr vattr;
2857
2858         /*
2859          * note: vget is required for any operation that might mod the vnode
2860          * so VINACTIVE is properly cleared.
2861          */
2862         if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2863                 VATTR_NULL(&vattr);
2864                 vattr.va_atime = ts[0];
2865                 vattr.va_mtime = ts[1];
2866                 if (nullflag)
2867                         vattr.va_vaflags |= VA_UTIMES_NULL;
2868                 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2869                 vput(vp);
2870         }
2871         return error;
2872 }
2873
2874 int
2875 kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
2876 {
2877         struct timespec ts[2];
2878         struct vnode *vp;
2879         int error;
2880
2881         if ((error = getutimes(tptr, ts)) != 0)
2882                 return (error);
2883         /* XXX Add NLC flag indicating modifying operation? */
2884         if ((error = nlookup(nd)) != 0)
2885                 return (error);
2886         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2887                 return (error);
2888         if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
2889                 return (error);
2890         error = setutimes(vp, ts, tptr == NULL);
2891         vrele(vp);
2892         return (error);
2893 }
2894
2895 /*
2896  * utimes_args(char *path, struct timeval *tptr)
2897  *
2898  * Set the access and modification times of a file.
2899  */
2900 int
2901 sys_utimes(struct utimes_args *uap)
2902 {
2903         struct timeval tv[2];
2904         struct nlookupdata nd;
2905         int error;
2906
2907         if (uap->tptr) {
2908                 error = copyin(uap->tptr, tv, sizeof(tv));
2909                 if (error)
2910                         return (error);
2911         }
2912         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2913         if (error == 0)
2914                 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2915         nlookup_done(&nd);
2916         return (error);
2917 }
2918
2919 /*
2920  * lutimes_args(char *path, struct timeval *tptr)
2921  *
2922  * Set the access and modification times of a file.
2923  */
2924 int
2925 sys_lutimes(struct lutimes_args *uap)
2926 {
2927         struct timeval tv[2];
2928         struct nlookupdata nd;
2929         int error;
2930
2931         if (uap->tptr) {
2932                 error = copyin(uap->tptr, tv, sizeof(tv));
2933                 if (error)
2934                         return (error);
2935         }
2936         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2937         if (error == 0)
2938                 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2939         nlookup_done(&nd);
2940         return (error);
2941 }
2942
2943 int
2944 kern_futimes(int fd, struct timeval *tptr)
2945 {
2946         struct thread *td = curthread;
2947         struct proc *p = td->td_proc;
2948         struct timespec ts[2];
2949         struct file *fp;
2950         int error;
2951
2952         error = getutimes(tptr, ts);
2953         if (error)
2954                 return (error);
2955         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
2956                 return (error);
2957         if (fp->f_nchandle.ncp)
2958                 error = ncp_writechk(&fp->f_nchandle);
2959         if (error == 0)
2960                 error =  setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
2961         fdrop(fp);
2962         return (error);
2963 }
2964
2965 /*
2966  * futimes_args(int fd, struct timeval *tptr)
2967  *
2968  * Set the access and modification times of a file.
2969  */
2970 int
2971 sys_futimes(struct futimes_args *uap)
2972 {
2973         struct timeval tv[2];
2974         int error;
2975
2976         if (uap->tptr) {
2977                 error = copyin(uap->tptr, tv, sizeof(tv));
2978                 if (error)
2979                         return (error);
2980         }
2981
2982         error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
2983
2984         return (error);
2985 }
2986
2987 int
2988 kern_truncate(struct nlookupdata *nd, off_t length)
2989 {
2990         struct vnode *vp;
2991         struct vattr vattr;
2992         int error;
2993
2994         if (length < 0)
2995                 return(EINVAL);
2996         /* XXX Add NLC flag indicating modifying operation? */
2997         if ((error = nlookup(nd)) != 0)
2998                 return (error);
2999         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
3000                 return (error);
3001         if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
3002                 return (error);
3003         if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) != 0) {
3004                 vrele(vp);
3005                 return (error);
3006         }
3007         if (vp->v_type == VDIR) {
3008                 error = EISDIR;
3009         } else if ((error = vn_writechk(vp, &nd->nl_nch)) == 0 &&
3010             (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred)) == 0) {
3011                 VATTR_NULL(&vattr);
3012                 vattr.va_size = length;
3013                 error = VOP_SETATTR(vp, &vattr, nd->nl_cred);
3014         }
3015         vput(vp);
3016         return (error);
3017 }
3018
3019 /*
3020  * truncate(char *path, int pad, off_t length)
3021  *
3022  * Truncate a file given its path name.
3023  */
3024 int
3025 sys_truncate(struct truncate_args *uap)
3026 {
3027         struct nlookupdata nd;
3028         int error;
3029
3030         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3031         if (error == 0)
3032                 error = kern_truncate(&nd, uap->length);
3033         nlookup_done(&nd);
3034         return error;
3035 }
3036
3037 int
3038 kern_ftruncate(int fd, off_t length)
3039 {
3040         struct thread *td = curthread;
3041         struct proc *p = td->td_proc;
3042         struct vattr vattr;
3043         struct vnode *vp;
3044         struct file *fp;
3045         int error;
3046
3047         if (length < 0)
3048                 return(EINVAL);
3049         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
3050                 return (error);
3051         if (fp->f_nchandle.ncp) {
3052                 error = ncp_writechk(&fp->f_nchandle);
3053                 if (error)
3054                         goto done;
3055         }
3056         if ((fp->f_flag & FWRITE) == 0) {
3057                 error = EINVAL;
3058                 goto done;
3059         }
3060         vp = (struct vnode *)fp->f_data;
3061         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3062         if (vp->v_type == VDIR) {
3063                 error = EISDIR;
3064         } else if ((error = vn_writechk(vp, NULL)) == 0) {
3065                 VATTR_NULL(&vattr);
3066                 vattr.va_size = length;
3067                 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3068         }
3069         vn_unlock(vp);
3070 done:
3071         fdrop(fp);
3072         return (error);
3073 }
3074
3075 /*
3076  * ftruncate_args(int fd, int pad, off_t length)
3077  *
3078  * Truncate a file given a file descriptor.
