vfs_syscalls.c

Version: - FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
    1 /*      $OpenBSD: vfs_syscalls.c,v 1.360 2022/08/14 01:58:28 jsg Exp $  */
    2 /*      $NetBSD: vfs_syscalls.c,v 1.71 1996/04/23 10:29:02 mycroft Exp $        */
    3 
    4 /*
    5  * Copyright (c) 1989, 1993
    6  *      The Regents of the University of California.  All rights reserved.
    7  * (c) UNIX System Laboratories, Inc.
    8  * All or some portions of this file are derived from material licensed
    9  * to the University of California by American Telephone and Telegraph
   10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
   11  * the permission of UNIX System Laboratories, Inc.
   12  *
   13  * Redistribution and use in source and binary forms, with or without
   14  * modification, are permitted provided that the following conditions
   15  * are met:
   16  * 1. Redistributions of source code must retain the above copyright
   17  *    notice, this list of conditions and the following disclaimer.
   18  * 2. Redistributions in binary form must reproduce the above copyright
   19  *    notice, this list of conditions and the following disclaimer in the
   20  *    documentation and/or other materials provided with the distribution.
   21  * 3. Neither the name of the University nor the names of its contributors
   22  *    may be used to endorse or promote products derived from this software
   23  *    without specific prior written permission.
   24  *
   25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   35  * SUCH DAMAGE.
   36  *
   37  *      @(#)vfs_syscalls.c      8.28 (Berkeley) 12/10/94
   38  */
   39 
   40 #include <sys/param.h>
   41 #include <sys/systm.h>
   42 #include <sys/namei.h>
   43 #include <sys/filedesc.h>
   44 #include <sys/conf.h>
   45 #include <sys/fcntl.h>
   46 #include <sys/file.h>
   47 #include <sys/stat.h>
   48 #include <sys/lock.h>
   49 #include <sys/vnode.h>
   50 #include <sys/mount.h>
   51 #include <sys/proc.h>
   52 #include <sys/pledge.h>
   53 #include <sys/uio.h>
   54 #include <sys/malloc.h>
   55 #include <sys/pool.h>
   56 #include <sys/ktrace.h>
   57 #include <sys/unistd.h>
   58 #include <sys/specdev.h>
   59 #include <sys/resourcevar.h>
   60 #include <sys/signalvar.h>
   61 
   62 #include <sys/syscallargs.h>
   63 
   64 extern int suid_clear;
   65 
   66 static int change_dir(struct nameidata *, struct proc *);
   67 
   68 void checkdirs(struct vnode *);
   69 
   70 int copyout_statfs(struct statfs *, void *, struct proc *);
   71 
   72 int doopenat(struct proc *, int, const char *, int, mode_t, register_t *);
   73 int domknodat(struct proc *, int, const char *, mode_t, dev_t);
   74 int dolinkat(struct proc *, int, const char *, int, const char *, int);
   75 int dosymlinkat(struct proc *, const char *, int, const char *);
   76 int dounlinkat(struct proc *, int, const char *, int);
   77 int dofaccessat(struct proc *, int, const char *, int, int);
   78 int dofstatat(struct proc *, int, const char *, struct stat *, int);
   79 int doreadlinkat(struct proc *, int, const char *, char *, size_t,
   80     register_t *);
   81 int dochflagsat(struct proc *, int, const char *, u_int, int);
   82 int dovchflags(struct proc *, struct vnode *, u_int);
   83 int dofchmodat(struct proc *, int, const char *, mode_t, int);
   84 int dofchownat(struct proc *, int, const char *, uid_t, gid_t, int);
   85 int dorenameat(struct proc *, int, const char *, int, const char *);
   86 int domkdirat(struct proc *, int, const char *, mode_t);
   87 int doutimensat(struct proc *, int, const char *, struct timespec [2], int);
   88 int dovutimens(struct proc *, struct vnode *, struct timespec [2]);
   89 int dofutimens(struct proc *, int, struct timespec [2]);
   90 int dounmount_leaf(struct mount *, int, struct proc *);
   91 
   92 /*
   93  * Virtual File System System Calls
   94  */
   95 
   96 /*
   97  * Mount a file system.
   98  */
   99 int
  100 sys_mount(struct proc *p, void *v, register_t *retval)
  101 {
  102         struct sys_mount_args /* {
  103                 syscallarg(const char *) type;
  104                 syscallarg(const char *) path;
  105                 syscallarg(int) flags;
  106                 syscallarg(void *) data;
  107         } */ *uap = v;
  108         struct vnode *vp;
  109         struct mount *mp;
  110         int error, mntflag = 0;
  111         char fstypename[MFSNAMELEN];
  112         char fspath[MNAMELEN];
  113         struct nameidata nd;
  114         struct vfsconf *vfsp;
  115         int flags = SCARG(uap, flags);
  116         void *args = NULL;
  117 
  118         if ((error = suser(p)))
  119                 return (error);
  120 
  121         /*
  122          * Mount points must fit in MNAMELEN, not MAXPATHLEN.
  123          */
  124         error = copyinstr(SCARG(uap, path), fspath, MNAMELEN, NULL);
  125         if (error)
  126                 return(error);
  127 
  128         /*
  129          * Get vnode to be covered
  130          */
  131         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, p);
  132         if ((error = namei(&nd)) != 0)
  133                 goto fail;
  134         vp = nd.ni_vp;
  135         if (flags & MNT_UPDATE) {
  136                 if ((vp->v_flag & VROOT) == 0) {
  137                         vput(vp);
  138                         error = EINVAL;
  139                         goto fail;
  140                 }
  141                 mp = vp->v_mount;
  142                 vfsp = mp->mnt_vfc;
  143 
  144                 args = malloc(vfsp->vfc_datasize, M_TEMP, M_WAITOK | M_ZERO);
  145                 error = copyin(SCARG(uap, data), args, vfsp->vfc_datasize);
  146                 if (error) {
  147                         vput(vp);
  148                         goto fail;
  149                 }
  150 
  151                 mntflag = mp->mnt_flag;
  152                 /*
  153                  * We only allow the filesystem to be reloaded if it
  154                  * is currently mounted read-only.
  155                  */
  156                 if ((flags & MNT_RELOAD) &&
  157                     ((mp->mnt_flag & MNT_RDONLY) == 0)) {
  158                         vput(vp);
  159                         error = EOPNOTSUPP;     /* Needs translation */
  160                         goto fail;
  161                 }
  162 
  163                 if ((error = vfs_busy(mp, VB_READ|VB_NOWAIT)) != 0) {
  164                         vput(vp);
  165                         goto fail;
  166                 }
  167                 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_UPDATE);
  168                 goto update;
  169         }
  170         /*
  171          * Do not allow disabling of permission checks unless exec and access to
  172          * device files is disabled too.
  173          */
  174         if ((flags & MNT_NOPERM) &&
  175             (flags & (MNT_NODEV | MNT_NOEXEC)) != (MNT_NODEV | MNT_NOEXEC)) {
  176                 vput(vp);
  177                 error = EPERM;
  178                 goto fail;
  179         }
  180         if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, INFSLP)) != 0) {
  181                 vput(vp);
  182                 goto fail;
  183         }
  184         if (vp->v_type != VDIR) {
  185                 vput(vp);
  186                 goto fail;
  187         }
  188         error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
  189         if (error) {
  190                 vput(vp);
  191                 goto fail;
  192         }
  193         vfsp = vfs_byname(fstypename);
  194         if (vfsp == NULL) {
  195                 vput(vp);
  196                 error = EOPNOTSUPP;
  197                 goto fail;
  198         }
  199 
  200         args = malloc(vfsp->vfc_datasize, M_TEMP, M_WAITOK | M_ZERO);
  201         error = copyin(SCARG(uap, data), args, vfsp->vfc_datasize);
  202         if (error) {
  203                 vput(vp);
  204                 goto fail;
  205         }
  206 
  207         if (vp->v_mountedhere != NULL) {
  208                 vput(vp);
  209                 error = EBUSY;
  210                 goto fail;
  211         }
  212 
  213         /*
  214          * Allocate and initialize the file system.
  215          */
  216         mp = vfs_mount_alloc(vp, vfsp);
  217         mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
  218 
  219 update:
  220         /* Ensure that the parent mountpoint does not get unmounted. */
  221         error = vfs_busy(vp->v_mount, VB_READ|VB_NOWAIT|VB_DUPOK);
  222         if (error) {
  223                 if (mp->mnt_flag & MNT_UPDATE) {
  224                         mp->mnt_flag = mntflag;
  225                         vfs_unbusy(mp);
  226                 } else {
  227                         vfs_unbusy(mp);
  228                         vfs_mount_free(mp);
  229                 }
  230                 vput(vp);
  231                 goto fail;
  232         }
  233 
  234         /*
  235          * Set the mount level flags.
  236          */
  237         if (flags & MNT_RDONLY)
  238                 mp->mnt_flag |= MNT_RDONLY;
  239         else if (mp->mnt_flag & MNT_RDONLY)
  240                 mp->mnt_flag |= MNT_WANTRDWR;
  241         mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_WXALLOWED | MNT_NODEV |
  242             MNT_SYNCHRONOUS | MNT_ASYNC | MNT_SOFTDEP | MNT_NOATIME |
  243             MNT_NOPERM | MNT_FORCE);
  244         mp->mnt_flag |= flags & (MNT_NOSUID | MNT_NOEXEC | MNT_WXALLOWED |
  245             MNT_NODEV | MNT_SYNCHRONOUS | MNT_ASYNC | MNT_SOFTDEP |
  246             MNT_NOATIME | MNT_NOPERM | MNT_FORCE);
  247         /*
  248          * Mount the filesystem.
  249          */
  250         error = VFS_MOUNT(mp, fspath, args, &nd, p);
  251         if (!error) {
  252                 mp->mnt_stat.f_ctime = gettime();
  253         }
  254         if (mp->mnt_flag & MNT_UPDATE) {
  255                 vfs_unbusy(vp->v_mount);
  256                 vput(vp);
  257                 if (mp->mnt_flag & MNT_WANTRDWR)
  258                         mp->mnt_flag &= ~MNT_RDONLY;
  259                 mp->mnt_flag &= ~MNT_OP_FLAGS;
  260                 if (error)
  261                         mp->mnt_flag = mntflag;
  262 
  263                 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
  264                         if (mp->mnt_syncer == NULL)
  265                                 error = vfs_allocate_syncvnode(mp);
  266                 } else {
  267                         if (mp->mnt_syncer != NULL)
  268                                 vgone(mp->mnt_syncer);
  269                         mp->mnt_syncer = NULL;
  270                 }
  271 
  272                 vfs_unbusy(mp);
  273                 goto fail;
  274         }
  275 
  276         mp->mnt_flag &= ~MNT_OP_FLAGS;
  277         vp->v_mountedhere = mp;
  278 
  279         /*
  280          * Put the new filesystem on the mount list after root.
  281          */
  282         cache_purge(vp);
  283         if (!error) {
  284                 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
  285                 checkdirs(vp);
  286                 vfs_unbusy(vp->v_mount);
  287                 VOP_UNLOCK(vp);
  288                 if ((mp->mnt_flag & MNT_RDONLY) == 0)
  289                         error = vfs_allocate_syncvnode(mp);
  290                 vfs_unbusy(mp);
  291                 (void) VFS_STATFS(mp, &mp->mnt_stat, p);
  292                 if ((error = VFS_START(mp, 0, p)) != 0)
  293                         vrele(vp);
  294         } else {
  295                 mp->mnt_vnodecovered->v_mountedhere = NULL;
  296                 vfs_unbusy(mp);
  297                 vfs_mount_free(mp);
  298                 vfs_unbusy(vp->v_mount);
  299                 vput(vp);
  300         }
  301 fail:
  302         if (args)
  303                 free(args, M_TEMP, vfsp->vfc_datasize);
  304         return (error);
  305 }
  306 
  307 /*
  308  * Scan all active processes to see if any of them have a current
  309  * or root directory onto which the new filesystem has just been
  310  * mounted. If so, replace them with the new mount point, keeping
  311  * track of how many were replaced.  That's the number of references
  312  * the old vnode had that we've replaced, so finish by vrele()'ing
  313  * it that many times.  This puts off any possible sleeping until
  314  * we've finished walking the allprocess list.
  315  */
  316 void
  317 checkdirs(struct vnode *olddp)
  318 {
  319         struct filedesc *fdp;
  320         struct vnode *newdp;
  321         struct process *pr;
  322         u_int  free_count = 0;
  323 
  324         if (olddp->v_usecount == 1)
  325                 return;
  326         if (VFS_ROOT(olddp->v_mountedhere, &newdp))
  327                 panic("mount: lost mount");
  328         LIST_FOREACH(pr, &allprocess, ps_list) {
  329                 fdp = pr->ps_fd;
  330                 if (fdp->fd_cdir == olddp) {
  331                         free_count++;
  332                         vref(newdp);
  333                         fdp->fd_cdir = newdp;
  334                 }
  335                 if (fdp->fd_rdir == olddp) {
  336                         free_count++;
  337                         vref(newdp);
  338                         fdp->fd_rdir = newdp;
  339                 }
  340         }
  341         if (rootvnode == olddp) {
  342                 free_count++;
  343                 vref(newdp);
  344                 rootvnode = newdp;
  345         }
  346         while (free_count-- > 0)
  347                 vrele(olddp);
  348         vput(newdp);
  349 }
  350 
  351 /*
  352  * Unmount a file system.
  353  *
  354  * Note: unmount takes a path to the vnode mounted on as argument,
  355  * not special file (as before).
  356  */
  357 int
  358 sys_unmount(struct proc *p, void *v, register_t *retval)
  359 {
  360         struct sys_unmount_args /* {
  361                 syscallarg(const char *) path;
  362                 syscallarg(int) flags;
  363         } */ *uap = v;
  364         struct vnode *vp;
  365         struct mount *mp;
  366         int error;
  367         struct nameidata nd;
  368 
  369         if ((error = suser(p)) != 0)
  370                 return (error);
  371 
  372         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
  373             SCARG(uap, path), p);
  374         if ((error = namei(&nd)) != 0)
  375                 return (error);
  376         vp = nd.ni_vp;
  377         mp = vp->v_mount;
  378 
  379         /*
  380          * Don't allow unmounting the root file system.
  381          */
  382         if (mp->mnt_flag & MNT_ROOTFS) {
  383                 vput(vp);
  384                 return (EINVAL);
  385         }
  386 
  387         /*
  388          * Must be the root of the filesystem
  389          */
  390         if ((vp->v_flag & VROOT) == 0) {
  391                 vput(vp);
  392                 return (EINVAL);
  393         }
  394         vput(vp);
  395 
  396         if (vfs_busy(mp, VB_WRITE|VB_WAIT))
  397                 return (EBUSY);
  398 
  399         return (dounmount(mp, SCARG(uap, flags) & MNT_FORCE, p));
  400 }
  401 
  402 /*
  403  * Do the actual file system unmount.
  404  */
  405 int
  406 dounmount(struct mount *mp, int flags, struct proc *p)
  407 {
  408         SLIST_HEAD(, mount) mplist;
  409         struct mount *nmp;
  410         int error;
  411 
  412         SLIST_INIT(&mplist);
  413         SLIST_INSERT_HEAD(&mplist, mp, mnt_dounmount);
  414 
  415         /*
  416          * Collect nested mount points. This takes advantage of the mount list
  417          * being ordered - nested mount points come after their parent.
