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. Neither the name of the University nor the names of its contributors
  19  *    may be used to endorse or promote products derived from this software
  20  *    without specific prior written permission.
  21  *
  22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  *      @(#)vfs_syscalls.c      8.13 (Berkeley) 4/15/94
  35  * $FreeBSD: src/sys/kern/vfs_syscalls.c,v 1.151.2.18 2003/04/04 20:35:58 tegge Exp $
  36  */
  37
  38 #include <sys/param.h>
  39 #include <sys/systm.h>
  40 #include <sys/buf.h>
  41 #include <sys/conf.h>
  42 #include <sys/sysent.h>
  43 #include <sys/malloc.h>
  44 #include <sys/mount.h>
  45 #include <sys/mountctl.h>
  46 #include <sys/sysproto.h>
  47 #include <sys/filedesc.h>
  48 #include <sys/kernel.h>
  49 #include <sys/fcntl.h>
  50 #include <sys/file.h>
  51 #include <sys/linker.h>
  52 #include <sys/stat.h>
  53 #include <sys/unistd.h>
  54 #include <sys/vnode.h>
  55 #include <sys/proc.h>
  56 #include <sys/priv.h>
  57 #include <sys/jail.h>
  58 #include <sys/namei.h>
  59 #include <sys/nlookup.h>
  60 #include <sys/dirent.h>
  61 #include <sys/extattr.h>
  62 #include <sys/spinlock.h>
  63 #include <sys/kern_syscall.h>
  64 #include <sys/objcache.h>
  65 #include <sys/sysctl.h>
  66
  67 #include <sys/buf2.h>
  68 #include <sys/file2.h>
  69 #include <sys/spinlock2.h>
  70 #include <sys/mplock2.h>
  71
  72 #include <vm/vm.h>
  73 #include <vm/vm_object.h>
  74 #include <vm/vm_page.h>
  75
  76 #include <machine/limits.h>
  77 #include <machine/stdarg.h>
  78
  79 static void mount_warning(struct mount *mp, const char *ctl, ...)
  80                 __printflike(2, 3);
  81 static int mount_path(struct proc *p, struct mount *mp, char **rb, char **fb);
  82 static int checkvp_chdir (struct vnode *vn, struct thread *td);
  83 static void checkdirs (struct nchandle *old_nch, struct nchandle *new_nch);
  84 static int chroot_refuse_vdir_fds (struct filedesc *fdp);
  85 static int chroot_visible_mnt(struct mount *mp, struct proc *p);
  86 static int getutimes (struct timeval *, struct timespec *);
  87 static int getutimens (const struct timespec *, struct timespec *, int *);
  88 static int setfown (struct mount *, struct vnode *, uid_t, gid_t);
  89 static int setfmode (struct vnode *, int);
  90 static int setfflags (struct vnode *, int);
  91 static int setutimes (struct vnode *, struct vattr *,
  92                         const struct timespec *, int);
  93 static int      usermount = 0;  /* if 1, non-root can mount fs. */
  94
  95 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
  96     "Allow non-root users to mount filesystems");
  97
  98 /*
  99  * Virtual File System System Calls
 100  */
 101
 102 /*
 103  * Mount a file system.
 104  *
 105  * mount_args(char *type, char *path, int flags, caddr_t data)
 106  *
 107  * MPALMOSTSAFE
 108  */
 109 int
 110 sys_mount(struct mount_args *uap)
 111 {
 112         struct thread *td = curthread;
 113         struct vnode *vp;
 114         struct nchandle nch;
 115         struct mount *mp, *nullmp;
 116         struct vfsconf *vfsp;
 117         int error, flag = 0, flag2 = 0;
 118         int hasmount;
 119         struct vattr va;
 120         struct nlookupdata nd;
 121         char fstypename[MFSNAMELEN];
 122         struct ucred *cred;
 123
 124         cred = td->td_ucred;
 125         if (jailed(cred)) {
 126                 error = EPERM;
 127                 goto done;
 128         }
 129         if (usermount == 0 && (error = priv_check(td, PRIV_ROOT)))
 130                 goto done;
 131
 132         /*
 133          * Do not allow NFS export by non-root users.
 134          */
 135         if (uap->flags & MNT_EXPORTED) {
 136                 error = priv_check(td, PRIV_ROOT);
 137                 if (error)
 138                         goto done;
 139         }
 140         /*
 141          * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
 142          */
 143         if (priv_check(td, PRIV_ROOT))
 144                 uap->flags |= MNT_NOSUID | MNT_NODEV;
 145
 146         /*
 147          * Lookup the requested path and extract the nch and vnode.
 148          */
 149         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
 150         if (error == 0) {
 151                 if ((error = nlookup(&nd)) == 0) {
 152                         if (nd.nl_nch.ncp->nc_vp == NULL)
 153                                 error = ENOENT;
 154                 }
 155         }
 156         if (error) {
 157                 nlookup_done(&nd);
 158                 goto done;
 159         }
 160
 161         /*
 162          * If the target filesystem is resolved via a nullfs mount, then
 163          * nd.nl_nch.mount will be pointing to the nullfs mount structure
 164          * instead of the target file system. We need it in case we are
 165          * doing an update.
 166          */
 167         nullmp = nd.nl_nch.mount;
 168
 169         /*
 170          * Extract the locked+refd ncp and cleanup the nd structure
 171          */
 172         nch = nd.nl_nch;
 173         cache_zero(&nd.nl_nch);
 174         nlookup_done(&nd);
 175
 176         if ((nch.ncp->nc_flag & NCF_ISMOUNTPT) &&
 177             (mp = cache_findmount(&nch)) != NULL) {
 178                 cache_dropmount(mp);
 179                 hasmount = 1;
 180         } else {
 181                 hasmount = 0;
 182         }
 183
 184
 185         /*
 186          * now we have the locked ref'd nch and unreferenced vnode.
 187          */
 188         vp = nch.ncp->nc_vp;
 189         if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
 190                 cache_put(&nch);
 191                 goto done;
 192         }
 193         cache_unlock(&nch);
 194
 195         /*
 196          * Extract the file system type. We need to know this early, to take
 197          * appropriate actions if we are dealing with a nullfs.
 198          */
 199         if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
 200                 cache_drop(&nch);
 201                 vput(vp);
 202                 goto done;
 203         }
 204
 205         /*
 206          * Now we have an unlocked ref'd nch and a locked ref'd vp
 207          */
 208         if (uap->flags & MNT_UPDATE) {
 209                 if ((vp->v_flag & (VROOT|VPFSROOT)) == 0) {
 210                         cache_drop(&nch);
 211                         vput(vp);
 212                         error = EINVAL;
 213                         goto done;
 214                 }
 215
 216                 if (strncmp(fstypename, "null", 5) == 0) {
 217                         KKASSERT(nullmp);
 218                         mp = nullmp;
 219                 } else {
 220                         mp = vp->v_mount;
 221                 }
 222
 223                 flag = mp->mnt_flag;
 224                 flag2 = mp->mnt_kern_flag;
 225                 /*
 226                  * We only allow the filesystem to be reloaded if it
 227                  * is currently mounted read-only.
 228                  */
 229                 if ((uap->flags & MNT_RELOAD) &&
 230                     ((mp->mnt_flag & MNT_RDONLY) == 0)) {
 231                         cache_drop(&nch);
 232                         vput(vp);
 233                         error = EOPNOTSUPP;     /* Needs translation */
 234                         goto done;
 235                 }
 236                 /*
 237                  * Only root, or the user that did the original mount is
 238                  * permitted to update it.
 239                  */
 240                 if (mp->mnt_stat.f_owner != cred->cr_uid &&
 241                     (error = priv_check(td, PRIV_ROOT))) {
 242                         cache_drop(&nch);
 243                         vput(vp);
 244                         goto done;
 245                 }
 246                 if (vfs_busy(mp, LK_NOWAIT)) {
 247                         cache_drop(&nch);
 248                         vput(vp);
 249                         error = EBUSY;
 250                         goto done;
 251                 }
 252                 if (hasmount) {
 253                         cache_drop(&nch);
 254                         vfs_unbusy(mp);
 255                         vput(vp);
 256                         error = EBUSY;
 257                         goto done;
 258                 }
 259                 mp->mnt_flag |=
 260                     uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
 261                 lwkt_gettoken(&mp->mnt_token);
 262                 vn_unlock(vp);
 263                 goto update;
 264         }
 265
 266         /*
 267          * If the user is not root, ensure that they own the directory
 268          * onto which we are attempting to mount.
 269          */
 270         if ((error = VOP_GETATTR(vp, &va)) ||
 271             (va.va_uid != cred->cr_uid &&
 272              (error = priv_check(td, PRIV_ROOT)))) {
 273                 cache_drop(&nch);
 274                 vput(vp);
 275                 goto done;
 276         }
 277         if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
 278                 cache_drop(&nch);
 279                 vput(vp);
 280                 goto done;
 281         }
 282         if (vp->v_type != VDIR) {
 283                 cache_drop(&nch);
 284                 vput(vp);
 285                 error = ENOTDIR;
 286                 goto done;
 287         }
 288         if (vp->v_mount->mnt_kern_flag & MNTK_NOSTKMNT) {
 289                 cache_drop(&nch);
 290                 vput(vp);
 291                 error = EPERM;
 292                 goto done;
 293         }
 294         vfsp = vfsconf_find_by_name(fstypename);
 295         if (vfsp == NULL) {
 296                 linker_file_t lf;
 297
 298                 /* Only load modules for root (very important!) */
 299                 if ((error = priv_check(td, PRIV_ROOT)) != 0) {
 300                         cache_drop(&nch);
 301                         vput(vp);
 302                         goto done;
 303                 }
 304                 error = linker_load_file(fstypename, &lf);
 305                 if (error || lf == NULL) {
 306                         cache_drop(&nch);
 307                         vput(vp);
 308                         if (lf == NULL)
 309                                 error = ENODEV;
 310                         goto done;
 311                 }
 312                 lf->userrefs++;
 313                 /* lookup again, see if the VFS was loaded */
 314                 vfsp = vfsconf_find_by_name(fstypename);
 315                 if (vfsp == NULL) {
 316                         lf->userrefs--;
 317                         linker_file_unload(lf);
 318                         cache_drop(&nch);
 319                         vput(vp);
 320                         error = ENODEV;
 321                         goto done;
 322                 }
 323         }
 324         if (hasmount) {
 325                 cache_drop(&nch);
 326                 vput(vp);
 327                 error = EBUSY;
 328                 goto done;
 329         }
 330
 331         /*
 332          * Allocate and initialize the filesystem.
 333          */
 334         mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
 335         mount_init(mp);
 336         vfs_busy(mp, LK_NOWAIT);
 337         mp->mnt_op = vfsp->vfc_vfsops;
 338         mp->mnt_vfc = vfsp;
 339         vfsp->vfc_refcount++;
 340         mp->mnt_stat.f_type = vfsp->vfc_typenum;
 341         mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
 342         strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
 343         mp->mnt_stat.f_owner = cred->cr_uid;
 344         lwkt_gettoken(&mp->mnt_token);
 345         vn_unlock(vp);
 346 update:
 347         /*
 348          * (per-mount token acquired at this point)
 349          *
 350          * Set the mount level flags.
 351          */
 352         if (uap->flags & MNT_RDONLY)
 353                 mp->mnt_flag |= MNT_RDONLY;
 354         else if (mp->mnt_flag & MNT_RDONLY)
 355                 mp->mnt_kern_flag |= MNTK_WANTRDWR;
 356         mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
 357             MNT_SYNCHRONOUS | MNT_ASYNC | MNT_NOATIME |
 358             MNT_NOSYMFOLLOW | MNT_IGNORE | MNT_TRIM |
 359             MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
 360         mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
 361             MNT_NODEV | MNT_SYNCHRONOUS | MNT_ASYNC | MNT_FORCE |
 362             MNT_NOSYMFOLLOW | MNT_IGNORE | MNT_TRIM |
 363             MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
 364         /*
 365          * Mount the filesystem.
 366          * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
 367          * get.
 368          */
 369         error = VFS_MOUNT(mp, uap->path, uap->data, cred);
 370         if (mp->mnt_flag & MNT_UPDATE) {
 371                 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
 372                         mp->mnt_flag &= ~MNT_RDONLY;
 373                 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
 374                 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
 375                 if (error) {
 376                         mp->mnt_flag = flag;
 377                         mp->mnt_kern_flag = flag2;
 378                 }
 379                 lwkt_reltoken(&mp->mnt_token);
 380                 vfs_unbusy(mp);
 381                 vrele(vp);
 382                 cache_drop(&nch);
 383                 goto done;
 384         }
 385         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 386
 387         /*
 388          * Put the new filesystem on the mount list after root.  The mount
 389          * point gets its own mnt_ncmountpt (unless the VFS already set one
 390          * up) which represents the root of the mount.  The lookup code
 391          * detects the mount point going forward and checks the root of
 392          * the mount going backwards.
 393          *
 394          * It is not necessary to invalidate or purge the vnode underneath
 395          * because elements under the mount will be given their own glue
 396          * namecache record.
 397          */
 398         if (!error) {
 399                 if (mp->mnt_ncmountpt.ncp == NULL) {
 400                         /*
 401                          * allocate, then unlock, but leave the ref intact
 402                          */
 403                         cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
 404                         cache_unlock(&mp->mnt_ncmountpt);
 405                 }
 406                 mp->mnt_ncmounton = nch;                /* inherits ref */
 407                 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
 408                 cache_ismounting(mp);
 409
 410                 mountlist_insert(mp, MNTINS_LAST);
 411                 vn_unlock(vp);
 412                 checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
 413                 error = vfs_allocate_syncvnode(mp);
 414                 lwkt_reltoken(&mp->mnt_token);
 415                 vfs_unbusy(mp);
 416                 error = VFS_START(mp, 0);
 417                 vrele(vp);
 418         } else {
 419                 vn_syncer_thr_stop(mp);
 420                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
 421                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
 422                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
 423                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
 424                 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
 425                 mp->mnt_vfc->vfc_refcount--;
 426                 lwkt_reltoken(&mp->mnt_token);
 427                 vfs_unbusy(mp);
 428                 kfree(mp, M_MOUNT);
 429                 cache_drop(&nch);
 430                 vput(vp);
 431         }
 432 done:
 433         return (error);
 434 }
 435
 436 /*
 437  * Scan all active processes to see if any of them have a current
 438  * or root directory onto which the new filesystem has just been
 439  * mounted. If so, replace them with the new mount point.
 440  *
 441  * Both old_nch and new_nch are ref'd on call but not locked.
 442  * new_nch must be temporarily locked so it can be associated with the
 443  * vnode representing the root of the mount point.
 444  */
 445 struct checkdirs_info {
 446         struct nchandle old_nch;
 447         struct nchandle new_nch;
 448         struct vnode *old_vp;
 449         struct vnode *new_vp;
 450 };
 451
 452 static int checkdirs_callback(struct proc *p, void *data);
 453
 454 static void
 455 checkdirs(struct nchandle *old_nch, struct nchandle *new_nch)
 456 {
 457         struct checkdirs_info info;
 458         struct vnode *olddp;
 459         struct vnode *newdp;
 460         struct mount *mp;
 461
 462         /*
 463          * If the old mount point's vnode has a usecount of 1, it is not
 464          * being held as a descriptor anywhere.
 465          */
 466         olddp = old_nch->ncp->nc_vp;
 467         if (olddp == NULL || VREFCNT(olddp) == 1)
 468                 return;
 469
 470         /*
 471          * Force the root vnode of the new mount point to be resolved
 472          * so we can update any matching processes.
 473          */
 474         mp = new_nch->mount;
 475         if (VFS_ROOT(mp, &newdp))
 476                 panic("mount: lost mount");
 477         vn_unlock(newdp);
 478         cache_lock(new_nch);
 479         vn_lock(newdp, LK_EXCLUSIVE | LK_RETRY);
 480         cache_setunresolved(new_nch);
 481         cache_setvp(new_nch, newdp);
 482         cache_unlock(new_nch);
 483
 484         /*
 485          * Special handling of the root node
 486          */
 487         if (rootvnode == olddp) {
 488                 vref(newdp);
 489                 vfs_cache_setroot(newdp, cache_hold(new_nch));
 490         }
 491
 492         /*
 493          * Pass newdp separately so the callback does not have to access
 494          * it via new_nch->ncp->nc_vp.
 495          */
 496         info.old_nch = *old_nch;
 497         info.new_nch = *new_nch;
 498         info.new_vp = newdp;
 499         allproc_scan(checkdirs_callback, &info);
 500         vput(newdp);
 501 }
 502
 503 /*
 504  * NOTE: callback is not MP safe because the scanned process's filedesc
 505  * structure can be ripped out from under us, amoung other things.
 506  */
 507 static int
 508 checkdirs_callback(struct proc *p, void *data)
 509 {
 510         struct checkdirs_info *info = data;
 511         struct filedesc *fdp;
 512         struct nchandle ncdrop1;
 513         struct nchandle ncdrop2;
 514         struct vnode *vprele1;
 515         struct vnode *vprele2;
 516
 517         if ((fdp = p->p_fd) != NULL) {
 518                 cache_zero(&ncdrop1);
 519                 cache_zero(&ncdrop2);
 520                 vprele1 = NULL;
 521                 vprele2 = NULL;
 522
 523                 /*
 524                  * MPUNSAFE - XXX fdp can be pulled out from under a
 525                  * foreign process.
 526                  *
 527                  * A shared filedesc is ok, we don't have to copy it
 528                  * because we are making this change globally.
 529                  */
 530                 spin_lock(&fdp->fd_spin);
 531                 if (fdp->fd_ncdir.mount == info->old_nch.mount &&
 532                     fdp->fd_ncdir.ncp == info->old_nch.ncp) {
 533                         vprele1 = fdp->fd_cdir;
 534                         vref(info->new_vp);
 535                         fdp->fd_cdir = info->new_vp;
 536                         ncdrop1 = fdp->fd_ncdir;
 537                         cache_copy(&info->new_nch, &fdp->fd_ncdir);
 538                 }
 539                 if (fdp->fd_nrdir.mount == info->old_nch.mount &&
 540                     fdp->fd_nrdir.ncp == info->old_nch.ncp) {
 541                         vprele2 = fdp->fd_rdir;
 542                         vref(info->new_vp);
 543                         fdp->fd_rdir = info->new_vp;
 544                         ncdrop2 = fdp->fd_nrdir;
 545                         cache_copy(&info->new_nch, &fdp->fd_nrdir);
 546                 }
 547                 spin_unlock(&fdp->fd_spin);
 548                 if (ncdrop1.ncp)
 549                         cache_drop(&ncdrop1);
 550                 if (ncdrop2.ncp)
 551                         cache_drop(&ncdrop2);
 552                 if (vprele1)
 553                         vrele(vprele1);
 554                 if (vprele2)
 555                         vrele(vprele2);
 556         }
 557         return(0);
 558 }
 559
 560 /*
 561  * Unmount a file system.
 562  *
 563  * Note: unmount takes a path to the vnode mounted on as argument,
 564  * not special file (as before).
