kern_descrip.c

Version: - FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
    1 /*      $OpenBSD: kern_descrip.c,v 1.207 2022/12/05 23:18:37 deraadt Exp $      */
    2 /*      $NetBSD: kern_descrip.c,v 1.42 1996/03/30 22:24:38 christos Exp $       */
    3 
    4 /*
    5  * Copyright (c) 1982, 1986, 1989, 1991, 1993
    6  *      The Regents of the University of California.  All rights reserved.
    7  * (c) UNIX System Laboratories, Inc.
    8  * All or some portions of this file are derived from material licensed
    9  * to the University of California by American Telephone and Telegraph
   10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
   11  * the permission of UNIX System Laboratories, Inc.
   12  *
   13  * Redistribution and use in source and binary forms, with or without
   14  * modification, are permitted provided that the following conditions
   15  * are met:
   16  * 1. Redistributions of source code must retain the above copyright
   17  *    notice, this list of conditions and the following disclaimer.
   18  * 2. Redistributions in binary form must reproduce the above copyright
   19  *    notice, this list of conditions and the following disclaimer in the
   20  *    documentation and/or other materials provided with the distribution.
   21  * 3. Neither the name of the University nor the names of its contributors
   22  *    may be used to endorse or promote products derived from this software
   23  *    without specific prior written permission.
   24  *
   25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   35  * SUCH DAMAGE.
   36  *
   37  *      @(#)kern_descrip.c      8.6 (Berkeley) 4/19/94
   38  */
   39 
   40 #include <sys/param.h>
   41 #include <sys/systm.h>
   42 #include <sys/filedesc.h>
   43 #include <sys/vnode.h>
   44 #include <sys/proc.h>
   45 #include <sys/file.h>
   46 #include <sys/socket.h>
   47 #include <sys/stat.h>
   48 #include <sys/ioctl.h>
   49 #include <sys/fcntl.h>
   50 #include <sys/lock.h>
   51 #include <sys/malloc.h>
   52 #include <sys/ucred.h>
   53 #include <sys/unistd.h>
   54 #include <sys/resourcevar.h>
   55 #include <sys/mount.h>
   56 #include <sys/syscallargs.h>
   57 #include <sys/event.h>
   58 #include <sys/pool.h>
   59 #include <sys/ktrace.h>
   60 #include <sys/pledge.h>
   61 
   62 /*
   63  * Descriptor management.
   64  *
   65  * We need to block interrupts as long as `fhdlk' is being taken
   66  * with and without the KERNEL_LOCK().
   67  */
   68 struct mutex fhdlk = MUTEX_INITIALIZER(IPL_MPFLOOR);
   69 struct filelist filehead;       /* head of list of open files */
   70 int numfiles;                   /* actual number of open files */
   71 
   72 static __inline void fd_used(struct filedesc *, int);
   73 static __inline void fd_unused(struct filedesc *, int);
   74 static __inline int find_next_zero(u_int *, int, u_int);
   75 static __inline int fd_inuse(struct filedesc *, int);
   76 int finishdup(struct proc *, struct file *, int, int, register_t *, int);
   77 int find_last_set(struct filedesc *, int);
   78 int dodup3(struct proc *, int, int, int, register_t *);
   79 
   80 #define DUPF_CLOEXEC    0x01
   81 #define DUPF_DUP2       0x02
   82 
   83 struct pool file_pool;
   84 struct pool fdesc_pool;
   85 
   86 void
   87 filedesc_init(void)
   88 {
   89         pool_init(&file_pool, sizeof(struct file), 0, IPL_MPFLOOR,
   90             PR_WAITOK, "filepl", NULL);
   91         pool_init(&fdesc_pool, sizeof(struct filedesc0), 0, IPL_NONE,
   92             PR_WAITOK, "fdescpl", NULL);
   93         LIST_INIT(&filehead);
   94 }
   95 
   96 static __inline int
   97 find_next_zero (u_int *bitmap, int want, u_int bits)
   98 {
   99         int i, off, maxoff;
  100         u_int sub;
  101 
  102         if (want > bits)
  103                 return -1;
  104 
  105         off = want >> NDENTRYSHIFT;
  106         i = want & NDENTRYMASK;
  107         if (i) {
  108                 sub = bitmap[off] | ((u_int)~0 >> (NDENTRIES - i));
  109                 if (sub != ~0)
  110                         goto found;
  111                 off++;
  112         }
  113 
  114         maxoff = NDLOSLOTS(bits);
  115         while (off < maxoff) {
  116                 if ((sub = bitmap[off]) != ~0)
  117                         goto found;
  118                 off++;
  119         }
  120 
  121         return -1;
  122 
  123  found:
  124         return (off << NDENTRYSHIFT) + ffs(~sub) - 1;
  125 }
  126 
  127 int
  128 find_last_set(struct filedesc *fd, int last)
  129 {
  130         int off, i;
  131         u_int *bitmap = fd->fd_lomap;
  132 
  133         off = (last - 1) >> NDENTRYSHIFT;
  134 
  135         while (off >= 0 && !bitmap[off])
  136                 off--;
  137         if (off < 0)
  138                 return 0;
  139 
  140         i = ((off + 1) << NDENTRYSHIFT) - 1;
  141         if (i >= last)
  142                 i = last - 1;
  143 
  144         while (i > 0 && !fd_inuse(fd, i))
  145                 i--;
  146         return i;
  147 }
  148 
  149 static __inline int
  150 fd_inuse(struct filedesc *fdp, int fd)
  151 {
  152         u_int off = fd >> NDENTRYSHIFT;
  153 
  154         if (fdp->fd_lomap[off] & (1U << (fd & NDENTRYMASK)))
  155                 return 1;
  156 
  157         return 0;
  158 }
  159 
  160 static __inline void
  161 fd_used(struct filedesc *fdp, int fd)
  162 {
  163         u_int off = fd >> NDENTRYSHIFT;
  164 
  165         fdp->fd_lomap[off] |= 1U << (fd & NDENTRYMASK);
  166         if (fdp->fd_lomap[off] == ~0)
  167                 fdp->fd_himap[off >> NDENTRYSHIFT] |= 1U << (off & NDENTRYMASK);
  168 
  169         if (fd > fdp->fd_lastfile)
  170                 fdp->fd_lastfile = fd;
  171         fdp->fd_openfd++;
  172 }
  173 
  174 static __inline void
  175 fd_unused(struct filedesc *fdp, int fd)
  176 {
  177         u_int off = fd >> NDENTRYSHIFT;
  178 
  179         if (fd < fdp->fd_freefile)
  180                 fdp->fd_freefile = fd;
  181 
  182         if (fdp->fd_lomap[off] == ~0)
  183                 fdp->fd_himap[off >> NDENTRYSHIFT] &= ~(1U << (off & NDENTRYMASK));
  184         fdp->fd_lomap[off] &= ~(1U << (fd & NDENTRYMASK));
  185 
  186 #ifdef DIAGNOSTIC
  187         if (fd > fdp->fd_lastfile)
  188                 panic("fd_unused: fd_lastfile inconsistent");
  189 #endif
  190         if (fd == fdp->fd_lastfile)
  191                 fdp->fd_lastfile = find_last_set(fdp, fd);
