1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1992, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software donated to Berkeley by 8 * Jan-Simon Pendry. 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 * 3. 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 * @(#)fdesc_vnops.c 8.9 (Berkeley) 1/21/94 35 * 36 * $FreeBSD$ 37 */ 38 39 /* 40 * /dev/fd Filesystem 41 */ 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/capsicum.h> 46 #include <sys/conf.h> 47 #include <sys/dirent.h> 48 #include <sys/filedesc.h> 49 #include <sys/kernel.h> /* boottime */ 50 #include <sys/lock.h> 51 #include <sys/mutex.h> 52 #include <sys/malloc.h> 53 #include <sys/file.h> /* Must come after sys/malloc.h */ 54 #include <sys/mount.h> 55 #include <sys/namei.h> 56 #include <sys/proc.h> 57 #include <sys/stat.h> 58 #include <sys/syscallsubr.h> 59 #include <sys/unistd.h> 60 #include <sys/vnode.h> 61 62 #include <fs/fdescfs/fdesc.h> 63 64 #define NFDCACHE 4 65 #define FD_NHASH(ix) \ 66 (&fdhashtbl[(ix) & fdhash]) 67 static LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl; 68 static u_long fdhash; 69 70 struct mtx fdesc_hashmtx; 71 72 static vop_getattr_t fdesc_getattr; 73 static vop_lookup_t fdesc_lookup; 74 static vop_open_t fdesc_open; 75 static vop_pathconf_t fdesc_pathconf; 76 static vop_readdir_t fdesc_readdir; 77 static vop_readlink_t fdesc_readlink; 78 static vop_reclaim_t fdesc_reclaim; 79 static vop_setattr_t fdesc_setattr; 80 81 static struct vop_vector fdesc_vnodeops = { 82 .vop_default = &default_vnodeops, 83 84 .vop_access = VOP_NULL, 85 .vop_getattr = fdesc_getattr, 86 .vop_lookup = fdesc_lookup, 87 .vop_open = fdesc_open, 88 .vop_pathconf = fdesc_pathconf, 89 .vop_readdir = fdesc_readdir, 90 .vop_readlink = fdesc_readlink, 91 .vop_reclaim = fdesc_reclaim, 92 .vop_setattr = fdesc_setattr, 93 }; 94 VFS_VOP_VECTOR_REGISTER(fdesc_vnodeops); 95 96 static void fdesc_insmntque_dtr(struct vnode *, void *); 97 static void fdesc_remove_entry(struct fdescnode *); 98 99 /* 100 * Initialise cache headers 101 */ 102 int 103 fdesc_init(struct vfsconf *vfsp) 104 { 105 106 mtx_init(&fdesc_hashmtx, "fdescfs_hash", NULL, MTX_DEF); 107 fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash); 108 return (0); 109 } 110 111 /* 112 * Uninit ready for unload. 113 */ 114 int 115 fdesc_uninit(struct vfsconf *vfsp) 116 { 117 118 hashdestroy(fdhashtbl, M_CACHE, fdhash); 119 mtx_destroy(&fdesc_hashmtx); 120 return (0); 121 } 122 123 /* 124 * If allocating vnode fails, call this. 125 */ 126 static void 127 fdesc_insmntque_dtr(struct vnode *vp, void *arg) 128 { 129 130 vgone(vp); 131 vput(vp); 132 } 133 134 /* 135 * Remove an entry from the hash if it exists. 