The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

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

Version: -  FREEBSD  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-2  -  FREEBSD-11-1  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-4  -  FREEBSD-10-3  -  FREEBSD-10-2  -  FREEBSD-10-1  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-3  -  FREEBSD-9-2  -  FREEBSD-9-1  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-4  -  FREEBSD-8-3  -  FREEBSD-8-2  -  FREEBSD-8-1  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-4  -  FREEBSD-7-3  -  FREEBSD-7-2  -  FREEBSD-7-1  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-4  -  FREEBSD-6-3  -  FREEBSD-6-2  -  FREEBSD-6-1  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-5  -  FREEBSD-5-4  -  FREEBSD-5-3  -  FREEBSD-5-2  -  FREEBSD-5-1  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1989, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  * (c) UNIX System Laboratories, Inc.
    5  * All or some portions of this file are derived from material licensed
    6  * to the University of California by American Telephone and Telegraph
    7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
    8  * the permission of UNIX System Laboratories, Inc.
    9  *
   10  * Redistribution and use in source and binary forms, with or without
   11  * modification, are permitted provided that the following conditions
   12  * are met:
   13  * 1. Redistributions of source code must retain the above copyright
   14  *    notice, this list of conditions and the following disclaimer.
   15  * 2. Redistributions in binary form must reproduce the above copyright
   16  *    notice, this list of conditions and the following disclaimer in the
   17  *    documentation and/or other materials provided with the distribution.
   18  * 4. Neither the name of the University nor the names of its contributors
   19  *    may be used to endorse or promote products derived from this software
   20  *    without specific prior written permission.
   21  *
   22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   32  * SUCH DAMAGE.
   33  *
   34  *      @(#)vfs_syscalls.c      8.13 (Berkeley) 4/15/94
   35  */
   36 
   37 #include <sys/cdefs.h>
   38 __FBSDID("$FreeBSD: releng/6.0/sys/kern/vfs_syscalls.c 150271 2005-09-18 03:31:35Z csjp $");
   39 
   40 #include "opt_compat.h"
   41 #include "opt_mac.h"
   42 
   43 #include <sys/param.h>
   44 #include <sys/systm.h>
   45 #include <sys/bio.h>
   46 #include <sys/buf.h>
   47 #include <sys/sysent.h>
   48 #include <sys/mac.h>
   49 #include <sys/malloc.h>
   50 #include <sys/mount.h>
   51 #include <sys/mutex.h>
   52 #include <sys/sysproto.h>
   53 #include <sys/namei.h>
   54 #include <sys/filedesc.h>
   55 #include <sys/kernel.h>
   56 #include <sys/fcntl.h>
   57 #include <sys/file.h>
   58 #include <sys/limits.h>
   59 #include <sys/linker.h>
   60 #include <sys/stat.h>
   61 #include <sys/sx.h>
   62 #include <sys/unistd.h>
   63 #include <sys/vnode.h>
   64 #include <sys/proc.h>
   65 #include <sys/dirent.h>
   66 #include <sys/extattr.h>
   67 #include <sys/jail.h>
   68 #include <sys/syscallsubr.h>
   69 #include <sys/sysctl.h>
   70 
   71 #include <machine/stdarg.h>
   72 
   73 #include <vm/vm.h>
   74 #include <vm/vm_object.h>
   75 #include <vm/vm_page.h>
   76 #include <vm/uma.h>
   77 
   78 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
   79 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
   80 static int setfown(struct thread *td, struct vnode *, uid_t, gid_t);
   81 static int setfmode(struct thread *td, struct vnode *, int);
   82 static int setfflags(struct thread *td, struct vnode *, int);
   83 static int setutimes(struct thread *td, struct vnode *,
   84     const struct timespec *, int, int);
   85 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
   86     struct thread *td);
   87 
   88 static int extattr_list_vp(struct vnode *vp, int attrnamespace, void *data,
   89     size_t nbytes, struct thread *td);
   90 
   91 int (*union_dircheckp)(struct thread *td, struct vnode **, struct file *);
   92 
   93 /*
   94  * The module initialization routine for POSIX asynchronous I/O will
   95  * set this to the version of AIO that it implements.  (Zero means
   96  * that it is not implemented.)  This value is used here by pathconf()
   97  * and in kern_descrip.c by fpathconf().
   98  */
   99 int async_io_version;
  100 
  101 /*
  102  * Sync each mounted filesystem.
  103  */
  104 #ifndef _SYS_SYSPROTO_H_
  105 struct sync_args {
  106         int     dummy;
  107 };
  108 #endif
  109 
  110 #ifdef DEBUG
  111 static int syncprt = 0;
  112 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
  113 #endif
  114 
  115 /* ARGSUSED */
  116 int
  117 sync(td, uap)
  118         struct thread *td;
  119         struct sync_args *uap;
  120 {
  121         struct mount *mp, *nmp;
  122         int asyncflag;
  123 
  124         mtx_lock(&Giant);
  125         mtx_lock(&mountlist_mtx);
  126         for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
  127                 if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td)) {
  128                         nmp = TAILQ_NEXT(mp, mnt_list);
  129                         continue;
  130                 }
  131                 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
  132                     vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
  133                         asyncflag = mp->mnt_flag & MNT_ASYNC;
  134                         mp->mnt_flag &= ~MNT_ASYNC;
  135                         vfs_msync(mp, MNT_NOWAIT);
  136                         VFS_SYNC(mp, MNT_NOWAIT, td);
  137                         mp->mnt_flag |= asyncflag;
  138                         vn_finished_write(mp);
  139                 }
  140                 mtx_lock(&mountlist_mtx);
  141                 nmp = TAILQ_NEXT(mp, mnt_list);
  142                 vfs_unbusy(mp, td);
  143         }
  144         mtx_unlock(&mountlist_mtx);
  145 #if 0
  146 /*
  147  * XXX don't call vfs_bufstats() yet because that routine
  148  * was not imported in the Lite2 merge.
  149  */
  150 #ifdef DIAGNOSTIC
  151         if (syncprt)
  152                 vfs_bufstats();
  153 #endif /* DIAGNOSTIC */
  154 #endif
  155         mtx_unlock(&Giant);
  156         return (0);
  157 }
  158 
  159 /* XXX PRISON: could be per prison flag */
  160 static int prison_quotas;
  161 #if 0
  162 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
  163 #endif
  164 
  165 /*
  166  * Change filesystem quotas.
  167  *
  168  * MP SAFE
  169  */
  170 #ifndef _SYS_SYSPROTO_H_
  171 struct quotactl_args {
  172         char *path;
  173         int cmd;
  174         int uid;
  175         caddr_t arg;
  176 };
  177 #endif
  178 int
  179 quotactl(td, uap)
  180         struct thread *td;
  181         register struct quotactl_args /* {
  182                 char *path;
  183                 int cmd;
  184                 int uid;
  185                 caddr_t arg;
  186         } */ *uap;
  187 {
  188         struct mount *mp, *vmp;
  189         int error;
  190         struct nameidata nd;
  191 
  192         if (jailed(td->td_ucred) && !prison_quotas)
  193                 return (EPERM);
  194         mtx_lock(&Giant);
  195         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
  196         if ((error = namei(&nd)) != 0) {
  197                 mtx_unlock(&Giant);
  198                 return (error);
  199         }
  200         NDFREE(&nd, NDF_ONLY_PNBUF);
  201         error = vn_start_write(nd.ni_vp, &vmp, V_WAIT | PCATCH);
  202         mp = nd.ni_vp->v_mount;
  203         vrele(nd.ni_vp);
  204         if (error) {
  205                 mtx_unlock(&Giant);
  206                 return (error);
  207         }
  208         error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, td);
  209         vn_finished_write(vmp);
  210         mtx_unlock(&Giant);
  211         return (error);
  212 }
  213 
  214 /*
  215  * Get filesystem statistics.
  216  */
  217 #ifndef _SYS_SYSPROTO_H_
  218 struct statfs_args {
  219         char *path;
  220         struct statfs *buf;
  221 };
  222 #endif
  223 int
  224 statfs(td, uap)
  225         struct thread *td;
  226         register struct statfs_args /* {
  227                 char *path;
  228                 struct statfs *buf;
  229         } */ *uap;
  230 {
  231         struct statfs sf;
  232         int error;
  233 
  234         error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
  235         if (error == 0)
  236                 error = copyout(&sf, uap->buf, sizeof(sf));
  237         return (error);
  238 }
  239 
  240 int
  241 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
  242     struct statfs *buf)
  243 {
  244         struct mount *mp;
  245         struct statfs *sp, sb;
  246         int error;
  247         struct nameidata nd;
  248 
  249         mtx_lock(&Giant);
  250         NDINIT(&nd, LOOKUP, FOLLOW, pathseg, path, td);
  251         error = namei(&nd);
  252         if (error) {
  253                 mtx_unlock(&Giant);
  254                 return (error);
  255         }
  256         mp = nd.ni_vp->v_mount;
  257         sp = &mp->mnt_stat;
  258         NDFREE(&nd, NDF_ONLY_PNBUF);
  259         vrele(nd.ni_vp);
  260 #ifdef MAC
  261         error = mac_check_mount_stat(td->td_ucred, mp);
  262         if (error) {
  263                 mtx_unlock(&Giant);
  264                 return (error);
  265         }
  266 #endif
  267         /*
  268          * Set these in case the underlying filesystem fails to do so.
  269          */
  270         sp->f_version = STATFS_VERSION;
  271         sp->f_namemax = NAME_MAX;
  272         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
  273         error = VFS_STATFS(mp, sp, td);
  274         if (error) {
  275                 mtx_unlock(&Giant);
  276                 return (error);
  277         }
  278         if (suser(td)) {
  279                 bcopy(sp, &sb, sizeof(sb));
  280                 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
  281                 prison_enforce_statfs(td->td_ucred, mp, &sb);
  282                 sp = &sb;
  283         }
  284         mtx_unlock(&Giant);
  285         *buf = *sp;
  286         return (0);
  287 }
  288 
  289 /*
  290  * Get filesystem statistics.
  291  */
  292 #ifndef _SYS_SYSPROTO_H_
  293 struct fstatfs_args {
  294         int fd;
  295         struct statfs *buf;
  296 };
  297 #endif
  298 int
  299 fstatfs(td, uap)
  300         struct thread *td;
  301         register struct fstatfs_args /* {
  302                 int fd;
  303                 struct statfs *buf;
  304         } */ *uap;
  305 {
  306         struct statfs sf;
  307         int error;
  308 
  309         error = kern_fstatfs(td, uap->fd, &sf);
  310         if (error == 0)
  311                 error = copyout(&sf, uap->buf, sizeof(sf));
  312         return (error);
  313 }
  314 
  315 int
  316 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
  317 {
  318         struct file *fp;
  319         struct mount *mp;
  320         struct statfs *sp, sb;
  321         int error;
  322 
  323         error = getvnode(td->td_proc->p_fd, fd, &fp);
  324         if (error)
  325                 return (error);
  326         mtx_lock(&Giant);
  327         mp = fp->f_vnode->v_mount;
  328         fdrop(fp, td);
  329         if (mp == NULL) {
  330                 mtx_unlock(&Giant);
  331                 return (EBADF);
  332         }
  333 #ifdef MAC
  334         error = mac_check_mount_stat(td->td_ucred, mp);
  335         if (error) {
  336                 mtx_unlock(&Giant);
  337                 return (error);
  338         }
  339 #endif
  340         sp = &mp->mnt_stat;
  341         /*
  342          * Set these in case the underlying filesystem fails to do so.
  343          */
  344         sp->f_version = STATFS_VERSION;
  345         sp->f_namemax = NAME_MAX;
  346         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
  347         error = VFS_STATFS(mp, sp, td);
  348         if (error) {
  349                 mtx_unlock(&Giant);
  350                 return (error);
  351         }
  352         if (suser(td)) {
  353                 bcopy(sp, &sb, sizeof(sb));
  354                 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
  355                 prison_enforce_statfs(td->td_ucred, mp, &sb);
  356                 sp = &sb;
  357         }
  358         mtx_unlock(&Giant);
  359         *buf = *sp;
  360         return (0);
  361 }
  362 
  363 /*
  364  * Get statistics on all filesystems.
  365  */
  366 #ifndef _SYS_SYSPROTO_H_
  367 struct getfsstat_args {
  368         struct statfs *buf;
  369         long bufsize;
  370         int flags;
  371 };
  372 #endif
  373 int
  374 getfsstat(td, uap)
  375         struct thread *td;
  376         register struct getfsstat_args /* {
  377                 struct statfs *buf;
  378                 long bufsize;
  379                 int flags;
  380         } */ *uap;
  381 {
  382 
  383         return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
  384             uap->flags));
  385 }
  386 
  387 /*
  388  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
  389  *      The caller is responsible for freeing memory which will be allocated
  390  *      in '*buf'.
  391  */
  392 int
  393 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
  394     enum uio_seg bufseg, int flags)
  395 {
  396         struct mount *mp, *nmp;
  397         struct statfs *sfsp, *sp, sb;
  398         size_t count, maxcount;
  399         int error;
  400 
  401         maxcount = bufsize / sizeof(struct statfs);
  402         if (bufsize == 0)
  403                 sfsp = NULL;
  404         else if (bufseg == UIO_USERSPACE)
  405                 sfsp = *buf;
  406         else /* if (bufseg == UIO_SYSSPACE) */ {
  407                 count = 0;
  408                 mtx_lock(&mountlist_mtx);
  409                 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
  410                         count++;
  411                 }
  412                 mtx_unlock(&mountlist_mtx);
  413                 if (maxcount > count)
  414                         maxcount = count;
  415                 sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
  416                     M_WAITOK);
  417         }
  418         count = 0;
  419         mtx_lock(&Giant);
  420         mtx_lock(&mountlist_mtx);
  421         for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
  422                 if (prison_canseemount(td->td_ucred, mp) != 0) {
  423                         nmp = TAILQ_NEXT(mp, mnt_list);
  424                         continue;
  425                 }
  426 #ifdef MAC
  427                 if (mac_check_mount_stat(td->td_ucred, mp) != 0) {
  428                         nmp = TAILQ_NEXT(mp, mnt_list);
  429                         continue;
  430                 }
  431 #endif
  432                 if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td)) {
  433                         nmp = TAILQ_NEXT(mp, mnt_list);
  434                         continue;
  435                 }
  436                 if (sfsp && count < maxcount) {
  437                         sp = &mp->mnt_stat;
  438                         /*
  439                          * Set these in case the underlying filesystem
  440                          * fails to do so.
  441                          */
  442                         sp->f_version = STATFS_VERSION;
  443                         sp->f_namemax = NAME_MAX;
  444                         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
  445                         /*
  446                          * If MNT_NOWAIT or MNT_LAZY is specified, do not
  447                          * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
  448                          * overrides MNT_WAIT.
  449                          */
  450                         if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
  451                             (flags & MNT_WAIT)) &&
  452                             (error = VFS_STATFS(mp, sp, td))) {
  453                                 mtx_lock(&mountlist_mtx);
  454                                 nmp = TAILQ_NEXT(mp, mnt_list);
  455                                 vfs_unbusy(mp, td);
  456                                 continue;
  457                         }
  458                         if (suser(td)) {
  459                                 bcopy(sp, &sb, sizeof(sb));
  460                                 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
  461                                 prison_enforce_statfs(td->td_ucred, mp, &sb);
  462                                 sp = &sb;
  463                         }
  464                         if (bufseg == UIO_SYSSPACE)
  465                                 bcopy(sp, sfsp, sizeof(*sp));
  466                         else /* if (bufseg == UIO_USERSPACE) */ {
  467                                 error = copyout(sp, sfsp, sizeof(*sp));
  468                                 if (error) {
  469                                         vfs_unbusy(mp, td);
  470                                         mtx_unlock(&Giant);
  471                                         return (error);
  472                                 }
  473                         }
  474                         sfsp++;
  475                 }
  476                 count++;
  477                 mtx_lock(&mountlist_mtx);
  478                 nmp = TAILQ_NEXT(mp, mnt_list);
  479                 vfs_unbusy(mp, td);
  480         }
  481         mtx_unlock(&mountlist_mtx);
  482         mtx_unlock(&Giant);
  483         if (sfsp && count > maxcount)
  484                 td->td_retval[0] = maxcount;
  485         else
  486                 td->td_retval[0] = count;
  487         return (0);
  488 }
  489 
  490 #ifdef COMPAT_FREEBSD4
  491 /*
  492  * Get old format filesystem statistics.
  493  */
  494 static void cvtstatfs(struct statfs *, struct ostatfs *);
  495 
  496 #ifndef _SYS_SYSPROTO_H_
  497 struct freebsd4_statfs_args {
  498         char *path;
  499         struct ostatfs *buf;
  500 };
  501 #endif
  502 int
  503 freebsd4_statfs(td, uap)
  504         struct thread *td;
  505         struct freebsd4_statfs_args /* {
  506                 char *path;
  507                 struct ostatfs *buf;
  508         } */ *uap;
  509 {
  510         struct ostatfs osb;
  511         struct statfs sf;
  512         int error;
  513 
  514         error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
  515         if (error)
  516                 return (error);
  517         cvtstatfs(&sf, &osb);
  518         return (copyout(&osb, uap->buf, sizeof(osb)));
  519 }
  520 
  521 /*
  522  * Get filesystem statistics.
  523  */
  524 #ifndef _SYS_SYSPROTO_H_
  525 struct freebsd4_fstatfs_args {
  526         int fd;
  527         struct ostatfs *buf;
  528 };
  529 #endif
  530 int
  531 freebsd4_fstatfs(td, uap)
  532         struct thread *td;
  533         struct freebsd4_fstatfs_args /* {
  534                 int fd;
  535                 struct ostatfs *buf;
  536         } */ *uap;
  537 {
  538         struct ostatfs osb;
  539         struct statfs sf;
  540         int error;
  541 
  542         error = kern_fstatfs(td, uap->fd, &sf);
  543         if (error)
  544                 return (error);
  545         cvtstatfs(&sf, &osb);
  546         return (copyout(&osb, uap->buf, sizeof(osb)));
  547 }
  548 
  549 /*
  550  * Get statistics on all filesystems.
  551  */
  552 #ifndef _SYS_SYSPROTO_H_
  553 struct freebsd4_getfsstat_args {
  554         struct ostatfs *buf;
  555         long bufsize;
  556         int flags;
  557 };
  558 #endif
  559 int
  560 freebsd4_getfsstat(td, uap)
  561         struct thread *td;
  562         register struct freebsd4_getfsstat_args /* {
  563                 struct ostatfs *buf;
  564                 long bufsize;
  565                 int flags;
  566         } */ *uap;
  567 {
  568         struct statfs *buf, *sp;
  569         struct ostatfs osb;
  570         size_t count, size;
  571         int error;
  572 
  573         count = uap->bufsize / sizeof(struct ostatfs);
  574         size = count * sizeof(struct statfs);
  575         error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
  576         if (size > 0) {
  577                 count = td->td_retval[0];
  578                 sp = buf;
  579                 while (count > 0 && error == 0) {
  580                         cvtstatfs(sp, &osb);
  581                         error = copyout(&osb, uap->buf, sizeof(osb));
  582                         sp++;
  583                         uap->buf++;
  584                         count--;
  585                 }
  586                 free(buf, M_TEMP);
  587         }
  588         return (error);
  589 }
  590 
  591 /*
  592  * Implement fstatfs() for (NFS) file handles.
  593  */
  594 #ifndef _SYS_SYSPROTO_H_
  595 struct freebsd4_fhstatfs_args {
  596         struct fhandle *u_fhp;
  597         struct ostatfs *buf;
  598 };
  599 #endif
  600 int
  601 freebsd4_fhstatfs(td, uap)
  602         struct thread *td;
  603         struct freebsd4_fhstatfs_args /* {
  604                 struct fhandle *u_fhp;
  605                 struct ostatfs *buf;
  606         } */ *uap;
  607 {
  608         struct ostatfs osb;
  609         struct statfs sf;
  610         fhandle_t fh;
  611         int error;
  612 
  613         error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
  614         if (error)
  615                 return (error);
  616         error = kern_fhstatfs(td, fh, &sf);
  617         if (error)
  618                 return (error);
  619         cvtstatfs(&sf, &osb);
  620         return (copyout(&osb, uap->buf, sizeof(osb)));
  621 }
  622 
  623 /*
  624  * Convert a new format statfs structure to an old format statfs structure.