3079  */
3080 int
3081 sys_ftruncate(struct ftruncate_args *uap)
3082 {
3083         int error;
3084
3085         error = kern_ftruncate(uap->fd, uap->length);
3086
3087         return (error);
3088 }
3089
3090 /*
3091  * fsync(int fd)
3092  *
3093  * Sync an open file.
3094  */
3095 /* ARGSUSED */
3096 int
3097 sys_fsync(struct fsync_args *uap)
3098 {
3099         struct thread *td = curthread;
3100         struct proc *p = td->td_proc;
3101         struct vnode *vp;
3102         struct file *fp;
3103         vm_object_t obj;
3104         int error;
3105
3106         if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
3107                 return (error);
3108         vp = (struct vnode *)fp->f_data;
3109         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3110         if ((obj = vp->v_object) != NULL)
3111                 vm_object_page_clean(obj, 0, 0, 0);
3112         if ((error = VOP_FSYNC(vp, MNT_WAIT)) == 0 && vp->v_mount)
3113                 error = buf_fsync(vp);
3114         vn_unlock(vp);
3115         fdrop(fp);
3116         return (error);
3117 }
3118
3119 int
3120 kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
3121 {
3122         struct nchandle fnchd;
3123         struct nchandle tnchd;
3124         struct namecache *ncp;
3125         struct vnode *fdvp;
3126         struct vnode *tdvp;
3127         struct mount *mp;
3128         int error;
3129
3130         bwillinode(1);
3131         fromnd->nl_flags |= NLC_REFDVP;
3132         if ((error = nlookup(fromnd)) != 0)
3133                 return (error);
3134         if ((fnchd.ncp = fromnd->nl_nch.ncp->nc_parent) == NULL)
3135                 return (ENOENT);
3136         fnchd.mount = fromnd->nl_nch.mount;
3137         cache_hold(&fnchd);
3138
3139         /*
3140          * unlock the source nch so we can lookup the target nch without
3141          * deadlocking.  The target may or may not exist so we do not check
3142          * for a target vp like kern_mkdir() and other creation functions do.
3143          *
3144          * The source and target directories are ref'd and rechecked after
3145          * everything is relocked to determine if the source or target file
3146          * has been renamed.
3147          */
3148         KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
3149         fromnd->nl_flags &= ~NLC_NCPISLOCKED;
3150         cache_unlock(&fromnd->nl_nch);
3151
3152         tond->nl_flags |= NLC_CREATE | NLC_REFDVP;
3153         if ((error = nlookup(tond)) != 0) {
3154                 cache_drop(&fnchd);
3155                 return (error);
3156         }
3157         if ((tnchd.ncp = tond->nl_nch.ncp->nc_parent) == NULL) {
3158                 cache_drop(&fnchd);
3159                 return (ENOENT);
3160         }
3161         tnchd.mount = tond->nl_nch.mount;
3162         cache_hold(&tnchd);
3163
3164         /*
3165          * If the source and target are the same there is nothing to do
3166          */
3167         if (fromnd->nl_nch.ncp == tond->nl_nch.ncp) {
3168                 cache_drop(&fnchd);
3169                 cache_drop(&tnchd);
3170                 return (0);
3171         }
3172
3173         /*
3174          * Mount points cannot be renamed or overwritten
3175          */
3176         if ((fromnd->nl_nch.ncp->nc_flag | tond->nl_nch.ncp->nc_flag) &
3177             NCF_ISMOUNTPT
3178         ) {
3179                 cache_drop(&fnchd);
3180                 cache_drop(&tnchd);
3181                 return (EINVAL);
3182         }
3183
3184         /*
3185          * relock the source ncp.  NOTE AFTER RELOCKING: the source ncp
3186          * may have become invalid while it was unlocked, nc_vp and nc_mount
3187          * could be NULL.
3188          */
3189         if (cache_lock_nonblock(&fromnd->nl_nch) == 0) {
3190                 cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
3191         } else if (fromnd->nl_nch.ncp > tond->nl_nch.ncp) {
3192                 cache_lock(&fromnd->nl_nch);
3193                 cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
3194         } else {
3195                 cache_unlock(&tond->nl_nch);
3196                 cache_lock(&fromnd->nl_nch);
3197                 cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
3198                 cache_lock(&tond->nl_nch);
3199                 cache_resolve(&tond->nl_nch, tond->nl_cred);
3200         }
3201         fromnd->nl_flags |= NLC_NCPISLOCKED;
3202
3203         /*
3204          * make sure the parent directories linkages are the same
3205          */
3206         if (fnchd.ncp != fromnd->nl_nch.ncp->nc_parent ||
3207             tnchd.ncp != tond->nl_nch.ncp->nc_parent) {
3208                 cache_drop(&fnchd);
3209                 cache_drop(&tnchd);
3210                 return (ENOENT);
3211         }
3212
3213         /*
3214          * Both the source and target must be within the same filesystem and
3215          * in the same filesystem as their parent directories within the
3216          * namecache topology.