  418          */
  419         while ((mp = TAILQ_NEXT(mp, mnt_list)) != NULL) {
  420                 SLIST_FOREACH(nmp, &mplist, mnt_dounmount) {
  421                         if (mp->mnt_vnodecovered == NULLVP ||
  422                             mp->mnt_vnodecovered->v_mount != nmp)
  423                                 continue;
  424 
  425                         if ((flags & MNT_FORCE) == 0) {
  426                                 error = EBUSY;
  427                                 goto err;
  428                         }
  429                         error = vfs_busy(mp, VB_WRITE|VB_WAIT|VB_DUPOK);
  430                         if (error) {
  431                                 if ((flags & MNT_DOOMED)) {
  432                                         /*
  433                                          * If the mount point was busy due to
  434                                          * being unmounted, it has been removed
  435                                          * from the mount list already.
  436                                          * Restart the iteration from the last
  437                                          * collected busy entry.
  438                                          */
  439                                         mp = SLIST_FIRST(&mplist);
  440                                         break;
  441                                 }
  442                                 goto err;
  443                         }
  444                         SLIST_INSERT_HEAD(&mplist, mp, mnt_dounmount);
  445                         break;
  446                 }
  447         }
  448 
  449         /*
  450          * Nested mount points cannot appear during this loop as mounting
  451          * requires a read lock for the parent mount point.
  452          */
  453         while ((mp = SLIST_FIRST(&mplist)) != NULL) {
  454                 SLIST_REMOVE(&mplist, mp, mount, mnt_dounmount);
  455                 error = dounmount_leaf(mp, flags, p);
  456                 if (error)
  457                         goto err;
  458         }
  459         return (0);
  460 
  461 err:
  462         while ((mp = SLIST_FIRST(&mplist)) != NULL) {
  463                 SLIST_REMOVE(&mplist, mp, mount, mnt_dounmount);
  464                 vfs_unbusy(mp);
  465         }
  466         return (error);
  467 }
  468 
  469 int
  470 dounmount_leaf(struct mount *mp, int flags, struct proc *p)
  471 {
  472         struct vnode *coveredvp;
  473         struct vnode *vp, *nvp;
  474         int error;
  475         int hadsyncer = 0;
  476 
  477         mp->mnt_flag &=~ MNT_ASYNC;
  478         cache_purgevfs(mp);     /* remove cache entries for this file sys */
  479         if (mp->mnt_syncer != NULL) {
  480                 hadsyncer = 1;
  481                 vgone(mp->mnt_syncer);
  482                 mp->mnt_syncer = NULL;
  483         }
  484 
  485         /*
  486          * Before calling file system unmount, make sure
  487          * all unveils to vnodes in here are dropped.
  488          */
  489         TAILQ_FOREACH_SAFE(vp , &mp->mnt_vnodelist, v_mntvnodes, nvp) {
  490                 unveil_removevnode(vp);
  491         }
  492 
  493         if (((mp->mnt_flag & MNT_RDONLY) ||
  494             (error = VFS_SYNC(mp, MNT_WAIT, 0, p->p_ucred, p)) == 0) ||
  495             (flags & MNT_FORCE))
  496                 error = VFS_UNMOUNT(mp, flags, p);
  497 
  498         if (error && !(flags & MNT_DOOMED)) {
  499                 if ((mp->mnt_flag & MNT_RDONLY) == 0 && hadsyncer)
  500                         (void) vfs_allocate_syncvnode(mp);
  501                 vfs_unbusy(mp);
  502                 return (error);
  503         }
  504 
  505         TAILQ_REMOVE(&mountlist, mp, mnt_list);
  506         if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
  507                 coveredvp->v_mountedhere = NULL;
  508                 vrele(coveredvp);
  509         }
  510 
  511         if (!TAILQ_EMPTY(&mp->mnt_vnodelist))
  512                 panic("unmount: dangling vnode");
  513 
  514         vfs_unbusy(mp);
  515         vfs_mount_free(mp);
  516 
  517         return (0);
  518 }
  519 
  520 /*
  521  * Sync each mounted filesystem.
  522  */
  523 int
  524 sys_sync(struct proc *p, void *v, register_t *retval)
  525 {
  526         struct mount *mp;
  527         int asyncflag;
  528 
  529         TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
  530                 if (vfs_busy(mp, VB_READ|VB_NOWAIT))
  531                         continue;
  532                 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
  533                         asyncflag = mp->mnt_flag & MNT_ASYNC;
  534                         mp->mnt_flag &= ~MNT_ASYNC;
  535                         uvm_vnp_sync(mp);
  536                         VFS_SYNC(mp, MNT_NOWAIT, 0, p->p_ucred, p);
  537                         if (asyncflag)
  538                                 mp->mnt_flag |= MNT_ASYNC;
  539                 }
  540                 vfs_unbusy(mp);
  541         }
  542 
  543         return (0);
  544 }
  545 
  546 /*
  547  * Change filesystem quotas.
  548  */
  549 int
  550 sys_quotactl(struct proc *p, void *v, register_t *retval)
  551 {
  552         struct sys_quotactl_args /* {
  553                 syscallarg(const char *) path;
  554                 syscallarg(int) cmd;
  555                 syscallarg(int) uid;
  556                 syscallarg(char *) arg;
  557         } */ *uap = v;
  558         struct mount *mp;
  559         int error;
  560         struct nameidata nd;
  561 
  562         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
  563         if ((error = namei(&nd)) != 0)
  564                 return (error);
  565         mp = nd.ni_vp->v_mount;
  566         vrele(nd.ni_vp);
  567         return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
  568             SCARG(uap, arg), p));
  569 }
  570 
  571 int
  572 copyout_statfs(struct statfs *sp, void *uaddr, struct proc *p)
  573 {
  574         size_t co_sz1 = offsetof(struct statfs, f_fsid);
  575         size_t co_off2 = co_sz1 + sizeof(fsid_t);
  576         size_t co_sz2 = sizeof(struct statfs) - co_off2;
  577         char *s, *d;
  578         int error;
  579 
  580         /* Don't let non-root see filesystem id (for NFS security) */
  581         if (suser(p)) {
  582                 fsid_t fsid;
  583 
  584                 s = (char *)sp;
  585                 d = (char *)uaddr;
  586 
  587                 memset(&fsid, 0, sizeof(fsid));
  588 
  589                 if ((error = copyout(s, d, co_sz1)) != 0)
  590                         return (error);
  591                 if ((error = copyout(&fsid, d + co_sz1, sizeof(fsid))) != 0)
  592                         return (error);
  593                 return (copyout(s + co_off2, d + co_off2, co_sz2));
  594         }
  595 
  596         return (copyout(sp, uaddr, sizeof(*sp)));
  597 }
  598 
  599 /*
  600  * Get filesystem statistics.
  601  */
  602 int
  603 sys_statfs(struct proc *p, void *v, register_t *retval)
  604 {
  605         struct sys_statfs_args /* {
  606                 syscallarg(const char *) path;
  607                 syscallarg(struct statfs *) buf;
  608         } */ *uap = v;
  609         struct mount *mp;
  610         struct statfs *sp;
  611         int error;
  612         struct nameidata nd;
  613 
  614         NDINIT(&nd, LOOKUP, FOLLOW | BYPASSUNVEIL, UIO_USERSPACE,
  615             SCARG(uap, path), p);
  616         nd.ni_pledge = PLEDGE_RPATH;
  617         nd.ni_unveil = UNVEIL_READ;
  618         if ((error = namei(&nd)) != 0)
  619                 return (error);
  620         mp = nd.ni_vp->v_mount;
  621         sp = &mp->mnt_stat;
  622         vrele(nd.ni_vp);
  623         if ((error = VFS_STATFS(mp, sp, p)) != 0)
  624                 return (error);
  625         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
  626 
  627         return (copyout_statfs(sp, SCARG(uap, buf), p));
  628 }
  629 
  630 /*
  631  * Get filesystem statistics.
  632  */
  633 int
  634 sys_fstatfs(struct proc *p, void *v, register_t *retval)
  635 {
  636         struct sys_fstatfs_args /* {
  637                 syscallarg(int) fd;
  638                 syscallarg(struct statfs *) buf;
  639         } */ *uap = v;
  640         struct file *fp;
  641         struct mount *mp;
  642         struct statfs *sp;
  643         int error;
  644 
  645         if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
  646                 return (error);
  647         mp = ((struct vnode *)fp->f_data)->v_mount;
  648         if (!mp) {
  649                 FRELE(fp, p);
  650                 return (ENOENT);
  651         }
  652         sp = &mp->mnt_stat;
  653         error = VFS_STATFS(mp, sp, p);
  654         FRELE(fp, p);
  655         if (error)
  656                 return (error);
  657         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
  658 
  659         return (copyout_statfs(sp, SCARG(uap, buf), p));
  660 }
  661 
  662 /*
  663  * Get statistics on all filesystems.
  664  */
  665 int
  666 sys_getfsstat(struct proc *p, void *v, register_t *retval)
  667 {
  668         struct sys_getfsstat_args /* {
  669                 syscallarg(struct statfs *) buf;
  670                 syscallarg(size_t) bufsize;
  671                 syscallarg(int) flags;
  672         } */ *uap = v;
  673         struct mount *mp;
  674         struct statfs *sp;
  675         struct statfs *sfsp;
  676         size_t count, maxcount;
  677         int error, flags = SCARG(uap, flags);
  678 
  679         maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
  680         sfsp = SCARG(uap, buf);
  681         count = 0;
  682 
  683         TAILQ_FOREACH(mp, &mountlist, mnt_list) {
  684                 if (vfs_busy(mp, VB_READ|VB_NOWAIT))
  685                         continue;
  686                 if (sfsp && count < maxcount) {
  687                         sp = &mp->mnt_stat;
  688 
  689                         /* Refresh stats unless MNT_NOWAIT is specified */
  690                         if (flags != MNT_NOWAIT &&
  691                             flags != MNT_LAZY &&
  692                             (flags == MNT_WAIT ||
  693                             flags == 0) &&
  694                             (error = VFS_STATFS(mp, sp, p))) {
  695                                 vfs_unbusy(mp);
  696                                 continue;
  697                         }
  698 
  699                         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
  700 #if notyet
  701                         if (mp->mnt_flag & MNT_SOFTDEP)
  702                                 sp->f_eflags = STATFS_SOFTUPD;
  703 #endif
  704                         error = (copyout_statfs(sp, sfsp, p));
  705                         if (error) {
  706                                 vfs_unbusy(mp);
  707                                 return (error);
  708                         }
  709                         sfsp++;
  710                 }
  711                 count++;
  712                 vfs_unbusy(mp);
  713         }
  714 
  715         if (sfsp && count > maxcount)
  716                 *retval = maxcount;
  717         else
  718                 *retval = count;
  719 
  720         return (0);
  721 }
  722 
  723 /*
  724  * Change current working directory to a given file descriptor.
  725  */
  726 int
  727 sys_fchdir(struct proc *p, void *v, register_t *retval)
  728 {
  729         struct sys_fchdir_args /* {
  730                 syscallarg(int) fd;
  731         } */ *uap = v;
  732         struct filedesc *fdp = p->p_fd;
  733         struct vnode *vp, *tdp, *old_cdir;
  734         struct mount *mp;
  735         struct file *fp;
  736         int error;
  737 
  738         if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
  739                 return (EBADF);
  740         vp = fp->f_data;
  741         if (fp->f_type != DTYPE_VNODE || vp->v_type != VDIR) {
  742                 FRELE(fp, p);
  743                 return (ENOTDIR);
  744         }
  745         vref(vp);
  746         FRELE(fp, p);
  747         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  748         error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
  749 
  750         while (!error && (mp = vp->v_mountedhere) != NULL) {
  751                 if (vfs_busy(mp, VB_READ|VB_WAIT))
  752                         continue;
  753                 error = VFS_ROOT(mp, &tdp);
  754                 vfs_unbusy(mp);
  755                 if (error)
  756                         break;
  757                 vput(vp);
  758                 vp = tdp;
  759         }
  760         if (error) {
  761                 vput(vp);
  762                 return (error);
  763         }
  764         VOP_UNLOCK(vp);
  765         old_cdir = fdp->fd_cdir;
  766         fdp->fd_cdir = vp;
  767         vrele(old_cdir);
  768         return (0);
  769 }
  770 
  771 /*
  772  * Change current working directory (``.'').
  773  */
  774 int
  775 sys_chdir(struct proc *p, void *v, register_t *retval)
  776 {
  777         struct sys_chdir_args /* {
  778                 syscallarg(const char *) path;
  779         } */ *uap = v;
  780         struct filedesc *fdp = p->p_fd;
  781         struct vnode *old_cdir;
  782         int error;
  783         struct nameidata nd;
  784 
  785         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
  786             SCARG(uap, path), p);
  787         nd.ni_pledge = PLEDGE_RPATH;
  788         nd.ni_unveil = UNVEIL_READ;
  789         if ((error = change_dir(&nd, p)) != 0)
  790                 return (error);
  791         old_cdir = fdp->fd_cdir;
  792         fdp->fd_cdir = nd.ni_vp;
  793         vrele(old_cdir);
  794         return (0);
  795 }
  796 
  797 /*
  798  * Change notion of root (``/'') directory.
  799  */
  800 int
  801 sys_chroot(struct proc *p, void *v, register_t *retval)
  802 {
  803         struct sys_chroot_args /* {
  804                 syscallarg(const char *) path;
  805         } */ *uap = v;
  806         struct filedesc *fdp = p->p_fd;
  807         struct vnode *old_cdir, *old_rdir;
  808         int error;
  809         struct nameidata nd;
  810 
  811         if ((error = suser(p)) != 0)
  812                 return (error);
  813         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
  814             SCARG(uap, path), p);
  815         if ((error = change_dir(&nd, p)) != 0)
  816                 return (error);
  817         if (fdp->fd_rdir != NULL) {
  818                 /*
  819                  * A chroot() done inside a changed root environment does
  820                  * an automatic chdir to avoid the out-of-tree experience.
  821                  */
  822                 vref(nd.ni_vp);
  823                 old_rdir = fdp->fd_rdir;
  824                 old_cdir = fdp->fd_cdir;
  825                 fdp->fd_rdir = fdp->fd_cdir = nd.ni_vp;
  826                 vrele(old_rdir);
  827                 vrele(old_cdir);
  828         } else
  829                 fdp->fd_rdir = nd.ni_vp;
  830         atomic_setbits_int(&p->p_p->ps_flags, PS_CHROOT);
  831         return (0);
  832 }
  833 
  834 /*
  835  * Common routine for chroot and chdir.
