sys/fs/fdescfs/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  * 4. Neither the name of the University nor the names of its contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  *
  32  *      @(#)fdesc_vnops.c       8.9 (Berkeley) 1/21/94
  33  *
  34  * $FreeBSD$
  35  */
  36
  37 /*
  38  * /dev/fd Filesystem
  39  */
  40
  41 #include <sys/param.h>
  42 #include <sys/systm.h>
  43 #include <sys/conf.h>
  44 #include <sys/dirent.h>
  45 #include <sys/filedesc.h>
  46 #include <sys/kernel.h> /* boottime */
  47 #include <sys/lock.h>
  48 #include <sys/mutex.h>
  49 #include <sys/malloc.h>
  50 #include <sys/file.h>   /* Must come after sys/malloc.h */
  51 #include <sys/mount.h>
  52 #include <sys/namei.h>
  53 #include <sys/proc.h>
  54 #include <sys/stat.h>
  55 #include <sys/vnode.h>
  56
  57 #include <fs/fdescfs/fdesc.h>
  58
  59 #define NFDCACHE 4
  60 #define FD_NHASH(ix) \
  61         (&fdhashtbl[(ix) & fdhash])
  62 static LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
  63 static u_long fdhash;
  64
  65 struct mtx fdesc_hashmtx;
  66
  67 static vop_getattr_t    fdesc_getattr;
  68 static vop_lookup_t     fdesc_lookup;
  69 static vop_open_t       fdesc_open;
  70 static vop_readdir_t    fdesc_readdir;
  71 static vop_reclaim_t    fdesc_reclaim;
  72 static vop_setattr_t    fdesc_setattr;
  73
  74 static struct vop_vector fdesc_vnodeops = {
  75         .vop_default =          &default_vnodeops,
  76
  77         .vop_access =           VOP_NULL,
  78         .vop_getattr =          fdesc_getattr,
  79         .vop_lookup =           fdesc_lookup,
  80         .vop_open =             fdesc_open,
  81         .vop_pathconf =         vop_stdpathconf,
  82         .vop_readdir =          fdesc_readdir,
  83         .vop_reclaim =          fdesc_reclaim,
  84         .vop_setattr =          fdesc_setattr,
  85 };
  86
  87 static void fdesc_insmntque_dtr(struct vnode *, void *);
  88 static void fdesc_remove_entry(struct fdescnode *);
  89
  90 /*
  91  * Initialise cache headers
  92  */
  93 int
  94 fdesc_init(vfsp)
  95         struct vfsconf *vfsp;
  96 {
  97
  98         mtx_init(&fdesc_hashmtx, "fdescfs_hash", NULL, MTX_DEF);
  99         fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash);
 100         return (0);
 101 }
 102
 103 /*
 104  * Uninit ready for unload.
 105  */
 106 int
 107 fdesc_uninit(vfsp)
 108         struct vfsconf *vfsp;
 109 {
 110
 111         hashdestroy(fdhashtbl, M_CACHE, fdhash);
 112         mtx_destroy(&fdesc_hashmtx);
 113         return (0);
 114 }
 115
 116 /*
 117  * If allocating vnode fails, call this.
 118  */
 119 static void
 120 fdesc_insmntque_dtr(struct vnode *vp, void *arg)
 121 {
 122
 123         vgone(vp);
 124         vput(vp);
 125 }
 126
 127 /*
 128  * Remove an entry from the hash if it exists.
 129  */
 130 static void
 131 fdesc_remove_entry(struct fdescnode *fd)
 132 {
 133         struct fdhashhead *fc;
 134         struct fdescnode *fd2;
 135
 136         fc = FD_NHASH(fd->fd_ix);
 137         mtx_lock(&fdesc_hashmtx);
 138         LIST_FOREACH(fd2, fc, fd_hash) {
 139                 if (fd == fd2) {
 140                         LIST_REMOVE(fd, fd_hash);
 141                         break;
 142                 }
 143         }
 144         mtx_unlock(&fdesc_hashmtx);
 145 }
 146
 147 int
 148 fdesc_allocvp(ftype, fd_fd, ix, mp, vpp, td)
 149         fdntype ftype;
 150         unsigned fd_fd;
 151         int ix;
 152         struct mount *mp;
 153         struct vnode **vpp;
 154         struct thread *td;
 155 {
 156         struct fdescmount *fmp;
 157         struct fdhashhead *fc;
 158         struct fdescnode *fd, *fd2;
 159         struct vnode *vp, *vp2;
 160         int error = 0;
 161
 162         fc = FD_NHASH(ix);
 163 loop:
 164         mtx_lock(&fdesc_hashmtx);
 165         /*
 166          * If a forced unmount is progressing, we need to drop it. The flags are
 167          * protected by the hashmtx.
