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_vnops.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 /*-
    2  * Copyright (c) 1982, 1986, 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_vnops.c 8.2 (Berkeley) 1/21/94
   35  */
   36 
   37 #include <sys/cdefs.h>
   38 __FBSDID("$FreeBSD: releng/8.3/sys/kern/vfs_vnops.c 229725 2012-01-06 19:32:39Z jhb $");
   39 
   40 #include <sys/param.h>
   41 #include <sys/systm.h>
   42 #include <sys/fcntl.h>
   43 #include <sys/file.h>
   44 #include <sys/kdb.h>
   45 #include <sys/stat.h>
   46 #include <sys/priv.h>
   47 #include <sys/proc.h>
   48 #include <sys/limits.h>
   49 #include <sys/lock.h>
   50 #include <sys/mount.h>
   51 #include <sys/mutex.h>
   52 #include <sys/namei.h>
   53 #include <sys/vnode.h>
   54 #include <sys/bio.h>
   55 #include <sys/buf.h>
   56 #include <sys/filio.h>
   57 #include <sys/resourcevar.h>
   58 #include <sys/sx.h>
   59 #include <sys/ttycom.h>
   60 #include <sys/conf.h>
   61 #include <sys/syslog.h>
   62 #include <sys/unistd.h>
   63 
   64 #include <security/mac/mac_framework.h>
   65 
   66 #include <vm/vm.h>
   67 #include <vm/vm_object.h>
   68 
   69 static fo_rdwr_t        vn_read;
   70 static fo_rdwr_t        vn_write;
   71 static fo_truncate_t    vn_truncate;
   72 static fo_ioctl_t       vn_ioctl;
   73 static fo_poll_t        vn_poll;
   74 static fo_kqfilter_t    vn_kqfilter;
   75 static fo_stat_t        vn_statfile;
   76 static fo_close_t       vn_closefile;
   77 
   78 struct  fileops vnops = {
   79         .fo_read = vn_read,
   80         .fo_write = vn_write,
   81         .fo_truncate = vn_truncate,
   82         .fo_ioctl = vn_ioctl,
   83         .fo_poll = vn_poll,
   84         .fo_kqfilter = vn_kqfilter,
   85         .fo_stat = vn_statfile,
   86         .fo_close = vn_closefile,
   87         .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE
   88 };
   89 
   90 int
   91 vn_open(ndp, flagp, cmode, fp)
   92         struct nameidata *ndp;
   93         int *flagp, cmode;
   94         struct file *fp;
   95 {
   96         struct thread *td = ndp->ni_cnd.cn_thread;
   97 
   98         return (vn_open_cred(ndp, flagp, cmode, 0, td->td_ucred, fp));
   99 }
  100 
  101 /*
  102  * Common code for vnode open operations.
  103  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
  104  * 
  105  * Note that this does NOT free nameidata for the successful case,
  106  * due to the NDINIT being done elsewhere.
  107  */
  108 int
  109 vn_open_cred(struct nameidata *ndp, int *flagp, int cmode, u_int vn_open_flags,
  110     struct ucred *cred, struct file *fp)
  111 {
  112         struct vnode *vp;
  113         struct mount *mp;
  114         struct thread *td = ndp->ni_cnd.cn_thread;
  115         struct vattr vat;
  116         struct vattr *vap = &vat;
  117         int fmode, error;
  118         accmode_t accmode;
  119         int vfslocked, mpsafe;
  120 
  121         mpsafe = ndp->ni_cnd.cn_flags & MPSAFE;
  122 restart:
  123         vfslocked = 0;
  124         fmode = *flagp;
  125         if (fmode & O_CREAT) {
  126                 ndp->ni_cnd.cn_nameiop = CREATE;
  127                 ndp->ni_cnd.cn_flags = ISOPEN | LOCKPARENT | LOCKLEAF |
  128                     MPSAFE;
  129                 if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
  130                         ndp->ni_cnd.cn_flags |= FOLLOW;
  131                 if (!(vn_open_flags & VN_OPEN_NOAUDIT))
  132                         ndp->ni_cnd.cn_flags |= AUDITVNODE1;
  133                 bwillwrite();
  134                 if ((error = namei(ndp)) != 0)
  135                         return (error);
  136                 vfslocked = NDHASGIANT(ndp);
  137                 if (!mpsafe)
  138                         ndp->ni_cnd.cn_flags &= ~MPSAFE;
  139                 if (ndp->ni_vp == NULL) {
  140                         VATTR_NULL(vap);
  141                         vap->va_type = VREG;
  142                         vap->va_mode = cmode;
  143                         if (fmode & O_EXCL)
  144                                 vap->va_vaflags |= VA_EXCLUSIVE;
  145                         if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) {
  146                                 NDFREE(ndp, NDF_ONLY_PNBUF);
  147                                 vput(ndp->ni_dvp);
  148                                 VFS_UNLOCK_GIANT(vfslocked);
  149                                 if ((error = vn_start_write(NULL, &mp,
  150                                     V_XSLEEP | PCATCH)) != 0)
  151                                         return (error);
  152                                 goto restart;
  153                         }
  154 #ifdef MAC
  155                         error = mac_vnode_check_create(cred, ndp->ni_dvp,
  156                             &ndp->ni_cnd, vap);
  157                         if (error == 0)
  158 #endif
  159                                 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
  160                                                    &ndp->ni_cnd, vap);
  161                         vput(ndp->ni_dvp);
  162                         vn_finished_write(mp);
  163                         if (error) {
  164                                 VFS_UNLOCK_GIANT(vfslocked);
  165                                 NDFREE(ndp, NDF_ONLY_PNBUF);
  166                                 return (error);
  167                         }
  168                         fmode &= ~O_TRUNC;
  169                         vp = ndp->ni_vp;
  170                 } else {
  171                         if (ndp->ni_dvp == ndp->ni_vp)
  172                                 vrele(ndp->ni_dvp);
  173                         else
  174                                 vput(ndp->ni_dvp);
  175                         ndp->ni_dvp = NULL;
  176                         vp = ndp->ni_vp;
  177                         if (fmode & O_EXCL) {
  178                                 error = EEXIST;
  179                                 goto bad;
  180                         }
  181                         fmode &= ~O_CREAT;
  182                 }
  183         } else {
  184                 ndp->ni_cnd.cn_nameiop = LOOKUP;
  185                 ndp->ni_cnd.cn_flags = ISOPEN |
  186                     ((fmode & O_NOFOLLOW) ? NOFOLLOW : FOLLOW) |