136 */ 137 static void 138 fdesc_remove_entry(struct fdescnode *fd) 139 { 140 struct fdhashhead *fc; 141 struct fdescnode *fd2; 142 143 fc = FD_NHASH(fd->fd_ix); 144 mtx_lock(&fdesc_hashmtx); 145 LIST_FOREACH(fd2, fc, fd_hash) { 146 if (fd == fd2) { 147 LIST_REMOVE(fd, fd_hash); 148 break; 149 } 150 } 151 mtx_unlock(&fdesc_hashmtx); 152 } 153 154 int 155 fdesc_allocvp(fdntype ftype, unsigned fd_fd, int ix, struct mount *mp, 156 struct vnode **vpp) 157 { 158 struct fdescmount *fmp; 159 struct fdhashhead *fc; 160 struct fdescnode *fd, *fd2; 161 struct vnode *vp, *vp2; 162 struct thread *td; 163 int error; 164 165 td = curthread; 166 fc = FD_NHASH(ix); 167 loop: 168 mtx_lock(&fdesc_hashmtx); 169 /* 170 * If a forced unmount is progressing, we need to drop it. The flags are 171 * protected by the hashmtx. 172 */ 173 fmp = mp->mnt_data; 174 if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) { 175 mtx_unlock(&fdesc_hashmtx); 176 return (-1); 177 } 178 179 LIST_FOREACH(fd, fc, fd_hash) { 180 if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) { 181 /* Get reference to vnode in case it's being free'd */ 182 vp = fd->fd_vnode; 183 VI_LOCK(vp); 184 mtx_unlock(&fdesc_hashmtx); 185 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK)) 186 goto loop; 187 *vpp = vp; 188 return (0); 189 } 190 } 191 mtx_unlock(&fdesc_hashmtx); 192 193 fd = malloc(sizeof(struct fdescnode), M_TEMP, M_WAITOK); 194 195 error = getnewvnode("fdescfs", mp, &fdesc_vnodeops, &vp); 196 if (error) { 197 free(fd, M_TEMP); 198 return (error); 199 } 200 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 201 vp->v_data = fd; 202 fd->fd_vnode = vp; 203 fd->fd_type = ftype; 204 fd->fd_fd = fd_fd; 205 fd->fd_ix = ix; 206 if (ftype == Fdesc && fmp->flags & FMNT_LINRDLNKF) 207 vp->v_vflag |= VV_READLINK; 208 error = insmntque1(vp, mp, fdesc_insmntque_dtr, NULL); 209 if (error != 0) { 210 *vpp = NULLVP; 211 return (error); 212 } 213 214 /* Make sure that someone didn't beat us when inserting the vnode. */ 215 mtx_lock(&fdesc_hashmtx); 216 /* 217 * If a forced unmount is progressing, we need to drop it. The flags are 218 * protected by the hashmtx. 219 */ 220 fmp = mp->mnt_data; 221 if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) { 222 mtx_unlock(&fdesc_hashmtx); 223 vgone(vp); 224 vput(vp); 225 *vpp = NULLVP; 226 return (-1); 227 } 228 229 LIST_FOREACH(fd2, fc, fd_hash) { 230 if (fd2->fd_ix == ix && fd2->fd_vnode->v_mount == mp) { 231 /* Get reference to vnode in case it's being free'd */ 232 vp2 = fd2->fd_vnode; 233 VI_LOCK(vp2); 234 mtx_unlock(&fdesc_hashmtx); 235 error = vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK); 236 /* Someone beat us, dec use count and wait for reclaim */ 237 vgone(vp); 238 vput(vp); 239 /* If we didn't get it, return no vnode. */ 240 if (error) 241 vp2 = NULLVP; 242 *vpp = vp2; 243 return (error); 244 } 245 } 246 247 /* If we came here, we can insert it safely. */ 248 LIST_INSERT_HEAD(fc, fd, fd_hash); 249 mtx_unlock(&fdesc_hashmtx); 250 *vpp = vp; 251 return (0); 252 } 253 254 struct fdesc_get_ino_args { 255 fdntype ftype; 256 unsigned fd_fd; 257 int ix; 258 struct file *fp; 259 struct thread *td; 260 }; 261 262 static int 263 fdesc_get_ino_alloc(struct mount *mp, void *arg, int lkflags, 264 struct vnode **rvp) 265 { 266 struct fdesc_get_ino_args *a; 267 struct fdescmount *fdm; 268 struct vnode *vp; 269 int error; 270 271 a = arg; 272 fdm = VFSTOFDESC(mp); 273 if ((fdm->flags & FMNT_NODUP) != 0 && a->fp->f_type == DTYPE_VNODE) { 274 vp = a->fp->f_vnode; 275 vget(vp, lkflags | LK_RETRY); 276 *rvp = vp; 277 error = 0; 278 } else { 279 error = fdesc_allocvp(a->ftype, a->fd_fd, a->ix, mp, rvp); 280 } 281 fdrop(a->fp, a->td); 282 return (error); 283 } 284 285 /* 286 * vp is the current namei directory 287 * ndp is the name to locate in that directory... 