  192         fdp->fd_openfd--;
  193 }
  194 
  195 struct file *
  196 fd_iterfile(struct file *fp, struct proc *p)
  197 {
  198         struct file *nfp;
  199         unsigned int count;
  200 
  201         mtx_enter(&fhdlk);
  202         if (fp == NULL)
  203                 nfp = LIST_FIRST(&filehead);
  204         else
  205                 nfp = LIST_NEXT(fp, f_list);
  206 
  207         /* don't refcount when f_count == 0 to avoid race in fdrop() */
  208         while (nfp != NULL) {
  209                 count = nfp->f_count;
  210                 if (count == 0) {
  211                         nfp = LIST_NEXT(nfp, f_list);
  212                         continue;
  213                 }
  214                 if (atomic_cas_uint(&nfp->f_count, count, count + 1) == count)
  215                         break;
  216         }
  217         mtx_leave(&fhdlk);
  218 
  219         if (fp != NULL)
  220                 FRELE(fp, p);
  221 
  222         return nfp;
  223 }
  224 
  225 struct file *
  226 fd_getfile(struct filedesc *fdp, int fd)
  227 {
  228         struct file *fp;
  229 
  230         vfs_stall_barrier();
  231 
  232         if ((u_int)fd >= fdp->fd_nfiles)
  233                 return (NULL);
  234 
  235         mtx_enter(&fdp->fd_fplock);
  236         fp = fdp->fd_ofiles[fd];
  237         if (fp != NULL)
  238                 atomic_inc_int(&fp->f_count);
  239         mtx_leave(&fdp->fd_fplock);
  240 
  241         return (fp);
  242 }
  243 
  244 struct file *
  245 fd_getfile_mode(struct filedesc *fdp, int fd, int mode)
  246 {
  247         struct file *fp;
  248 
  249         KASSERT(mode != 0);
  250 
  251         fp = fd_getfile(fdp, fd);
  252         if (fp == NULL)
  253                 return (NULL);
  254 
  255         if ((fp->f_flag & mode) == 0) {
  256                 FRELE(fp, curproc);
  257                 return (NULL);
  258         }
  259 
  260         return (fp);
  261 }
  262 
  263 int
  264 fd_checkclosed(struct filedesc *fdp, int fd, struct file *fp)
  265 {
  266         int closed;
  267 
  268         mtx_enter(&fdp->fd_fplock);
  269         KASSERT(fd < fdp->fd_nfiles);
  270         closed = (fdp->fd_ofiles[fd] != fp);
  271         mtx_leave(&fdp->fd_fplock);
  272         return (closed);
  273 }
  274 
  275 /*
  276  * System calls on descriptors.
  277  */
  278 
  279 /*
  280  * Duplicate a file descriptor.
  281  */
  282 int
  283 sys_dup(struct proc *p, void *v, register_t *retval)
  284 {
  285         struct sys_dup_args /* {
  286                 syscallarg(int) fd;
  287         } */ *uap = v;
  288         struct filedesc *fdp = p->p_fd;
  289         int old = SCARG(uap, fd);
  290         struct file *fp;
  291         int new;
  292         int error;
  293 
  294 restart:
  295         if ((fp = fd_getfile(fdp, old)) == NULL)
  296                 return (EBADF);
  297         fdplock(fdp);
  298         if ((error = fdalloc(p, 0, &new)) != 0) {
  299                 if (error == ENOSPC) {
  300                         fdexpand(p);
  301                         fdpunlock(fdp);
  302                         FRELE(fp, p);
  303                         goto restart;
  304                 }
  305                 fdpunlock(fdp);
  306                 FRELE(fp, p);
  307                 return (error);
  308         }
  309         /* No need for FRELE(), finishdup() uses current ref. */
  310         return (finishdup(p, fp, old, new, retval, 0));
  311 }
  312 
  313 /*
  314  * Duplicate a file descriptor to a particular value.
  315  */
  316 int
  317 sys_dup2(struct proc *p, void *v, register_t *retval)
  318 {
  319         struct sys_dup2_args /* {
  320                 syscallarg(int) from;
  321                 syscallarg(int) to;
  322         } */ *uap = v;
  323 
  324         return (dodup3(p, SCARG(uap, from), SCARG(uap, to), 0, retval));
  325 }
  326 
  327 int
  328 sys_dup3(struct proc *p, void *v, register_t *retval)
  329 {
  330         struct sys_dup3_args /* {
  331                 syscallarg(int) from;
  332                 syscallarg(int) to;
  333                 syscallarg(int) flags;
  334         } */ *uap = v;
  335 
  336         if (SCARG(uap, from) == SCARG(uap, to))
  337                 return (EINVAL);
  338         if (SCARG(uap, flags) & ~O_CLOEXEC)
  339                 return (EINVAL);
  340         return (dodup3(p, SCARG(uap, from), SCARG(uap, to),
  341             SCARG(uap, flags), retval));
  342 }
  343 
  344 int
  345 dodup3(struct proc *p, int old, int new, int flags, register_t *retval)
  346 {
  347         struct filedesc *fdp = p->p_fd;
  348         struct file *fp;
  349         int dupflags, error, i;
  350 
  351 restart:
  352         if ((fp = fd_getfile(fdp, old)) == NULL)
  353                 return (EBADF);
  354         if (old == new) {
  355                 /*
  356                  * NOTE! This doesn't clear the close-on-exec flag. This might
  357                  * or might not be the intended behavior from the start, but
  358                  * this is what everyone else does.
  359                  */
  360                 *retval = new;
  361                 FRELE(fp, p);
  362                 return (0);
  363         }
  364         if ((u_int)new >= lim_cur(RLIMIT_NOFILE) ||
  365             (u_int)new >= maxfiles) {
  366                 FRELE(fp, p);
  367                 return (EBADF);
  368         }
  369         fdplock(fdp);
  370         if (new >= fdp->fd_nfiles) {
  371                 if ((error = fdalloc(p, new, &i)) != 0) {
  372                         if (error == ENOSPC) {
  373                                 fdexpand(p);
  374                                 fdpunlock(fdp);
  375                                 FRELE(fp, p);
  376                                 goto restart;
  377                         }
  378                         fdpunlock(fdp);
  379                         FRELE(fp, p);
  380                         return (error);
  381                 }
  382                 if (new != i)
  383                         panic("dup2: fdalloc");
  384                 fd_unused(fdp, new);
  385         }
  386 
  387         dupflags = DUPF_DUP2;
  388         if (flags & O_CLOEXEC)
  389                 dupflags |= DUPF_CLOEXEC;
  390 
  391         /* No need for FRELE(), finishdup() uses current ref. */
  392         return (finishdup(p, fp, old, new, retval, dupflags));
  393 }
  394 
  395 /*
  396  * The file control system call.
