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_remove_entry(struct fdescnode *); 97 98 /* 99 * Initialise cache headers 100 */ 101 int 102 fdesc_init(struct vfsconf *vfsp) 103 { 104 105 mtx_init(&fdesc_hashmtx, "fdescfs_hash", NULL, MTX_DEF); 106 fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash); 107 return (0); 108 } 109 110 /* 111 * Uninit ready for unload. 112 */ 113 int 114 fdesc_uninit(struct vfsconf *vfsp) 115 { 116 117 hashdestroy(fdhashtbl, M_CACHE, fdhash); 118 mtx_destroy(&fdesc_hashmtx); 119 return (0); 120 } 121 122 /* 123 * Remove an entry from the hash if it exists. 124 */ 125 static void 126 fdesc_remove_entry(struct fdescnode *fd) 127 { 128 struct fdhashhead *fc; 129 struct fdescnode *fd2; 130 131 fc = FD_NHASH(fd->fd_ix); 132 mtx_lock(&fdesc_hashmtx); 133 LIST_FOREACH(fd2, fc, fd_hash) { 134 if (fd == fd2) { 135 LIST_REMOVE(fd, fd_hash); 136 break; 137 } 138 } 139 mtx_unlock(&fdesc_hashmtx); 140 } 141 142 int 143 fdesc_allocvp(fdntype ftype, unsigned fd_fd, int ix, struct mount *mp, 144 struct vnode **vpp) 145 { 146 struct fdescmount *fmp; 147 struct fdhashhead *fc; 148 struct fdescnode *fd, *fd2; 149 struct vnode *vp, *vp2; 150 int error; 151 152 fc = FD_NHASH(ix); 153 loop: 154 mtx_lock(&fdesc_hashmtx); 155 /* 156 * If a forced unmount is progressing, we need to drop it. The flags are 157 * protected by the hashmtx. 158 */ 159 fmp = mp->mnt_data; 160 if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) { 161 mtx_unlock(&fdesc_hashmtx); 162 return (-1); 163 } 164 165 LIST_FOREACH(fd, fc, fd_hash) { 166 if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) { 167 /* Get reference to vnode in case it's being free'd */ 168 vp = fd->fd_vnode; 169 VI_LOCK(vp); 170 mtx_unlock(&fdesc_hashmtx); 171 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK)) 172 goto loop; 173 *vpp = vp; 174 return (0); 175 } 176 } 177 mtx_unlock(&fdesc_hashmtx); 178 179 fd = malloc(sizeof(struct fdescnode), M_TEMP, M_WAITOK); 180 181 error = getnewvnode("fdescfs", mp, &fdesc_vnodeops, &vp); 182 if (error) { 183 free(fd, M_TEMP); 184 return (error); 185 } 186 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 187 vp->v_data = fd; 188 fd->fd_vnode = vp; 189 fd->fd_type = ftype; 190 fd->fd_fd = fd_fd; 191 fd->fd_ix = ix; 192 if (ftype == Fdesc && fmp->flags & FMNT_LINRDLNKF) 193 vp->v_vflag |= VV_READLINK; 194 error = insmntque1(vp, mp); 195 if (error != 0) { 196 vgone(vp); 197 vput(vp); 198 *vpp = NULLVP; 199 return (error); 200 } 201 202 /* Make sure that someone didn't beat us when inserting the vnode. */ 203 mtx_lock(&fdesc_hashmtx); 204 /* 205 * If a forced unmount is progressing, we need to drop it. The flags are 206 * protected by the hashmtx. 207 */ 208 fmp = mp->mnt_data; 209 if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) { 210 mtx_unlock(&fdesc_hashmtx); 211 vgone(vp); 212 vput(vp); 213 *vpp = NULLVP; 214 return (-1); 215 } 216 217 LIST_FOREACH(fd2, fc, fd_hash) { 218 if (fd2->fd_ix == ix && fd2->fd_vnode->v_mount == mp) { 219 /* Get reference to vnode in case it's being free'd */ 220 vp2 = fd2->fd_vnode; 221 VI_LOCK(vp2); 222 mtx_unlock(&fdesc_hashmtx); 223 error = vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK); 224 /* Someone beat us, dec use count and wait for reclaim */ 225 vgone(vp); 226 vput(vp); 227 /* If we didn't get it, return no vnode. */ 228 if (error) 229 vp2 = NULLVP; 230 *vpp = vp2; 231 return (error); 232 } 233 } 234 235 /* If we came here, we can insert