288 */ 289 static int 290 fdesc_lookup(struct vop_lookup_args *ap) 291 { 292 struct vnode **vpp = ap->a_vpp; 293 struct vnode *dvp = ap->a_dvp; 294 struct componentname *cnp = ap->a_cnp; 295 char *pname = cnp->cn_nameptr; 296 struct thread *td = curthread; 297 struct file *fp; 298 struct fdesc_get_ino_args arg; 299 int nlen = cnp->cn_namelen; 300 u_int fd, fd1; 301 int error; 302 struct vnode *fvp; 303 304 if ((cnp->cn_flags & ISLASTCN) && 305 (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) { 306 error = EROFS; 307 goto bad; 308 } 309 310 if (cnp->cn_namelen == 1 && *pname == '.') { 311 *vpp = dvp; 312 VREF(dvp); 313 return (0); 314 } 315 316 if (VTOFDESC(dvp)->fd_type != Froot) { 317 error = ENOTDIR; 318 goto bad; 319 } 320 321 fd = 0; 322 /* the only time a leading 0 is acceptable is if it's "0" */ 323 if (*pname == '0' && nlen != 1) { 324 error = ENOENT; 325 goto bad; 326 } 327 while (nlen--) { 328 if (*pname < '0' || *pname > '9') { 329 error = ENOENT; 330 goto bad; 331 } 332 fd1 = 10 * fd + *pname++ - '0'; 333 if (fd1 < fd) { 334 error = ENOENT; 335 goto bad; 336 } 337 fd = fd1; 338 } 339 340 /* 341 * No rights to check since 'fp' isn't actually used. 342 */ 343 if ((error = fget(td, fd, &cap_no_rights, &fp)) != 0) 344 goto bad; 345 346 /* Check if we're looking up ourselves. */ 347 if (VTOFDESC(dvp)->fd_ix == FD_DESC + fd) { 348 /* 349 * In case we're holding the last reference to the file, the dvp 350 * will be re-acquired. 351 */ 352 vhold(dvp); 353 VOP_UNLOCK(dvp); 354 fdrop(fp, td); 355 356 /* Re-aquire the lock afterwards. */ 357 vn_lock(dvp, LK_RETRY | LK_EXCLUSIVE); 358 vdrop(dvp); 359 fvp = dvp; 360 if (VN_IS_DOOMED(dvp)) 361 error = ENOENT; 362 } else { 363 /* 364 * Unlock our root node (dvp) when doing this, since we might 365 * deadlock since the vnode might be locked by another thread 366 * and the root vnode lock will be obtained afterwards (in case 367 * we're looking up the fd of the root vnode), which will be the 368 * opposite lock order. Vhold the root vnode first so we don't 369 * lose it. 370 */ 371 arg.ftype = Fdesc; 372 arg.fd_fd = fd; 373 arg.ix = FD_DESC + fd; 374 arg.fp = fp; 375 arg.td = td; 376 error = vn_vget_ino_gen(dvp, fdesc_get_ino_alloc, &arg, 377 LK_EXCLUSIVE, &fvp); 378 } 379 380 if (error) 381 goto bad; 382 *vpp = fvp; 383 return (0); 384 385 bad: 386 *vpp = NULL; 387 return (error); 388 } 389 390 static int 391 fdesc_open(struct vop_open_args *ap) 392 { 393 struct vnode *vp = ap->a_vp; 394 395 if (VTOFDESC(vp)->fd_type == Froot) 396 return (0); 397 398 /* 399 * XXX Kludge: set td->td_proc->p_dupfd to contain the value of the file 400 * descriptor being sought for duplication. The error return ensures 401 * that the vnode for this device will be released by vn_open. Open 402 * will detect this special error and take the actions in dupfdopen. 