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. 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/capsicum.h> 44 #include <sys/conf.h> 45 #include <sys/dirent.h> 46 #include <sys/filedesc.h> 47 #include <sys/kernel.h> /* boottime */ 48 #include <sys/lock.h> 49 #include <sys/mutex.h> 50 #include <sys/malloc.h> 51 #include <sys/file.h> /* Must come after sys/malloc.h */ 52 #include <sys/mount.h> 53 #include <sys/namei.h> 54 #include <sys/proc.h> 55 #include <sys/stat.h> 56 #include <sys/vnode.h> 57 58 #include <fs/fdescfs/fdesc.h> 59 60 #define NFDCACHE 4 61 #define FD_NHASH(ix) \ 62 (&fdhashtbl[(ix) & fdhash]) 63 static LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl; 64 static u_long fdhash; 65 66 struct mtx fdesc_hashmtx; 67 68 static vop_getattr_t fdesc_getattr; 69 static vop_lookup_t fdesc_lookup; 70 static vop_open_t fdesc_open; 71 static vop_readdir_t fdesc_readdir; 72 static vop_reclaim_t fdesc_reclaim; 73 static vop_setattr_t fdesc_setattr; 74 75 static struct vop_vector fdesc_vnodeops = { 76 .vop_default = &default_vnodeops, 77 78 .vop_access = VOP_NULL, 79 .vop_getattr = fdesc_getattr, 80 .vop_lookup = fdesc_lookup, 81 .vop_open = fdesc_open, 82 .vop_pathconf = vop_stdpathconf, 83 .vop_readdir = fdesc_readdir, 84 .vop_reclaim = fdesc_reclaim, 85 .vop_setattr = fdesc_setattr, 86 }; 87 88 static void fdesc_insmntque_dtr(struct vnode *, void *); 89 static void fdesc_remove_entry(struct fdescnode *); 90 91 /* 92 * Initialise cache headers 93 */ 94 int 95 fdesc_init(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(struct vfsconf *vfsp) 108 { 109 110 hashdestroy(fdhashtbl, M_CACHE, fdhash); 111 mtx_destroy(&fdesc_hashmtx); 112 return (0); 113 } 114 115 /* 116 * If allocating vnode fails, call this. 117 */ 118 static void 119 fdesc_insmntque_dtr(struct vnode *vp, void *arg) 120 { 121 122 vgone(vp); 123 vput(vp); 124 } 125 126 /* 127 * Remove an entry from the hash if it exists. 128 */ 129 static void 130 fdesc_remove_entry(struct fdescnode *fd) 131 { 132 struct fdhashhead *fc; 133 struct fdescnode *fd2; 134 135 fc = FD_NHASH(fd->fd_ix); 136 mtx_lock(&fdesc_hashmtx); 137 LIST_FOREACH(fd2, fc, fd_hash) { 138 if (fd == fd2) { 139 LIST_REMOVE(fd, fd_hash); 140 break; 141 } 142 } 143 mtx_unlock(&fdesc_hashmtx); 144 } 145 146 int 147 fdesc_allocvp(fdntype ftype, unsigned fd_fd, int ix, struct mount *mp, 148 struct vnode **vpp) 149 { 150 struct fdescmount *fmp; 151 struct fdhashhead *fc; 152 struct fdescnode *fd, *fd2; 153 struct vnode *vp, *vp2; 154 struct thread *td; 155 int error = 0; 156 157 td = curthread; 158 fc = FD_NHASH(ix); 159 loop: 160 mtx_lock(&fdesc_hashmtx); 161 /* 162 * If a forced unmount is progressing, we need to drop it. The flags are 163 * protected by the hashmtx. 