  836  */
  837 static int
  838 change_dir(struct nameidata *ndp, struct proc *p)
  839 {
  840         struct vnode *vp;
  841         int error;
  842 
  843         if ((error = namei(ndp)) != 0)
  844                 return (error);
  845         vp = ndp->ni_vp;
  846         if (vp->v_type != VDIR)
  847                 error = ENOTDIR;
  848         else
  849                 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
  850         if (error)
  851                 vput(vp);
  852         else
  853                 VOP_UNLOCK(vp);
  854         return (error);
  855 }
  856 
  857 int
  858 sys___realpath(struct proc *p, void *v, register_t *retval)
  859 {
  860         struct sys___realpath_args /* {
  861                 syscallarg(const char *) pathname;
  862                 syscallarg(char *) resolved;
  863         } */ *uap = v;
  864         char *pathname;
  865         char *rpbuf;
  866         struct nameidata nd;
  867         size_t pathlen;
  868         int error = 0;
  869 
  870         if (SCARG(uap, pathname) == NULL)
  871                 return (EINVAL);
  872 
  873         pathname = pool_get(&namei_pool, PR_WAITOK);
  874         rpbuf = pool_get(&namei_pool, PR_WAITOK);
  875 
  876         if ((error = copyinstr(SCARG(uap, pathname), pathname, MAXPATHLEN,
  877             &pathlen)))
  878                 goto end;
  879 
  880         if (pathlen == 1) { /* empty string "" */
  881                 error = ENOENT;
  882                 goto end;
  883         }
  884         if (pathlen < 2) {
  885                 error = EINVAL;
  886                 goto end;
  887         }
  888 
  889         /* Get cwd for relative path if needed, prepend to rpbuf */
  890         rpbuf[0] = '\0';
  891         if (pathname[0] != '/') {
  892                 int cwdlen = MAXPATHLEN * 4; /* for vfs_getcwd_common */
  893                 char *cwdbuf, *bp;
  894 
  895                 cwdbuf = malloc(cwdlen, M_TEMP, M_WAITOK);
  896 
  897                 /* vfs_getcwd_common fills this in backwards */
  898                 bp = &cwdbuf[cwdlen - 1];
  899                 *bp = '\0';
  900 
  901                 error = vfs_getcwd_common(p->p_fd->fd_cdir, NULL, &bp, cwdbuf,
  902                     cwdlen/2, GETCWD_CHECK_ACCESS, p);
  903 
  904                 if (error) {
  905                         free(cwdbuf, M_TEMP, cwdlen);
  906                         goto end;
  907                 }
  908 
  909                 if (strlcpy(rpbuf, bp, MAXPATHLEN) >= MAXPATHLEN) {
  910                         free(cwdbuf, M_TEMP, cwdlen);
  911                         error = ENAMETOOLONG;
  912                         goto end;
  913                 }
  914 
  915                 free(cwdbuf, M_TEMP, cwdlen);
  916         }
  917 
  918         NDINIT(&nd, LOOKUP, FOLLOW | SAVENAME | REALPATH, UIO_SYSSPACE,
  919             pathname, p);
  920 
  921         nd.ni_cnd.cn_rpbuf = rpbuf;
  922         nd.ni_cnd.cn_rpi = strlen(rpbuf);
  923 
  924         nd.ni_pledge = PLEDGE_RPATH;
  925         nd.ni_unveil = UNVEIL_READ;
  926         if ((error = namei(&nd)) != 0)
  927                 goto end;
  928 
  929         /* release reference from namei */
  930         if (nd.ni_vp)
  931                 vrele(nd.ni_vp);
  932 
  933         error = copyoutstr(nd.ni_cnd.cn_rpbuf, SCARG(uap, resolved),
  934             MAXPATHLEN, NULL);
  935 
  936 #ifdef KTRACE
  937         if (KTRPOINT(p, KTR_NAMEI))
  938                 ktrnamei(p, nd.ni_cnd.cn_rpbuf);
  939 #endif
  940         pool_put(&namei_pool, nd.ni_cnd.cn_pnbuf);
  941 end:
  942         pool_put(&namei_pool, rpbuf);
  943         pool_put(&namei_pool, pathname);
  944         return (error);
  945 }
  946 
  947 int
  948 sys_unveil(struct proc *p, void *v, register_t *retval)
  949 {
  950         struct sys_unveil_args /* {
  951                 syscallarg(const char *) path;
  952                 syscallarg(const char *) permissions;
  953         } */ *uap = v;
  954         struct process *pr = p->p_p;
  955         char *pathname, *c;
  956         struct nameidata nd;
  957         size_t pathlen;
  958         char permissions[5];
  959         int error, allow;
  960 
  961         if (SCARG(uap, path) == NULL && SCARG(uap, permissions) == NULL) {
  962                 pr->ps_uvdone = 1;
  963                 return (0);
  964         }
  965 
  966         if (pr->ps_uvdone != 0)
  967                 return EPERM;
  968 
  969         error = copyinstr(SCARG(uap, permissions), permissions,
  970             sizeof(permissions), NULL);
  971         if (error)
  972                 return (error);
  973         pathname = pool_get(&namei_pool, PR_WAITOK);
  974         error = copyinstr(SCARG(uap, path), pathname, MAXPATHLEN, &pathlen);
  975         if (error)
  976                 goto end;
  977 
  978 #ifdef KTRACE
  979         if (KTRPOINT(p, KTR_STRUCT))
  980                 ktrstruct(p, "unveil", permissions, strlen(permissions));
  981 #endif
  982         if (pathlen < 2) {
  983                 error = EINVAL;
  984                 goto end;
  985         }
  986 
  987         /* find root "/" or "//" */
  988         for (c = pathname; *c != '\0'; c++) {
  989                 if (*c != '/')
  990                         break;
  991         }
  992         if (*c == '\0')
  993                 /* root directory */
  994                 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | SAVENAME,
  995                     UIO_SYSSPACE, pathname, p);
  996         else
  997                 NDINIT(&nd, CREATE, FOLLOW | LOCKLEAF | LOCKPARENT | SAVENAME,
  998                     UIO_SYSSPACE, pathname, p);
  999 
 1000         nd.ni_pledge = PLEDGE_UNVEIL;
 1001         if ((error = namei(&nd)) != 0)
 1002                 goto end;
 1003 
 1004         /*
 1005          * XXX Any access to the file or directory will allow us to
 1006          * pledge path it
 1007          */
 1008         allow = ((nd.ni_vp &&
 1009             (VOP_ACCESS(nd.ni_vp, VREAD, p->p_ucred, p) == 0 ||
 1010             VOP_ACCESS(nd.ni_vp, VWRITE, p->p_ucred, p) == 0 ||
 1011             VOP_ACCESS(nd.ni_vp, VEXEC, p->p_ucred, p) == 0)) ||
 1012             (nd.ni_dvp &&
 1013             (VOP_ACCESS(nd.ni_dvp, VREAD, p->p_ucred, p) == 0 ||
 1014             VOP_ACCESS(nd.ni_dvp, VWRITE, p->p_ucred, p) == 0 ||
 1015             VOP_ACCESS(nd.ni_dvp, VEXEC, p->p_ucred, p) == 0)));
 1016 
 1017         /* release lock from namei, but keep ref */
 1018         if (nd.ni_vp)
 1019                 VOP_UNLOCK(nd.ni_vp);
 1020         if (nd.ni_dvp && nd.ni_dvp != nd.ni_vp)
 1021                 VOP_UNLOCK(nd.ni_dvp);
 1022 
 1023         if (allow)
 1024                 error = unveil_add(p, &nd, permissions);
 1025         else
 1026                 error = EPERM;
 1027 
 1028         /* release vref from namei, but not vref from unveil_add */
 1029         if (nd.ni_vp)
 1030                 vrele(nd.ni_vp);
 1031         if (nd.ni_dvp)
 1032                 vrele(nd.ni_dvp);
 1033 
 1034         pool_put(&namei_pool, nd.ni_cnd.cn_pnbuf);
 1035 end:
 1036         pool_put(&namei_pool, pathname);
 1037 
 1038         return (error);
 1039 }
 1040 
 1041 /*
 1042  * Check permissions, allocate an open file structure,
 1043  * and call the device open routine if any.
 1044  */
 1045 int
 1046 sys_open(struct proc *p, void *v, register_t *retval)
 1047 {
 1048         struct sys_open_args /* {
 1049                 syscallarg(const char *) path;
 1050                 syscallarg(int) flags;
 1051                 syscallarg(mode_t) mode;
 1052         } */ *uap = v;
 1053 
 1054         return (doopenat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, flags),
 1055             SCARG(uap, mode), retval));
 1056 }
 1057 
 1058 int
 1059 sys_openat(struct proc *p, void *v, register_t *retval)
 1060 {
 1061         struct sys_openat_args /* {
 1062                 syscallarg(int) fd;
 1063                 syscallarg(const char *) path;
 1064                 syscallarg(int) flags;
 1065                 syscallarg(mode_t) mode;
 1066         } */ *uap = v;
 1067 
 1068         return (doopenat(p, SCARG(uap, fd), SCARG(uap, path),
 1069             SCARG(uap, flags), SCARG(uap, mode), retval));
 1070 }
 1071 
 1072 int
 1073 doopenat(struct proc *p, int fd, const char *path, int oflags, mode_t mode,
 1074     register_t *retval)
 1075 {
 1076         struct filedesc *fdp = p->p_fd;
 1077         struct file *fp;
 1078         struct vnode *vp;
 1079         struct vattr vattr;
 1080         int flags, cloexec, cmode;
 1081         int type, indx, error, localtrunc = 0;
 1082         struct flock lf;
 1083         struct nameidata nd;
 1084         uint64_t ni_pledge = 0;
 1085         u_char ni_unveil = 0;
 1086 
 1087         if (oflags & (O_EXLOCK | O_SHLOCK)) {
 1088                 error = pledge_flock(p);
 1089                 if (error != 0)
 1090                         return (error);
 1091         }
 1092 
 1093         cloexec = (oflags & O_CLOEXEC) ? UF_EXCLOSE : 0;
 1094 
 1095         fdplock(fdp);
 1096         if ((error = falloc(p, &fp, &indx)) != 0) {
 1097                 fdpunlock(fdp);
 1098                 return (error);
 1099         }
 1100         fdpunlock(fdp);
 1101 
 1102         flags = FFLAGS(oflags);
 1103         if (flags & FREAD) {
 1104                 ni_pledge |= PLEDGE_RPATH;
 1105                 ni_unveil |= UNVEIL_READ;
 1106         }
 1107         if (flags & FWRITE) {
 1108                 ni_pledge |= PLEDGE_WPATH;
 1109                 ni_unveil |= UNVEIL_WRITE;
 1110         }
 1111         if (oflags & O_CREAT) {
 1112                 ni_pledge |= PLEDGE_CPATH;
 1113                 ni_unveil |= UNVEIL_CREATE;
 1114         }
 1115 
 1116         cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
 1117         if ((p->p_p->ps_flags & PS_PLEDGE))
 1118                 cmode &= ACCESSPERMS;
 1119         NDINITAT(&nd, 0, 0, UIO_USERSPACE, fd, path, p);
 1120         nd.ni_pledge = ni_pledge;
 1121         nd.ni_unveil = ni_unveil;
 1122         p->p_dupfd = -1;                        /* XXX check for fdopen */
 1123         if ((flags & O_TRUNC) && (flags & (O_EXLOCK | O_SHLOCK))) {
 1124                 localtrunc = 1;
 1125                 flags &= ~O_TRUNC;      /* Must do truncate ourselves */
 1126         }
 1127         if ((error = vn_open(&nd, flags, cmode)) != 0) {
 1128                 fdplock(fdp);
 1129                 if (error == ENODEV &&
 1130                     p->p_dupfd >= 0 &&                  /* XXX from fdopen */
 1131                     (error =
 1132                         dupfdopen(p, indx, flags)) == 0) {
 1133                         fdpunlock(fdp);
 1134                         closef(fp, p);
 1135                         *retval = indx;
 1136                         return (error);
 1137                 }
 1138                 if (error == ERESTART)
 1139                         error = EINTR;
 1140                 fdremove(fdp, indx);
 1141                 fdpunlock(fdp);
 1142                 closef(fp, p);
 1143                 return (error);
 1144         }
 1145         p->p_dupfd = 0;
 1146         vp = nd.ni_vp;
 1147         fp->f_flag = flags & FMASK;
 1148         fp->f_type = DTYPE_VNODE;
 1149         fp->f_ops = &vnops;
 1150         fp->f_data = vp;
 1151         if (flags & (O_EXLOCK | O_SHLOCK)) {
 1152                 lf.l_whence = SEEK_SET;
 1153                 lf.l_start = 0;
 1154                 lf.l_len = 0;
 1155                 if (flags & O_EXLOCK)
 1156                         lf.l_type = F_WRLCK;
 1157                 else
 1158                         lf.l_type = F_RDLCK;
 1159                 type = F_FLOCK;
 1160                 if ((flags & FNONBLOCK) == 0)
 1161                         type |= F_WAIT;
 1162                 VOP_UNLOCK(vp);
 1163                 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
 1164                 if (error) {
 1165                         fdplock(fdp);
 1166                         /* closef will vn_close the file for us. */
 1167                         fdremove(fdp, indx);
 1168                         fdpunlock(fdp);
 1169                         closef(fp, p);
 1170                         return (error);
 1171                 }
 1172                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 1173                 atomic_setbits_int(&fp->f_iflags, FIF_HASLOCK);
 1174         }
 1175         if (localtrunc) {
 1176                 if ((fp->f_flag & FWRITE) == 0)
 1177                         error = EACCES;
 1178                 else if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_RDONLY))
 1179                         error = EROFS;
 1180                 else if (vp->v_type == VDIR)
 1181                         error = EISDIR;
 1182                 else if ((error = vn_writechk(vp)) == 0) {
 1183                         VATTR_NULL(&vattr);
 1184                         vattr.va_size = 0;
 1185                         error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
 1186                 }
 1187                 if (error) {
 1188                         VOP_UNLOCK(vp);
 1189                         fdplock(fdp);
 1190                         /* closef will close the file for us. */
 1191                         fdremove(fdp, indx);
 1192                         fdpunlock(fdp);
 1193                         closef(fp, p);
 1194                         return (error);
 1195                 }
 1196         }
 1197         VOP_UNLOCK(vp);
 1198         *retval = indx;
 1199         fdplock(fdp);
 1200         fdinsert(fdp, indx, cloexec, fp);
 1201         fdpunlock(fdp);
 1202         FRELE(fp, p);
 1203         return (error);
 1204 }
 1205 
 1206 /*
 1207  * Open a new created file (in /tmp) suitable for mmaping.
 1208  */
 1209 int
 1210 sys___tmpfd(struct proc *p, void *v, register_t *retval)
 1211 {
 1212         struct sys___tmpfd_args /* {
 1213                 syscallarg(int) flags;
 1214         } */ *uap = v;
 1215         struct filedesc *fdp = p->p_fd;
 1216         struct file *fp;
 1217         struct vnode *vp;
 1218         int oflags = SCARG(uap, flags);
 1219         int flags, cloexec, cmode;
 1220         int indx, error;
 1221         unsigned int i;
 1222         struct nameidata nd;
 1223         char path[64];
 1224         static const char *letters = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789_-";
 1225 
 1226         /* most flags are hardwired */
 1227         oflags = O_RDWR | O_CREAT | O_EXCL | O_NOFOLLOW | (oflags & O_CLOEXEC);
 1228 
 1229         cloexec = (oflags & O_CLOEXEC) ? UF_EXCLOSE : 0;
 1230 
 1231         fdplock(fdp);
 1232         if ((error = falloc(p, &fp, &indx)) != 0) {
 1233                 fdpunlock(fdp);
 1234                 return (error);
 1235         }
 1236         fdpunlock(fdp);
 1237 
 1238         flags = FFLAGS(oflags);
 1239 
 1240         arc4random_buf(path, sizeof(path));
 1241         memcpy(path, "/tmp/", 5);
 1242         for (i = 5; i < sizeof(path) - 1; i++)
 1243                 path[i] = letters[(unsigned char)path[i] & 63];
 1244         path[sizeof(path)-1] = 0;
 1245 
 1246         cmode = 0600;
 1247         NDINITAT(&nd, 0, KERNELPATH, UIO_SYSSPACE, AT_FDCWD, path, p);
 1248         if ((error = vn_open(&nd, flags, cmode)) != 0) {
 1249                 if (error == ERESTART)
 1250                         error = EINTR;
 1251                 fdplock(fdp);
 1252                 fdremove(fdp, indx);
 1253                 fdpunlock(fdp);
 1254                 closef(fp, p);
 1255                 return (error);
 1256         }
 1257         vp = nd.ni_vp;
 1258         fp->f_flag = flags & FMASK;
 1259         fp->f_type = DTYPE_VNODE;
 1260         fp->f_ops = &vnops;
 1261         fp->f_data = vp;
 1262         VOP_UNLOCK(vp);
 1263         *retval = indx;
 1264         fdplock(fdp);
 1265         fdinsert(fdp, indx, cloexec, fp);
 1266         fdpunlock(fdp);
 1267         FRELE(fp, p);
 1268 
 1269         /* unlink it */
 1270         /* XXX
 1271          * there is a wee race here, although it is mostly inconsequential.
 1272          * perhaps someday we can create a file like object without a name...