3217          *
3218          * NOTE: fromnd's nc_mount or nc_vp could be NULL.
3219          */
3220         mp = fnchd.mount;
3221         if (mp != tnchd.mount || mp != fromnd->nl_nch.mount ||
3222             mp != tond->nl_nch.mount) {
3223                 cache_drop(&fnchd);
3224                 cache_drop(&tnchd);
3225                 return (EXDEV);
3226         }
3227
3228         /*
3229          * Make sure the mount point is writable
3230          */
3231         if ((error = ncp_writechk(&tond->nl_nch)) != 0) {
3232                 cache_drop(&fnchd);
3233                 cache_drop(&tnchd);
3234                 return (error);
3235         }
3236
3237         /*
3238          * If the target exists and either the source or target is a directory,
3239          * then both must be directories.
3240          *
3241          * Due to relocking of the source, fromnd->nl_nch.ncp->nc_vp might h
3242          * have become NULL.
3243          */
3244         if (tond->nl_nch.ncp->nc_vp) {
3245                 if (fromnd->nl_nch.ncp->nc_vp == NULL) {
3246                         error = ENOENT;
3247                 } else if (fromnd->nl_nch.ncp->nc_vp->v_type == VDIR) {
3248                         if (tond->nl_nch.ncp->nc_vp->v_type != VDIR)
3249                                 error = ENOTDIR;
3250                 } else if (tond->nl_nch.ncp->nc_vp->v_type == VDIR) {
3251                         error = EISDIR;
3252                 }
3253         }
3254
3255         /*
3256          * You cannot rename a source into itself or a subdirectory of itself.
3257          * We check this by travsersing the target directory upwards looking
3258          * for a match against the source.
3259          */
3260         if (error == 0) {
3261                 for (ncp = tnchd.ncp; ncp; ncp = ncp->nc_parent) {
3262                         if (fromnd->nl_nch.ncp == ncp) {
3263                                 error = EINVAL;
3264                                 break;
3265                         }
3266                 }
3267         }
3268
3269         cache_drop(&fnchd);
3270         cache_drop(&tnchd);
3271
3272         /*
3273          * Even though the namespaces are different, they may still represent
3274          * hardlinks to the same file.  The filesystem might have a hard time
3275          * with this so we issue a NREMOVE of the source instead of a NRENAME
3276          * when we detect the situation.
3277          */
3278         if (error == 0) {
3279                 fdvp = fromnd->nl_dvp;
3280                 tdvp = tond->nl_dvp;
3281                 if (fdvp == NULL || tdvp == NULL) {
3282                         error = EPERM;
3283                 } else if (fromnd->nl_nch.ncp->nc_vp == tond->nl_nch.ncp->nc_vp) {
3284                         error = VOP_NREMOVE(&fromnd->nl_nch, fdvp,
3285                                             fromnd->nl_cred);
3286                 } else {
3287                         error = VOP_NRENAME(&fromnd->nl_nch, &tond->nl_nch,
3288                                             fdvp, tdvp, tond->nl_cred);
3289                 }
3290         }
3291         return (error);
3292 }
3293
3294 /*
3295  * rename_args(char *from, char *to)
3296  *
3297  * Rename files.  Source and destination must either both be directories,
3298  * or both not be directories.  If target is a directory, it must be empty.
3299  */
3300 int
3301 sys_rename(struct rename_args *uap)
3302 {
3303         struct nlookupdata fromnd, tond;
3304         int error;
3305
3306         error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
3307         if (error == 0) {
3308                 error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
3309                 if (error == 0)
3310                         error = kern_rename(&fromnd, &tond);
3311                 nlookup_done(&tond);
3312         }
3313         nlookup_done(&fromnd);
3314         return (error);
3315 }
3316
3317 int
3318 kern_mkdir(struct nlookupdata *nd, int mode)
3319 {
3320         struct thread *td = curthread;
3321         struct proc *p = td->td_proc;
3322         struct vnode *vp;
3323         struct vattr vattr;
3324         int error;
3325
3326         bwillinode(1);
3327         nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE | NLC_REFDVP;
3328         if ((error = nlookup(nd)) != 0)
3329                 return (error);
3330
3331         if (nd->nl_nch.ncp->nc_vp)
3332                 return (EEXIST);
3333         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
3334                 return (error);
3335         VATTR_NULL(&vattr);
3336         vattr.va_type = VDIR;
3337         vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
3338
3339         vp = NULL;
3340         error = VOP_NMKDIR(&nd->nl_nch, nd->nl_dvp, &vp, p->p_ucred, &vattr);
3341         if (error == 0)
3342                 vput(vp);
3343         return (error);
3344 }
3345
3346 /*
3347  * mkdir_args(char *path, int mode)
3348  *
3349  * Make a directory file.