164 */ 165 fmp = (struct fdescmount *)mp->mnt_data; 166 if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) { 167 mtx_unlock(&fdesc_hashmtx); 168 return (-1); 169 } 170 171 LIST_FOREACH(fd, fc, fd_hash) { 172 if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) { 173 /* Get reference to vnode in case it's being free'd */ 174 vp = fd->fd_vnode; 175 VI_LOCK(vp); 176 mtx_unlock(&fdesc_hashmtx); 177 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) 178 goto loop; 179 *vpp = vp; 180 return (0); 181 } 182 } 183 mtx_unlock(&fdesc_hashmtx); 184 185 fd = malloc(sizeof(struct fdescnode), M_TEMP, M_WAITOK); 186 187 error = getnewvnode("fdescfs", mp, &fdesc_vnodeops, &vp); 188 if (error) { 189 free(fd, M_TEMP); 190 return (error); 191 } 192 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 193 vp->v_data = fd; 194 fd->fd_vnode = vp; 195 fd->fd_type = ftype; 196 fd->fd_fd = fd_fd; 197 fd->fd_ix = ix; 198 error = insmntque1(vp, mp, fdesc_insmntque_dtr, NULL); 199 if (error != 0) { 200 *vpp = NULLVP; 201 return (error); 202 } 203 204 /* Make sure that someone didn't beat us when inserting the vnode. */ 205 mtx_lock(&fdesc_hashmtx); 206 /* 207 * If a forced unmount is progressing, we need to drop it. The flags are 208 * protected by the hashmtx. 209 */ 210 fmp = (struct fdescmount *)mp->mnt_data; 211 if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) { 212 mtx_unlock(&fdesc_hashmtx); 213 vgone(vp); 214 vput(vp); 215 *vpp = NULLVP; 216 return (-1); 217 } 218 219 LIST_FOREACH(fd2, fc, fd_hash) { 220 if (fd2->fd_ix == ix && fd2->fd_vnode->v_mount == mp) { 221 /* Get reference to vnode in case it's being free'd */ 222 vp2 = fd2->fd_vnode; 223 VI_LOCK(vp2); 224 mtx_unlock(&fdesc_hashmtx); 225 error = vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK, td); 226 /* Someone beat us, dec use count and wait for reclaim */ 227 vgone(vp); 228 vput(vp); 229 /* If we didn't get it, return no vnode. */ 230 if (error) 231 vp2 = NULLVP; 232 *vpp = vp2; 233 return (error); 234 } 235 } 236 237 /* If we came here, we can insert it safely. */ 238 LIST_INSERT_HEAD(fc, fd, fd_hash); 239 mtx_unlock(&fdesc_hashmtx); 240 *vpp = vp; 241 return (0); 242 } 243 244 struct fdesc_get_ino_args { 245 fdntype ftype; 246 unsigned fd_fd; 247 int ix; 248 struct file *fp; 249 struct thread *td; 250 }; 251 252 static int 253 fdesc_get_ino_alloc(struct mount *mp, void *arg, int lkflags, 254 struct vnode **rvp) 255 { 256 struct fdesc_get_ino_args *a; 257 int error; 258 259 a = arg; 260 error = fdesc_allocvp(a->ftype, a->fd_fd, a->ix, mp, rvp); 261 fdrop(a->fp, a->td); 262 return (error); 263 } 264 265 266 /* 267 * vp is the current namei directory 268 * ndp is the name to locate in that directory... 269 */ 270 static int 271 fdesc_lookup(struct vop_lookup_args *ap) 272 { 273 struct vnode **vpp = ap->a_vpp; 274 struct vnode *dvp = ap->a_dvp; 275 struct componentname *cnp = ap->a_cnp; 276 char *pname = cnp->cn_nameptr; 277 struct thread *td = cnp->cn_thread; 278 struct file *fp; 279 struct fdesc_get_ino_args arg; 280 cap_rights_t rights; 281 int nlen = cnp->cn_namelen; 282 u_int fd, fd1; 283 int error; 284 struct vnode *fvp; 285 286 if ((cnp->cn_flags & ISLASTCN) && 287 (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) { 288 error = EROFS; 289 goto bad; 290 } 291 292 if (cnp->cn_namelen == 1 && *pname == '.') { 293 *vpp = dvp; 294 VREF(dvp); 295 return (0); 296 } 297 298 if (VTOFDESC(dvp)->fd_type != Froot) { 299 error = ENOTDIR; 300 goto bad; 301 } 302 303 fd = 0; 304 /* the only time a leading 0 is acceptable is if it's "0" */ 305 if (*pname == '0' && nlen != 