  625  */
  626 static void
  627 cvtstatfs(nsp, osp)
  628         struct statfs *nsp;
  629         struct ostatfs *osp;
  630 {
  631 
  632         bzero(osp, sizeof(*osp));
  633         osp->f_bsize = MIN(nsp->f_bsize, LONG_MAX);
  634         osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
  635         osp->f_blocks = MIN(nsp->f_blocks, LONG_MAX);
  636         osp->f_bfree = MIN(nsp->f_bfree, LONG_MAX);
  637         osp->f_bavail = MIN(nsp->f_bavail, LONG_MAX);
  638         osp->f_files = MIN(nsp->f_files, LONG_MAX);
  639         osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
  640         osp->f_owner = nsp->f_owner;
  641         osp->f_type = nsp->f_type;
  642         osp->f_flags = nsp->f_flags;
  643         osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
  644         osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
  645         osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
  646         osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
  647         bcopy(nsp->f_fstypename, osp->f_fstypename,
  648             MIN(MFSNAMELEN, OMNAMELEN));
  649         bcopy(nsp->f_mntonname, osp->f_mntonname,
  650             MIN(MFSNAMELEN, OMNAMELEN));
  651         bcopy(nsp->f_mntfromname, osp->f_mntfromname,
  652             MIN(MFSNAMELEN, OMNAMELEN));
  653         osp->f_fsid = nsp->f_fsid;
  654 }
  655 #endif /* COMPAT_FREEBSD4 */
  656 
  657 /*
  658  * Change current working directory to a given file descriptor.
  659  */
  660 #ifndef _SYS_SYSPROTO_H_
  661 struct fchdir_args {
  662         int     fd;
  663 };
  664 #endif
  665 int
  666 fchdir(td, uap)
  667         struct thread *td;
  668         struct fchdir_args /* {
  669                 int fd;
  670         } */ *uap;
  671 {
  672         register struct filedesc *fdp = td->td_proc->p_fd;
  673         struct vnode *vp, *tdp, *vpold;
  674         struct mount *mp;
  675         struct file *fp;
  676         int vfslocked;
  677         int error;
  678 
  679         if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
  680                 return (error);
  681         vp = fp->f_vnode;
  682         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
  683         VREF(vp);
  684         fdrop(fp, td);
  685         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
  686         if (vp->v_type != VDIR)
  687                 error = ENOTDIR;
  688 #ifdef MAC
  689         else if ((error = mac_check_vnode_chdir(td->td_ucred, vp)) != 0) {
  690         }
  691 #endif
  692         else
  693                 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
  694         while (!error && (mp = vp->v_mountedhere) != NULL) {
  695                 int tvfslocked;
  696                 if (vfs_busy(mp, 0, 0, td))
  697                         continue;
  698                 tvfslocked = VFS_LOCK_GIANT(mp);
  699                 error = VFS_ROOT(mp, LK_EXCLUSIVE, &tdp, td);
  700                 vfs_unbusy(mp, td);
  701                 if (error) {
  702                         VFS_UNLOCK_GIANT(tvfslocked);
  703                         break;
  704                 }
  705                 vput(vp);
  706                 VFS_UNLOCK_GIANT(vfslocked);
  707                 vp = tdp;
  708                 vfslocked = tvfslocked;
  709         }
  710         if (error) {
  711                 vput(vp);
  712                 VFS_UNLOCK_GIANT(vfslocked);
  713                 return (error);
  714         }
  715         VOP_UNLOCK(vp, 0, td);
  716         FILEDESC_LOCK_FAST(fdp);
  717         vpold = fdp->fd_cdir;
  718         fdp->fd_cdir = vp;
  719         FILEDESC_UNLOCK_FAST(fdp);
  720         vrele(vpold);
  721         VFS_UNLOCK_GIANT(vfslocked);
  722         return (0);
  723 }
  724 
  725 /*
  726  * Change current working directory (``.'').
  727  */
  728 #ifndef _SYS_SYSPROTO_H_
  729 struct chdir_args {
  730         char    *path;
  731 };
  732 #endif
  733 int
  734 chdir(td, uap)
  735         struct thread *td;
  736         struct chdir_args /* {
  737                 char *path;
  738         } */ *uap;
  739 {
  740 
  741         return (kern_chdir(td, uap->path, UIO_USERSPACE));
  742 }
  743 
  744 int
  745 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
  746 {
  747         register struct filedesc *fdp = td->td_proc->p_fd;
  748         int error;
  749         struct nameidata nd;
  750         struct vnode *vp;
  751         int vfslocked;
  752 
  753         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, pathseg, path, td);
  754         if ((error = namei(&nd)) != 0)
  755                 return (error);
  756         vfslocked = NDHASGIANT(&nd);
  757         if ((error = change_dir(nd.ni_vp, td)) != 0) {
  758                 vput(nd.ni_vp);
  759                 VFS_UNLOCK_GIANT(vfslocked);
  760                 NDFREE(&nd, NDF_ONLY_PNBUF);
  761                 return (error);
  762         }
  763         VOP_UNLOCK(nd.ni_vp, 0, td);
  764         NDFREE(&nd, NDF_ONLY_PNBUF);
  765         FILEDESC_LOCK_FAST(fdp);
  766         vp = fdp->fd_cdir;
  767         fdp->fd_cdir = nd.ni_vp;
  768         FILEDESC_UNLOCK_FAST(fdp);
  769         vrele(vp);
  770         VFS_UNLOCK_GIANT(vfslocked);
  771         return (0);
  772 }
  773 
  774 /*
  775  * Helper function for raised chroot(2) security function:  Refuse if
  776  * any filedescriptors are open directories.
  777  */
  778 static int
  779 chroot_refuse_vdir_fds(fdp)
  780         struct filedesc *fdp;
  781 {
  782         struct vnode *vp;
  783         struct file *fp;
  784         int fd;
  785 
  786         FILEDESC_LOCK_ASSERT(fdp, MA_OWNED);
  787         for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
  788                 fp = fget_locked(fdp, fd);
  789                 if (fp == NULL)
  790                         continue;
  791                 if (fp->f_type == DTYPE_VNODE) {
  792                         vp = fp->f_vnode;
  793                         if (vp->v_type == VDIR)
  794                                 return (EPERM);
  795                 }
  796         }
  797         return (0);
  798 }
  799 
  800 /*
  801  * This sysctl determines if we will allow a process to chroot(2) if it
  802  * has a directory open:
  803  *      0: disallowed for all processes.
  804  *      1: allowed for processes that were not already chroot(2)'ed.
  805  *      2: allowed for all processes.
  806  */
  807 
  808 static int chroot_allow_open_directories = 1;
  809 
  810 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
  811      &chroot_allow_open_directories, 0, "");
  812 
  813 /*
  814  * Change notion of root (``/'') directory.
  815  */
  816 #ifndef _SYS_SYSPROTO_H_
  817 struct chroot_args {
  818         char    *path;
  819 };
  820 #endif
  821 int
  822 chroot(td, uap)
  823         struct thread *td;
  824         struct chroot_args /* {
  825                 char *path;
  826         } */ *uap;
  827 {
  828         int error;
  829         struct nameidata nd;
  830         int vfslocked;
  831 
  832         error = suser_cred(td->td_ucred, SUSER_ALLOWJAIL);
  833         if (error)
  834                 return (error);
  835         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE,
  836             UIO_USERSPACE, uap->path, td);
  837         error = namei(&nd);
  838         if (error)
  839                 goto error;
  840         vfslocked = NDHASGIANT(&nd);
  841         if ((error = change_dir(nd.ni_vp, td)) != 0)
  842                 goto e_vunlock;
  843 #ifdef MAC
  844         if ((error = mac_check_vnode_chroot(td->td_ucred, nd.ni_vp)))
  845                 goto e_vunlock;
  846 #endif
  847         VOP_UNLOCK(nd.ni_vp, 0, td);
  848         error = change_root(nd.ni_vp, td);
  849         vrele(nd.ni_vp);
  850         VFS_UNLOCK_GIANT(vfslocked);
  851         NDFREE(&nd, NDF_ONLY_PNBUF);
  852         return (error);
  853 e_vunlock:
  854         vput(nd.ni_vp);
  855         VFS_UNLOCK_GIANT(vfslocked);
  856 error:
  857         NDFREE(&nd, NDF_ONLY_PNBUF);
  858         return (error);
  859 }
  860 
  861 /*
  862  * Common routine for chroot and chdir.  Callers must provide a locked vnode
  863  * instance.
  864  */
  865 int
  866 change_dir(vp, td)
  867         struct vnode *vp;
  868         struct thread *td;
  869 {
  870         int error;
  871 
  872         ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
  873         if (vp->v_type != VDIR)
  874                 return (ENOTDIR);
  875 #ifdef MAC
  876         error = mac_check_vnode_chdir(td->td_ucred, vp);
  877         if (error)
  878                 return (error);
  879 #endif
  880         error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
  881         return (error);
  882 }
  883 
  884 /*
  885  * Common routine for kern_chroot() and jail_attach().  The caller is
  886  * responsible for invoking suser() and mac_check_chroot() to authorize this
  887  * operation.
  888  */
  889 int
  890 change_root(vp, td)
  891         struct vnode *vp;
  892         struct thread *td;
  893 {
  894         struct filedesc *fdp;
  895         struct vnode *oldvp;
  896         int error;
  897 
  898         VFS_ASSERT_GIANT(vp->v_mount);
  899         fdp = td->td_proc->p_fd;
  900         FILEDESC_LOCK(fdp);
  901         if (chroot_allow_open_directories == 0 ||
  902             (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
  903                 error = chroot_refuse_vdir_fds(fdp);
  904                 if (error) {
  905                         FILEDESC_UNLOCK(fdp);
  906                         return (error);
  907                 }
  908         }
  909         oldvp = fdp->fd_rdir;
  910         fdp->fd_rdir = vp;
  911         VREF(fdp->fd_rdir);
  912         if (!fdp->fd_jdir) {
  913                 fdp->fd_jdir = vp;
  914                 VREF(fdp->fd_jdir);
  915         }
  916         FILEDESC_UNLOCK(fdp);
  917         vrele(oldvp);
  918         return (0);
  919 }
  920 
  921 /*
  922  * Check permissions, allocate an open file structure,
  923  * and call the device open routine if any.
  924  *
  925  * MP SAFE
  926  */
  927 #ifndef _SYS_SYSPROTO_H_
  928 struct open_args {
  929         char    *path;
  930         int     flags;
  931         int     mode;
  932 };
  933 #endif
  934 int
  935 open(td, uap)
  936         struct thread *td;
  937         register struct open_args /* {
  938                 char *path;
  939                 int flags;
  940                 int mode;
  941         } */ *uap;
  942 {
  943         int error;
  944 
  945         error = kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode);
  946         if (mtx_owned(&Giant))
  947                 printf("open: %s: %d\n", uap->path, error);
  948         return (error);
  949 }
  950 
  951 int
  952 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
  953     int mode)
  954 {
  955         struct proc *p = td->td_proc;
  956         struct filedesc *fdp = p->p_fd;
  957         struct file *fp;
  958         struct vnode *vp;
  959         struct vattr vat;
  960         struct mount *mp;
  961         int cmode;
  962         struct file *nfp;
  963         int type, indx, error;
  964         struct flock lf;
  965         struct nameidata nd;
  966         int vfslocked;
  967 
  968         if ((flags & O_ACCMODE) == O_ACCMODE)
  969                 return (EINVAL);
  970         flags = FFLAGS(flags);
  971         error = falloc(td, &nfp, &indx);
  972         if (error)
  973                 return (error);
  974         /* An extra reference on `nfp' has been held for us by falloc(). */
  975         fp = nfp;
  976         cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
  977         NDINIT(&nd, LOOKUP, FOLLOW, pathseg, path, td);
  978         td->td_dupfd = -1;              /* XXX check for fdopen */
  979         error = vn_open(&nd, &flags, cmode, indx);
  980         if (error) {
  981                 /*
  982                  * If the vn_open replaced the method vector, something
  983                  * wonderous happened deep below and we just pass it up
  984                  * pretending we know what we do.
  985                  */
  986                 if (error == ENXIO && fp->f_ops != &badfileops) {
  987                         fdrop(fp, td);
  988                         td->td_retval[0] = indx;
  989                         return (0);
  990                 }
  991 
  992                 /*
  993                  * release our own reference
  994                  */
  995                 fdrop(fp, td);
  996 
  997                 /*
  998                  * handle special fdopen() case.  bleh.  dupfdopen() is
  999                  * responsible for dropping the old contents of ofiles[indx]
 1000                  * if it succeeds.
 1001                  */
 1002                 if ((error == ENODEV || error == ENXIO) &&
 1003                     td->td_dupfd >= 0 &&                /* XXX from fdopen */
 1004                     (error =
 1005                         dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
 1006                         td->td_retval[0] = indx;
 1007                         return (0);
 1008                 }
 1009                 /*
 1010                  * Clean up the descriptor, but only if another thread hadn't
 1011                  * replaced or closed it.
 1012                  */
 1013                 fdclose(fdp, fp, indx, td);
 1014 
 1015                 if (error == ERESTART)
 1016                         error = EINTR;
 1017                 return (error);
 1018         }
 1019         td->td_dupfd = 0;
 1020         vfslocked = NDHASGIANT(&nd);
 1021         NDFREE(&nd, NDF_ONLY_PNBUF);
 1022         vp = nd.ni_vp;
 1023 
 1024         /*
 1025          * There should be 2 references on the file, one from the descriptor
 1026          * table, and one for us.
 1027          *
 1028          * Handle the case where someone closed the file (via its file
 1029          * descriptor) while we were blocked.  The end result should look
 1030          * like opening the file succeeded but it was immediately closed.
 1031          * We call vn_close() manually because we haven't yet hooked up
 1032          * the various 'struct file' fields.
 1033          */
 1034         FILEDESC_LOCK(fdp);
 1035         FILE_LOCK(fp);
 1036         if (fp->f_count == 1) {
 1037                 mp = vp->v_mount;
 1038                 KASSERT(fdp->fd_ofiles[indx] != fp,
 1039                     ("Open file descriptor lost all refs"));
 1040                 FILE_UNLOCK(fp);
 1041                 FILEDESC_UNLOCK(fdp);
 1042                 VOP_UNLOCK(vp, 0, td);
 1043                 vn_close(vp, flags & FMASK, fp->f_cred, td);
 1044                 VFS_UNLOCK_GIANT(vfslocked);
 1045                 fdrop(fp, td);
 1046                 td->td_retval[0] = indx;
 1047                 return (0);
 1048         }
 1049         fp->f_vnode = vp;
 1050         if (fp->f_data == NULL)
 1051                 fp->f_data = vp;
 1052         fp->f_flag = flags & FMASK;
 1053         if (fp->f_ops == &badfileops)
 1054                 fp->f_ops = &vnops;
 1055         fp->f_seqcount = 1;
 1056         fp->f_type = (vp->v_type == VFIFO ? DTYPE_FIFO : DTYPE_VNODE);
 1057         FILE_UNLOCK(fp);
 1058         FILEDESC_UNLOCK(fdp);
 1059 
 1060         VOP_UNLOCK(vp, 0, td);
 1061         if (flags & (O_EXLOCK | O_SHLOCK)) {
 1062                 lf.l_whence = SEEK_SET;
 1063                 lf.l_start = 0;
 1064                 lf.l_len = 0;
 1065                 if (flags & O_EXLOCK)
 1066                         lf.l_type = F_WRLCK;
 1067                 else
 1068                         lf.l_type = F_RDLCK;
 1069                 type = F_FLOCK;
 1070                 if ((flags & FNONBLOCK) == 0)
 1071                         type |= F_WAIT;
 1072                 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
 1073                             type)) != 0)
 1074                         goto bad;
 1075                 fp->f_flag |= FHASLOCK;
 1076         }
 1077         if (flags & O_TRUNC) {
 1078                 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 1079                         goto bad;
 1080                 VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 1081                 VATTR_NULL(&vat);
 1082                 vat.va_size = 0;
 1083                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 1084 #ifdef MAC
 1085                 error = mac_check_vnode_write(td->td_ucred, fp->f_cred, vp);
 1086                 if (error == 0)
 1087 #endif
 1088                         error = VOP_SETATTR(vp, &vat, td->td_ucred, td);
 1089                 VOP_UNLOCK(vp, 0, td);
 1090                 vn_finished_write(mp);
 1091                 if (error)
 1092                         goto bad;
 1093         }
 1094         VFS_UNLOCK_GIANT(vfslocked);
 1095         /*
 1096          * Release our private reference, leaving the one associated with
 1097          * the descriptor table intact.
 1098          */
 1099         fdrop(fp, td);
 1100         td->td_retval[0] = indx;
 1101         return (0);
 1102 bad:
 1103         VFS_UNLOCK_GIANT(vfslocked);
 1104         fdclose(fdp, fp, indx, td);
 1105         fdrop(fp, td);
 1106         return (error);
 1107 }
 1108 
 1109 #ifdef COMPAT_43
 1110 /*
 1111  * Create a file.
 1112  *
 1113  * MP SAFE
 1114  */
 1115 #ifndef _SYS_SYSPROTO_H_
 1116 struct ocreat_args {
 1117         char    *path;
 1118         int     mode;
 1119 };
 1120 #endif
 1121 int
 1122 ocreat(td, uap)
 1123         struct thread *td;
 1124         register struct ocreat_args /* {
 1125                 char *path;
 1126                 int mode;
 1127         } */ *uap;
 1128 {
 1129 
 1130         return (kern_open(td, uap->path, UIO_USERSPACE,
 1131             O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
 1132 }
 1133 #endif /* COMPAT_43 */
 1134 
 1135 /*
 1136  * Create a special file.
 1137  */
 1138 #ifndef _SYS_SYSPROTO_H_
 1139 struct mknod_args {
 1140         char    *path;
 1141         int     mode;
 1142         int     dev;
 1143 };
 1144 #endif
 1145 int
 1146 mknod(td, uap)
 1147         struct thread *td;
 1148         register struct mknod_args /* {
 1149                 char *path;
 1150                 int mode;
 1151                 int dev;
 1152         } */ *uap;
 1153 {
 1154 
 1155         return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
 1156 }
 1157 
 1158 int
 1159 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
 1160     int dev)
 1161 {
 1162         struct vnode *vp;
 1163         struct mount *mp;
 1164         struct vattr vattr;
 1165         int error;
 1166         int whiteout = 0;
 1167         struct nameidata nd;
 1168         int vfslocked;
 1169 
 1170         switch (mode & S_IFMT) {
 1171         case S_IFCHR:
 1172         case S_IFBLK:
 1173                 error = suser(td);
 1174                 break;
 1175         default:
 1176                 error = suser_cred(td->td_ucred, SUSER_ALLOWJAIL);
 1177                 break;
 1178         }
 1179         if (error)
 1180                 return (error);
 1181 restart:
 1182         bwillwrite();
 1183         NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE, pathseg, path, td);
 1184         if ((error = namei(&nd)) != 0)
 1185                 return (error);
 1186         vfslocked = NDHASGIANT(&nd);
 1187         vp = nd.ni_vp;
 1188         if (vp != NULL) {
 1189                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1190                 vrele(vp);
 1191                 if (vp == nd.ni_dvp)
 1192                         vrele(nd.ni_dvp);
 1193                 else
 1194                         vput(nd.ni_dvp);
 1195                 VFS_UNLOCK_GIANT(vfslocked);
 1196                 return (EEXIST);
 1197         } else {
 1198                 VATTR_NULL(&vattr);
 1199                 FILEDESC_LOCK_FAST(td->td_proc->p_fd);
 1200                 vattr.va_mode = (mode & ALLPERMS) &
 1201                     ~td->td_proc->p_fd->fd_cmask;
 1202                 FILEDESC_UNLOCK_FAST(td->td_proc->p_fd);
 1203                 vattr.va_rdev = dev;
 1204                 whiteout = 0;
 1205 
 1206                 switch (mode & S_IFMT) {
 1207                 case S_IFMT:    /* used by badsect to flag bad sectors */
 1208                         vattr.va_type = VBAD;
 1209                         break;
 1210                 case S_IFCHR:
 1211                         vattr.va_type = VCHR;
 1212                         break;
 1213                 case S_IFBLK:
 1214                         vattr.va_type = VBLK;
 1215                         break;
 1216                 case S_IFWHT:
 1217                         whiteout = 1;
 1218                         break;
 1219                 default:
 1220                         error = EINVAL;
 1221                         break;
 1222                 }
 1223         }
 1224         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 1225                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1226                 vput(nd.ni_dvp);
 1227                 VFS_UNLOCK_GIANT(vfslocked);
 1228                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 1229                         return (error);
 1230                 goto restart;
 1231         }
 1232 #ifdef MAC
 1233         if (error == 0 && !whiteout)
 1234                 error = mac_check_vnode_create(td->td_ucred, nd.ni_dvp,
 1235                     &nd.ni_cnd, &vattr);
 1236 #endif
 1237         if (!error) {
 1238                 VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1239                 if (whiteout)
 1240                         error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
 1241                 else {
 1242                         error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
 1243                                                 &nd.ni_cnd, &vattr);
 1244                         if (error == 0)
 1245                                 vput(nd.ni_vp);
 1246                 }
 1247         }
 1248         NDFREE(&nd, NDF_ONLY_PNBUF);
 1249         vput(nd.ni_dvp);
 1250         vn_finished_write(mp);
 1251         VFS_UNLOCK_GIANT(vfslocked);
 1252         return (error);
 1253 }
 1254 
 1255 /*
 1256  * Create a named pipe.