 565  *
 566  * umount_args(char *path, int flags)
 567  *
 568  * MPALMOSTSAFE
 569  */
 570 int
 571 sys_unmount(struct unmount_args *uap)
 572 {
 573         struct thread *td = curthread;
 574         struct proc *p __debugvar = td->td_proc;
 575         struct mount *mp = NULL;
 576         struct nlookupdata nd;
 577         int error;
 578
 579         KKASSERT(p);
 580         get_mplock();
 581         if (td->td_ucred->cr_prison != NULL) {
 582                 error = EPERM;
 583                 goto done;
 584         }
 585         if (usermount == 0 && (error = priv_check(td, PRIV_ROOT)))
 586                 goto done;
 587
 588         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
 589         if (error == 0)
 590                 error = nlookup(&nd);
 591         if (error)
 592                 goto out;
 593
 594         mp = nd.nl_nch.mount;
 595
 596         /*
 597          * Only root, or the user that did the original mount is
 598          * permitted to unmount this filesystem.
 599          */
 600         if ((mp->mnt_stat.f_owner != td->td_ucred->cr_uid) &&
 601             (error = priv_check(td, PRIV_ROOT)))
 602                 goto out;
 603
 604         /*
 605          * Don't allow unmounting the root file system.
 606          */
 607         if (mp->mnt_flag & MNT_ROOTFS) {
 608                 error = EINVAL;
 609                 goto out;
 610         }
 611
 612         /*
 613          * Must be the root of the filesystem
 614          */
 615         if (nd.nl_nch.ncp != mp->mnt_ncmountpt.ncp) {
 616                 error = EINVAL;
 617                 goto out;
 618         }
 619
 620 out:
 621         nlookup_done(&nd);
 622         if (error == 0)
 623                 error = dounmount(mp, uap->flags);
 624 done:
 625         rel_mplock();
 626         return (error);
 627 }
 628
 629 /*
 630  * Do the actual file system unmount.
 631  */
 632 static int
 633 dounmount_interlock(struct mount *mp)
 634 {
 635         if (mp->mnt_kern_flag & MNTK_UNMOUNT)
 636                 return (EBUSY);
 637         mp->mnt_kern_flag |= MNTK_UNMOUNT;
 638         return(0);
 639 }
 640
 641 static int
 642 unmount_allproc_cb(struct proc *p, void *arg)
 643 {
 644         struct mount *mp;
 645
 646         if (p->p_textnch.ncp == NULL)
 647                 return 0;
 648
 649         mp = (struct mount *)arg;
 650         if (p->p_textnch.mount == mp)
 651                 cache_drop(&p->p_textnch);
 652
 653         return 0;
 654 }
 655
 656 int
 657 dounmount(struct mount *mp, int flags)
 658 {
 659         struct namecache *ncp;
 660         struct nchandle nch;
 661         struct vnode *vp;
 662         int error;
 663         int async_flag;
 664         int lflags;
 665         int freeok = 1;
 666         int retry;
 667
 668         lwkt_gettoken(&mp->mnt_token);
 669         /*
 670          * Exclusive access for unmounting purposes
 671          */
 672         if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
 673                 goto out;
 674
 675         /*
 676          * Allow filesystems to detect that a forced unmount is in progress.
 677          */
 678         if (flags & MNT_FORCE)
 679                 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
 680         lflags = LK_EXCLUSIVE | ((flags & MNT_FORCE) ? 0 : LK_TIMELOCK);
 681         error = lockmgr(&mp->mnt_lock, lflags);
 682         if (error) {
 683                 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
 684                 if (mp->mnt_kern_flag & MNTK_MWAIT) {
 685                         mp->mnt_kern_flag &= ~MNTK_MWAIT;
 686                         wakeup(mp);
 687                 }
 688                 goto out;
 689         }
 690
 691         if (mp->mnt_flag & MNT_EXPUBLIC)
 692                 vfs_setpublicfs(NULL, NULL, NULL);
 693
 694         vfs_msync(mp, MNT_WAIT);
 695         async_flag = mp->mnt_flag & MNT_ASYNC;
 696         mp->mnt_flag &=~ MNT_ASYNC;
 697
 698         /*
 699          * If this filesystem isn't aliasing other filesystems,
 700          * try to invalidate any remaining namecache entries and
 701          * check the count afterwords.
 702          */
 703         if ((mp->mnt_kern_flag & MNTK_NCALIASED) == 0) {
 704                 cache_lock(&mp->mnt_ncmountpt);
 705                 cache_inval(&mp->mnt_ncmountpt, CINV_DESTROY|CINV_CHILDREN);
 706                 cache_unlock(&mp->mnt_ncmountpt);
 707
 708                 if ((ncp = mp->mnt_ncmountpt.ncp) != NULL &&
 709                     (ncp->nc_refs != 1 || TAILQ_FIRST(&ncp->nc_list))) {
 710                         allproc_scan(&unmount_allproc_cb, mp);
 711                 }
 712
 713                 if ((ncp = mp->mnt_ncmountpt.ncp) != NULL &&
 714                     (ncp->nc_refs != 1 || TAILQ_FIRST(&ncp->nc_list))) {
 715
 716                         if ((flags & MNT_FORCE) == 0) {
 717                                 error = EBUSY;
 718                                 mount_warning(mp, "Cannot unmount: "
 719                                                   "%d namecache "
 720                                                   "references still "
 721                                                   "present",
 722                                                   ncp->nc_refs - 1);
 723                         } else {
 724                                 mount_warning(mp, "Forced unmount: "
 725                                                   "%d namecache "
 726                                                   "references still "
 727                                                   "present",
 728                                                   ncp->nc_refs - 1);
 729                                 freeok = 0;
 730                         }
 731                 }
 732         }
 733
 734         /*
 735          * Decomission our special mnt_syncer vnode.  This also stops
 736          * the vnlru code.  If we are unable to unmount we recommission
 737          * the vnode.
 738          *
 739          * Then sync the filesystem.
 740          */
 741         if ((vp = mp->mnt_syncer) != NULL) {
 742                 mp->mnt_syncer = NULL;
 743                 atomic_set_int(&vp->v_refcnt, VREF_FINALIZE);
 744                 vrele(vp);
 745         }
 746         if ((mp->mnt_flag & MNT_RDONLY) == 0)
 747                 VFS_SYNC(mp, MNT_WAIT);
 748
 749         /*
 750          * nchandle records ref the mount structure.  Expect a count of 1
 751          * (our mount->mnt_ncmountpt).
 752          *
 753          * Scans can get temporary refs on a mountpoint (thought really
 754          * heavy duty stuff like cache_findmount() do not).
 755          */
 756         for (retry = 0; retry < 10 && mp->mnt_refs != 1; ++retry) {
 757                 cache_unmounting(mp);
 758                 tsleep(&mp->mnt_refs, 0, "mntbsy", hz / 10 + 1);
 759         }
 760         if (mp->mnt_refs != 1) {
 761                 if ((flags & MNT_FORCE) == 0) {
 762                         mount_warning(mp, "Cannot unmount: "
 763                                           "%d mount refs still present",
 764                                           mp->mnt_refs);
 765                         error = EBUSY;
 766                 } else {
 767                         mount_warning(mp, "Forced unmount: "
 768                                           "%d mount refs still present",
 769                                           mp->mnt_refs);
 770                         freeok = 0;
 771                 }
 772         }
 773
 774         /*
 775          * So far so good, sync the filesystem once more and
 776          * call the VFS unmount code if the sync succeeds.
 777          */
 778         if (error == 0) {
 779                 if (((mp->mnt_flag & MNT_RDONLY) ||
 780                      (error = VFS_SYNC(mp, MNT_WAIT)) == 0) ||
 781                     (flags & MNT_FORCE)) {
 782                         error = VFS_UNMOUNT(mp, flags);
 783                 }
 784         }
 785
 786         /*
 787          * If an error occurred we can still recover, restoring the
 788          * syncer vnode and misc flags.
 789          */
 790         if (error) {
 791                 if (mp->mnt_syncer == NULL)
 792                         vfs_allocate_syncvnode(mp);
 793                 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
 794                 mp->mnt_flag |= async_flag;
 795                 lockmgr(&mp->mnt_lock, LK_RELEASE);
 796                 if (mp->mnt_kern_flag & MNTK_MWAIT) {
 797                         mp->mnt_kern_flag &= ~MNTK_MWAIT;
 798                         wakeup(mp);
 799                 }
 800                 goto out;
 801         }
 802         /*
 803          * Clean up any journals still associated with the mount after
 804          * filesystem activity has ceased.
 805          */
 806         journal_remove_all_journals(mp,
 807             ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));
 808
 809         mountlist_remove(mp);
 810
 811         /*
 812          * Remove any installed vnode ops here so the individual VFSs don't
 813          * have to.
 814          */
 815         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
 816         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
 817         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
 818         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
 819         vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
 820
 821         if (mp->mnt_ncmountpt.ncp != NULL) {
 822                 nch = mp->mnt_ncmountpt;
 823                 cache_zero(&mp->mnt_ncmountpt);
 824                 cache_clrmountpt(&nch);
 825                 cache_drop(&nch);
 826         }
 827         if (mp->mnt_ncmounton.ncp != NULL) {
 828                 cache_unmounting(mp);
 829                 nch = mp->mnt_ncmounton;
 830                 cache_zero(&mp->mnt_ncmounton);
 831                 cache_clrmountpt(&nch);
 832                 cache_drop(&nch);
 833         }
 834
 835         mp->mnt_vfc->vfc_refcount--;
 836         if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
 837                 panic("unmount: dangling vnode");
 838         lockmgr(&mp->mnt_lock, LK_RELEASE);
 839         if (mp->mnt_kern_flag & MNTK_MWAIT) {
 840                 mp->mnt_kern_flag &= ~MNTK_MWAIT;
 841                 wakeup(mp);
 842         }
 843
 844         /*
 845          * If we reach here and freeok != 0 we must free the mount.
 846          * If refs > 1 cycle and wait, just in case someone tried
 847          * to busy the mount after we decided to do the unmount.
 848          */
 849         if (freeok) {
 850                 while (mp->mnt_refs > 1) {
 851                         cache_unmounting(mp);
 852                         wakeup(mp);
 853                         tsleep(&mp->mnt_refs, 0, "umntrwait", hz / 10 + 1);
 854                 }
 855                 lwkt_reltoken(&mp->mnt_token);
 856                 kfree(mp, M_MOUNT);
 857                 mp = NULL;
 858         }
 859         error = 0;
 860 out:
 861         if (mp)
 862                 lwkt_reltoken(&mp->mnt_token);
 863         return (error);
 864 }
 865
 866 static
 867 void
 868 mount_warning(struct mount *mp, const char *ctl, ...)
 869 {
 870         char *ptr;
 871         char *buf;
 872         __va_list va;
 873
 874         __va_start(va, ctl);
 875         if (cache_fullpath(NULL, &mp->mnt_ncmounton, NULL,
 876                            &ptr, &buf, 0) == 0) {
 877                 kprintf("unmount(%s): ", ptr);
 878                 kvprintf(ctl, va);
 879                 kprintf("\n");
 880                 kfree(buf, M_TEMP);
 881         } else {
 882                 kprintf("unmount(%p", mp);
 883                 if (mp->mnt_ncmounton.ncp && mp->mnt_ncmounton.ncp->nc_name)
 884                         kprintf(",%s", mp->mnt_ncmounton.ncp->nc_name);
 885                 kprintf("): ");
 886                 kvprintf(ctl, va);
 887                 kprintf("\n");
 888         }
 889         __va_end(va);
 890 }
 891
 892 /*
 893  * Shim cache_fullpath() to handle the case where a process is chrooted into
 894  * a subdirectory of a mount.  In this case if the root mount matches the
 895  * process root directory's mount we have to specify the process's root
 896  * directory instead of the mount point, because the mount point might
 897  * be above the root directory.
 898  */
 899 static
 900 int
 901 mount_path(struct proc *p, struct mount *mp, char **rb, char **fb)
 902 {
 903         struct nchandle *nch;
 904
 905         if (p && p->p_fd->fd_nrdir.mount == mp)
 906                 nch = &p->p_fd->fd_nrdir;
 907         else
 908                 nch = &mp->mnt_ncmountpt;
 909         return(cache_fullpath(p, nch, NULL, rb, fb, 0));
 910 }
 911
 912 /*
 913  * Sync each mounted filesystem.
 914  */
 915
 916 #ifdef DEBUG
 917 static int syncprt = 0;
 918 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
 919 #endif /* DEBUG */
 920
 921 static int sync_callback(struct mount *mp, void *data);
 922
 923 int
 924 sys_sync(struct sync_args *uap)
 925 {
 926         mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
 927         return (0);
 928 }
 929
 930 static
 931 int
 932 sync_callback(struct mount *mp, void *data __unused)
 933 {
 934         int asyncflag;
 935
 936         if ((mp->mnt_flag & MNT_RDONLY) == 0) {
 937                 asyncflag = mp->mnt_flag & MNT_ASYNC;
 938                 mp->mnt_flag &= ~MNT_ASYNC;
 939                 vfs_msync(mp, MNT_NOWAIT);
 940                 VFS_SYNC(mp, MNT_NOWAIT);
 941                 mp->mnt_flag |= asyncflag;
 942         }
 943         return(0);
 944 }
 945
 946 /* XXX PRISON: could be per prison flag */
 947 static int prison_quotas;
 948 #if 0
 949 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
 950 #endif
 951
 952 /*
 953  *  quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
 954  *
 955  * Change filesystem quotas.
 956  *
 957  * MPALMOSTSAFE
 958  */
 959 int
 960 sys_quotactl(struct quotactl_args *uap)
 961 {
 962         struct nlookupdata nd;
 963         struct thread *td;
 964         struct mount *mp;
 965         int error;
 966
 967         get_mplock();
 968         td = curthread;
 969         if (td->td_ucred->cr_prison && !prison_quotas) {
 970                 error = EPERM;
 971                 goto done;
 972         }
 973
 974         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
 975         if (error == 0)
 976                 error = nlookup(&nd);
 977         if (error == 0) {
 978                 mp = nd.nl_nch.mount;
 979                 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
 980                                     uap->arg, nd.nl_cred);
 981         }
 982         nlookup_done(&nd);
 983 done:
 984         rel_mplock();
 985         return (error);
 986 }
 987
 988 /*
 989  * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
 990  *              void *buf, int buflen)
 991  *
 992  * This function operates on a mount point and executes the specified
 993  * operation using the specified control data, and possibly returns data.
 994  *
 995  * The actual number of bytes stored in the result buffer is returned, 0
 996  * if none, otherwise an error is returned.
 997  *
 998  * MPALMOSTSAFE
 999  */
1000 int
1001 sys_mountctl(struct mountctl_args *uap)
1002 {
1003         struct thread *td = curthread;
1004         struct proc *p = td->td_proc;
1005         struct file *fp;
1006         void *ctl = NULL;
1007         void *buf = NULL;
1008         char *path = NULL;
1009         int error;
1010
1011         /*
1012          * Sanity and permissions checks.  We must be root.
1013          */
1014         KKASSERT(p);
1015         if (td->td_ucred->cr_prison != NULL)
1016                 return (EPERM);
1017         if ((uap->op != MOUNTCTL_MOUNTFLAGS) &&
1018             (error = priv_check(td, PRIV_ROOT)) != 0)
1019                 return (error);
1020
1021         /*
1022          * Argument length checks
1023          */
1024         if (uap->ctllen < 0 || uap->ctllen > 1024)
1025                 return (EINVAL);
1026         if (uap->buflen < 0 || uap->buflen > 16 * 1024)
1027                 return (EINVAL);
1028         if (uap->path == NULL)
1029                 return (EINVAL);
1030
1031         /*
1032          * Allocate the necessary buffers and copyin data
1033          */
1034         path = objcache_get(namei_oc, M_WAITOK);
1035         error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
1036         if (error)
1037                 goto done;
1038
1039         if (uap->ctllen) {
1040                 ctl = kmalloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
1041                 error = copyin(uap->ctl, ctl, uap->ctllen);
1042                 if (error)
1043                         goto done;
1044         }
1045         if (uap->buflen)
1046                 buf = kmalloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
1047
1048         /*
1049          * Validate the descriptor
1050          */
1051         if (uap->fd >= 0) {
1052                 fp = holdfp(p->p_fd, uap->fd, -1);
1053                 if (fp == NULL) {
1054                         error = EBADF;
1055                         goto done;
1056                 }
1057         } else {
1058                 fp = NULL;
1059         }
1060
1061         /*
1062          * Execute the internal kernel function and clean up.
1063          */
1064         get_mplock();
1065         error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
1066         rel_mplock();
1067         if (fp)
1068                 fdrop(fp);
1069         if (error == 0 && uap->sysmsg_result > 0)
1070                 error = copyout(buf, uap->buf, uap->sysmsg_result);
1071 done:
1072         if (path)
1073                 objcache_put(namei_oc, path);
1074         if (ctl)
1075                 kfree(ctl, M_TEMP);
1076         if (buf)
1077                 kfree(buf, M_TEMP);
1078         return (error);
1079 }
1080
1081 /*
1082  * Execute a mount control operation by resolving the path to a mount point
1083  * and calling vop_mountctl().
1084  *
1085  * Use the mount point from the nch instead of the vnode so nullfs mounts
1086  * can properly spike the VOP.
1087  */
1088 int
1089 kern_mountctl(const char *path, int op, struct file *fp,
1090                 const void *ctl, int ctllen,
1091                 void *buf, int buflen, int *res)
1092 {
1093         struct vnode *vp;
1094         struct mount *mp;
1095         struct nlookupdata nd;
1096         int error;
1097
1098         *res = 0;
1099         vp = NULL;
1100         error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
1101         if (error == 0)
1102                 error = nlookup(&nd);
1103         if (error == 0)
1104                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
1105         mp = nd.nl_nch.mount;
1106         nlookup_done(&nd);
1107         if (error)
1108                 return (error);
1109         vn_unlock(vp);
1110
1111         /*
1112          * Must be the root of the filesystem
1113          */
1114         if ((vp->v_flag & (VROOT|VPFSROOT)) == 0) {
1115                 vrele(vp);
1116                 return (EINVAL);
1117         }
1118         error = vop_mountctl(mp->mnt_vn_use_ops, vp, op, fp, ctl, ctllen,
1119                              buf, buflen, res);
1120         vrele(vp);
1121         return (error);
1122 }
1123
1124 int
1125 kern_statfs(struct nlookupdata *nd, struct statfs *buf)
1126 {
1127         struct thread *td = curthread;
1128         struct proc *p = td->td_proc;
1129         struct mount *mp;
1130         struct statfs *sp;
1131         char *fullpath, *freepath;
1132         int error;
1133
1134         if ((error = nlookup(nd)) != 0)
1135                 return (error);
1136         mp = nd->nl_nch.mount;
1137         sp = &mp->mnt_stat;
1138         if ((error = VFS_STATFS(mp, sp, nd->nl_cred)) != 0)
1139                 return (error);
1140
1141         error = mount_path(p, mp, &fullpath, &freepath);
1142         if (error)
1143                 return(error);
1144         bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1145         strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1146         kfree(freepath, M_TEMP);
1147
1148         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1149         bcopy(sp, buf, sizeof(*buf));
1150         /* Only root should have access to the fsid's. */
1151         if (priv_check(td, PRIV_ROOT))
1152                 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
1153         return (0);
1154 }
1155
1156 /*
1157  * statfs_args(char *path, struct statfs *buf)
1158  *
1159  * Get filesystem statistics.