  187                     LOCKLEAF | MPSAFE;
  188                 if (!(fmode & FWRITE))
  189                         ndp->ni_cnd.cn_flags |= LOCKSHARED;
  190                 if (!(vn_open_flags & VN_OPEN_NOAUDIT))
  191                         ndp->ni_cnd.cn_flags |= AUDITVNODE1;
  192                 if ((error = namei(ndp)) != 0)
  193                         return (error);
  194                 if (!mpsafe)
  195                         ndp->ni_cnd.cn_flags &= ~MPSAFE;
  196                 vfslocked = NDHASGIANT(ndp);
  197                 vp = ndp->ni_vp;
  198         }
  199         if (vp->v_type == VLNK) {
  200                 error = EMLINK;
  201                 goto bad;
  202         }
  203         if (vp->v_type == VSOCK) {
  204                 error = EOPNOTSUPP;
  205                 goto bad;
  206         }
  207         accmode = 0;
  208         if (fmode & (FWRITE | O_TRUNC)) {
  209                 if (vp->v_type == VDIR) {
  210                         error = EISDIR;
  211                         goto bad;
  212                 }
  213                 accmode |= VWRITE;
  214         }
  215         if (fmode & FREAD)
  216                 accmode |= VREAD;
  217         if (fmode & FEXEC)
  218                 accmode |= VEXEC;
  219         if ((fmode & O_APPEND) && (fmode & FWRITE))
  220                 accmode |= VAPPEND;
  221 #ifdef MAC
  222         error = mac_vnode_check_open(cred, vp, accmode);
  223         if (error)
  224                 goto bad;
  225 #endif
  226         if ((fmode & O_CREAT) == 0) {
  227                 if (accmode & VWRITE) {
  228                         error = vn_writechk(vp);
  229                         if (error)
  230                                 goto bad;
  231                 }
  232                 if (accmode) {
  233                         error = VOP_ACCESS(vp, accmode, cred, td);
  234                         if (error)
  235                                 goto bad;
  236                 }
  237         }
  238         if ((error = VOP_OPEN(vp, fmode, cred, td, fp)) != 0)
  239                 goto bad;
  240 
  241         if (fmode & FWRITE)
  242                 vp->v_writecount++;
  243         *flagp = fmode;
  244         ASSERT_VOP_LOCKED(vp, "vn_open_cred");
  245         if (!mpsafe)
  246                 VFS_UNLOCK_GIANT(vfslocked);
  247         return (0);
  248 bad:
  249         NDFREE(ndp, NDF_ONLY_PNBUF);
  250         vput(vp);
  251         VFS_UNLOCK_GIANT(vfslocked);
  252         *flagp = fmode;
  253         ndp->ni_vp = NULL;
  254         return (error);
  255 }
  256 
  257 /*
  258  * Check for write permissions on the specified vnode.
  259  * Prototype text segments cannot be written.
  260  */
  261 int
  262 vn_writechk(vp)
  263         register struct vnode *vp;
  264 {
  265 
  266         ASSERT_VOP_LOCKED(vp, "vn_writechk");
  267         /*
  268          * If there's shared text associated with
  269          * the vnode, try to free it up once.  If
  270          * we fail, we can't allow writing.
  271          */
  272         if (vp->v_vflag & VV_TEXT)
  273                 return (ETXTBSY);
  274 
  275         return (0);
  276 }
  277 
  278 /*
  279  * Vnode close call
  280  */
  281 int
  282 vn_close(vp, flags, file_cred, td)
  283         register struct vnode *vp;
  284         int flags;
  285         struct ucred *file_cred;
  286         struct thread *td;
  287 {
  288         struct mount *mp;
  289         int error, lock_flags;
  290 
  291         if (!(flags & FWRITE) && vp->v_mount != NULL &&
  292             vp->v_mount->mnt_kern_flag & MNTK_EXTENDED_SHARED)
  293                 lock_flags = LK_SHARED;
  294         else
  295                 lock_flags = LK_EXCLUSIVE;
  296 
  297         VFS_ASSERT_GIANT(vp->v_mount);
  298 
  299         vn_start_write(vp, &mp, V_WAIT);
  300         vn_lock(vp, lock_flags | LK_RETRY);
  301         if (flags & FWRITE) {
  302                 VNASSERT(vp->v_writecount > 0, vp, 
  303                     ("vn_close: negative writecount"));
  304                 vp->v_writecount--;
  305         }
  306         error = VOP_CLOSE(vp, flags, file_cred, td);
  307         vput(vp);
  308         vn_finished_write(mp);
  309         return (error);
  310 }
  311 
  312 /*
  313  * Heuristic to detect sequential operation.
  314  */
  315 static int
  316 sequential_heuristic(struct uio *uio, struct file *fp)
  317 {
  318 
  319         if (atomic_load_acq_int(&(fp->f_flag)) & FRDAHEAD)
  320                 return (fp->f_seqcount << IO_SEQSHIFT);
  321 
  322         /*
  323          * Offset 0 is handled specially.  open() sets f_seqcount to 1 so
  324          * that the first I/O is normally considered to be slightly
  325          * sequential.  Seeking to offset 0 doesn't change sequentiality
  326          * unless previous seeks have reduced f_seqcount to 0, in which
  327          * case offset 0 is not special.
  328          */
  329         if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
  330             uio->uio_offset == fp->f_nextoff) {
  331                 /*
  332                  * f_seqcount is in units of fixed-size blocks so that it
  333                  * depends mainly on the amount of sequential I/O and not
  334                  * much on the number of sequential I/O's.  The fixed size
  335                  * of 16384 is hard-coded here since it is (not quite) just
  336                  * a magic size that works well here.  This size is more
  337                  * closely related to the best I/O size for real disks than
  338                  * to any block size used by software.
  339                  */
  340                 fp->f_seqcount += howmany(uio->uio_resid, 16384);
  341                 if (fp->f_seqcount > IO_SEQMAX)
  342                         fp->f_seqcount = IO_SEQMAX;
  343                 return (fp->f_seqcount << IO_SEQSHIFT);
  344         }
  345 
  346         /* Not sequential.  Quickly draw-down sequentiality. */
  347         if (fp->f_seqcount > 1)
  348                 fp->f_seqcount = 1;
  349         else
  350                 fp->f_seqcount = 0;
  351         return (0);
  352 }
  353 
  354 /*
  355  * Package up an I/O request on a vnode into a uio and do it.