1) { 306 error = ENOENT; 307 goto bad; 308 } 309 while (nlen--) { 310 if (*pname < '0' || *pname > '9') { 311 error = ENOENT; 312 goto bad; 313 } 314 fd1 = 10 * fd + *pname++ - '0'; 315 if (fd1 < fd) { 316 error = ENOENT; 317 goto bad; 318 } 319 fd = fd1; 320 } 321 322 /* 323 * No rights to check since 'fp' isn't actually used. 324 */ 325 if ((error = fget(td, fd, cap_rights_init(&rights), &fp)) != 0) 326 goto bad; 327 328 /* Check if we're looking up ourselves. */ 329 if (VTOFDESC(dvp)->fd_ix == FD_DESC + fd) { 330 /* 331 * In case we're holding the last reference to the file, the dvp 332 * will be re-acquired. 333 */ 334 vhold(dvp); 335 VOP_UNLOCK(dvp, 0); 336 fdrop(fp, td); 337 338 /* Re-aquire the lock afterwards. */ 339 vn_lock(dvp, LK_RETRY | LK_EXCLUSIVE); 340 vdrop(dvp); 341 fvp = dvp; 342 if ((dvp->v_iflag & VI_DOOMED) != 0) 343 error = ENOENT; 344 } else { 345 /* 346 * Unlock our root node (dvp) when doing this, since we might 347 * deadlock since the vnode might be locked by another thread 348 * and the root vnode lock will be obtained afterwards (in case 349 * we're looking up the fd of the root vnode), which will be the 350 * opposite lock order. Vhold the root vnode first so we don't 351 * lose it. 352 */ 353 arg.ftype = Fdesc; 354 arg.fd_fd = fd; 355 arg.ix = FD_DESC + fd; 356 arg.fp = fp; 357 arg.td = td; 358 error = vn_vget_ino_gen(dvp, fdesc_get_ino_alloc, &arg, 359 LK_EXCLUSIVE, &fvp); 360 } 361 362 if (error) 363 goto bad; 364 *vpp = fvp; 365 return (0); 366 367 bad: 368 *vpp = NULL; 369 return (error); 370 } 371 372 static int 373 fdesc_open(struct vop_open_args *ap) 374 { 375 struct vnode *vp = ap->a_vp; 376 377 if (VTOFDESC(vp)->fd_type == Froot) 378 return (0); 379 380 /* 381 * XXX Kludge: set td->td_proc->p_dupfd to contain the value of the file 382 * descriptor being sought for duplication. The error return ensures 383 * that the vnode for this device will be released by vn_open. Open 384 * will detect this special error and take the actions in dupfdopen. 385 * Other callers of vn_open or VOP_OPEN will simply report the 386 * error. 387 */ 388 ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd; /* XXX */ 389 return (ENODEV); 390 } 391 392 static int 393 fdesc_getattr(struct vop_getattr_args *ap) 394 { 395 struct vnode *vp = ap->a_vp; 396 struct vattr *vap = ap->a_vap; 397 struct timeval boottime; 398 399 getboottime(&boottime); 400 vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH; 401 vap->va_fileid = VTOFDESC(vp)->fd_ix; 402 vap->va_uid = 0; 403 vap->va_gid = 0; 404 vap->va_blocksize = DEV_BSIZE; 405 vap->va_atime.tv_sec = boottime.tv_sec; 406 vap->va_atime.tv_nsec = 0; 407 vap->va_mtime = vap->va_atime; 408 vap->va_ctime = vap->va_mtime; 409 vap->va_gen = 0; 410 vap->va_flags = 0; 411 vap->va_bytes = 0; 412 vap->va_filerev = 0; 413 414 switch (VTOFDESC(vp)->fd_type) { 415 case Froot: 416 vap->va_type = VDIR; 417 vap->va_nlink = 2; 418 vap->va_size = DEV_BSIZE; 419 vap->va_rdev = NODEV; 420 break; 421 422 case Fdesc: 423 vap->va_type = VCHR; 424 vap->va_nlink = 1; 425 vap->va_size = 0; 426 vap->va_rdev = makedev(0, vap->va_fileid); 427 break; 428 429 default: 430 panic("fdesc_getattr"); 