1160  */
1161 int
1162 sys_statfs(struct statfs_args *uap)
1163 {
1164         struct nlookupdata nd;
1165         struct statfs buf;
1166         int error;
1167
1168         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1169         if (error == 0)
1170                 error = kern_statfs(&nd, &buf);
1171         nlookup_done(&nd);
1172         if (error == 0)
1173                 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1174         return (error);
1175 }
1176
1177 int
1178 kern_fstatfs(int fd, struct statfs *buf)
1179 {
1180         struct thread *td = curthread;
1181         struct proc *p = td->td_proc;
1182         struct file *fp;
1183         struct mount *mp;
1184         struct statfs *sp;
1185         char *fullpath, *freepath;
1186         int error;
1187
1188         KKASSERT(p);
1189         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
1190                 return (error);
1191
1192         /*
1193          * Try to use mount info from any overlays rather than the
1194          * mount info for the underlying vnode, otherwise we will
1195          * fail when operating on null-mounted paths inside a chroot.
1196          */
1197         if ((mp = fp->f_nchandle.mount) == NULL)
1198                 mp = ((struct vnode *)fp->f_data)->v_mount;
1199         if (mp == NULL) {
1200                 error = EBADF;
1201                 goto done;
1202         }
1203         if (fp->f_cred == NULL) {
1204                 error = EINVAL;
1205                 goto done;
1206         }
1207         sp = &mp->mnt_stat;
1208         if ((error = VFS_STATFS(mp, sp, fp->f_cred)) != 0)
1209                 goto done;
1210
1211         if ((error = mount_path(p, mp, &fullpath, &freepath)) != 0)
1212                 goto done;
1213         bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1214         strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1215         kfree(freepath, M_TEMP);
1216
1217         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1218         bcopy(sp, buf, sizeof(*buf));
1219
1220         /* Only root should have access to the fsid's. */
1221         if (priv_check(td, PRIV_ROOT))
1222                 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
1223         error = 0;
1224 done:
1225         fdrop(fp);
1226         return (error);
1227 }
1228
1229 /*
1230  * fstatfs_args(int fd, struct statfs *buf)
1231  *
1232  * Get filesystem statistics.
1233  */
1234 int
1235 sys_fstatfs(struct fstatfs_args *uap)
1236 {
1237         struct statfs buf;
1238         int error;
1239
1240         error = kern_fstatfs(uap->fd, &buf);
1241
1242         if (error == 0)
1243                 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1244         return (error);
1245 }
1246
1247 int
1248 kern_statvfs(struct nlookupdata *nd, struct statvfs *buf)
1249 {
1250         struct mount *mp;
1251         struct statvfs *sp;
1252         int error;
1253
1254         if ((error = nlookup(nd)) != 0)
1255                 return (error);
1256         mp = nd->nl_nch.mount;
1257         sp = &mp->mnt_vstat;
1258         if ((error = VFS_STATVFS(mp, sp, nd->nl_cred)) != 0)
1259                 return (error);
1260
1261         sp->f_flag = 0;
1262         if (mp->mnt_flag & MNT_RDONLY)
1263                 sp->f_flag |= ST_RDONLY;
1264         if (mp->mnt_flag & MNT_NOSUID)
1265                 sp->f_flag |= ST_NOSUID;
1266         bcopy(sp, buf, sizeof(*buf));
1267         return (0);
1268 }
1269
1270 /*
1271  * statfs_args(char *path, struct statfs *buf)
1272  *
1273  * Get filesystem statistics.
1274  */
1275 int
1276 sys_statvfs(struct statvfs_args *uap)
1277 {
1278         struct nlookupdata nd;
1279         struct statvfs buf;
1280         int error;
1281
1282         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1283         if (error == 0)
1284                 error = kern_statvfs(&nd, &buf);
1285         nlookup_done(&nd);
1286         if (error == 0)
1287                 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1288         return (error);
1289 }
1290
1291 int
1292 kern_fstatvfs(int fd, struct statvfs *buf)
1293 {
1294         struct thread *td = curthread;
1295         struct proc *p = td->td_proc;
1296         struct file *fp;
1297         struct mount *mp;
1298         struct statvfs *sp;
1299         int error;
1300
1301         KKASSERT(p);
1302         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
1303                 return (error);
1304         if ((mp = fp->f_nchandle.mount) == NULL)
1305                 mp = ((struct vnode *)fp->f_data)->v_mount;
1306         if (mp == NULL) {
1307                 error = EBADF;
1308                 goto done;
1309         }
1310         if (fp->f_cred == NULL) {
1311                 error = EINVAL;
1312                 goto done;
1313         }
1314         sp = &mp->mnt_vstat;
1315         if ((error = VFS_STATVFS(mp, sp, fp->f_cred)) != 0)
1316                 goto done;
1317
1318         sp->f_flag = 0;
1319         if (mp->mnt_flag & MNT_RDONLY)
1320                 sp->f_flag |= ST_RDONLY;
1321         if (mp->mnt_flag & MNT_NOSUID)
1322                 sp->f_flag |= ST_NOSUID;
1323
1324         bcopy(sp, buf, sizeof(*buf));
1325         error = 0;
1326 done:
1327         fdrop(fp);
1328         return (error);
1329 }
1330
1331 /*
1332  * fstatfs_args(int fd, struct statfs *buf)
1333  *
1334  * Get filesystem statistics.
1335  */
1336 int
1337 sys_fstatvfs(struct fstatvfs_args *uap)
1338 {
1339         struct statvfs buf;
1340         int error;
1341
1342         error = kern_fstatvfs(uap->fd, &buf);
1343
1344         if (error == 0)
1345                 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1346         return (error);
1347 }
1348
1349 /*
1350  * getfsstat_args(struct statfs *buf, long bufsize, int flags)
1351  *
1352  * Get statistics on all filesystems.
1353  */
1354
1355 struct getfsstat_info {
1356         struct statfs *sfsp;
1357         long count;
1358         long maxcount;
1359         int error;
1360         int flags;
1361         struct thread *td;
1362 };
1363
1364 static int getfsstat_callback(struct mount *, void *);
1365
1366 int
1367 sys_getfsstat(struct getfsstat_args *uap)
1368 {
1369         struct thread *td = curthread;
1370         struct getfsstat_info info;
1371
1372         bzero(&info, sizeof(info));
1373
1374         info.maxcount = uap->bufsize / sizeof(struct statfs);
1375         info.sfsp = uap->buf;
1376         info.count = 0;
1377         info.flags = uap->flags;
1378         info.td = td;
1379
1380         mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
1381         if (info.sfsp && info.count > info.maxcount)
1382                 uap->sysmsg_result = info.maxcount;
1383         else
1384                 uap->sysmsg_result = info.count;
1385         return (info.error);
1386 }
1387
1388 static int
1389 getfsstat_callback(struct mount *mp, void *data)
1390 {
1391         struct getfsstat_info *info = data;
1392         struct statfs *sp;
1393         char *freepath;
1394         char *fullpath;
1395         int error;
1396
1397         if (info->sfsp && info->count < info->maxcount) {
1398                 if (info->td->td_proc &&
1399                     !chroot_visible_mnt(mp, info->td->td_proc)) {
1400                         return(0);
1401                 }
1402                 sp = &mp->mnt_stat;
1403
1404                 /*
1405                  * If MNT_NOWAIT or MNT_LAZY is specified, do not
1406                  * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1407                  * overrides MNT_WAIT.
1408                  */
1409                 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1410                     (info->flags & MNT_WAIT)) &&
1411                     (error = VFS_STATFS(mp, sp, info->td->td_ucred))) {
1412                         return(0);
1413                 }
1414                 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1415
1416                 error = mount_path(info->td->td_proc, mp, &fullpath, &freepath);
1417                 if (error) {
1418                         info->error = error;
1419                         return(-1);
1420                 }
1421                 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1422                 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1423                 kfree(freepath, M_TEMP);
1424
1425                 error = copyout(sp, info->sfsp, sizeof(*sp));
1426                 if (error) {
1427                         info->error = error;
1428                         return (-1);
1429                 }
1430                 ++info->sfsp;
1431         }
1432         info->count++;
1433         return(0);
1434 }
1435
1436 /*
1437  * getvfsstat_args(struct statfs *buf, struct statvfs *vbuf,
1438                    long bufsize, int flags)
1439  *
1440  * Get statistics on all filesystems.
1441  */
1442
1443 struct getvfsstat_info {
1444         struct statfs *sfsp;
1445         struct statvfs *vsfsp;
1446         long count;
1447         long maxcount;
1448         int error;
1449         int flags;
1450         struct thread *td;
1451 };
1452
1453 static int getvfsstat_callback(struct mount *, void *);
1454
1455 int
1456 sys_getvfsstat(struct getvfsstat_args *uap)
1457 {
1458         struct thread *td = curthread;
1459         struct getvfsstat_info info;
1460
1461         bzero(&info, sizeof(info));
1462
1463         info.maxcount = uap->vbufsize / sizeof(struct statvfs);
1464         info.sfsp = uap->buf;
1465         info.vsfsp = uap->vbuf;
1466         info.count = 0;
1467         info.flags = uap->flags;
1468         info.td = td;
1469
1470         mountlist_scan(getvfsstat_callback, &info, MNTSCAN_FORWARD);
1471         if (info.vsfsp && info.count > info.maxcount)
1472                 uap->sysmsg_result = info.maxcount;
1473         else
1474                 uap->sysmsg_result = info.count;
1475         return (info.error);
1476 }
1477
1478 static int
1479 getvfsstat_callback(struct mount *mp, void *data)
1480 {
1481         struct getvfsstat_info *info = data;
1482         struct statfs *sp;
1483         struct statvfs *vsp;
1484         char *freepath;
1485         char *fullpath;
1486         int error;
1487
1488         if (info->vsfsp && info->count < info->maxcount) {
1489                 if (info->td->td_proc &&
1490                     !chroot_visible_mnt(mp, info->td->td_proc)) {
1491                         return(0);
1492                 }
1493                 sp = &mp->mnt_stat;
1494                 vsp = &mp->mnt_vstat;
1495
1496                 /*
1497                  * If MNT_NOWAIT or MNT_LAZY is specified, do not
1498                  * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1499                  * overrides MNT_WAIT.
1500                  */
1501                 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1502                     (info->flags & MNT_WAIT)) &&
1503                     (error = VFS_STATFS(mp, sp, info->td->td_ucred))) {
1504                         return(0);
1505                 }
1506                 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1507
1508                 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1509                     (info->flags & MNT_WAIT)) &&
1510                     (error = VFS_STATVFS(mp, vsp, info->td->td_ucred))) {
1511                         return(0);
1512                 }
1513                 vsp->f_flag = 0;
1514                 if (mp->mnt_flag & MNT_RDONLY)
1515                         vsp->f_flag |= ST_RDONLY;
1516                 if (mp->mnt_flag & MNT_NOSUID)
1517                         vsp->f_flag |= ST_NOSUID;
1518
1519                 error = mount_path(info->td->td_proc, mp, &fullpath, &freepath);
1520                 if (error) {
1521                         info->error = error;
1522                         return(-1);
1523                 }
1524                 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1525                 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1526                 kfree(freepath, M_TEMP);
1527
1528                 error = copyout(sp, info->sfsp, sizeof(*sp));
1529                 if (error == 0)
1530                         error = copyout(vsp, info->vsfsp, sizeof(*vsp));
1531                 if (error) {
1532                         info->error = error;
1533                         return (-1);
1534                 }
1535                 ++info->sfsp;
1536                 ++info->vsfsp;
1537         }
1538         info->count++;
1539         return(0);
1540 }
1541
1542
1543 /*
1544  * fchdir_args(int fd)
1545  *
1546  * Change current working directory to a given file descriptor.
1547  */
1548 int
1549 sys_fchdir(struct fchdir_args *uap)
1550 {
1551         struct thread *td = curthread;
1552         struct proc *p = td->td_proc;
1553         struct filedesc *fdp = p->p_fd;
1554         struct vnode *vp, *ovp;
1555         struct mount *mp;
1556         struct file *fp;
1557         struct nchandle nch, onch, tnch;
1558         int error;
1559
1560         if ((error = holdvnode(fdp, uap->fd, &fp)) != 0)
1561                 return (error);
1562         lwkt_gettoken(&p->p_token);
1563         vp = (struct vnode *)fp->f_data;
1564         vref(vp);
1565         vn_lock(vp, LK_SHARED | LK_RETRY);
1566         if (fp->f_nchandle.ncp == NULL)
1567                 error = ENOTDIR;
1568         else
1569                 error = checkvp_chdir(vp, td);
1570         if (error) {
1571                 vput(vp);
1572                 goto done;
1573         }
1574         cache_copy(&fp->f_nchandle, &nch);
1575
1576         /*
1577          * If the ncp has become a mount point, traverse through
1578          * the mount point.
1579          */
1580
1581         while (!error && (nch.ncp->nc_flag & NCF_ISMOUNTPT) &&
1582                (mp = cache_findmount(&nch)) != NULL
1583         ) {
1584                 error = nlookup_mp(mp, &tnch);
1585                 if (error == 0) {
1586                         cache_unlock(&tnch);    /* leave ref intact */
1587                         vput(vp);
1588                         vp = tnch.ncp->nc_vp;
1589                         error = vget(vp, LK_SHARED);
1590                         KKASSERT(error == 0);
1591                         cache_drop(&nch);
1592                         nch = tnch;
1593                 }
1594                 cache_dropmount(mp);
1595         }
1596         if (error == 0) {
1597                 ovp = fdp->fd_cdir;
1598                 onch = fdp->fd_ncdir;
1599                 vn_unlock(vp);          /* leave ref intact */
1600                 fdp->fd_cdir = vp;
1601                 fdp->fd_ncdir = nch;
1602                 cache_drop(&onch);
1603                 vrele(ovp);
1604         } else {
1605                 cache_drop(&nch);
1606                 vput(vp);
1607         }
1608         fdrop(fp);
1609 done:
1610         lwkt_reltoken(&p->p_token);
1611         return (error);
1612 }
1613
1614 int
1615 kern_chdir(struct nlookupdata *nd)
1616 {
1617         struct thread *td = curthread;
1618         struct proc *p = td->td_proc;
1619         struct filedesc *fdp = p->p_fd;
1620         struct vnode *vp, *ovp;
1621         struct nchandle onch;
1622         int error;
1623
1624         nd->nl_flags |= NLC_SHAREDLOCK;
1625         if ((error = nlookup(nd)) != 0)
1626                 return (error);
1627         if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
1628                 return (ENOENT);
1629         if ((error = vget(vp, LK_SHARED)) != 0)
1630                 return (error);
1631
1632         lwkt_gettoken(&p->p_token);
1633         error = checkvp_chdir(vp, td);
1634         vn_unlock(vp);
1635         if (error == 0) {
1636                 ovp = fdp->fd_cdir;
1637                 onch = fdp->fd_ncdir;
1638                 cache_unlock(&nd->nl_nch);      /* leave reference intact */
1639                 fdp->fd_ncdir = nd->nl_nch;
1640                 fdp->fd_cdir = vp;
1641                 cache_drop(&onch);
1642                 vrele(ovp);
1643                 cache_zero(&nd->nl_nch);
1644         } else {
1645                 vrele(vp);
1646         }
1647         lwkt_reltoken(&p->p_token);
1648         return (error);
1649 }
1650
1651 /*
1652  * chdir_args(char *path)
1653  *
1654  * Change current working directory (``.'').
1655  */
1656 int
1657 sys_chdir(struct chdir_args *uap)
1658 {
1659         struct nlookupdata nd;
1660         int error;
1661
1662         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1663         if (error == 0)
1664                 error = kern_chdir(&nd);
1665         nlookup_done(&nd);
1666         return (error);
1667 }
1668
1669 /*
1670  * Helper function for raised chroot(2) security function:  Refuse if
1671  * any filedescriptors are open directories.
1672  */
1673 static int
1674 chroot_refuse_vdir_fds(struct filedesc *fdp)
1675 {
1676         struct vnode *vp;
1677         struct file *fp;
1678         int error;
1679         int fd;
1680
1681         for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
1682                 if ((error = holdvnode(fdp, fd, &fp)) != 0)
1683                         continue;
1684                 vp = (struct vnode *)fp->f_data;
1685                 if (vp->v_type != VDIR) {
1686                         fdrop(fp);
1687                         continue;
1688                 }
1689                 fdrop(fp);
1690                 return(EPERM);
1691         }
1692         return (0);
1693 }
1694
1695 /*
1696  * This sysctl determines if we will allow a process to chroot(2) if it
1697  * has a directory open:
1698  *      0: disallowed for all processes.
1699  *      1: allowed for processes that were not already chroot(2)'ed.
1700  *      2: allowed for all processes.
1701  */
1702
1703 static int chroot_allow_open_directories = 1;
1704
1705 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1706      &chroot_allow_open_directories, 0, "");
1707
1708 /*
1709  * chroot to the specified namecache entry.  We obtain the vp from the
1710  * namecache data.  The passed ncp must be locked and referenced and will
1711  * remain locked and referenced on return.
1712  */
1713 int
1714 kern_chroot(struct nchandle *nch)
1715 {
1716         struct thread *td = curthread;
1717         struct proc *p = td->td_proc;
1718         struct filedesc *fdp = p->p_fd;
1719         struct vnode *vp;
1720         int error;
1721
1722         /*
1723          * Only privileged user can chroot
1724          */
1725         error = priv_check_cred(td->td_ucred, PRIV_VFS_CHROOT, 0);
1726         if (error)
1727                 return (error);
1728
1729         /*
1730          * Disallow open directory descriptors (fchdir() breakouts).
1731          */
1732         if (chroot_allow_open_directories == 0 ||
1733            (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1734                 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1735                         return (error);
1736         }
1737         if ((vp = nch->ncp->nc_vp) == NULL)
1738                 return (ENOENT);
1739
1740         if ((error = vget(vp, LK_SHARED)) != 0)
1741                 return (error);
1742
1743         /*
1744          * Check the validity of vp as a directory to change to and
1745          * associate it with rdir/jdir.
1746          */
1747         error = checkvp_chdir(vp, td);
1748         vn_unlock(vp);                  /* leave reference intact */
1749         if (error == 0) {
1750                 vrele(fdp->fd_rdir);
1751                 fdp->fd_rdir = vp;      /* reference inherited by fd_rdir */
1752                 cache_drop(&fdp->fd_nrdir);
1753                 cache_copy(nch, &fdp->fd_nrdir);
1754                 if (fdp->fd_jdir == NULL) {
1755                         fdp->fd_jdir = vp;
1756                         vref(fdp->fd_jdir);
1757                         cache_copy(nch, &fdp->fd_njdir);
1758                 }
1759         } else {
1760                 vrele(vp);
1761         }
1762         return (error);
1763 }
1764
1765 /*
1766  * chroot_args(char *path)
1767  *
1768  * Change notion of root (``/'') directory.
1769  */
1770 int
1771 sys_chroot(struct chroot_args *uap)
1772 {
1773         struct thread *td __debugvar = curthread;
1774         struct nlookupdata nd;
1775         int error;
1776
1777         KKASSERT(td->td_proc);
1778         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1779         if (error == 0) {
1780                 nd.nl_flags |= NLC_EXEC;
1781                 error = nlookup(&nd);
1782                 if (error == 0)
1783                         error = kern_chroot(&nd.nl_nch);
1784         }
1785         nlookup_done(&nd);
1786         return(error);
1787 }
1788
1789 int
1790 sys_chroot_kernel(struct chroot_kernel_args *uap)
1791 {
1792         struct thread *td = curthread;
1793         struct nlookupdata nd;
1794         struct nchandle *nch;
1795         struct vnode *vp;
1796         int error;
1797
1798         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1799         if (error)
1800                 goto error_nond;
1801
1802         error = nlookup(&nd);
1803         if (error)
1804                 goto error_out;
1805
1806         nch = &nd.nl_nch;
1807
1808         error = priv_check_cred(td->td_ucred, PRIV_VFS_CHROOT, 0);
1809         if (error)
1810                 goto error_out;
1811
1812         if ((vp = nch->ncp->nc_vp) == NULL) {
1813                 error = ENOENT;
1814                 goto error_out;
1815         }
1816
1817         if ((error = cache_vref(nch, nd.nl_cred, &vp)) != 0)
1818                 goto error_out;
1819
1820         kprintf("chroot_kernel: set new rootnch/rootvnode to %s\n", uap->path);
1821         get_mplock();
1822         vfs_cache_setroot(vp, cache_hold(nch));
1823         rel_mplock();
1824
1825 error_out:
1826         nlookup_done(&nd);
1827 error_nond:
1828         return(error);
1829 }
1830
1831 /*
1832  * Common routine for chroot and chdir.  Given a locked, referenced vnode,
1833  * determine whether it is legal to chdir to the vnode.  The vnode's state
1834  * is not changed by this call.