 1273          */
 1274         NDINITAT(&nd, DELETE, KERNELPATH | LOCKPARENT | LOCKLEAF, UIO_SYSSPACE,
 1275             AT_FDCWD, path, p);
 1276         if ((error = namei(&nd)) != 0) {
 1277                 printf("can't unlink temp file! %d\n", error);
 1278                 error = 0;
 1279         } else {
 1280                 vp = nd.ni_vp;
 1281                 uvm_vnp_uncache(vp);
 1282                 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
 1283                 if (error) {
 1284                         printf("error removing vop: %d\n", error);
 1285                         error = 0;
 1286                 }
 1287         }
 1288 
 1289         return (error);
 1290 }
 1291 
 1292 /*
 1293  * Get file handle system call
 1294  */
 1295 int
 1296 sys_getfh(struct proc *p, void *v, register_t *retval)
 1297 {
 1298         struct sys_getfh_args /* {
 1299                 syscallarg(const char *) fname;
 1300                 syscallarg(fhandle_t *) fhp;
 1301         } */ *uap = v;
 1302         struct vnode *vp;
 1303         fhandle_t fh;
 1304         int error;
 1305         struct nameidata nd;
 1306 
 1307         /*
 1308          * Must be super user
 1309          */
 1310         error = suser(p);
 1311         if (error)
 1312                 return (error);
 1313         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
 1314             SCARG(uap, fname), p);
 1315         error = namei(&nd);
 1316         if (error)
 1317                 return (error);
 1318         vp = nd.ni_vp;
 1319         memset(&fh, 0, sizeof(fh));
 1320         fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
 1321         error = VFS_VPTOFH(vp, &fh.fh_fid);
 1322         vput(vp);
 1323         if (error)
 1324                 return (error);
 1325         error = copyout(&fh, SCARG(uap, fhp), sizeof(fh));
 1326         return (error);
 1327 }
 1328 
 1329 /*
 1330  * Open a file given a file handle.
 1331  *
 1332  * Check permissions, allocate an open file structure,
 1333  * and call the device open routine if any.
 1334  */
 1335 int
 1336 sys_fhopen(struct proc *p, void *v, register_t *retval)
 1337 {
 1338         struct sys_fhopen_args /* {
 1339                 syscallarg(const fhandle_t *) fhp;
 1340                 syscallarg(int) flags;
 1341         } */ *uap = v;
 1342         struct filedesc *fdp = p->p_fd;
 1343         struct file *fp;
 1344         struct vnode *vp = NULL;
 1345         struct mount *mp;
 1346         struct ucred *cred = p->p_ucred;
 1347         int flags, cloexec;
 1348         int type, indx, error=0;
 1349         struct flock lf;
 1350         struct vattr va;
 1351         fhandle_t fh;
 1352 
 1353         /*
 1354          * Must be super user
 1355          */
 1356         if ((error = suser(p)))
 1357                 return (error);
 1358 
 1359         flags = FFLAGS(SCARG(uap, flags));
 1360         if ((flags & (FREAD | FWRITE)) == 0)
 1361                 return (EINVAL);
 1362         if ((flags & O_CREAT))
 1363                 return (EINVAL);
 1364 
 1365         cloexec = (flags & O_CLOEXEC) ? UF_EXCLOSE : 0;
 1366 
 1367         fdplock(fdp);
 1368         if ((error = falloc(p, &fp, &indx)) != 0) {
 1369                 fdpunlock(fdp);
 1370                 fp = NULL;
 1371                 goto bad;
 1372         }
 1373         fdpunlock(fdp);
 1374 
 1375         if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
 1376                 goto bad;
 1377 
 1378         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
 1379                 error = ESTALE;
 1380                 goto bad;
 1381         }
 1382 
 1383         if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
 1384                 vp = NULL;      /* most likely unnecessary sanity for bad: */
 1385                 goto bad;
 1386         }
 1387 
 1388         /* Now do an effective vn_open */
 1389 
 1390         if (vp->v_type == VSOCK) {
 1391                 error = EOPNOTSUPP;
 1392                 goto bad;
 1393         }
 1394         if ((flags & O_DIRECTORY) && vp->v_type != VDIR) {
 1395                 error = ENOTDIR;
 1396                 goto bad;
 1397         }
 1398         if (flags & FREAD) {
 1399                 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
 1400                         goto bad;
 1401         }
 1402         if (flags & (FWRITE | O_TRUNC)) {
 1403                 if (vp->v_type == VDIR) {
 1404                         error = EISDIR;
 1405                         goto bad;
 1406                 }
 1407                 if ((error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0 ||
 1408                     (error = vn_writechk(vp)) != 0)
 1409                         goto bad;
 1410         }
 1411         if (flags & O_TRUNC) {
 1412                 VATTR_NULL(&va);
 1413                 va.va_size = 0;
 1414                 if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
 1415                         goto bad;
 1416         }
 1417         if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
 1418                 goto bad;
 1419         if (flags & FWRITE)
 1420                 vp->v_writecount++;
 1421 
 1422         /* done with modified vn_open, now finish what sys_open does. */
 1423 
 1424         fp->f_flag = flags & FMASK;
 1425         fp->f_type = DTYPE_VNODE;
 1426         fp->f_ops = &vnops;
 1427         fp->f_data = vp;
 1428         if (flags & (O_EXLOCK | O_SHLOCK)) {
 1429                 lf.l_whence = SEEK_SET;
 1430                 lf.l_start = 0;
 1431                 lf.l_len = 0;
 1432                 if (flags & O_EXLOCK)
 1433                         lf.l_type = F_WRLCK;
 1434                 else
 1435                         lf.l_type = F_RDLCK;
 1436                 type = F_FLOCK;
 1437                 if ((flags & FNONBLOCK) == 0)
 1438                         type |= F_WAIT;
 1439                 VOP_UNLOCK(vp);
 1440                 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
 1441                 if (error) {
 1442                         vp = NULL;      /* closef will vn_close the file */
 1443                         goto bad;
 1444                 }
 1445                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 1446                 atomic_setbits_int(&fp->f_iflags, FIF_HASLOCK);
 1447         }
 1448         VOP_UNLOCK(vp);
 1449         *retval = indx;
 1450         fdplock(fdp);
 1451         fdinsert(fdp, indx, cloexec, fp);
 1452         fdpunlock(fdp);
 1453         FRELE(fp, p);
 1454         return (0);
 1455 
 1456 bad:
 1457         if (fp) {
 1458                 fdplock(fdp);
 1459                 fdremove(fdp, indx);
 1460                 fdpunlock(fdp);
 1461                 closef(fp, p);
 1462                 if (vp != NULL)
 1463                         vput(vp);
 1464         }
 1465         return (error);
 1466 }
 1467 
 1468 int
 1469 sys_fhstat(struct proc *p, void *v, register_t *retval)
 1470 {
 1471         struct sys_fhstat_args /* {
 1472                 syscallarg(const fhandle_t *) fhp;
 1473                 syscallarg(struct stat *) sb;
 1474         } */ *uap = v;
 1475         struct stat sb;
 1476         int error;
 1477         fhandle_t fh;
 1478         struct mount *mp;
 1479         struct vnode *vp;
 1480 
 1481         /*
 1482          * Must be super user
 1483          */
 1484         if ((error = suser(p)))
 1485                 return (error);
 1486 
 1487         if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
 1488                 return (error);
 1489 
 1490         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
 1491                 return (ESTALE);
 1492         if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
 1493                 return (error);
 1494         error = vn_stat(vp, &sb, p);
 1495         vput(vp);
 1496         if (error)
 1497                 return (error);
 1498         error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
 1499         return (error);
 1500 }
 1501 
 1502 int
 1503 sys_fhstatfs(struct proc *p, void *v, register_t *retval)
 1504 {
 1505         struct sys_fhstatfs_args /* {
 1506                 syscallarg(const fhandle_t *) fhp;
 1507                 syscallarg(struct statfs *) buf;
 1508         } */ *uap = v;
 1509         struct statfs *sp;
 1510         fhandle_t fh;
 1511         struct mount *mp;
 1512         struct vnode *vp;
 1513         int error;
 1514 
 1515         /*
 1516          * Must be super user
 1517          */
 1518         if ((error = suser(p)))
 1519                 return (error);
 1520 
 1521         if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
 1522                 return (error);
 1523 
 1524         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
 1525                 return (ESTALE);
 1526         if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
 1527                 return (error);
 1528         mp = vp->v_mount;
 1529         sp = &mp->mnt_stat;
 1530         vput(vp);
 1531         if ((error = VFS_STATFS(mp, sp, p)) != 0)
 1532                 return (error);
 1533         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
 1534         return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
 1535 }
 1536 
 1537 /*
 1538  * Create a special file or named pipe.
 1539  */
 1540 int
 1541 sys_mknod(struct proc *p, void *v, register_t *retval)
 1542 {
 1543         struct sys_mknod_args /* {
 1544                 syscallarg(const char *) path;
 1545                 syscallarg(mode_t) mode;
 1546                 syscallarg(int) dev;
 1547         } */ *uap = v;
 1548 
 1549         return (domknodat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode),
 1550             SCARG(uap, dev)));
 1551 }
 1552 
 1553 int
 1554 sys_mknodat(struct proc *p, void *v, register_t *retval)
 1555 {
 1556         struct sys_mknodat_args /* {
 1557                 syscallarg(int) fd;
 1558                 syscallarg(const char *) path;
 1559                 syscallarg(mode_t) mode;
 1560                 syscallarg(dev_t) dev;
 1561         } */ *uap = v;
 1562 
 1563         return (domknodat(p, SCARG(uap, fd), SCARG(uap, path),
 1564             SCARG(uap, mode), SCARG(uap, dev)));
 1565 }
 1566 
 1567 int
 1568 domknodat(struct proc *p, int fd, const char *path, mode_t mode, dev_t dev)
 1569 {
 1570         struct vnode *vp;
 1571         struct vattr vattr;
 1572         int error;
 1573         struct nameidata nd;
 1574 
 1575         if (dev == VNOVAL)
 1576                 return (EINVAL);
 1577         NDINITAT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, fd, path, p);
 1578         nd.ni_pledge = PLEDGE_DPATH;
 1579         nd.ni_unveil = UNVEIL_CREATE;
 1580         if ((error = namei(&nd)) != 0)
 1581                 return (error);
 1582         vp = nd.ni_vp;
 1583         if (!S_ISFIFO(mode) || dev != 0) {
 1584                 if (!vnoperm(nd.ni_dvp) && (error = suser(p)) != 0)
 1585                         goto out;
 1586                 if (p->p_fd->fd_rdir) {
 1587                         error = EINVAL;
 1588                         goto out;
 1589                 }
 1590         }
 1591         if (vp != NULL)
 1592                 error = EEXIST;
 1593         else {
 1594                 VATTR_NULL(&vattr);
 1595                 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
 1596                 if ((p->p_p->ps_flags & PS_PLEDGE))
 1597                         vattr.va_mode &= ACCESSPERMS;
 1598                 vattr.va_rdev = dev;
 1599 
 1600                 switch (mode & S_IFMT) {
 1601                 case S_IFMT:    /* used by badsect to flag bad sectors */
 1602                         vattr.va_type = VBAD;
 1603                         break;
 1604                 case S_IFCHR:
 1605                         vattr.va_type = VCHR;
 1606                         break;
 1607                 case S_IFBLK:
 1608                         vattr.va_type = VBLK;
 1609                         break;
 1610                 case S_IFIFO:
 1611 #ifndef FIFO
 1612                         error = EOPNOTSUPP;
 1613                         break;
 1614 #else
 1615                         if (dev == 0) {
 1616                                 vattr.va_type = VFIFO;
 1617                                 break;
 1618                         }
 1619                         /* FALLTHROUGH */
 1620 #endif /* FIFO */
 1621                 default:
 1622                         error = EINVAL;
 1623                         break;
 1624                 }
 1625         }
 1626 out:
 1627         if (!error) {
 1628                 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
 1629                 vput(nd.ni_dvp);
 1630         } else {
 1631                 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
 1632                 if (nd.ni_dvp == vp)
 1633                         vrele(nd.ni_dvp);
 1634                 else
 1635                         vput(nd.ni_dvp);
 1636                 if (vp)
 1637                         vrele(vp);
 1638         }
 1639         return (error);
 1640 }
 1641 
 1642 /*
 1643  * Create a named pipe.
 1644  */
 1645 int
 1646 sys_mkfifo(struct proc *p, void *v, register_t *retval)
 1647 {
 1648         struct sys_mkfifo_args /* {
 1649                 syscallarg(const char *) path;
 1650                 syscallarg(mode_t) mode;
 1651         } */ *uap = v;
 1652 
 1653         return (domknodat(p, AT_FDCWD, SCARG(uap, path),
 1654             (SCARG(uap, mode) & ALLPERMS) | S_IFIFO, 0));
 1655 }
 1656 
 1657 int
 1658 sys_mkfifoat(struct proc *p, void *v, register_t *retval)
 1659 {
 1660         struct sys_mkfifoat_args /* {
 1661                 syscallarg(int) fd;
 1662                 syscallarg(const char *) path;
 1663                 syscallarg(mode_t) mode;
 1664         } */ *uap = v;
 1665 
 1666         return (domknodat(p, SCARG(uap, fd), SCARG(uap, path),
 1667             (SCARG(uap, mode) & ALLPERMS) | S_IFIFO, 0));
 1668 }
 1669 
 1670 /*
 1671  * Make a hard file link.
 1672  */
 1673 int
 1674 sys_link(struct proc *p, void *v, register_t *retval)
 1675 {
 1676         struct sys_link_args /* {
 1677                 syscallarg(const char *) path;
 1678                 syscallarg(const char *) link;
 1679         } */ *uap = v;
 1680 
 1681         return (dolinkat(p, AT_FDCWD, SCARG(uap, path), AT_FDCWD,
 1682             SCARG(uap, link), AT_SYMLINK_FOLLOW));
 1683 }
 1684 
 1685 int
 1686 sys_linkat(struct proc *p, void *v, register_t *retval)
 1687 {
 1688         struct sys_linkat_args /* {
 1689                 syscallarg(int) fd1;
 1690                 syscallarg(const char *) path1;
 1691                 syscallarg(int) fd2;
 1692                 syscallarg(const char *) path2;
 1693                 syscallarg(int) flag;
 1694         } */ *uap = v;
 1695 
 1696         return (dolinkat(p, SCARG(uap, fd1), SCARG(uap, path1),
 1697             SCARG(uap, fd2), SCARG(uap, path2), SCARG(uap, flag)));
 1698 }
 1699 
 1700 int
 1701 dolinkat(struct proc *p, int fd1, const char *path1, int fd2,
 1702     const char *path2, int flag)
 1703 {
 1704         struct vnode *vp;
 1705         struct nameidata nd;
 1706         int error, follow;
 1707         int flags;
 1708 
 1709         if (flag & ~AT_SYMLINK_FOLLOW)
 1710                 return (EINVAL);
 1711 
 1712         follow = (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW;
 1713         NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd1, path1, p);
 1714         nd.ni_pledge = PLEDGE_RPATH;
 1715         nd.ni_unveil = UNVEIL_READ;
 1716         if ((error = namei(&nd)) != 0)
 1717                 return (error);
 1718         vp = nd.ni_vp;
 1719 
 1720         flags = LOCKPARENT;
 1721         if (vp->v_type == VDIR) {
 1722                 flags |= STRIPSLASHES;
 1723         }
 1724 
 1725         NDINITAT(&nd, CREATE, flags, UIO_USERSPACE, fd2, path2, p);
 1726         nd.ni_pledge = PLEDGE_CPATH;
 1727         nd.ni_unveil = UNVEIL_CREATE;
 1728         if ((error = namei(&nd)) != 0)
 1729                 goto out;
 1730         if (nd.ni_vp) {
 1731                 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
 1732                 if (nd.ni_dvp == nd.ni_vp)
 1733                         vrele(nd.ni_dvp);
 1734                 else
 1735                         vput(nd.ni_dvp);
 1736                 vrele(nd.ni_vp);
 1737                 error = EEXIST;
 1738                 goto out;
 1739         }
 1740         error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
 1741 out:
 1742         vrele(vp);
 1743         return (error);
 1744 }
 1745 
 1746 /*
 1747  * Make a symbolic link.