 168          */
 169         fmp = (struct fdescmount *)mp->mnt_data;
 170         if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) {
 171                 mtx_unlock(&fdesc_hashmtx);
 172                 return (-1);
 173         }
 174
 175         LIST_FOREACH(fd, fc, fd_hash) {
 176                 if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
 177                         /* Get reference to vnode in case it's being free'd */
 178                         vp = fd->fd_vnode;
 179                         VI_LOCK(vp);
 180                         mtx_unlock(&fdesc_hashmtx);
 181                         if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td))
 182                                 goto loop;
 183                         *vpp = vp;
 184                         return (0);
 185                 }
 186         }
 187         mtx_unlock(&fdesc_hashmtx);
 188
 189         MALLOC(fd, struct fdescnode *, sizeof(struct fdescnode), M_TEMP, M_WAITOK);
 190
 191         error = getnewvnode("fdescfs", mp, &fdesc_vnodeops, &vp);
 192         if (error) {
 193                 FREE(fd, M_TEMP);
 194                 return (error);
 195         }
 196         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 197         vp->v_data = fd;
 198         fd->fd_vnode = vp;
 199         fd->fd_type = ftype;
 200         fd->fd_fd = fd_fd;
 201         fd->fd_ix = ix;
 202         error = insmntque1(vp, mp, fdesc_insmntque_dtr, NULL);
 203         if (error != 0) {
 204                 *vpp = NULLVP;
 205                 return (error);
 206         }
 207
 208         /* Make sure that someone didn't beat us when inserting the vnode. */
 209         mtx_lock(&fdesc_hashmtx);
 210         /*
 211          * If a forced unmount is progressing, we need to drop it. The flags are
 212          * protected by the hashmtx.
 213          */
 214         fmp = (struct fdescmount *)mp->mnt_data;
 215         if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) {
 216                 mtx_unlock(&fdesc_hashmtx);
 217                 vgone(vp);
 218                 vput(vp);
 219                 *vpp = NULLVP;
 220                 return (-1);
 221         }
 222
 223         LIST_FOREACH(fd2, fc, fd_hash) {
 224                 if (fd2->fd_ix == ix && fd2->fd_vnode->v_mount == mp) {
 225                         /* Get reference to vnode in case it's being free'd */
 226                         vp2 = fd2->fd_vnode;
 227                         VI_LOCK(vp2);
 228                         mtx_unlock(&fdesc_hashmtx);
 229                         error = vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK, td);
 230                         /* Someone beat us, dec use count and wait for reclaim */
 231                         vgone(vp);
 232                         vput(vp);
 233                         /* If we didn't get it, return no vnode. */
 234                         if (error)
 235                                 vp2 = NULLVP;
 236                         *vpp = vp2;
 237                         return (error);
 238                 }
 239         }
 240
 241         /* If we came here, we can insert it safely. */
 242         LIST_INSERT_HEAD(fc, fd, fd_hash);
 243         mtx_unlock(&fdesc_hashmtx);
 244         *vpp = vp;
 245         return (0);
 246 }
 247
 248 /*
 249  * vp is the current namei directory
 250  * ndp is the name to locate in that directory...