3350  */
3351 /* ARGSUSED */
3352 int
3353 sys_mkdir(struct mkdir_args *uap)
3354 {
3355         struct nlookupdata nd;
3356         int error;
3357
3358         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3359         if (error == 0)
3360                 error = kern_mkdir(&nd, uap->mode);
3361         nlookup_done(&nd);
3362         return (error);
3363 }
3364
3365 int
3366 kern_rmdir(struct nlookupdata *nd)
3367 {
3368         int error;
3369
3370         bwillinode(1);
3371         nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
3372         if ((error = nlookup(nd)) != 0)
3373                 return (error);
3374
3375         /*
3376          * Do not allow directories representing mount points to be
3377          * deleted, even if empty.  Check write perms on mount point
3378          * in case the vnode is aliased (aka nullfs).
3379          */
3380         if (nd->nl_nch.ncp->nc_flag & (NCF_ISMOUNTPT))
3381                 return (EINVAL);
3382         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
3383                 return (error);
3384         error = VOP_NRMDIR(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
3385         return (error);
3386 }
3387
3388 /*
3389  * rmdir_args(char *path)
3390  *
3391  * Remove a directory file.
3392  */
3393 /* ARGSUSED */
3394 int
3395 sys_rmdir(struct rmdir_args *uap)
3396 {
3397         struct nlookupdata nd;
3398         int error;
3399
3400         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3401         if (error == 0)
3402                 error = kern_rmdir(&nd);
3403         nlookup_done(&nd);
3404         return (error);
3405 }
3406
3407 int
3408 kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res,
3409     enum uio_seg direction)
3410 {
3411         struct thread *td = curthread;
3412         struct proc *p = td->td_proc;
3413         struct vnode *vp;
3414         struct file *fp;
3415         struct uio auio;
3416         struct iovec aiov;
3417         off_t loff;
3418         int error, eofflag;
3419
3420         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
3421                 return (error);
3422         if ((fp->f_flag & FREAD) == 0) {
3423                 error = EBADF;
3424                 goto done;
3425         }
3426         vp = (struct vnode *)fp->f_data;
3427 unionread:
3428         if (vp->v_type != VDIR) {
3429                 error = EINVAL;
3430                 goto done;
3431         }
3432         aiov.iov_base = buf;
3433         aiov.iov_len = count;
3434         auio.uio_iov = &aiov;
3435         auio.uio_iovcnt = 1;
3436         auio.uio_rw = UIO_READ;
3437         auio.uio_segflg = direction;
3438         auio.uio_td = td;
3439         auio.uio_resid = count;
3440         loff = auio.uio_offset = fp->f_offset;
3441         error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
3442         fp->f_offset = auio.uio_offset;
3443         if (error)
3444                 goto done;
3445         if (count == auio.uio_resid) {
3446                 if (union_dircheckp) {
3447                         error = union_dircheckp(td, &vp, fp);
3448                         if (error == -1)
3449                                 goto unionread;
3450                         if (error)
3451                                 goto done;
3452                 }
3453 #if 0
3454                 if ((vp->v_flag & VROOT) &&
3455                     (vp->v_mount->mnt_flag & MNT_UNION)) {
3456                         struct vnode *tvp = vp;
3457                         vp = vp->v_mount->mnt_vnodecovered;
3458                         vref(vp);
3459                         fp->f_data = vp;
3460                         fp->f_offset = 0;
3461                         vrele(tvp);
3462                         goto unionread;
3463                 }
3464 #endif
3465         }
3466
3467         /*
3468          * WARNING!  *basep may not be wide enough to accomodate the
3469          * seek offset.   XXX should we hack this to return the upper 32 bits
3470          * for offsets greater then 4G?
3471          */
3472         if (basep) {
3473                 *basep = (long)loff;
3474         }
3475         *res = count - auio.uio_resid;
3476 done:
3477         fdrop(fp);
3478         return (error);
3479 }
3480
3481 /*
3482  * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
3483  *
3484  * Read a block of directory entries in a file system independent format.
3485  */
3486 int
3487 sys_getdirentries(struct getdirentries_args *uap)
3488 {
3489         long base;
3490         int error;
3491
3492         error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base,
3493                                    &uap->sysmsg_result, UIO_USERSPACE);
3494
3495         if (error == 0 && uap->basep)
3496                 error = copyout(&base, uap->basep, sizeof(*uap->basep));
3497         return (error);
3498 }
3499
3500 /*
3501  * getdents_args(int fd, char *buf, size_t count)
3502  */
3503 int
3504 sys_getdents(struct getdents_args *uap)
3505 {
3506         int error;
3507
3508         error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL,
3509             &uap->sysmsg_result, UIO_USERSPACE);
3510
3511         return (error);
3512 }
3513
3514 /*
3515  * umask(int newmask)
3516  *
3517  * Set the mode mask for creation of filesystem nodes.
3518  *
3519  * MP SAFE
3520  */
3521 int
3522 sys_umask(struct umask_args *uap)
3523 {
3524         struct thread *td = curthread;
3525         struct proc *p = td->td_proc;
3526         struct filedesc *fdp;
3527
3528         fdp = p->p_fd;
3529         uap->sysmsg_result = fdp->fd_cmask;
3530         fdp->fd_cmask = uap->newmask & ALLPERMS;
3531         return (0);
3532 }
3533
3534 /*
3535  * revoke(char *path)
3536  *
3537  * Void all references to file by ripping underlying filesystem
3538  * away from vnode.