 1257  */
 1258 #ifndef _SYS_SYSPROTO_H_
 1259 struct mkfifo_args {
 1260         char    *path;
 1261         int     mode;
 1262 };
 1263 #endif
 1264 int
 1265 mkfifo(td, uap)
 1266         struct thread *td;
 1267         register struct mkfifo_args /* {
 1268                 char *path;
 1269                 int mode;
 1270         } */ *uap;
 1271 {
 1272 
 1273         return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
 1274 }
 1275 
 1276 int
 1277 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
 1278 {
 1279         struct mount *mp;
 1280         struct vattr vattr;
 1281         int error;
 1282         struct nameidata nd;
 1283         int vfslocked;
 1284 
 1285 restart:
 1286         bwillwrite();
 1287         NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE, pathseg, path, td);
 1288         if ((error = namei(&nd)) != 0)
 1289                 return (error);
 1290         vfslocked = NDHASGIANT(&nd);
 1291         if (nd.ni_vp != NULL) {
 1292                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1293                 vrele(nd.ni_vp);
 1294                 if (nd.ni_vp == nd.ni_dvp)
 1295                         vrele(nd.ni_dvp);
 1296                 else
 1297                         vput(nd.ni_dvp);
 1298                 VFS_UNLOCK_GIANT(vfslocked);
 1299                 return (EEXIST);
 1300         }
 1301         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 1302                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1303                 vput(nd.ni_dvp);
 1304                 VFS_UNLOCK_GIANT(vfslocked);
 1305                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 1306                         return (error);
 1307                 goto restart;
 1308         }
 1309         VATTR_NULL(&vattr);
 1310         vattr.va_type = VFIFO;
 1311         FILEDESC_LOCK_FAST(td->td_proc->p_fd);
 1312         vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
 1313         FILEDESC_UNLOCK_FAST(td->td_proc->p_fd);
 1314 #ifdef MAC
 1315         error = mac_check_vnode_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
 1316             &vattr);
 1317         if (error)
 1318                 goto out;
 1319 #endif
 1320         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1321         error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
 1322         if (error == 0)
 1323                 vput(nd.ni_vp);
 1324 #ifdef MAC
 1325 out:
 1326 #endif
 1327         vput(nd.ni_dvp);
 1328         vn_finished_write(mp);
 1329         VFS_UNLOCK_GIANT(vfslocked);
 1330         NDFREE(&nd, NDF_ONLY_PNBUF);
 1331         return (error);
 1332 }
 1333 
 1334 /*
 1335  * Make a hard file link.
 1336  */
 1337 #ifndef _SYS_SYSPROTO_H_
 1338 struct link_args {
 1339         char    *path;
 1340         char    *link;
 1341 };
 1342 #endif
 1343 int
 1344 link(td, uap)
 1345         struct thread *td;
 1346         register struct link_args /* {
 1347                 char *path;
 1348                 char *link;
 1349         } */ *uap;
 1350 {
 1351         int error;
 1352 
 1353         error = kern_link(td, uap->path, uap->link, UIO_USERSPACE);
 1354         return (error);
 1355 }
 1356 
 1357 SYSCTL_DECL(_security_bsd);
 1358 
 1359 static int hardlink_check_uid = 0;
 1360 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
 1361     &hardlink_check_uid, 0,
 1362     "Unprivileged processes cannot create hard links to files owned by other "
 1363     "users");
 1364 static int hardlink_check_gid = 0;
 1365 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
 1366     &hardlink_check_gid, 0,
 1367     "Unprivileged processes cannot create hard links to files owned by other "
 1368     "groups");
 1369 
 1370 static int
 1371 can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
 1372 {
 1373         struct vattr va;
 1374         int error;
 1375 
 1376         if (suser_cred(cred, SUSER_ALLOWJAIL) == 0)
 1377                 return (0);
 1378 
 1379         if (!hardlink_check_uid && !hardlink_check_gid)
 1380                 return (0);
 1381 
 1382         error = VOP_GETATTR(vp, &va, cred, td);
 1383         if (error != 0)
 1384                 return (error);
 1385 
 1386         if (hardlink_check_uid) {
 1387                 if (cred->cr_uid != va.va_uid)
 1388                         return (EPERM);
 1389         }
 1390 
 1391         if (hardlink_check_gid) {
 1392                 if (!groupmember(va.va_gid, cred))
 1393                         return (EPERM);
 1394         }
 1395 
 1396         return (0);
 1397 }
 1398 
 1399 int
 1400 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
 1401 {
 1402         struct vnode *vp;
 1403         struct mount *mp;
 1404         struct nameidata nd;
 1405         int vfslocked;
 1406         int lvfslocked;
 1407         int error;
 1408 
 1409         bwillwrite();
 1410         NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, segflg, path, td);
 1411         if ((error = namei(&nd)) != 0)
 1412                 return (error);
 1413         vfslocked = NDHASGIANT(&nd);
 1414         NDFREE(&nd, NDF_ONLY_PNBUF);
 1415         vp = nd.ni_vp;
 1416         if (vp->v_type == VDIR) {
 1417                 vrele(vp);
 1418                 VFS_UNLOCK_GIANT(vfslocked);
 1419                 return (EPERM);         /* POSIX */
 1420         }
 1421         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
 1422                 vrele(vp);
 1423                 VFS_UNLOCK_GIANT(vfslocked);
 1424                 return (error);
 1425         }
 1426         NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE, segflg, link, td);
 1427         if ((error = namei(&nd)) == 0) {
 1428                 lvfslocked = NDHASGIANT(&nd);
 1429                 if (nd.ni_vp != NULL) {
 1430                         vrele(nd.ni_vp);
 1431                         if (nd.ni_dvp == nd.ni_vp)
 1432                                 vrele(nd.ni_dvp);
 1433                         else
 1434                                 vput(nd.ni_dvp);
 1435                         error = EEXIST;
 1436                 } else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td))
 1437                     == 0) {
 1438                         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1439                         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 1440                         error = can_hardlink(vp, td, td->td_ucred);
 1441                         if (error == 0)
 1442 #ifdef MAC
 1443                                 error = mac_check_vnode_link(td->td_ucred,
 1444                                     nd.ni_dvp, vp, &nd.ni_cnd);
 1445                         if (error == 0)
 1446 #endif
 1447                                 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
 1448                         VOP_UNLOCK(vp, 0, td);
 1449                         vput(nd.ni_dvp);
 1450                 }
 1451                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1452                 VFS_UNLOCK_GIANT(lvfslocked);
 1453         }
 1454         vrele(vp);
 1455         vn_finished_write(mp);
 1456         VFS_UNLOCK_GIANT(vfslocked);
 1457         return (error);
 1458 }
 1459 
 1460 /*
 1461  * Make a symbolic link.
 1462  */
 1463 #ifndef _SYS_SYSPROTO_H_
 1464 struct symlink_args {
 1465         char    *path;
 1466         char    *link;
 1467 };
 1468 #endif
 1469 int
 1470 symlink(td, uap)
 1471         struct thread *td;
 1472         register struct symlink_args /* {
 1473                 char *path;
 1474                 char *link;
 1475         } */ *uap;
 1476 {
 1477 
 1478         return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
 1479 }
 1480 
 1481 int
 1482 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
 1483 {
 1484         struct mount *mp;
 1485         struct vattr vattr;
 1486         char *syspath;
 1487         int error;
 1488         struct nameidata nd;
 1489         int vfslocked;
 1490 
 1491         if (segflg == UIO_SYSSPACE) {
 1492                 syspath = path;
 1493         } else {
 1494                 syspath = uma_zalloc(namei_zone, M_WAITOK);
 1495                 if ((error = copyinstr(path, syspath, MAXPATHLEN, NULL)) != 0)
 1496                         goto out;
 1497         }
 1498 restart:
 1499         bwillwrite();
 1500         NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE,
 1501             segflg, link, td);
 1502         if ((error = namei(&nd)) != 0)
 1503                 goto out;
 1504         vfslocked = NDHASGIANT(&nd);
 1505         if (nd.ni_vp) {
 1506                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1507                 vrele(nd.ni_vp);
 1508                 if (nd.ni_vp == nd.ni_dvp)
 1509                         vrele(nd.ni_dvp);
 1510                 else
 1511                         vput(nd.ni_dvp);
 1512                 VFS_UNLOCK_GIANT(vfslocked);
 1513                 error = EEXIST;
 1514                 goto out;
 1515         }
 1516         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 1517                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1518                 vput(nd.ni_dvp);
 1519                 VFS_UNLOCK_GIANT(vfslocked);
 1520                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 1521                         goto out;
 1522                 goto restart;
 1523         }
 1524         VATTR_NULL(&vattr);
 1525         FILEDESC_LOCK_FAST(td->td_proc->p_fd);
 1526         vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
 1527         FILEDESC_UNLOCK_FAST(td->td_proc->p_fd);
 1528 #ifdef MAC
 1529         vattr.va_type = VLNK;
 1530         error = mac_check_vnode_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
 1531             &vattr);
 1532         if (error)
 1533                 goto out2;
 1534 #endif
 1535         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1536         error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
 1537         if (error == 0)
 1538                 vput(nd.ni_vp);
 1539 #ifdef MAC
 1540 out2:
 1541 #endif
 1542         NDFREE(&nd, NDF_ONLY_PNBUF);
 1543         vput(nd.ni_dvp);
 1544         vn_finished_write(mp);
 1545         VFS_UNLOCK_GIANT(vfslocked);
 1546 out:
 1547         if (segflg != UIO_SYSSPACE)
 1548                 uma_zfree(namei_zone, syspath);
 1549         return (error);
 1550 }
 1551 
 1552 /*
 1553  * Delete a whiteout from the filesystem.
 1554  */
 1555 int
 1556 undelete(td, uap)
 1557         struct thread *td;
 1558         register struct undelete_args /* {
 1559                 char *path;
 1560         } */ *uap;
 1561 {
 1562         int error;
 1563         struct mount *mp;
 1564         struct nameidata nd;
 1565         int vfslocked;
 1566 
 1567 restart:
 1568         bwillwrite();
 1569         NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE, UIO_USERSPACE,
 1570             uap->path, td);
 1571         error = namei(&nd);
 1572         if (error)
 1573                 return (error);
 1574         vfslocked = NDHASGIANT(&nd);
 1575 
 1576         if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
 1577                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1578                 if (nd.ni_vp)
 1579                         vrele(nd.ni_vp);
 1580                 if (nd.ni_vp == nd.ni_dvp)
 1581                         vrele(nd.ni_dvp);
 1582                 else
 1583                         vput(nd.ni_dvp);
 1584                 VFS_UNLOCK_GIANT(vfslocked);
 1585                 return (EEXIST);
 1586         }
 1587         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 1588                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1589                 vput(nd.ni_dvp);
 1590                 VFS_UNLOCK_GIANT(vfslocked);
 1591                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 1592                         return (error);
 1593                 goto restart;
 1594         }
 1595         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1596         error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
 1597         NDFREE(&nd, NDF_ONLY_PNBUF);
 1598         vput(nd.ni_dvp);
 1599         vn_finished_write(mp);
 1600         VFS_UNLOCK_GIANT(vfslocked);
 1601         return (error);
 1602 }
 1603 
 1604 /*
 1605  * Delete a name from the filesystem.
 1606  */
 1607 #ifndef _SYS_SYSPROTO_H_
 1608 struct unlink_args {
 1609         char    *path;
 1610 };
 1611 #endif
 1612 int
 1613 unlink(td, uap)
 1614         struct thread *td;
 1615         struct unlink_args /* {
 1616                 char *path;
 1617         } */ *uap;
 1618 {
 1619         int error;
 1620 
 1621         error = kern_unlink(td, uap->path, UIO_USERSPACE);
 1622         return (error);
 1623 }
 1624 
 1625 int
 1626 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
 1627 {
 1628         struct mount *mp;
 1629         struct vnode *vp;
 1630         int error;
 1631         struct nameidata nd;
 1632         int vfslocked;
 1633 
 1634 restart:
 1635         bwillwrite();
 1636         NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE, pathseg, path, td);
 1637         if ((error = namei(&nd)) != 0)
 1638                 return (error);
 1639         vfslocked = NDHASGIANT(&nd);
 1640         vp = nd.ni_vp;
 1641         if (vp->v_type == VDIR)
 1642                 error = EPERM;          /* POSIX */
 1643         else {
 1644                 /*
 1645                  * The root of a mounted filesystem cannot be deleted.
 1646                  *
 1647                  * XXX: can this only be a VDIR case?
 1648                  */
 1649                 if (vp->v_vflag & VV_ROOT)
 1650                         error = EBUSY;
 1651         }
 1652         if (error == 0) {
 1653                 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 1654                         NDFREE(&nd, NDF_ONLY_PNBUF);
 1655                         if (vp == nd.ni_dvp)
 1656                                 vrele(vp);
 1657                         else
 1658                                 vput(vp);
 1659                         vput(nd.ni_dvp);
 1660                         VFS_UNLOCK_GIANT(vfslocked);
 1661                         if ((error = vn_start_write(NULL, &mp,
 1662                             V_XSLEEP | PCATCH)) != 0)
 1663                                 return (error);
 1664                         goto restart;
 1665                 }
 1666 #ifdef MAC
 1667                 error = mac_check_vnode_delete(td->td_ucred, nd.ni_dvp, vp,
 1668                     &nd.ni_cnd);
 1669                 if (error)
 1670                         goto out;
 1671 #endif
 1672                 VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1673                 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
 1674 #ifdef MAC
 1675 out:
 1676 #endif
 1677                 vn_finished_write(mp);
 1678         }
 1679         NDFREE(&nd, NDF_ONLY_PNBUF);
 1680         if (vp == nd.ni_dvp)
 1681                 vrele(vp);
 1682         else
 1683                 vput(vp);
 1684         vput(nd.ni_dvp);
 1685         VFS_UNLOCK_GIANT(vfslocked);
 1686         return (error);
 1687 }
 1688 
 1689 /*
 1690  * Reposition read/write file offset.
 1691  */
 1692 #ifndef _SYS_SYSPROTO_H_
 1693 struct lseek_args {
 1694         int     fd;
 1695         int     pad;
 1696         off_t   offset;
 1697         int     whence;
 1698 };
 1699 #endif
 1700 int
 1701 lseek(td, uap)
 1702         struct thread *td;
 1703         register struct lseek_args /* {
 1704                 int fd;
 1705                 int pad;
 1706                 off_t offset;
 1707                 int whence;
 1708         } */ *uap;
 1709 {
 1710         struct ucred *cred = td->td_ucred;
 1711         struct file *fp;
 1712         struct vnode *vp;
 1713         struct vattr vattr;
 1714         off_t offset;
 1715         int error, noneg;
 1716         int vfslocked;
 1717 
 1718         if ((error = fget(td, uap->fd, &fp)) != 0)
 1719                 return (error);
 1720         if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
 1721                 fdrop(fp, td);
 1722                 return (ESPIPE);
 1723         }
 1724         vp = fp->f_vnode;
 1725         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
 1726         noneg = (vp->v_type != VCHR);
 1727         offset = uap->offset;
 1728         switch (uap->whence) {
 1729         case L_INCR:
 1730                 if (noneg &&
 1731                     (fp->f_offset < 0 ||
 1732                     (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
 1733                         error = EOVERFLOW;
 1734                         break;
 1735                 }
 1736                 offset += fp->f_offset;
 1737                 break;
 1738         case L_XTND:
 1739                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 1740                 error = VOP_GETATTR(vp, &vattr, cred, td);
 1741                 VOP_UNLOCK(vp, 0, td);
 1742                 if (error)
 1743                         break;
 1744                 if (noneg &&
 1745                     (vattr.va_size > OFF_MAX ||
 1746                     (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
 1747                         error = EOVERFLOW;
 1748                         break;
 1749                 }
 1750                 offset += vattr.va_size;
 1751                 break;
 1752         case L_SET:
 1753                 break;
 1754         default:
 1755                 error = EINVAL;
 1756         }
 1757         if (error == 0 && noneg && offset < 0)
 1758                 error = EINVAL;
 1759         if (error != 0)
 1760                 goto drop;
 1761         fp->f_offset = offset;
 1762         *(off_t *)(td->td_retval) = fp->f_offset;
 1763 drop:
 1764         fdrop(fp, td);
 1765         VFS_UNLOCK_GIANT(vfslocked);
 1766         return (error);
 1767 }
 1768 
 1769 #if defined(COMPAT_43)
 1770 /*
 1771  * Reposition read/write file offset.
 1772  */
 1773 #ifndef _SYS_SYSPROTO_H_
 1774 struct olseek_args {
 1775         int     fd;
 1776         long    offset;
 1777         int     whence;
 1778 };
 1779 #endif
 1780 int
 1781 olseek(td, uap)
 1782         struct thread *td;
 1783         register struct olseek_args /* {
 1784                 int fd;
 1785                 long offset;
 1786                 int whence;
 1787         } */ *uap;
 1788 {
 1789         struct lseek_args /* {
 1790                 int fd;
 1791                 int pad;
 1792                 off_t offset;
 1793                 int whence;
 1794         } */ nuap;
 1795         int error;
 1796 
 1797         nuap.fd = uap->fd;
 1798         nuap.offset = uap->offset;
 1799         nuap.whence = uap->whence;
 1800         error = lseek(td, &nuap);
 1801         return (error);
 1802 }
 1803 #endif /* COMPAT_43 */
 1804 
 1805 /*
 1806  * Check access permissions using passed credentials.
 1807  */
 1808 static int
 1809 vn_access(vp, user_flags, cred, td)
 1810         struct vnode    *vp;
 1811         int             user_flags;
 1812         struct ucred    *cred;
 1813         struct thread   *td;
 1814 {
 1815         int error, flags;
 1816 
 1817         /* Flags == 0 means only check for existence. */
 1818         error = 0;
 1819         if (user_flags) {
 1820                 flags = 0;
 1821                 if (user_flags & R_OK)
 1822                         flags |= VREAD;
 1823                 if (user_flags & W_OK)
 1824                         flags |= VWRITE;
 1825                 if (user_flags & X_OK)
 1826                         flags |= VEXEC;
 1827 #ifdef MAC
 1828                 error = mac_check_vnode_access(cred, vp, flags);
 1829                 if (error)
 1830                         return (error);
 1831 #endif
 1832                 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
 1833                         error = VOP_ACCESS(vp, flags, cred, td);
 1834         }
 1835         return (error);
 1836 }
 1837 
 1838 /*
 1839  * Check access permissions using "real" credentials.