it safely. */ 236 LIST_INSERT_HEAD(fc, fd, fd_hash); 237 mtx_unlock(&fdesc_hashmtx); 238 vn_set_state(vp, VSTATE_CONSTRUCTED); 239 *vpp = vp; 240 return (0); 241 } 242 243 struct fdesc_get_ino_args { 244 fdntype ftype; 245 unsigned fd_fd; 246 int ix; 247 struct file *fp; 248 struct thread *td; 249 }; 250 251 static int 252 fdesc_get_ino_alloc(struct mount *mp, void *arg, int lkflags, 253 struct vnode **rvp) 254 { 255 struct fdesc_get_ino_args *a; 256 struct fdescmount *fdm; 257 struct vnode *vp; 258 int error; 259 260 a = arg; 261 fdm = VFSTOFDESC(mp); 262 if ((fdm->flags & FMNT_NODUP) != 0 && a->fp->f_type == DTYPE_VNODE) { 263 vp = a->fp->f_vnode; 264 vget(vp, lkflags | LK_RETRY); 265 *rvp = vp; 266 error = 0; 267 } else { 268 error = fdesc_allocvp(a->ftype, a->fd_fd, a->ix, mp, rvp); 269 } 270 fdrop(a->fp, a->td); 271 return (error); 272 } 273 274 /* 275 * vp is the current namei directory 276 * ndp is the name to locate in that directory... 277 */ 278 static int 279 fdesc_lookup(struct vop_lookup_args *ap) 280 { 281 struct vnode **vpp = ap->a_vpp; 282 struct vnode *dvp = ap->a_dvp; 283 struct componentname *cnp = ap->a_cnp; 284 char *pname = cnp->cn_nameptr; 285 struct thread *td = curthread; 286 struct file *fp; 287 struct fdesc_get_ino_args arg; 288 int nlen = cnp->cn_namelen; 289 u_int fd, fd1; 290 int error; 291 struct vnode *fvp; 292 293 if ((cnp->cn_flags & ISLASTCN) && 294 (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) { 295 error = EROFS; 296 goto bad; 297 } 298 299 if (cnp->cn_namelen == 1 && *pname == '.') { 300 *vpp = dvp; 301 VREF(dvp); 302 return (0); 303 } 304 305 if (VTOFDESC(dvp)->fd_type != Froot) { 306 error = ENOTDIR; 307 goto bad; 308 } 309 310 fd = 0; 311 /* the only time a leading 0 is acceptable is if it's "0" */ 312 if (*pname == '0' && nlen != 1) { 313 error = ENOENT; 314 goto bad; 315 } 316 while (nlen--) { 317 if (*pname < '0' || *pname > '9') { 318 error = ENOENT; 319 goto bad; 320 } 321 fd1 = 10 * fd + *pname++ - '0'; 322 if (fd1 < fd) { 323 error = ENOENT; 324 goto bad; 325 } 326 fd = fd1; 327 } 328 329 /* 330 * No rights to check since 'fp' isn't actually used. 331 */ 332 if ((error = fget(td, fd, &cap_no_rights, &fp)) != 0) 333 goto bad; 334 335 /* Check if we're looking up ourselves. */ 336 if (VTOFDESC(dvp)->fd_ix == FD_DESC + fd) { 337 /* 338 * In case we're holding the last reference to the file, the dvp 339 * will be re-acquired. 340 */ 341 vhold(dvp); 342 VOP_UNLOCK(dvp); 343 fdrop(fp, td); 344 345 /* Re-aquire the lock afterwards. */ 346 vn_lock(dvp, LK_RETRY | LK_EXCLUSIVE); 347 vdrop(dvp); 348 fvp = dvp; 349 if (VN_IS_DOOMED(dvp)) 350 error = ENOENT; 351 } else { 352 /* 353 * Unlock our root node (dvp) when doing this, since we might 354 * deadlock since the vnode might be locked by another thread 355 * and the root vnode lock will be obtained afterwards (in case 356 * we're looking up the fd of the root vnode), which will be the 357 * opposite lock order. Vhold the root vnode first so we don't 358 * lose it. 359 */ 360 arg.ftype = Fdesc; 361 arg.fd_fd = fd; 362 arg.ix = FD_DESC + fd; 363 arg.fp = fp; 364 arg.td = td; 365 error = vn_vget_ino_gen(dvp, fdesc_get_ino_alloc, &arg, 366 LK_EXCLUSIVE, &fvp); 367 } 368 369 if (error) 370 goto bad; 371 *vpp = fvp; 372 return (0); 373 374 bad: 375 *vpp = NULL; 376 return (error); 377 } 378 379 static int 380 fdesc_open(struct vop_open_args *ap) 381 { 382 struct vnode *vp = ap->a_vp; 383 384 if (VTOFDESC(vp)->fd_type == Froot) 385 return (0); 386 387 /* 388 * XXX Kludge: set td->td_proc->p_dupfd to contain the value of the file 389 * descriptor being sought for duplication. The error return ensures 390 * that the vnode for this device will be released by vn_open. Open 391 * will detect this special error and take the actions in dupfdopen. 