1835  */
1836 static int
1837 checkvp_chdir(struct vnode *vp, struct thread *td)
1838 {
1839         int error;
1840
1841         if (vp->v_type != VDIR)
1842                 error = ENOTDIR;
1843         else
1844                 error = VOP_EACCESS(vp, VEXEC, td->td_ucred);
1845         return (error);
1846 }
1847
1848 int
1849 kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
1850 {
1851         struct thread *td = curthread;
1852         struct proc *p = td->td_proc;
1853         struct lwp *lp = td->td_lwp;
1854         struct filedesc *fdp = p->p_fd;
1855         int cmode, flags;
1856         struct file *nfp;
1857         struct file *fp;
1858         struct vnode *vp;
1859         int type, indx, error = 0;
1860         struct flock lf;
1861
1862         if ((oflags & O_ACCMODE) == O_ACCMODE)
1863                 return (EINVAL);
1864         flags = FFLAGS(oflags);
1865         error = falloc(lp, &nfp, NULL);
1866         if (error)
1867                 return (error);
1868         fp = nfp;
1869         cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1870
1871         /*
1872          * XXX p_dupfd is a real mess.  It allows a device to return a
1873          * file descriptor to be duplicated rather then doing the open
1874          * itself.
1875          */
1876         lp->lwp_dupfd = -1;
1877
1878         /*
1879          * Call vn_open() to do the lookup and assign the vnode to the
1880          * file pointer.  vn_open() does not change the ref count on fp
1881          * and the vnode, on success, will be inherited by the file pointer
1882          * and unlocked.
1883          *
1884          * Request a shared lock on the vnode if possible.
1885          */
1886         nd->nl_flags |= NLC_LOCKVP;
1887         if ((flags & (O_CREAT|O_TRUNC)) == 0)
1888                 nd->nl_flags |= NLC_SHAREDLOCK;
1889
1890         error = vn_open(nd, fp, flags, cmode);
1891         nlookup_done(nd);
1892
1893         if (error) {
1894                 /*
1895                  * handle special fdopen() case.  bleh.  dupfdopen() is
1896                  * responsible for dropping the old contents of ofiles[indx]
1897                  * if it succeeds.
1898                  *
1899                  * Note that fsetfd() will add a ref to fp which represents
1900                  * the fd_files[] assignment.  We must still drop our
1901                  * reference.
1902                  */
1903                 if ((error == ENODEV || error == ENXIO) && lp->lwp_dupfd >= 0) {
1904                         if (fdalloc(p, 0, &indx) == 0) {
1905                                 error = dupfdopen(fdp, indx, lp->lwp_dupfd, flags, error);
1906                                 if (error == 0) {
1907                                         *res = indx;
1908                                         fdrop(fp);      /* our ref */
1909                                         return (0);
1910                                 }
1911                                 fsetfd(fdp, NULL, indx);
1912                         }
1913                 }
1914                 fdrop(fp);      /* our ref */
1915                 if (error == ERESTART)
1916                         error = EINTR;
1917                 return (error);
1918         }
1919
1920         /*
1921          * ref the vnode for ourselves so it can't be ripped out from under
1922          * is.  XXX need an ND flag to request that the vnode be returned
1923          * anyway.
1924          *
1925          * Reserve a file descriptor but do not assign it until the open
1926          * succeeds.
1927          */
1928         vp = (struct vnode *)fp->f_data;
1929         vref(vp);
1930         if ((error = fdalloc(p, 0, &indx)) != 0) {
1931                 fdrop(fp);
1932                 vrele(vp);
1933                 return (error);
1934         }
1935
1936         /*
1937          * If no error occurs the vp will have been assigned to the file
1938          * pointer.
1939          */
1940         lp->lwp_dupfd = 0;
1941
1942         if (flags & (O_EXLOCK | O_SHLOCK)) {
1943                 lf.l_whence = SEEK_SET;
1944                 lf.l_start = 0;
1945                 lf.l_len = 0;
1946                 if (flags & O_EXLOCK)
1947                         lf.l_type = F_WRLCK;
1948                 else
1949                         lf.l_type = F_RDLCK;
1950                 if (flags & FNONBLOCK)
1951                         type = 0;
1952                 else
1953                         type = F_WAIT;
1954
1955                 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1956                         /*
1957                          * lock request failed.  Clean up the reserved
1958                          * descriptor.
1959                          */
1960                         vrele(vp);
1961                         fsetfd(fdp, NULL, indx);
1962                         fdrop(fp);
1963                         return (error);
1964                 }
1965                 atomic_set_int(&fp->f_flag, FHASLOCK); /* race ok */
1966         }
1967 #if 0
1968         /*
1969          * Assert that all regular file vnodes were created with a object.
1970          */
1971         KASSERT(vp->v_type != VREG || vp->v_object != NULL,
1972                 ("open: regular file has no backing object after vn_open"));
1973 #endif
1974
1975         vrele(vp);
1976
1977         /*
1978          * release our private reference, leaving the one associated with the
1979          * descriptor table intact.
1980          */
1981         if (oflags & O_CLOEXEC)
1982                 fdp->fd_files[indx].fileflags |= UF_EXCLOSE;
1983         fsetfd(fdp, fp, indx);
1984         fdrop(fp);
1985         *res = indx;
1986         return (error);
1987 }
1988
1989 /*
1990  * open_args(char *path, int flags, int mode)
1991  *
1992  * Check permissions, allocate an open file structure,
1993  * and call the device open routine if any.
1994  */
1995 int
1996 sys_open(struct open_args *uap)
1997 {
1998         struct nlookupdata nd;
1999         int error;
2000
2001         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2002         if (error == 0) {
2003                 error = kern_open(&nd, uap->flags,
2004                                     uap->mode, &uap->sysmsg_result);
2005         }
2006         nlookup_done(&nd);
2007         return (error);
2008 }
2009
2010 /*
2011  * openat_args(int fd, char *path, int flags, int mode)
2012  */
2013 int
2014 sys_openat(struct openat_args *uap)
2015 {
2016         struct nlookupdata nd;
2017         int error;
2018         struct file *fp;
2019
2020         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
2021         if (error == 0) {
2022                 error = kern_open(&nd, uap->flags, uap->mode,
2023                                         &uap->sysmsg_result);
2024         }
2025         nlookup_done_at(&nd, fp);
2026         return (error);
2027 }
2028
2029 int
2030 kern_mknod(struct nlookupdata *nd, int mode, int rmajor, int rminor)
2031 {
2032         struct thread *td = curthread;
2033         struct proc *p = td->td_proc;
2034         struct vnode *vp;
2035         struct vattr vattr;
2036         int error;
2037         int whiteout = 0;
2038
2039         KKASSERT(p);
2040
2041         VATTR_NULL(&vattr);
2042         vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
2043         vattr.va_rmajor = rmajor;
2044         vattr.va_rminor = rminor;
2045
2046         switch (mode & S_IFMT) {
2047         case S_IFMT:    /* used by badsect to flag bad sectors */
2048                 error = priv_check_cred(td->td_ucred, PRIV_VFS_MKNOD_BAD, 0);
2049                 vattr.va_type = VBAD;
2050                 break;
2051         case S_IFCHR:
2052                 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
2053                 vattr.va_type = VCHR;
2054                 break;
2055         case S_IFBLK:
2056                 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
2057                 vattr.va_type = VBLK;
2058                 break;
2059         case S_IFWHT:
2060                 error = priv_check_cred(td->td_ucred, PRIV_VFS_MKNOD_WHT, 0);
2061                 whiteout = 1;
2062                 break;
2063         case S_IFDIR:   /* special directories support for HAMMER */
2064                 error = priv_check_cred(td->td_ucred, PRIV_VFS_MKNOD_DIR, 0);
2065                 vattr.va_type = VDIR;
2066                 break;
2067         default:
2068                 error = EINVAL;
2069                 break;
2070         }
2071
2072         if (error)
2073                 return (error);
2074
2075         bwillinode(1);
2076         nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
2077         if ((error = nlookup(nd)) != 0)
2078                 return (error);
2079         if (nd->nl_nch.ncp->nc_vp)
2080                 return (EEXIST);
2081         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2082                 return (error);
2083
2084         if (whiteout) {
2085                 error = VOP_NWHITEOUT(&nd->nl_nch, nd->nl_dvp,
2086                                       nd->nl_cred, NAMEI_CREATE);
2087         } else {
2088                 vp = NULL;
2089                 error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp,
2090                                    &vp, nd->nl_cred, &vattr);
2091                 if (error == 0)
2092                         vput(vp);
2093         }
2094         return (error);
2095 }
2096
2097 /*
2098  * mknod_args(char *path, int mode, int dev)
2099  *
2100  * Create a special file.
2101  */
2102 int
2103 sys_mknod(struct mknod_args *uap)
2104 {
2105         struct nlookupdata nd;
2106         int error;
2107
2108         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2109         if (error == 0) {
2110                 error = kern_mknod(&nd, uap->mode,
2111                                    umajor(uap->dev), uminor(uap->dev));
2112         }
2113         nlookup_done(&nd);
2114         return (error);
2115 }
2116
2117 /*
2118  * mknodat_args(int fd, char *path, mode_t mode, dev_t dev)
2119  *
2120  * Create a special file.  The path is relative to the directory associated
2121  * with fd.
2122  */
2123 int
2124 sys_mknodat(struct mknodat_args *uap)
2125 {
2126         struct nlookupdata nd;
2127         struct file *fp;
2128         int error;
2129
2130         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
2131         if (error == 0) {
2132                 error = kern_mknod(&nd, uap->mode,
2133                                    umajor(uap->dev), uminor(uap->dev));
2134         }
2135         nlookup_done_at(&nd, fp);
2136         return (error);
2137 }
2138
2139 int
2140 kern_mkfifo(struct nlookupdata *nd, int mode)
2141 {
2142         struct thread *td = curthread;
2143         struct proc *p = td->td_proc;
2144         struct vattr vattr;
2145         struct vnode *vp;
2146         int error;
2147
2148         bwillinode(1);
2149
2150         nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
2151         if ((error = nlookup(nd)) != 0)
2152                 return (error);
2153         if (nd->nl_nch.ncp->nc_vp)
2154                 return (EEXIST);
2155         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2156                 return (error);
2157
2158         VATTR_NULL(&vattr);
2159         vattr.va_type = VFIFO;
2160         vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
2161         vp = NULL;
2162         error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp, &vp, nd->nl_cred, &vattr);
2163         if (error == 0)
2164                 vput(vp);
2165         return (error);
2166 }
2167
2168 /*
2169  * mkfifo_args(char *path, int mode)
2170  *
2171  * Create a named pipe.
2172  */
2173 int
2174 sys_mkfifo(struct mkfifo_args *uap)
2175 {
2176         struct nlookupdata nd;
2177         int error;
2178
2179         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2180         if (error == 0)
2181                 error = kern_mkfifo(&nd, uap->mode);
2182         nlookup_done(&nd);
2183         return (error);
2184 }
2185
2186 /*
2187  * mkfifoat_args(int fd, char *path, mode_t mode)
2188  *
2189  * Create a named pipe.  The path is relative to the directory associated
2190  * with fd.
2191  */
2192 int
2193 sys_mkfifoat(struct mkfifoat_args *uap)
2194 {
2195         struct nlookupdata nd;
2196         struct file *fp;
2197         int error;
2198
2199         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
2200         if (error == 0)
2201                 error = kern_mkfifo(&nd, uap->mode);
2202         nlookup_done_at(&nd, fp);
2203         return (error);
2204 }
2205
2206 static int hardlink_check_uid = 0;
2207 SYSCTL_INT(_security, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
2208     &hardlink_check_uid, 0,
2209     "Unprivileged processes cannot create hard links to files owned by other "
2210     "users");
2211 static int hardlink_check_gid = 0;
2212 SYSCTL_INT(_security, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
2213     &hardlink_check_gid, 0,
2214     "Unprivileged processes cannot create hard links to files owned by other "
2215     "groups");
2216
2217 static int
2218 can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
2219 {
2220         struct vattr va;
2221         int error;
2222
2223         /*
2224          * Shortcut if disabled
2225          */
2226         if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
2227                 return (0);
2228
2229         /*
2230          * Privileged user can always hardlink
2231          */
2232         if (priv_check_cred(cred, PRIV_VFS_LINK, 0) == 0)
2233                 return (0);
2234
2235         /*
2236          * Otherwise only if the originating file is owned by the
2237          * same user or group.  Note that any group is allowed if
2238          * the file is owned by the caller.
2239          */
2240         error = VOP_GETATTR(vp, &va);
2241         if (error != 0)
2242                 return (error);
2243
2244         if (hardlink_check_uid) {
2245                 if (cred->cr_uid != va.va_uid)
2246                         return (EPERM);
2247         }
2248
2249         if (hardlink_check_gid) {
2250                 if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
2251                         return (EPERM);
2252         }
2253
2254         return (0);
2255 }
2256
2257 int
2258 kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
2259 {
2260         struct thread *td = curthread;
2261         struct vnode *vp;
2262         int error;
2263
2264         /*
2265          * Lookup the source and obtained a locked vnode.
2266          *
2267          * You may only hardlink a file which you have write permission
2268          * on or which you own.
2269          *
2270          * XXX relookup on vget failure / race ?
2271          */
2272         bwillinode(1);
2273         nd->nl_flags |= NLC_WRITE | NLC_OWN | NLC_HLINK;
2274         if ((error = nlookup(nd)) != 0)
2275                 return (error);
2276         vp = nd->nl_nch.ncp->nc_vp;
2277         KKASSERT(vp != NULL);
2278         if (vp->v_type == VDIR)
2279                 return (EPERM);         /* POSIX */
2280         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2281                 return (error);
2282         if ((error = vget(vp, LK_EXCLUSIVE)) != 0)
2283                 return (error);
2284
2285         /*
2286          * Unlock the source so we can lookup the target without deadlocking
2287          * (XXX vp is locked already, possible other deadlock?).  The target
2288          * must not exist.
2289          */
2290         KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
2291         nd->nl_flags &= ~NLC_NCPISLOCKED;
2292         cache_unlock(&nd->nl_nch);
2293         vn_unlock(vp);
2294
2295         linknd->nl_flags |= NLC_CREATE | NLC_REFDVP;
2296         if ((error = nlookup(linknd)) != 0) {
2297                 vrele(vp);
2298                 return (error);
2299         }
2300         if (linknd->nl_nch.ncp->nc_vp) {
2301                 vrele(vp);
2302                 return (EEXIST);
2303         }
2304         error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_FAILRECLAIM);
2305         if (error) {
2306                 vrele(vp);
2307                 return (error);
2308         }
2309
2310         /*
2311          * Finally run the new API VOP.
2312          */
2313         error = can_hardlink(vp, td, td->td_ucred);
2314         if (error == 0) {
2315                 error = VOP_NLINK(&linknd->nl_nch, linknd->nl_dvp,
2316                                   vp, linknd->nl_cred);
2317         }
2318         vput(vp);
2319         return (error);
2320 }
2321
2322 /*
2323  * link_args(char *path, char *link)
2324  *
2325  * Make a hard file link.
2326  */
2327 int
2328 sys_link(struct link_args *uap)
2329 {
2330         struct nlookupdata nd, linknd;
2331         int error;
2332
2333         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2334         if (error == 0) {
2335                 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
2336                 if (error == 0)
2337                         error = kern_link(&nd, &linknd);
2338                 nlookup_done(&linknd);
2339         }
2340         nlookup_done(&nd);
2341         return (error);
2342 }
2343
2344 /*
2345  * linkat_args(int fd1, char *path1, int fd2, char *path2, int flags)
2346  *
2347  * Make a hard file link. The path1 argument is relative to the directory
2348  * associated with fd1, and similarly the path2 argument is relative to
2349  * the directory associated with fd2.
2350  */
2351 int
2352 sys_linkat(struct linkat_args *uap)
2353 {
2354         struct nlookupdata nd, linknd;
2355         struct file *fp1, *fp2;
2356         int error;
2357
2358         error = nlookup_init_at(&nd, &fp1, uap->fd1, uap->path1, UIO_USERSPACE,
2359             (uap->flags & AT_SYMLINK_FOLLOW) ? NLC_FOLLOW : 0);
2360         if (error == 0) {
2361                 error = nlookup_init_at(&linknd, &fp2, uap->fd2,
2362                     uap->path2, UIO_USERSPACE, 0);
2363                 if (error == 0)
2364                         error = kern_link(&nd, &linknd);
2365                 nlookup_done_at(&linknd, fp2);
2366         }
2367         nlookup_done_at(&nd, fp1);
2368         return (error);
2369 }
2370
2371 int
2372 kern_symlink(struct nlookupdata *nd, char *path, int mode)
2373 {
2374         struct vattr vattr;
2375         struct vnode *vp;
2376         struct vnode *dvp;
2377         int error;
2378
2379         bwillinode(1);
2380         nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
2381         if ((error = nlookup(nd)) != 0)
2382                 return (error);
2383         if (nd->nl_nch.ncp->nc_vp)
2384                 return (EEXIST);
2385         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2386                 return (error);
2387         dvp = nd->nl_dvp;
2388         VATTR_NULL(&vattr);
2389         vattr.va_mode = mode;
2390         error = VOP_NSYMLINK(&nd->nl_nch, dvp, &vp, nd->nl_cred, &vattr, path);
2391         if (error == 0)
2392                 vput(vp);
2393         return (error);
2394 }
2395
2396 /*
2397  * symlink(char *path, char *link)
2398  *
2399  * Make a symbolic link.
2400  */
2401 int
2402 sys_symlink(struct symlink_args *uap)
2403 {
2404         struct thread *td = curthread;
2405         struct nlookupdata nd;
2406         char *path;
2407         int error;
2408         int mode;
2409
2410         path = objcache_get(namei_oc, M_WAITOK);
2411         error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
2412         if (error == 0) {
2413                 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
2414                 if (error == 0) {
2415                         mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
2416                         error = kern_symlink(&nd, path, mode);
2417                 }
2418                 nlookup_done(&nd);
2419         }
2420         objcache_put(namei_oc, path);
2421         return (error);
2422 }
2423
2424 /*
2425  * symlinkat_args(char *path1, int fd, char *path2)
2426  *
2427  * Make a symbolic link.  The path2 argument is relative to the directory
2428  * associated with fd.
