sys/vfs/fdesc/fdesc_vnops.c

   1 /*
   2  * Copyright (c) 1992, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * This code is derived from software donated to Berkeley by
   6  * Jan-Simon Pendry.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. All advertising materials mentioning features or use of this software
  17  *    must display the following acknowledgement:
  18  *      This product includes software developed by the University of
  19  *      California, Berkeley and its contributors.
  20  * 4. Neither the name of the University nor the names of its contributors
  21  *    may be used to endorse or promote products derived from this software
  22  *    without specific prior written permission.
  23  *
  24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  34  * SUCH DAMAGE.
  35  *
  36  *      @(#)fdesc_vnops.c       8.9 (Berkeley) 1/21/94
  37  *
  38  * $FreeBSD: src/sys/miscfs/fdesc/fdesc_vnops.c,v 1.47.2.1 2001/10/22 22:49:26 chris Exp $
  39  * $DragonFly: src/sys/vfs/fdesc/fdesc_vnops.c,v 1.20 2005/08/16 22:40:03 joerg Exp $
  40  */
  41
  42 /*
  43  * /dev/fd Filesystem
  44  */
  45
  46 #include <sys/param.h>
  47 #include <sys/systm.h>
  48 #include <sys/conf.h>
  49 #include <sys/dirent.h>
  50 #include <sys/filedesc.h>
  51 #include <sys/kernel.h> /* boottime */
  52 #include <sys/lock.h>
  53 #include <sys/malloc.h>
  54 #include <sys/file.h>   /* Must come after sys/malloc.h */
  55 #include <sys/mount.h>
  56 #include <sys/proc.h>
  57 #include <sys/namei.h>
  58 #include <sys/socket.h>
  59 #include <sys/stat.h>
  60 #include <sys/vnode.h>
  61 #include <sys/file2.h>
  62
  63 #include <machine/limits.h>
  64
  65 #include "fdesc.h"
  66
  67 #define FDL_WANT        0x01
  68 #define FDL_LOCKED      0x02
  69 static int fdcache_lock;
  70
  71 #define NFDCACHE 4
  72 #define FD_NHASH(ix) \
  73         (&fdhashtbl[(ix) & fdhash])
  74 static LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
  75 static u_long fdhash;
  76
  77 static int      fdesc_getattr (struct vop_getattr_args *ap);
  78 static int      fdesc_inactive (struct vop_inactive_args *ap);
  79 static int      fdesc_lookup (struct vop_lookup_args *ap);
  80 static int      fdesc_open (struct vop_open_args *ap);
  81 static int      fdesc_print (struct vop_print_args *ap);
  82 static int      fdesc_readdir (struct vop_readdir_args *ap);
  83 static int      fdesc_reclaim (struct vop_reclaim_args *ap);
  84 static int      fdesc_poll (struct vop_poll_args *ap);
  85 static int      fdesc_setattr (struct vop_setattr_args *ap);
  86
  87 /*
  88  * Initialise cache headers
  89  */
  90 int
  91 fdesc_init(struct vfsconf *vfsp)
  92 {
  93
  94         fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash);
  95         return (0);
  96 }
  97
  98 int
  99 fdesc_uninit(struct vfsconf *vfsp)
 100 {
 101         if (fdhashtbl)
 102                 free(fdhashtbl, M_CACHE);
 103         return (0);
 104 }
 105 int
 106 fdesc_allocvp(fdntype ftype, int ix, struct mount *mp, struct vnode **vpp,
 107               struct thread *td)
 108 {
 109         struct fdhashhead *fc;
 110         struct fdescnode *fd;
 111         int error = 0;
 112
 113         fc = FD_NHASH(ix);
 114 loop:
 115         LIST_FOREACH(fd, fc, fd_hash) {
 116                 if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
 117                         if (vget(fd->fd_vnode, LK_EXCLUSIVE|LK_SLEEPFAIL, td))
 118                                 goto loop;
 119                         *vpp = fd->fd_vnode;
 120                         return (error);
 121                 }
 122         }
 123
 124         /*
 125          * otherwise lock the array while we call getnewvnode
 126          * since that can block.