 251  */
 252 static int
 253 fdesc_lookup(ap)
 254         struct vop_lookup_args /* {
 255                 struct vnode * a_dvp;
 256                 struct vnode ** a_vpp;
 257                 struct componentname * a_cnp;
 258         } */ *ap;
 259 {
 260         struct vnode **vpp = ap->a_vpp;
 261         struct vnode *dvp = ap->a_dvp;
 262         struct componentname *cnp = ap->a_cnp;
 263         char *pname = cnp->cn_nameptr;
 264         struct thread *td = cnp->cn_thread;
 265         struct file *fp;
 266         int nlen = cnp->cn_namelen;
 267         u_int fd;
 268         int error;
 269         struct vnode *fvp;
 270
 271         if ((cnp->cn_flags & ISLASTCN) &&
 272             (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
 273                 error = EROFS;
 274                 goto bad;
 275         }
 276
 277         if (cnp->cn_namelen == 1 && *pname == '.') {
 278                 *vpp = dvp;
 279                 VREF(dvp);
 280                 return (0);
 281         }
 282
 283         if (VTOFDESC(dvp)->fd_type != Froot) {
 284                 error = ENOTDIR;
 285                 goto bad;
 286         }
 287
 288         fd = 0;
 289         /* the only time a leading 0 is acceptable is if it's "0" */
 290         if (*pname == '0' && nlen != 1) {
 291                 error = ENOENT;
 292                 goto bad;
 293         }
 294         while (nlen--) {
 295                 if (*pname < '0' || *pname > '9') {
 296                         error = ENOENT;
 297                         goto bad;
 298                 }
 299                 fd = 10 * fd + *pname++ - '0';
 300         }
 301
 302         if ((error = fget(td, fd, &fp)) != 0)
 303                 goto bad;
 304
 305         /* Check if we're looking up ourselves. */
 306         if (VTOFDESC(dvp)->fd_ix == FD_DESC + fd) {
 307                 /*
 308                  * In case we're holding the last reference to the file, the dvp
 309                  * will be re-acquired.
 310                  */
 311                 vhold(dvp);
 312                 VOP_UNLOCK(dvp, 0);
 313                 fdrop(fp, td);
 314
 315                 /* Re-aquire the lock afterwards. */
 316                 vn_lock(dvp, LK_RETRY | LK_EXCLUSIVE);
 317                 vdrop(dvp);
 318                 fvp = dvp;
 319         } else {
 320                 /*
 321                  * Unlock our root node (dvp) when doing this, since we might
 322                  * deadlock since the vnode might be locked by another thread
 323                  * and the root vnode lock will be obtained afterwards (in case
 324                  * we're looking up the fd of the root vnode), which will be the
 325                  * opposite lock order. Vhold the root vnode first so we don't
 326                  * loose it.
 327                  */
 328                 vhold(dvp);
 329                 VOP_UNLOCK(dvp, 0);
 330                 error = fdesc_allocvp(Fdesc, fd, FD_DESC + fd, dvp->v_mount,
 331                     &fvp, td);
 332                 fdrop(fp, td);
 333                 /*
 334                  * The root vnode must be locked last to prevent deadlock condition.
 335                  */
 336                 vn_lock(dvp, LK_RETRY | LK_EXCLUSIVE);
 337                 vdrop(dvp);
 338         }
 339
 340         if (error)
 341                 goto bad;
 342         *vpp = fvp;
 343         return (0);
 344
 345 bad:
 346         *vpp = NULL;
 347         return (error);
 348 }
 349
 350 static int
 351 fdesc_open(ap)
 352         struct vop_open_args /* {
 353                 struct vnode *a_vp;
 354                 int  a_mode;
 355                 struct ucred *a_cred;
 356                 struct thread *a_td;
 357         } */ *ap;
 358 {
 359         struct vnode *vp = ap->a_vp;
 360
 361         if (VTOFDESC(vp)->fd_type == Froot)
 362                 return (0);
 363
 364         /*
 365          * XXX Kludge: set td->td_proc->p_dupfd to contain the value of the the file
 366          * descriptor being sought for duplication. The error return ensures
 367          * that the vnode for this device will be released by vn_open. Open
 368          * will detect this special error and take the actions in dupfdopen.
 369          * Other callers of vn_open or VOP_OPEN will simply report the
 370          * error.