392 * Other callers of vn_open or VOP_OPEN will simply report the 393 * error. 394 */ 395 ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd; /* XXX */ 396 return (ENODEV); 397 } 398 399 static int 400 fdesc_pathconf(struct vop_pathconf_args *ap) 401 { 402 struct vnode *vp = ap->a_vp; 403 int error; 404 405 switch (ap->a_name) { 406 case _PC_NAME_MAX: 407 *ap->a_retval = NAME_MAX; 408 return (0); 409 case _PC_LINK_MAX: 410 if (VTOFDESC(vp)->fd_type == Froot) 411 *ap->a_retval = 2; 412 else 413 *ap->a_retval = 1; 414 return (0); 415 default: 416 if (VTOFDESC(vp)->fd_type == Froot) 417 return (vop_stdpathconf(ap)); 418 vref(vp); 419 VOP_UNLOCK(vp); 420 error = kern_fpathconf(curthread, VTOFDESC(vp)->fd_fd, 421 ap->a_name, ap->a_retval); 422 vn_lock(vp, LK_SHARED | LK_RETRY); 423 vunref(vp); 424 return (error); 425 } 426 } 427 428 static int 429 fdesc_getattr(struct vop_getattr_args *ap) 430 { 431 struct vnode *vp = ap->a_vp; 432 struct vattr *vap = ap->a_vap; 433 struct timeval boottime; 434 435 getboottime(&boottime); 436 vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH; 437 vap->va_fileid = VTOFDESC(vp)->fd_ix; 438 vap->va_uid = 0; 439 vap->va_gid = 0; 440 vap->va_blocksize = DEV_BSIZE; 441 vap->va_atime.tv_sec = boottime.tv_sec; 442 vap->va_atime.tv_nsec = 0; 443 vap->va_mtime = vap->va_atime; 444 vap->va_ctime = vap->va_mtime; 445 vap->va_gen = 0; 446 vap->va_flags = 0; 447 vap->va_bytes = 0; 448 vap->va_filerev = 0; 449 450 switch (VTOFDESC(vp)->fd_type) { 451 case Froot: 452 vap->va_type = VDIR; 453 vap->va_nlink = 2; 454 vap->va_size = DEV_BSIZE; 455 vap->va_rdev = NODEV; 456 break; 457 458 case Fdesc: 459 vap->va_type = (vp->v_vflag & VV_READLINK) == 0 ? VCHR : VLNK; 460 vap->va_nlink = 1; 461 vap->va_size = 0; 462 vap->va_rdev = makedev(0, vap->va_fileid); 463 break; 464 465 default: 466 panic("fdesc_getattr"); 467 break; 468 } 469 470 vp->v_type = vap->va_type; 471 return (0); 472 } 473 474 static int 475 fdesc_setattr(struct vop_setattr_args *ap) 476 { 477 struct vattr *vap = ap->a_vap; 478 struct vnode *vp; 479 struct mount *mp; 480 struct file *fp; 481 struct thread *td = curthread; 482 cap_rights_t rights; 483 unsigned fd; 484 int error; 485 486 /* 487 * Can't mess with the root vnode 488 */ 489 if (VTOFDESC(ap->a_vp)->fd_type == Froot) 490 return (EACCES); 491 492 fd = VTOFDESC(ap->a_vp)->fd_fd; 493 494 /* 495 * Allow setattr where there is an underlying vnode. 496 * For O_PATH descriptors, disallow truncate. 497 */ 498 if (vap->va_size != VNOVAL) { 499 error = getvnode(td, fd, 500 cap_rights_init_one(&rights, CAP_EXTATTR_SET), &fp); 501 } else { 502 error = getvnode_path(td, fd, 503 cap_rights_init_one(&rights, CAP_EXTATTR_SET), &fp); 504 } 505 if (error) { 506 /* 507 * getvnode() returns EINVAL if the file descriptor is not 508 * backed by a vnode. Silently drop all changes except 509 * chflags(2) in this case. 510 */ 511 if (error == EINVAL) { 512 if (vap->va_flags != VNOVAL) 513 error = EOPNOTSUPP; 514 else 515 error = 0; 516 } 517 return (error); 518 } 519 vp = fp->f_vnode; 520 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) == 0) { 521 