1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed 8 * to Berkeley by John Heidemann of the UCLA Ficus project. 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 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/bio.h> 38 #include <sys/buf.h> 39 #include <sys/conf.h> 40 #include <sys/event.h> 41 #include <sys/filio.h> 42 #include <sys/kernel.h> 43 #include <sys/limits.h> 44 #include <sys/lock.h> 45 #include <sys/lockf.h> 46 #include <sys/malloc.h> 47 #include <sys/mount.h> 48 #include <sys/namei.h> 49 #include <sys/rwlock.h> 50 #include <sys/fcntl.h> 51 #include <sys/unistd.h> 52 #include <sys/vnode.h> 53 #include <sys/dirent.h> 54 #include <sys/poll.h> 55 #include <sys/stat.h> 56 #include <security/audit/audit.h> 57 #include <sys/priv.h> 58 59 #include <security/mac/mac_framework.h> 60 61 #include <vm/vm.h> 62 #include <vm/vm_object.h> 63 #include <vm/vm_extern.h> 64 #include <vm/pmap.h> 65 #include <vm/vm_map.h> 66 #include <vm/vm_page.h> 67 #include <vm/vm_pager.h> 68 #include <vm/vnode_pager.h> 69 70 static int vop_nolookup(struct vop_lookup_args *); 71 static int vop_norename(struct vop_rename_args *); 72 static int vop_nostrategy(struct vop_strategy_args *); 73 static int dirent_exists(struct vnode *vp, const char *dirname, 74 struct thread *td); 75 76 static int vop_stdis_text(struct vop_is_text_args *ap); 77 static int vop_stdunset_text(struct vop_unset_text_args *ap); 78 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap); 79 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap); 80 static int vop_stdfdatasync(struct vop_fdatasync_args *ap); 81 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap); 82 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap); 83 static int vop_stdstat(struct vop_stat_args *ap); 84 static int vop_stdvput_pair(struct vop_vput_pair_args *ap); 85 static int vop_stdgetlowvnode(struct vop_getlowvnode_args *ap); 86 87 /* 88 * This vnode table stores what we want to do if the filesystem doesn't 89 * implement a particular VOP. 90 * 91 * If there is no specific entry here, we will return EOPNOTSUPP. 92 * 93 * Note that every filesystem has to implement either vop_access 94 * or vop_accessx; failing to do so will result in immediate crash 95 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(), 96 * which calls vop_stdaccess() etc. 97 */ 98 99 struct vop_vector default_vnodeops = { 100 .vop_default = NULL, 101 .vop_bypass = VOP_EOPNOTSUPP, 102 103 .vop_access = vop_stdaccess, 104 .vop_accessx = vop_stdaccessx, 105 .vop_advise = vop_stdadvise, 106 .vop_advlock = vop_stdadvlock, 107 .vop_advlockasync = vop_stdadvlockasync, 108 .vop_advlockpurge = vop_stdadvlockpurge, 109 .vop_allocate = vop_stdallocate, 110 .vop_deallocate = vop_stddeallocate, 111 .vop_bmap = vop_stdbmap, 112 .vop_close = VOP_NULL, 113 .vop_fsync = VOP_NULL, 114 .vop_stat = vop_stdstat, 115 .vop_fdatasync = vop_stdfdatasync, 116 .vop_getlowvnode = vop_stdgetlowvnode, 117 .vop_getpages = vop_stdgetpages, 118 .vop_getpages_async = vop_stdgetpages_async, 119 .vop_getwritemount = vop_stdgetwritemount, 120 .vop_inactive = VOP_NULL, 121 .vop_need_inactive = vop_stdneed_inactive, 122 .vop_ioctl = vop_stdioctl, 123 .vop_kqfilter = vop_stdkqfilter, 124 .vop_islocked = vop_stdislocked, 125 .vop_lock1 = vop_stdlock, 126 .vop_lookup = vop_nolookup, 127 .vop_open = VOP_NULL, 128 .vop_pathconf = VOP_EINVAL, 129 .vop_poll = vop_nopoll, 130 .vop_putpages = vop_stdputpages, 131 .vop_readlink = VOP_EINVAL, 132 .vop_read_pgcache = vop_stdread_pgcache, 133 .vop_rename = vop_norename, 134 .vop_revoke = VOP_PANIC, 135 .vop_strategy = vop_nostrategy, 136 .vop_unlock = vop_stdunlock, 137 .vop_vptocnp = vop_stdvptocnp, 138 .vop_vptofh = vop_stdvptofh, 139 .vop_unp_bind = vop_stdunp_bind, 140 .vop_unp_connect = vop_stdunp_connect, 141 .vop_unp_detach = vop_stdunp_detach, 142 .vop_is_text = vop_stdis_text, 143 .vop_set_text = vop_stdset_text, 144 .vop_unset_text = vop_stdunset_text, 