 127          */
 128         if (fdcache_lock & FDL_LOCKED) {
 129                 fdcache_lock |= FDL_WANT;
 130                 (void) tsleep((caddr_t) &fdcache_lock, 0, "fdalvp", 0);
 131                 goto loop;
 132         }
 133         fdcache_lock |= FDL_LOCKED;
 134
 135         /*
 136          * Do the MALLOC before the getnewvnode since doing so afterward
 137          * might cause a bogus v_data pointer to get dereferenced
 138          * elsewhere if MALLOC should block.
 139          */
 140         MALLOC(fd, struct fdescnode *, sizeof(struct fdescnode), M_TEMP, M_WAITOK);
 141
 142         error = getnewvnode(VT_FDESC, mp, vpp, 0, 0);
 143         if (error) {
 144                 FREE(fd, M_TEMP);
 145                 goto out;
 146         }
 147         (*vpp)->v_data = fd;
 148         fd->fd_vnode = *vpp;
 149         fd->fd_type = ftype;
 150         fd->fd_fd = -1;
 151         fd->fd_ix = ix;
 152         LIST_INSERT_HEAD(fc, fd, fd_hash);
 153         vx_unlock(*vpp);
 154
 155 out:
 156         fdcache_lock &= ~FDL_LOCKED;
 157
 158         if (fdcache_lock & FDL_WANT) {
 159                 fdcache_lock &= ~FDL_WANT;
 160                 wakeup((caddr_t) &fdcache_lock);
 161         }
 162
 163         return (error);
 164 }
 165
 166 /*
 167  * vp is the current namei directory
 168  * ndp is the name to locate in that directory...
 169  *
 170  * fdesc_lookup(struct vnode *a_dvp, struct vnode **a_vpp,
 171  *              struct componentname *a_cnp)
 172  */
 173 static int
 174 fdesc_lookup(struct vop_lookup_args *ap)
 175 {
 176         struct componentname *cnp = ap->a_cnp;
 177         struct thread *td = cnp->cn_td;
 178         struct proc *p = td->td_proc;
 179         struct vnode **vpp = ap->a_vpp;
 180         struct vnode *dvp = ap->a_dvp;
 181         char *pname = cnp->cn_nameptr;
 182         int nlen = cnp->cn_namelen;
 183         int nfiles;
 184         u_int fd;
 185         int error;
 186         struct vnode *fvp;
 187
 188         KKASSERT(p);
 189         nfiles = p->p_fd->fd_nfiles;
 190         if (cnp->cn_nameiop == NAMEI_DELETE || cnp->cn_nameiop == NAMEI_RENAME) {
 191                 error = EROFS;
 192                 goto bad;
 193         }
 194
 195         VOP_UNLOCK(dvp, 0, td);
 196         if (cnp->cn_namelen == 1 && *pname == '.') {
 197                 *vpp = dvp;
 198                 vref(dvp);
 199                 vn_lock(dvp, LK_SHARED | LK_RETRY, td);
 200                 return (0);
 201         }
 202
 203         if (VTOFDESC(dvp)->fd_type != Froot) {
 204                 error = ENOTDIR;
 205                 goto bad;
 206         }
 207
 208         fd = 0;
 209         /* the only time a leading 0 is acceptable is if it's "0" */
 210         if (*pname == '0' && nlen != 1) {
 211                 error = ENOENT;
 212                 goto bad;
 213         }
 214         while (nlen--) {
 215                 if (*pname < '0' || *pname > '9') {
 216                         error = ENOENT;
 217                         goto bad;
 218                 }
 219                 fd = 10 * fd + *pname++ - '0';
 220         }
 221
 222         if (fd >= nfiles || p->p_fd->fd_files[fd].fp == NULL) {
 223                 error = EBADF;
 224                 goto bad;
 225         }
 226
 227         error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp, td);
 228         if (error)
 229                 goto bad;
 230         VTOFDESC(fvp)->fd_fd = fd;
 231         vn_lock(fvp, LK_SHARED | LK_RETRY, td);
 232         *vpp = fvp;
 233         return (0);
 234
 235 bad:
 236         vn_lock(dvp, LK_SHARED | LK_RETRY, td);
 237         *vpp = NULL;
 238         return (error);
 239 }
 240
 241 /*
 242  * fdesc_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
 243  *            struct thread *a_td)
 244  */
 245 static int
 246 fdesc_open(struct vop_open_args *ap)
 247 {
 248         struct vnode *vp = ap->a_vp;
 249         struct proc *p = ap->a_td->td_proc;
 250
 251         KKASSERT(p);
 252
 253         if (VTOFDESC(vp)->fd_type == Froot)
 254                 return (0);
 255
 256         /*
 257          * XXX Kludge: set p->p_dupfd to contain the value of the the file
 258          * descriptor being sought for duplication. The error return ensures
 259          * that the vnode for this device will be released by vn_open. Open
 260          * will detect this special error and take the actions in dupfdopen.