  356  */
  357 int
  358 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, active_cred, file_cred,
  359     aresid, td)
  360         enum uio_rw rw;
  361         struct vnode *vp;
  362         void *base;
  363         int len;
  364         off_t offset;
  365         enum uio_seg segflg;
  366         int ioflg;
  367         struct ucred *active_cred;
  368         struct ucred *file_cred;
  369         int *aresid;
  370         struct thread *td;
  371 {
  372         struct uio auio;
  373         struct iovec aiov;
  374         struct mount *mp;
  375         struct ucred *cred;
  376         int error, lock_flags;
  377 
  378         VFS_ASSERT_GIANT(vp->v_mount);
  379 
  380         if ((ioflg & IO_NODELOCKED) == 0) {
  381                 mp = NULL;
  382                 if (rw == UIO_WRITE) { 
  383                         if (vp->v_type != VCHR &&
  384                             (error = vn_start_write(vp, &mp, V_WAIT | PCATCH))
  385                             != 0)
  386                                 return (error);
  387                         if (MNT_SHARED_WRITES(mp) ||
  388                             ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
  389                                 lock_flags = LK_SHARED;
  390                         } else {
  391                                 lock_flags = LK_EXCLUSIVE;
  392                         }
  393                         vn_lock(vp, lock_flags | LK_RETRY);
  394                 } else
  395                         vn_lock(vp, LK_SHARED | LK_RETRY);
  396 
  397         }
  398         ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
  399         auio.uio_iov = &aiov;
  400         auio.uio_iovcnt = 1;
  401         aiov.iov_base = base;
  402         aiov.iov_len = len;
  403         auio.uio_resid = len;
  404         auio.uio_offset = offset;
  405         auio.uio_segflg = segflg;
  406         auio.uio_rw = rw;
  407         auio.uio_td = td;
  408         error = 0;
  409 #ifdef MAC
  410         if ((ioflg & IO_NOMACCHECK) == 0) {
  411                 if (rw == UIO_READ)
  412                         error = mac_vnode_check_read(active_cred, file_cred,
  413                             vp);
  414                 else
  415                         error = mac_vnode_check_write(active_cred, file_cred,
  416                             vp);
  417         }
  418 #endif
  419         if (error == 0) {
  420                 if (file_cred)
  421                         cred = file_cred;
  422                 else
  423                         cred = active_cred;
  424                 if (rw == UIO_READ)
  425                         error = VOP_READ(vp, &auio, ioflg, cred);
  426                 else
  427                         error = VOP_WRITE(vp, &auio, ioflg, cred);
  428         }
  429         if (aresid)
  430                 *aresid = auio.uio_resid;
  431         else
  432                 if (auio.uio_resid && error == 0)
  433                         error = EIO;
  434         if ((ioflg & IO_NODELOCKED) == 0) {
  435                 if (rw == UIO_WRITE && vp->v_type != VCHR)
  436                         vn_finished_write(mp);
  437                 VOP_UNLOCK(vp, 0);
  438         }
  439         return (error);
  440 }
  441 
  442 /*
  443  * Package up an I/O request on a vnode into a uio and do it.  The I/O
  444  * request is split up into smaller chunks and we try to avoid saturating
  445  * the buffer cache while potentially holding a vnode locked, so we 
  446  * check bwillwrite() before calling vn_rdwr().  We also call uio_yield()
  447  * to give other processes a chance to lock the vnode (either other processes
  448  * core'ing the same binary, or unrelated processes scanning the directory).
  449  */
  450 int
  451 vn_rdwr_inchunks(rw, vp, base, len, offset, segflg, ioflg, active_cred,
  452     file_cred, aresid, td)
  453         enum uio_rw rw;
  454         struct vnode *vp;
  455         void *base;
  456         size_t len;
  457         off_t offset;
  458         enum uio_seg segflg;
  459         int ioflg;
  460         struct ucred *active_cred;
  461         struct ucred *file_cred;
  462         size_t *aresid;
  463         struct thread *td;
  464 {
  465         int error = 0;
  466         int iaresid;
  467 
  468         VFS_ASSERT_GIANT(vp->v_mount);
  469 
  470         do {
  471                 int chunk;
  472 
  473                 /*
  474                  * Force `offset' to a multiple of MAXBSIZE except possibly
  475                  * for the first chunk, so that filesystems only need to
  476                  * write full blocks except possibly for the first and last
  477                  * chunks.
  478                  */
  479                 chunk = MAXBSIZE - (uoff_t)offset % MAXBSIZE;
  480 
  481                 if (chunk > len)
  482                         chunk = len;
  483                 if (rw != UIO_READ && vp->v_type == VREG)
  484                         bwillwrite();
  485                 iaresid = 0;
  486                 error = vn_rdwr(rw, vp, base, chunk, offset, segflg,
  487                     ioflg, active_cred, file_cred, &iaresid, td);
  488                 len -= chunk;   /* aresid calc already includes length */
  489                 if (error)
  490                         break;
  491                 offset += chunk;
  492                 base = (char *)base + chunk;
  493                 uio_yield();
  494         } while (len);
  495         if (aresid)
  496                 *aresid = len + iaresid;
  497         return (error);
  498 }
  499 
  500 /*
  501  * File table vnode read routine.
  502  */
  503 static int
  504 vn_read(fp, uio, active_cred, flags, td)
  505         struct file *fp;
  506         struct uio *uio;
  507         struct ucred *active_cred;
  508         struct thread *td;
  509         int flags;
  510 {
  511         struct vnode *vp;
  512         int error, ioflag;
  513         struct mtx *mtxp;
  514         int advice, vfslocked;
  515 
  516         KASSERT(uio->uio_td == td, ("uio_td %p is not td %p",
  517             uio->uio_td, td));
  518         mtxp = NULL;
  519         vp = fp->f_vnode;
  520         ioflag = 0;
  521         if (fp->f_flag & FNONBLOCK)
  522                 ioflag |= IO_NDELAY;
  523         if (fp->f_flag & O_DIRECT)
  524                 ioflag |= IO_DIRECT;
  525         advice = POSIX_FADV_NORMAL;
  526         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
  527         /*
  528          * According to McKusick the vn lock was protecting f_offset here.
  529          * It is now protected by the FOFFSET_LOCKED flag.