 1840  */
 1841 #ifndef _SYS_SYSPROTO_H_
 1842 struct access_args {
 1843         char    *path;
 1844         int     flags;
 1845 };
 1846 #endif
 1847 int
 1848 access(td, uap)
 1849         struct thread *td;
 1850         register struct access_args /* {
 1851                 char *path;
 1852                 int flags;
 1853         } */ *uap;
 1854 {
 1855 
 1856         return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
 1857 }
 1858 
 1859 int
 1860 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int flags)
 1861 {
 1862         struct ucred *cred, *tmpcred;
 1863         register struct vnode *vp;
 1864         struct nameidata nd;
 1865         int vfslocked;
 1866         int error;
 1867 
 1868         /*
 1869          * Create and modify a temporary credential instead of one that
 1870          * is potentially shared.  This could also mess up socket
 1871          * buffer accounting which can run in an interrupt context.
 1872          */
 1873         cred = td->td_ucred;
 1874         tmpcred = crdup(cred);
 1875         tmpcred->cr_uid = cred->cr_ruid;
 1876         tmpcred->cr_groups[0] = cred->cr_rgid;
 1877         td->td_ucred = tmpcred;
 1878         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, pathseg, path, td);
 1879         if ((error = namei(&nd)) != 0)
 1880                 goto out1;
 1881         vfslocked = NDHASGIANT(&nd);
 1882         vp = nd.ni_vp;
 1883 
 1884         error = vn_access(vp, flags, tmpcred, td);
 1885         NDFREE(&nd, NDF_ONLY_PNBUF);
 1886         vput(vp);
 1887         VFS_UNLOCK_GIANT(vfslocked);
 1888 out1:
 1889         td->td_ucred = cred;
 1890         crfree(tmpcred);
 1891         return (error);
 1892 }
 1893 
 1894 /*
 1895  * Check access permissions using "effective" credentials.
 1896  */
 1897 #ifndef _SYS_SYSPROTO_H_
 1898 struct eaccess_args {
 1899         char    *path;
 1900         int     flags;
 1901 };
 1902 #endif
 1903 int
 1904 eaccess(td, uap)
 1905         struct thread *td;
 1906         register struct eaccess_args /* {
 1907                 char *path;
 1908                 int flags;
 1909         } */ *uap;
 1910 {
 1911         struct nameidata nd;
 1912         struct vnode *vp;
 1913         int vfslocked;
 1914         int error;
 1915 
 1916         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, UIO_USERSPACE,
 1917             uap->path, td);
 1918         if ((error = namei(&nd)) != 0)
 1919                 return (error);
 1920         vp = nd.ni_vp;
 1921         vfslocked = NDHASGIANT(&nd);
 1922         error = vn_access(vp, uap->flags, td->td_ucred, td);
 1923         NDFREE(&nd, NDF_ONLY_PNBUF);
 1924         vput(vp);
 1925         VFS_UNLOCK_GIANT(vfslocked);
 1926         return (error);
 1927 }
 1928 
 1929 #if defined(COMPAT_43)
 1930 /*
 1931  * Get file status; this version follows links.
 1932  */
 1933 #ifndef _SYS_SYSPROTO_H_
 1934 struct ostat_args {
 1935         char    *path;
 1936         struct ostat *ub;
 1937 };
 1938 #endif
 1939 int
 1940 ostat(td, uap)
 1941         struct thread *td;
 1942         register struct ostat_args /* {
 1943                 char *path;
 1944                 struct ostat *ub;
 1945         } */ *uap;
 1946 {
 1947         struct stat sb;
 1948         struct ostat osb;
 1949         int error;
 1950 
 1951         error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
 1952         if (error)
 1953                 return (error);
 1954         cvtstat(&sb, &osb);
 1955         error = copyout(&osb, uap->ub, sizeof (osb));
 1956         return (error);
 1957 }
 1958 
 1959 /*
 1960  * Get file status; this version does not follow links.
 1961  */
 1962 #ifndef _SYS_SYSPROTO_H_
 1963 struct olstat_args {
 1964         char    *path;
 1965         struct ostat *ub;
 1966 };
 1967 #endif
 1968 int
 1969 olstat(td, uap)
 1970         struct thread *td;
 1971         register struct olstat_args /* {
 1972                 char *path;
 1973                 struct ostat *ub;
 1974         } */ *uap;
 1975 {
 1976         struct stat sb;
 1977         struct ostat osb;
 1978         int error;
 1979 
 1980         error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
 1981         if (error)
 1982                 return (error);
 1983         cvtstat(&sb, &osb);
 1984         error = copyout(&osb, uap->ub, sizeof (osb));
 1985         return (error);
 1986 }
 1987 
 1988 /*
 1989  * Convert from an old to a new stat structure.
 1990  */
 1991 void
 1992 cvtstat(st, ost)
 1993         struct stat *st;
 1994         struct ostat *ost;
 1995 {
 1996 
 1997         ost->st_dev = st->st_dev;
 1998         ost->st_ino = st->st_ino;
 1999         ost->st_mode = st->st_mode;
 2000         ost->st_nlink = st->st_nlink;
 2001         ost->st_uid = st->st_uid;
 2002         ost->st_gid = st->st_gid;
 2003         ost->st_rdev = st->st_rdev;
 2004         if (st->st_size < (quad_t)1 << 32)
 2005                 ost->st_size = st->st_size;
 2006         else
 2007                 ost->st_size = -2;
 2008         ost->st_atime = st->st_atime;
 2009         ost->st_mtime = st->st_mtime;
 2010         ost->st_ctime = st->st_ctime;
 2011         ost->st_blksize = st->st_blksize;
 2012         ost->st_blocks = st->st_blocks;
 2013         ost->st_flags = st->st_flags;
 2014         ost->st_gen = st->st_gen;
 2015 }
 2016 #endif /* COMPAT_43 */
 2017 
 2018 /*
 2019  * Get file status; this version follows links.
 2020  */
 2021 #ifndef _SYS_SYSPROTO_H_
 2022 struct stat_args {
 2023         char    *path;
 2024         struct stat *ub;
 2025 };
 2026 #endif
 2027 int
 2028 stat(td, uap)
 2029         struct thread *td;
 2030         register struct stat_args /* {
 2031                 char *path;
 2032                 struct stat *ub;
 2033         } */ *uap;
 2034 {
 2035         struct stat sb;
 2036         int error;
 2037 
 2038         error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
 2039         if (error == 0)
 2040                 error = copyout(&sb, uap->ub, sizeof (sb));
 2041         return (error);
 2042 }
 2043 
 2044 int
 2045 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
 2046 {
 2047         struct nameidata nd;
 2048         struct stat sb;
 2049         int error, vfslocked;
 2050 
 2051         NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE,
 2052             pathseg, path, td);
 2053         if ((error = namei(&nd)) != 0)
 2054                 return (error);
 2055         vfslocked = NDHASGIANT(&nd);
 2056         error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
 2057         NDFREE(&nd, NDF_ONLY_PNBUF);
 2058         vput(nd.ni_vp);
 2059         VFS_UNLOCK_GIANT(vfslocked);
 2060         if (error)
 2061                 return (error);
 2062         *sbp = sb;
 2063         return (0);
 2064 }
 2065 
 2066 /*
 2067  * Get file status; this version does not follow links.
 2068  */
 2069 #ifndef _SYS_SYSPROTO_H_
 2070 struct lstat_args {
 2071         char    *path;
 2072         struct stat *ub;
 2073 };
 2074 #endif
 2075 int
 2076 lstat(td, uap)
 2077         struct thread *td;
 2078         register struct lstat_args /* {
 2079                 char *path;
 2080                 struct stat *ub;
 2081         } */ *uap;
 2082 {
 2083         struct stat sb;
 2084         int error;
 2085 
 2086         error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
 2087         if (error == 0)
 2088                 error = copyout(&sb, uap->ub, sizeof (sb));
 2089         return (error);
 2090 }
 2091 
 2092 int
 2093 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
 2094 {
 2095         struct vnode *vp;
 2096         struct stat sb;
 2097         struct nameidata nd;
 2098         int error, vfslocked;
 2099 
 2100         NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE,
 2101             pathseg, path, td);
 2102         if ((error = namei(&nd)) != 0)
 2103                 return (error);
 2104         vfslocked = NDHASGIANT(&nd);
 2105         vp = nd.ni_vp;
 2106         error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
 2107         NDFREE(&nd, NDF_ONLY_PNBUF);
 2108         vput(vp);
 2109         VFS_UNLOCK_GIANT(vfslocked);
 2110         if (error)
 2111                 return (error);
 2112         *sbp = sb;
 2113         return (0);
 2114 }
 2115 
 2116 /*
 2117  * Implementation of the NetBSD [l]stat() functions.
 2118  */
 2119 void
 2120 cvtnstat(sb, nsb)
 2121         struct stat *sb;
 2122         struct nstat *nsb;
 2123 {
 2124         bzero(nsb, sizeof *nsb);
 2125         nsb->st_dev = sb->st_dev;
 2126         nsb->st_ino = sb->st_ino;
 2127         nsb->st_mode = sb->st_mode;
 2128         nsb->st_nlink = sb->st_nlink;
 2129         nsb->st_uid = sb->st_uid;
 2130         nsb->st_gid = sb->st_gid;
 2131         nsb->st_rdev = sb->st_rdev;
 2132         nsb->st_atimespec = sb->st_atimespec;
 2133         nsb->st_mtimespec = sb->st_mtimespec;
 2134         nsb->st_ctimespec = sb->st_ctimespec;
 2135         nsb->st_size = sb->st_size;
 2136         nsb->st_blocks = sb->st_blocks;
 2137         nsb->st_blksize = sb->st_blksize;
 2138         nsb->st_flags = sb->st_flags;
 2139         nsb->st_gen = sb->st_gen;
 2140         nsb->st_birthtimespec = sb->st_birthtimespec;
 2141 }
 2142 
 2143 #ifndef _SYS_SYSPROTO_H_
 2144 struct nstat_args {
 2145         char    *path;
 2146         struct nstat *ub;
 2147 };
 2148 #endif
 2149 int
 2150 nstat(td, uap)
 2151         struct thread *td;
 2152         register struct nstat_args /* {
 2153                 char *path;
 2154                 struct nstat *ub;
 2155         } */ *uap;
 2156 {
 2157         struct stat sb;
 2158         struct nstat nsb;
 2159         int error;
 2160 
 2161         error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
 2162         if (error)
 2163                 return (error);
 2164         cvtnstat(&sb, &nsb);
 2165         error = copyout(&nsb, uap->ub, sizeof (nsb));
 2166         return (error);
 2167 }
 2168 
 2169 /*
 2170  * NetBSD lstat.  Get file status; this version does not follow links.
 2171  */
 2172 #ifndef _SYS_SYSPROTO_H_
 2173 struct lstat_args {
 2174         char    *path;
 2175         struct stat *ub;
 2176 };
 2177 #endif
 2178 int
 2179 nlstat(td, uap)
 2180         struct thread *td;
 2181         register struct nlstat_args /* {
 2182                 char *path;
 2183                 struct nstat *ub;
 2184         } */ *uap;
 2185 {
 2186         struct stat sb;
 2187         struct nstat nsb;
 2188         int error;
 2189 
 2190         error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
 2191         if (error)
 2192                 return (error);
 2193         cvtnstat(&sb, &nsb);
 2194         error = copyout(&nsb, uap->ub, sizeof (nsb));
 2195         return (error);
 2196 }
 2197 
 2198 /*
 2199  * Get configurable pathname variables.
 2200  */
 2201 #ifndef _SYS_SYSPROTO_H_
 2202 struct pathconf_args {
 2203         char    *path;
 2204         int     name;
 2205 };
 2206 #endif
 2207 int
 2208 pathconf(td, uap)
 2209         struct thread *td;
 2210         register struct pathconf_args /* {
 2211                 char *path;
 2212                 int name;
 2213         } */ *uap;
 2214 {
 2215 
 2216         return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name));
 2217 }
 2218 
 2219 int
 2220 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name)
 2221 {
 2222         struct nameidata nd;
 2223         int error, vfslocked;
 2224 
 2225         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, pathseg, path, td);
 2226         if ((error = namei(&nd)) != 0)
 2227                 return (error);
 2228         vfslocked = NDHASGIANT(&nd);
 2229         NDFREE(&nd, NDF_ONLY_PNBUF);
 2230 
 2231         /* If asynchronous I/O is available, it works for all files. */
 2232         if (name == _PC_ASYNC_IO)
 2233                 td->td_retval[0] = async_io_version;
 2234         else
 2235                 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
 2236         vput(nd.ni_vp);
 2237         VFS_UNLOCK_GIANT(vfslocked);
 2238         return (error);
 2239 }
 2240 
 2241 /*
 2242  * Return target name of a symbolic link.
 2243  */
 2244 #ifndef _SYS_SYSPROTO_H_
 2245 struct readlink_args {
 2246         char    *path;
 2247         char    *buf;
 2248         int     count;
 2249 };
 2250 #endif
 2251 int
 2252 readlink(td, uap)
 2253         struct thread *td;
 2254         register struct readlink_args /* {
 2255                 char *path;
 2256                 char *buf;
 2257                 int count;
 2258         } */ *uap;
 2259 {
 2260 
 2261         return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
 2262             UIO_USERSPACE, uap->count));
 2263 }
 2264 
 2265 int
 2266 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
 2267     enum uio_seg bufseg, int count)
 2268 {
 2269         register struct vnode *vp;
 2270         struct iovec aiov;
 2271         struct uio auio;
 2272         int error;
 2273         struct nameidata nd;
 2274         int vfslocked;
 2275 
 2276         NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE, pathseg, path, td);
 2277         if ((error = namei(&nd)) != 0)
 2278                 return (error);
 2279         NDFREE(&nd, NDF_ONLY_PNBUF);
 2280         vfslocked = NDHASGIANT(&nd);
 2281         vp = nd.ni_vp;
 2282 #ifdef MAC
 2283         error = mac_check_vnode_readlink(td->td_ucred, vp);
 2284         if (error) {
 2285                 vput(vp);
 2286                 VFS_UNLOCK_GIANT(vfslocked);
 2287                 return (error);
 2288         }
 2289 #endif
 2290         if (vp->v_type != VLNK)
 2291                 error = EINVAL;
 2292         else {
 2293                 aiov.iov_base = buf;
 2294                 aiov.iov_len = count;
 2295                 auio.uio_iov = &aiov;
 2296                 auio.uio_iovcnt = 1;
 2297                 auio.uio_offset = 0;
 2298                 auio.uio_rw = UIO_READ;
 2299                 auio.uio_segflg = bufseg;
 2300                 auio.uio_td = td;
 2301                 auio.uio_resid = count;
 2302                 error = VOP_READLINK(vp, &auio, td->td_ucred);
 2303         }
 2304         vput(vp);
 2305         VFS_UNLOCK_GIANT(vfslocked);
 2306         td->td_retval[0] = count - auio.uio_resid;
 2307         return (error);
 2308 }
 2309 
 2310 /*
 2311  * Common implementation code for chflags() and fchflags().
 2312  */
 2313 static int
 2314 setfflags(td, vp, flags)
 2315         struct thread *td;
 2316         struct vnode *vp;
 2317         int flags;
 2318 {
 2319         int error;
 2320         struct mount *mp;
 2321         struct vattr vattr;
 2322 
 2323         /*
 2324          * Prevent non-root users from setting flags on devices.  When
 2325          * a device is reused, users can retain ownership of the device
 2326          * if they are allowed to set flags and programs assume that
 2327          * chown can't fail when done as root.
 2328          */
 2329         if (vp->v_type == VCHR || vp->v_type == VBLK) {
 2330                 error = suser_cred(td->td_ucred, SUSER_ALLOWJAIL);
 2331                 if (error)
 2332                         return (error);
 2333         }
 2334 
 2335         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 2336                 return (error);
 2337         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2338         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2339         VATTR_NULL(&vattr);
 2340         vattr.va_flags = flags;
 2341 #ifdef MAC
 2342         error = mac_check_vnode_setflags(td->td_ucred, vp, vattr.va_flags);
 2343         if (error == 0)
 2344 #endif
 2345                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 2346         VOP_UNLOCK(vp, 0, td);
 2347         vn_finished_write(mp);
 2348         return (error);
 2349 }
 2350 
 2351 /*
 2352  * Change flags of a file given a path name.
 2353  */
 2354 #ifndef _SYS_SYSPROTO_H_
 2355 struct chflags_args {
 2356         char    *path;
 2357         int     flags;
 2358 };
 2359 #endif
 2360 int
 2361 chflags(td, uap)
 2362         struct thread *td;
 2363         register struct chflags_args /* {
 2364                 char *path;
 2365                 int flags;
 2366         } */ *uap;
 2367 {
 2368         int error;
 2369         struct nameidata nd;
 2370         int vfslocked;
 2371 
 2372         NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, UIO_USERSPACE, uap->path, td);
 2373         if ((error = namei(&nd)) != 0)
 2374                 return (error);
 2375         NDFREE(&nd, NDF_ONLY_PNBUF);
 2376         vfslocked = NDHASGIANT(&nd);
 2377         error = setfflags(td, nd.ni_vp, uap->flags);
 2378         vrele(nd.ni_vp);
 2379         VFS_UNLOCK_GIANT(vfslocked);
 2380         return (error);
 2381 }
 2382 
 2383 /*
 2384  * Same as chflags() but doesn't follow symlinks.
 2385  */
 2386 int
 2387 lchflags(td, uap)
 2388         struct thread *td;
 2389         register struct lchflags_args /* {
 2390                 char *path;
 2391                 int flags;
 2392         } */ *uap;
 2393 {
 2394         int error;
 2395         struct nameidata nd;
 2396         int vfslocked;
 2397 
 2398         NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_USERSPACE, uap->path, td);
 2399         if ((error = namei(&nd)) != 0)
 2400                 return (error);
 2401         vfslocked = NDHASGIANT(&nd);
 2402         NDFREE(&nd, NDF_ONLY_PNBUF);
 2403         error = setfflags(td, nd.ni_vp, uap->flags);
 2404         vrele(nd.ni_vp);
 2405         VFS_UNLOCK_GIANT(vfslocked);
 2406         return (error);
 2407 }
 2408 
 2409 /*
 2410  * Change flags of a file given a file descriptor.
 2411  */
 2412 #ifndef _SYS_SYSPROTO_H_
 2413 struct fchflags_args {
 2414         int     fd;
 2415         int     flags;
 2416 };
 2417 #endif
 2418 int
 2419 fchflags(td, uap)
 2420         struct thread *td;
 2421         register struct fchflags_args /* {
 2422                 int fd;
 2423                 int flags;
 2424         } */ *uap;
 2425 {
 2426         struct file *fp;
 2427         int vfslocked;
 2428         int error;
 2429 
 2430         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 2431                 return (error);
 2432         vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
 2433         error = setfflags(td, fp->f_vnode, uap->flags);
 2434         fdrop(fp, td);
 2435         VFS_UNLOCK_GIANT(vfslocked);
 2436         return (error);
 2437 }
 2438 
 2439 /*
 2440  * Common implementation code for chmod(), lchmod() and fchmod().
 2441  */
 2442 static int
 2443 setfmode(td, vp, mode)
 2444         struct thread *td;
 2445         struct vnode *vp;
 2446         int mode;
 2447 {
 2448         int error;
 2449         struct mount *mp;
 2450         struct vattr vattr;
 2451 
 2452         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 2453                 return (error);
 2454         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2455         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2456         VATTR_NULL(&vattr);
 2457         vattr.va_mode = mode & ALLPERMS;
 2458 #ifdef MAC
 2459         error = mac_check_vnode_setmode(td->td_ucred, vp, vattr.va_mode);
 2460         if (error == 0)
 2461 #endif
 2462                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 2463         VOP_UNLOCK(vp, 0, td);
 2464         vn_finished_write(mp);
 2465         return (error);
 2466 }
 2467 
 2468 /*
 2469  * Change mode of a file given path name.