 261          * Other callers of vn_open or VOP_OPEN will simply report the
 262          * error.
 263          */
 264         p->p_dupfd = VTOFDESC(vp)->fd_fd;       /* XXX */
 265         return (ENODEV);
 266 }
 267
 268 /*
 269  * fdesc_getattr(struct vnode *a_vp, struct vattr *a_vap,
 270  *               struct ucred *a_cred, struct thread *a_td)
 271  */
 272 static int
 273 fdesc_getattr(struct vop_getattr_args *ap)
 274 {
 275         struct proc *p = ap->a_td->td_proc;
 276         struct vnode *vp = ap->a_vp;
 277         struct vattr *vap = ap->a_vap;
 278         struct filedesc *fdp;
 279         struct file *fp;
 280         struct stat stb;
 281         u_int fd;
 282         int error = 0;
 283
 284         KKASSERT(p);
 285         fdp = p->p_fd;
 286         switch (VTOFDESC(vp)->fd_type) {
 287         case Froot:
 288                 VATTR_NULL(vap);
 289
 290                 vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
 291                 vap->va_type = VDIR;
 292                 vap->va_nlink = 2;
 293                 vap->va_size = DEV_BSIZE;
 294                 vap->va_fileid = VTOFDESC(vp)->fd_ix;
 295                 vap->va_uid = 0;
 296                 vap->va_gid = 0;
 297                 vap->va_blocksize = DEV_BSIZE;
 298                 vap->va_atime.tv_sec = boottime.tv_sec;
 299                 vap->va_atime.tv_nsec = 0;
 300                 vap->va_mtime = vap->va_atime;
 301                 vap->va_ctime = vap->va_mtime;
 302                 vap->va_gen = 0;
 303                 vap->va_flags = 0;
 304                 vap->va_rdev = 0;
 305                 vap->va_bytes = 0;
 306                 break;
 307
 308         case Fdesc:
 309                 fd = VTOFDESC(vp)->fd_fd;
 310
 311                 if (fd >= fdp->fd_nfiles || (fp = fdp->fd_files[fd].fp) == NULL)
 312                         return (EBADF);
 313
 314                 bzero(&stb, sizeof(stb));
 315                 error = fo_stat(fp, &stb, ap->a_td);
 316                 if (error == 0) {
 317                         VATTR_NULL(vap);
 318                         vap->va_type = IFTOVT(stb.st_mode);
 319                         vap->va_mode = stb.st_mode;
 320 #define FDRX (VREAD|VEXEC)
 321                         if (vap->va_type == VDIR)
 322                                 vap->va_mode &= ~((FDRX)|(FDRX>>3)|(FDRX>>6));
 323 #undef FDRX
 324                         vap->va_nlink = 1;
 325                         vap->va_flags = 0;
 326                         vap->va_bytes = stb.st_blocks * stb.st_blksize;
 327                         vap->va_fileid = VTOFDESC(vp)->fd_ix;
 328                         vap->va_size = stb.st_size;
 329                         vap->va_blocksize = stb.st_blksize;
 330                         vap->va_rdev = stb.st_rdev;
 331
 332                         /*
 333                          * If no time data is provided, use the current time.