  530          */
  531         if ((flags & FOF_OFFSET) == 0 || fp->f_advice != NULL) {
  532                 mtxp = mtx_pool_find(mtxpool_sleep, fp);
  533                 mtx_lock(mtxp);
  534                 if ((flags & FOF_OFFSET) == 0) {
  535                         while (fp->f_vnread_flags & FOFFSET_LOCKED) {
  536                                 fp->f_vnread_flags |= FOFFSET_LOCK_WAITING;
  537                                 msleep(&fp->f_vnread_flags, mtxp, PUSER -1,
  538                                     "vnread offlock", 0);
  539                         }
  540                         fp->f_vnread_flags |= FOFFSET_LOCKED;
  541                         uio->uio_offset = fp->f_offset;
  542                 }
  543                 if (fp->f_advice != NULL &&
  544                     uio->uio_offset >= fp->f_advice->fa_start &&
  545                     uio->uio_offset + uio->uio_resid <= fp->f_advice->fa_end)
  546                         advice = fp->f_advice->fa_advice;
  547                 mtx_unlock(mtxp);
  548         }
  549         vn_lock(vp, LK_SHARED | LK_RETRY);
  550 
  551         switch (advice) {
  552         case POSIX_FADV_NORMAL:
  553         case POSIX_FADV_SEQUENTIAL:
  554                 ioflag |= sequential_heuristic(uio, fp);
  555                 break;
  556         case POSIX_FADV_RANDOM:
  557                 /* Disable read-ahead for random I/O. */
  558                 break;
  559         case POSIX_FADV_NOREUSE:
  560                 /*
  561                  * Request the underlying FS to discard the buffers
  562                  * and pages after the I/O is complete.
  563                  */
  564                 ioflag |= IO_DIRECT;
  565                 break;
  566         }
  567 
  568 #ifdef MAC
  569         error = mac_vnode_check_read(active_cred, fp->f_cred, vp);
  570         if (error == 0)
  571 #endif
  572                 error = VOP_READ(vp, uio, ioflag, fp->f_cred);
  573         if ((flags & FOF_OFFSET) == 0) {
  574                 fp->f_offset = uio->uio_offset;
  575                 mtx_lock(mtxp);
  576                 if (fp->f_vnread_flags & FOFFSET_LOCK_WAITING)
  577                         wakeup(&fp->f_vnread_flags);
  578                 fp->f_vnread_flags = 0;
  579                 mtx_unlock(mtxp);
  580         }
  581         fp->f_nextoff = uio->uio_offset;
  582         VOP_UNLOCK(vp, 0);
  583         VFS_UNLOCK_GIANT(vfslocked);
  584         return (error);
  585 }
  586 
  587 /*
  588  * File table vnode write routine.
  589  */
  590 static int
  591 vn_write(fp, uio, active_cred, flags, td)
  592         struct file *fp;
  593         struct uio *uio;
  594         struct ucred *active_cred;
  595         struct thread *td;
  596         int flags;
  597 {
  598         struct vnode *vp;
  599         struct mount *mp;
  600         int error, ioflag, lock_flags;
  601         struct mtx *mtxp;
  602         int advice, vfslocked;
  603 
  604         KASSERT(uio->uio_td == td, ("uio_td %p is not td %p",
  605             uio->uio_td, td));
  606         vp = fp->f_vnode;
  607         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
  608         if (vp->v_type == VREG)
  609                 bwillwrite();
  610         ioflag = IO_UNIT;
  611         if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
  612                 ioflag |= IO_APPEND;
  613         if (fp->f_flag & FNONBLOCK)
  614                 ioflag |= IO_NDELAY;
  615         if (fp->f_flag & O_DIRECT)
  616                 ioflag |= IO_DIRECT;
  617         if ((fp->f_flag & O_FSYNC) ||
  618             (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
  619                 ioflag |= IO_SYNC;
  620         mp = NULL;
  621         if (vp->v_type != VCHR &&
  622             (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
  623                 goto unlock;
  624  
  625         if ((MNT_SHARED_WRITES(mp) ||
  626             ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) &&
  627             (flags & FOF_OFFSET) != 0) {
  628                 lock_flags = LK_SHARED;
  629         } else {
  630                 lock_flags = LK_EXCLUSIVE;
  631         }
  632 
  633         vn_lock(vp, lock_flags | LK_RETRY);
  634         if ((flags & FOF_OFFSET) == 0)
  635                 uio->uio_offset = fp->f_offset;
  636         advice = POSIX_FADV_NORMAL;
  637         if (fp->f_advice != NULL) {
  638                 mtxp = mtx_pool_find(mtxpool_sleep, fp);
  639                 mtx_lock(mtxp);
  640                 if (fp->f_advice != NULL &&
  641                     uio->uio_offset >= fp->f_advice->fa_start &&
  642                     uio->uio_offset + uio->uio_resid <= fp->f_advice->fa_end)
  643                         advice = fp->f_advice->fa_advice;
  644                 mtx_unlock(mtxp);
  645         }
  646         switch (advice) {
  647         case POSIX_FADV_NORMAL:
  648         case POSIX_FADV_SEQUENTIAL:
  649                 ioflag |= sequential_heuristic(uio, fp);
  650                 break;
  651         case POSIX_FADV_RANDOM:
  652                 /* XXX: Is this correct? */
  653                 break;
  654         case POSIX_FADV_NOREUSE:
  655                 /*
  656                  * Request the underlying FS to discard the buffers
  657                  * and pages after the I/O is complete.
  658                  */
  659                 ioflag |= IO_DIRECT;
  660                 break;
  661         }
  662 
  663 #ifdef MAC
  664         error = mac_vnode_check_write(active_cred, fp->f_cred, vp);
  665         if (error == 0)
  666 #endif
  667                 error = VOP_WRITE(vp, uio, ioflag, fp->f_cred);
  668         if ((flags & FOF_OFFSET) == 0)
  669                 fp->f_offset = uio->uio_offset;
  670         fp->f_nextoff = uio->uio_offset;
  671         VOP_UNLOCK(vp, 0);
  672         if (vp->v_type != VCHR)
  673                 vn_finished_write(mp);
  674 unlock:
  675         VFS_UNLOCK_GIANT(vfslocked);
  676         return (error);
  677 }
  678 
  679 /*
  680  * File table truncate routine.