 2470  */
 2471 #ifndef _SYS_SYSPROTO_H_
 2472 struct chmod_args {
 2473         char    *path;
 2474         int     mode;
 2475 };
 2476 #endif
 2477 int
 2478 chmod(td, uap)
 2479         struct thread *td;
 2480         register struct chmod_args /* {
 2481                 char *path;
 2482                 int mode;
 2483         } */ *uap;
 2484 {
 2485 
 2486         return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
 2487 }
 2488 
 2489 int
 2490 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
 2491 {
 2492         int error;
 2493         struct nameidata nd;
 2494         int vfslocked;
 2495 
 2496         NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, pathseg, path, td);
 2497         if ((error = namei(&nd)) != 0)
 2498                 return (error);
 2499         vfslocked = NDHASGIANT(&nd);
 2500         NDFREE(&nd, NDF_ONLY_PNBUF);
 2501         error = setfmode(td, nd.ni_vp, mode);
 2502         vrele(nd.ni_vp);
 2503         VFS_UNLOCK_GIANT(vfslocked);
 2504         return (error);
 2505 }
 2506 
 2507 /*
 2508  * Change mode of a file given path name (don't follow links.)
 2509  */
 2510 #ifndef _SYS_SYSPROTO_H_
 2511 struct lchmod_args {
 2512         char    *path;
 2513         int     mode;
 2514 };
 2515 #endif
 2516 int
 2517 lchmod(td, uap)
 2518         struct thread *td;
 2519         register struct lchmod_args /* {
 2520                 char *path;
 2521                 int mode;
 2522         } */ *uap;
 2523 {
 2524         int error;
 2525         struct nameidata nd;
 2526         int vfslocked;
 2527 
 2528         NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_USERSPACE, uap->path, td);
 2529         if ((error = namei(&nd)) != 0)
 2530                 return (error);
 2531         vfslocked = NDHASGIANT(&nd);
 2532         NDFREE(&nd, NDF_ONLY_PNBUF);
 2533         error = setfmode(td, nd.ni_vp, uap->mode);
 2534         vrele(nd.ni_vp);
 2535         VFS_UNLOCK_GIANT(vfslocked);
 2536         return (error);
 2537 }
 2538 
 2539 /*
 2540  * Change mode of a file given a file descriptor.
 2541  */
 2542 #ifndef _SYS_SYSPROTO_H_
 2543 struct fchmod_args {
 2544         int     fd;
 2545         int     mode;
 2546 };
 2547 #endif
 2548 int
 2549 fchmod(td, uap)
 2550         struct thread *td;
 2551         register struct fchmod_args /* {
 2552                 int fd;
 2553                 int mode;
 2554         } */ *uap;
 2555 {
 2556         struct file *fp;
 2557         int vfslocked;
 2558         int error;
 2559 
 2560         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 2561                 return (error);
 2562         vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
 2563         error = setfmode(td, fp->f_vnode, uap->mode);
 2564         fdrop(fp, td);
 2565         VFS_UNLOCK_GIANT(vfslocked);
 2566         return (error);
 2567 }
 2568 
 2569 /*
 2570  * Common implementation for chown(), lchown(), and fchown()
 2571  */
 2572 static int
 2573 setfown(td, vp, uid, gid)
 2574         struct thread *td;
 2575         struct vnode *vp;
 2576         uid_t uid;
 2577         gid_t gid;
 2578 {
 2579         int error;
 2580         struct mount *mp;
 2581         struct vattr vattr;
 2582 
 2583         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 2584                 return (error);
 2585         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2586         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2587         VATTR_NULL(&vattr);
 2588         vattr.va_uid = uid;
 2589         vattr.va_gid = gid;
 2590 #ifdef MAC
 2591         error = mac_check_vnode_setowner(td->td_ucred, vp, vattr.va_uid,
 2592             vattr.va_gid);
 2593         if (error == 0)
 2594 #endif
 2595                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 2596         VOP_UNLOCK(vp, 0, td);
 2597         vn_finished_write(mp);
 2598         return (error);
 2599 }
 2600 
 2601 /*
 2602  * Set ownership given a path name.
 2603  */
 2604 #ifndef _SYS_SYSPROTO_H_
 2605 struct chown_args {
 2606         char    *path;
 2607         int     uid;
 2608         int     gid;
 2609 };
 2610 #endif
 2611 int
 2612 chown(td, uap)
 2613         struct thread *td;
 2614         register struct chown_args /* {
 2615                 char *path;
 2616                 int uid;
 2617                 int gid;
 2618         } */ *uap;
 2619 {
 2620 
 2621         return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
 2622 }
 2623 
 2624 int
 2625 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
 2626     int gid)
 2627 {
 2628         int error;
 2629         struct nameidata nd;
 2630         int vfslocked;
 2631 
 2632         NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, pathseg, path, td);
 2633         if ((error = namei(&nd)) != 0)
 2634                 return (error);
 2635         vfslocked = NDHASGIANT(&nd);
 2636         NDFREE(&nd, NDF_ONLY_PNBUF);
 2637         error = setfown(td, nd.ni_vp, uid, gid);
 2638         vrele(nd.ni_vp);
 2639         VFS_UNLOCK_GIANT(vfslocked);
 2640         return (error);
 2641 }
 2642 
 2643 /*
 2644  * Set ownership given a path name, do not cross symlinks.
 2645  */
 2646 #ifndef _SYS_SYSPROTO_H_
 2647 struct lchown_args {
 2648         char    *path;
 2649         int     uid;
 2650         int     gid;
 2651 };
 2652 #endif
 2653 int
 2654 lchown(td, uap)
 2655         struct thread *td;
 2656         register struct lchown_args /* {
 2657                 char *path;
 2658                 int uid;
 2659                 int gid;
 2660         } */ *uap;
 2661 {
 2662 
 2663         return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
 2664 }
 2665 
 2666 int
 2667 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
 2668     int gid)
 2669 {
 2670         int error;
 2671         struct nameidata nd;
 2672         int vfslocked;
 2673 
 2674         NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE, pathseg, path, td);
 2675         if ((error = namei(&nd)) != 0)
 2676                 return (error);
 2677         vfslocked = NDHASGIANT(&nd);
 2678         NDFREE(&nd, NDF_ONLY_PNBUF);
 2679         error = setfown(td, nd.ni_vp, uid, gid);
 2680         vrele(nd.ni_vp);
 2681         VFS_UNLOCK_GIANT(vfslocked);
 2682         return (error);
 2683 }
 2684 
 2685 /*
 2686  * Set ownership given a file descriptor.
 2687  */
 2688 #ifndef _SYS_SYSPROTO_H_
 2689 struct fchown_args {
 2690         int     fd;
 2691         int     uid;
 2692         int     gid;
 2693 };
 2694 #endif
 2695 int
 2696 fchown(td, uap)
 2697         struct thread *td;
 2698         register struct fchown_args /* {
 2699                 int fd;
 2700                 int uid;
 2701                 int gid;
 2702         } */ *uap;
 2703 {
 2704         struct file *fp;
 2705         int vfslocked;
 2706         int error;
 2707 
 2708         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 2709                 return (error);
 2710         vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
 2711         error = setfown(td, fp->f_vnode, uap->uid, uap->gid);
 2712         fdrop(fp, td);
 2713         VFS_UNLOCK_GIANT(vfslocked);
 2714         return (error);
 2715 }
 2716 
 2717 /*
 2718  * Common implementation code for utimes(), lutimes(), and futimes().
 2719  */
 2720 static int
 2721 getutimes(usrtvp, tvpseg, tsp)
 2722         const struct timeval *usrtvp;
 2723         enum uio_seg tvpseg;
 2724         struct timespec *tsp;
 2725 {
 2726         struct timeval tv[2];
 2727         const struct timeval *tvp;
 2728         int error;
 2729 
 2730         if (usrtvp == NULL) {
 2731                 microtime(&tv[0]);
 2732                 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
 2733                 tsp[1] = tsp[0];
 2734         } else {
 2735                 if (tvpseg == UIO_SYSSPACE) {
 2736                         tvp = usrtvp;
 2737                 } else {
 2738                         if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
 2739                                 return (error);
 2740                         tvp = tv;
 2741                 }
 2742 
 2743                 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
 2744                 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
 2745         }
 2746         return (0);
 2747 }
 2748 
 2749 /*
 2750  * Common implementation code for utimes(), lutimes(), and futimes().
 2751  */
 2752 static int
 2753 setutimes(td, vp, ts, numtimes, nullflag)
 2754         struct thread *td;
 2755         struct vnode *vp;
 2756         const struct timespec *ts;
 2757         int numtimes;
 2758         int nullflag;
 2759 {
 2760         int error, setbirthtime;
 2761         struct mount *mp;
 2762         struct vattr vattr;
 2763 
 2764         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 2765                 return (error);
 2766         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2767         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2768         setbirthtime = 0;
 2769         if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred, td) == 0 &&
 2770             timespeccmp(&ts[1], &vattr.va_birthtime, < ))
 2771                 setbirthtime = 1;
 2772         VATTR_NULL(&vattr);
 2773         vattr.va_atime = ts[0];
 2774         vattr.va_mtime = ts[1];
 2775         if (setbirthtime)
 2776                 vattr.va_birthtime = ts[1];
 2777         if (numtimes > 2)
 2778                 vattr.va_birthtime = ts[2];
 2779         if (nullflag)
 2780                 vattr.va_vaflags |= VA_UTIMES_NULL;
 2781 #ifdef MAC
 2782         error = mac_check_vnode_setutimes(td->td_ucred, vp, vattr.va_atime,
 2783             vattr.va_mtime);
 2784 #endif
 2785         if (error == 0)
 2786                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 2787         VOP_UNLOCK(vp, 0, td);
 2788         vn_finished_write(mp);
 2789         return (error);
 2790 }
 2791 
 2792 /*
 2793  * Set the access and modification times of a file.
 2794  */
 2795 #ifndef _SYS_SYSPROTO_H_
 2796 struct utimes_args {
 2797         char    *path;
 2798         struct  timeval *tptr;
 2799 };
 2800 #endif
 2801 int
 2802 utimes(td, uap)
 2803         struct thread *td;
 2804         register struct utimes_args /* {
 2805                 char *path;
 2806                 struct timeval *tptr;
 2807         } */ *uap;
 2808 {
 2809 
 2810         return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
 2811             UIO_USERSPACE));
 2812 }
 2813 
 2814 int
 2815 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
 2816     struct timeval *tptr, enum uio_seg tptrseg)
 2817 {
 2818         struct timespec ts[2];
 2819         int error;
 2820         struct nameidata nd;
 2821         int vfslocked;
 2822 
 2823         if ((error = getutimes(tptr, tptrseg, ts)) != 0)
 2824                 return (error);
 2825         NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, pathseg, path, td);
 2826         if ((error = namei(&nd)) != 0)
 2827                 return (error);
 2828         vfslocked = NDHASGIANT(&nd);
 2829         NDFREE(&nd, NDF_ONLY_PNBUF);
 2830         error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
 2831         vrele(nd.ni_vp);
 2832         VFS_UNLOCK_GIANT(vfslocked);
 2833         return (error);
 2834 }
 2835 
 2836 /*
 2837  * Set the access and modification times of a file.
 2838  */
 2839 #ifndef _SYS_SYSPROTO_H_
 2840 struct lutimes_args {
 2841         char    *path;
 2842         struct  timeval *tptr;
 2843 };
 2844 #endif
 2845 int
 2846 lutimes(td, uap)
 2847         struct thread *td;
 2848         register struct lutimes_args /* {
 2849                 char *path;
 2850                 struct timeval *tptr;
 2851         } */ *uap;
 2852 {
 2853 
 2854         return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
 2855             UIO_USERSPACE));
 2856 }
 2857 
 2858 int
 2859 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
 2860     struct timeval *tptr, enum uio_seg tptrseg)
 2861 {
 2862         struct timespec ts[2];
 2863         int error;
 2864         struct nameidata nd;
 2865         int vfslocked;
 2866 
 2867         if ((error = getutimes(tptr, tptrseg, ts)) != 0)
 2868                 return (error);
 2869         NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE, pathseg, path, td);
 2870         if ((error = namei(&nd)) != 0)
 2871                 return (error);
 2872         vfslocked = NDHASGIANT(&nd);
 2873         NDFREE(&nd, NDF_ONLY_PNBUF);
 2874         error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
 2875         vrele(nd.ni_vp);
 2876         VFS_UNLOCK_GIANT(vfslocked);
 2877         return (error);
 2878 }
 2879 
 2880 /*
 2881  * Set the access and modification times of a file.
 2882  */
 2883 #ifndef _SYS_SYSPROTO_H_
 2884 struct futimes_args {
 2885         int     fd;
 2886         struct  timeval *tptr;
 2887 };
 2888 #endif
 2889 int
 2890 futimes(td, uap)
 2891         struct thread *td;
 2892         register struct futimes_args /* {
 2893                 int  fd;
 2894                 struct timeval *tptr;
 2895         } */ *uap;
 2896 {
 2897 
 2898         return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
 2899 }
 2900 
 2901 int
 2902 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
 2903     enum uio_seg tptrseg)
 2904 {
 2905         struct timespec ts[2];
 2906         struct file *fp;
 2907         int vfslocked;
 2908         int error;
 2909 
 2910         if ((error = getutimes(tptr, tptrseg, ts)) != 0)
 2911                 return (error);
 2912         if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
 2913                 return (error);
 2914         vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
 2915         error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
 2916         fdrop(fp, td);
 2917         VFS_UNLOCK_GIANT(vfslocked);
 2918         return (error);
 2919 }
 2920 
 2921 /*
 2922  * Truncate a file given its path name.
 2923  */
 2924 #ifndef _SYS_SYSPROTO_H_
 2925 struct truncate_args {
 2926         char    *path;
 2927         int     pad;
 2928         off_t   length;
 2929 };
 2930 #endif
 2931 int
 2932 truncate(td, uap)
 2933         struct thread *td;
 2934         register struct truncate_args /* {
 2935                 char *path;
 2936                 int pad;
 2937                 off_t length;
 2938         } */ *uap;
 2939 {
 2940 
 2941         return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
 2942 }
 2943 
 2944 int
 2945 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
 2946 {
 2947         struct mount *mp;
 2948         struct vnode *vp;
 2949         struct vattr vattr;
 2950         int error;
 2951         struct nameidata nd;
 2952         int vfslocked;
 2953 
 2954         if (length < 0)
 2955                 return(EINVAL);
 2956         NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, pathseg, path, td);
 2957         if ((error = namei(&nd)) != 0)
 2958                 return (error);
 2959         vfslocked = NDHASGIANT(&nd);
 2960         vp = nd.ni_vp;
 2961         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
 2962                 vrele(vp);
 2963                 VFS_UNLOCK_GIANT(vfslocked);
 2964                 return (error);
 2965         }
 2966         NDFREE(&nd, NDF_ONLY_PNBUF);
 2967         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2968         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2969         if (vp->v_type == VDIR)
 2970                 error = EISDIR;
 2971 #ifdef MAC
 2972         else if ((error = mac_check_vnode_write(td->td_ucred, NOCRED, vp))) {
 2973         }
 2974 #endif
 2975         else if ((error = vn_writechk(vp)) == 0 &&
 2976             (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
 2977                 VATTR_NULL(&vattr);
 2978                 vattr.va_size = length;
 2979                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 2980         }
 2981         vput(vp);
 2982         vn_finished_write(mp);
 2983         VFS_UNLOCK_GIANT(vfslocked);
 2984         return (error);
 2985 }
 2986 
 2987 /*
 2988  * Truncate a file given a file descriptor.
 2989  */
 2990 #ifndef _SYS_SYSPROTO_H_
 2991 struct ftruncate_args {
 2992         int     fd;
 2993         int     pad;
 2994         off_t   length;
 2995 };
 2996 #endif
 2997 int
 2998 ftruncate(td, uap)
 2999         struct thread *td;
 3000         register struct ftruncate_args /* {
 3001                 int fd;
 3002                 int pad;
 3003                 off_t length;
 3004         } */ *uap;
 3005 {
 3006         struct mount *mp;
 3007         struct vattr vattr;
 3008         struct vnode *vp;
 3009         struct file *fp;
 3010         int vfslocked;
 3011         int error;
 3012 
 3013         if (uap->length < 0)
 3014                 return(EINVAL);
 3015         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 3016                 return (error);
 3017         if ((fp->f_flag & FWRITE) == 0) {
 3018                 fdrop(fp, td);
 3019                 return (EINVAL);
 3020         }
 3021         vp = fp->f_vnode;
 3022         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
 3023         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 3024                 goto drop;
 3025         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 3026         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3027         if (vp->v_type == VDIR)
 3028                 error = EISDIR;
 3029 #ifdef MAC
 3030         else if ((error = mac_check_vnode_write(td->td_ucred, fp->f_cred,
 3031             vp))) {
 3032         }
 3033 #endif
 3034         else if ((error = vn_writechk(vp)) == 0) {
 3035                 VATTR_NULL(&vattr);
 3036                 vattr.va_size = uap->length;
 3037                 error = VOP_SETATTR(vp, &vattr, fp->f_cred, td);
 3038         }
 3039         VOP_UNLOCK(vp, 0, td);
 3040         vn_finished_write(mp);
 3041 drop:
 3042         VFS_UNLOCK_GIANT(vfslocked);
 3043         fdrop(fp, td);
 3044         return (error);
 3045 }
 3046 
 3047 #if defined(COMPAT_43)
 3048 /*
 3049  * Truncate a file given its path name.
 3050  */
 3051 #ifndef _SYS_SYSPROTO_H_
 3052 struct otruncate_args {
 3053         char    *path;
 3054         long    length;
 3055 };
 3056 #endif
 3057 int
 3058 otruncate(td, uap)
 3059         struct thread *td;
 3060         register struct otruncate_args /* {
 3061                 char *path;
 3062                 long length;
 3063         } */ *uap;
 3064 {
 3065         struct truncate_args /* {
 3066                 char *path;
 3067                 int pad;
 3068                 off_t length;
 3069         } */ nuap;
 3070 
 3071         nuap.path = uap->path;
 3072         nuap.length = uap->length;
 3073         return (truncate(td, &nuap));
 3074 }
 3075 
 3076 /*
 3077  * Truncate a file given a file descriptor.
 3078  */
 3079 #ifndef _SYS_SYSPROTO_H_
 3080 struct oftruncate_args {
 3081         int     fd;
 3082         long    length;
 3083 };
 3084 #endif
 3085 int
 3086 oftruncate(td, uap)
 3087         struct thread *td;
 3088         register struct oftruncate_args /* {
 3089                 int fd;
 3090                 long length;
 3091         } */ *uap;
 3092 {
 3093         struct ftruncate_args /* {
 3094                 int fd;
 3095                 int pad;
 3096                 off_t length;
 3097         } */ nuap;
 3098 
 3099         nuap.fd = uap->fd;
 3100         nuap.length = uap->length;
 3101         return (ftruncate(td, &nuap));
 3102 }
 3103 #endif /* COMPAT_43 */
 3104 
 3105 /*
 3106  * Sync an open file.
 3107  */
 3108 #ifndef _SYS_SYSPROTO_H_
 3109 struct fsync_args {
 3110         int     fd;
 3111 };
 3112 #endif
 3113 int
 3114 fsync(td, uap)
 3115         struct thread *td;
 3116         struct fsync_args /* {
 3117                 int fd;
 3118         } */ *uap;
 3119 {
 3120         struct vnode *vp;
 3121         struct mount *mp;
 3122         struct file *fp;
 3123         int vfslocked;
 3124         int error;
 3125 
 3126         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 3127                 return (error);
 3128         vp = fp->f_vnode;
 3129         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
 3130         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 3131                 goto drop;
 3132         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3133         if (vp->v_object != NULL) {
 3134                 VM_OBJECT_LOCK(vp->v_object);
 3135                 vm_object_page_clean(vp->v_object, 0, 0, 0);
 3136                 VM_OBJECT_UNLOCK(vp->v_object);
 3137         }
 3138         error = VOP_FSYNC(vp, MNT_WAIT, td);
 3139 
 3140         VOP_UNLOCK(vp, 0, td);
 3141         vn_finished_write(mp);
 3142 drop:
 3143         VFS_UNLOCK_GIANT(vfslocked);
 3144         fdrop(fp, td);
 3145         return (error);
 3146 }
 3147 
 3148 /*
 3149  * Rename files.  Source and destination must either both be directories,
 3150  * or both not be directories.  If target is a directory, it must be empty.