 371          */
 372         ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd;       /* XXX */
 373         return (ENODEV);
 374 }
 375
 376 static int
 377 fdesc_getattr(ap)
 378         struct vop_getattr_args /* {
 379                 struct vnode *a_vp;
 380                 struct vattr *a_vap;
 381                 struct ucred *a_cred;
 382         } */ *ap;
 383 {
 384         struct vnode *vp = ap->a_vp;
 385         struct vattr *vap = ap->a_vap;
 386         struct thread *td = curthread;
 387         struct file *fp;
 388         struct stat stb;
 389         u_int fd;
 390         int error = 0;
 391
 392         switch (VTOFDESC(vp)->fd_type) {
 393         case Froot:
 394                 VATTR_NULL(vap);
 395
 396                 vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
 397                 vap->va_type = VDIR;
 398                 vap->va_nlink = 2;
 399                 vap->va_size = DEV_BSIZE;
 400                 vap->va_fileid = VTOFDESC(vp)->fd_ix;
 401                 vap->va_uid = 0;
 402                 vap->va_gid = 0;
 403                 vap->va_blocksize = DEV_BSIZE;
 404                 vap->va_atime.tv_sec = boottime.tv_sec;
 405                 vap->va_atime.tv_nsec = 0;
 406                 vap->va_mtime = vap->va_atime;
 407                 vap->va_ctime = vap->va_mtime;
 408                 vap->va_gen = 0;
 409                 vap->va_flags = 0;
 410                 vap->va_rdev = 0;
 411                 vap->va_bytes = 0;
 412                 break;
 413
 414         case Fdesc:
 415                 fd = VTOFDESC(vp)->fd_fd;
 416
 417                 if ((error = fget(td, fd, &fp)) != 0)
 418                         return (error);
 419
 420                 bzero(&stb, sizeof(stb));
 421                 error = fo_stat(fp, &stb, td->td_ucred, td);
 422                 fdrop(fp, td);
 423                 if (error == 0) {
 424                         VATTR_NULL(vap);
 425                         vap->va_type = IFTOVT(stb.st_mode);
 426                         vap->va_mode = stb.st_mode;
 427 #define FDRX (VREAD|VEXEC)
 428                         if (vap->va_type == VDIR)
 429                                 vap->va_mode &= ~((FDRX)|(FDRX>>3)|(FDRX>>6));
 430 #undef FDRX
 431                         vap->va_nlink = 1;
 432                         vap->va_flags = 0;
 433                         vap->va_bytes = stb.st_blocks * stb.st_blksize;
 434                         vap->va_fileid = VTOFDESC(vp)->fd_ix;
 435                         vap->va_size = stb.st_size;
 436                         vap->va_blocksize = stb.st_blksize;
 437                         vap->va_rdev = stb.st_rdev;
 438
 439                         /*
 440                          * If no time data is provided, use the current time.
 441                          */
 442                         if (stb.st_atimespec.tv_sec == 0 &&
 443                             stb.st_atimespec.tv_nsec == 0)
 444                                 nanotime(&stb.st_atimespec);
 445
 446                         if (stb.st_ctimespec.tv_sec == 0 &&
 447                             stb.st_ctimespec.tv_nsec == 0)
 448                                 nanotime(&stb.st_ctimespec);
 449
 450                         if (stb.st_mtimespec.tv_sec == 0 &&
 451                             stb.st_mtimespec.tv_nsec == 0)
 452                                 nanotime(&stb.st_mtimespec);
 453
 454                         vap->va_atime = stb.st_atimespec;
 455                         vap->va_mtime = stb.st_mtimespec;
 456                         vap->va_ctime = stb.st_ctimespec;
 457                         vap->va_uid = stb.st_uid;
 458                         vap->va_gid = stb.st_gid;
 459                 }
 460                 break;
 461
 462         default:
 463                 panic("fdesc_getattr");
 464                 break;
 465         }
 466
 467         if (error == 0)
 468                 vp->v_type = vap->va_type;
 469         return (error);
 470 }
 471
 472 static int
 473 fdesc_setattr(ap)
 474         struct vop_setattr_args /* {
 475                 struct vnode *a_vp;
 476                 struct vattr *a_vap;
 477                 struct ucred *a_cred;
 478         } */ *ap;
 479 {
 480         struct vattr *vap = ap->a_vap;
 481         struct vnode *vp;
 482         struct mount *mp;
 483         struct file *fp;
 484         struct thread *td = curthread;
 485         unsigned fd;
 486         int error;
 487
 488         /*
 489          * Can't mess with the root vnode
 490          */
 491         if (VTOFDESC(ap->a_vp)->fd_type == Froot)
 492                 return (EACCES);
 493
 494         fd = VTOFDESC(ap->a_vp)->fd_fd;
 495
 496         /*
 497          * Allow setattr where there is an underlying vnode.