3539  */
3540 /* ARGSUSED */
3541 int
3542 sys_revoke(struct revoke_args *uap)
3543 {
3544         struct nlookupdata nd;
3545         struct vattr vattr;
3546         struct vnode *vp;
3547         struct ucred *cred;
3548         int error;
3549
3550         vp = NULL;
3551         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3552         if (error == 0)
3553                 error = nlookup(&nd);
3554         if (error == 0)
3555                 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
3556         cred = crhold(nd.nl_cred);
3557         nlookup_done(&nd);
3558         if (error == 0) {
3559                 if (vp->v_type != VCHR && vp->v_type != VBLK)
3560                         error = EINVAL;
3561                 if (error == 0)
3562                         error = VOP_GETATTR(vp, &vattr);
3563                 if (error == 0 && cred->cr_uid != vattr.va_uid)
3564                         error = priv_check_cred(cred, PRIV_ROOT, PRISON_ROOT);
3565                 if (error == 0 && count_udev(vp->v_umajor, vp->v_uminor) > 0) {
3566                         error = 0;
3567                         vx_lock(vp);
3568                         VOP_REVOKE(vp, REVOKEALL);
3569                         vx_unlock(vp);
3570                 }
3571                 vrele(vp);
3572         }
3573         if (cred)
3574                 crfree(cred);
3575         return (error);
3576 }
3577
3578 /*
3579  * getfh_args(char *fname, fhandle_t *fhp)
3580  *
3581  * Get (NFS) file handle
3582  *
3583  * NOTE: We use the fsid of the covering mount, even if it is a nullfs
3584  * mount.  This allows nullfs mounts to be explicitly exported.
3585  *
3586  * WARNING: nullfs mounts of HAMMER PFS ROOTs are safe.
3587  *
3588  *          nullfs mounts of subdirectories are not safe.  That is, it will
3589  *          work, but you do not really have protection against access to
3590  *          the related parent directories.
3591  */
3592 int
3593 sys_getfh(struct getfh_args *uap)
3594 {
3595         struct thread *td = curthread;
3596         struct nlookupdata nd;
3597         fhandle_t fh;
3598         struct vnode *vp;
3599         struct mount *mp;
3600         int error;
3601
3602         /*
3603          * Must be super user
3604          */
3605         if ((error = priv_check(td, PRIV_ROOT)) != 0)
3606                 return (error);
3607
3608         vp = NULL;
3609         error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
3610         if (error == 0)
3611                 error = nlookup(&nd);
3612         if (error == 0)
3613                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
3614         mp = nd.nl_nch.mount;
3615         nlookup_done(&nd);
3616         if (error == 0) {
3617                 bzero(&fh, sizeof(fh));
3618                 fh.fh_fsid = mp->mnt_stat.f_fsid;
3619                 error = VFS_VPTOFH(vp, &fh.fh_fid);
3620                 vput(vp);
3621                 if (error == 0)
3622                         error = copyout(&fh, uap->fhp, sizeof(fh));
3623         }
3624         return (error);
3625 }
3626
3627 /*
3628  * fhopen_args(const struct fhandle *u_fhp, int flags)
3629  *
3630  * syscall for the rpc.lockd to use to translate a NFS file handle into
3631  * an open descriptor.
3632  *
3633  * warning: do not remove the priv_check() call or this becomes one giant
3634  * security hole.
3635  */
3636 int
3637 sys_fhopen(struct fhopen_args *uap)
3638 {
3639         struct thread *td = curthread;
3640         struct proc *p = td->td_proc;
3641         struct mount *mp;
3642         struct vnode *vp;
3643         struct fhandle fhp;
3644         struct vattr vat;
3645         struct vattr *vap = &vat;
3646         struct flock lf;
3647         int fmode, mode, error, type;
3648         struct file *nfp;
3649         struct file *fp;
3650         int indx;
3651
3652         /*
3653          * Must be super user
3654          */
3655         error = priv_check(td, PRIV_ROOT);
3656         if (error)
3657                 return (error);
3658
3659         fmode = FFLAGS(uap->flags);
3660         /* why not allow a non-read/write open for our lockd? */
3661         if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3662                 return (EINVAL);
3663         error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
3664         if (error)
3665                 return(error);
3666         /* find the mount point */
3667         mp = vfs_getvfs(&fhp.fh_fsid);
3668         if (mp == NULL)
3669                 return (ESTALE);
3670         /* now give me my vnode, it gets returned to me locked */
3671         error = VFS_FHTOVP(mp, NULL, &fhp.fh_fid, &vp);
3672         if (error)
3673                 return (error);
3674         /*
3675          * from now on we have to make sure not
3676          * to forget about the vnode
3677          * any error that causes an abort must vput(vp)
3678          * just set error = err and 'goto bad;'.
3679          */
3680
3681         /*
3682          * from vn_open
3683          */
3684         if (vp->v_type == VLNK) {
3685                 error = EMLINK;
3686                 goto bad;
3687         }
3688         if (vp->v_type == VSOCK) {
3689                 error = EOPNOTSUPP;
3690                 goto bad;
3691         }
3692         mode = 0;
3693         if (fmode & (FWRITE | O_TRUNC)) {
3694                 if (vp->v_type == VDIR) {
3695                         error = EISDIR;
3696                         goto bad;
3697                 }
3698                 error = vn_writechk(vp, NULL);
3699                 if (error)
3700                         goto bad;
3701                 mode |= VWRITE;
3702         }
3703         if (fmode & FREAD)
3704                 mode |= VREAD;
3705         if (mode) {
3706                 error = VOP_ACCESS(vp, mode, p->p_ucred);
3707                 if (error)
3708                         goto bad;
3709         }
3710         if (fmode & O_TRUNC) {
3711                 vn_unlock(vp);                          /* XXX */
3712                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
3713                 VATTR_NULL(vap);
3714                 vap->va_size = 0;
3715                 error = VOP_SETATTR(vp, vap, p->p_ucred);
3716                 if (error)
3717                         goto bad;
3718         }
3719
3720         /*
3721          * VOP_OPEN needs the file pointer so it can potentially override
3722          * it.