 3151  */
 3152 #ifndef _SYS_SYSPROTO_H_
 3153 struct rename_args {
 3154         char    *from;
 3155         char    *to;
 3156 };
 3157 #endif
 3158 int
 3159 rename(td, uap)
 3160         struct thread *td;
 3161         register struct rename_args /* {
 3162                 char *from;
 3163                 char *to;
 3164         } */ *uap;
 3165 {
 3166 
 3167         return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
 3168 }
 3169 
 3170 int
 3171 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
 3172 {
 3173         struct mount *mp = NULL;
 3174         struct vnode *tvp, *fvp, *tdvp;
 3175         struct nameidata fromnd, tond;
 3176         int tvfslocked;
 3177         int fvfslocked;
 3178         int error;
 3179 
 3180         bwillwrite();
 3181 #ifdef MAC
 3182         NDINIT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | MPSAFE,
 3183             pathseg, from, td);
 3184 #else
 3185         NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE,
 3186             pathseg, from, td);
 3187 #endif
 3188         if ((error = namei(&fromnd)) != 0)
 3189                 return (error);
 3190         fvfslocked = NDHASGIANT(&fromnd);
 3191         tvfslocked = 0;
 3192 #ifdef MAC
 3193         error = mac_check_vnode_rename_from(td->td_ucred, fromnd.ni_dvp,
 3194             fromnd.ni_vp, &fromnd.ni_cnd);
 3195         VOP_UNLOCK(fromnd.ni_dvp, 0, td);
 3196         VOP_UNLOCK(fromnd.ni_vp, 0, td);
 3197 #endif
 3198         fvp = fromnd.ni_vp;
 3199         if (error == 0)
 3200                 error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
 3201         if (error != 0) {
 3202                 NDFREE(&fromnd, NDF_ONLY_PNBUF);
 3203                 vrele(fromnd.ni_dvp);
 3204                 vrele(fvp);
 3205                 goto out1;
 3206         }
 3207         NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
 3208             MPSAFE, pathseg, to, td);
 3209         if (fromnd.ni_vp->v_type == VDIR)
 3210                 tond.ni_cnd.cn_flags |= WILLBEDIR;
 3211         if ((error = namei(&tond)) != 0) {
 3212                 /* Translate error code for rename("dir1", "dir2/."). */
 3213                 if (error == EISDIR && fvp->v_type == VDIR)
 3214                         error = EINVAL;
 3215                 NDFREE(&fromnd, NDF_ONLY_PNBUF);
 3216                 vrele(fromnd.ni_dvp);
 3217                 vrele(fvp);
 3218                 vn_finished_write(mp);
 3219                 goto out1;
 3220         }
 3221         tvfslocked = NDHASGIANT(&tond);
 3222         tdvp = tond.ni_dvp;
 3223         tvp = tond.ni_vp;
 3224         if (tvp != NULL) {
 3225                 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
 3226                         error = ENOTDIR;
 3227                         goto out;
 3228                 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
 3229                         error = EISDIR;
 3230                         goto out;
 3231                 }
 3232         }
 3233         if (fvp == tdvp)
 3234                 error = EINVAL;
 3235         /*
 3236          * If the source is the same as the destination (that is, if they
 3237          * are links to the same vnode), then there is nothing to do.
 3238          */
 3239         if (fvp == tvp)
 3240                 error = -1;
 3241 #ifdef MAC
 3242         else
 3243                 error = mac_check_vnode_rename_to(td->td_ucred, tdvp,
 3244                     tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
 3245 #endif
 3246 out:
 3247         if (!error) {
 3248                 VOP_LEASE(tdvp, td, td->td_ucred, LEASE_WRITE);
 3249                 if (fromnd.ni_dvp != tdvp) {
 3250                         VOP_LEASE(fromnd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 3251                 }
 3252                 if (tvp) {
 3253                         VOP_LEASE(tvp, td, td->td_ucred, LEASE_WRITE);
 3254                 }
 3255                 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
 3256                                    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
 3257                 NDFREE(&fromnd, NDF_ONLY_PNBUF);
 3258                 NDFREE(&tond, NDF_ONLY_PNBUF);
 3259         } else {
 3260                 NDFREE(&fromnd, NDF_ONLY_PNBUF);
 3261                 NDFREE(&tond, NDF_ONLY_PNBUF);
 3262                 if (tvp)
 3263                         vput(tvp);
 3264                 if (tdvp == tvp)
 3265                         vrele(tdvp);
 3266                 else
 3267                         vput(tdvp);
 3268                 vrele(fromnd.ni_dvp);
 3269                 vrele(fvp);
 3270         }
 3271         vrele(tond.ni_startdir);
 3272         vn_finished_write(mp);
 3273 out1:
 3274         if (fromnd.ni_startdir)
 3275                 vrele(fromnd.ni_startdir);
 3276         VFS_UNLOCK_GIANT(fvfslocked);
 3277         VFS_UNLOCK_GIANT(tvfslocked);
 3278         if (error == -1)
 3279                 return (0);
 3280         return (error);
 3281 }
 3282 
 3283 /*
 3284  * Make a directory file.
 3285  */
 3286 #ifndef _SYS_SYSPROTO_H_
 3287 struct mkdir_args {
 3288         char    *path;
 3289         int     mode;
 3290 };
 3291 #endif
 3292 int
 3293 mkdir(td, uap)
 3294         struct thread *td;
 3295         register struct mkdir_args /* {
 3296                 char *path;
 3297                 int mode;
 3298         } */ *uap;
 3299 {
 3300 
 3301         return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
 3302 }
 3303 
 3304 int
 3305 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
 3306 {
 3307         struct mount *mp;
 3308         struct vnode *vp;
 3309         struct vattr vattr;
 3310         int error;
 3311         struct nameidata nd;
 3312         int vfslocked;
 3313 
 3314 restart:
 3315         bwillwrite();
 3316         NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE, segflg, path, td);
 3317         nd.ni_cnd.cn_flags |= WILLBEDIR;
 3318         if ((error = namei(&nd)) != 0)
 3319                 return (error);
 3320         vfslocked = NDHASGIANT(&nd);
 3321         vp = nd.ni_vp;
 3322         if (vp != NULL) {
 3323                 NDFREE(&nd, NDF_ONLY_PNBUF);
 3324                 vrele(vp);
 3325                 /*
 3326                  * XXX namei called with LOCKPARENT but not LOCKLEAF has
 3327                  * the strange behaviour of leaving the vnode unlocked
 3328                  * if the target is the same vnode as the parent.
 3329                  */
 3330                 if (vp == nd.ni_dvp)
 3331                         vrele(nd.ni_dvp);
 3332                 else
 3333                         vput(nd.ni_dvp);
 3334                 VFS_UNLOCK_GIANT(vfslocked);
 3335                 return (EEXIST);
 3336         }
 3337         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 3338                 NDFREE(&nd, NDF_ONLY_PNBUF);
 3339                 vput(nd.ni_dvp);
 3340                 VFS_UNLOCK_GIANT(vfslocked);
 3341                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 3342                         return (error);
 3343                 goto restart;
 3344         }
 3345         VATTR_NULL(&vattr);
 3346         vattr.va_type = VDIR;
 3347         FILEDESC_LOCK_FAST(td->td_proc->p_fd);
 3348         vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
 3349         FILEDESC_UNLOCK_FAST(td->td_proc->p_fd);
 3350 #ifdef MAC
 3351         error = mac_check_vnode_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
 3352             &vattr);
 3353         if (error)
 3354                 goto out;
 3355 #endif
 3356         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 3357         error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
 3358 #ifdef MAC
 3359 out:
 3360 #endif
 3361         NDFREE(&nd, NDF_ONLY_PNBUF);
 3362         vput(nd.ni_dvp);
 3363         if (!error)
 3364                 vput(nd.ni_vp);
 3365         vn_finished_write(mp);
 3366         VFS_UNLOCK_GIANT(vfslocked);
 3367         return (error);
 3368 }
 3369 
 3370 /*
 3371  * Remove a directory file.
 3372  */
 3373 #ifndef _SYS_SYSPROTO_H_
 3374 struct rmdir_args {
 3375         char    *path;
 3376 };
 3377 #endif
 3378 int
 3379 rmdir(td, uap)
 3380         struct thread *td;
 3381         struct rmdir_args /* {
 3382                 char *path;
 3383         } */ *uap;
 3384 {
 3385 
 3386         return (kern_rmdir(td, uap->path, UIO_USERSPACE));
 3387 }
 3388 
 3389 int
 3390 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
 3391 {
 3392         struct mount *mp;
 3393         struct vnode *vp;
 3394         int error;
 3395         struct nameidata nd;
 3396         int vfslocked;
 3397 
 3398 restart:
 3399         bwillwrite();
 3400         NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE, pathseg, path, td);
 3401         if ((error = namei(&nd)) != 0)
 3402                 return (error);
 3403         vfslocked = NDHASGIANT(&nd);
 3404         vp = nd.ni_vp;
 3405         if (vp->v_type != VDIR) {
 3406                 error = ENOTDIR;
 3407                 goto out;
 3408         }
 3409         /*
 3410          * No rmdir "." please.
 3411          */
 3412         if (nd.ni_dvp == vp) {
 3413                 error = EINVAL;
 3414                 goto out;
 3415         }
 3416         /*
 3417          * The root of a mounted filesystem cannot be deleted.
 3418          */
 3419         if (vp->v_vflag & VV_ROOT) {
 3420                 error = EBUSY;
 3421                 goto out;
 3422         }
 3423 #ifdef MAC
 3424         error = mac_check_vnode_delete(td->td_ucred, nd.ni_dvp, vp,
 3425             &nd.ni_cnd);
 3426         if (error)
 3427                 goto out;
 3428 #endif
 3429         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 3430                 NDFREE(&nd, NDF_ONLY_PNBUF);
 3431                 if (nd.ni_dvp == vp)
 3432                         vrele(nd.ni_dvp);
 3433                 else
 3434                         vput(nd.ni_dvp);
 3435                 vput(vp);
 3436                 VFS_UNLOCK_GIANT(vfslocked);
 3437                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 3438                         return (error);
 3439                 goto restart;
 3440         }
 3441         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 3442         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 3443         error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
 3444         vn_finished_write(mp);
 3445 out:
 3446         NDFREE(&nd, NDF_ONLY_PNBUF);
 3447         if (nd.ni_dvp == vp)
 3448                 vrele(nd.ni_dvp);
 3449         else
 3450                 vput(nd.ni_dvp);
 3451         vput(vp);
 3452         VFS_UNLOCK_GIANT(vfslocked);
 3453         return (error);
 3454 }
 3455 
 3456 #ifdef COMPAT_43
 3457 /*
 3458  * Read a block of directory entries in a filesystem independent format.
 3459  */
 3460 #ifndef _SYS_SYSPROTO_H_
 3461 struct ogetdirentries_args {
 3462         int     fd;
 3463         char    *buf;
 3464         u_int   count;
 3465         long    *basep;
 3466 };
 3467 #endif
 3468 int
 3469 ogetdirentries(td, uap)
 3470         struct thread *td;
 3471         register struct ogetdirentries_args /* {
 3472                 int fd;
 3473                 char *buf;
 3474                 u_int count;
 3475                 long *basep;
 3476         } */ *uap;
 3477 {
 3478         struct vnode *vp;
 3479         struct file *fp;
 3480         struct uio auio, kuio;
 3481         struct iovec aiov, kiov;
 3482         struct dirent *dp, *edp;
 3483         caddr_t dirbuf;
 3484         int error, eofflag, readcnt;
 3485         long loff;
 3486 
 3487         /* XXX arbitrary sanity limit on `count'. */
 3488         if (uap->count > 64 * 1024)
 3489                 return (EINVAL);
 3490         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 3491                 return (error);
 3492         if ((fp->f_flag & FREAD) == 0) {
 3493                 fdrop(fp, td);
 3494                 return (EBADF);
 3495         }
 3496         vp = fp->f_vnode;
 3497 unionread:
 3498         if (vp->v_type != VDIR) {
 3499                 fdrop(fp, td);
 3500                 return (EINVAL);
 3501         }
 3502         aiov.iov_base = uap->buf;
 3503         aiov.iov_len = uap->count;
 3504         auio.uio_iov = &aiov;
 3505         auio.uio_iovcnt = 1;
 3506         auio.uio_rw = UIO_READ;
 3507         auio.uio_segflg = UIO_USERSPACE;
 3508         auio.uio_td = td;
 3509         auio.uio_resid = uap->count;
 3510         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3511         loff = auio.uio_offset = fp->f_offset;
 3512 #ifdef MAC
 3513         error = mac_check_vnode_readdir(td->td_ucred, vp);
 3514         if (error) {
 3515                 VOP_UNLOCK(vp, 0, td);
 3516                 fdrop(fp, td);
 3517                 return (error);
 3518         }
 3519 #endif
 3520 #       if (BYTE_ORDER != LITTLE_ENDIAN)
 3521                 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
 3522                         error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
 3523                             NULL, NULL);
 3524                         fp->f_offset = auio.uio_offset;
 3525                 } else
 3526 #       endif
 3527         {
 3528                 kuio = auio;
 3529                 kuio.uio_iov = &kiov;
 3530                 kuio.uio_segflg = UIO_SYSSPACE;
 3531                 kiov.iov_len = uap->count;
 3532                 MALLOC(dirbuf, caddr_t, uap->count, M_TEMP, M_WAITOK);
 3533                 kiov.iov_base = dirbuf;
 3534                 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
 3535                             NULL, NULL);
 3536                 fp->f_offset = kuio.uio_offset;
 3537                 if (error == 0) {
 3538                         readcnt = uap->count - kuio.uio_resid;
 3539                         edp = (struct dirent *)&dirbuf[readcnt];
 3540                         for (dp = (struct dirent *)dirbuf; dp < edp; ) {
 3541 #                               if (BYTE_ORDER == LITTLE_ENDIAN)
 3542                                         /*
 3543                                          * The expected low byte of
 3544                                          * dp->d_namlen is our dp->d_type.
 3545                                          * The high MBZ byte of dp->d_namlen
 3546                                          * is our dp->d_namlen.
 3547                                          */
 3548                                         dp->d_type = dp->d_namlen;
 3549                                         dp->d_namlen = 0;
 3550 #                               else
 3551                                         /*
 3552                                          * The dp->d_type is the high byte
 3553                                          * of the expected dp->d_namlen,
 3554                                          * so must be zero'ed.
 3555                                          */
 3556                                         dp->d_type = 0;
 3557 #                               endif
 3558                                 if (dp->d_reclen > 0) {
 3559                                         dp = (struct dirent *)
 3560                                             ((char *)dp + dp->d_reclen);
 3561                                 } else {
 3562                                         error = EIO;
 3563                                         break;
 3564                                 }
 3565                         }
 3566                         if (dp >= edp)
 3567                                 error = uiomove(dirbuf, readcnt, &auio);
 3568                 }
 3569                 FREE(dirbuf, M_TEMP);
 3570         }
 3571         VOP_UNLOCK(vp, 0, td);
 3572         if (error) {
 3573                 fdrop(fp, td);
 3574                 return (error);
 3575         }
 3576         if (uap->count == auio.uio_resid) {
 3577                 if (union_dircheckp) {
 3578                         error = union_dircheckp(td, &vp, fp);
 3579                         if (error == -1)
 3580                                 goto unionread;
 3581                         if (error) {
 3582                                 fdrop(fp, td);
 3583                                 return (error);
 3584                         }
 3585                 }
 3586                 /*
 3587                  * XXX We could delay dropping the lock above but
 3588                  * union_dircheckp complicates things.
 3589                  */
 3590                 vn_lock(vp, LK_EXCLUSIVE|LK_RETRY, td);
 3591                 if ((vp->v_vflag & VV_ROOT) &&
 3592                     (vp->v_mount->mnt_flag & MNT_UNION)) {
 3593                         struct vnode *tvp = vp;
 3594                         vp = vp->v_mount->mnt_vnodecovered;
 3595                         VREF(vp);
 3596                         fp->f_vnode = vp;
 3597                         fp->f_data = vp;
 3598                         fp->f_offset = 0;
 3599                         vput(tvp);
 3600                         goto unionread;
 3601                 }
 3602                 VOP_UNLOCK(vp, 0, td);
 3603         }
 3604         error = copyout(&loff, uap->basep, sizeof(long));
 3605         fdrop(fp, td);
 3606         td->td_retval[0] = uap->count - auio.uio_resid;
 3607         return (error);
 3608 }
 3609 #endif /* COMPAT_43 */
 3610 
 3611 /*
 3612  * Read a block of directory entries in a filesystem independent format.
 3613  */
 3614 #ifndef _SYS_SYSPROTO_H_
 3615 struct getdirentries_args {
 3616         int     fd;
 3617         char    *buf;
 3618         u_int   count;
 3619         long    *basep;
 3620 };
 3621 #endif
 3622 int
 3623 getdirentries(td, uap)
 3624         struct thread *td;
 3625         register struct getdirentries_args /* {
 3626                 int fd;
 3627                 char *buf;
 3628                 u_int count;
 3629                 long *basep;
 3630         } */ *uap;
 3631 {
 3632         struct vnode *vp;
 3633         struct file *fp;
 3634         struct uio auio;
 3635         struct iovec aiov;
 3636         int vfslocked;
 3637         long loff;
 3638         int error, eofflag;
 3639 
 3640         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 3641                 return (error);
 3642         if ((fp->f_flag & FREAD) == 0) {
 3643                 fdrop(fp, td);
 3644                 return (EBADF);
 3645         }
 3646         vp = fp->f_vnode;
 3647 unionread:
 3648         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
 3649         if (vp->v_type != VDIR) {
 3650                 error = EINVAL;
 3651                 goto fail;
 3652         }
 3653         aiov.iov_base = uap->buf;
 3654         aiov.iov_len = uap->count;
 3655         auio.uio_iov = &aiov;
 3656         auio.uio_iovcnt = 1;
 3657         auio.uio_rw = UIO_READ;
 3658         auio.uio_segflg = UIO_USERSPACE;
 3659         auio.uio_td = td;
 3660         auio.uio_resid = uap->count;
 3661         /* vn_lock(vp, LK_SHARED | LK_RETRY, td); */
 3662         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3663         loff = auio.uio_offset = fp->f_offset;
 3664 #ifdef MAC
 3665         error = mac_check_vnode_readdir(td->td_ucred, vp);
 3666         if (error == 0)
 3667 #endif
 3668                 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
 3669                     NULL);
 3670         fp->f_offset = auio.uio_offset;
 3671         VOP_UNLOCK(vp, 0, td);
 3672         if (error)
 3673                 goto fail;
 3674         if (uap->count == auio.uio_resid) {
 3675                 if (union_dircheckp) {
 3676                         error = union_dircheckp(td, &vp, fp);
 3677                         if (error == -1) {
 3678                                 VFS_UNLOCK_GIANT(vfslocked);
 3679                                 goto unionread;
 3680                         }
 3681                         if (error)
 3682                                 goto fail;
 3683                 }
 3684                 /*
 3685                  * XXX We could delay dropping the lock above but
 3686                  * union_dircheckp complicates things.