  397  */
  398 int
  399 sys_fcntl(struct proc *p, void *v, register_t *retval)
  400 {
  401         struct sys_fcntl_args /* {
  402                 syscallarg(int) fd;
  403                 syscallarg(int) cmd;
  404                 syscallarg(void *) arg;
  405         } */ *uap = v;
  406         int fd = SCARG(uap, fd);
  407         struct filedesc *fdp = p->p_fd;
  408         struct file *fp;
  409         struct vnode *vp;
  410         int i, prev, tmp, newmin, flg = F_POSIX;
  411         struct flock fl;
  412         int error = 0;
  413 
  414         error = pledge_fcntl(p, SCARG(uap, cmd));
  415         if (error)
  416                 return (error);
  417 
  418 restart:
  419         if ((fp = fd_getfile(fdp, fd)) == NULL)
  420                 return (EBADF);
  421         switch (SCARG(uap, cmd)) {
  422 
  423         case F_DUPFD:
  424         case F_DUPFD_CLOEXEC:
  425                 newmin = (long)SCARG(uap, arg);
  426                 if ((u_int)newmin >= lim_cur(RLIMIT_NOFILE) ||
  427                     (u_int)newmin >= maxfiles) {
  428                         error = EINVAL;
  429                         break;
  430                 }
  431                 fdplock(fdp);
  432                 if ((error = fdalloc(p, newmin, &i)) != 0) {
  433                         if (error == ENOSPC) {
  434                                 fdexpand(p);
  435                                 fdpunlock(fdp);
  436                                 FRELE(fp, p);
  437                                 goto restart;
  438                         }
  439                         fdpunlock(fdp);
  440                         FRELE(fp, p);
  441                 } else {
  442                         int dupflags = 0;
  443 
  444                         if (SCARG(uap, cmd) == F_DUPFD_CLOEXEC)
  445                                 dupflags |= DUPF_CLOEXEC;
  446 
  447                         /* No need for FRELE(), finishdup() uses current ref. */
  448                         error = finishdup(p, fp, fd, i, retval, dupflags);
  449                 }
  450                 return (error);
  451 
  452         case F_GETFD:
  453                 fdplock(fdp);
  454                 *retval = fdp->fd_ofileflags[fd] & UF_EXCLOSE ? 1 : 0;
  455                 fdpunlock(fdp);
  456                 break;
  457 
  458         case F_SETFD:
  459                 fdplock(fdp);
  460                 if ((long)SCARG(uap, arg) & 1)
  461                         fdp->fd_ofileflags[fd] |= UF_EXCLOSE;
  462                 else
  463                         fdp->fd_ofileflags[fd] &= ~UF_EXCLOSE;
  464                 fdpunlock(fdp);
  465                 break;
  466 
  467         case F_GETFL:
  468                 *retval = OFLAGS(fp->f_flag);
  469                 break;
  470 
  471         case F_ISATTY:
  472                 vp = fp->f_data;
  473                 if (fp->f_type == DTYPE_VNODE && (vp->v_flag & VISTTY))
  474                         *retval = 1;
  475                 else {
  476                         *retval = 0;
  477                         error = ENOTTY;
  478                 }
  479                 break;
  480 
  481         case F_SETFL:
  482                 do {
  483                         tmp = prev = fp->f_flag;
  484                         tmp &= ~FCNTLFLAGS;
  485                         tmp |= FFLAGS((long)SCARG(uap, arg)) & FCNTLFLAGS;
  486                 } while (atomic_cas_uint(&fp->f_flag, prev, tmp) != prev);
  487                 tmp = fp->f_flag & FNONBLOCK;
  488                 error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p);
  489                 if (error)
  490                         break;
  491                 tmp = fp->f_flag & FASYNC;
  492                 error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p);
  493                 if (!error)
  494                         break;
  495                 atomic_clearbits_int(&fp->f_flag, FNONBLOCK);
  496                 tmp = 0;
  497                 (void) (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p);
  498                 break;
  499 
  500         case F_GETOWN:
  501                 tmp = 0;
  502                 error = (*fp->f_ops->fo_ioctl)
  503                         (fp, FIOGETOWN, (caddr_t)&tmp, p);
  504                 *retval = tmp;
  505                 break;
  506 
  507         case F_SETOWN:
  508                 tmp = (long)SCARG(uap, arg);
  509                 error = ((*fp->f_ops->fo_ioctl)
  510                         (fp, FIOSETOWN, (caddr_t)&tmp, p));
  511                 break;
  512 
  513         case F_SETLKW:
  514                 flg |= F_WAIT;
  515                 /* FALLTHROUGH */
  516 
  517         case F_SETLK:
  518                 error = pledge_flock(p);
  519                 if (error != 0)
  520                         break;
  521 
  522                 if (fp->f_type != DTYPE_VNODE) {
  523                         error = EINVAL;
  524                         break;
  525                 }
  526                 vp = fp->f_data;
  527                 /* Copy in the lock structure */
  528                 error = copyin((caddr_t)SCARG(uap, arg), (caddr_t)&fl,
  529                     sizeof (fl));
  530                 if (error)
  531                         break;
  532 #ifdef KTRACE
  533                 if (KTRPOINT(p, KTR_STRUCT))
  534                         ktrflock(p, &fl);
  535 #endif
  536                 if (fl.l_whence == SEEK_CUR) {
  537                         off_t offset = foffset(fp);
  538 
  539                         if (fl.l_start == 0 && fl.l_len < 0) {
  540                                 /* lockf(3) compliance hack */
  541                                 fl.l_len = -fl.l_len;
  542                                 fl.l_start = offset - fl.l_len;
  543                         } else
  544                                 fl.l_start += offset;
  545                 }
  546                 switch (fl.l_type) {
  547 
  548                 case F_RDLCK:
  549                         if ((fp->f_flag & FREAD) == 0) {
  550                                 error = EBADF;
  551                                 goto out;
  552                         }
  553                         atomic_setbits_int(&fdp->fd_flags, FD_ADVLOCK);
  554                         error = VOP_ADVLOCK(vp, fdp, F_SETLK, &fl, flg);
  555                         break;
  556 
  557                 case F_WRLCK:
  558                         if ((fp->f_flag & FWRITE) == 0) {
  559                                 error = EBADF;
  560                                 goto out;
  561                         }
  562                         atomic_setbits_int(&fdp->fd_flags, FD_ADVLOCK);
  563                         error = VOP_ADVLOCK(vp, fdp, F_SETLK, &fl, flg);
  564                         break;
  565 
  566                 case F_UNLCK:
  567                         error = VOP_ADVLOCK(vp, fdp, F_UNLCK, &fl, F_POSIX);
  568                         goto out;
  569 
  570                 default:
  571                         error = EINVAL;