 1748  */
 1749 int
 1750 sys_symlink(struct proc *p, void *v, register_t *retval)
 1751 {
 1752         struct sys_symlink_args /* {
 1753                 syscallarg(const char *) path;
 1754                 syscallarg(const char *) link;
 1755         } */ *uap = v;
 1756 
 1757         return (dosymlinkat(p, SCARG(uap, path), AT_FDCWD, SCARG(uap, link)));
 1758 }
 1759 
 1760 int
 1761 sys_symlinkat(struct proc *p, void *v, register_t *retval)
 1762 {
 1763         struct sys_symlinkat_args /* {
 1764                 syscallarg(const char *) path;
 1765                 syscallarg(int) fd;
 1766                 syscallarg(const char *) link;
 1767         } */ *uap = v;
 1768 
 1769         return (dosymlinkat(p, SCARG(uap, path), SCARG(uap, fd),
 1770             SCARG(uap, link)));
 1771 }
 1772 
 1773 int
 1774 dosymlinkat(struct proc *p, const char *upath, int fd, const char *link)
 1775 {
 1776         struct vattr vattr;
 1777         char *path;
 1778         int error;
 1779         struct nameidata nd;
 1780 
 1781         path = pool_get(&namei_pool, PR_WAITOK);
 1782         error = copyinstr(upath, path, MAXPATHLEN, NULL);
 1783         if (error)
 1784                 goto out;
 1785         NDINITAT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, fd, link, p);
 1786         nd.ni_pledge = PLEDGE_CPATH;
 1787         nd.ni_unveil = UNVEIL_CREATE;
 1788         if ((error = namei(&nd)) != 0)
 1789                 goto out;
 1790         if (nd.ni_vp) {
 1791                 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
 1792                 if (nd.ni_dvp == nd.ni_vp)
 1793                         vrele(nd.ni_dvp);
 1794                 else
 1795                         vput(nd.ni_dvp);
 1796                 vrele(nd.ni_vp);
 1797                 error = EEXIST;
 1798                 goto out;
 1799         }
 1800         VATTR_NULL(&vattr);
 1801         vattr.va_mode = ACCESSPERMS &~ p->p_fd->fd_cmask;
 1802         error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
 1803 out:
 1804         pool_put(&namei_pool, path);
 1805         return (error);
 1806 }
 1807 
 1808 /*
 1809  * Delete a name from the filesystem.
 1810  */
 1811 int
 1812 sys_unlink(struct proc *p, void *v, register_t *retval)
 1813 {
 1814         struct sys_unlink_args /* {
 1815                 syscallarg(const char *) path;
 1816         } */ *uap = v;
 1817 
 1818         return (dounlinkat(p, AT_FDCWD, SCARG(uap, path), 0));
 1819 }
 1820 
 1821 int
 1822 sys_unlinkat(struct proc *p, void *v, register_t *retval)
 1823 {
 1824         struct sys_unlinkat_args /* {
 1825                 syscallarg(int) fd;
 1826                 syscallarg(const char *) path;
 1827                 syscallarg(int) flag;
 1828         } */ *uap = v;
 1829 
 1830         return (dounlinkat(p, SCARG(uap, fd), SCARG(uap, path),
 1831             SCARG(uap, flag)));
 1832 }
 1833 
 1834 int
 1835 dounlinkat(struct proc *p, int fd, const char *path, int flag)
 1836 {
 1837         struct vnode *vp;
 1838         int error;
 1839         struct nameidata nd;
 1840 
 1841         if (flag & ~AT_REMOVEDIR)
 1842                 return (EINVAL);
 1843 
 1844         NDINITAT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
 1845             fd, path, p);
 1846         nd.ni_pledge = PLEDGE_CPATH;
 1847         nd.ni_unveil = UNVEIL_CREATE;
 1848         if ((error = namei(&nd)) != 0)
 1849                 return (error);
 1850         vp = nd.ni_vp;
 1851 
 1852         if (flag & AT_REMOVEDIR) {
 1853                 if (vp->v_type != VDIR) {
 1854                         error = ENOTDIR;
 1855                         goto out;
 1856                 }
 1857                 /*
 1858                  * No rmdir "." please.
 1859                  */
 1860                 if (nd.ni_dvp == vp) {
 1861                         error = EINVAL;
 1862                         goto out;
 1863                 }
 1864                 /*
 1865                  * A mounted on directory cannot be deleted.
 1866                  */
 1867                 if (vp->v_mountedhere != NULL) {
 1868                         error = EBUSY;
 1869                         goto out;
 1870                 }
 1871         }
 1872 
 1873         /*
 1874          * The root of a mounted filesystem cannot be deleted.
 1875          */
 1876         if (vp->v_flag & VROOT)
 1877                 error = EBUSY;
 1878 out:
 1879         if (!error) {
 1880                 if (flag & AT_REMOVEDIR) {
 1881                         error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
 1882                 } else {
 1883                         (void)uvm_vnp_uncache(vp);
 1884                         error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
 1885                 }
 1886         } else {
 1887                 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
 1888                 if (nd.ni_dvp == vp)
 1889                         vrele(nd.ni_dvp);
 1890                 else
 1891                         vput(nd.ni_dvp);
 1892                 vput(vp);
 1893         }
 1894         return (error);
 1895 }
 1896 
 1897 /*
 1898  * Reposition read/write file offset.
 1899  */
 1900 int
 1901 sys_lseek(struct proc *p, void *v, register_t *retval)
 1902 {
 1903         struct sys_lseek_args /* {
 1904                 syscallarg(int) fd;
 1905                 syscallarg(off_t) offset;
 1906                 syscallarg(int) whence;
 1907         } */ *uap = v;
 1908         struct filedesc *fdp = p->p_fd;
 1909         struct file *fp;
 1910         off_t offset;
 1911         int error;
 1912 
 1913         if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
 1914                 return (EBADF);
 1915         if (fp->f_ops->fo_seek == NULL) {
 1916                 error = ESPIPE;
 1917                 goto bad;
 1918         }
 1919         offset = SCARG(uap, offset);
 1920 
 1921         error = (*fp->f_ops->fo_seek)(fp, &offset, SCARG(uap, whence), p);
 1922         if (error)
 1923                 goto bad;
 1924 
 1925         *(off_t *)retval = offset;
 1926         mtx_enter(&fp->f_mtx);
 1927         fp->f_seek++;
 1928         mtx_leave(&fp->f_mtx);
 1929         error = 0;
 1930  bad:
 1931         FRELE(fp, p);
 1932         return (error);
 1933 }
 1934 
 1935 #if 1
 1936 int
 1937 sys_pad_lseek(struct proc *p, void *v, register_t *retval)
 1938 {
 1939         struct sys_pad_lseek_args *uap = v;
 1940         struct sys_lseek_args unpad;
 1941 
 1942         SCARG(&unpad, fd) = SCARG(uap, fd);
 1943         SCARG(&unpad, offset) = SCARG(uap, offset);
 1944         SCARG(&unpad, whence) = SCARG(uap, whence);
 1945         return sys_lseek(p, &unpad, retval);
 1946 }
 1947 #endif
 1948 
 1949 /*
 1950  * Check access permissions.
 1951  */
 1952 int
 1953 sys_access(struct proc *p, void *v, register_t *retval)
 1954 {
 1955         struct sys_access_args /* {
 1956                 syscallarg(const char *) path;
 1957                 syscallarg(int) amode;
 1958         } */ *uap = v;
 1959 
 1960         return (dofaccessat(p, AT_FDCWD, SCARG(uap, path),
 1961             SCARG(uap, amode), 0));
 1962 }
 1963 
 1964 int
 1965 sys_faccessat(struct proc *p, void *v, register_t *retval)
 1966 {
 1967         struct sys_faccessat_args /* {
 1968                 syscallarg(int) fd;
 1969                 syscallarg(const char *) path;
 1970                 syscallarg(int) amode;
 1971                 syscallarg(int) flag;
 1972         } */ *uap = v;
 1973 
 1974         return (dofaccessat(p, SCARG(uap, fd), SCARG(uap, path),
 1975             SCARG(uap, amode), SCARG(uap, flag)));
 1976 }
 1977 
 1978 int
 1979 dofaccessat(struct proc *p, int fd, const char *path, int amode, int flag)
 1980 {
 1981         struct vnode *vp;
 1982         struct ucred *newcred, *oldcred;
 1983         struct nameidata nd;
 1984         int error;
 1985 
 1986         if (amode & ~(R_OK | W_OK | X_OK))
 1987                 return (EINVAL);
 1988         if (flag & ~AT_EACCESS)
 1989                 return (EINVAL);
 1990 
 1991         newcred = NULL;
 1992         oldcred = p->p_ucred;
 1993 
 1994         /*
 1995          * If access as real ids was requested and they really differ,
 1996          * give the thread new creds with them reset
 1997          */
 1998         if ((flag & AT_EACCESS) == 0 &&
 1999             (oldcred->cr_uid != oldcred->cr_ruid ||
 2000             (oldcred->cr_gid != oldcred->cr_rgid))) {
 2001                 p->p_ucred = newcred = crdup(oldcred);
 2002                 newcred->cr_uid = newcred->cr_ruid;
 2003                 newcred->cr_gid = newcred->cr_rgid;
 2004         }
 2005 
 2006         NDINITAT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fd, path, p);
 2007         nd.ni_pledge = PLEDGE_RPATH;
 2008         nd.ni_unveil = UNVEIL_READ;
 2009         if ((error = namei(&nd)) != 0)
 2010                 goto out;
 2011         vp = nd.ni_vp;
 2012 
 2013         /* Flags == 0 means only check for existence. */
 2014         if (amode) {
 2015                 int vflags = 0;
 2016 
 2017                 if (amode & R_OK)
 2018                         vflags |= VREAD;
 2019                 if (amode & W_OK)
 2020                         vflags |= VWRITE;
 2021                 if (amode & X_OK)
 2022                         vflags |= VEXEC;
 2023 
 2024                 error = VOP_ACCESS(vp, vflags, p->p_ucred, p);
 2025                 if (!error && (vflags & VWRITE))
 2026                         error = vn_writechk(vp);
 2027         }
 2028         vput(vp);
 2029 out:
 2030         if (newcred != NULL) {
 2031                 p->p_ucred = oldcred;
 2032                 crfree(newcred);
 2033         }
 2034         return (error);
 2035 }
 2036 
 2037 /*
 2038  * Get file status; this version follows links.
 2039  */
 2040 int
 2041 sys_stat(struct proc *p, void *v, register_t *retval)
 2042 {
 2043         struct sys_stat_args /* {
 2044                 syscallarg(const char *) path;
 2045                 syscallarg(struct stat *) ub;
 2046         } */ *uap = v;
 2047 
 2048         return (dofstatat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, ub), 0));
 2049 }
 2050 
 2051 int
 2052 sys_fstatat(struct proc *p, void *v, register_t *retval)
 2053 {
 2054         struct sys_fstatat_args /* {
 2055                 syscallarg(int) fd;
 2056                 syscallarg(const char *) path;
 2057                 syscallarg(struct stat *) buf;
 2058                 syscallarg(int) flag;
 2059         } */ *uap = v;
 2060 
 2061         return (dofstatat(p, SCARG(uap, fd), SCARG(uap, path),
 2062             SCARG(uap, buf), SCARG(uap, flag)));
 2063 }
 2064 
 2065 int
 2066 dofstatat(struct proc *p, int fd, const char *path, struct stat *buf, int flag)
 2067 {
 2068         struct stat sb;
 2069         int error, follow;
 2070         struct nameidata nd;
 2071 
 2072         if (flag & ~AT_SYMLINK_NOFOLLOW)
 2073                 return (EINVAL);
 2074 
 2075 
 2076         follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
 2077         NDINITAT(&nd, LOOKUP, follow | LOCKLEAF, UIO_USERSPACE, fd, path, p);
 2078         nd.ni_pledge = PLEDGE_RPATH;
 2079         nd.ni_unveil = UNVEIL_READ;
 2080         if ((error = namei(&nd)) != 0)
 2081                 return (error);
 2082         error = vn_stat(nd.ni_vp, &sb, p);
 2083         vput(nd.ni_vp);
 2084         if (error)
 2085                 return (error);
 2086         /* Don't let non-root see generation numbers (for NFS security) */
 2087         if (suser(p))
 2088                 sb.st_gen = 0;
 2089         error = copyout(&sb, buf, sizeof(sb));
 2090 #ifdef KTRACE
 2091         if (error == 0 && KTRPOINT(p, KTR_STRUCT))
 2092                 ktrstat(p, &sb);
 2093 #endif
 2094         return (error);
 2095 }
 2096 
 2097 /*
 2098  * Get file status; this version does not follow links.
 2099  */
 2100 int
 2101 sys_lstat(struct proc *p, void *v, register_t *retval)
 2102 {
 2103         struct sys_lstat_args /* {
 2104                 syscallarg(const char *) path;
 2105                 syscallarg(struct stat *) ub;
 2106         } */ *uap = v;
 2107 
 2108         return (dofstatat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, ub),
 2109             AT_SYMLINK_NOFOLLOW));
 2110 }
 2111 
 2112 /*
 2113  * Get configurable pathname variables.
 2114  */
 2115 int
 2116 sys_pathconf(struct proc *p, void *v, register_t *retval)
 2117 {
 2118         struct sys_pathconf_args /* {
 2119                 syscallarg(const char *) path;
 2120                 syscallarg(int) name;
 2121         } */ *uap = v;
 2122         int error;
 2123         struct nameidata nd;
 2124 
 2125         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
 2126             SCARG(uap, path), p);
 2127         nd.ni_pledge = PLEDGE_RPATH;
 2128         nd.ni_unveil = UNVEIL_READ;
 2129         if ((error = namei(&nd)) != 0)
 2130                 return (error);
 2131         error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
 2132         vput(nd.ni_vp);
 2133         return (error);
 2134 }
 2135 
 2136 /*
 2137  * Return target name of a symbolic link.
 2138  */
 2139 int
 2140 sys_readlink(struct proc *p, void *v, register_t *retval)
 2141 {
 2142         struct sys_readlink_args /* {
 2143                 syscallarg(const char *) path;
 2144                 syscallarg(char *) buf;
 2145                 syscallarg(size_t) count;
 2146         } */ *uap = v;
 2147 
 2148         return (doreadlinkat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, buf),
 2149             SCARG(uap, count), retval));
 2150 }
 2151 
 2152 int
 2153 sys_readlinkat(struct proc *p, void *v, register_t *retval)
 2154 {
 2155         struct sys_readlinkat_args /* {
 2156                 syscallarg(int) fd;
 2157                 syscallarg(const char *) path;
 2158                 syscallarg(char *) buf;
 2159                 syscallarg(size_t) count;
 2160         } */ *uap = v;
 2161 
 2162         return (doreadlinkat(p, SCARG(uap, fd), SCARG(uap, path),
 2163             SCARG(uap, buf), SCARG(uap, count), retval));
 2164 }
 2165 
 2166 int
 2167 doreadlinkat(struct proc *p, int fd, const char *path, char *buf,
 2168     size_t count, register_t *retval)
 2169 {
 2170         struct vnode *vp;
 2171         struct iovec aiov;
 2172         struct uio auio;
 2173         int error;
 2174         struct nameidata nd;
 2175 
 2176         NDINITAT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, fd, path, p);
 2177         nd.ni_pledge = PLEDGE_RPATH;
 2178         nd.ni_unveil = UNVEIL_READ;
 2179         if ((error = namei(&nd)) != 0)
 2180                 return (error);
 2181         vp = nd.ni_vp;
 2182         if (vp->v_type != VLNK)
 2183                 error = EINVAL;
 2184         else {
 2185                 aiov.iov_base = buf;
 2186                 aiov.iov_len = count;
 2187                 auio.uio_iov = &aiov;
 2188                 auio.uio_iovcnt = 1;
 2189                 auio.uio_offset = 0;
 2190                 auio.uio_rw = UIO_READ;
 2191                 auio.uio_segflg = UIO_USERSPACE;
 2192                 auio.uio_procp = p;
 2193                 auio.uio_resid = count;
 2194                 error = VOP_READLINK(vp, &auio, p->p_ucred);
 2195                 *retval = count - auio.uio_resid;
 2196         }
 2197         vput(vp);
 2198         return (error);
 2199 }
 2200 
 2201 /*
 2202  * Change flags of a file given a path name.