 334                          */
 335                         if (stb.st_atimespec.tv_sec == 0 &&
 336                             stb.st_atimespec.tv_nsec == 0)
 337                                 nanotime(&stb.st_atimespec);
 338
 339                         if (stb.st_ctimespec.tv_sec == 0 &&
 340                             stb.st_ctimespec.tv_nsec == 0)
 341                                 nanotime(&stb.st_ctimespec);
 342
 343                         if (stb.st_mtimespec.tv_sec == 0 &&
 344                             stb.st_mtimespec.tv_nsec == 0)
 345                                 nanotime(&stb.st_mtimespec);
 346
 347                         vap->va_atime = stb.st_atimespec;
 348                         vap->va_mtime = stb.st_mtimespec;
 349                         vap->va_ctime = stb.st_ctimespec;
 350                         vap->va_uid = stb.st_uid;
 351                         vap->va_gid = stb.st_gid;
 352                 }
 353                 break;
 354
 355         default:
 356                 panic("fdesc_getattr");
 357                 break;
 358         }
 359
 360         if (error == 0)
 361                 vp->v_type = vap->va_type;
 362         return (error);
 363 }
 364
 365 /*
 366  * fdesc_setattr(struct vnode *a_vp, struct vattr *a_vap,
 367  *               struct ucred *a_cred, struct thread *a_td)
 368  */
 369 static int
 370 fdesc_setattr(struct vop_setattr_args *ap)
 371 {
 372         struct proc *p = ap->a_td->td_proc;
 373         struct vattr *vap = ap->a_vap;
 374         struct file *fp;
 375         unsigned fd;
 376         int error;
 377
 378         /*
 379          * Can't mess with the root vnode
 380          */
 381         if (VTOFDESC(ap->a_vp)->fd_type == Froot)
 382                 return (EACCES);
 383
 384         fd = VTOFDESC(ap->a_vp)->fd_fd;
 385         KKASSERT(p);
 386
 387         /*
 388          * Allow setattr where there is an underlying vnode.
 389          */
 390         error = getvnode(p->p_fd, fd, &fp);
 391         if (error) {
 392                 /*
 393                  * getvnode() returns EINVAL if the file descriptor is not
 394                  * backed by a vnode.  Silently drop all changes except
 395                  * chflags(2) in this case.
 396                  */
 397                 if (error == EINVAL) {
 398                         if (vap->va_flags != VNOVAL)
 399                                 error = EOPNOTSUPP;
 400                         else
 401                                 error = 0;
 402                 }
 403                 return (error);
 404         }
 405         return (error);
 406 }
 407
 408 #define UIO_MX 16
 409
 410 /*
 411  * fdesc_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred,
 412  *               int *a_eofflag, u_long *a_cookies, int a_ncookies)
 413  */
 414 static int
 415 fdesc_readdir(struct vop_readdir_args *ap)
 416 {
 417         struct uio *uio = ap->a_uio;
 418         struct filedesc *fdp;
 419         int error, i, fcnt;
 420         size_t namelen;
 421         char name[20]; /* enough for %d */
 422
 423         /*
 424          * We don't allow exporting fdesc mounts, and currently local
 425          * requests do not need cookies.
 426          */
 427         if (ap->a_ncookies)
 428                 panic("fdesc_readdir: not hungry");
 429
 430         if (VTOFDESC(ap->a_vp)->fd_type != Froot)
 431                 panic("fdesc_readdir: not dir");
 432
 433         if (uio->uio_offset < 0 || uio->uio_offset > INT_MAX ||
 434             uio->uio_resid < 0)
 435                 return(EINVAL);
 436         i = uio->uio_offset;
 437         KKASSERT(uio->uio_td->td_proc);
 438         fdp = uio->uio_td->td_proc->p_fd;
 439         error = 0;
 440
 441         fcnt = i - 2;           /* The first two nodes are `.' and `..' */
 442
 443         while (fcnt < fdp->fd_nfiles && uio->uio_resid > 0 && !error) {
 444                 switch (i) {
 445                 case 0: /* `.' */
 446                         if (vop_write_dirent(&error, uio, FD_ROOT + i, DT_DIR,
 447                                          1, "."))