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 522 error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred); 523 VOP_UNLOCK(vp); 524 vn_finished_write(mp); 525 } 526 fdrop(fp, td); 527 return (error); 528 } 529 530 #define UIO_MX _GENERIC_DIRLEN(10) /* number of symbols in INT_MAX printout */ 531 532 static int 533 fdesc_readdir(struct vop_readdir_args *ap) 534 { 535 struct fdescmount *fmp; 536 struct uio *uio = ap->a_uio; 537 struct filedesc *fdp; 538 struct dirent d; 539 struct dirent *dp = &d; 540 int error, i, off, fcnt; 541 542 if (VTOFDESC(ap->a_vp)->fd_type != Froot) 543 panic("fdesc_readdir: not dir"); 544 545 fmp = VFSTOFDESC(ap->a_vp->v_mount); 546 if (ap->a_ncookies != NULL) 547 *ap->a_ncookies = 0; 548 549 off = (int)uio->uio_offset; 550 if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 || 551 uio->uio_resid < UIO_MX) 552 return (EINVAL); 553 i = (u_int)off / UIO_MX; 554 fdp = uio->uio_td->td_proc->p_fd; 555 error = 0; 556 557 fcnt = i - 2; /* The first two nodes are `.' and `..' */ 558 559 FILEDESC_SLOCK(fdp); 560 while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) { 561 bzero((caddr_t)dp, UIO_MX); 562 switch (i) { 563 case 0: /* `.' */ 564 case 1: /* `..' */ 565 dp->d_fileno = i + FD_ROOT; 566 dp->d_namlen = i + 1; 567 dp->d_reclen = UIO_MX; 568 bcopy("..", dp->d_name, dp->d_namlen); 569 dp->d_type = DT_DIR; 570 dirent_terminate(dp); 571 break; 572 default: 573 if (fdp->fd_ofiles[fcnt].fde_file == NULL) 574 break; 575 dp->d_namlen = sprintf(dp->d_name, "%d", fcnt); 576 dp->d_reclen = UIO_MX; 577 dp->d_type = (fmp->flags & FMNT_LINRDLNKF) == 0 ? 578 DT_CHR : DT_LNK; 579 dp->d_fileno = i + FD_DESC; 580 dirent_terminate(dp); 581 break; 582 } 583 /* NOTE: d_off is the offset of the *next* entry. */ 584 dp->d_off = UIO_MX * (i + 1); 585 if (dp->d_namlen != 0) { 586 /* 587 * And ship to userland 588 */ 589 FILEDESC_SUNLOCK(fdp); 590 error = uiomove(dp, UIO_MX, uio); 591 if (error) 592 goto done; 593 FILEDESC_SLOCK(fdp); 594 } 595 i++; 596 fcnt++; 597 } 598 FILEDESC_SUNLOCK(fdp); 599 600 done: 601 uio->uio_offset = i * UIO_MX; 602 return (error); 603 } 604 605 static int 606 fdesc_reclaim(struct vop_reclaim_args *ap) 607 { 608 struct vnode *vp; 609 struct fdescnode *fd; 610 611 vp = ap->a_vp; 612 fd = VTOFDESC(vp); 613 fdesc_remove_entry(fd); 614 free(vp->v_data, M_TEMP); 615 vp->v_data = NULL; 616 return (0); 617 } 618 619 static int 620 fdesc_readlink(struct vop_readlink_args *va) 621 { 622 struct vnode *vp, *vn; 623 struct thread *td; 624 struct uio *uio; 625 struct file *fp; 626 char *freepath, *fullpath; 627 size_t pathlen; 628 int lockflags, fd_fd; 629 int error; 630 631 freepath = NULL; 632 vn = va->a_vp; 633 if (VTOFDESC(vn)->fd_type != Fdesc) 634 panic("fdesc_readlink: not fdescfs link"); 635 fd_fd = ((struct fdescnode *)vn->v_data)->fd_fd; 636 lockflags = VOP_ISLOCKED(vn); 637 VOP_UNLOCK(vn); 638 639 td = curthread; 640 error = fget_cap(td, fd_fd, &cap_no_rights, &fp, NULL); 641 if (error != 0) 642 goto out; 643 644 switch (fp->f_type) { 645 case DTYPE_VNODE: 646 vp = fp->f_vnode; 647 error = vn_fullpath(vp, &fullpath, &freepath); 648 break; 649 default: 650 fullpath = "anon_inode:[unknown]"; 651 break; 652 } 653 if (error == 0) { 654 uio = va->a_uio; 655 pathlen = strlen(fullpath); 656 error = uiomove(fullpath, pathlen, uio); 657 } 658 if (freepath != NULL) 659 free(freepath, M_TEMP); 660 fdrop(fp, td); 661 662 out: 663 vn_lock(vn, lockflags | LK_RETRY); 664 return (error); 665 } 666