  681  */
  682 static int
  683 vn_truncate(fp, length, active_cred, td)
  684         struct file *fp;
  685         off_t length;
  686         struct ucred *active_cred;
  687         struct thread *td;
  688 {
  689         struct vattr vattr;
  690         struct mount *mp;
  691         struct vnode *vp;
  692         int vfslocked;
  693         int error;
  694 
  695         vp = fp->f_vnode;
  696         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
  697         error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
  698         if (error) {
  699                 VFS_UNLOCK_GIANT(vfslocked);
  700                 return (error);
  701         }
  702         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  703         if (vp->v_type == VDIR) {
  704                 error = EISDIR;
  705                 goto out;
  706         }
  707 #ifdef MAC
  708         error = mac_vnode_check_write(active_cred, fp->f_cred, vp);
  709         if (error)
  710                 goto out;
  711 #endif
  712         error = vn_writechk(vp);
  713         if (error == 0) {
  714                 VATTR_NULL(&vattr);
  715                 vattr.va_size = length;
  716                 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
  717         }
  718 out:
  719         VOP_UNLOCK(vp, 0);
  720         vn_finished_write(mp);
  721         VFS_UNLOCK_GIANT(vfslocked);
  722         return (error);
  723 }
  724 
  725 /*
  726  * File table vnode stat routine.
  727  */
  728 static int
  729 vn_statfile(fp, sb, active_cred, td)
  730         struct file *fp;
  731         struct stat *sb;
  732         struct ucred *active_cred;
  733         struct thread *td;
  734 {
  735         struct vnode *vp = fp->f_vnode;
  736         int vfslocked;
  737         int error;
  738 
  739         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
  740         vn_lock(vp, LK_SHARED | LK_RETRY);
  741         error = vn_stat(vp, sb, active_cred, fp->f_cred, td);
  742         VOP_UNLOCK(vp, 0);
  743         VFS_UNLOCK_GIANT(vfslocked);
  744 
  745         return (error);
  746 }
  747 
  748 /*
  749  * Stat a vnode; implementation for the stat syscall
  750  */
  751 int
  752 vn_stat(vp, sb, active_cred, file_cred, td)
  753         struct vnode *vp;
  754         register struct stat *sb;
  755         struct ucred *active_cred;
  756         struct ucred *file_cred;
  757         struct thread *td;
  758 {
  759         struct vattr vattr;
  760         register struct vattr *vap;
  761         int error;
  762         u_short mode;
  763 
  764 #ifdef MAC
  765         error = mac_vnode_check_stat(active_cred, file_cred, vp);
  766         if (error)
  767                 return (error);
  768 #endif
  769 
  770         vap = &vattr;
  771 
  772         /*
  773          * Initialize defaults for new and unusual fields, so that file
  774          * systems which don't support these fields don't need to know
  775          * about them.
  776          */
  777         vap->va_birthtime.tv_sec = -1;
  778         vap->va_birthtime.tv_nsec = 0;
  779         vap->va_fsid = VNOVAL;
  780         vap->va_rdev = NODEV;
  781 
  782         error = VOP_GETATTR(vp, vap, active_cred);
  783         if (error)
  784                 return (error);
  785 
  786         /*
  787          * Zero the spare stat fields
  788          */
  789         bzero(sb, sizeof *sb);
  790 
  791         /*
  792          * Copy from vattr table
  793          */
  794         if (vap->va_fsid != VNOVAL)
  795                 sb->st_dev = vap->va_fsid;
  796         else
  797                 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
  798         sb->st_ino = vap->va_fileid;
  799         mode = vap->va_mode;
  800         switch (vap->va_type) {
  801         case VREG:
  802                 mode |= S_IFREG;
  803                 break;
  804         case VDIR:
  805                 mode |= S_IFDIR;
  806                 break;
  807         case VBLK:
  808                 mode |= S_IFBLK;
  809                 break;
  810         case VCHR:
  811                 mode |= S_IFCHR;
  812                 break;
  813         case VLNK:
  814                 mode |= S_IFLNK;
  815                 break;
  816         case VSOCK:
  817                 mode |= S_IFSOCK;
  818                 break;
  819         case VFIFO:
  820                 mode |= S_IFIFO;
  821                 break;
  822         default:
  823                 return (EBADF);
  824         };
  825         sb->st_mode = mode;
  826         sb->st_nlink = vap->va_nlink;
  827         sb->st_uid = vap->va_uid;
  828         sb->st_gid = vap->va_gid;
  829         sb->st_rdev = vap->va_rdev;
  830         if (vap->va_size > OFF_MAX)
  831                 return (EOVERFLOW);
  832         sb->st_size = vap->va_size;
  833         sb->st_atimespec = vap->va_atime;
  834         sb->st_mtimespec = vap->va_mtime;
  835         sb->st_ctimespec = vap->va_ctime;
  836         sb->st_birthtimespec = vap->va_birthtime;
  837 
  838         /*
  839          * According to www.opengroup.org, the meaning of st_blksize is 
  840          *   "a filesystem-specific preferred I/O block size for this 
  841          *    object.  In some filesystem types, this may vary from file
  842          *    to file"
  843          * Use miminum/default of PAGE_SIZE (e.g. for VCHR).
  844          */
  845 
  846         sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
  847         
  848         sb->st_flags = vap->va_flags;
  849         if (priv_check(td, PRIV_VFS_GENERATION))
  850                 sb->st_gen = 0;
  851         else
  852                 sb->st_gen = vap->va_gen;
  853 
  854         sb->st_blocks = vap->va_bytes / S_BLKSIZE;
  855         return (0);
  856 }
  857 
  858 /*
  859  * File table vnode ioctl routine.
  860  */
  861 static int
  862 vn_ioctl(fp, com, data, active_cred, td)
  863         struct file *fp;
  864         u_long com;
  865         void *data;
  866         struct ucred *active_cred;
  867         struct thread *td;
  868 {
  869         struct vnode *vp = fp->f_vnode;
  870         struct vattr vattr;
  871         int vfslocked;
  872         int error;
  873 
  874         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
  875         error = ENOTTY;
  876         switch (vp->v_type) {
  877         case VREG:
  878         case VDIR:
  879                 if (com == FIONREAD) {
  880                         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  881                         error = VOP_GETATTR(vp, &vattr, active_cred);
  882                         VOP_UNLOCK(vp, 0);
  883                         if (!error)
  884                                 *(int *)data = vattr.va_size - fp->f_offset;
  885                 }
  886                 if (com == FIONBIO || com == FIOASYNC)  /* XXX */
  887                         error = 0;
  888                 else
  889                         error = VOP_IOCTL(vp, com, data, fp->f_flag,
  890                             active_cred, td);
  891                 break;
  892 
  893         default:
  894                 break;
  895         }
  896         VFS_UNLOCK_GIANT(vfslocked);
  897         return (error);
  898 }
  899 
  900 /*
  901  * File table vnode poll routine.