2429  */
2430 int
2431 sys_symlinkat(struct symlinkat_args *uap)
2432 {
2433         struct thread *td = curthread;
2434         struct nlookupdata nd;
2435         struct file *fp;
2436         char *path1;
2437         int error;
2438         int mode;
2439
2440         path1 = objcache_get(namei_oc, M_WAITOK);
2441         error = copyinstr(uap->path1, path1, MAXPATHLEN, NULL);
2442         if (error == 0) {
2443                 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path2,
2444                     UIO_USERSPACE, 0);
2445                 if (error == 0) {
2446                         mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
2447                         error = kern_symlink(&nd, path1, mode);
2448                 }
2449                 nlookup_done_at(&nd, fp);
2450         }
2451         objcache_put(namei_oc, path1);
2452         return (error);
2453 }
2454
2455 /*
2456  * undelete_args(char *path)
2457  *
2458  * Delete a whiteout from the filesystem.
2459  */
2460 int
2461 sys_undelete(struct undelete_args *uap)
2462 {
2463         struct nlookupdata nd;
2464         int error;
2465
2466         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2467         bwillinode(1);
2468         nd.nl_flags |= NLC_DELETE | NLC_REFDVP;
2469         if (error == 0)
2470                 error = nlookup(&nd);
2471         if (error == 0)
2472                 error = ncp_writechk(&nd.nl_nch);
2473         if (error == 0) {
2474                 error = VOP_NWHITEOUT(&nd.nl_nch, nd.nl_dvp, nd.nl_cred,
2475                                       NAMEI_DELETE);
2476         }
2477         nlookup_done(&nd);
2478         return (error);
2479 }
2480
2481 int
2482 kern_unlink(struct nlookupdata *nd)
2483 {
2484         int error;
2485
2486         bwillinode(1);
2487         nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
2488         if ((error = nlookup(nd)) != 0)
2489                 return (error);
2490         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2491                 return (error);
2492         error = VOP_NREMOVE(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
2493         return (error);
2494 }
2495
2496 /*
2497  * unlink_args(char *path)
2498  *
2499  * Delete a name from the filesystem.
2500  */
2501 int
2502 sys_unlink(struct unlink_args *uap)
2503 {
2504         struct nlookupdata nd;
2505         int error;
2506
2507         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2508         if (error == 0)
2509                 error = kern_unlink(&nd);
2510         nlookup_done(&nd);
2511         return (error);
2512 }
2513
2514
2515 /*
2516  * unlinkat_args(int fd, char *path, int flags)
2517  *
2518  * Delete the file or directory entry pointed to by fd/path.
2519  */
2520 int
2521 sys_unlinkat(struct unlinkat_args *uap)
2522 {
2523         struct nlookupdata nd;
2524         struct file *fp;
2525         int error;
2526
2527         if (uap->flags & ~AT_REMOVEDIR)
2528                 return (EINVAL);
2529
2530         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
2531         if (error == 0) {
2532                 if (uap->flags & AT_REMOVEDIR)
2533                         error = kern_rmdir(&nd);
2534                 else
2535                         error = kern_unlink(&nd);
2536         }
2537         nlookup_done_at(&nd, fp);
2538         return (error);
2539 }
2540
2541 int
2542 kern_lseek(int fd, off_t offset, int whence, off_t *res)
2543 {
2544         struct thread *td = curthread;
2545         struct proc *p = td->td_proc;
2546         struct file *fp;
2547         struct vnode *vp;
2548         struct vattr vattr;
2549         off_t new_offset;
2550         int error;
2551
2552         fp = holdfp(p->p_fd, fd, -1);
2553         if (fp == NULL)
2554                 return (EBADF);
2555         if (fp->f_type != DTYPE_VNODE) {
2556                 error = ESPIPE;
2557                 goto done;
2558         }
2559         vp = (struct vnode *)fp->f_data;
2560
2561         switch (whence) {
2562         case L_INCR:
2563                 spin_lock(&fp->f_spin);
2564                 new_offset = fp->f_offset + offset;
2565                 error = 0;
2566                 break;
2567         case L_XTND:
2568                 error = VOP_GETATTR(vp, &vattr);
2569                 spin_lock(&fp->f_spin);
2570                 new_offset = offset + vattr.va_size;
2571                 break;
2572         case L_SET:
2573                 new_offset = offset;
2574                 error = 0;
2575                 spin_lock(&fp->f_spin);
2576                 break;
2577         default:
2578                 new_offset = 0;
2579                 error = EINVAL;
2580                 spin_lock(&fp->f_spin);
2581                 break;
2582         }
2583
2584         /*
2585          * Validate the seek position.  Negative offsets are not allowed
2586          * for regular files or directories.
2587          *
2588          * Normally we would also not want to allow negative offsets for
2589          * character and block-special devices.  However kvm addresses
2590          * on 64 bit architectures might appear to be negative and must
2591          * be allowed.
2592          */
2593         if (error == 0) {
2594                 if (new_offset < 0 &&
2595                     (vp->v_type == VREG || vp->v_type == VDIR)) {
2596                         error = EINVAL;
2597                 } else {
2598                         fp->f_offset = new_offset;
2599                 }
2600         }
2601         *res = fp->f_offset;
2602         spin_unlock(&fp->f_spin);
2603 done:
2604         fdrop(fp);
2605         return (error);
2606 }
2607
2608 /*
2609  * lseek_args(int fd, int pad, off_t offset, int whence)
2610  *
2611  * Reposition read/write file offset.
2612  */
2613 int
2614 sys_lseek(struct lseek_args *uap)
2615 {
2616         int error;
2617
2618         error = kern_lseek(uap->fd, uap->offset, uap->whence,
2619                            &uap->sysmsg_offset);
2620
2621         return (error);
2622 }
2623
2624 /*
2625  * Check if current process can access given file.  amode is a bitmask of *_OK
2626  * access bits.  flags is a bitmask of AT_* flags.
2627  */
2628 int
2629 kern_access(struct nlookupdata *nd, int amode, int flags)
2630 {
2631         struct vnode *vp;
2632         int error, mode;
2633
2634         if (flags & ~AT_EACCESS)
2635                 return (EINVAL);
2636         nd->nl_flags |= NLC_SHAREDLOCK;
2637         if ((error = nlookup(nd)) != 0)
2638                 return (error);
2639 retry:
2640         error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_SHARED, &vp);
2641         if (error)
2642                 return (error);
2643
2644         /* Flags == 0 means only check for existence. */
2645         if (amode) {
2646                 mode = 0;
2647                 if (amode & R_OK)
2648                         mode |= VREAD;
2649                 if (amode & W_OK)
2650                         mode |= VWRITE;
2651                 if (amode & X_OK)
2652                         mode |= VEXEC;
2653                 if ((mode & VWRITE) == 0 ||
2654                     (error = vn_writechk(vp, &nd->nl_nch)) == 0)
2655                         error = VOP_ACCESS_FLAGS(vp, mode, flags, nd->nl_cred);
2656
2657                 /*
2658                  * If the file handle is stale we have to re-resolve the
2659                  * entry with the ncp held exclusively.  This is a hack
2660                  * at the moment.
2661                  */
2662                 if (error == ESTALE) {
2663                         vput(vp);
2664                         cache_unlock(&nd->nl_nch);
2665                         cache_lock(&nd->nl_nch);
2666                         cache_setunresolved(&nd->nl_nch);
2667                         error = cache_resolve(&nd->nl_nch, nd->nl_cred);
2668                         if (error == 0) {
2669                                 vp = NULL;
2670                                 goto retry;
2671                         }
2672                         return(error);
2673                 }
2674         }
2675         vput(vp);
2676         return (error);
2677 }
2678
2679 /*
2680  * access_args(char *path, int flags)
2681  *
2682  * Check access permissions.
2683  */
2684 int
2685 sys_access(struct access_args *uap)
2686 {
2687         struct nlookupdata nd;
2688         int error;
2689
2690         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2691         if (error == 0)
2692                 error = kern_access(&nd, uap->flags, 0);
2693         nlookup_done(&nd);
2694         return (error);
2695 }
2696
2697
2698 /*
2699  * eaccess_args(char *path, int flags)
2700  *
2701  * Check access permissions.
2702  */
2703 int
2704 sys_eaccess(struct eaccess_args *uap)
2705 {
2706         struct nlookupdata nd;
2707         int error;
2708
2709         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2710         if (error == 0)
2711                 error = kern_access(&nd, uap->flags, AT_EACCESS);
2712         nlookup_done(&nd);
2713         return (error);
2714 }
2715
2716
2717 /*
2718  * faccessat_args(int fd, char *path, int amode, int flags)
2719  *
2720  * Check access permissions.
2721  */
2722 int
2723 sys_faccessat(struct faccessat_args *uap)
2724 {
2725         struct nlookupdata nd;
2726         struct file *fp;
2727         int error;
2728
2729         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE,
2730                                 NLC_FOLLOW);
2731         if (error == 0)
2732                 error = kern_access(&nd, uap->amode, uap->flags);
2733         nlookup_done_at(&nd, fp);
2734         return (error);
2735 }
2736
2737 int
2738 kern_stat(struct nlookupdata *nd, struct stat *st)
2739 {
2740         int error;
2741         struct vnode *vp;
2742
2743         nd->nl_flags |= NLC_SHAREDLOCK;
2744         if ((error = nlookup(nd)) != 0)
2745                 return (error);
2746 again:
2747         if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
2748                 return (ENOENT);
2749
2750         if ((error = vget(vp, LK_SHARED)) != 0)
2751                 return (error);
2752         error = vn_stat(vp, st, nd->nl_cred);
2753
2754         /*
2755          * If the file handle is stale we have to re-resolve the
2756          * entry with the ncp held exclusively.  This is a hack
2757          * at the moment.
2758          */
2759         if (error == ESTALE) {
2760                 vput(vp);
2761                 cache_unlock(&nd->nl_nch);
2762                 cache_lock(&nd->nl_nch);
2763                 cache_setunresolved(&nd->nl_nch);
2764                 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
2765                 if (error == 0)
2766                         goto again;
2767         } else {
2768                 vput(vp);
2769         }
2770         return (error);
2771 }
2772
2773 /*
2774  * stat_args(char *path, struct stat *ub)
2775  *
2776  * Get file status; this version follows links.
2777  */
2778 int
2779 sys_stat(struct stat_args *uap)
2780 {
2781         struct nlookupdata nd;
2782         struct stat st;
2783         int error;
2784
2785         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2786         if (error == 0) {
2787                 error = kern_stat(&nd, &st);
2788                 if (error == 0)
2789                         error = copyout(&st, uap->ub, sizeof(*uap->ub));
2790         }
2791         nlookup_done(&nd);
2792         return (error);
2793 }
2794
2795 /*
2796  * lstat_args(char *path, struct stat *ub)
2797  *
2798  * Get file status; this version does not follow links.
2799  */
2800 int
2801 sys_lstat(struct lstat_args *uap)
2802 {
2803         struct nlookupdata nd;
2804         struct stat st;
2805         int error;
2806
2807         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2808         if (error == 0) {
2809                 error = kern_stat(&nd, &st);
2810                 if (error == 0)
2811                         error = copyout(&st, uap->ub, sizeof(*uap->ub));
2812         }
2813         nlookup_done(&nd);
2814         return (error);
2815 }
2816
2817 /*
2818  * fstatat_args(int fd, char *path, struct stat *sb, int flags)
2819  *
2820  * Get status of file pointed to by fd/path.
2821  */
2822 int
2823 sys_fstatat(struct fstatat_args *uap)
2824 {
2825         struct nlookupdata nd;
2826         struct stat st;
2827         int error;
2828         int flags;
2829         struct file *fp;
2830
2831         if (uap->flags & ~AT_SYMLINK_NOFOLLOW)
2832                 return (EINVAL);
2833
2834         flags = (uap->flags & AT_SYMLINK_NOFOLLOW) ? 0 : NLC_FOLLOW;
2835
2836         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path,
2837                                 UIO_USERSPACE, flags);
2838         if (error == 0) {
2839                 error = kern_stat(&nd, &st);
2840                 if (error == 0)
2841                         error = copyout(&st, uap->sb, sizeof(*uap->sb));
2842         }
2843         nlookup_done_at(&nd, fp);
2844         return (error);
2845 }
2846
2847 static int
2848 kern_pathconf(char *path, int name, int flags, register_t *sysmsg_regp)
2849 {
2850         struct nlookupdata nd;
2851         struct vnode *vp;
2852         int error;
2853
2854         vp = NULL;
2855         error = nlookup_init(&nd, path, UIO_USERSPACE, flags);
2856         if (error == 0)
2857                 error = nlookup(&nd);
2858         if (error == 0)
2859                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
2860         nlookup_done(&nd);
2861         if (error == 0) {
2862                 error = VOP_PATHCONF(vp, name, sysmsg_regp);
2863                 vput(vp);
2864         }
2865         return (error);
2866 }
2867
2868 /*
2869  * pathconf_Args(char *path, int name)
2870  *
2871  * Get configurable pathname variables.
2872  */
2873 int
2874 sys_pathconf(struct pathconf_args *uap)
2875 {
2876         return (kern_pathconf(uap->path, uap->name, NLC_FOLLOW,
2877                 &uap->sysmsg_reg));
2878 }
2879
2880 /*
2881  * lpathconf_Args(char *path, int name)
2882  *
2883  * Get configurable pathname variables, but don't follow symlinks.
2884  */
2885 int
2886 sys_lpathconf(struct lpathconf_args *uap)
2887 {
2888         return (kern_pathconf(uap->path, uap->name, 0, &uap->sysmsg_reg));
2889 }
2890
2891 /*
2892  * XXX: daver
2893  * kern_readlink isn't properly split yet.  There is a copyin burried
2894  * in VOP_READLINK().
2895  */
2896 int
2897 kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
2898 {
2899         struct thread *td = curthread;
2900         struct vnode *vp;
2901         struct iovec aiov;
2902         struct uio auio;
2903         int error;
2904
2905         nd->nl_flags |= NLC_SHAREDLOCK;
2906         if ((error = nlookup(nd)) != 0)
2907                 return (error);
2908         error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_SHARED, &vp);
2909         if (error)
2910                 return (error);
2911         if (vp->v_type != VLNK) {
2912                 error = EINVAL;
2913         } else {
2914                 aiov.iov_base = buf;
2915                 aiov.iov_len = count;
2916                 auio.uio_iov = &aiov;
2917                 auio.uio_iovcnt = 1;
2918                 auio.uio_offset = 0;
2919                 auio.uio_rw = UIO_READ;
2920                 auio.uio_segflg = UIO_USERSPACE;
2921                 auio.uio_td = td;
2922                 auio.uio_resid = count;
2923                 error = VOP_READLINK(vp, &auio, td->td_ucred);
2924         }
2925         vput(vp);
2926         *res = count - auio.uio_resid;
2927         return (error);
2928 }
2929
2930 /*
2931  * readlink_args(char *path, char *buf, int count)
2932  *
2933  * Return target name of a symbolic link.
2934  */
2935 int
2936 sys_readlink(struct readlink_args *uap)
2937 {
2938         struct nlookupdata nd;
2939         int error;
2940
2941         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2942         if (error == 0) {
2943                 error = kern_readlink(&nd, uap->buf, uap->count,
2944                                         &uap->sysmsg_result);
2945         }
2946         nlookup_done(&nd);
2947         return (error);
2948 }
2949
2950 /*
2951  * readlinkat_args(int fd, char *path, char *buf, size_t bufsize)
2952  *
2953  * Return target name of a symbolic link.  The path is relative to the
2954  * directory associated with fd.
2955  */
2956 int
2957 sys_readlinkat(struct readlinkat_args *uap)
2958 {
2959         struct nlookupdata nd;
2960         struct file *fp;
2961         int error;
2962
2963         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
2964         if (error == 0) {
2965                 error = kern_readlink(&nd, uap->buf, uap->bufsize,
2966                                         &uap->sysmsg_result);
2967         }
2968         nlookup_done_at(&nd, fp);
2969         return (error);
2970 }
2971
2972 static int
2973 setfflags(struct vnode *vp, int flags)
2974 {
2975         struct thread *td = curthread;
2976         int error;
2977         struct vattr vattr;
2978
2979         /*
2980          * Prevent non-root users from setting flags on devices.  When
2981          * a device is reused, users can retain ownership of the device
2982          * if they are allowed to set flags and programs assume that
2983          * chown can't fail when done as root.
2984          */
2985         if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
2986             ((error = priv_check_cred(td->td_ucred, PRIV_VFS_CHFLAGS_DEV, 0)) != 0))
2987                 return (error);
2988
2989         /*
2990          * note: vget is required for any operation that might mod the vnode
2991          * so VINACTIVE is properly cleared.
2992          */
2993         if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2994                 VATTR_NULL(&vattr);
2995                 vattr.va_flags = flags;
2996                 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2997                 vput(vp);
2998         }
2999         return (error);
3000 }
3001
3002 /*
3003  * chflags(char *path, int flags)
3004  *
3005  * Change flags of a file given a path name.
3006  */
3007 int
3008 sys_chflags(struct chflags_args *uap)
3009 {
3010         struct nlookupdata nd;
3011         struct vnode *vp;
3012         int error;
3013
3014         vp = NULL;
3015         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3016         if (error == 0)
3017                 error = nlookup(&nd);
3018         if (error == 0)
3019                 error = ncp_writechk(&nd.nl_nch);
3020         if (error == 0)
3021                 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
3022         nlookup_done(&nd);
3023         if (error == 0) {
3024                 error = setfflags(vp, uap->flags);
3025                 vrele(vp);
3026         }
3027         return (error);
3028 }
3029
3030 /*
3031  * lchflags(char *path, int flags)
3032  *
3033  * Change flags of a file given a path name, but don't follow symlinks.
3034  */
3035 int
3036 sys_lchflags(struct lchflags_args *uap)
3037 {
3038         struct nlookupdata nd;
3039         struct vnode *vp;
3040         int error;
3041
3042         vp = NULL;
3043         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3044         if (error == 0)
3045                 error = nlookup(&nd);
3046         if (error == 0)
3047                 error = ncp_writechk(&nd.nl_nch);
3048         if (error == 0)
3049                 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
3050         nlookup_done(&nd);
3051         if (error == 0) {
3052                 error = setfflags(vp, uap->flags);
3053                 vrele(vp);
3054         }
3055         return (error);
3056 }
3057
3058 /*
3059  * fchflags_args(int fd, int flags)
3060  *
3061  * Change flags of a file given a file descriptor.
3062  */
3063 int
3064 sys_fchflags(struct fchflags_args *uap)
3065 {
3066         struct thread *td = curthread;
3067         struct proc *p = td->td_proc;
3068         struct file *fp;
3069         int error;
3070
3071         if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
3072                 return (error);
3073         if (fp->f_nchandle.ncp)
3074                 error = ncp_writechk(&fp->f_nchandle);
3075         if (error == 0)
3076                 error = setfflags((struct vnode *) fp->f_data, uap->flags);
3077         fdrop(fp);
3078         return (error);
3079 }
3080
3081 /*
3082  * chflagsat_args(int fd, const char *path, int flags, int atflags)
3083  * change flags given a pathname relative to a filedescriptor
3084  */
3085 int sys_chflagsat(struct chflagsat_args *uap)
3086 {
3087         struct nlookupdata nd;
3088         struct vnode *vp;
3089         struct file *fp;
3090         int error;
3091         int lookupflags;
3092
3093         if (uap->atflags & ~AT_SYMLINK_NOFOLLOW)
3094                 return (EINVAL);
3095
3096         lookupflags = (uap->atflags & AT_SYMLINK_NOFOLLOW) ? 0 : NLC_FOLLOW;
3097
3098         vp = NULL;
3099         error = nlookup_init_at(&nd, &fp, uap->fd,  uap->path, UIO_USERSPACE, lookupflags);
3100         if (error == 0)
3101                 error = nlookup(&nd);
3102         if (error == 0)
3103                 error = ncp_writechk(&nd.nl_nch);
3104         if (error == 0)
3105                 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
3106         nlookup_done_at(&nd, fp);
3107         if (error == 0) {
3108                 error = setfflags(vp, uap->flags);
3109                 vrele(vp);
3110         }
3111         return (error);
3112 }
3113
3114
3115 static int
3116 setfmode(struct vnode *vp, int mode)
3117 {
3118         struct thread *td = curthread;
3119         int error;
3120         struct vattr vattr;
3121
3122         /*
3123          * note: vget is required for any operation that might mod the vnode
3124          * so VINACTIVE is properly cleared.