  572                         goto out;
  573                 }
  574 
  575                 if (fd_checkclosed(fdp, fd, fp)) {
  576                         /*
  577                          * We have lost the race with close() or dup2();
  578                          * unlock, pretend that we've won the race and that
  579                          * lock had been removed by close()
  580                          */
  581                         fl.l_whence = SEEK_SET;
  582                         fl.l_start = 0;
  583                         fl.l_len = 0;
  584                         VOP_ADVLOCK(vp, fdp, F_UNLCK, &fl, F_POSIX);
  585                         fl.l_type = F_UNLCK;
  586                 }
  587                 goto out;
  588 
  589 
  590         case F_GETLK:
  591                 error = pledge_flock(p);
  592                 if (error != 0)
  593                         break;
  594 
  595                 if (fp->f_type != DTYPE_VNODE) {
  596                         error = EINVAL;
  597                         break;
  598                 }
  599                 vp = fp->f_data;
  600                 /* Copy in the lock structure */
  601                 error = copyin((caddr_t)SCARG(uap, arg), (caddr_t)&fl,
  602                     sizeof (fl));
  603                 if (error)
  604                         break;
  605                 if (fl.l_whence == SEEK_CUR) {
  606                         off_t offset = foffset(fp);
  607 
  608                         if (fl.l_start == 0 && fl.l_len < 0) {
  609                                 /* lockf(3) compliance hack */
  610                                 fl.l_len = -fl.l_len;
  611                                 fl.l_start = offset - fl.l_len;
  612                         } else
  613                                 fl.l_start += offset;
  614                 }
  615                 if (fl.l_type != F_RDLCK &&
  616                     fl.l_type != F_WRLCK &&
  617                     fl.l_type != F_UNLCK &&
  618                     fl.l_type != 0) {
  619                         error = EINVAL;
  620                         break;
  621                 }
  622                 error = VOP_ADVLOCK(vp, fdp, F_GETLK, &fl, F_POSIX);
  623                 if (error)
  624                         break;
  625 #ifdef KTRACE
  626                 if (KTRPOINT(p, KTR_STRUCT))
  627                         ktrflock(p, &fl);
  628 #endif
  629                 error = (copyout((caddr_t)&fl, (caddr_t)SCARG(uap, arg),
  630                     sizeof (fl)));
  631                 break;
  632 
  633         default:
  634                 error = EINVAL;
  635                 break;
  636         }
  637 out:
  638         FRELE(fp, p);
  639         return (error);
  640 }
  641 
  642 /*
  643  * Common code for dup, dup2, and fcntl(F_DUPFD).
  644  */
  645 int
  646 finishdup(struct proc *p, struct file *fp, int old, int new,
  647     register_t *retval, int dupflags)
  648 {
  649         struct file *oldfp;
  650         struct filedesc *fdp = p->p_fd;
  651         int error;
  652 
  653         fdpassertlocked(fdp);
  654         KASSERT(fp->f_iflags & FIF_INSERTED);
  655 
  656         if (fp->f_count >= FDUP_MAX_COUNT) {
  657                 error = EDEADLK;
  658                 goto fail;
  659         }
  660 
  661         oldfp = fd_getfile(fdp, new);
  662         if ((dupflags & DUPF_DUP2) && oldfp == NULL) {
  663                 if (fd_inuse(fdp, new)) {
  664                         error = EBUSY;
  665                         goto fail;
  666                 }
  667                 fd_used(fdp, new);
  668         }
  669 
  670         /*
  671          * Use `fd_fplock' to synchronize with fd_getfile() so that
  672          * the function no longer creates a new reference to the old file.
  673          */
  674         mtx_enter(&fdp->fd_fplock);
  675         fdp->fd_ofiles[new] = fp;
  676         mtx_leave(&fdp->fd_fplock);
  677 
  678         fdp->fd_ofileflags[new] = fdp->fd_ofileflags[old] & ~UF_EXCLOSE;
  679         if (dupflags & DUPF_CLOEXEC)
  680                 fdp->fd_ofileflags[new] |= UF_EXCLOSE;
  681         *retval = new;
  682 
  683         if (oldfp != NULL) {
  684                 knote_fdclose(p, new);
  685                 fdpunlock(fdp);
  686                 closef(oldfp, p);
  687         } else {
  688                 fdpunlock(fdp);
  689         }
  690 
  691         return (0);
  692 
  693 fail:
  694         fdpunlock(fdp);
  695         FRELE(fp, p);
  696         return (error);
  697 }
  698 
  699 void
  700 fdinsert(struct filedesc *fdp, int fd, int flags, struct file *fp)
  701 {
  702         struct file *fq;
  703 
  704         fdpassertlocked(fdp);
  705 
  706         mtx_enter(&fhdlk);
  707         if ((fp->f_iflags & FIF_INSERTED) == 0) {
  708                 atomic_setbits_int(&fp->f_iflags, FIF_INSERTED);
  709                 if ((fq = fdp->fd_ofiles[0]) != NULL) {
  710                         LIST_INSERT_AFTER(fq, fp, f_list);
  711                 } else {
  712                         LIST_INSERT_HEAD(&filehead, fp, f_list);
  713                 }
  714         }
  715         mtx_leave(&fhdlk);
  716 
  717         mtx_enter(&fdp->fd_fplock);
  718         KASSERT(fdp->fd_ofiles[fd] == NULL);
  719         fdp->fd_ofiles[fd] = fp;
  720         mtx_leave(&fdp->fd_fplock);
  721 
  722         fdp->fd_ofileflags[fd] |= (flags & UF_EXCLOSE);
  723 }
  724 
  725 void
  726 fdremove(struct filedesc *fdp, int fd)
  727 {
  728         fdpassertlocked(fdp);
  729 
  730         /*
  731          * Use `fd_fplock' to synchronize with fd_getfile() so that
  732          * the function no longer creates a new reference to the file.
  733          */
  734         mtx_enter(&fdp->fd_fplock);
  735         fdp->fd_ofiles[fd] = NULL;
  736         mtx_leave(&fdp->fd_fplock);
  737 
  738         fdp->fd_ofileflags[fd] = 0;
  739 
  740         fd_unused(fdp, fd);
  741 }
  742 
  743 int
  744 fdrelease(struct proc *p, int fd)
  745 {
  746         struct filedesc *fdp = p->p_fd;
  747         struct file *fp;
  748 
  749         fdpassertlocked(fdp);
  750 
  751         fp = fd_getfile(fdp, fd);
  752         if (fp == NULL) {
  753                 fdpunlock(fdp);
  754                 return (EBADF);
  755         }
  756         fdremove(fdp, fd);
  757         knote_fdclose(p, fd);
  758         fdpunlock(fdp);
  759         return (closef(fp, p));
  760 }
  761 
  762 /*
  763  * Close a file descriptor.
  764  */
  765 int
  766 sys_close(struct proc *p, void *v, register_t *retval)
  767 {
  768         struct sys_close_args /* {
  769                 syscallarg(int) fd;
  770         } */ *uap = v;
  771         int fd = SCARG(uap, fd), error;
  772         struct filedesc *fdp = p->p_fd;
  773 
  774         fdplock(fdp);
  775         /* fdrelease unlocks fdp. */
  776         error = fdrelease(p, fd);
  777 
  778         return (error);
  779 }
  780 
  781 /*
  782  * Return status information about a file descriptor.