  902  */
  903 static int
  904 vn_poll(fp, events, active_cred, td)
  905         struct file *fp;
  906         int events;
  907         struct ucred *active_cred;
  908         struct thread *td;
  909 {
  910         struct vnode *vp;
  911         int vfslocked;
  912         int error;
  913 
  914         vp = fp->f_vnode;
  915         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
  916 #ifdef MAC
  917         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  918         error = mac_vnode_check_poll(active_cred, fp->f_cred, vp);
  919         VOP_UNLOCK(vp, 0);
  920         if (!error)
  921 #endif
  922 
  923         error = VOP_POLL(vp, events, fp->f_cred, td);
  924         VFS_UNLOCK_GIANT(vfslocked);
  925         return (error);
  926 }
  927 
  928 /*
  929  * Acquire the requested lock and then check for validity.  LK_RETRY
  930  * permits vn_lock to return doomed vnodes.
  931  */
  932 int
  933 _vn_lock(struct vnode *vp, int flags, char *file, int line)
  934 {
  935         int error;
  936 
  937         VNASSERT((flags & LK_TYPE_MASK) != 0, vp,
  938             ("vn_lock called with no locktype."));
  939         do {
  940 #ifdef DEBUG_VFS_LOCKS
  941                 KASSERT(vp->v_holdcnt != 0,
  942                     ("vn_lock %p: zero hold count", vp));
  943 #endif
  944                 error = VOP_LOCK1(vp, flags, file, line);
  945                 flags &= ~LK_INTERLOCK; /* Interlock is always dropped. */
  946                 KASSERT((flags & LK_RETRY) == 0 || error == 0,
  947                     ("LK_RETRY set with incompatible flags (0x%x) or an error occured (%d)",
  948                     flags, error));
  949                 /*
  950                  * Callers specify LK_RETRY if they wish to get dead vnodes.
  951                  * If RETRY is not set, we return ENOENT instead.
  952                  */
  953                 if (error == 0 && vp->v_iflag & VI_DOOMED &&
  954                     (flags & LK_RETRY) == 0) {
  955                         VOP_UNLOCK(vp, 0);
  956                         error = ENOENT;
  957                         break;
  958                 }
  959         } while (flags & LK_RETRY && error != 0);
  960         return (error);
  961 }
  962 
  963 /*
  964  * File table vnode close routine.
  965  */
  966 static int
  967 vn_closefile(fp, td)
  968         struct file *fp;
  969         struct thread *td;
  970 {
  971         struct vnode *vp;
  972         struct flock lf;
  973         int vfslocked;
  974         int error;
  975 
  976         vp = fp->f_vnode;
  977 
  978         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
  979         if (fp->f_type == DTYPE_VNODE && fp->f_flag & FHASLOCK) {
  980                 lf.l_whence = SEEK_SET;
  981                 lf.l_start = 0;
  982                 lf.l_len = 0;
  983                 lf.l_type = F_UNLCK;
  984                 (void) VOP_ADVLOCK(vp, fp, F_UNLCK, &lf, F_FLOCK);
  985         }
  986 
  987         fp->f_ops = &badfileops;
  988 
  989         error = vn_close(vp, fp->f_flag, fp->f_cred, td);
  990         VFS_UNLOCK_GIANT(vfslocked);
  991         return (error);
  992 }
  993 
  994 /*
  995  * Preparing to start a filesystem write operation. If the operation is
  996  * permitted, then we bump the count of operations in progress and
  997  * proceed. If a suspend request is in progress, we wait until the
  998  * suspension is over, and then proceed.
  999  */
 1000 int
 1001 vn_start_write(vp, mpp, flags)
 1002         struct vnode *vp;
 1003         struct mount **mpp;
 1004         int flags;
 1005 {
 1006         struct mount *mp;
 1007         int error;
 1008 
 1009         error = 0;
 1010         /*
 1011          * If a vnode is provided, get and return the mount point that
 1012          * to which it will write.
 1013          */
 1014         if (vp != NULL) {
 1015                 if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) {
 1016                         *mpp = NULL;
 1017                         if (error != EOPNOTSUPP)
 1018                                 return (error);
 1019                         return (0);
 1020                 }
 1021         }
 1022         if ((mp = *mpp) == NULL)
 1023                 return (0);
 1024 
 1025         /*
 1026          * VOP_GETWRITEMOUNT() returns with the mp refcount held through
 1027          * a vfs_ref().
 1028          * As long as a vnode is not provided we need to acquire a
 1029          * refcount for the provided mountpoint too, in order to
 1030          * emulate a vfs_ref().
 1031          */
 1032         MNT_ILOCK(mp);
 1033         if (vp == NULL)
 1034                 MNT_REF(mp);
 1035 
 1036         /*
 1037          * Check on status of suspension.
 1038          */
 1039         if ((curthread->td_pflags & TDP_IGNSUSP) == 0 ||
 1040             mp->mnt_susp_owner != curthread) {
 1041                 while ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
 1042                         if (flags & V_NOWAIT) {
 1043                                 error = EWOULDBLOCK;
 1044                                 goto unlock;
 1045                         }
 1046                         error = msleep(&mp->mnt_flag, MNT_MTX(mp),
 1047                             (PUSER - 1) | (flags & PCATCH), "suspfs", 0);
 1048                         if (error)
 1049                                 goto unlock;
 1050                 }
 1051         }
 1052         if (flags & V_XSLEEP)
 1053                 goto unlock;
 1054         mp->mnt_writeopcount++;
 1055 unlock:
 1056         if (error != 0 || (flags & V_XSLEEP) != 0)
 1057                 MNT_REL(mp);
 1058         MNT_IUNLOCK(mp);
 1059         return (error);
 1060 }
 1061 
 1062 /*
 1063  * Secondary suspension. Used by operations such as vop_inactive
 1064  * routines that are needed by the higher level functions. These
 1065  * are allowed to proceed until all the higher level functions have
 1066  * completed (indicated by mnt_writeopcount dropping to zero). At that
 1067  * time, these operations are halted until the suspension is over.