403 * Other callers of vn_open or VOP_OPEN will simply report the 404 * error. 405 */ 406 ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd; /* XXX */ 407 return (ENODEV); 408 } 409 410 static int 411 fdesc_pathconf(struct vop_pathconf_args *ap) 412 { 413 struct vnode *vp = ap->a_vp; 414 int error; 415 416 switch (ap->a_name) { 417 case _PC_NAME_MAX: 418 *ap->a_retval = NAME_MAX; 419 return (0); 420 case _PC_LINK_MAX: 421 if (VTOFDESC(vp)->fd_type == Froot) 422 *ap->a_retval = 2; 423 else 424 *ap->a_retval = 1; 425 return (0); 426 default: 427 if (VTOFDESC(vp)->fd_type == Froot) 428 return (vop_stdpathconf(ap)); 429 vref(vp); 430 VOP_UNLOCK(vp); 431 error = kern_fpathconf(curthread, VTOFDESC(vp)->fd_fd, 432 ap->a_name, ap->a_retval); 433 vn_lock(vp, LK_SHARED | LK_RETRY); 434 vunref(vp); 435 return (error); 436 } 437 } 438 439 static int 440 fdesc_getattr(struct vop_getattr_args *ap) 441 { 442 struct vnode *vp = ap->a_vp; 443 struct vattr *vap = ap->a_vap; 444 struct timeval boottime; 445 446 getboottime(&boottime); 447 vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH; 448 vap->va_fileid = VTOFDESC(vp)->fd_ix; 449 vap->va_uid = 0; 450 vap->va_gid = 0; 451 vap->va_blocksize = DEV_BSIZE; 452 vap->va_atime.tv_sec = boottime.tv_sec; 453 vap->va_atime.tv_nsec = 0; 454 vap->va_mtime = vap->va_atime; 455 vap->va_ctime = vap->va_mtime; 456 vap->va_gen = 0; 457 vap->va_flags = 0; 458 vap->va_bytes = 0; 459 vap->va_filerev = 0; 460 461 switch (VTOFDESC(vp)->fd_type) { 462 case Froot: 463 vap->va_type = VDIR; 464 vap->va_nlink = 2; 465 vap->va_size = DEV_BSIZE; 466 vap->va_rdev = NODEV; 467 break; 468 469 case Fdesc: 470 vap->va_type = (vp->v_vflag & VV_READLINK) == 0 ? VCHR : VLNK; 471 vap->va_nlink = 1; 472 vap->va_size = 0; 473 vap->va_rdev = makedev(0, vap->va_fileid); 474 break; 475 476 default: 477 panic("fdesc_getattr"); 478 break; 479 } 480 481 vp->v_type = vap->va_type; 482 return (0); 483 } 484 485 static int 486 fdesc_setattr(struct vop_setattr_args *ap) 487 { 488 struct vattr *vap = ap->a_vap; 489 struct vnode *vp; 490 struct mount *mp; 491 struct file *fp; 492 struct thread *td = curthread; 493 cap_rights_t rights; 494 unsigned fd; 495 int error; 496 497 /* 498 * Can't mess with the root vnode 499 */ 500 if (VTOFDESC(ap->a_vp)->fd_type == Froot) 501 return (EACCES); 502 503 fd = VTOFDESC(ap->a_vp)->fd_fd; 504 505 /* 506 * Allow setattr where there is an underlying vnode. 507 */ 508 error = getvnode(td, fd, 509 cap_rights_init_one(&rights, CAP_EXTATTR_SET), &fp); 510 if (error) { 511 /* 512 * getvnode() returns EINVAL if the file descriptor is not 513 * backed by a vnode. Silently drop all changes except 514 * chflags(2) in this case. 515 */ 516 if (error == EINVAL) { 517 if (vap->va_flags != VNOVAL) 518 error = EOPNOTSUPP; 519 else 520 error = 0; 521 } 522 return (error); 523 } 524 vp = fp->f_vnode; 525 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) == 