 3687                  */
 3688                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3689                 if ((vp->v_vflag & VV_ROOT) &&
 3690                     (vp->v_mount->mnt_flag & MNT_UNION)) {
 3691                         struct vnode *tvp = vp;
 3692                         vp = vp->v_mount->mnt_vnodecovered;
 3693                         VREF(vp);
 3694                         fp->f_vnode = vp;
 3695                         fp->f_data = vp;
 3696                         fp->f_offset = 0;
 3697                         vput(tvp);
 3698                         VFS_UNLOCK_GIANT(vfslocked);
 3699                         goto unionread;
 3700                 }
 3701                 VOP_UNLOCK(vp, 0, td);
 3702         }
 3703         if (uap->basep != NULL) {
 3704                 error = copyout(&loff, uap->basep, sizeof(long));
 3705         }
 3706         td->td_retval[0] = uap->count - auio.uio_resid;
 3707 fail:
 3708         VFS_UNLOCK_GIANT(vfslocked);
 3709         fdrop(fp, td);
 3710         return (error);
 3711 }
 3712 #ifndef _SYS_SYSPROTO_H_
 3713 struct getdents_args {
 3714         int fd;
 3715         char *buf;
 3716         size_t count;
 3717 };
 3718 #endif
 3719 int
 3720 getdents(td, uap)
 3721         struct thread *td;
 3722         register struct getdents_args /* {
 3723                 int fd;
 3724                 char *buf;
 3725                 u_int count;
 3726         } */ *uap;
 3727 {
 3728         struct getdirentries_args ap;
 3729         ap.fd = uap->fd;
 3730         ap.buf = uap->buf;
 3731         ap.count = uap->count;
 3732         ap.basep = NULL;
 3733         return (getdirentries(td, &ap));
 3734 }
 3735 
 3736 /*
 3737  * Set the mode mask for creation of filesystem nodes.
 3738  *
 3739  * MP SAFE
 3740  */
 3741 #ifndef _SYS_SYSPROTO_H_
 3742 struct umask_args {
 3743         int     newmask;
 3744 };
 3745 #endif
 3746 int
 3747 umask(td, uap)
 3748         struct thread *td;
 3749         struct umask_args /* {
 3750                 int newmask;
 3751         } */ *uap;
 3752 {
 3753         register struct filedesc *fdp;
 3754 
 3755         FILEDESC_LOCK_FAST(td->td_proc->p_fd);
 3756         fdp = td->td_proc->p_fd;
 3757         td->td_retval[0] = fdp->fd_cmask;
 3758         fdp->fd_cmask = uap->newmask & ALLPERMS;
 3759         FILEDESC_UNLOCK_FAST(td->td_proc->p_fd);
 3760         return (0);
 3761 }
 3762 
 3763 /*
 3764  * Void all references to file by ripping underlying filesystem
 3765  * away from vnode.
 3766  */
 3767 #ifndef _SYS_SYSPROTO_H_
 3768 struct revoke_args {
 3769         char    *path;
 3770 };
 3771 #endif
 3772 int
 3773 revoke(td, uap)
 3774         struct thread *td;
 3775         register struct revoke_args /* {
 3776                 char *path;
 3777         } */ *uap;
 3778 {
 3779         struct vnode *vp;
 3780         struct vattr vattr;
 3781         int error;
 3782         struct nameidata nd;
 3783         int vfslocked;
 3784 
 3785         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, UIO_USERSPACE,
 3786             uap->path, td);
 3787         if ((error = namei(&nd)) != 0)
 3788                 return (error);
 3789         vfslocked = NDHASGIANT(&nd);
 3790         vp = nd.ni_vp;
 3791         NDFREE(&nd, NDF_ONLY_PNBUF);
 3792         if (vp->v_type != VCHR) {
 3793                 error = EINVAL;
 3794                 goto out;
 3795         }
 3796 #ifdef MAC
 3797         error = mac_check_vnode_revoke(td->td_ucred, vp);
 3798         if (error)
 3799                 goto out;
 3800 #endif
 3801         error = VOP_GETATTR(vp, &vattr, td->td_ucred, td);
 3802         if (error)
 3803                 goto out;
 3804         if (td->td_ucred->cr_uid != vattr.va_uid) {
 3805                 error = suser_cred(td->td_ucred, SUSER_ALLOWJAIL);
 3806                 if (error)
 3807                         goto out;
 3808         }
 3809         if (vcount(vp) > 1)
 3810                 VOP_REVOKE(vp, REVOKEALL);
 3811 out:
 3812         vput(vp);
 3813         VFS_UNLOCK_GIANT(vfslocked);
 3814         return (error);
 3815 }
 3816 
 3817 /*
 3818  * Convert a user file descriptor to a kernel file entry.
 3819  * A reference on the file entry is held upon returning.
 3820  */
 3821 int
 3822 getvnode(fdp, fd, fpp)
 3823         struct filedesc *fdp;
 3824         int fd;
 3825         struct file **fpp;
 3826 {
 3827         int error;
 3828         struct file *fp;
 3829 
 3830         fp = NULL;
 3831         if (fdp == NULL)
 3832                 error = EBADF;
 3833         else {
 3834                 FILEDESC_LOCK(fdp);
 3835                 if ((u_int)fd >= fdp->fd_nfiles ||
 3836                     (fp = fdp->fd_ofiles[fd]) == NULL)
 3837                         error = EBADF;
 3838                 else if (fp->f_vnode == NULL) {
 3839                         fp = NULL;
 3840                         error = EINVAL;
 3841                 } else {
 3842                         fhold(fp);
 3843                         error = 0;
 3844                 }
 3845                 FILEDESC_UNLOCK(fdp);
 3846         }
 3847         *fpp = fp;
 3848         return (error);
 3849 }
 3850 
 3851 /*
 3852  * Get (NFS) file handle
 3853  */
 3854 #ifndef _SYS_SYSPROTO_H_
 3855 struct lgetfh_args {
 3856         char    *fname;
 3857         fhandle_t *fhp;
 3858 };
 3859 #endif
 3860 int
 3861 lgetfh(td, uap)
 3862         struct thread *td;
 3863         register struct lgetfh_args *uap;
 3864 {
 3865         struct nameidata nd;
 3866         fhandle_t fh;
 3867         register struct vnode *vp;
 3868         int vfslocked;
 3869         int error;
 3870 
 3871         error = suser(td);
 3872         if (error)
 3873                 return (error);
 3874         NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE,
 3875             UIO_USERSPACE, uap->fname, td);
 3876         error = namei(&nd);
 3877         if (error)
 3878                 return (error);
 3879         vfslocked = NDHASGIANT(&nd);
 3880         NDFREE(&nd, NDF_ONLY_PNBUF);
 3881         vp = nd.ni_vp;
 3882         bzero(&fh, sizeof(fh));
 3883         fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
 3884         error = VFS_VPTOFH(vp, &fh.fh_fid);
 3885         vput(vp);
 3886         VFS_UNLOCK_GIANT(vfslocked);
 3887         if (error)
 3888                 return (error);
 3889         error = copyout(&fh, uap->fhp, sizeof (fh));
 3890         return (error);
 3891 }
 3892 
 3893 #ifndef _SYS_SYSPROTO_H_
 3894 struct getfh_args {
 3895         char    *fname;
 3896         fhandle_t *fhp;
 3897 };
 3898 #endif
 3899 int
 3900 getfh(td, uap)
 3901         struct thread *td;
 3902         register struct getfh_args *uap;
 3903 {
 3904         struct nameidata nd;
 3905         fhandle_t fh;
 3906         register struct vnode *vp;
 3907         int vfslocked;
 3908         int error;
 3909 
 3910         error = suser(td);
 3911         if (error)
 3912                 return (error);
 3913         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE,
 3914             UIO_USERSPACE, uap->fname, td);
 3915         error = namei(&nd);
 3916         if (error)
 3917                 return (error);
 3918         vfslocked = NDHASGIANT(&nd);
 3919         NDFREE(&nd, NDF_ONLY_PNBUF);
 3920         vp = nd.ni_vp;
 3921         bzero(&fh, sizeof(fh));
 3922         fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
 3923         error = VFS_VPTOFH(vp, &fh.fh_fid);
 3924         vput(vp);
 3925         VFS_UNLOCK_GIANT(vfslocked);
 3926         if (error)
 3927                 return (error);
 3928         error = copyout(&fh, uap->fhp, sizeof (fh));
 3929         return (error);
 3930 }
 3931 
 3932 /*
 3933  * syscall for the rpc.lockd to use to translate a NFS file handle into
 3934  * an open descriptor.
 3935  *
 3936  * warning: do not remove the suser() call or this becomes one giant
 3937  * security hole.
 3938  *
 3939  * MP SAFE
 3940  */
 3941 #ifndef _SYS_SYSPROTO_H_
 3942 struct fhopen_args {
 3943         const struct fhandle *u_fhp;
 3944         int flags;
 3945 };
 3946 #endif
 3947 int
 3948 fhopen(td, uap)
 3949         struct thread *td;
 3950         struct fhopen_args /* {
 3951                 const struct fhandle *u_fhp;
 3952                 int flags;
 3953         } */ *uap;
 3954 {
 3955         struct proc *p = td->td_proc;
 3956         struct mount *mp;
 3957         struct vnode *vp;
 3958         struct fhandle fhp;
 3959         struct vattr vat;
 3960         struct vattr *vap = &vat;
 3961         struct flock lf;
 3962         struct file *fp;
 3963         register struct filedesc *fdp = p->p_fd;
 3964         int fmode, mode, error, type;
 3965         struct file *nfp;
 3966         int indx;
 3967 
 3968         error = suser(td);
 3969         if (error)
 3970                 return (error);
 3971         fmode = FFLAGS(uap->flags);
 3972         /* why not allow a non-read/write open for our lockd? */
 3973         if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
 3974                 return (EINVAL);
 3975         error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
 3976         if (error)
 3977                 return(error);
 3978         /* find the mount point */
 3979         mtx_lock(&Giant);
 3980         mp = vfs_getvfs(&fhp.fh_fsid);
 3981         if (mp == NULL) {
 3982                 error = ESTALE;
 3983                 goto out;
 3984         }
 3985         /* now give me my vnode, it gets returned to me locked */
 3986         error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
 3987         if (error)
 3988                 goto out;
 3989         /*
 3990          * from now on we have to make sure not
 3991          * to forget about the vnode
 3992          * any error that causes an abort must vput(vp)
 3993          * just set error = err and 'goto bad;'.
 3994          */
 3995 
 3996         /*
 3997          * from vn_open
 3998          */
 3999         if (vp->v_type == VLNK) {
 4000                 error = EMLINK;
 4001                 goto bad;
 4002         }
 4003         if (vp->v_type == VSOCK) {
 4004                 error = EOPNOTSUPP;
 4005                 goto bad;
 4006         }
 4007         mode = 0;
 4008         if (fmode & (FWRITE | O_TRUNC)) {
 4009                 if (vp->v_type == VDIR) {
 4010                         error = EISDIR;
 4011                         goto bad;
 4012                 }
 4013                 error = vn_writechk(vp);
 4014                 if (error)
 4015                         goto bad;
 4016                 mode |= VWRITE;
 4017         }
 4018         if (fmode & FREAD)
 4019                 mode |= VREAD;
 4020         if (fmode & O_APPEND)
 4021                 mode |= VAPPEND;
 4022 #ifdef MAC
 4023         error = mac_check_vnode_open(td->td_ucred, vp, mode);
 4024         if (error)
 4025                 goto bad;
 4026 #endif
 4027         if (mode) {
 4028                 error = VOP_ACCESS(vp, mode, td->td_ucred, td);
 4029                 if (error)
 4030                         goto bad;
 4031         }
 4032         if (fmode & O_TRUNC) {
 4033                 VOP_UNLOCK(vp, 0, td);                          /* XXX */
 4034                 if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
 4035                         vrele(vp);
 4036                         goto out;
 4037                 }
 4038                 VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 4039                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);       /* XXX */
 4040 #ifdef MAC
 4041                 /*
 4042                  * We don't yet have fp->f_cred, so use td->td_ucred, which
 4043                  * should be right.
 4044                  */
 4045                 error = mac_check_vnode_write(td->td_ucred, td->td_ucred, vp);
 4046                 if (error == 0) {
 4047 #endif
 4048                         VATTR_NULL(vap);
 4049                         vap->va_size = 0;
 4050                         error = VOP_SETATTR(vp, vap, td->td_ucred, td);
 4051 #ifdef MAC
 4052                 }
 4053 #endif
 4054                 vn_finished_write(mp);
 4055                 if (error)
 4056                         goto bad;
 4057         }
 4058         error = VOP_OPEN(vp, fmode, td->td_ucred, td, -1);
 4059         if (error)
 4060                 goto bad;
 4061 
 4062         if (fmode & FWRITE)
 4063                 vp->v_writecount++;
 4064 
 4065         /*
 4066          * end of vn_open code
 4067          */
 4068 
 4069         if ((error = falloc(td, &nfp, &indx)) != 0) {
 4070                 if (fmode & FWRITE)
 4071                         vp->v_writecount--;
 4072                 goto bad;
 4073         }
 4074         /* An extra reference on `nfp' has been held for us by falloc(). */
 4075         fp = nfp;
 4076 
 4077         nfp->f_vnode = vp;
 4078         nfp->f_data = vp;
 4079         nfp->f_flag = fmode & FMASK;
 4080         nfp->f_ops = &vnops;
 4081         nfp->f_type = DTYPE_VNODE;
 4082         if (fmode & (O_EXLOCK | O_SHLOCK)) {
 4083                 lf.l_whence = SEEK_SET;
 4084                 lf.l_start = 0;
 4085                 lf.l_len = 0;
 4086                 if (fmode & O_EXLOCK)
 4087                         lf.l_type = F_WRLCK;
 4088                 else
 4089                         lf.l_type = F_RDLCK;
 4090                 type = F_FLOCK;
 4091                 if ((fmode & FNONBLOCK) == 0)
 4092                         type |= F_WAIT;
 4093                 VOP_UNLOCK(vp, 0, td);
 4094                 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
 4095                             type)) != 0) {
 4096                         /*
 4097                          * The lock request failed.  Normally close the
 4098                          * descriptor but handle the case where someone might
 4099                          * have dup()d or close()d it when we weren't looking.
 4100                          */
 4101                         fdclose(fdp, fp, indx, td);
 4102 
 4103                         /*
 4104                          * release our private reference
 4105                          */
 4106                         fdrop(fp, td);
 4107                         goto out;
 4108                 }
 4109                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 4110                 fp->f_flag |= FHASLOCK;
 4111         }
 4112 
 4113         VOP_UNLOCK(vp, 0, td);
 4114         fdrop(fp, td);
 4115         mtx_unlock(&Giant);
 4116         td->td_retval[0] = indx;
 4117         return (0);
 4118 
 4119 bad:
 4120         vput(vp);
 4121 out:
 4122         mtx_unlock(&Giant);
 4123         return (error);
 4124 }
 4125 
 4126 /*
 4127  * Stat an (NFS) file handle.
 4128  *
 4129  * MP SAFE
 4130  */
 4131 #ifndef _SYS_SYSPROTO_H_
 4132 struct fhstat_args {
 4133         struct fhandle *u_fhp;
 4134         struct stat *sb;
 4135 };
 4136 #endif
 4137 int
 4138 fhstat(td, uap)
 4139         struct thread *td;
 4140         register struct fhstat_args /* {
 4141                 struct fhandle *u_fhp;
 4142                 struct stat *sb;
 4143         } */ *uap;
 4144 {
 4145         struct stat sb;
 4146         fhandle_t fh;
 4147         struct mount *mp;
 4148         struct vnode *vp;
 4149         int error;
 4150 
 4151         error = suser(td);
 4152         if (error)
 4153                 return (error);
 4154         error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
 4155         if (error)
 4156                 return (error);
 4157         mtx_lock(&Giant);
 4158         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
 4159                 mtx_unlock(&Giant);
 4160                 return (ESTALE);
 4161         }
 4162         if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) {
 4163                 mtx_unlock(&Giant);
 4164                 return (error);
 4165         }
 4166         error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
 4167         vput(vp);
 4168         mtx_unlock(&Giant);
 4169         if (error)
 4170                 return (error);
 4171         error = copyout(&sb, uap->sb, sizeof(sb));
 4172         return (error);
 4173 }
 4174 
 4175 /*
 4176  * Implement fstatfs() for (NFS) file handles.
 4177  *
 4178  * MP SAFE
 4179  */
 4180 #ifndef _SYS_SYSPROTO_H_
 4181 struct fhstatfs_args {
 4182         struct fhandle *u_fhp;
 4183         struct statfs *buf;
 4184 };
 4185 #endif
 4186 int
 4187 fhstatfs(td, uap)
 4188         struct thread *td;
 4189         struct fhstatfs_args /* {
 4190                 struct fhandle *u_fhp;
 4191                 struct statfs *buf;
 4192         } */ *uap;
 4193 {
 4194         struct statfs sf;
 4195         fhandle_t fh;
 4196         int error;
 4197 
 4198         error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
 4199         if (error)
 4200                 return (error);
 4201         error = kern_fhstatfs(td, fh, &sf);
 4202         if (error)
 4203                 return (error);
 4204         return (copyout(&sf, uap->buf, sizeof(sf)));
 4205 }
 4206 
 4207 int
 4208 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
 4209 {
 4210         struct statfs *sp;
 4211         struct mount *mp;
 4212         struct vnode *vp;
 4213         int error;
 4214 
 4215         error = suser(td);
 4216         if (error)
 4217                 return (error);
 4218         mtx_lock(&Giant);
 4219         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
 4220                 mtx_unlock(&Giant);
 4221                 return (ESTALE);
 4222         }
 4223         error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
 4224         if (error) {
 4225                 mtx_unlock(&Giant);
 4226                 return (error);
 4227         }
 4228         mp = vp->v_mount;
 4229         sp = &mp->mnt_stat;
 4230         vput(vp);
 4231         error = prison_canseemount(td->td_ucred, mp);
 4232         if (error)
 4233                 return (error);
 4234 #ifdef MAC
 4235         error = mac_check_mount_stat(td->td_ucred, mp);
 4236         if (error) {
 4237                 mtx_unlock(&Giant);
 4238                 return (error);
 4239         }
 4240 #endif
 4241         /*
 4242          * Set these in case the underlying filesystem fails to do so.
 4243          */
 4244         sp->f_version = STATFS_VERSION;
 4245         sp->f_namemax = NAME_MAX;
 4246         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
 4247         error = VFS_STATFS(mp, sp, td);
 4248         mtx_unlock(&Giant);
 4249         if (error)
 4250                 return (error);
 4251         *buf = *sp;
 4252         return (0);
 4253 }
 4254 
 4255 /*
 4256  * Syscall to push extended attribute configuration information into the
 4257  * VFS.  Accepts a path, which it converts to a mountpoint, as well as
 4258  * a command (int cmd), and attribute name and misc data.  For now, the
 4259  * attribute name is left in userspace for consumption by the VFS_op.
 4260  * It will probably be changed to be copied into sysspace by the
 4261  * syscall in the future, once issues with various consumers of the
 4262  * attribute code have raised their hands.
 4263  *
 4264  * Currently this is used only by UFS Extended Attributes.
 4265  */
 4266 int
 4267 extattrctl(td, uap)
 4268         struct thread *td;
 4269         struct extattrctl_args /* {
 4270                 const char *path;
 4271                 int cmd;
 4272                 const char *filename;
 4273                 int attrnamespace;
 4274                 const char *attrname;
 4275         } */ *uap;
 4276 {
 4277         struct vnode *filename_vp;
 4278         struct nameidata nd;
 4279         struct mount *mp, *mp_writable;
 4280         char attrname[EXTATTR_MAXNAMELEN];
 4281         int error;
 4282 
 4283         /*
 4284          * uap->attrname is not always defined.  We check again later when we
 4285          * invoke the VFS call so as to pass in NULL there if needed.
 4286          */
 4287         if (uap->attrname != NULL) {
 4288                 error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN,
 4289                     NULL);
 4290                 if (error)
 4291                         return (error);
 4292         }
 4293 
 4294         /*
 4295          * uap->filename is not always defined.  If it is, grab a vnode lock,
 4296          * which VFS_EXTATTRCTL() will later release.