 1068  */
 1069 int
 1070 vn_start_secondary_write(vp, mpp, flags)
 1071         struct vnode *vp;
 1072         struct mount **mpp;
 1073         int flags;
 1074 {
 1075         struct mount *mp;
 1076         int error;
 1077 
 1078  retry:
 1079         if (vp != NULL) {
 1080                 if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) {
 1081                         *mpp = NULL;
 1082                         if (error != EOPNOTSUPP)
 1083                                 return (error);
 1084                         return (0);
 1085                 }
 1086         }
 1087         /*
 1088          * If we are not suspended or have not yet reached suspended
 1089          * mode, then let the operation proceed.
 1090          */
 1091         if ((mp = *mpp) == NULL)
 1092                 return (0);
 1093 
 1094         /*
 1095          * VOP_GETWRITEMOUNT() returns with the mp refcount held through
 1096          * a vfs_ref().
 1097          * As long as a vnode is not provided we need to acquire a
 1098          * refcount for the provided mountpoint too, in order to
 1099          * emulate a vfs_ref().
 1100          */
 1101         MNT_ILOCK(mp);
 1102         if (vp == NULL)
 1103                 MNT_REF(mp);
 1104         if ((mp->mnt_kern_flag & (MNTK_SUSPENDED | MNTK_SUSPEND2)) == 0) {
 1105                 mp->mnt_secondary_writes++;
 1106                 mp->mnt_secondary_accwrites++;
 1107                 MNT_IUNLOCK(mp);
 1108                 return (0);
 1109         }
 1110         if (flags & V_NOWAIT) {
 1111                 MNT_REL(mp);
 1112                 MNT_IUNLOCK(mp);
 1113                 return (EWOULDBLOCK);
 1114         }
 1115         /*
 1116          * Wait for the suspension to finish.
 1117          */
 1118         error = msleep(&mp->mnt_flag, MNT_MTX(mp),
 1119                        (PUSER - 1) | (flags & PCATCH) | PDROP, "suspfs", 0);
 1120         vfs_rel(mp);
 1121         if (error == 0)
 1122                 goto retry;
 1123         return (error);
 1124 }
 1125 
 1126 /*
 1127  * Filesystem write operation has completed. If we are suspending and this
 1128  * operation is the last one, notify the suspender that the suspension is
 1129  * now in effect.
 1130  */
 1131 void
 1132 vn_finished_write(mp)
 1133         struct mount *mp;
 1134 {
 1135         if (mp == NULL)
 1136                 return;
 1137         MNT_ILOCK(mp);
 1138         MNT_REL(mp);
 1139         mp->mnt_writeopcount--;
 1140         if (mp->mnt_writeopcount < 0)
 1141                 panic("vn_finished_write: neg cnt");
 1142         if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 &&
 1143             mp->mnt_writeopcount <= 0)
 1144                 wakeup(&mp->mnt_writeopcount);
 1145         MNT_IUNLOCK(mp);
 1146 }
 1147 
 1148 
 1149 /*
 1150  * Filesystem secondary write operation has completed. If we are
 1151  * suspending and this operation is the last one, notify the suspender
 1152  * that the suspension is now in effect.
 1153  */
 1154 void
 1155 vn_finished_secondary_write(mp)
 1156         struct mount *mp;
 1157 {
 1158         if (mp == NULL)
 1159                 return;
 1160         MNT_ILOCK(mp);
 1161         MNT_REL(mp);
 1162         mp->mnt_secondary_writes--;
 1163         if (mp->mnt_secondary_writes < 0)
 1164                 panic("vn_finished_secondary_write: neg cnt");
 1165         if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 &&
 1166             mp->mnt_secondary_writes <= 0)
 1167                 wakeup(&mp->mnt_secondary_writes);
 1168         MNT_IUNLOCK(mp);
 1169 }
 1170 
 1171 
 1172 
 1173 /*
 1174  * Request a filesystem to suspend write operations.
 1175  */
 1176 int
 1177 vfs_write_suspend(mp)
 1178         struct mount *mp;
 1179 {
 1180         int error;
 1181 
 1182         MNT_ILOCK(mp);
 1183         if (mp->mnt_susp_owner == curthread) {
 1184                 MNT_IUNLOCK(mp);
 1185                 return (EALREADY);
 1186         }
 1187         while (mp->mnt_kern_flag & MNTK_SUSPEND)
 1188                 msleep(&mp->mnt_flag, MNT_MTX(mp), PUSER - 1, "wsuspfs", 0);
 1189         mp->mnt_kern_flag |= MNTK_SUSPEND;
 1190         mp->mnt_susp_owner = curthread;
 1191         if (mp->mnt_writeopcount > 0)
 1192                 (void) msleep(&mp->mnt_writeopcount, 
 1193                     MNT_MTX(mp), (PUSER - 1)|PDROP, "suspwt", 0);
 1194         else
 1195                 MNT_IUNLOCK(mp);
 1196         if ((error = VFS_SYNC(mp, MNT_SUSPEND)) != 0)
 1197                 vfs_write_resume(mp);
 1198         return (error);
 1199 }
 1200 
 1201 /*
 1202  * Request a filesystem to resume write operations.
 1203  */
 1204 void
 1205 vfs_write_resume(mp)
 1206         struct mount *mp;
 1207 {
 1208 
 1209         MNT_ILOCK(mp);
 1210         if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
 1211                 KASSERT(mp->mnt_susp_owner == curthread, ("mnt_susp_owner"));
 1212                 mp->mnt_kern_flag &= ~(MNTK_SUSPEND | MNTK_SUSPEND2 |
 1213                                        MNTK_SUSPENDED);
 1214                 mp->mnt_susp_owner = NULL;
 1215                 wakeup(&mp->mnt_writeopcount);
 1216                 wakeup(&mp->mnt_flag);
 1217                 curthread->td_pflags &= ~TDP_IGNSUSP;
 1218                 MNT_IUNLOCK(mp);
 1219                 VFS_SUSP_CLEAN(mp);
 1220         } else
 1221                 MNT_IUNLOCK(mp);
 1222 }
 1223 
 1224 /*
 1225  * Implement kqueues for files by translating it to vnode operation.
 1226  */
 1227 static int
 1228 vn_kqfilter(struct file *fp, struct knote *kn)
 1229 {
 1230         int vfslocked;
 1231         int error;
 1232 
 1233         vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
 1234         error = VOP_KQFILTER(fp->f_vnode, kn);
 1235         VFS_UNLOCK_GIANT(vfslocked);
 1236 
 1237         return error;
 1238 }
 1239 
 1240 /*
 1241  * Simplified in-kernel wrapper calls for extended attribute access.