3125          */
3126         if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
3127                 VATTR_NULL(&vattr);
3128                 vattr.va_mode = mode & ALLPERMS;
3129                 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3130                 vput(vp);
3131         }
3132         return error;
3133 }
3134
3135 int
3136 kern_chmod(struct nlookupdata *nd, int mode)
3137 {
3138         struct vnode *vp;
3139         int error;
3140
3141         if ((error = nlookup(nd)) != 0)
3142                 return (error);
3143         if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
3144                 return (error);
3145         if ((error = ncp_writechk(&nd->nl_nch)) == 0)
3146                 error = setfmode(vp, mode);
3147         vrele(vp);
3148         return (error);
3149 }
3150
3151 /*
3152  * chmod_args(char *path, int mode)
3153  *
3154  * Change mode of a file given path name.
3155  */
3156 int
3157 sys_chmod(struct chmod_args *uap)
3158 {
3159         struct nlookupdata nd;
3160         int error;
3161
3162         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3163         if (error == 0)
3164                 error = kern_chmod(&nd, uap->mode);
3165         nlookup_done(&nd);
3166         return (error);
3167 }
3168
3169 /*
3170  * lchmod_args(char *path, int mode)
3171  *
3172  * Change mode of a file given path name (don't follow links.)
3173  */
3174 int
3175 sys_lchmod(struct lchmod_args *uap)
3176 {
3177         struct nlookupdata nd;
3178         int error;
3179
3180         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3181         if (error == 0)
3182                 error = kern_chmod(&nd, uap->mode);
3183         nlookup_done(&nd);
3184         return (error);
3185 }
3186
3187 /*
3188  * fchmod_args(int fd, int mode)
3189  *
3190  * Change mode of a file given a file descriptor.
3191  */
3192 int
3193 sys_fchmod(struct fchmod_args *uap)
3194 {
3195         struct thread *td = curthread;
3196         struct proc *p = td->td_proc;
3197         struct file *fp;
3198         int error;
3199
3200         if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
3201                 return (error);
3202         if (fp->f_nchandle.ncp)
3203                 error = ncp_writechk(&fp->f_nchandle);
3204         if (error == 0)
3205                 error = setfmode((struct vnode *)fp->f_data, uap->mode);
3206         fdrop(fp);
3207         return (error);
3208 }
3209
3210 /*
3211  * fchmodat_args(char *path, int mode)
3212  *
3213  * Change mode of a file pointed to by fd/path.
3214  */
3215 int
3216 sys_fchmodat(struct fchmodat_args *uap)
3217 {
3218         struct nlookupdata nd;
3219         struct file *fp;
3220         int error;
3221         int flags;
3222
3223         if (uap->flags & ~AT_SYMLINK_NOFOLLOW)
3224                 return (EINVAL);
3225         flags = (uap->flags & AT_SYMLINK_NOFOLLOW) ? 0 : NLC_FOLLOW;
3226
3227         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path,
3228                                 UIO_USERSPACE, flags);
3229         if (error == 0)
3230                 error = kern_chmod(&nd, uap->mode);
3231         nlookup_done_at(&nd, fp);
3232         return (error);
3233 }
3234
3235 static int
3236 setfown(struct mount *mp, struct vnode *vp, uid_t uid, gid_t gid)
3237 {
3238         struct thread *td = curthread;
3239         int error;
3240         struct vattr vattr;
3241         uid_t o_uid;
3242         gid_t o_gid;
3243         uint64_t size;
3244
3245         /*
3246          * note: vget is required for any operation that might mod the vnode
3247          * so VINACTIVE is properly cleared.
3248          */
3249         if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
3250                 if ((error = VOP_GETATTR(vp, &vattr)) != 0)
3251                         return error;
3252                 o_uid = vattr.va_uid;
3253                 o_gid = vattr.va_gid;
3254                 size = vattr.va_size;
3255
3256                 VATTR_NULL(&vattr);
3257                 vattr.va_uid = uid;
3258                 vattr.va_gid = gid;
3259                 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3260                 vput(vp);
3261         }
3262
3263         if (error == 0) {
3264                 if (uid == -1)
3265                         uid = o_uid;
3266                 if (gid == -1)
3267                         gid = o_gid;
3268                 VFS_ACCOUNT(mp, o_uid, o_gid, -size);
3269                 VFS_ACCOUNT(mp,   uid,   gid,  size);
3270         }
3271
3272         return error;
3273 }
3274
3275 int
3276 kern_chown(struct nlookupdata *nd, int uid, int gid)
3277 {
3278         struct vnode *vp;
3279         int error;
3280
3281         if ((error = nlookup(nd)) != 0)
3282                 return (error);
3283         if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
3284                 return (error);
3285         if ((error = ncp_writechk(&nd->nl_nch)) == 0)
3286                 error = setfown(nd->nl_nch.mount, vp, uid, gid);
3287         vrele(vp);
3288         return (error);
3289 }
3290
3291 /*
3292  * chown(char *path, int uid, int gid)
3293  *
3294  * Set ownership given a path name.
3295  */
3296 int
3297 sys_chown(struct chown_args *uap)
3298 {
3299         struct nlookupdata nd;
3300         int error;
3301
3302         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3303         if (error == 0)
3304                 error = kern_chown(&nd, uap->uid, uap->gid);
3305         nlookup_done(&nd);
3306         return (error);
3307 }
3308
3309 /*
3310  * lchown_args(char *path, int uid, int gid)
3311  *
3312  * Set ownership given a path name, do not cross symlinks.
3313  */
3314 int
3315 sys_lchown(struct lchown_args *uap)
3316 {
3317         struct nlookupdata nd;
3318         int error;
3319
3320         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3321         if (error == 0)
3322                 error = kern_chown(&nd, uap->uid, uap->gid);
3323         nlookup_done(&nd);
3324         return (error);
3325 }
3326
3327 /*
3328  * fchown_args(int fd, int uid, int gid)
3329  *
3330  * Set ownership given a file descriptor.
3331  */
3332 int
3333 sys_fchown(struct fchown_args *uap)
3334 {
3335         struct thread *td = curthread;
3336         struct proc *p = td->td_proc;
3337         struct file *fp;
3338         int error;
3339
3340         if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
3341                 return (error);
3342         if (fp->f_nchandle.ncp)
3343                 error = ncp_writechk(&fp->f_nchandle);
3344         if (error == 0)
3345                 error = setfown(p->p_fd->fd_ncdir.mount,
3346                         (struct vnode *)fp->f_data, uap->uid, uap->gid);
3347         fdrop(fp);
3348         return (error);
3349 }
3350
3351 /*
3352  * fchownat(int fd, char *path, int uid, int gid, int flags)
3353  *
3354  * Set ownership of file pointed to by fd/path.
3355  */
3356 int
3357 sys_fchownat(struct fchownat_args *uap)
3358 {
3359         struct nlookupdata nd;
3360         struct file *fp;
3361         int error;
3362         int flags;
3363
3364         if (uap->flags & ~AT_SYMLINK_NOFOLLOW)
3365                 return (EINVAL);
3366         flags = (uap->flags & AT_SYMLINK_NOFOLLOW) ? 0 : NLC_FOLLOW;
3367
3368         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path,
3369                                 UIO_USERSPACE, flags);
3370         if (error == 0)
3371                 error = kern_chown(&nd, uap->uid, uap->gid);
3372         nlookup_done_at(&nd, fp);
3373         return (error);
3374 }
3375
3376
3377 static int
3378 getutimes(struct timeval *tvp, struct timespec *tsp)
3379 {
3380         struct timeval tv[2];
3381         int error;
3382
3383         if (tvp == NULL) {
3384                 microtime(&tv[0]);
3385                 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
3386                 tsp[1] = tsp[0];
3387         } else {
3388                 if ((error = itimerfix(tvp)) != 0)
3389                         return (error);
3390                 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3391                 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3392         }
3393         return 0;
3394 }
3395
3396 static int
3397 getutimens(const struct timespec *ts, struct timespec *newts, int *nullflag)
3398 {
3399         struct timespec tsnow;
3400         int error;
3401
3402         *nullflag = 0;
3403         nanotime(&tsnow);
3404         if (ts == NULL) {
3405                 newts[0] = tsnow;
3406                 newts[1] = tsnow;
3407                 *nullflag = 1;
3408                 return (0);
3409         }
3410
3411         newts[0] = ts[0];
3412         newts[1] = ts[1];
3413         if (newts[0].tv_nsec == UTIME_OMIT && newts[1].tv_nsec == UTIME_OMIT)
3414                 return (0);
3415         if (newts[0].tv_nsec == UTIME_NOW && newts[1].tv_nsec == UTIME_NOW)
3416                 *nullflag = 1;
3417
3418         if (newts[0].tv_nsec == UTIME_OMIT)
3419                 newts[0].tv_sec = VNOVAL;
3420         else if (newts[0].tv_nsec == UTIME_NOW)
3421                 newts[0] = tsnow;
3422         else if ((error = itimespecfix(&newts[0])) != 0)
3423                 return (error);
3424
3425         if (newts[1].tv_nsec == UTIME_OMIT)
3426                 newts[1].tv_sec = VNOVAL;
3427         else if (newts[1].tv_nsec == UTIME_NOW)
3428                 newts[1] = tsnow;
3429         else if ((error = itimespecfix(&newts[1])) != 0)
3430                 return (error);
3431
3432         return (0);
3433 }
3434
3435 static int
3436 setutimes(struct vnode *vp, struct vattr *vattr,
3437           const struct timespec *ts, int nullflag)
3438 {
3439         struct thread *td = curthread;
3440         int error;
3441
3442         VATTR_NULL(vattr);
3443         vattr->va_atime = ts[0];
3444         vattr->va_mtime = ts[1];
3445         if (nullflag)
3446                 vattr->va_vaflags |= VA_UTIMES_NULL;
3447         error = VOP_SETATTR(vp, vattr, td->td_ucred);
3448
3449         return error;
3450 }
3451
3452 int
3453 kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
3454 {
3455         struct timespec ts[2];
3456         int error;
3457
3458         if (tptr) {
3459                 if ((error = getutimes(tptr, ts)) != 0)
3460                         return (error);
3461         }
3462         error = kern_utimensat(nd, tptr ? ts : NULL, 0);
3463         return (error);
3464 }
3465
3466 /*
3467  * utimes_args(char *path, struct timeval *tptr)
3468  *
3469  * Set the access and modification times of a file.
3470  */
3471 int
3472 sys_utimes(struct utimes_args *uap)
3473 {
3474         struct timeval tv[2];
3475         struct nlookupdata nd;
3476         int error;
3477
3478         if (uap->tptr) {
3479                 error = copyin(uap->tptr, tv, sizeof(tv));
3480                 if (error)
3481                         return (error);
3482         }
3483         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3484         if (error == 0)
3485                 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
3486         nlookup_done(&nd);
3487         return (error);
3488 }
3489
3490 /*
3491  * lutimes_args(char *path, struct timeval *tptr)
3492  *
3493  * Set the access and modification times of a file.
3494  */
3495 int
3496 sys_lutimes(struct lutimes_args *uap)
3497 {
3498         struct timeval tv[2];
3499         struct nlookupdata nd;
3500         int error;
3501
3502         if (uap->tptr) {
3503                 error = copyin(uap->tptr, tv, sizeof(tv));
3504                 if (error)
3505                         return (error);
3506         }
3507         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3508         if (error == 0)
3509                 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
3510         nlookup_done(&nd);
3511         return (error);
3512 }
3513
3514 /*
3515  * Set utimes on a file descriptor.  The creds used to open the
3516  * file are used to determine whether the operation is allowed
3517  * or not.
3518  */
3519 int
3520 kern_futimens(int fd, struct timespec *ts)
3521 {
3522         struct thread *td = curthread;
3523         struct proc *p = td->td_proc;
3524         struct timespec newts[2];
3525         struct file *fp;
3526         struct vnode *vp;
3527         struct vattr vattr;
3528         int nullflag;
3529         int error;
3530
3531         error = getutimens(ts, newts, &nullflag);
3532         if (error)
3533                 return (error);
3534         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
3535                 return (error);
3536         if (fp->f_nchandle.ncp)
3537                 error = ncp_writechk(&fp->f_nchandle);
3538         if (error == 0) {
3539                 vp = fp->f_data;
3540                 error = vget(vp, LK_EXCLUSIVE);
3541                 if (error == 0) {
3542                         error = VOP_GETATTR(vp, &vattr);
3543                         if (error == 0) {
3544                                 error = naccess_va(&vattr, NLC_OWN | NLC_WRITE,
3545                                                    fp->f_cred);
3546                         }
3547                         if (error == 0) {
3548                                 error = setutimes(vp, &vattr, newts, nullflag);
3549                         }
3550                         vput(vp);
3551                 }
3552         }
3553         fdrop(fp);
3554         return (error);
3555 }
3556
3557 /*
3558  * futimens_args(int fd, struct timespec *ts)
3559  *
3560  * Set the access and modification times of a file.
3561  */
3562 int
3563 sys_futimens(struct futimens_args *uap)
3564 {
3565         struct timespec ts[2];
3566         int error;
3567
3568         if (uap->ts) {
3569                 error = copyin(uap->ts, ts, sizeof(ts));
3570                 if (error)
3571                         return (error);
3572         }
3573         error = kern_futimens(uap->fd, uap->ts ? ts : NULL);
3574         return (error);
3575 }
3576
3577 int
3578 kern_futimes(int fd, struct timeval *tptr)
3579 {
3580         struct timespec ts[2];
3581         int error;
3582
3583         if (tptr) {
3584                 if ((error = getutimes(tptr, ts)) != 0)
3585                         return (error);
3586         }
3587         error = kern_futimens(fd, tptr ? ts : NULL);
3588         return (error);
3589 }
3590
3591 /*
3592  * futimes_args(int fd, struct timeval *tptr)
3593  *
3594  * Set the access and modification times of a file.
3595  */
3596 int
3597 sys_futimes(struct futimes_args *uap)
3598 {
3599         struct timeval tv[2];
3600         int error;
3601
3602         if (uap->tptr) {
3603                 error = copyin(uap->tptr, tv, sizeof(tv));
3604                 if (error)
3605                         return (error);
3606         }
3607         error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
3608         return (error);
3609 }
3610
3611 int
3612 kern_utimensat(struct nlookupdata *nd, const struct timespec *ts, int flags)
3613 {
3614         struct timespec newts[2];
3615         struct vnode *vp;
3616         struct vattr vattr;
3617         int nullflag;
3618         int error;
3619
3620         if (flags & ~AT_SYMLINK_NOFOLLOW)
3621                 return (EINVAL);
3622
3623         error = getutimens(ts, newts, &nullflag);
3624         if (error)
3625                 return (error);
3626
3627         nd->nl_flags |= NLC_OWN | NLC_WRITE;
3628         if ((error = nlookup(nd)) != 0)
3629                 return (error);
3630         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
3631                 return (error);
3632         if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
3633                 return (error);
3634         if ((error = vn_writechk(vp, &nd->nl_nch)) == 0) {
3635                 error = vget(vp, LK_EXCLUSIVE);
3636                 if (error == 0) {
3637                         error = setutimes(vp, &vattr, newts, nullflag);
3638                         vput(vp);
3639                 }
3640         }
3641         vrele(vp);
3642         return (error);
3643 }
3644
3645 /*
3646  * utimensat_args(int fd, const char *path, const struct timespec *ts, int flags);
3647  *
3648  * Set file access and modification times of a file.
3649  */
3650 int
3651 sys_utimensat(struct utimensat_args *uap)
3652 {
3653         struct timespec ts[2];
3654         struct nlookupdata nd;
3655         struct file *fp;
3656         int error;
3657         int flags;
3658
3659         if (uap->ts) {
3660                 error = copyin(uap->ts, ts, sizeof(ts));
3661                 if (error)
3662                         return (error);
3663         }
3664
3665         flags = (uap->flags & AT_SYMLINK_NOFOLLOW) ? 0 : NLC_FOLLOW;
3666         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path,
3667                                 UIO_USERSPACE, flags);
3668         if (error == 0)
3669                 error = kern_utimensat(&nd, uap->ts ? ts : NULL, uap->flags);
3670         nlookup_done_at(&nd, fp);
3671         return (error);
3672 }
3673
3674 int
3675 kern_truncate(struct nlookupdata *nd, off_t length)
3676 {
3677         struct vnode *vp;
3678         struct vattr vattr;
3679         int error;
3680         uid_t uid = 0;
3681         gid_t gid = 0;
3682         uint64_t old_size = 0;
3683
3684         if (length < 0)
3685                 return(EINVAL);
3686         nd->nl_flags |= NLC_WRITE | NLC_TRUNCATE;
3687         if ((error = nlookup(nd)) != 0)
3688                 return (error);
3689         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
3690                 return (error);
3691         if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
3692                 return (error);
3693         error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_FAILRECLAIM);
3694         if (error) {
3695                 vrele(vp);
3696                 return (error);
3697         }
3698         if (vp->v_type == VDIR) {
3699                 error = EISDIR;
3700                 goto done;
3701         }
3702         if (vfs_quota_enabled) {
3703                 error = VOP_GETATTR(vp, &vattr);
3704                 KASSERT(error == 0, ("kern_truncate(): VOP_GETATTR didn't return 0"));
3705                 uid = vattr.va_uid;
3706                 gid = vattr.va_gid;
3707                 old_size = vattr.va_size;
3708         }
3709
3710         if ((error = vn_writechk(vp, &nd->nl_nch)) == 0) {
3711                 VATTR_NULL(&vattr);
3712                 vattr.va_size = length;
3713                 error = VOP_SETATTR(vp, &vattr, nd->nl_cred);
3714                 VFS_ACCOUNT(nd->nl_nch.mount, uid, gid, length - old_size);
3715         }
3716 done:
3717         vput(vp);
3718         return (error);
3719 }
3720
3721 /*
3722  * truncate(char *path, int pad, off_t length)
3723  *
3724  * Truncate a file given its path name.