0) { 526 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 527 error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred); 528 VOP_UNLOCK(vp); 529 vn_finished_write(mp); 530 } 531 fdrop(fp, td); 532 return (error); 533 } 534 535 #define UIO_MX _GENERIC_DIRLEN(10) /* number of symbols in INT_MAX printout */ 536 537 static int 538 fdesc_readdir(struct vop_readdir_args *ap) 539 { 540 struct fdescmount *fmp; 541 struct uio *uio = ap->a_uio; 542 struct filedesc *fdp; 543 struct dirent d; 544 struct dirent *dp = &d; 545 int error, i, off, fcnt; 546 547 if (VTOFDESC(ap->a_vp)->fd_type != Froot) 548 panic("fdesc_readdir: not dir"); 549 550 fmp = VFSTOFDESC(ap->a_vp->v_mount); 551 if (ap->a_ncookies != NULL) 552 *ap->a_ncookies = 0; 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 bzero((caddr_t)dp, UIO_MX); 567 switch (i) { 568 case 0: /* `.' */ 569 case 1: /* `..' */ 570 dp->d_fileno = i + FD_ROOT; 571 dp->d_namlen = i + 1; 572 dp->d_reclen = UIO_MX; 573 bcopy("..", dp->d_name, dp->d_namlen); 574 dp->d_type = DT_DIR; 575 dirent_terminate(dp); 576 break; 577 default: 578 if (fdp->fd_ofiles[fcnt].fde_file == NULL) 579 break; 580 dp->d_namlen = sprintf(dp->d_name, "%d", fcnt); 581 dp->d_reclen = UIO_MX; 582 dp->d_type = (fmp->flags & FMNT_LINRDLNKF) == 0 ? 583 DT_CHR : DT_LNK; 584 dp->d_fileno = i + FD_DESC; 585 dirent_terminate(dp); 586 break; 587 } 588 /* NOTE: d_off is the offset of the *next* entry. */ 589 dp->d_off = UIO_MX * (i + 1); 590 if (dp->d_namlen != 0) { 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 } 600 i++; 601 fcnt++; 602 } 603 FILEDESC_SUNLOCK(fdp); 604 605 done: 606 uio->uio_offset = i * UIO_MX; 607 return (error); 608 } 609 610 static int 611 fdesc_reclaim(struct vop_reclaim_args *ap) 612 { 613 struct vnode *vp; 614 struct fdescnode *fd; 615 616 vp = ap->a_vp; 617 fd = VTOFDESC(vp); 618 fdesc_remove_entry(fd); 619 free(vp->v_data, M_TEMP); 620 vp->v_data = NULL; 621 return (0); 622 } 623 624 static int 625 fdesc_readlink(struct vop_readlink_args *va) 626 { 627 struct vnode *vp, *vn; 628 struct thread *td; 629 struct uio *uio; 630 struct file *fp; 631 char *freepath, *fullpath; 632 size_t pathlen; 633 int lockflags, fd_fd; 634 int error; 635 636 freepath = NULL; 637 vn = va->a_vp; 638 if (VTOFDESC(vn)->fd_type != Fdesc) 639 panic("fdesc_readlink: not fdescfs link"); 640 fd_fd = ((struct fdescnode *)vn->v_data)->fd_fd; 641 lockflags = VOP_ISLOCKED(vn); 642 VOP_UNLOCK(vn); 643 644 td = curthread; 645 error = fget_cap(td, fd_fd, &cap_no_rights, &fp, NULL); 646 if (error != 0) 647 goto out; 648 649 switch (fp->f_type) { 650 case DTYPE_VNODE: 651 vp = fp->f_vnode; 652 error = vn_fullpath(vp, &fullpath, &freepath); 653 break; 654 default: 655 fullpath = "anon_inode:[unknown]"; 656 break; 657 } 658 if (error == 0) { 659 uio = va->a_uio; 660 pathlen = strlen(fullpath); 661 error = uiomove(fullpath, pathlen, uio); 662 } 663 if (freepath != NULL) 664 free(freepath, M_TEMP); 665 fdrop(fp, td); 666 667 out: 668 vn_lock(vn, lockflags | LK_RETRY); 669 return (error); 670 } 671