3723          *
3724          * WARNING! no f_nchandle will be associated when fhopen()ing a
3725          * directory.  XXX
3726          */
3727         if ((error = falloc(p, &nfp, &indx)) != 0)
3728                 goto bad;
3729         fp = nfp;
3730
3731         error = VOP_OPEN(vp, fmode, p->p_ucred, fp);
3732         if (error) {
3733                 /*
3734                  * setting f_ops this way prevents VOP_CLOSE from being
3735                  * called or fdrop() releasing the vp from v_data.   Since
3736                  * the VOP_OPEN failed we don't want to VOP_CLOSE.
3737                  */
3738                 fp->f_ops = &badfileops;
3739                 fp->f_data = NULL;
3740                 goto bad_drop;
3741         }
3742
3743         /*
3744          * The fp is given its own reference, we still have our ref and lock.
3745          *
3746          * Assert that all regular files must be created with a VM object.
3747          */
3748         if (vp->v_type == VREG && vp->v_object == NULL) {
3749                 kprintf("fhopen: regular file did not have VM object: %p\n", vp);
3750                 goto bad_drop;
3751         }
3752
3753         /*
3754          * The open was successful.  Handle any locking requirements.
3755          */
3756         if (fmode & (O_EXLOCK | O_SHLOCK)) {
3757                 lf.l_whence = SEEK_SET;
3758                 lf.l_start = 0;
3759                 lf.l_len = 0;
3760                 if (fmode & O_EXLOCK)
3761                         lf.l_type = F_WRLCK;
3762                 else
3763                         lf.l_type = F_RDLCK;
3764                 if (fmode & FNONBLOCK)
3765                         type = 0;
3766                 else
3767                         type = F_WAIT;
3768                 vn_unlock(vp);
3769                 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
3770                         /*
3771                          * release our private reference.
3772                          */
3773                         fsetfd(p, NULL, indx);
3774                         fdrop(fp);
3775                         vrele(vp);
3776                         return (error);
3777                 }
3778                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3779                 fp->f_flag |= FHASLOCK;
3780         }
3781
3782         /*
3783          * Clean up.  Associate the file pointer with the previously
3784          * reserved descriptor and return it.
3785          */
3786         vput(vp);
3787         fsetfd(p, fp, indx);
3788         fdrop(fp);
3789         uap->sysmsg_result = indx;
3790         return (0);
3791
3792 bad_drop:
3793         fsetfd(p, NULL, indx);
3794         fdrop(fp);
3795 bad:
3796         vput(vp);
3797         return (error);
3798 }
3799
3800 /*
3801  * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
3802  */
3803 int
3804 sys_fhstat(struct fhstat_args *uap)
3805 {
3806         struct thread *td = curthread;
3807         struct stat sb;
3808         fhandle_t fh;
3809         struct mount *mp;
3810         struct vnode *vp;
3811         int error;
3812
3813         /*
3814          * Must be super user
3815          */
3816         error = priv_check(td, PRIV_ROOT);
3817         if (error)
3818                 return (error);
3819
3820         error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
3821         if (error)
3822                 return (error);
3823
3824         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3825                 return (ESTALE);
3826         if ((error = VFS_FHTOVP(mp, NULL, &fh.fh_fid, &vp)))
3827                 return (error);
3828         error = vn_stat(vp, &sb, td->td_proc->p_ucred);
3829         vput(vp);
3830         if (error)
3831                 return (error);
3832         error = copyout(&sb, uap->sb, sizeof(sb));
3833         return (error);
3834 }
3835
3836 /*
3837  * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
3838  */
3839 int
3840 sys_fhstatfs(struct fhstatfs_args *uap)
3841 {
3842         struct thread *td = curthread;
3843         struct proc *p = td->td_proc;
3844         struct statfs *sp;
3845         struct mount *mp;
3846         struct vnode *vp;
3847         struct statfs sb;
3848         char *fullpath, *freepath;
3849         fhandle_t fh;
3850         int error;
3851
3852         /*
3853          * Must be super user
3854          */
3855         if ((error = priv_check(td, PRIV_ROOT)))
3856                 return (error);
3857
3858         if ((error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t))) != 0)
3859                 return (error);
3860
3861         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3862                 return (ESTALE);
3863
3864         if (p != NULL && !chroot_visible_mnt(mp, p))
3865                 return (ESTALE);
3866
3867         if ((error = VFS_FHTOVP(mp, NULL, &fh.fh_fid, &vp)))
3868                 return (error);
3869         mp = vp->v_mount;
3870         sp = &mp->mnt_stat;
3871         vput(vp);
3872         if ((error = VFS_STATFS(mp, sp, p->p_ucred)) != 0)
3873                 return (error);
3874
3875         error = mount_path(p, mp, &fullpath, &freepath);
3876         if (error)
3877                 return(error);
3878         bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3879         strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
3880         kfree(freepath, M_TEMP);
3881
3882         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
3883         if (priv_check(td, PRIV_ROOT)) {
3884                 bcopy(sp, &sb, sizeof(sb));
3885                 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
3886                 sp = &sb;
3887         }