 2203  */
 2204 int
 2205 sys_chflags(struct proc *p, void *v, register_t *retval)
 2206 {
 2207         struct sys_chflags_args /* {
 2208                 syscallarg(const char *) path;
 2209                 syscallarg(u_int) flags;
 2210         } */ *uap = v;
 2211 
 2212         return (dochflagsat(p, AT_FDCWD, SCARG(uap, path),
 2213             SCARG(uap, flags), 0));
 2214 }
 2215 
 2216 int
 2217 sys_chflagsat(struct proc *p, void *v, register_t *retval)
 2218 {
 2219         struct sys_chflagsat_args /* {
 2220                 syscallarg(int) fd;
 2221                 syscallarg(const char *) path;
 2222                 syscallarg(u_int) flags;
 2223                 syscallarg(int) atflags;
 2224         } */ *uap = v;
 2225 
 2226         return (dochflagsat(p, SCARG(uap, fd), SCARG(uap, path),
 2227             SCARG(uap, flags), SCARG(uap, atflags)));
 2228 }
 2229 
 2230 int
 2231 dochflagsat(struct proc *p, int fd, const char *path, u_int flags, int atflags)
 2232 {
 2233         struct nameidata nd;
 2234         int error, follow;
 2235 
 2236         if (atflags & ~AT_SYMLINK_NOFOLLOW)
 2237                 return (EINVAL);
 2238 
 2239         follow = (atflags & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
 2240         NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd, path, p);
 2241         nd.ni_pledge = PLEDGE_FATTR | PLEDGE_RPATH;
 2242         nd.ni_unveil = UNVEIL_WRITE;
 2243         if ((error = namei(&nd)) != 0)
 2244                 return (error);
 2245         return (dovchflags(p, nd.ni_vp, flags));
 2246 }
 2247 
 2248 /*
 2249  * Change flags of a file given a file descriptor.
 2250  */
 2251 int
 2252 sys_fchflags(struct proc *p, void *v, register_t *retval)
 2253 {
 2254         struct sys_fchflags_args /* {
 2255                 syscallarg(int) fd;
 2256                 syscallarg(u_int) flags;
 2257         } */ *uap = v;
 2258         struct file *fp;
 2259         struct vnode *vp;
 2260         int error;
 2261 
 2262         if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
 2263                 return (error);
 2264         vp = fp->f_data;
 2265         vref(vp);
 2266         FRELE(fp, p);
 2267         return (dovchflags(p, vp, SCARG(uap, flags)));
 2268 }
 2269 
 2270 int
 2271 dovchflags(struct proc *p, struct vnode *vp, u_int flags)
 2272 {
 2273         struct vattr vattr;
 2274         int error;
 2275 
 2276         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2277         if (vp->v_mount && vp->v_mount->mnt_flag & MNT_RDONLY)
 2278                 error = EROFS;
 2279         else if (flags == VNOVAL)
 2280                 error = EINVAL;
 2281         else {
 2282                 if (suser(p)) {
 2283                         if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p))
 2284                             != 0)
 2285                                 goto out;
 2286                         if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
 2287                                 error = EINVAL;
 2288                                 goto out;
 2289                         }
 2290                 }
 2291                 VATTR_NULL(&vattr);
 2292                 vattr.va_flags = flags;
 2293                 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
 2294         }
 2295 out:
 2296         vput(vp);
 2297         return (error);
 2298 }
 2299 
 2300 /*
 2301  * Change mode of a file given path name.
 2302  */
 2303 int
 2304 sys_chmod(struct proc *p, void *v, register_t *retval)
 2305 {
 2306         struct sys_chmod_args /* {
 2307                 syscallarg(const char *) path;
 2308                 syscallarg(mode_t) mode;
 2309         } */ *uap = v;
 2310 
 2311         return (dofchmodat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 0));
 2312 }
 2313 
 2314 int
 2315 sys_fchmodat(struct proc *p, void *v, register_t *retval)
 2316 {
 2317         struct sys_fchmodat_args /* {
 2318                 syscallarg(int) fd;
 2319                 syscallarg(const char *) path;
 2320                 syscallarg(mode_t) mode;
 2321                 syscallarg(int) flag;
 2322         } */ *uap = v;
 2323 
 2324         return (dofchmodat(p, SCARG(uap, fd), SCARG(uap, path),
 2325             SCARG(uap, mode), SCARG(uap, flag)));
 2326 }
 2327 
 2328 int
 2329 dofchmodat(struct proc *p, int fd, const char *path, mode_t mode, int flag)
 2330 {
 2331         struct vnode *vp;
 2332         struct vattr vattr;
 2333         int error, follow;
 2334         struct nameidata nd;
 2335 
 2336         if (mode & ~(S_IFMT | ALLPERMS))
 2337                 return (EINVAL);
 2338         if ((p->p_p->ps_flags & PS_PLEDGE))
 2339                 mode &= ACCESSPERMS;
 2340         if (flag & ~AT_SYMLINK_NOFOLLOW)
 2341                 return (EINVAL);
 2342 
 2343         follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
 2344         NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd, path, p);
 2345         nd.ni_pledge = PLEDGE_FATTR | PLEDGE_RPATH;
 2346         nd.ni_unveil = UNVEIL_WRITE;
 2347         if ((error = namei(&nd)) != 0)
 2348                 return (error);
 2349         vp = nd.ni_vp;
 2350         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2351         if (vp->v_mount->mnt_flag & MNT_RDONLY)
 2352                 error = EROFS;
 2353         else {
 2354                 VATTR_NULL(&vattr);
 2355                 vattr.va_mode = mode & ALLPERMS;
 2356                 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
 2357         }
 2358         vput(vp);
 2359         return (error);
 2360 }
 2361 
 2362 /*
 2363  * Change mode of a file given a file descriptor.
 2364  */
 2365 int
 2366 sys_fchmod(struct proc *p, void *v, register_t *retval)
 2367 {
 2368         struct sys_fchmod_args /* {
 2369                 syscallarg(int) fd;
 2370                 syscallarg(mode_t) mode;
 2371         } */ *uap = v;
 2372         struct vattr vattr;
 2373         struct vnode *vp;
 2374         struct file *fp;
 2375         mode_t mode = SCARG(uap, mode);
 2376         int error;
 2377 
 2378         if (mode & ~(S_IFMT | ALLPERMS))
 2379                 return (EINVAL);
 2380         if ((p->p_p->ps_flags & PS_PLEDGE))
 2381                 mode &= ACCESSPERMS;
 2382 
 2383         if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
 2384                 return (error);
 2385         vp = fp->f_data;
 2386         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2387         if (vp->v_mount && vp->v_mount->mnt_flag & MNT_RDONLY)
 2388                 error = EROFS;
 2389         else {
 2390                 VATTR_NULL(&vattr);
 2391                 vattr.va_mode = mode & ALLPERMS;
 2392                 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
 2393         }
 2394         VOP_UNLOCK(vp);
 2395         FRELE(fp, p);
 2396         return (error);
 2397 }
 2398 
 2399 /*
 2400  * Set ownership given a path name.
 2401  */
 2402 int
 2403 sys_chown(struct proc *p, void *v, register_t *retval)
 2404 {
 2405         struct sys_chown_args /* {
 2406                 syscallarg(const char *) path;
 2407                 syscallarg(uid_t) uid;
 2408                 syscallarg(gid_t) gid;
 2409         } */ *uap = v;
 2410 
 2411         return (dofchownat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, uid),
 2412             SCARG(uap, gid), 0));
 2413 }
 2414 
 2415 int
 2416 sys_fchownat(struct proc *p, void *v, register_t *retval)
 2417 {
 2418         struct sys_fchownat_args /* {
 2419                 syscallarg(int) fd;
 2420                 syscallarg(const char *) path;
 2421                 syscallarg(uid_t) uid;
 2422                 syscallarg(gid_t) gid;
 2423                 syscallarg(int) flag;
 2424         } */ *uap = v;
 2425 
 2426         return (dofchownat(p, SCARG(uap, fd), SCARG(uap, path),
 2427             SCARG(uap, uid), SCARG(uap, gid), SCARG(uap, flag)));
 2428 }
 2429 
 2430 int
 2431 dofchownat(struct proc *p, int fd, const char *path, uid_t uid, gid_t gid,
 2432     int flag)
 2433 {
 2434         struct vnode *vp;
 2435         struct vattr vattr;
 2436         int error, follow;
 2437         struct nameidata nd;
 2438         mode_t mode;
 2439 
 2440         if (flag & ~AT_SYMLINK_NOFOLLOW)
 2441                 return (EINVAL);
 2442 
 2443         follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
 2444         NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd, path, p);
 2445         nd.ni_pledge = PLEDGE_CHOWN | PLEDGE_RPATH;
 2446         nd.ni_unveil = UNVEIL_WRITE;
 2447         if ((error = namei(&nd)) != 0)
 2448                 return (error);
 2449         vp = nd.ni_vp;
 2450         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2451         if (vp->v_mount->mnt_flag & MNT_RDONLY)
 2452                 error = EROFS;
 2453         else {
 2454                 if ((error = pledge_chown(p, uid, gid)))
 2455                         goto out;
 2456                 if ((uid != -1 || gid != -1) &&
 2457                     !vnoperm(vp) &&
 2458                     (suser(p) || suid_clear)) {
 2459                         error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
 2460                         if (error)
 2461                                 goto out;
 2462                         mode = vattr.va_mode & ~(VSUID | VSGID);
 2463                         if (mode == vattr.va_mode)
 2464                                 mode = VNOVAL;
 2465                 } else
 2466                         mode = VNOVAL;
 2467                 VATTR_NULL(&vattr);
 2468                 vattr.va_uid = uid;
 2469                 vattr.va_gid = gid;
 2470                 vattr.va_mode = mode;
 2471                 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
 2472         }
 2473 out:
 2474         vput(vp);
 2475         return (error);
 2476 }
 2477 
 2478 /*
 2479  * Set ownership given a path name, without following links.
 2480  */
 2481 int
 2482 sys_lchown(struct proc *p, void *v, register_t *retval)
 2483 {
 2484         struct sys_lchown_args /* {
 2485                 syscallarg(const char *) path;
 2486                 syscallarg(uid_t) uid;
 2487                 syscallarg(gid_t) gid;
 2488         } */ *uap = v;
 2489         struct vnode *vp;
 2490         struct vattr vattr;
 2491         int error;
 2492         struct nameidata nd;
 2493         mode_t mode;
 2494         uid_t uid = SCARG(uap, uid);
 2495         gid_t gid = SCARG(uap, gid);
 2496 
 2497         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
 2498         nd.ni_pledge = PLEDGE_CHOWN | PLEDGE_RPATH;
 2499         nd.ni_unveil = UNVEIL_WRITE;
 2500         if ((error = namei(&nd)) != 0)
 2501                 return (error);
 2502         vp = nd.ni_vp;
 2503         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2504         if (vp->v_mount->mnt_flag & MNT_RDONLY)
 2505                 error = EROFS;
 2506         else {
 2507                 if ((error = pledge_chown(p, uid, gid)))
 2508                         goto out;
 2509                 if ((uid != -1 || gid != -1) &&
 2510                     !vnoperm(vp) &&
 2511                     (suser(p) || suid_clear)) {
 2512                         error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
 2513                         if (error)
 2514                                 goto out;
 2515                         mode = vattr.va_mode & ~(VSUID | VSGID);
 2516                         if (mode == vattr.va_mode)
 2517                                 mode = VNOVAL;
 2518                 } else
 2519                         mode = VNOVAL;
 2520                 VATTR_NULL(&vattr);
 2521                 vattr.va_uid = uid;
 2522                 vattr.va_gid = gid;
 2523                 vattr.va_mode = mode;
 2524                 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
 2525         }
 2526 out:
 2527         vput(vp);
 2528         return (error);
 2529 }
 2530 
 2531 /*
 2532  * Set ownership given a file descriptor.
 2533  */
 2534 int
 2535 sys_fchown(struct proc *p, void *v, register_t *retval)
 2536 {
 2537         struct sys_fchown_args /* {
 2538                 syscallarg(int) fd;
 2539                 syscallarg(uid_t) uid;
 2540                 syscallarg(gid_t) gid;
 2541         } */ *uap = v;
 2542         struct vnode *vp;
 2543         struct vattr vattr;
 2544         int error;
 2545         struct file *fp;
 2546         mode_t mode;
 2547         uid_t uid = SCARG(uap, uid);
 2548         gid_t gid = SCARG(uap, gid);
 2549 
 2550         if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
 2551                 return (error);
 2552         vp = fp->f_data;
 2553         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2554         if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_RDONLY))
 2555                 error = EROFS;
 2556         else {
 2557                 if ((error = pledge_chown(p, uid, gid)))
 2558                         goto out;
 2559                 if ((uid != -1 || gid != -1) &&
 2560                     !vnoperm(vp) &&
 2561                     (suser(p) || suid_clear)) {
 2562                         error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
 2563                         if (error)
 2564                                 goto out;
 2565                         mode = vattr.va_mode & ~(VSUID | VSGID);
 2566                         if (mode == vattr.va_mode)
 2567                                 mode = VNOVAL;
 2568                 } else
 2569                         mode = VNOVAL;
 2570                 VATTR_NULL(&vattr);
 2571                 vattr.va_uid = uid;
 2572                 vattr.va_gid = gid;
 2573                 vattr.va_mode = mode;
 2574                 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
 2575         }
 2576 out:
 2577         VOP_UNLOCK(vp);
 2578         FRELE(fp, p);
 2579         return (error);
 2580 }
 2581 
 2582 /*
 2583  * Set the access and modification times given a path name.