  783  */
  784 int
  785 sys_fstat(struct proc *p, void *v, register_t *retval)
  786 {
  787         struct sys_fstat_args /* {
  788                 syscallarg(int) fd;
  789                 syscallarg(struct stat *) sb;
  790         } */ *uap = v;
  791         int fd = SCARG(uap, fd);
  792         struct filedesc *fdp = p->p_fd;
  793         struct file *fp;
  794         struct stat ub;
  795         int error;
  796 
  797         if ((fp = fd_getfile(fdp, fd)) == NULL)
  798                 return (EBADF);
  799         error = (*fp->f_ops->fo_stat)(fp, &ub, p);
  800         FRELE(fp, p);
  801         if (error == 0) {
  802                 /* 
  803                  * Don't let non-root see generation numbers
  804                  * (for NFS security)
  805                  */
  806                 if (suser(p))
  807                         ub.st_gen = 0;
  808                 error = copyout((caddr_t)&ub, (caddr_t)SCARG(uap, sb),
  809                     sizeof (ub));
  810         }
  811 #ifdef KTRACE
  812         if (error == 0 && KTRPOINT(p, KTR_STRUCT))
  813                 ktrstat(p, &ub);
  814 #endif
  815         return (error);
  816 }
  817 
  818 /*
  819  * Return pathconf information about a file descriptor.
  820  */
  821 int
  822 sys_fpathconf(struct proc *p, void *v, register_t *retval)
  823 {
  824         struct sys_fpathconf_args /* {
  825                 syscallarg(int) fd;
  826                 syscallarg(int) name;
  827         } */ *uap = v;
  828         int fd = SCARG(uap, fd);
  829         struct filedesc *fdp = p->p_fd;
  830         struct file *fp;
  831         struct vnode *vp;
  832         int error;
  833 
  834         if ((fp = fd_getfile(fdp, fd)) == NULL)
  835                 return (EBADF);
  836         switch (fp->f_type) {
  837         case DTYPE_PIPE:
  838         case DTYPE_SOCKET:
  839                 if (SCARG(uap, name) != _PC_PIPE_BUF) {
  840                         error = EINVAL;
  841                         break;
  842                 }
  843                 *retval = PIPE_BUF;
  844                 error = 0;
  845                 break;
  846 
  847         case DTYPE_VNODE:
  848                 vp = fp->f_data;
  849                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  850                 error = VOP_PATHCONF(vp, SCARG(uap, name), retval);
  851                 VOP_UNLOCK(vp);
  852                 break;
  853 
  854         default:
  855                 error = EOPNOTSUPP;
  856                 break;
  857         }
  858         FRELE(fp, p);
  859         return (error);
  860 }
  861 
  862 /*
  863  * Allocate a file descriptor for the process.
  864  */
  865 int
  866 fdalloc(struct proc *p, int want, int *result)
  867 {
  868         struct filedesc *fdp = p->p_fd;
  869         int lim, last, i;
  870         u_int new, off;
  871 
  872         fdpassertlocked(fdp);
  873 
  874         /*
  875          * Search for a free descriptor starting at the higher
  876          * of want or fd_freefile.  If that fails, consider
  877          * expanding the ofile array.
  878          */
  879 restart:
  880         lim = min((int)lim_cur(RLIMIT_NOFILE), maxfiles);
  881         last = min(fdp->fd_nfiles, lim);
  882         if ((i = want) < fdp->fd_freefile)
  883                 i = fdp->fd_freefile;
  884         off = i >> NDENTRYSHIFT;
  885         new = find_next_zero(fdp->fd_himap, off,
  886             (last + NDENTRIES - 1) >> NDENTRYSHIFT);
  887         if (new != -1) {
  888                 i = find_next_zero(&fdp->fd_lomap[new], 
  889                                    new > off ? 0 : i & NDENTRYMASK,
  890                                    NDENTRIES);
  891                 if (i == -1) {
  892                         /*
  893                          * Free file descriptor in this block was
  894                          * below want, try again with higher want.
  895                          */
  896                         want = (new + 1) << NDENTRYSHIFT;
  897                         goto restart;
  898                 }
  899                 i += (new << NDENTRYSHIFT);
  900                 if (i < last) {
  901                         fd_used(fdp, i);
  902                         if (want <= fdp->fd_freefile)
  903                                 fdp->fd_freefile = i;
  904                         *result = i;
  905                         fdp->fd_ofileflags[i] = 0;
  906                         if (ISSET(p->p_p->ps_flags, PS_PLEDGE))
  907                                 fdp->fd_ofileflags[i] |= UF_PLEDGED;
  908                         return (0);
  909                 }
  910         }
  911         if (fdp->fd_nfiles >= lim)
  912                 return (EMFILE);
  913 
  914         return (ENOSPC);
  915 }
  916 
  917 void
  918 fdexpand(struct proc *p)
  919 {
  920         struct filedesc *fdp = p->p_fd;
  921         int nfiles, oldnfiles;
  922         size_t copylen;
  923         struct file **newofile, **oldofile;
  924         char *newofileflags;
  925         u_int *newhimap, *newlomap;
  926 
  927         fdpassertlocked(fdp);
  928 
  929         oldnfiles = fdp->fd_nfiles;
  930         oldofile = fdp->fd_ofiles;
  931 
  932         /*
  933          * No space in current array.
  934          */
  935         if (fdp->fd_nfiles < NDEXTENT)
  936                 nfiles = NDEXTENT;
  937         else
  938                 nfiles = 2 * fdp->fd_nfiles;
  939 
  940         newofile = mallocarray(nfiles, OFILESIZE, M_FILEDESC, M_WAITOK);
  941         /*
  942          * Allocate all required chunks before calling free(9) to make
  943          * sure that ``fd_ofiles'' stays valid if we go to sleep.
  944          */
  945         if (NDHISLOTS(nfiles) > NDHISLOTS(fdp->fd_nfiles)) {
  946                 newhimap = mallocarray(NDHISLOTS(nfiles), sizeof(u_int),
  947                     M_FILEDESC, M_WAITOK);
  948                 newlomap = mallocarray(NDLOSLOTS(nfiles), sizeof(u_int),
  949                     M_FILEDESC, M_WAITOK);
  950         }
  951         newofileflags = (char *) &newofile[nfiles];
  952 
  953         /*
  954          * Copy the existing ofile and ofileflags arrays
  955          * and zero the new portion of each array.