 4297          */
 4298         filename_vp = NULL;
 4299         if (uap->filename != NULL) {
 4300                 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
 4301                     uap->filename, td);
 4302                 error = namei(&nd);
 4303                 if (error)
 4304                         return (error);
 4305                 filename_vp = nd.ni_vp;
 4306                 NDFREE(&nd, NDF_NO_VP_RELE | NDF_NO_VP_UNLOCK);
 4307         }
 4308 
 4309         /* uap->path is always defined. */
 4310         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 4311         error = namei(&nd);
 4312         if (error) {
 4313                 if (filename_vp != NULL)
 4314                         vput(filename_vp);
 4315                 return (error);
 4316         }
 4317         mp = nd.ni_vp->v_mount;
 4318         error = vn_start_write(nd.ni_vp, &mp_writable, V_WAIT | PCATCH);
 4319         NDFREE(&nd, 0);
 4320         if (error) {
 4321                 if (filename_vp != NULL)
 4322                         vput(filename_vp);
 4323                 return (error);
 4324         }
 4325 
 4326         error = VFS_EXTATTRCTL(mp, uap->cmd, filename_vp, uap->attrnamespace,
 4327             uap->attrname != NULL ? attrname : NULL, td);
 4328 
 4329         vn_finished_write(mp_writable);
 4330         /*
 4331          * VFS_EXTATTRCTL will have unlocked, but not de-ref'd,
 4332          * filename_vp, so vrele it if it is defined.
 4333          */
 4334         if (filename_vp != NULL)
 4335                 vrele(filename_vp);
 4336         return (error);
 4337 }
 4338 
 4339 /*-
 4340  * Set a named extended attribute on a file or directory
 4341  *
 4342  * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
 4343  *            kernelspace string pointer "attrname", userspace buffer
 4344  *            pointer "data", buffer length "nbytes", thread "td".
 4345  * Returns: 0 on success, an error number otherwise
 4346  * Locks: none
 4347  * References: vp must be a valid reference for the duration of the call
 4348  */
 4349 static int
 4350 extattr_set_vp(struct vnode *vp, int attrnamespace, const char *attrname,
 4351     void *data, size_t nbytes, struct thread *td)
 4352 {
 4353         struct mount *mp;
 4354         struct uio auio;
 4355         struct iovec aiov;
 4356         ssize_t cnt;
 4357         int error;
 4358 
 4359         error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
 4360         if (error)
 4361                 return (error);
 4362         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 4363         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 4364 
 4365         aiov.iov_base = data;
 4366         aiov.iov_len = nbytes;
 4367         auio.uio_iov = &aiov;
 4368         auio.uio_iovcnt = 1;
 4369         auio.uio_offset = 0;
 4370         if (nbytes > INT_MAX) {
 4371                 error = EINVAL;
 4372                 goto done;
 4373         }
 4374         auio.uio_resid = nbytes;
 4375         auio.uio_rw = UIO_WRITE;
 4376         auio.uio_segflg = UIO_USERSPACE;
 4377         auio.uio_td = td;
 4378         cnt = nbytes;
 4379 
 4380 #ifdef MAC
 4381         error = mac_check_vnode_setextattr(td->td_ucred, vp, attrnamespace,
 4382             attrname, &auio);
 4383         if (error)
 4384                 goto done;
 4385 #endif
 4386 
 4387         error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio,
 4388             td->td_ucred, td);
 4389         cnt -= auio.uio_resid;
 4390         td->td_retval[0] = cnt;
 4391 
 4392 done:
 4393         VOP_UNLOCK(vp, 0, td);
 4394         vn_finished_write(mp);
 4395         return (error);
 4396 }
 4397 
 4398 int
 4399 extattr_set_fd(td, uap)
 4400         struct thread *td;
 4401         struct extattr_set_fd_args /* {
 4402                 int fd;
 4403                 int attrnamespace;
 4404                 const char *attrname;
 4405                 void *data;
 4406                 size_t nbytes;
 4407         } */ *uap;
 4408 {
 4409         struct file *fp;
 4410         char attrname[EXTATTR_MAXNAMELEN];
 4411         int error;
 4412 
 4413         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4414         if (error)
 4415                 return (error);
 4416 
 4417         error = getvnode(td->td_proc->p_fd, uap->fd, &fp);
 4418         if (error)
 4419                 return (error);
 4420 
 4421         error = extattr_set_vp(fp->f_vnode, uap->attrnamespace,
 4422             attrname, uap->data, uap->nbytes, td);
 4423         fdrop(fp, td);
 4424 
 4425         return (error);
 4426 }
 4427 
 4428 int
 4429 extattr_set_file(td, uap)
 4430         struct thread *td;
 4431         struct extattr_set_file_args /* {
 4432                 const char *path;
 4433                 int attrnamespace;
 4434                 const char *attrname;
 4435                 void *data;
 4436                 size_t nbytes;
 4437         } */ *uap;
 4438 {
 4439         struct nameidata nd;
 4440         char attrname[EXTATTR_MAXNAMELEN];
 4441         int error;
 4442 
 4443         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4444         if (error)
 4445                 return (error);
 4446 
 4447         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 4448         error = namei(&nd);
 4449         if (error)
 4450                 return (error);
 4451         NDFREE(&nd, NDF_ONLY_PNBUF);
 4452 
 4453         error = extattr_set_vp(nd.ni_vp, uap->attrnamespace, attrname,
 4454             uap->data, uap->nbytes, td);
 4455 
 4456         vrele(nd.ni_vp);
 4457         return (error);
 4458 }
 4459 
 4460 int
 4461 extattr_set_link(td, uap)
 4462         struct thread *td;
 4463         struct extattr_set_link_args /* {
 4464                 const char *path;
 4465                 int attrnamespace;
 4466                 const char *attrname;
 4467                 void *data;
 4468                 size_t nbytes;
 4469         } */ *uap;
 4470 {
 4471         struct nameidata nd;
 4472         char attrname[EXTATTR_MAXNAMELEN];
 4473         int error;
 4474 
 4475         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4476         if (error)
 4477                 return (error);
 4478 
 4479         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
 4480         error = namei(&nd);
 4481         if (error)
 4482                 return (error);
 4483         NDFREE(&nd, NDF_ONLY_PNBUF);
 4484 
 4485         error = extattr_set_vp(nd.ni_vp, uap->attrnamespace, attrname,
 4486             uap->data, uap->nbytes, td);
 4487 
 4488         vrele(nd.ni_vp);
 4489         return (error);
 4490 }
 4491 
 4492 /*-
 4493  * Get a named extended attribute on a file or directory
 4494  *
 4495  * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
 4496  *            kernelspace string pointer "attrname", userspace buffer
 4497  *            pointer "data", buffer length "nbytes", thread "td".
 4498  * Returns: 0 on success, an error number otherwise
 4499  * Locks: none
 4500  * References: vp must be a valid reference for the duration of the call
 4501  */
 4502 static int
 4503 extattr_get_vp(struct vnode *vp, int attrnamespace, const char *attrname,
 4504     void *data, size_t nbytes, struct thread *td)
 4505 {
 4506         struct uio auio, *auiop;
 4507         struct iovec aiov;
 4508         ssize_t cnt;
 4509         size_t size, *sizep;
 4510         int error;
 4511 
 4512         VOP_LEASE(vp, td, td->td_ucred, LEASE_READ);
 4513         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 4514 
 4515         /*
 4516          * Slightly unusual semantics: if the user provides a NULL data
 4517          * pointer, they don't want to receive the data, just the
 4518          * maximum read length.
 4519          */
 4520         auiop = NULL;
 4521         sizep = NULL;
 4522         cnt = 0;
 4523         if (data != NULL) {
 4524                 aiov.iov_base = data;
 4525                 aiov.iov_len = nbytes;
 4526                 auio.uio_iov = &aiov;
 4527                 auio.uio_iovcnt = 1;
 4528                 auio.uio_offset = 0;
 4529                 if (nbytes > INT_MAX) {
 4530                         error = EINVAL;
 4531                         goto done;
 4532                 }
 4533                 auio.uio_resid = nbytes;
 4534                 auio.uio_rw = UIO_READ;
 4535                 auio.uio_segflg = UIO_USERSPACE;
 4536                 auio.uio_td = td;
 4537                 auiop = &auio;
 4538                 cnt = nbytes;
 4539         } else
 4540                 sizep = &size;
 4541 
 4542 #ifdef MAC
 4543         error = mac_check_vnode_getextattr(td->td_ucred, vp, attrnamespace,
 4544             attrname, &auio);
 4545         if (error)
 4546                 goto done;
 4547 #endif
 4548 
 4549         error = VOP_GETEXTATTR(vp, attrnamespace, attrname, auiop, sizep,
 4550             td->td_ucred, td);
 4551 
 4552         if (auiop != NULL) {
 4553                 cnt -= auio.uio_resid;
 4554                 td->td_retval[0] = cnt;
 4555         } else
 4556                 td->td_retval[0] = size;
 4557 
 4558 done:
 4559         VOP_UNLOCK(vp, 0, td);
 4560         return (error);
 4561 }
 4562 
 4563 int
 4564 extattr_get_fd(td, uap)
 4565         struct thread *td;
 4566         struct extattr_get_fd_args /* {
 4567                 int fd;
 4568                 int attrnamespace;
 4569                 const char *attrname;
 4570                 void *data;
 4571                 size_t nbytes;
 4572         } */ *uap;
 4573 {
 4574         struct file *fp;
 4575         char attrname[EXTATTR_MAXNAMELEN];
 4576         int error;
 4577 
 4578         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4579         if (error)
 4580                 return (error);
 4581 
 4582         error = getvnode(td->td_proc->p_fd, uap->fd, &fp);
 4583         if (error)
 4584                 return (error);
 4585 
 4586         error = extattr_get_vp(fp->f_vnode, uap->attrnamespace,
 4587             attrname, uap->data, uap->nbytes, td);
 4588 
 4589         fdrop(fp, td);
 4590         return (error);
 4591 }
 4592 
 4593 int
 4594 extattr_get_file(td, uap)
 4595         struct thread *td;
 4596         struct extattr_get_file_args /* {
 4597                 const char *path;
 4598                 int attrnamespace;
 4599                 const char *attrname;
 4600                 void *data;
 4601                 size_t nbytes;
 4602         } */ *uap;
 4603 {
 4604         struct nameidata nd;
 4605         char attrname[EXTATTR_MAXNAMELEN];
 4606         int error;
 4607 
 4608         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4609         if (error)
 4610                 return (error);
 4611 
 4612         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 4613         error = namei(&nd);
 4614         if (error)
 4615                 return (error);
 4616         NDFREE(&nd, NDF_ONLY_PNBUF);
 4617 
 4618         error = extattr_get_vp(nd.ni_vp, uap->attrnamespace, attrname,
 4619             uap->data, uap->nbytes, td);
 4620 
 4621         vrele(nd.ni_vp);
 4622         return (error);
 4623 }
 4624 
 4625 int
 4626 extattr_get_link(td, uap)
 4627         struct thread *td;
 4628         struct extattr_get_link_args /* {
 4629                 const char *path;
 4630                 int attrnamespace;
 4631                 const char *attrname;
 4632                 void *data;
 4633                 size_t nbytes;
 4634         } */ *uap;
 4635 {
 4636         struct nameidata nd;
 4637         char attrname[EXTATTR_MAXNAMELEN];
 4638         int error;
 4639 
 4640         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4641         if (error)
 4642                 return (error);
 4643 
 4644         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
 4645         error = namei(&nd);
 4646         if (error)
 4647                 return (error);
 4648         NDFREE(&nd, NDF_ONLY_PNBUF);
 4649 
 4650         error = extattr_get_vp(nd.ni_vp, uap->attrnamespace, attrname,
 4651             uap->data, uap->nbytes, td);
 4652 
 4653         vrele(nd.ni_vp);
 4654         return (error);
 4655 }
 4656 
 4657 /*
 4658  * extattr_delete_vp(): Delete a named extended attribute on a file or
 4659  *                      directory
 4660  *
 4661  * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
 4662  *            kernelspace string pointer "attrname", proc "p"
 4663  * Returns: 0 on success, an error number otherwise
 4664  * Locks: none
 4665  * References: vp must be a valid reference for the duration of the call
 4666  */
 4667 static int
 4668 extattr_delete_vp(struct vnode *vp, int attrnamespace, const char *attrname,
 4669     struct thread *td)
 4670 {
 4671         struct mount *mp;
 4672         int error;
 4673 
 4674         error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
 4675         if (error)
 4676                 return (error);
 4677         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 4678         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 4679 
 4680 #ifdef MAC
 4681         error = mac_check_vnode_deleteextattr(td->td_ucred, vp, attrnamespace,
 4682             attrname);
 4683         if (error)
 4684                 goto done;
 4685 #endif
 4686 
 4687         error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, td->td_ucred,
 4688             td);
 4689         if (error == EOPNOTSUPP)
 4690                 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
 4691                     td->td_ucred, td);
 4692 #ifdef MAC
 4693 done:
 4694 #endif
 4695         VOP_UNLOCK(vp, 0, td);
 4696         vn_finished_write(mp);
 4697         return (error);
 4698 }
 4699 
 4700 int
 4701 extattr_delete_fd(td, uap)
 4702         struct thread *td;
 4703         struct extattr_delete_fd_args /* {
 4704                 int fd;
 4705                 int attrnamespace;
 4706                 const char *attrname;
 4707         } */ *uap;
 4708 {
 4709         struct file *fp;
 4710         struct vnode *vp;
 4711         char attrname[EXTATTR_MAXNAMELEN];
 4712         int error;
 4713 
 4714         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4715         if (error)
 4716                 return (error);
 4717 
 4718         error = getvnode(td->td_proc->p_fd, uap->fd, &fp);
 4719         if (error)
 4720                 return (error);
 4721         vp = fp->f_vnode;
 4722 
 4723         error = extattr_delete_vp(vp, uap->attrnamespace, attrname, td);
 4724         fdrop(fp, td);
 4725         return (error);
 4726 }
 4727 
 4728 int
 4729 extattr_delete_file(td, uap)
 4730         struct thread *td;
 4731         struct extattr_delete_file_args /* {
 4732                 const char *path;
 4733                 int attrnamespace;
 4734                 const char *attrname;
 4735         } */ *uap;
 4736 {
 4737         struct nameidata nd;
 4738         char attrname[EXTATTR_MAXNAMELEN];
 4739         int error;
 4740 
 4741         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4742         if (error)
 4743                 return(error);
 4744 
 4745         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 4746         error = namei(&nd);
 4747         if (error)
 4748                 return(error);
 4749         NDFREE(&nd, NDF_ONLY_PNBUF);
 4750 
 4751         error = extattr_delete_vp(nd.ni_vp, uap->attrnamespace, attrname, td);
 4752         vrele(nd.ni_vp);
 4753         return(error);
 4754 }
 4755 
 4756 int
 4757 extattr_delete_link(td, uap)
 4758         struct thread *td;
 4759         struct extattr_delete_link_args /* {
 4760                 const char *path;
 4761                 int attrnamespace;
 4762                 const char *attrname;
 4763         } */ *uap;
 4764 {
 4765         struct nameidata nd;
 4766         char attrname[EXTATTR_MAXNAMELEN];
 4767         int error;
 4768 
 4769         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4770         if (error)
 4771                 return(error);
 4772 
 4773         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
 4774         error = namei(&nd);
 4775         if (error)
 4776                 return(error);
 4777         NDFREE(&nd, NDF_ONLY_PNBUF);
 4778 
 4779         error = extattr_delete_vp(nd.ni_vp, uap->attrnamespace, attrname, td);
 4780         vrele(nd.ni_vp);
 4781         return(error);
 4782 }
 4783 
 4784 /*-
 4785  * Retrieve a list of extended attributes on a file or directory.
 4786  *
 4787  * Arguments: unlocked vnode "vp", attribute namespace 'attrnamespace",
 4788  *            userspace buffer pointer "data", buffer length "nbytes",
 4789  *            thread "td".
 4790  * Returns: 0 on success, an error number otherwise
 4791  * Locks: none
 4792  * References: vp must be a valid reference for the duration of the call
 4793  */
 4794 static int
 4795 extattr_list_vp(struct vnode *vp, int attrnamespace, void *data,
 4796     size_t nbytes, struct thread *td)
 4797 {
 4798         struct uio auio, *auiop;
 4799         size_t size, *sizep;
 4800         struct iovec aiov;
 4801         ssize_t cnt;
 4802         int error;
 4803 
 4804         VOP_LEASE(vp, td, td->td_ucred, LEASE_READ);
 4805         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 4806 
 4807         auiop = NULL;
 4808         sizep = NULL;
 4809         cnt = 0;
 4810         if (data != NULL) {
 4811                 aiov.iov_base = data;
 4812                 aiov.iov_len = nbytes;
 4813                 auio.uio_iov = &aiov;
 4814                 auio.uio_iovcnt = 1;
 4815                 auio.uio_offset = 0;
 4816                 if (nbytes > INT_MAX) {
 4817                         error = EINVAL;
 4818                         goto done;
 4819                 }
 4820                 auio.uio_resid = nbytes;
 4821                 auio.uio_rw = UIO_READ;
 4822                 auio.uio_segflg = UIO_USERSPACE;
 4823                 auio.uio_td = td;
 4824                 auiop = &auio;
 4825                 cnt = nbytes;
 4826         } else
 4827                 sizep = &size;
 4828 
 4829 #ifdef MAC
 4830         error = mac_check_vnode_listextattr(td->td_ucred, vp, attrnamespace);
 4831         if (error)
 4832                 goto done;
 4833 #endif
 4834 
 4835         error = VOP_LISTEXTATTR(vp, attrnamespace, auiop, sizep,
 4836             td->td_ucred, td);
 4837 
 4838         if (auiop != NULL) {
 4839                 cnt -= auio.uio_resid;
 4840                 td->td_retval[0] = cnt;
 4841         } else
 4842                 td->td_retval[0] = size;
 4843 
 4844 done:
 4845         VOP_UNLOCK(vp, 0, td);
 4846         return (error);
 4847 }
 4848 
 4849 
 4850 int
 4851 extattr_list_fd(td, uap)
 4852         struct thread *td;
 4853         struct extattr_list_fd_args /* {
 4854                 int fd;
 4855                 int attrnamespace;
 4856                 void *data;
 4857                 size_t nbytes;
 4858         } */ *uap;
 4859 {
 4860         struct file *fp;
 4861         int error;
 4862 
 4863         error = getvnode(td->td_proc->p_fd, uap->fd, &fp);
 4864         if (error)
 4865                 return (error);
 4866 
 4867         error = extattr_list_vp(fp->f_vnode, uap->attrnamespace, uap->data,
 4868             uap->nbytes, td);
 4869 
 4870         fdrop(fp, td);
 4871         return (error);
 4872 }
 4873 
 4874 int
 4875 extattr_list_file(td, uap)
 4876         struct thread*td;
 4877         struct extattr_list_file_args /* {
 4878                 const char *path;
 4879                 int attrnamespace;
 4880                 void *data;
 4881                 size_t nbytes;
 4882         } */ *uap;
 4883 {
 4884         struct nameidata nd;
 4885         int error;
 4886 
 4887         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 4888         error = namei(&nd);
 4889         if (error)
 4890                 return (error);
 4891         NDFREE(&nd, NDF_ONLY_PNBUF);
 4892 
 4893         error = extattr_list_vp(nd.ni_vp, uap->attrnamespace, uap->data,
 4894             uap->nbytes, td);
 4895 
 4896         vrele(nd.ni_vp);
 4897         return (error);
 4898 }
 4899 
 4900 int
 4901 extattr_list_link(td, uap)
 4902         struct thread*td;
 4903         struct extattr_list_link_args /* {
 4904                 const char *path;
 4905                 int attrnamespace;
 4906                 void *data;
 4907                 size_t nbytes;
 4908         } */ *uap;
 4909 {
 4910         struct nameidata nd;
 4911         int error;
 4912 
 4913         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
 4914         error = namei(&nd);
 4915         if (error)
 4916                 return (error);
 4917         NDFREE(&nd, NDF_ONLY_PNBUF);
 4918 
 4919         error = extattr_list_vp(nd.ni_vp, uap->attrnamespace, uap->data,
 4920             uap->nbytes, td);
 4921 
 4922         vrele(nd.ni_vp);
 4923         return (error);
 4924 }

Cache object: 8ab62f11cca6363b85e85ccc0afb8a44


[ 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.