3888         return (copyout(sp, uap->buf, sizeof(*sp)));
3889 }
3890
3891 /*
3892  * fhstatvfs_args(struct fhandle *u_fhp, struct statvfs *buf)
3893  */
3894 int
3895 sys_fhstatvfs(struct fhstatvfs_args *uap)
3896 {
3897         struct thread *td = curthread;
3898         struct proc *p = td->td_proc;
3899         struct statvfs *sp;
3900         struct mount *mp;
3901         struct vnode *vp;
3902         fhandle_t fh;
3903         int error;
3904
3905         /*
3906          * Must be super user
3907          */
3908         if ((error = priv_check(td, PRIV_ROOT)))
3909                 return (error);
3910
3911         if ((error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t))) != 0)
3912                 return (error);
3913
3914         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3915                 return (ESTALE);
3916
3917         if (p != NULL && !chroot_visible_mnt(mp, p))
3918                 return (ESTALE);
3919
3920         if ((error = VFS_FHTOVP(mp, NULL, &fh.fh_fid, &vp)))
3921                 return (error);
3922         mp = vp->v_mount;
3923         sp = &mp->mnt_vstat;
3924         vput(vp);
3925         if ((error = VFS_STATVFS(mp, sp, p->p_ucred)) != 0)
3926                 return (error);
3927
3928         sp->f_flag = 0;
3929         if (mp->mnt_flag & MNT_RDONLY)
3930                 sp->f_flag |= ST_RDONLY;
3931         if (mp->mnt_flag & MNT_NOSUID)
3932                 sp->f_flag |= ST_NOSUID;
3933
3934         return (copyout(sp, uap->buf, sizeof(*sp)));
3935 }
3936
3937
3938 /*
3939  * Syscall to push extended attribute configuration information into the
3940  * VFS.  Accepts a path, which it converts to a mountpoint, as well as
3941  * a command (int cmd), and attribute name and misc data.  For now, the
3942  * attribute name is left in userspace for consumption by the VFS_op.
3943  * It will probably be changed to be copied into sysspace by the
3944  * syscall in the future, once issues with various consumers of the
3945  * attribute code have raised their hands.
3946  *
3947  * Currently this is used only by UFS Extended Attributes.
3948  */
3949 int
3950 sys_extattrctl(struct extattrctl_args *uap)
3951 {
3952         struct nlookupdata nd;
3953         struct mount *mp;
3954         struct vnode *vp;
3955         int error;
3956
3957         vp = NULL;
3958         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3959         if (error == 0)
3960                 error = nlookup(&nd);
3961         if (error == 0) {
3962                 mp = nd.nl_nch.mount;
3963                 error = VFS_EXTATTRCTL(mp, uap->cmd,
3964                                 uap->attrname, uap->arg,
3965                                 nd.nl_cred);
3966         }
3967         nlookup_done(&nd);
3968         return (error);
3969 }
3970
3971 /*
3972  * Syscall to set a named extended attribute on a file or directory.
3973  * Accepts attribute name, and a uio structure pointing to the data to set.
3974  * The uio is consumed in the style of writev().  The real work happens
3975  * in VOP_SETEXTATTR().
3976  */
3977 int
3978 sys_extattr_set_file(struct extattr_set_file_args *uap)
3979 {
3980         char attrname[EXTATTR_MAXNAMELEN];
3981         struct iovec aiov[UIO_SMALLIOV];
3982         struct iovec *needfree;
3983         struct nlookupdata nd;
3984         struct iovec *iov;
3985         struct vnode *vp;
3986         struct uio auio;
3987         u_int iovlen;
3988         u_int cnt;
3989         int error;
3990         int i;
3991
3992         error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
3993         if (error)
3994                 return (error);
3995
3996         vp = NULL;
3997         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3998         if (error == 0)
3999                 error = nlookup(&nd);
4000         if (error == 0)
4001                 error = ncp_writechk(&nd.nl_nch);
4002         if (error == 0)
4003                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
4004         if (error) {
4005                 nlookup_done(&nd);
4006                 return (error);
4007         }
4008
4009         needfree = NULL;
4010         iovlen = uap->iovcnt * sizeof(struct iovec);
4011         if (uap->iovcnt > UIO_SMALLIOV) {
4012                 if (uap->iovcnt > UIO_MAXIOV) {
4013                         error = EINVAL;
4014                         goto done;
4015                 }
4016                 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
4017                 needfree = iov;
4018         } else {
4019                 iov = aiov;
4020         }
4021         auio.uio_iov = iov;
4022         auio.uio_iovcnt = uap->iovcnt;
4023         auio.uio_rw = UIO_WRITE;
4024         auio.uio_segflg = UIO_USERSPACE;
4025         auio.uio_td = nd.nl_td;
4026         auio.uio_offset = 0;
4027         if ((error = copyin(uap->iovp, iov, iovlen)))
4028                 goto done;
4029         auio.uio_resid = 0;
4030         for (i = 0; i < uap->iovcnt; i++) {
4031                 if (iov->iov_len > INT_MAX - auio.uio_resid) {
4032                         error = EINVAL;
4033                         goto done;
4034                 }
4035                 auio.uio_resid += iov->iov_len;
4036                 iov++;
4037         }
4038         cnt = auio.uio_resid;
4039         error = VOP_SETEXTATTR(vp, attrname, &auio, nd.nl_cred);
4040         cnt -= auio.uio_resid;
4041         uap->sysmsg_result = cnt;
4042 done:
4043         vput(vp);
4044         nlookup_done(&nd);
4045         if (needfree)
4046                 FREE(needfree, M_IOV);
4047         return (error);
4048 }
4049
4050 /*
4051  * Syscall to get a named extended attribute on a file or directory.