3725  */
3726 int
3727 sys_truncate(struct truncate_args *uap)
3728 {
3729         struct nlookupdata nd;
3730         int error;
3731
3732         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3733         if (error == 0)
3734                 error = kern_truncate(&nd, uap->length);
3735         nlookup_done(&nd);
3736         return error;
3737 }
3738
3739 int
3740 kern_ftruncate(int fd, off_t length)
3741 {
3742         struct thread *td = curthread;
3743         struct proc *p = td->td_proc;
3744         struct vattr vattr;
3745         struct vnode *vp;
3746         struct file *fp;
3747         int error;
3748         uid_t uid = 0;
3749         gid_t gid = 0;
3750         uint64_t old_size = 0;
3751         struct mount *mp;
3752
3753         if (length < 0)
3754                 return(EINVAL);
3755         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
3756                 return (error);
3757         if (fp->f_nchandle.ncp) {
3758                 error = ncp_writechk(&fp->f_nchandle);
3759                 if (error)
3760                         goto done;
3761         }
3762         if ((fp->f_flag & FWRITE) == 0) {
3763                 error = EINVAL;
3764                 goto done;
3765         }
3766         if (fp->f_flag & FAPPENDONLY) { /* inode was set s/uapnd */
3767                 error = EINVAL;
3768                 goto done;
3769         }
3770         vp = (struct vnode *)fp->f_data;
3771         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3772         if (vp->v_type == VDIR) {
3773                 error = EISDIR;
3774                 vn_unlock(vp);
3775                 goto done;
3776         }
3777
3778         if (vfs_quota_enabled) {
3779                 error = VOP_GETATTR(vp, &vattr);
3780                 KASSERT(error == 0, ("kern_ftruncate(): VOP_GETATTR didn't return 0"));
3781                 uid = vattr.va_uid;
3782                 gid = vattr.va_gid;
3783                 old_size = vattr.va_size;
3784         }
3785
3786         if ((error = vn_writechk(vp, NULL)) == 0) {
3787                 VATTR_NULL(&vattr);
3788                 vattr.va_size = length;
3789                 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3790                 mp = vq_vptomp(vp);
3791                 VFS_ACCOUNT(mp, uid, gid, length - old_size);
3792         }
3793         vn_unlock(vp);
3794 done:
3795         fdrop(fp);
3796         return (error);
3797 }
3798
3799 /*
3800  * ftruncate_args(int fd, int pad, off_t length)
3801  *
3802  * Truncate a file given a file descriptor.
3803  */
3804 int
3805 sys_ftruncate(struct ftruncate_args *uap)
3806 {
3807         int error;
3808
3809         error = kern_ftruncate(uap->fd, uap->length);
3810
3811         return (error);
3812 }
3813
3814 /*
3815  * fsync(int fd)
3816  *
3817  * Sync an open file.
3818  */
3819 int
3820 sys_fsync(struct fsync_args *uap)
3821 {
3822         struct thread *td = curthread;
3823         struct proc *p = td->td_proc;
3824         struct vnode *vp;
3825         struct file *fp;
3826         vm_object_t obj;
3827         int error;
3828
3829         if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
3830                 return (error);
3831         vp = (struct vnode *)fp->f_data;
3832         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3833         if ((obj = vp->v_object) != NULL) {
3834                 if (vp->v_mount == NULL ||
3835                     (vp->v_mount->mnt_kern_flag & MNTK_NOMSYNC) == 0) {
3836                         vm_object_page_clean(obj, 0, 0, 0);
3837                 }
3838         }
3839         error = VOP_FSYNC(vp, MNT_WAIT, VOP_FSYNC_SYSCALL);
3840         if (error == 0 && vp->v_mount)
3841                 error = buf_fsync(vp);
3842         vn_unlock(vp);
3843         fdrop(fp);
3844
3845         return (error);
3846 }
3847
3848 int
3849 kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
3850 {
3851         struct nchandle fnchd;
3852         struct nchandle tnchd;
3853         struct namecache *ncp;
3854         struct vnode *fdvp;
3855         struct vnode *tdvp;
3856         struct mount *mp;
3857         int error;
3858         u_int fncp_gen;
3859         u_int tncp_gen;
3860
3861         bwillinode(1);
3862         fromnd->nl_flags |= NLC_REFDVP | NLC_RENAME_SRC;
3863         if ((error = nlookup(fromnd)) != 0)
3864                 return (error);
3865         if ((fnchd.ncp = fromnd->nl_nch.ncp->nc_parent) == NULL)
3866                 return (ENOENT);
3867         fnchd.mount = fromnd->nl_nch.mount;
3868         cache_hold(&fnchd);
3869
3870         /*
3871          * unlock the source nch so we can lookup the target nch without
3872          * deadlocking.  The target may or may not exist so we do not check
3873          * for a target vp like kern_mkdir() and other creation functions do.
3874          *
3875          * The source and target directories are ref'd and rechecked after
3876          * everything is relocked to determine if the source or target file
3877          * has been renamed.
3878          */
3879         KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
3880         fromnd->nl_flags &= ~NLC_NCPISLOCKED;
3881
3882         fncp_gen = fromnd->nl_nch.ncp->nc_generation;
3883
3884         cache_unlock(&fromnd->nl_nch);
3885
3886         tond->nl_flags |= NLC_RENAME_DST | NLC_REFDVP;
3887         if ((error = nlookup(tond)) != 0) {
3888                 cache_drop(&fnchd);
3889                 return (error);
3890         }
3891         tncp_gen = tond->nl_nch.ncp->nc_generation;
3892
3893         if ((tnchd.ncp = tond->nl_nch.ncp->nc_parent) == NULL) {
3894                 cache_drop(&fnchd);
3895                 return (ENOENT);
3896         }
3897         tnchd.mount = tond->nl_nch.mount;
3898         cache_hold(&tnchd);
3899
3900         /*
3901          * If the source and target are the same there is nothing to do
3902          */
3903         if (fromnd->nl_nch.ncp == tond->nl_nch.ncp) {
3904                 cache_drop(&fnchd);
3905                 cache_drop(&tnchd);
3906                 return (0);
3907         }
3908
3909         /*
3910          * Mount points cannot be renamed or overwritten
3911          */
3912         if ((fromnd->nl_nch.ncp->nc_flag | tond->nl_nch.ncp->nc_flag) &
3913             NCF_ISMOUNTPT
3914         ) {
3915                 cache_drop(&fnchd);
3916                 cache_drop(&tnchd);
3917                 return (EINVAL);
3918         }
3919
3920         /*
3921          * Relock the source ncp.  cache_relock() will deal with any
3922          * deadlocks against the already-locked tond and will also
3923          * make sure both are resolved.
3924          *
3925          * NOTE AFTER RELOCKING: The source or target ncp may have become
3926          * invalid while they were unlocked, nc_vp and nc_mount could
3927          * be NULL.
3928          */
3929         cache_relock(&fromnd->nl_nch, fromnd->nl_cred,
3930                      &tond->nl_nch, tond->nl_cred);
3931         fromnd->nl_flags |= NLC_NCPISLOCKED;
3932
3933         /*
3934          * If the namecache generation changed for either fromnd or tond,
3935          * we must retry.
3936          */
3937         if (fromnd->nl_nch.ncp->nc_generation != fncp_gen ||
3938             tond->nl_nch.ncp->nc_generation != tncp_gen) {
3939                 kprintf("kern_rename: retry due to gen on: "
3940                         "\"%s\" -> \"%s\"\n",
3941                         fromnd->nl_nch.ncp->nc_name,
3942                         tond->nl_nch.ncp->nc_name);
3943                 cache_drop(&fnchd);
3944                 cache_drop(&tnchd);
3945                 return (EAGAIN);
3946         }
3947
3948         /*
3949          * If either fromnd or tond are marked destroyed a ripout occured
3950          * out from under us and we must retry.
3951          */
3952         if ((fromnd->nl_nch.ncp->nc_flag & (NCF_DESTROYED | NCF_UNRESOLVED)) ||
3953             fromnd->nl_nch.ncp->nc_vp == NULL ||
3954             (tond->nl_nch.ncp->nc_flag & NCF_DESTROYED)) {
3955                 kprintf("kern_rename: retry due to ripout on: "
3956                         "\"%s\" -> \"%s\"\n",
3957                         fromnd->nl_nch.ncp->nc_name,
3958                         tond->nl_nch.ncp->nc_name);
3959                 cache_drop(&fnchd);
3960                 cache_drop(&tnchd);
3961                 return (EAGAIN);
3962         }
3963
3964         /*
3965          * Make sure the parent directories linkages are the same.
3966          * XXX shouldn't be needed any more w/ generation check above.
3967          */
3968         if (fnchd.ncp != fromnd->nl_nch.ncp->nc_parent ||
3969             tnchd.ncp != tond->nl_nch.ncp->nc_parent) {
3970                 cache_drop(&fnchd);
3971                 cache_drop(&tnchd);
3972                 return (ENOENT);
3973         }
3974
3975         /*
3976          * Both the source and target must be within the same filesystem and
3977          * in the same filesystem as their parent directories within the
3978          * namecache topology.
3979          *
3980          * NOTE: fromnd's nc_mount or nc_vp could be NULL.
3981          */
3982         mp = fnchd.mount;
3983         if (mp != tnchd.mount || mp != fromnd->nl_nch.mount ||
3984             mp != tond->nl_nch.mount) {
3985                 cache_drop(&fnchd);
3986                 cache_drop(&tnchd);
3987                 return (EXDEV);
3988         }
3989
3990         /*
3991          * Make sure the mount point is writable
3992          */
3993         if ((error = ncp_writechk(&tond->nl_nch)) != 0) {
3994                 cache_drop(&fnchd);
3995                 cache_drop(&tnchd);
3996                 return (error);
3997         }
3998
3999         /*
4000          * If the target exists and either the source or target is a directory,
4001          * then both must be directories.
4002          *
4003          * Due to relocking of the source, fromnd->nl_nch.ncp->nc_vp might h
4004          * have become NULL.
4005          */
4006         if (tond->nl_nch.ncp->nc_vp) {
4007                 if (fromnd->nl_nch.ncp->nc_vp == NULL) {
4008                         error = ENOENT;
4009                 } else if (fromnd->nl_nch.ncp->nc_vp->v_type == VDIR) {
4010                         if (tond->nl_nch.ncp->nc_vp->v_type != VDIR)
4011                                 error = ENOTDIR;
4012                 } else if (tond->nl_nch.ncp->nc_vp->v_type == VDIR) {
4013                         error = EISDIR;
4014                 }
4015         }
4016
4017         /*
4018          * You cannot rename a source into itself or a subdirectory of itself.
4019          * We check this by travsersing the target directory upwards looking
4020          * for a match against the source.
4021          *
4022          * XXX MPSAFE
4023          */
4024         if (error == 0) {
4025                 for (ncp = tnchd.ncp; ncp; ncp = ncp->nc_parent) {
4026                         if (fromnd->nl_nch.ncp == ncp) {
4027                                 error = EINVAL;
4028                                 break;
4029                         }
4030                 }
4031         }
4032
4033         cache_drop(&fnchd);
4034         cache_drop(&tnchd);
4035
4036         /*
4037          * Even though the namespaces are different, they may still represent
4038          * hardlinks to the same file.  The filesystem might have a hard time
4039          * with this so we issue a NREMOVE of the source instead of a NRENAME
4040          * when we detect the situation.
4041          */
4042         if (error == 0) {
4043                 fdvp = fromnd->nl_dvp;
4044                 tdvp = tond->nl_dvp;
4045                 if (fdvp == NULL || tdvp == NULL) {
4046                         error = EPERM;
4047                 } else if (fromnd->nl_nch.ncp->nc_vp == tond->nl_nch.ncp->nc_vp) {
4048                         error = VOP_NREMOVE(&fromnd->nl_nch, fdvp,
4049                                             fromnd->nl_cred);
4050                 } else {
4051                         error = VOP_NRENAME(&fromnd->nl_nch, &tond->nl_nch,
4052                                             fdvp, tdvp, tond->nl_cred);
4053                 }
4054         }
4055         return (error);
4056 }
4057
4058 /*
4059  * rename_args(char *from, char *to)
4060  *
4061  * Rename files.  Source and destination must either both be directories,
4062  * or both not be directories.  If target is a directory, it must be empty.
4063  */
4064 int
4065 sys_rename(struct rename_args *uap)
4066 {
4067         struct nlookupdata fromnd, tond;
4068         int error;
4069
4070         do {
4071                 error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
4072                 if (error == 0) {
4073                         error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
4074                         if (error == 0)
4075                                 error = kern_rename(&fromnd, &tond);
4076                         nlookup_done(&tond);
4077                 }
4078                 nlookup_done(&fromnd);
4079         } while (error == EAGAIN);
4080         return (error);
4081 }
4082
4083 /*
4084  * renameat_args(int oldfd, char *old, int newfd, char *new)
4085  *
4086  * Rename files using paths relative to the directories associated with
4087  * oldfd and newfd.  Source and destination must either both be directories,
4088  * or both not be directories.  If target is a directory, it must be empty.
4089  */
4090 int
4091 sys_renameat(struct renameat_args *uap)
4092 {
4093         struct nlookupdata oldnd, newnd;
4094         struct file *oldfp, *newfp;
4095         int error;
4096
4097         do {
4098                 error = nlookup_init_at(&oldnd, &oldfp,
4099                                         uap->oldfd, uap->old,
4100                                         UIO_USERSPACE, 0);
4101                 if (error == 0) {
4102                         error = nlookup_init_at(&newnd, &newfp,
4103                                                 uap->newfd, uap->new,
4104                                                 UIO_USERSPACE, 0);
4105                         if (error == 0)
4106                                 error = kern_rename(&oldnd, &newnd);
4107                         nlookup_done_at(&newnd, newfp);
4108                 }
4109                 nlookup_done_at(&oldnd, oldfp);
4110         } while (error == EAGAIN);
4111         return (error);
4112 }
4113
4114 int
4115 kern_mkdir(struct nlookupdata *nd, int mode)
4116 {
4117         struct thread *td = curthread;
4118         struct proc *p = td->td_proc;
4119         struct vnode *vp;
4120         struct vattr vattr;
4121         int error;
4122
4123         bwillinode(1);
4124         nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE | NLC_REFDVP;
4125         if ((error = nlookup(nd)) != 0)
4126                 return (error);
4127
4128         if (nd->nl_nch.ncp->nc_vp)
4129                 return (EEXIST);
4130         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
4131                 return (error);
4132         VATTR_NULL(&vattr);
4133         vattr.va_type = VDIR;
4134         vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
4135
4136         vp = NULL;
4137         error = VOP_NMKDIR(&nd->nl_nch, nd->nl_dvp, &vp, td->td_ucred, &vattr);
4138         if (error == 0)
4139                 vput(vp);
4140         return (error);
4141 }
4142
4143 /*
4144  * mkdir_args(char *path, int mode)
4145  *
4146  * Make a directory file.
4147  */
4148 int
4149 sys_mkdir(struct mkdir_args *uap)
4150 {
4151         struct nlookupdata nd;
4152         int error;
4153
4154         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
4155         if (error == 0)
4156                 error = kern_mkdir(&nd, uap->mode);
4157         nlookup_done(&nd);
4158         return (error);
4159 }
4160
4161 /*
4162  * mkdirat_args(int fd, char *path, mode_t mode)
4163  *
4164  * Make a directory file.  The path is relative to the directory associated
4165  * with fd.
4166  */
4167 int
4168 sys_mkdirat(struct mkdirat_args *uap)
4169 {
4170         struct nlookupdata nd;
4171         struct file *fp;
4172         int error;
4173
4174         error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
4175         if (error == 0)
4176                 error = kern_mkdir(&nd, uap->mode);
4177         nlookup_done_at(&nd, fp);
4178         return (error);
4179 }
4180
4181 int
4182 kern_rmdir(struct nlookupdata *nd)
4183 {
4184         int error;
4185
4186         bwillinode(1);
4187         nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
4188         if ((error = nlookup(nd)) != 0)
4189                 return (error);
4190
4191         /*
4192          * Do not allow directories representing mount points to be
4193          * deleted, even if empty.  Check write perms on mount point
4194          * in case the vnode is aliased (aka nullfs).
4195          */
4196         if (nd->nl_nch.ncp->nc_flag & (NCF_ISMOUNTPT))
4197                 return (EBUSY);
4198         if ((error = ncp_writechk(&nd->nl_nch)) != 0)
4199                 return (error);
4200         error = VOP_NRMDIR(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
4201         return (error);
4202 }
4203
4204 /*
4205  * rmdir_args(char *path)
4206  *
4207  * Remove a directory file.
4208  */
4209 int
4210 sys_rmdir(struct rmdir_args *uap)
4211 {
4212         struct nlookupdata nd;
4213         int error;
4214
4215         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
4216         if (error == 0)
4217                 error = kern_rmdir(&nd);
4218         nlookup_done(&nd);
4219         return (error);
4220 }
4221
4222 int
4223 kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res,
4224                    enum uio_seg direction)
4225 {
4226         struct thread *td = curthread;
4227         struct proc *p = td->td_proc;
4228         struct vnode *vp;
4229         struct file *fp;
4230         struct uio auio;
4231         struct iovec aiov;
4232         off_t loff;
4233         int error, eofflag;
4234
4235         if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
4236                 return (error);
4237         if ((fp->f_flag & FREAD) == 0) {
4238                 error = EBADF;
4239                 goto done;
4240         }
4241         vp = (struct vnode *)fp->f_data;
4242         if (vp->v_type != VDIR) {
4243                 error = EINVAL;
4244                 goto done;
4245         }
4246         aiov.iov_base = buf;
4247         aiov.iov_len = count;
4248         auio.uio_iov = &aiov;
4249         auio.uio_iovcnt = 1;
4250         auio.uio_rw = UIO_READ;
4251         auio.uio_segflg = direction;
4252         auio.uio_td = td;
4253         auio.uio_resid = count;
4254         loff = auio.uio_offset = fp->f_offset;
4255         error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
4256         fp->f_offset = auio.uio_offset;
4257         if (error)
4258                 goto done;
4259
4260         /*
4261          * WARNING!  *basep may not be wide enough to accomodate the
4262          * seek offset.   XXX should we hack this to return the upper 32 bits
4263          * for offsets greater then 4G?
4264          */
4265         if (basep) {
4266                 *basep = (long)loff;
4267         }
4268         *res = count - auio.uio_resid;
4269 done:
4270         fdrop(fp);
4271         return (error);
4272 }
4273
4274 /*
4275  * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
4276  *
4277  * Read a block of directory entries in a file system independent format.
4278  */
4279 int
4280 sys_getdirentries(struct getdirentries_args *uap)
4281 {
4282         long base;
4283         int error;
4284
4285         error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base,
4286                                    &uap->sysmsg_result, UIO_USERSPACE);
4287
4288         if (error == 0 && uap->basep)
4289                 error = copyout(&base, uap->basep, sizeof(*uap->basep));
4290         return (error);
4291 }
4292
4293 /*
4294  * getdents_args(int fd, char *buf, size_t count)
4295  */
4296 int
4297 sys_getdents(struct getdents_args *uap)
4298 {
4299         int error;
4300
4301         error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL,
4302                                    &uap->sysmsg_result, UIO_USERSPACE);
4303
4304         return (error);
4305 }
4306
4307 /*
4308  * Set the mode mask for creation of filesystem nodes.
4309  *
4310  * umask(int newmask)
4311  */
4312 int
4313 sys_umask(struct umask_args *uap)
4314 {
4315         struct thread *td = curthread;
4316         struct proc *p = td->td_proc;
4317         struct filedesc *fdp;
4318
4319         fdp = p->p_fd;
4320         uap->sysmsg_result = fdp->fd_cmask;
4321         fdp->fd_cmask = uap->newmask & ALLPERMS;
4322         return (0);
4323 }
4324
4325 /*
4326  * revoke(char *path)
4327  *
4328  * Void all references to file by ripping underlying filesystem
4329  * away from vnode.