  956          */
  957         copylen = sizeof(struct file *) * fdp->fd_nfiles;
  958         memcpy(newofile, fdp->fd_ofiles, copylen);
  959         memset((char *)newofile + copylen, 0,
  960             nfiles * sizeof(struct file *) - copylen);
  961         copylen = sizeof(char) * fdp->fd_nfiles;
  962         memcpy(newofileflags, fdp->fd_ofileflags, copylen);
  963         memset(newofileflags + copylen, 0, nfiles * sizeof(char) - copylen);
  964 
  965         if (NDHISLOTS(nfiles) > NDHISLOTS(fdp->fd_nfiles)) {
  966                 copylen = NDHISLOTS(fdp->fd_nfiles) * sizeof(u_int);
  967                 memcpy(newhimap, fdp->fd_himap, copylen);
  968                 memset((char *)newhimap + copylen, 0,
  969                     NDHISLOTS(nfiles) * sizeof(u_int) - copylen);
  970 
  971                 copylen = NDLOSLOTS(fdp->fd_nfiles) * sizeof(u_int);
  972                 memcpy(newlomap, fdp->fd_lomap, copylen);
  973                 memset((char *)newlomap + copylen, 0,
  974                     NDLOSLOTS(nfiles) * sizeof(u_int) - copylen);
  975 
  976                 if (NDHISLOTS(fdp->fd_nfiles) > NDHISLOTS(NDFILE)) {
  977                         free(fdp->fd_himap, M_FILEDESC,
  978                             NDHISLOTS(fdp->fd_nfiles) * sizeof(u_int));
  979                         free(fdp->fd_lomap, M_FILEDESC,
  980                             NDLOSLOTS(fdp->fd_nfiles) * sizeof(u_int));
  981                 }
  982                 fdp->fd_himap = newhimap;
  983                 fdp->fd_lomap = newlomap;
  984         }
  985 
  986         mtx_enter(&fdp->fd_fplock);
  987         fdp->fd_ofiles = newofile;
  988         mtx_leave(&fdp->fd_fplock);
  989 
  990         fdp->fd_ofileflags = newofileflags;
  991         fdp->fd_nfiles = nfiles;
  992 
  993         if (oldnfiles > NDFILE)
  994                 free(oldofile, M_FILEDESC, oldnfiles * OFILESIZE);
  995 }
  996 
  997 /*
  998  * Create a new open file structure and allocate
  999  * a file descriptor for the process that refers to it.
 1000  */
 1001 int
 1002 falloc(struct proc *p, struct file **resultfp, int *resultfd)
 1003 {
 1004         struct file *fp;
 1005         int error, i;
 1006 
 1007         KASSERT(resultfp != NULL);
 1008         KASSERT(resultfd != NULL);
 1009 
 1010         fdpassertlocked(p->p_fd);
 1011 restart:
 1012         if ((error = fdalloc(p, 0, &i)) != 0) {
 1013                 if (error == ENOSPC) {
 1014                         fdexpand(p);
 1015                         goto restart;
 1016                 }
 1017                 return (error);
 1018         }
 1019 
 1020         fp = fnew(p);
 1021         if (fp == NULL) {
 1022                 fd_unused(p->p_fd, i);
 1023                 return (ENFILE);
 1024         }
 1025 
 1026         FREF(fp);
 1027         *resultfp = fp;
 1028         *resultfd = i;
 1029 
 1030         return (0);
 1031 }
 1032 
 1033 struct file *
 1034 fnew(struct proc *p)
 1035 {
 1036         struct file *fp;
 1037         int nfiles;
 1038 
 1039         nfiles = atomic_inc_int_nv(&numfiles);
 1040         if (nfiles > maxfiles) {
 1041                 atomic_dec_int(&numfiles);
 1042                 tablefull("file");
 1043                 return (NULL);
 1044         }
 1045 
 1046         fp = pool_get(&file_pool, PR_WAITOK|PR_ZERO);
 1047         /*
 1048          * We need to block interrupts as long as `f_mtx' is being taken
 1049          * with and without the KERNEL_LOCK().
 1050          */
 1051         mtx_init(&fp->f_mtx, IPL_MPFLOOR);
 1052         fp->f_count = 1;
 1053         fp->f_cred = p->p_ucred;
 1054         crhold(fp->f_cred);
 1055 
 1056         return (fp);
 1057 }
 1058 
 1059 /*
 1060  * Build a new filedesc structure.
 1061  */
 1062 struct filedesc *
 1063 fdinit(void)
 1064 {
 1065         struct filedesc0 *newfdp;
 1066 
 1067         newfdp = pool_get(&fdesc_pool, PR_WAITOK|PR_ZERO);
 1068         rw_init(&newfdp->fd_fd.fd_lock, "fdlock");
 1069         mtx_init(&newfdp->fd_fd.fd_fplock, IPL_MPFLOOR);
 1070         LIST_INIT(&newfdp->fd_fd.fd_kqlist);
 1071 
 1072         /* Create the file descriptor table. */
 1073         newfdp->fd_fd.fd_refcnt = 1;
 1074         newfdp->fd_fd.fd_cmask = S_IWGRP|S_IWOTH;
 1075         newfdp->fd_fd.fd_ofiles = newfdp->fd_dfiles;
 1076         newfdp->fd_fd.fd_ofileflags = newfdp->fd_dfileflags;
 1077         newfdp->fd_fd.fd_nfiles = NDFILE;
 1078         newfdp->fd_fd.fd_himap = newfdp->fd_dhimap;
 1079         newfdp->fd_fd.fd_lomap = newfdp->fd_dlomap;
 1080 
 1081         newfdp->fd_fd.fd_freefile = 0;
 1082         newfdp->fd_fd.fd_lastfile = 0;
 1083 
 1084         return (&newfdp->fd_fd);
 1085 }
 1086 
 1087 /*
 1088  * Share a filedesc structure.
 1089  */
 1090 struct filedesc *
 1091 fdshare(struct process *pr)
 1092 {
 1093         pr->ps_fd->fd_refcnt++;
 1094         return (pr->ps_fd);
 1095 }
 1096 
 1097 /*
 1098  * Copy a filedesc structure.
 1099  */
 1100 struct filedesc *
 1101 fdcopy(struct process *pr)
 1102 {
 1103         struct filedesc *newfdp, *fdp = pr->ps_fd;
 1104         int i;
 1105 
 1106         newfdp = fdinit();
 1107 
 1108         fdplock(fdp);
 1109         if (fdp->fd_cdir) {
 1110                 vref(fdp->fd_cdir);
 1111                 newfdp->fd_cdir = fdp->fd_cdir;
 1112         }
 1113         if (fdp->fd_rdir) {
 1114                 vref(fdp->fd_rdir);
 1115                 newfdp->fd_rdir = fdp->fd_rdir;
 1116         }
 1117 
 1118         /*
 1119          * If the number of open files fits in the internal arrays
 1120          * of the open file structure, use them, otherwise allocate
 1121          * additional memory for the number of descriptors currently
 1122          * in use.
 1123          */
 1124         if (fdp->fd_lastfile >= NDFILE) {
 1125                 /*
 1126                  * Compute the smallest multiple of NDEXTENT needed
 1127                  * for the file descriptors currently in use,
 1128                  * allowing the table to shrink.