 2584  */
 2585 int
 2586 sys_utimes(struct proc *p, void *v, register_t *retval)
 2587 {
 2588         struct sys_utimes_args /* {
 2589                 syscallarg(const char *) path;
 2590                 syscallarg(const struct timeval *) tptr;
 2591         } */ *uap = v;
 2592 
 2593         struct timespec ts[2];
 2594         struct timeval tv[2];
 2595         const struct timeval *tvp;
 2596         int error;
 2597 
 2598         tvp = SCARG(uap, tptr);
 2599         if (tvp != NULL) {
 2600                 error = copyin(tvp, tv, sizeof(tv));
 2601                 if (error)
 2602                         return (error);
 2603 #ifdef KTRACE
 2604                 if (KTRPOINT(p, KTR_STRUCT))
 2605                         ktrabstimeval(p, &tv);
 2606 #endif
 2607                 if (!timerisvalid(&tv[0]) || !timerisvalid(&tv[1]))
 2608                         return (EINVAL);
 2609                 TIMEVAL_TO_TIMESPEC(&tv[0], &ts[0]);
 2610                 TIMEVAL_TO_TIMESPEC(&tv[1], &ts[1]);
 2611         } else
 2612                 ts[0].tv_nsec = ts[1].tv_nsec = UTIME_NOW;
 2613 
 2614         return (doutimensat(p, AT_FDCWD, SCARG(uap, path), ts, 0));
 2615 }
 2616 
 2617 int
 2618 sys_utimensat(struct proc *p, void *v, register_t *retval)
 2619 {
 2620         struct sys_utimensat_args /* {
 2621                 syscallarg(int) fd;
 2622                 syscallarg(const char *) path;
 2623                 syscallarg(const struct timespec *) times;
 2624                 syscallarg(int) flag;
 2625         } */ *uap = v;
 2626 
 2627         struct timespec ts[2];
 2628         const struct timespec *tsp;
 2629         int error, i;
 2630 
 2631         tsp = SCARG(uap, times);
 2632         if (tsp != NULL) {
 2633                 error = copyin(tsp, ts, sizeof(ts));
 2634                 if (error)
 2635                         return (error);
 2636                 for (i = 0; i < nitems(ts); i++) {
 2637                         if (ts[i].tv_nsec == UTIME_NOW)
 2638                                 continue;
 2639                         if (ts[i].tv_nsec == UTIME_OMIT)
 2640                                 continue;
 2641 #ifdef KTRACE
 2642                         if (KTRPOINT(p, KTR_STRUCT))
 2643                                 ktrabstimespec(p, &ts[i]);
 2644 #endif
 2645                         if (!timespecisvalid(&ts[i]))
 2646                                 return (EINVAL);
 2647                 }
 2648         } else
 2649                 ts[0].tv_nsec = ts[1].tv_nsec = UTIME_NOW;
 2650 
 2651         return (doutimensat(p, SCARG(uap, fd), SCARG(uap, path), ts,
 2652             SCARG(uap, flag)));
 2653 }
 2654 
 2655 int
 2656 doutimensat(struct proc *p, int fd, const char *path,
 2657     struct timespec ts[2], int flag)
 2658 {
 2659         struct vnode *vp;
 2660         int error, follow;
 2661         struct nameidata nd;
 2662 
 2663         if (flag & ~AT_SYMLINK_NOFOLLOW)
 2664                 return (EINVAL);
 2665 
 2666         follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
 2667         NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd, path, p);
 2668         nd.ni_pledge = PLEDGE_FATTR | PLEDGE_RPATH;
 2669         nd.ni_unveil = UNVEIL_WRITE;
 2670         if ((error = namei(&nd)) != 0)
 2671                 return (error);
 2672         vp = nd.ni_vp;
 2673 
 2674         return (dovutimens(p, vp, ts));
 2675 }
 2676 
 2677 int
 2678 dovutimens(struct proc *p, struct vnode *vp, struct timespec ts[2])
 2679 {
 2680         struct vattr vattr;
 2681         struct timespec now;
 2682         int error;
 2683 
 2684 #ifdef KTRACE
 2685         /* if they're both UTIME_NOW, then don't report either */
 2686         if ((ts[0].tv_nsec != UTIME_NOW || ts[1].tv_nsec != UTIME_NOW) &&
 2687             KTRPOINT(p, KTR_STRUCT)) {
 2688                 ktrabstimespec(p, &ts[0]);
 2689                 ktrabstimespec(p, &ts[1]);
 2690         }
 2691 #endif
 2692 
 2693         VATTR_NULL(&vattr);
 2694 
 2695         /*  make sure ctime is updated even if neither mtime nor atime is */
 2696         vattr.va_vaflags = VA_UTIMES_CHANGE;
 2697 
 2698         if (ts[0].tv_nsec == UTIME_NOW || ts[1].tv_nsec == UTIME_NOW) {
 2699                 if (ts[0].tv_nsec == UTIME_NOW && ts[1].tv_nsec == UTIME_NOW)
 2700                         vattr.va_vaflags |= VA_UTIMES_NULL;
 2701 
 2702                 getnanotime(&now);
 2703                 if (ts[0].tv_nsec == UTIME_NOW)
 2704                         ts[0] = now;
 2705                 if (ts[1].tv_nsec == UTIME_NOW)
 2706                         ts[1] = now;
 2707         }
 2708 
 2709         if (ts[0].tv_nsec != UTIME_OMIT)
 2710                 vattr.va_atime = ts[0];
 2711         if (ts[1].tv_nsec != UTIME_OMIT)
 2712                 vattr.va_mtime = ts[1];
 2713 
 2714         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2715         if (vp->v_mount->mnt_flag & MNT_RDONLY)
 2716                 error = EROFS;
 2717         else
 2718                 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
 2719         vput(vp);
 2720         return (error);
 2721 }
 2722 
 2723 /*
 2724  * Set the access and modification times given a file descriptor.
 2725  */
 2726 int
 2727 sys_futimes(struct proc *p, void *v, register_t *retval)
 2728 {
 2729         struct sys_futimes_args /* {
 2730                 syscallarg(int) fd;
 2731                 syscallarg(const struct timeval *) tptr;
 2732         } */ *uap = v;
 2733         struct timeval tv[2];
 2734         struct timespec ts[2];
 2735         const struct timeval *tvp;
 2736         int error;
 2737 
 2738         tvp = SCARG(uap, tptr);
 2739         if (tvp != NULL) {
 2740                 error = copyin(tvp, tv, sizeof(tv));
 2741                 if (error)
 2742                         return (error);
 2743 #ifdef KTRACE
 2744                 if (KTRPOINT(p, KTR_STRUCT)) {
 2745                         ktrabstimeval(p, &tv[0]);
 2746                         ktrabstimeval(p, &tv[1]);
 2747                 }
 2748 #endif
 2749                 if (!timerisvalid(&tv[0]) || !timerisvalid(&tv[1]))
 2750                         return (EINVAL);
 2751                 TIMEVAL_TO_TIMESPEC(&tv[0], &ts[0]);
 2752                 TIMEVAL_TO_TIMESPEC(&tv[1], &ts[1]);
 2753         } else
 2754                 ts[0].tv_nsec = ts[1].tv_nsec = UTIME_NOW;
 2755 
 2756         return (dofutimens(p, SCARG(uap, fd), ts));
 2757 }
 2758 
 2759 int
 2760 sys_futimens(struct proc *p, void *v, register_t *retval)
 2761 {
 2762         struct sys_futimens_args /* {
 2763                 syscallarg(int) fd;
 2764                 syscallarg(const struct timespec *) times;
 2765         } */ *uap = v;
 2766         struct timespec ts[2];
 2767         const struct timespec *tsp;
 2768         int error, i;
 2769 
 2770         tsp = SCARG(uap, times);
 2771         if (tsp != NULL) {
 2772                 error = copyin(tsp, ts, sizeof(ts));
 2773                 if (error)
 2774                         return (error);
 2775                 for (i = 0; i < nitems(ts); i++) {
 2776                         if (ts[i].tv_nsec == UTIME_NOW)
 2777                                 continue;
 2778                         if (ts[i].tv_nsec == UTIME_OMIT)
 2779                                 continue;
 2780 #ifdef KTRACE
 2781                         if (KTRPOINT(p, KTR_STRUCT))
 2782                                 ktrabstimespec(p, &ts[i]);
 2783 #endif
 2784                         if (!timespecisvalid(&ts[i]))
 2785                                 return (EINVAL);
 2786                 }
 2787         } else
 2788                 ts[0].tv_nsec = ts[1].tv_nsec = UTIME_NOW;
 2789 
 2790         return (dofutimens(p, SCARG(uap, fd), ts));
 2791 }
 2792 
 2793 int
 2794 dofutimens(struct proc *p, int fd, struct timespec ts[2])
 2795 {
 2796         struct file *fp;
 2797         struct vnode *vp;
 2798         int error;
 2799 
 2800         if ((error = getvnode(p, fd, &fp)) != 0)
 2801                 return (error);
 2802         vp = fp->f_data;
 2803         vref(vp);
 2804         FRELE(fp, p);
 2805 
 2806         return (dovutimens(p, vp, ts));
 2807 }
 2808 
 2809 /*
 2810  * Truncate a file given a vnode.
 2811  */
 2812 int
 2813 dotruncate(struct proc *p, struct vnode *vp, off_t len)
 2814 {
 2815         struct vattr vattr;
 2816         int error;
 2817 
 2818         if (len < 0)
 2819                 return EINVAL;
 2820         if (vp->v_type == VDIR)
 2821                 return EISDIR;
 2822         if ((error = vn_writechk(vp)) != 0)
 2823                 return error;
 2824         if (vp->v_type == VREG && len > lim_cur_proc(p, RLIMIT_FSIZE)) {
 2825                 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
 2826                         return error;
 2827                 if (len > vattr.va_size) {
 2828                         /* if extending over the limit, send signal and fail */
 2829                         psignal(p, SIGXFSZ);
 2830                         return EFBIG;
 2831                 }
 2832         }
 2833         VATTR_NULL(&vattr);
 2834         vattr.va_size = len;
 2835         return VOP_SETATTR(vp, &vattr, p->p_ucred, p);
 2836 }
 2837 
 2838 /*
 2839  * Truncate a file given its path name.
 2840  */
 2841 int
 2842 sys_truncate(struct proc *p, void *v, register_t *retval)
 2843 {
 2844         struct sys_truncate_args /* {
 2845                 syscallarg(const char *) path;
 2846                 syscallarg(off_t) length;
 2847         } */ *uap = v;
 2848         struct vnode *vp;
 2849         int error;
 2850         struct nameidata nd;
 2851 
 2852         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
 2853         nd.ni_pledge = PLEDGE_FATTR | PLEDGE_RPATH;
 2854         nd.ni_unveil = UNVEIL_WRITE;
 2855         if ((error = namei(&nd)) != 0)
 2856                 return (error);
 2857         vp = nd.ni_vp;
 2858         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2859         if ((error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0)
 2860                 error = dotruncate(p, vp, SCARG(uap, length));
 2861         vput(vp);
 2862         return (error);
 2863 }
 2864 
 2865 /*
 2866  * Truncate a file given a file descriptor.
 2867  */
 2868 int
 2869 sys_ftruncate(struct proc *p, void *v, register_t *retval)
 2870 {
 2871         struct sys_ftruncate_args /* {
 2872                 syscallarg(int) fd;
 2873                 syscallarg(off_t) length;
 2874         } */ *uap = v;
 2875         struct vnode *vp;
 2876         struct file *fp;
 2877         int error;
 2878 
 2879         if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
 2880                 return (error);
 2881         if ((fp->f_flag & FWRITE) == 0) {
 2882                 error = EINVAL;
 2883                 goto bad;
 2884         }
 2885         vp = fp->f_data;
 2886         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2887         error = dotruncate(p, vp, SCARG(uap, length));
 2888         VOP_UNLOCK(vp);
 2889 bad:
 2890         FRELE(fp, p);
 2891         return (error);
 2892 }
 2893 
 2894 #if 1
 2895 int
 2896 sys_pad_truncate(struct proc *p, void *v, register_t *retval)
 2897 {
 2898         struct sys_pad_truncate_args *uap = v;
 2899         struct sys_truncate_args unpad;
 2900 
 2901         SCARG(&unpad, path) = SCARG(uap, path);
 2902         SCARG(&unpad, length) = SCARG(uap, length);
 2903         return sys_truncate(p, &unpad, retval);
 2904 }
 2905 
 2906 int
 2907 sys_pad_ftruncate(struct proc *p, void *v, register_t *retval)
 2908 {
 2909         struct sys_pad_ftruncate_args *uap = v;
 2910         struct sys_ftruncate_args unpad;
 2911 
 2912         SCARG(&unpad, fd) = SCARG(uap, fd);
 2913         SCARG(&unpad, length) = SCARG(uap, length);
 2914         return sys_ftruncate(p, &unpad, retval);
 2915 }
 2916 #endif
 2917 
 2918 /*
 2919  * Sync an open file.
 2920  */
 2921 int
 2922 sys_fsync(struct proc *p, void *v, register_t *retval)
 2923 {
 2924         struct sys_fsync_args /* {
 2925                 syscallarg(int) fd;
 2926         } */ *uap = v;
 2927         struct vnode *vp;
 2928         struct file *fp;
 2929         int error;
 2930 
 2931         if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
 2932                 return (error);
 2933         vp = fp->f_data;
 2934         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 2935         error = VOP_FSYNC(vp, fp->f_cred, MNT_WAIT, p);
 2936 #ifdef FFS_SOFTUPDATES
 2937         if (error == 0 && vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
 2938                 error = softdep_fsync(vp);
 2939 #endif
 2940 
 2941         VOP_UNLOCK(vp);
 2942         FRELE(fp, p);
 2943         return (error);
 2944 }
 2945 
 2946 /*
 2947  * Rename files.  Source and destination must either both be directories,
 2948  * or both not be directories.  If target is a directory, it must be empty.
 2949  */
 2950 int
 2951 sys_rename(struct proc *p, void *v, register_t *retval)
 2952 {
 2953         struct sys_rename_args /* {
 2954                 syscallarg(const char *) from;
 2955                 syscallarg(const char *) to;
 2956         } */ *uap = v;
 2957 
 2958         return (dorenameat(p, AT_FDCWD, SCARG(uap, from), AT_FDCWD,
 2959             SCARG(uap, to)));
 2960 }
 2961 
 2962 int
 2963 sys_renameat(struct proc *p, void *v, register_t *retval)
 2964 {
 2965         struct sys_renameat_args /* {
 2966                 syscallarg(int) fromfd;
 2967                 syscallarg(const char *) from;
 2968                 syscallarg(int) tofd;
 2969                 syscallarg(const char *) to;
 2970         } */ *uap = v;
 2971 
 2972         return (dorenameat(p, SCARG(uap, fromfd), SCARG(uap, from),
 2973             SCARG(uap, tofd), SCARG(uap, to)));
 2974 }
 2975 
 2976 int
 2977 dorenameat(struct proc *p, int fromfd, const char *from, int tofd,
 2978     const char *to)
 2979 {
 2980         struct vnode *tvp, *fvp, *tdvp;
 2981         struct nameidata fromnd, tond;
 2982         int error;
 2983         int flags;
 2984 
 2985         NDINITAT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
 2986             fromfd, from, p);
 2987         fromnd.ni_pledge = PLEDGE_RPATH | PLEDGE_CPATH;
 2988         fromnd.ni_unveil = UNVEIL_READ | UNVEIL_CREATE;
 2989         if ((error = namei(&fromnd)) != 0)
 2990                 return (error);
 2991         fvp = fromnd.ni_vp;
 2992 
 2993         flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
 2994         /*
 2995          * rename("foo/", "bar/");  is  OK
 2996          */
 2997         if (fvp->v_type == VDIR)
 2998                 flags |= STRIPSLASHES;
 2999 
 3000         NDINITAT(&tond, RENAME, flags, UIO_USERSPACE, tofd, to, p);
 3001         tond.ni_pledge = PLEDGE_CPATH;
 3002         tond.ni_unveil = UNVEIL_CREATE;
 3003         if ((error = namei(&tond)) != 0) {
 3004                 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
 3005                 vrele(fromnd.ni_dvp);
 3006                 vrele(fvp);
 3007                 goto out1;
 3008         }
 3009         tdvp = tond.ni_dvp;
 3010         tvp = tond.ni_vp;
 3011         if (tvp != NULL) {
 3012                 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
 3013                         error = ENOTDIR;
 3014                         goto out;
 3015                 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
 3016                         error = EISDIR;
 3017                         goto out;
 3018                 }
 3019         }
 3020         if (fvp == tdvp)
 3021                 error = EINVAL;
 3022         /*
 3023          * If source is the same as the destination (that is the
 3024          * same inode number)
 3025          */
 3026         if (fvp == tvp)
 3027                 error = -1;
 3028 out:
 3029         if (!error) {
 3030                 if (tvp) {
 3031                         (void)uvm_vnp_uncache(tvp);
 3032                 }
 3033                 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
 3034                                    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
 3035         } else {
 3036                 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
 3037                 if (tdvp == tvp)
 3038                         vrele(tdvp);
 3039                 else
 3040                         vput(tdvp);
 3041                 if (tvp)
 3042                         vput(tvp);
 3043                 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
 3044                 vrele(fromnd.ni_dvp);
 3045                 vrele(fvp);
 3046         }
 3047         vrele(tond.ni_startdir);
 3048         pool_put(&namei_pool, tond.ni_cnd.cn_pnbuf);
 3049 out1:
 3050         if (fromnd.ni_startdir)
 3051                 vrele(fromnd.ni_startdir);
 3052         pool_put(&namei_pool, fromnd.ni_cnd.cn_pnbuf);
 3053         if (error == -1)
 3054                 return (0);
 3055         return (error);
 3056 }
 3057 
 3058 /*
 3059  * Make a directory file.