 498          */
 499         error = getvnode(td->td_proc->p_fd, fd, &fp);
 500         if (error) {
 501                 /*
 502                  * getvnode() returns EINVAL if the file descriptor is not
 503                  * backed by a vnode.  Silently drop all changes except
 504                  * chflags(2) in this case.
 505                  */
 506                 if (error == EINVAL) {
 507                         if (vap->va_flags != VNOVAL)
 508                                 error = EOPNOTSUPP;
 509                         else
 510                                 error = 0;
 511                 }
 512                 return (error);
 513         }
 514         vp = fp->f_vnode;
 515         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) == 0) {
 516                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 517                 error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred);
 518                 VOP_UNLOCK(vp, 0);
 519                 vn_finished_write(mp);
 520         }
 521         fdrop(fp, td);
 522         return (error);
 523 }
 524
 525 #define UIO_MX 16
 526
 527 static int
 528 fdesc_readdir(ap)
 529         struct vop_readdir_args /* {
 530                 struct vnode *a_vp;
 531                 struct uio *a_uio;
 532                 struct ucred *a_cred;
 533                 int *a_eofflag;
 534                 u_long *a_cookies;
 535                 int a_ncookies;
 536         } */ *ap;
 537 {
 538         struct uio *uio = ap->a_uio;
 539         struct filedesc *fdp;
 540         struct dirent d;
 541         struct dirent *dp = &d;
 542         int error, i, off, fcnt;
 543
 544         /*
 545          * We don't allow exporting fdesc mounts, and currently local
 546          * requests do not need cookies.
 547          */
 548         if (ap->a_ncookies)
 549                 panic("fdesc_readdir: not hungry");
 550
 551         if (VTOFDESC(ap->a_vp)->fd_type != Froot)
 552                 panic("fdesc_readdir: not dir");
 553
 554         off = (int)uio->uio_offset;
 555         if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 ||
 556             uio->uio_resid < UIO_MX)
 557                 return (EINVAL);
 558         i = (u_int)off / UIO_MX;
 559         fdp = uio->uio_td->td_proc->p_fd;
 560         error = 0;
 561
 562         fcnt = i - 2;           /* The first two nodes are `.' and `..' */
 563
 564         FILEDESC_SLOCK(fdp);
 565         while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) {
 566                 switch (i) {
 567                 case 0: /* `.' */
 568                 case 1: /* `..' */
 569                         bzero((caddr_t)dp, UIO_MX);
 570
 571                         dp->d_fileno = i + FD_ROOT;
 572                         dp->d_namlen = i + 1;
 573                         dp->d_reclen = UIO_MX;
 574                         bcopy("..", dp->d_name, dp->d_namlen);
 575                         dp->d_name[i + 1] = '\0';
 576                         dp->d_type = DT_DIR;
 577                         break;
 578                 default:
 579                         if (fdp->fd_ofiles[fcnt] == NULL) {
 580                                 FILEDESC_SUNLOCK(fdp);
 581                                 goto done;
 582                         }
 583
 584                         bzero((caddr_t) dp, UIO_MX);
 585                         dp->d_namlen = sprintf(dp->d_name, "%d", fcnt);
 586                         dp->d_reclen = UIO_MX;
 587                         dp->d_type = DT_UNKNOWN;
 588                         dp->d_fileno = i + FD_DESC;
 589                         break;
 590                 }
 591                 /*
 592                  * And ship to userland
 593                  */
 594                 FILEDESC_SUNLOCK(fdp);
 595                 error = uiomove(dp, UIO_MX, uio);
 596                 if (error)
 597                         goto done;
 598                 FILEDESC_SLOCK(fdp);
 599                 i++;
 600                 fcnt++;
 601         }
 602         FILEDESC_SUNLOCK(fdp);
 603
 604 done:
 605         uio->uio_offset = i * UIO_MX;
 606         return (error);
 607 }
 608
 609 static int
 610 fdesc_reclaim(ap)
 611         struct vop_reclaim_args /* {
 612                 struct vnode *a_vp;
 613         } */ *ap;
 614 {
 615         struct vnode *vp;
 616         struct fdescnode *fd;
 617
 618         vp = ap->a_vp;
 619         fd = VTOFDESC(vp);
 620         fdesc_remove_entry(fd);
 621         FREE(vp->v_data, M_TEMP);
 622         vp->v_data = NULL;
 623         return (0);
 624 }