4052  * Accepts attribute name, and a uio structure pointing to a buffer for the
4053  * data.  The uio is consumed in the style of readv().  The real work
4054  * happens in VOP_GETEXTATTR();
4055  */
4056 int
4057 sys_extattr_get_file(struct extattr_get_file_args *uap)
4058 {
4059         char attrname[EXTATTR_MAXNAMELEN];
4060         struct iovec aiov[UIO_SMALLIOV];
4061         struct iovec *needfree;
4062         struct nlookupdata nd;
4063         struct iovec *iov;
4064         struct vnode *vp;
4065         struct uio auio;
4066         u_int iovlen;
4067         u_int cnt;
4068         int error;
4069         int i;
4070
4071         error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
4072         if (error)
4073                 return (error);
4074
4075         vp = NULL;
4076         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
4077         if (error == 0)
4078                 error = nlookup(&nd);
4079         if (error == 0)
4080                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
4081         if (error) {
4082                 nlookup_done(&nd);
4083                 return (error);
4084         }
4085
4086         iovlen = uap->iovcnt * sizeof (struct iovec);
4087         needfree = NULL;
4088         if (uap->iovcnt > UIO_SMALLIOV) {
4089                 if (uap->iovcnt > UIO_MAXIOV) {
4090                         error = EINVAL;
4091                         goto done;
4092                 }
4093                 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
4094                 needfree = iov;
4095         } else {
4096                 iov = aiov;
4097         }
4098         auio.uio_iov = iov;
4099         auio.uio_iovcnt = uap->iovcnt;
4100         auio.uio_rw = UIO_READ;
4101         auio.uio_segflg = UIO_USERSPACE;
4102         auio.uio_td = nd.nl_td;
4103         auio.uio_offset = 0;
4104         if ((error = copyin(uap->iovp, iov, iovlen)))
4105                 goto done;
4106         auio.uio_resid = 0;
4107         for (i = 0; i < uap->iovcnt; i++) {
4108                 if (iov->iov_len > INT_MAX - auio.uio_resid) {
4109                         error = EINVAL;
4110                         goto done;
4111                 }
4112                 auio.uio_resid += iov->iov_len;
4113                 iov++;
4114         }
4115         cnt = auio.uio_resid;
4116         error = VOP_GETEXTATTR(vp, attrname, &auio, nd.nl_cred);
4117         cnt -= auio.uio_resid;
4118         uap->sysmsg_result = cnt;
4119 done:
4120         vput(vp);
4121         nlookup_done(&nd);
4122         if (needfree)
4123                 FREE(needfree, M_IOV);
4124         return(error);
4125 }
4126
4127 /*
4128  * Syscall to delete a named extended attribute from a file or directory.
4129  * Accepts attribute name.  The real work happens in VOP_SETEXTATTR().
4130  */
4131 int
4132 sys_extattr_delete_file(struct extattr_delete_file_args *uap)
4133 {
4134         char attrname[EXTATTR_MAXNAMELEN];
4135         struct nlookupdata nd;
4136         struct vnode *vp;
4137         int error;
4138
4139         error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
4140         if (error)
4141                 return(error);
4142
4143         vp = NULL;
4144         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
4145         if (error == 0)
4146                 error = nlookup(&nd);
4147         if (error == 0)
4148                 error = ncp_writechk(&nd.nl_nch);
4149         if (error == 0)
4150                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
4151         if (error) {
4152                 nlookup_done(&nd);
4153                 return (error);
4154         }
4155
4156         error = VOP_SETEXTATTR(vp, attrname, NULL, nd.nl_cred);
4157         vput(vp);
4158         nlookup_done(&nd);
4159         return(error);
4160 }
4161
4162 /*
4163  * Determine if the mount is visible to the process.
4164  */
4165 static int
4166 chroot_visible_mnt(struct mount *mp, struct proc *p)
4167 {
4168         struct nchandle nch;
4169
4170         /*
4171          * Traverse from the mount point upwards.  If we hit the process
4172          * root then the mount point is visible to the process.
4173          */
4174         nch = mp->mnt_ncmountpt;
4175         while (nch.ncp) {
4176                 if (nch.mount == p->p_fd->fd_nrdir.mount &&
4177                     nch.ncp == p->p_fd->fd_nrdir.ncp) {
4178                         return(1);
4179                 }
4180                 if (nch.ncp == nch.mount->mnt_ncmountpt.ncp) {
4181                         nch = nch.mount->mnt_ncmounton;
4182                 } else {
4183                         nch.ncp = nch.ncp->nc_parent;
4184                 }
4185         }
4186
4187         /*
4188          * If the mount point is not visible to the process, but the
4189          * process root is in a subdirectory of the mount, return
4190          * TRUE anyway.
4191          */
4192         if (p->p_fd->fd_nrdir.mount == mp)
4193                 return(1);
4194
4195         return(0);
4196 }
4197