 1242  * Both calls pass in a NULL credential, authorizing as "kernel" access.
 1243  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
 1244  */
 1245 int
 1246 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
 1247     const char *attrname, int *buflen, char *buf, struct thread *td)
 1248 {
 1249         struct uio      auio;
 1250         struct iovec    iov;
 1251         int     error;
 1252 
 1253         iov.iov_len = *buflen;
 1254         iov.iov_base = buf;
 1255 
 1256         auio.uio_iov = &iov;
 1257         auio.uio_iovcnt = 1;
 1258         auio.uio_rw = UIO_READ;
 1259         auio.uio_segflg = UIO_SYSSPACE;
 1260         auio.uio_td = td;
 1261         auio.uio_offset = 0;
 1262         auio.uio_resid = *buflen;
 1263 
 1264         if ((ioflg & IO_NODELOCKED) == 0)
 1265                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 1266 
 1267         ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
 1268 
 1269         /* authorize attribute retrieval as kernel */
 1270         error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL,
 1271             td);
 1272 
 1273         if ((ioflg & IO_NODELOCKED) == 0)
 1274                 VOP_UNLOCK(vp, 0);
 1275 
 1276         if (error == 0) {
 1277                 *buflen = *buflen - auio.uio_resid;
 1278         }
 1279 
 1280         return (error);
 1281 }
 1282 
 1283 /*
 1284  * XXX failure mode if partially written?
 1285  */
 1286 int
 1287 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
 1288     const char *attrname, int buflen, char *buf, struct thread *td)
 1289 {
 1290         struct uio      auio;
 1291         struct iovec    iov;
 1292         struct mount    *mp;
 1293         int     error;
 1294 
 1295         iov.iov_len = buflen;
 1296         iov.iov_base = buf;
 1297 
 1298         auio.uio_iov = &iov;
 1299         auio.uio_iovcnt = 1;
 1300         auio.uio_rw = UIO_WRITE;
 1301         auio.uio_segflg = UIO_SYSSPACE;
 1302         auio.uio_td = td;
 1303         auio.uio_offset = 0;
 1304         auio.uio_resid = buflen;
 1305 
 1306         if ((ioflg & IO_NODELOCKED) == 0) {
 1307                 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
 1308                         return (error);
 1309                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 1310         }
 1311 
 1312         ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
 1313 
 1314         /* authorize attribute setting as kernel */
 1315         error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, td);
 1316 
 1317         if ((ioflg & IO_NODELOCKED) == 0) {
 1318                 vn_finished_write(mp);
 1319                 VOP_UNLOCK(vp, 0);
 1320         }
 1321 
 1322         return (error);
 1323 }
 1324 
 1325 int
 1326 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
 1327     const char *attrname, struct thread *td)
 1328 {
 1329         struct mount    *mp;
 1330         int     error;
 1331 
 1332         if ((ioflg & IO_NODELOCKED) == 0) {
 1333                 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
 1334                         return (error);
 1335                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 1336         }
 1337 
 1338         ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held");
 1339 
 1340         /* authorize attribute removal as kernel */
 1341         error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, td);
 1342         if (error == EOPNOTSUPP)
 1343                 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
 1344                     NULL, td);
 1345 
 1346         if ((ioflg & IO_NODELOCKED) == 0) {
 1347                 vn_finished_write(mp);
 1348                 VOP_UNLOCK(vp, 0);
 1349         }
 1350 
 1351         return (error);
 1352 }
 1353 
 1354 int
 1355 vn_vget_ino(struct vnode *vp, ino_t ino, int lkflags, struct vnode **rvp)
 1356 {
 1357         struct mount *mp;
 1358         int ltype, error;
 1359 
 1360         mp = vp->v_mount;
 1361         ltype = VOP_ISLOCKED(vp);
 1362         KASSERT(ltype == LK_EXCLUSIVE || ltype == LK_SHARED,
 1363             ("vn_vget_ino: vp not locked"));
 1364         error = vfs_busy(mp, MBF_NOWAIT);
 1365         if (error != 0) {
 1366                 vfs_ref(mp);
 1367                 VOP_UNLOCK(vp, 0);
 1368                 error = vfs_busy(mp, 0);
 1369                 vn_lock(vp, ltype | LK_RETRY);
 1370                 vfs_rel(mp);
 1371                 if (error != 0)
 1372                         return (ENOENT);
 1373                 if (vp->v_iflag & VI_DOOMED) {
 1374                         vfs_unbusy(mp);
 1375                         return (ENOENT);
 1376                 }
 1377         }
 1378         VOP_UNLOCK(vp, 0);
 1379         error = VFS_VGET(mp, ino, lkflags, rvp);
 1380         vfs_unbusy(mp);
 1381         vn_lock(vp, ltype | LK_RETRY);
 1382         if (vp->v_iflag & VI_DOOMED) {
 1383                 if (error == 0)
 1384                         vput(*rvp);
 1385                 error = ENOENT;
 1386         }
 1387         return (error);
 1388 }
 1389 
 1390 int
 1391 vn_rlimit_fsize(const struct vnode *vp, const struct uio *uio, const struct thread *td)
 1392 {
 1393         if (vp->v_type != VREG || td == NULL)
 1394                 return (0);
 1395 
 1396         PROC_LOCK(td->td_proc);
 1397         if (uio->uio_offset + uio->uio_resid >
 1398             lim_cur(td->td_proc, RLIMIT_FSIZE)) {
 1399                 psignal(td->td_proc, SIGXFSZ);
 1400                 PROC_UNLOCK(td->td_proc);
 1401                 return (EFBIG);
 1402         }
 1403         PROC_UNLOCK(td->td_proc);
 1404 
 1405         return (0);
 1406 }
 1407 
 1408 void
 1409 vn_pages_remove(struct vnode *vp, vm_pindex_t start, vm_pindex_t end)
 1410 {
 1411         vm_object_t object;
 1412 
 1413         if ((object = vp->v_object) == NULL)
 1414                 return;
 1415         VM_OBJECT_LOCK(object);
 1416         vm_object_page_remove(object, start, end, 0);
 1417         VM_OBJECT_UNLOCK(object);
 1418 }

Cache object: 3f5c4c5d14558d102e9c4b209f0da9f5


[ 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.