 3060  */
 3061 int
 3062 sys_mkdir(struct proc *p, void *v, register_t *retval)
 3063 {
 3064         struct sys_mkdir_args /* {
 3065                 syscallarg(const char *) path;
 3066                 syscallarg(mode_t) mode;
 3067         } */ *uap = v;
 3068 
 3069         return (domkdirat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)));
 3070 }
 3071 
 3072 int
 3073 sys_mkdirat(struct proc *p, void *v, register_t *retval)
 3074 {
 3075         struct sys_mkdirat_args /* {
 3076                 syscallarg(int) fd;
 3077                 syscallarg(const char *) path;
 3078                 syscallarg(mode_t) mode;
 3079         } */ *uap = v;
 3080 
 3081         return (domkdirat(p, SCARG(uap, fd), SCARG(uap, path),
 3082             SCARG(uap, mode)));
 3083 }
 3084 
 3085 int
 3086 domkdirat(struct proc *p, int fd, const char *path, mode_t mode)
 3087 {
 3088         struct vnode *vp;
 3089         struct vattr vattr;
 3090         int error;
 3091         struct nameidata nd;
 3092 
 3093         NDINITAT(&nd, CREATE, LOCKPARENT | STRIPSLASHES, UIO_USERSPACE,
 3094             fd, path, p);
 3095         nd.ni_pledge = PLEDGE_CPATH;
 3096         nd.ni_unveil = UNVEIL_CREATE;
 3097         if ((error = namei(&nd)) != 0)
 3098                 return (error);
 3099         vp = nd.ni_vp;
 3100         if (vp != NULL) {
 3101                 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
 3102                 if (nd.ni_dvp == vp)
 3103                         vrele(nd.ni_dvp);
 3104                 else
 3105                         vput(nd.ni_dvp);
 3106                 vrele(vp);
 3107                 return (EEXIST);
 3108         }
 3109         VATTR_NULL(&vattr);
 3110         vattr.va_type = VDIR;
 3111         vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
 3112         error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
 3113         if (!error)
 3114                 vput(nd.ni_vp);
 3115         return (error);
 3116 }
 3117 
 3118 /*
 3119  * Remove a directory file.
 3120  */
 3121 int
 3122 sys_rmdir(struct proc *p, void *v, register_t *retval)
 3123 {
 3124         struct sys_rmdir_args /* {
 3125                 syscallarg(const char *) path;
 3126         } */ *uap = v;
 3127 
 3128         return (dounlinkat(p, AT_FDCWD, SCARG(uap, path), AT_REMOVEDIR));
 3129 }
 3130 
 3131 /*
 3132  * Read a block of directory entries in a file system independent format.
 3133  */
 3134 int
 3135 sys_getdents(struct proc *p, void *v, register_t *retval)
 3136 {
 3137         struct sys_getdents_args /* {
 3138                 syscallarg(int) fd;
 3139                 syscallarg(void *) buf;
 3140                 syscallarg(size_t) buflen;
 3141         } */ *uap = v;
 3142         struct vnode *vp;
 3143         struct file *fp;
 3144         struct uio auio;
 3145         struct iovec aiov;
 3146         size_t buflen;
 3147         int error, eofflag;
 3148 
 3149         buflen = SCARG(uap, buflen);
 3150 
 3151         if (buflen > INT_MAX)
 3152                 return (EINVAL);
 3153         if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
 3154                 return (error);
 3155         if ((fp->f_flag & FREAD) == 0) {
 3156                 error = EBADF;
 3157                 goto bad;
 3158         }
 3159         vp = fp->f_data;
 3160         if (vp->v_type != VDIR) {
 3161                 error = EINVAL;
 3162                 goto bad;
 3163         }
 3164 
 3165         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 3166 
 3167         if (fp->f_offset < 0) {
 3168                 VOP_UNLOCK(vp);
 3169                 error = EINVAL;
 3170                 goto bad;
 3171         }
 3172 
 3173         aiov.iov_base = SCARG(uap, buf);
 3174         aiov.iov_len = buflen;
 3175         auio.uio_iov = &aiov;
 3176         auio.uio_iovcnt = 1;
 3177         auio.uio_rw = UIO_READ;
 3178         auio.uio_segflg = UIO_USERSPACE;
 3179         auio.uio_procp = p;
 3180         auio.uio_resid = buflen;
 3181         auio.uio_offset = fp->f_offset;
 3182         error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag);
 3183         mtx_enter(&fp->f_mtx);
 3184         fp->f_offset = auio.uio_offset;
 3185         mtx_leave(&fp->f_mtx);
 3186         VOP_UNLOCK(vp);
 3187         if (error)
 3188                 goto bad;
 3189         *retval = buflen - auio.uio_resid;
 3190 bad:
 3191         FRELE(fp, p);
 3192         return (error);
 3193 }
 3194 
 3195 /*
 3196  * Set the mode mask for creation of filesystem nodes.
 3197  */
 3198 int
 3199 sys_umask(struct proc *p, void *v, register_t *retval)
 3200 {
 3201         struct sys_umask_args /* {
 3202                 syscallarg(mode_t) newmask;
 3203         } */ *uap = v;
 3204         struct filedesc *fdp = p->p_fd;
 3205 
 3206         fdplock(fdp);
 3207         *retval = fdp->fd_cmask;
 3208         fdp->fd_cmask = SCARG(uap, newmask) & ACCESSPERMS;
 3209         fdpunlock(fdp);
 3210         return (0);
 3211 }
 3212 
 3213 /*
 3214  * Void all references to file by ripping underlying filesystem
 3215  * away from vnode.
 3216  */
 3217 int
 3218 sys_revoke(struct proc *p, void *v, register_t *retval)
 3219 {
 3220         struct sys_revoke_args /* {
 3221                 syscallarg(const char *) path;
 3222         } */ *uap = v;
 3223         struct vnode *vp;
 3224         struct vattr vattr;
 3225         int error;
 3226         struct nameidata nd;
 3227 
 3228         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
 3229         nd.ni_pledge = PLEDGE_RPATH | PLEDGE_TTY;
 3230         nd.ni_unveil = UNVEIL_READ;
 3231         if ((error = namei(&nd)) != 0)
 3232                 return (error);
 3233         vp = nd.ni_vp;
 3234         if (vp->v_type != VCHR || (u_int)major(vp->v_rdev) >= nchrdev ||
 3235             cdevsw[major(vp->v_rdev)].d_type != D_TTY) {
 3236                 error = ENOTTY;
 3237                 goto out;
 3238         }
 3239         if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
 3240                 goto out;
 3241         if (p->p_ucred->cr_uid != vattr.va_uid &&
 3242             (error = suser(p)))
 3243                 goto out;
 3244         if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED)))
 3245                 VOP_REVOKE(vp, REVOKEALL);
 3246 out:
 3247         vrele(vp);
 3248         return (error);
 3249 }
 3250 
 3251 /*
 3252  * Convert a user file descriptor to a kernel file entry.
 3253  *
 3254  * On return *fpp is FREF:ed.
 3255  */
 3256 int
 3257 getvnode(struct proc *p, int fd, struct file **fpp)
 3258 {
 3259         struct file *fp;
 3260         struct vnode *vp;
 3261 
 3262         if ((fp = fd_getfile(p->p_fd, fd)) == NULL)
 3263                 return (EBADF);
 3264 
 3265         if (fp->f_type != DTYPE_VNODE) {
 3266                 FRELE(fp, p);
 3267                 return (EINVAL);
 3268         }
 3269 
 3270         vp = fp->f_data;
 3271         if (vp->v_type == VBAD) {
 3272                 FRELE(fp, p);
 3273                 return (EBADF);
 3274         }
 3275 
 3276         *fpp = fp;
 3277 
 3278         return (0);
 3279 }
 3280 
 3281 /*
 3282  * Positional read system call.
 3283  */
 3284 int
 3285 sys_pread(struct proc *p, void *v, register_t *retval)
 3286 {
 3287         struct sys_pread_args /* {
 3288                 syscallarg(int) fd;
 3289                 syscallarg(void *) buf;
 3290                 syscallarg(size_t) nbyte;
 3291                 syscallarg(off_t) offset;
 3292         } */ *uap = v;
 3293         struct iovec iov;
 3294         struct uio auio;
 3295 
 3296         iov.iov_base = SCARG(uap, buf);
 3297         iov.iov_len = SCARG(uap, nbyte);
 3298         if (iov.iov_len > SSIZE_MAX)
 3299                 return (EINVAL);
 3300 
 3301         auio.uio_iov = &iov;
 3302         auio.uio_iovcnt = 1;
 3303         auio.uio_resid = iov.iov_len;
 3304         auio.uio_offset = SCARG(uap, offset);
 3305 
 3306         return (dofilereadv(p, SCARG(uap, fd), &auio, FO_POSITION, retval));
 3307 }
 3308 
 3309 /*
 3310  * Positional scatter read system call.
 3311  */
 3312 int
 3313 sys_preadv(struct proc *p, void *v, register_t *retval)
 3314 {
 3315         struct sys_preadv_args /* {
 3316                 syscallarg(int) fd;
 3317                 syscallarg(const struct iovec *) iovp;
 3318                 syscallarg(int) iovcnt;
 3319                 syscallarg(off_t) offset;
 3320         } */ *uap = v;
 3321         struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
 3322         int error, iovcnt = SCARG(uap, iovcnt);
 3323         struct uio auio;
 3324         size_t resid;
 3325 
 3326         error = iovec_copyin(SCARG(uap, iovp), &iov, aiov, iovcnt, &resid);
 3327         if (error)
 3328                 goto done;
 3329 
 3330         auio.uio_iov = iov;
 3331         auio.uio_iovcnt = iovcnt;
 3332         auio.uio_resid = resid;
 3333         auio.uio_offset = SCARG(uap, offset);
 3334 
 3335         error = dofilereadv(p, SCARG(uap, fd), &auio, FO_POSITION, retval);
 3336  done:
 3337         iovec_free(iov, iovcnt);
 3338         return (error);
 3339 }
 3340 
 3341 /*
 3342  * Positional write system call.
 3343  */
 3344 int
 3345 sys_pwrite(struct proc *p, void *v, register_t *retval)
 3346 {
 3347         struct sys_pwrite_args /* {
 3348                 syscallarg(int) fd;
 3349                 syscallarg(const void *) buf;
 3350                 syscallarg(size_t) nbyte;
 3351                 syscallarg(off_t) offset;
 3352         } */ *uap = v;
 3353         struct iovec iov;
 3354         struct uio auio;
 3355 
 3356         iov.iov_base = (void *)SCARG(uap, buf);
 3357         iov.iov_len = SCARG(uap, nbyte);
 3358         if (iov.iov_len > SSIZE_MAX)
 3359                 return (EINVAL);
 3360 
 3361         auio.uio_iov = &iov;
 3362         auio.uio_iovcnt = 1;
 3363         auio.uio_resid = iov.iov_len;
 3364         auio.uio_offset = SCARG(uap, offset);
 3365 
 3366         return (dofilewritev(p, SCARG(uap, fd), &auio, FO_POSITION, retval));
 3367 }
 3368 
 3369 /*
 3370  * Positional gather write system call.
 3371  */
 3372 int
 3373 sys_pwritev(struct proc *p, void *v, register_t *retval)
 3374 {
 3375         struct sys_pwritev_args /* {
 3376                 syscallarg(int) fd;
 3377                 syscallarg(const struct iovec *) iovp;
 3378                 syscallarg(int) iovcnt;
 3379                 syscallarg(off_t) offset;
 3380         } */ *uap = v;
 3381         struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
 3382         int error, iovcnt = SCARG(uap, iovcnt);
 3383         struct uio auio;
 3384         size_t resid;
 3385 
 3386         error = iovec_copyin(SCARG(uap, iovp), &iov, aiov, iovcnt, &resid);
 3387         if (error)
 3388                 goto done;
 3389 
 3390         auio.uio_iov = iov;
 3391         auio.uio_iovcnt = iovcnt;
 3392         auio.uio_resid = resid;
 3393         auio.uio_offset = SCARG(uap, offset);
 3394 
 3395         error = dofilewritev(p, SCARG(uap, fd), &auio, FO_POSITION, retval);
 3396  done:
 3397         iovec_free(iov, iovcnt);
 3398         return (error);
 3399 }
 3400 
 3401 #if 1
 3402 int
 3403 sys_pad_pread(struct proc *p, void *v, register_t *retval)
 3404 {
 3405         struct sys_pad_pread_args *uap = v;
 3406         struct sys_pread_args unpad;
 3407 
 3408         SCARG(&unpad, fd) = SCARG(uap, fd);
 3409         SCARG(&unpad, buf) = SCARG(uap, buf);
 3410         SCARG(&unpad, nbyte) = SCARG(uap, nbyte);
 3411         SCARG(&unpad, offset) = SCARG(uap, offset);
 3412         return sys_pread(p, &unpad, retval);
 3413 }
 3414 
 3415 int
 3416 sys_pad_preadv(struct proc *p, void *v, register_t *retval)
 3417 {
 3418         struct sys_pad_preadv_args *uap = v;
 3419         struct sys_preadv_args unpad;
 3420 
 3421         SCARG(&unpad, fd) = SCARG(uap, fd);
 3422         SCARG(&unpad, iovp) = SCARG(uap, iovp);
 3423         SCARG(&unpad, iovcnt) = SCARG(uap, iovcnt);
 3424         SCARG(&unpad, offset) = SCARG(uap, offset);
 3425         return sys_preadv(p, &unpad, retval);
 3426 }
 3427 
 3428 int
 3429 sys_pad_pwrite(struct proc *p, void *v, register_t *retval)
 3430 {
 3431         struct sys_pad_pwrite_args *uap = v;
 3432         struct sys_pwrite_args unpad;
 3433 
 3434         SCARG(&unpad, fd) = SCARG(uap, fd);
 3435         SCARG(&unpad, buf) = SCARG(uap, buf);
 3436         SCARG(&unpad, nbyte) = SCARG(uap, nbyte);
 3437         SCARG(&unpad, offset) = SCARG(uap, offset);
 3438         return sys_pwrite(p, &unpad, retval);
 3439 }
 3440 
 3441 int
 3442 sys_pad_pwritev(struct proc *p, void *v, register_t *retval)
 3443 {
 3444         struct sys_pad_pwritev_args *uap = v;
 3445         struct sys_pwritev_args unpad;
 3446 
 3447         SCARG(&unpad, fd) = SCARG(uap, fd);
 3448         SCARG(&unpad, iovp) = SCARG(uap, iovp);
 3449         SCARG(&unpad, iovcnt) = SCARG(uap, iovcnt);
 3450         SCARG(&unpad, offset) = SCARG(uap, offset);
 3451         return sys_pwritev(p, &unpad, retval);
 3452 }
 3453 #endif
Cache object: 200d5e9eb734aefd46cee12d10114a7c
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.
FreeBSD/Linux Kernel Cross Reference sys/kern/vfs_syscalls.c

FreeBSD/Linux Kernel Cross Reference
sys/kern/vfs_syscalls.c