4330  */
4331 int
4332 sys_revoke(struct revoke_args *uap)
4333 {
4334         struct nlookupdata nd;
4335         struct vattr vattr;
4336         struct vnode *vp;
4337         struct ucred *cred;
4338         int error;
4339
4340         vp = NULL;
4341         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
4342         if (error == 0)
4343                 error = nlookup(&nd);
4344         if (error == 0)
4345                 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
4346         cred = crhold(nd.nl_cred);
4347         nlookup_done(&nd);
4348         if (error == 0) {
4349                 if (error == 0)
4350                         error = VOP_GETATTR(vp, &vattr);
4351                 if (error == 0 && cred->cr_uid != vattr.va_uid)
4352                         error = priv_check_cred(cred, PRIV_VFS_REVOKE, 0);
4353                 if (error == 0 && (vp->v_type == VCHR || vp->v_type == VBLK)) {
4354                         if (vcount(vp) > 0)
4355                                 error = vrevoke(vp, cred);
4356                 } else if (error == 0) {
4357                         error = vrevoke(vp, cred);
4358                 }
4359                 vrele(vp);
4360         }
4361         if (cred)
4362                 crfree(cred);
4363         return (error);
4364 }
4365
4366 /*
4367  * getfh_args(char *fname, fhandle_t *fhp)
4368  *
4369  * Get (NFS) file handle
4370  *
4371  * NOTE: We use the fsid of the covering mount, even if it is a nullfs
4372  * mount.  This allows nullfs mounts to be explicitly exported.
4373  *
4374  * WARNING: nullfs mounts of HAMMER PFS ROOTs are safe.
4375  *
4376  *          nullfs mounts of subdirectories are not safe.  That is, it will
4377  *          work, but you do not really have protection against access to
4378  *          the related parent directories.
4379  */
4380 int
4381 sys_getfh(struct getfh_args *uap)
4382 {
4383         struct thread *td = curthread;
4384         struct nlookupdata nd;
4385         fhandle_t fh;
4386         struct vnode *vp;
4387         struct mount *mp;
4388         int error;
4389
4390         /*
4391          * Must be super user
4392          */
4393         if ((error = priv_check(td, PRIV_ROOT)) != 0)
4394                 return (error);
4395
4396         vp = NULL;
4397         error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
4398         if (error == 0)
4399                 error = nlookup(&nd);
4400         if (error == 0)
4401                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
4402         mp = nd.nl_nch.mount;
4403         nlookup_done(&nd);
4404         if (error == 0) {
4405                 bzero(&fh, sizeof(fh));
4406                 fh.fh_fsid = mp->mnt_stat.f_fsid;
4407                 error = VFS_VPTOFH(vp, &fh.fh_fid);
4408                 vput(vp);
4409                 if (error == 0)
4410                         error = copyout(&fh, uap->fhp, sizeof(fh));
4411         }
4412         return (error);
4413 }
4414
4415 /*
4416  * fhopen_args(const struct fhandle *u_fhp, int flags)
4417  *
4418  * syscall for the rpc.lockd to use to translate a NFS file handle into
4419  * an open descriptor.
4420  *
4421  * warning: do not remove the priv_check() call or this becomes one giant
4422  * security hole.
4423  */
4424 int
4425 sys_fhopen(struct fhopen_args *uap)
4426 {
4427         struct thread *td = curthread;
4428         struct filedesc *fdp = td->td_proc->p_fd;
4429         struct mount *mp;
4430         struct vnode *vp;
4431         struct fhandle fhp;
4432         struct vattr vat;
4433         struct vattr *vap = &vat;
4434         struct flock lf;
4435         int fmode, mode, error = 0, type;
4436         struct file *nfp;
4437         struct file *fp;
4438         int indx;
4439
4440         /*
4441          * Must be super user
4442          */
4443         error = priv_check(td, PRIV_ROOT);
4444         if (error)
4445                 return (error);
4446
4447         fmode = FFLAGS(uap->flags);
4448
4449         /*
4450          * Why not allow a non-read/write open for our lockd?
4451          */
4452         if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4453                 return (EINVAL);
4454         error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4455         if (error)
4456                 return(error);
4457
4458         /*
4459          * Find the mount point
4460          */
4461         mp = vfs_getvfs(&fhp.fh_fsid);
4462         if (mp == NULL) {
4463                 error = ESTALE;
4464                 goto  done;
4465         }
4466         /* now give me my vnode, it gets returned to me locked */
4467         error = VFS_FHTOVP(mp, NULL, &fhp.fh_fid, &vp);
4468         if (error)
4469                 goto done;
4470         /*
4471          * from now on we have to make sure not
4472          * to forget about the vnode
4473          * any error that causes an abort must vput(vp)
4474          * just set error = err and 'goto bad;'.
4475          */
4476
4477         /*
4478          * from vn_open
4479          */
4480         if (vp->v_type == VLNK) {
4481                 error = EMLINK;
4482                 goto bad;
4483         }
4484         if (vp->v_type == VSOCK) {
4485                 error = EOPNOTSUPP;
4486                 goto bad;
4487         }
4488         mode = 0;
4489         if (fmode & (FWRITE | O_TRUNC)) {
4490                 if (vp->v_type == VDIR) {
4491                         error = EISDIR;
4492                         goto bad;
4493                 }
4494                 error = vn_writechk(vp, NULL);
4495                 if (error)
4496                         goto bad;
4497                 mode |= VWRITE;
4498         }
4499         if (fmode & FREAD)
4500                 mode |= VREAD;
4501         if (mode) {
4502                 error = VOP_ACCESS(vp, mode, td->td_ucred);
4503                 if (error)
4504                         goto bad;
4505         }
4506         if (fmode & O_TRUNC) {
4507                 vn_unlock(vp);                          /* XXX */
4508                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
4509                 VATTR_NULL(vap);
4510                 vap->va_size = 0;
4511                 error = VOP_SETATTR(vp, vap, td->td_ucred);
4512                 if (error)
4513                         goto bad;
4514         }
4515
4516         /*
4517          * VOP_OPEN needs the file pointer so it can potentially override
4518          * it.
4519          *
4520          * WARNING! no f_nchandle will be associated when fhopen()ing a
4521          * directory.  XXX
4522          */
4523         if ((error = falloc(td->td_lwp, &nfp, &indx)) != 0)
4524                 goto bad;
4525         fp = nfp;
4526
4527         error = VOP_OPEN(vp, fmode, td->td_ucred, fp);
4528         if (error) {
4529                 /*
4530                  * setting f_ops this way prevents VOP_CLOSE from being
4531                  * called or fdrop() releasing the vp from v_data.   Since
4532                  * the VOP_OPEN failed we don't want to VOP_CLOSE.
4533                  */
4534                 fp->f_ops = &badfileops;
4535                 fp->f_data = NULL;
4536                 goto bad_drop;
4537         }
4538
4539         /*
4540          * The fp is given its own reference, we still have our ref and lock.
4541          *
4542          * Assert that all regular files must be created with a VM object.
4543          */
4544         if (vp->v_type == VREG && vp->v_object == NULL) {
4545                 kprintf("fhopen: regular file did not have VM object: %p\n", vp);
4546                 goto bad_drop;
4547         }
4548
4549         /*
4550          * The open was successful.  Handle any locking requirements.
4551          */
4552         if (fmode & (O_EXLOCK | O_SHLOCK)) {
4553                 lf.l_whence = SEEK_SET;
4554                 lf.l_start = 0;
4555                 lf.l_len = 0;
4556                 if (fmode & O_EXLOCK)
4557                         lf.l_type = F_WRLCK;
4558                 else
4559                         lf.l_type = F_RDLCK;
4560                 if (fmode & FNONBLOCK)
4561                         type = 0;
4562                 else
4563                         type = F_WAIT;
4564                 vn_unlock(vp);
4565                 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
4566                         /*
4567                          * release our private reference.
4568                          */
4569                         fsetfd(fdp, NULL, indx);
4570                         fdrop(fp);
4571                         vrele(vp);
4572                         goto done;
4573                 }
4574                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4575                 atomic_set_int(&fp->f_flag, FHASLOCK);  /* race ok */
4576         }
4577
4578         /*
4579          * Clean up.  Associate the file pointer with the previously
4580          * reserved descriptor and return it.
4581          */
4582         vput(vp);
4583         if (uap->flags & O_CLOEXEC)
4584                 fdp->fd_files[indx].fileflags |= UF_EXCLOSE;
4585         fsetfd(fdp, fp, indx);
4586         fdrop(fp);
4587         uap->sysmsg_result = indx;
4588         return (error);
4589
4590 bad_drop:
4591         fsetfd(fdp, NULL, indx);
4592         fdrop(fp);
4593 bad:
4594         vput(vp);
4595 done:
4596         return (error);
4597 }
4598
4599 /*
4600  * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
4601  */
4602 int
4603 sys_fhstat(struct fhstat_args *uap)
4604 {
4605         struct thread *td = curthread;
4606         struct stat sb;
4607         fhandle_t fh;
4608         struct mount *mp;
4609         struct vnode *vp;
4610         int error;
4611
4612         /*
4613          * Must be super user
4614          */
4615         error = priv_check(td, PRIV_ROOT);
4616         if (error)
4617                 return (error);
4618
4619         error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4620         if (error)
4621                 return (error);
4622
4623         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
4624                 error = ESTALE;
4625         if (error == 0) {
4626                 if ((error = VFS_FHTOVP(mp, NULL, &fh.fh_fid, &vp)) == 0) {
4627                         error = vn_stat(vp, &sb, td->td_ucred);
4628                         vput(vp);
4629                 }
4630         }
4631         if (error == 0)
4632                 error = copyout(&sb, uap->sb, sizeof(sb));
4633         return (error);
4634 }
4635
4636 /*
4637  * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
4638  */
4639 int
4640 sys_fhstatfs(struct fhstatfs_args *uap)
4641 {
4642         struct thread *td = curthread;
4643         struct proc *p = td->td_proc;
4644         struct statfs *sp;
4645         struct mount *mp;
4646         struct vnode *vp;
4647         struct statfs sb;
4648         char *fullpath, *freepath;
4649         fhandle_t fh;
4650         int error;
4651
4652         /*
4653          * Must be super user
4654          */
4655         if ((error = priv_check(td, PRIV_ROOT)))
4656                 return (error);
4657
4658         if ((error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t))) != 0)
4659                 return (error);
4660
4661         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
4662                 error = ESTALE;
4663                 goto done;
4664         }
4665         if (p != NULL && !chroot_visible_mnt(mp, p)) {
4666                 error = ESTALE;
4667                 goto done;
4668         }
4669
4670         if ((error = VFS_FHTOVP(mp, NULL, &fh.fh_fid, &vp)) != 0)
4671                 goto done;
4672         mp = vp->v_mount;
4673         sp = &mp->mnt_stat;
4674         vput(vp);
4675         if ((error = VFS_STATFS(mp, sp, td->td_ucred)) != 0)
4676                 goto done;
4677
4678         error = mount_path(p, mp, &fullpath, &freepath);
4679         if (error)
4680                 goto done;
4681         bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
4682         strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
4683         kfree(freepath, M_TEMP);
4684
4685         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4686         if (priv_check(td, PRIV_ROOT)) {
4687                 bcopy(sp, &sb, sizeof(sb));
4688                 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
4689                 sp = &sb;
4690         }
4691         error = copyout(sp, uap->buf, sizeof(*sp));
4692 done:
4693         return (error);
4694 }
4695
4696 /*
4697  * fhstatvfs_args(struct fhandle *u_fhp, struct statvfs *buf)
4698  */
4699 int
4700 sys_fhstatvfs(struct fhstatvfs_args *uap)
4701 {
4702         struct thread *td = curthread;
4703         struct proc *p = td->td_proc;
4704         struct statvfs *sp;
4705         struct mount *mp;
4706         struct vnode *vp;
4707         fhandle_t fh;
4708         int error;
4709
4710         /*
4711          * Must be super user
4712          */
4713         if ((error = priv_check(td, PRIV_ROOT)))
4714                 return (error);
4715
4716         if ((error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t))) != 0)
4717                 return (error);
4718
4719         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
4720                 error = ESTALE;
4721                 goto done;
4722         }
4723         if (p != NULL && !chroot_visible_mnt(mp, p)) {
4724                 error = ESTALE;
4725                 goto done;
4726         }
4727
4728         if ((error = VFS_FHTOVP(mp, NULL, &fh.fh_fid, &vp)))
4729                 goto done;
4730         mp = vp->v_mount;
4731         sp = &mp->mnt_vstat;
4732         vput(vp);
4733         if ((error = VFS_STATVFS(mp, sp, td->td_ucred)) != 0)
4734                 goto done;
4735
4736         sp->f_flag = 0;
4737         if (mp->mnt_flag & MNT_RDONLY)
4738                 sp->f_flag |= ST_RDONLY;
4739         if (mp->mnt_flag & MNT_NOSUID)
4740                 sp->f_flag |= ST_NOSUID;
4741         error = copyout(sp, uap->buf, sizeof(*sp));
4742 done:
4743         return (error);
4744 }
4745
4746
4747 /*
4748  * Syscall to push extended attribute configuration information into the
4749  * VFS.  Accepts a path, which it converts to a mountpoint, as well as
4750  * a command (int cmd), and attribute name and misc data.  For now, the
4751  * attribute name is left in userspace for consumption by the VFS_op.
4752  * It will probably be changed to be copied into sysspace by the
4753  * syscall in the future, once issues with various consumers of the
4754  * attribute code have raised their hands.
4755  *
4756  * Currently this is used only by UFS Extended Attributes.
4757  */
4758 int
4759 sys_extattrctl(struct extattrctl_args *uap)
4760 {
4761         struct nlookupdata nd;
4762         struct vnode *vp;
4763         char attrname[EXTATTR_MAXNAMELEN];
4764         int error;
4765         size_t size;
4766
4767         attrname[0] = 0;
4768         vp = NULL;
4769         error = 0;
4770
4771         if (error == 0 && uap->filename) {
4772                 error = nlookup_init(&nd, uap->filename, UIO_USERSPACE,
4773                                      NLC_FOLLOW);
4774                 if (error == 0)
4775                         error = nlookup(&nd);
4776                 if (error == 0)
4777                         error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
4778                 nlookup_done(&nd);
4779         }
4780
4781         if (error == 0 && uap->attrname) {
4782                 error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN,
4783                                   &size);
4784         }
4785
4786         if (error == 0) {
4787                 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
4788                 if (error == 0)
4789                         error = nlookup(&nd);
4790                 if (error == 0)
4791                         error = ncp_writechk(&nd.nl_nch);
4792                 if (error == 0) {
4793                         error = VFS_EXTATTRCTL(nd.nl_nch.mount, uap->cmd, vp,
4794                                                uap->attrnamespace,
4795                                                uap->attrname, nd.nl_cred);
4796                 }
4797                 nlookup_done(&nd);
4798         }
4799
4800         return (error);
4801 }
4802
4803 /*
4804  * Syscall to get a named extended attribute on a file or directory.
4805  */
4806 int
4807 sys_extattr_set_file(struct extattr_set_file_args *uap)
4808 {
4809         char attrname[EXTATTR_MAXNAMELEN];
4810         struct nlookupdata nd;
4811         struct vnode *vp;
4812         struct uio auio;
4813         struct iovec aiov;
4814         int error;
4815
4816         error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
4817         if (error)
4818                 return (error);
4819
4820         vp = NULL;
4821
4822         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
4823         if (error == 0)
4824                 error = nlookup(&nd);
4825         if (error == 0)
4826                 error = ncp_writechk(&nd.nl_nch);
4827         if (error == 0)
4828                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
4829         if (error) {
4830                 nlookup_done(&nd);
4831                 return (error);
4832         }
4833
4834         bzero(&auio, sizeof(auio));
4835         aiov.iov_base = uap->data;
4836         aiov.iov_len = uap->nbytes;
4837         auio.uio_iov = &aiov;
4838         auio.uio_iovcnt = 1;
4839         auio.uio_offset = 0;
4840         auio.uio_resid = uap->nbytes;
4841         auio.uio_rw = UIO_WRITE;
4842         auio.uio_td = curthread;
4843
4844         error = VOP_SETEXTATTR(vp, uap->attrnamespace, attrname,
4845                                &auio, nd.nl_cred);
4846
4847         vput(vp);
4848         nlookup_done(&nd);
4849         return (error);
4850 }
4851
4852 /*
4853  * Syscall to get a named extended attribute on a file or directory.
4854  */
4855 int
4856 sys_extattr_get_file(struct extattr_get_file_args *uap)
4857 {
4858         char attrname[EXTATTR_MAXNAMELEN];
4859         struct nlookupdata nd;
4860         struct uio auio;
4861         struct iovec aiov;
4862         struct vnode *vp;
4863         int error;
4864
4865         error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
4866         if (error)
4867                 return (error);
4868
4869         vp = NULL;
4870
4871         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
4872         if (error == 0)
4873                 error = nlookup(&nd);
4874         if (error == 0)
4875                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_SHARED, &vp);
4876         if (error) {
4877                 nlookup_done(&nd);
4878                 return (error);
4879         }
4880
4881         bzero(&auio, sizeof(auio));
4882         aiov.iov_base = uap->data;
4883         aiov.iov_len = uap->nbytes;
4884         auio.uio_iov = &aiov;
4885         auio.uio_iovcnt = 1;
4886         auio.uio_offset = 0;
4887         auio.uio_resid = uap->nbytes;
4888         auio.uio_rw = UIO_READ;
4889         auio.uio_td = curthread;
4890
4891         error = VOP_GETEXTATTR(vp, uap->attrnamespace, attrname,
4892                                 &auio, nd.nl_cred);
4893         uap->sysmsg_result = uap->nbytes - auio.uio_resid;
4894
4895         vput(vp);
4896         nlookup_done(&nd);
4897         return(error);
4898 }
4899
4900 /*
4901  * Syscall to delete a named extended attribute from a file or directory.
4902  * Accepts attribute name.  The real work happens in VOP_SETEXTATTR().
4903  */
4904 int
4905 sys_extattr_delete_file(struct extattr_delete_file_args *uap)
4906 {
4907         char attrname[EXTATTR_MAXNAMELEN];
4908         struct nlookupdata nd;
4909         struct vnode *vp;
4910         int error;
4911
4912         error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
4913         if (error)
4914                 return(error);
4915
4916         error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
4917         if (error == 0)
4918                 error = nlookup(&nd);
4919         if (error == 0)
4920                 error = ncp_writechk(&nd.nl_nch);
4921         if (error == 0) {
4922                 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
4923                 if (error == 0) {
4924                         error = VOP_SETEXTATTR(vp, uap->attrnamespace,
4925                                                attrname, NULL, nd.nl_cred);
4926                         vput(vp);
4927                 }
4928         }
4929         nlookup_done(&nd);
4930         return(error);
4931 }
4932
4933 /*
4934  * Determine if the mount is visible to the process.
4935  */
4936 static int
4937 chroot_visible_mnt(struct mount *mp, struct proc *p)
4938 {
4939         struct nchandle nch;
4940
4941         /*
4942          * Traverse from the mount point upwards.  If we hit the process
4943          * root then the mount point is visible to the process.
4944          */
4945         nch = mp->mnt_ncmountpt;
4946         while (nch.ncp) {
4947                 if (nch.mount == p->p_fd->fd_nrdir.mount &&
4948                     nch.ncp == p->p_fd->fd_nrdir.ncp) {
4949                         return(1);
4950                 }
4951                 if (nch.ncp == nch.mount->mnt_ncmountpt.ncp) {
4952                         nch = nch.mount->mnt_ncmounton;
4953                 } else {
4954                         nch.ncp = nch.ncp->nc_parent;
4955                 }
4956         }
4957
4958         /*
4959          * If the mount point is not visible to the process, but the
4960          * process root is in a subdirectory of the mount, return
4961          * TRUE anyway.
4962          */
4963         if (p->p_fd->fd_nrdir.mount == mp)
4964                 return(1);
4965
4966         return(0);
4967 }
4968