431 break; 432 } 433 434 vp->v_type = vap->va_type; 435 return (0); 436 } 437 438 static int 439 fdesc_setattr(struct vop_setattr_args *ap) 440 { 441 struct vattr *vap = ap->a_vap; 442 struct vnode *vp; 443 struct mount *mp; 444 struct file *fp; 445 struct thread *td = curthread; 446 cap_rights_t rights; 447 unsigned fd; 448 int error; 449 450 /* 451 * Can't mess with the root vnode 452 */ 453 if (VTOFDESC(ap->a_vp)->fd_type == Froot) 454 return (EACCES); 455 456 fd = VTOFDESC(ap->a_vp)->fd_fd; 457 458 /* 459 * Allow setattr where there is an underlying vnode. 460 */ 461 error = getvnode(td, fd, 462 cap_rights_init(&rights, CAP_EXTATTR_SET), &fp); 463 if (error) { 464 /* 465 * getvnode() returns EINVAL if the file descriptor is not 466 * backed by a vnode. Silently drop all changes except 467 * chflags(2) in this case. 468 */ 469 if (error == EINVAL) { 470 if (vap->va_flags != VNOVAL) 471 error = EOPNOTSUPP; 472 else 473 error = 0; 474 } 475 return (error); 476 } 477 vp = fp->f_vnode; 478 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) == 0) { 479 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 480 error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred); 481 VOP_UNLOCK(vp, 0); 482 vn_finished_write(mp); 483 } 484 fdrop(fp, td); 485 return (error); 486 } 487 488 #define UIO_MX 16 489 490 static int 491 fdesc_readdir(struct vop_readdir_args *ap) 492 { 493 struct uio *uio = ap->a_uio; 494 struct filedesc *fdp; 495 struct dirent d; 496 struct dirent *dp = &d; 497 int error, i, off, fcnt; 498 499 if (VTOFDESC(ap->a_vp)->fd_type != Froot) 500 panic("fdesc_readdir: not dir"); 501 502 if (ap->a_ncookies != NULL) 503 *ap->a_ncookies = 0; 504 505 off = (int)uio->uio_offset; 506 if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 || 507 uio->uio_resid < UIO_MX) 508 return (EINVAL); 509 i = (u_int)off / UIO_MX; 510 fdp = uio->uio_td->td_proc->p_fd; 511 error = 0; 512 513 fcnt = i - 2; /* The first two nodes are `.' and `..' */ 514 515 FILEDESC_SLOCK(fdp); 516 while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) { 517 bzero((caddr_t)dp, UIO_MX); 518 switch (i) { 519 case 0: /* `.' */ 520 case 1: /* `..' */ 521 dp->d_fileno = i + FD_ROOT; 522 dp->d_namlen = i + 1; 523 dp->d_reclen = UIO_MX; 524 bcopy("..", dp->d_name, dp->d_namlen); 525 dp->d_name[i + 1] = '\0'; 526 dp->d_type = DT_DIR; 527 break; 528 default: 529 if (fdp->fd_ofiles[fcnt].fde_file == NULL) 530 break; 531 dp->d_namlen = sprintf(dp->d_name, "%d", fcnt); 532 dp->d_reclen = UIO_MX; 533 dp->d_type = DT_CHR; 534 dp->d_fileno = i + FD_DESC; 535 break; 536 } 537 if (dp->d_namlen != 0) { 538 /* 539 * And ship to userland 540 */ 541 FILEDESC_SUNLOCK(fdp); 542 error = uiomove(dp, UIO_MX, uio); 543 if (error) 544 goto done; 545 FILEDESC_SLOCK(fdp); 546 } 547 i++; 548 fcnt++; 549 } 550 FILEDESC_SUNLOCK(fdp); 551 552 done: 553 uio->uio_offset = i * UIO_MX; 554 return (error); 555 } 556 557 static int 558 fdesc_reclaim(struct vop_reclaim_args *ap) 559 { 560 struct vnode *vp; 561 struct fdescnode *fd; 562 563 vp = ap->a_vp; 564 fd = VTOFDESC(vp); 565 fdesc_remove_entry(fd); 566 free(vp->v_data, M_TEMP); 567 vp->v_data = NULL; 568 return (0); 569 } 570