 1129                  */
 1130                 i = fdp->fd_nfiles;
 1131                 while (i >= 2 * NDEXTENT && i > fdp->fd_lastfile * 2)
 1132                         i /= 2;
 1133                 newfdp->fd_ofiles = mallocarray(i, OFILESIZE, M_FILEDESC,
 1134                     M_WAITOK | M_ZERO);
 1135                 newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
 1136                 newfdp->fd_nfiles = i;
 1137         }
 1138         if (NDHISLOTS(newfdp->fd_nfiles) > NDHISLOTS(NDFILE)) {
 1139                 newfdp->fd_himap = mallocarray(NDHISLOTS(newfdp->fd_nfiles),
 1140                     sizeof(u_int), M_FILEDESC, M_WAITOK | M_ZERO);
 1141                 newfdp->fd_lomap = mallocarray(NDLOSLOTS(newfdp->fd_nfiles),
 1142                     sizeof(u_int), M_FILEDESC, M_WAITOK | M_ZERO);
 1143         }
 1144         newfdp->fd_freefile = fdp->fd_freefile;
 1145         newfdp->fd_flags = fdp->fd_flags;
 1146         newfdp->fd_cmask = fdp->fd_cmask;
 1147 
 1148         for (i = 0; i <= fdp->fd_lastfile; i++) {
 1149                 struct file *fp = fdp->fd_ofiles[i];
 1150 
 1151                 if (fp != NULL) {
 1152                         /*
 1153                          * XXX Gruesome hack. If count gets too high, fail
 1154                          * to copy an fd, since fdcopy()'s callers do not
 1155                          * permit it to indicate failure yet.
 1156                          * Meanwhile, kqueue files have to be
 1157                          * tied to the process that opened them to enforce
 1158                          * their internal consistency, so close them here.
 1159                          */
 1160                         if (fp->f_count >= FDUP_MAX_COUNT ||
 1161                             fp->f_type == DTYPE_KQUEUE) {
 1162                                 if (i < newfdp->fd_freefile)
 1163                                         newfdp->fd_freefile = i;
 1164                                 continue;
 1165                         }
 1166 
 1167                         FREF(fp);
 1168                         newfdp->fd_ofiles[i] = fp;
 1169                         newfdp->fd_ofileflags[i] = fdp->fd_ofileflags[i];
 1170                         fd_used(newfdp, i);
 1171                 }
 1172         }
 1173         fdpunlock(fdp);
 1174 
 1175         return (newfdp);
 1176 }
 1177 
 1178 /*
 1179  * Release a filedesc structure.
 1180  */
 1181 void
 1182 fdfree(struct proc *p)
 1183 {
 1184         struct filedesc *fdp = p->p_fd;
 1185         struct file *fp;
 1186         int fd;
 1187 
 1188         if (--fdp->fd_refcnt > 0)
 1189                 return;
 1190         for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
 1191                 fp = fdp->fd_ofiles[fd];
 1192                 if (fp != NULL) {
 1193                         fdp->fd_ofiles[fd] = NULL;
 1194                         knote_fdclose(p, fd);
 1195                          /* closef() expects a refcount of 2 */
 1196                         FREF(fp);
 1197                         (void) closef(fp, p);
 1198                 }
 1199         }
 1200         p->p_fd = NULL;
 1201         if (fdp->fd_nfiles > NDFILE)
 1202                 free(fdp->fd_ofiles, M_FILEDESC, fdp->fd_nfiles * OFILESIZE);
 1203         if (NDHISLOTS(fdp->fd_nfiles) > NDHISLOTS(NDFILE)) {
 1204                 free(fdp->fd_himap, M_FILEDESC,
 1205                     NDHISLOTS(fdp->fd_nfiles) * sizeof(u_int));
 1206                 free(fdp->fd_lomap, M_FILEDESC,
 1207                     NDLOSLOTS(fdp->fd_nfiles) * sizeof(u_int));
 1208         }
 1209         if (fdp->fd_cdir)
 1210                 vrele(fdp->fd_cdir);
 1211         if (fdp->fd_rdir)
 1212                 vrele(fdp->fd_rdir);
 1213         pool_put(&fdesc_pool, fdp);
 1214 }
 1215 
 1216 /*
 1217  * Internal form of close.
 1218  * Decrement reference count on file structure.
 1219  * Note: p may be NULL when closing a file
 1220  * that was being passed in a message.
 1221  *
 1222  * The fp must have its usecount bumped and will be FRELEd here.
 1223  */
 1224 int
 1225 closef(struct file *fp, struct proc *p)
 1226 {
 1227         struct filedesc *fdp;
 1228 
 1229         if (fp == NULL)
 1230                 return (0);
 1231 
 1232         KASSERTMSG(fp->f_count >= 2, "count (%u) < 2", fp->f_count);
 1233 
 1234         atomic_dec_int(&fp->f_count);
 1235 
 1236         /*
 1237          * POSIX record locking dictates that any close releases ALL
 1238          * locks owned by this process.  This is handled by setting
 1239          * a flag in the unlock to free ONLY locks obeying POSIX
 1240          * semantics, and not to free BSD-style file locks.
 1241          * If the descriptor was in a message, POSIX-style locks
 1242          * aren't passed with the descriptor.
 1243          */
 1244 
 1245         if (p && ((fdp = p->p_fd) != NULL) &&
 1246             (fdp->fd_flags & FD_ADVLOCK) &&
 1247             fp->f_type == DTYPE_VNODE) {
 1248                 struct vnode *vp = fp->f_data;
 1249                 struct flock lf;
 1250 
 1251                 lf.l_whence = SEEK_SET;
 1252                 lf.l_start = 0;
 1253                 lf.l_len = 0;
 1254                 lf.l_type = F_UNLCK;
 1255                 (void) VOP_ADVLOCK(vp, fdp, F_UNLCK, &lf, F_POSIX);
 1256         }
 1257 
 1258         return (FRELE(fp, p));
 1259 }
 1260 
 1261 int
 1262 fdrop(struct file *fp, struct proc *p)
 1263 {
 1264         int error;
 1265 
 1266         KASSERTMSG(fp->f_count == 0, "count (%u) != 0", fp->f_count);
 1267 
 1268         mtx_enter(&fhdlk);
 1269         if (fp->f_iflags & FIF_INSERTED)
 1270                 LIST_REMOVE(fp, f_list);
 1271         mtx_leave(&fhdlk);
 1272 
 1273         if (fp->f_ops)
 1274                 error = (*fp->f_ops->fo_close)(fp, p);
 1275         else
 1276                 error = 0;
 1277 
 1278         crfree(fp->f_cred);
 1279         atomic_dec_int(&numfiles);
 1280         pool_put(&file_pool, fp);
 1281 
 1282         return (error);
 1283 }
 1284 
 1285 /*
 1286  * Apply an advisory lock on a file descriptor.