 448                                 goto done;
 449                         if (error)
 450                                 return (error);
 451                         break;
 452                 case 1: /* `..' */
 453                         if (vop_write_dirent(&error, uio, FD_ROOT + i, DT_DIR,
 454                                              2, ".."))
 455                                 goto done;
 456                         if (error)
 457                                 return (error);
 458                         break;
 459                 default:
 460                         if (fdp->fd_files[fcnt].fp == NULL) {
 461                                 fcnt++;
 462                                 continue;
 463                         }
 464
 465                         namelen = snprintf(name, sizeof(name), "%d", fcnt);
 466                         if (vop_write_dirent(&error, uio, FD_ROOT + i,
 467                                              DT_UNKNOWN, namelen, name))
 468                                 goto done;
 469                         if (error)
 470                                 return (error);
 471                         break;
 472                 }
 473                 i++;
 474                 fcnt++;
 475         }
 476
 477 done:
 478         if (i >= 2)
 479                 uio->uio_offset = fcnt + 2;
 480         else
 481                 uio->uio_offset = i;
 482         return (error);
 483 }
 484
 485 /*
 486  * fdesc_poll(struct vnode *a_vp, int a_events, struct ucred *a_cred,
 487  *            struct thread *a_td)
 488  */
 489 static int
 490 fdesc_poll(struct vop_poll_args *ap)
 491 {
 492         return seltrue(0, ap->a_events, ap->a_td);
 493 }
 494
 495 /*
 496  * fdesc_inactive(struct vnode *a_vp, struct thread *a_td)
 497  */
 498 static int
 499 fdesc_inactive(struct vop_inactive_args *ap)
 500 {
 501         struct vnode *vp = ap->a_vp;
 502
 503         /*
 504          * Clear out the v_type field to avoid
 505          * nasty things happening in vgone().
 506          */
 507         vp->v_type = VNON;
 508         return (0);
 509 }
 510
 511 /*
 512  * fdesc_reclaim(struct vnode *a_vp)
 513  */
 514 static int
 515 fdesc_reclaim(struct vop_reclaim_args *ap)
 516 {
 517         struct vnode *vp = ap->a_vp;
 518         struct fdescnode *fd = VTOFDESC(vp);
 519
 520         LIST_REMOVE(fd, fd_hash);
 521         FREE(vp->v_data, M_TEMP);
 522         vp->v_data = 0;
 523
 524         return (0);
 525 }
 526
 527 /*
 528  * Print out the contents of a /dev/fd vnode.
 529  *
 530  * fdesc_print(struct vnode *a_vp)
 531  */
 532 /* ARGSUSED */
 533 static int
 534 fdesc_print(struct vop_print_args *ap)
 535 {
 536         printf("tag VT_NON, fdesc vnode\n");
 537         return (0);
 538 }
 539
 540 struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = {
 541         { &vop_default_desc,            vop_defaultop },
 542         { &vop_access_desc,             vop_null },
 543         { &vop_getattr_desc,            (vnodeopv_entry_t) fdesc_getattr },
 544         { &vop_inactive_desc,           (vnodeopv_entry_t) fdesc_inactive },
 545         { &vop_lookup_desc,             (vnodeopv_entry_t) fdesc_lookup },
 546         { &vop_open_desc,               (vnodeopv_entry_t) fdesc_open },
 547         { &vop_pathconf_desc,           (vnodeopv_entry_t) vop_stdpathconf },
 548         { &vop_poll_desc,               (vnodeopv_entry_t) fdesc_poll },
 549         { &vop_print_desc,              (vnodeopv_entry_t) fdesc_print },
 550         { &vop_readdir_desc,            (vnodeopv_entry_t) fdesc_readdir },
 551         { &vop_reclaim_desc,            (vnodeopv_entry_t) fdesc_reclaim },
 552         { &vop_setattr_desc,            (vnodeopv_entry_t) fdesc_setattr },
 553         { NULL, NULL }
 554 };
 555