 1287  *
 1288  * Just attempt to get a record lock of the requested type on
 1289  * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
 1290  */
 1291 int
 1292 sys_flock(struct proc *p, void *v, register_t *retval)
 1293 {
 1294         struct sys_flock_args /* {
 1295                 syscallarg(int) fd;
 1296                 syscallarg(int) how;
 1297         } */ *uap = v;
 1298         int fd = SCARG(uap, fd);
 1299         int how = SCARG(uap, how);
 1300         struct filedesc *fdp = p->p_fd;
 1301         struct file *fp;
 1302         struct vnode *vp;
 1303         struct flock lf;
 1304         int error;
 1305 
 1306         if ((fp = fd_getfile(fdp, fd)) == NULL)
 1307                 return (EBADF);
 1308         if (fp->f_type != DTYPE_VNODE) {
 1309                 error = EOPNOTSUPP;
 1310                 goto out;
 1311         }
 1312         vp = fp->f_data;
 1313         lf.l_whence = SEEK_SET;
 1314         lf.l_start = 0;
 1315         lf.l_len = 0;
 1316         if (how & LOCK_UN) {
 1317                 lf.l_type = F_UNLCK;
 1318                 atomic_clearbits_int(&fp->f_iflags, FIF_HASLOCK);
 1319                 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK);
 1320                 goto out;
 1321         }
 1322         if (how & LOCK_EX)
 1323                 lf.l_type = F_WRLCK;
 1324         else if (how & LOCK_SH)
 1325                 lf.l_type = F_RDLCK;
 1326         else {
 1327                 error = EINVAL;
 1328                 goto out;
 1329         }
 1330         atomic_setbits_int(&fp->f_iflags, FIF_HASLOCK);
 1331         if (how & LOCK_NB)
 1332                 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK);
 1333         else
 1334                 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK|F_WAIT);
 1335 out:
 1336         FRELE(fp, p);
 1337         return (error);
 1338 }
 1339 
 1340 /*
 1341  * File Descriptor pseudo-device driver (/dev/fd/).
 1342  *
 1343  * Opening minor device N dup()s the file (if any) connected to file
 1344  * descriptor N belonging to the calling process.  Note that this driver
 1345  * consists of only the ``open()'' routine, because all subsequent
 1346  * references to this file will be direct to the other driver.
 1347  */
 1348 int
 1349 filedescopen(dev_t dev, int mode, int type, struct proc *p)
 1350 {
 1351 
 1352         /*
 1353          * XXX Kludge: set curproc->p_dupfd to contain the value of the
 1354          * the file descriptor being sought for duplication. The error
 1355          * return ensures that the vnode for this device will be released
 1356          * by vn_open. Open will detect this special error and take the
 1357          * actions in dupfdopen below. Other callers of vn_open or VOP_OPEN
 1358          * will simply report the error.
 1359          */
 1360         p->p_dupfd = minor(dev);
 1361         return (ENODEV);
 1362 }
 1363 
 1364 /*
 1365  * Duplicate the specified descriptor to a free descriptor.
 1366  */
 1367 int
 1368 dupfdopen(struct proc *p, int indx, int mode)
 1369 {
 1370         struct filedesc *fdp = p->p_fd;
 1371         int dupfd = p->p_dupfd;
 1372         struct file *wfp;
 1373 
 1374         fdpassertlocked(fdp);
 1375 
 1376         /*
 1377          * Assume that the filename was user-specified; applications do
 1378          * not tend to open /dev/fd/# when they can just call dup()
 1379          */
 1380         if ((p->p_p->ps_flags & (PS_SUGIDEXEC | PS_SUGID))) {
 1381                 if (p->p_descfd == 255)
 1382                         return (EPERM);
 1383                 if (p->p_descfd != dupfd)
 1384                         return (EPERM);
 1385         }
 1386 
 1387         /*
 1388          * If the to-be-dup'd fd number is greater than the allowed number
 1389          * of file descriptors, or the fd to be dup'd has already been
 1390          * closed, reject. Note, there is no need to check for new == old
 1391          * because fd_getfile will return NULL if the file at indx is
 1392          * newly created by falloc.
 1393          */
 1394         if ((wfp = fd_getfile(fdp, dupfd)) == NULL)
 1395                 return (EBADF);
 1396 
 1397         /*
 1398          * Check that the mode the file is being opened for is a
 1399          * subset of the mode of the existing descriptor.
 1400          */
 1401         if (((mode & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag) {
 1402                 FRELE(wfp, p);
 1403                 return (EACCES);
 1404         }
 1405         if (wfp->f_count >= FDUP_MAX_COUNT) {
 1406                 FRELE(wfp, p);
 1407                 return (EDEADLK);
 1408         }
 1409 
 1410         KASSERT(wfp->f_iflags & FIF_INSERTED);
 1411 
 1412         mtx_enter(&fdp->fd_fplock);
 1413         KASSERT(fdp->fd_ofiles[indx] == NULL);
 1414         fdp->fd_ofiles[indx] = wfp;
 1415         mtx_leave(&fdp->fd_fplock);
 1416 
 1417         fdp->fd_ofileflags[indx] = (fdp->fd_ofileflags[indx] & UF_EXCLOSE) |
 1418             (fdp->fd_ofileflags[dupfd] & ~UF_EXCLOSE);
 1419 
 1420         return (0);
 1421 }
 1422 
 1423 /*
 1424  * Close any files on exec?
 1425  */
 1426 void
 1427 fdcloseexec(struct proc *p)
 1428 {
 1429         struct filedesc *fdp = p->p_fd;
 1430         int fd;
 1431 
 1432         fdplock(fdp);
 1433         for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
 1434                 fdp->fd_ofileflags[fd] &= ~UF_PLEDGED;
 1435                 if (fdp->fd_ofileflags[fd] & UF_EXCLOSE) {
 1436                         /* fdrelease() unlocks fdp. */
 1437                         (void) fdrelease(p, fd);
 1438                         fdplock(fdp);
 1439                 }
 1440         }
 1441         fdpunlock(fdp);
 1442 }
 1443 
 1444 int
 1445 sys_closefrom(struct proc *p, void *v, register_t *retval)
 1446 {
 1447         struct sys_closefrom_args *uap = v;
 1448         struct filedesc *fdp = p->p_fd;
 1449         u_int startfd, i;
 1450 
 1451         startfd = SCARG(uap, fd);
 1452         fdplock(fdp);
 1453 
 1454         if (startfd > fdp->fd_lastfile) {
 1455                 fdpunlock(fdp);
 1456                 return (EBADF);
 1457         }
 1458 
 1459         for (i = startfd; i <= fdp->fd_lastfile; i++) {
 1460                 /* fdrelease() unlocks fdp. */
 1461                 fdrelease(p, i);
 1462                 fdplock(fdp);
 1463         }
 1464 
 1465         fdpunlock(fdp);
 1466         return (0);
 1467 }
 1468 
 1469 int
 1470 sys_getdtablecount(struct proc *p, void *v, register_t *retval)
 1471 {
 1472         *retval = p->p_fd->fd_openfd;
 1473         return (0);
 1474 }
Cache object: 3dff1000ccf759deff1d68b91b70304a
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.
FreeBSD/Linux Kernel Cross Reference sys/kern/kern_descrip.c

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