145 .vop_add_writecount = vop_stdadd_writecount, 146 .vop_copy_file_range = vop_stdcopy_file_range, 147 .vop_vput_pair = vop_stdvput_pair, 148 }; 149 VFS_VOP_VECTOR_REGISTER(default_vnodeops); 150 151 /* 152 * Series of placeholder functions for various error returns for 153 * VOPs. 154 */ 155 156 int 157 vop_eopnotsupp(struct vop_generic_args *ap) 158 { 159 /* 160 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 161 */ 162 163 return (EOPNOTSUPP); 164 } 165 166 int 167 vop_ebadf(struct vop_generic_args *ap) 168 { 169 170 return (EBADF); 171 } 172 173 int 174 vop_enotty(struct vop_generic_args *ap) 175 { 176 177 return (ENOTTY); 178 } 179 180 int 181 vop_einval(struct vop_generic_args *ap) 182 { 183 184 return (EINVAL); 185 } 186 187 int 188 vop_enoent(struct vop_generic_args *ap) 189 { 190 191 return (ENOENT); 192 } 193 194 int 195 vop_eagain(struct vop_generic_args *ap) 196 { 197 198 return (EAGAIN); 199 } 200 201 int 202 vop_null(struct vop_generic_args *ap) 203 { 204 205 return (0); 206 } 207 208 /* 209 * Helper function to panic on some bad VOPs in some filesystems. 210 */ 211 int 212 vop_panic(struct vop_generic_args *ap) 213 { 214 215 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 216 } 217 218 /* 219 * vop_std<something> and vop_no<something> are default functions for use by 220 * filesystems that need the "default reasonable" implementation for a 221 * particular operation. 222 * 223 * The documentation for the operations they implement exists (if it exists) 224 * in the VOP_<SOMETHING>(9) manpage (all uppercase). 225 */ 226 227 /* 228 * Default vop for filesystems that do not support name lookup 229 */ 230 static int 231 vop_nolookup(struct vop_lookup_args *ap) 232 { 233 234 *ap->a_vpp = NULL; 235 return (ENOTDIR); 236 } 237 238 /* 239 * vop_norename: 240 * 241 * Handle unlock and reference counting for arguments of vop_rename 242 * for filesystems that do not implement rename operation. 243 */ 244 static int 245 vop_norename(struct vop_rename_args *ap) 246 { 247 248 vop_rename_fail(ap); 249 return (EOPNOTSUPP); 250 } 251 252 /* 253 * vop_nostrategy: 254 * 255 * Strategy routine for VFS devices that have none. 256 * 257 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 258 * routine. Typically this is done for a BIO_READ strategy call. 259 * Typically B_INVAL is assumed to already be clear prior to a write 260 * and should not be cleared manually unless you just made the buffer 261 * invalid. BIO_ERROR should be cleared either way. 262 */ 263 264 static int 265 vop_nostrategy (struct vop_strategy_args *ap) 266 { 267 printf("No strategy for buffer at %p\n", ap->a_bp); 268 vn_printf(ap->a_vp, "vnode "); 269 ap->a_bp->b_ioflags |= BIO_ERROR; 270 ap->a_bp->b_error = EOPNOTSUPP; 271 bufdone(ap->a_bp); 272 return (EOPNOTSUPP); 273 } 274 275 /* 276 * Check if a named file exists in a given directory vnode 277 * 278 * Returns 0 if the file exists, ENOENT if it doesn't, or errors returned by 279 * vn_dir_next_dirent(). 280 */ 281 static int 282 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td) 283 { 284 char *dirbuf; 285 int error, eofflag; 286 size_t dirbuflen, len; 287 off_t off; 288 struct dirent *dp; 289 struct vattr va; 290 291 ASSERT_VOP_LOCKED(vp, "vnode not locked"); 292 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp)); 293 294 error = VOP_GETATTR(vp, &va, td->td_ucred); 295 if (error != 0) 296 return (error); 297 298 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ); 299 if (dirbuflen < va.va_blocksize) 300 dirbuflen = va.va_blocksize; 301 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK); 302 303 len = 0; 304 off = 0; 305 eofflag = 0; 306 307 for (;;) { 308 error = vn_dir_next_dirent(vp, td, dirbuf, dirbuflen, 309 &dp, &len, &off, &eofflag); 310 if (error != 0) 311 goto out; 312 313 if (len == 0) 314 break; 315 316 if (dp->d_type != DT_WHT && dp->d_fileno != 0 && 317 strcmp(dp->d_name, dirname) == 0) 318 goto out; 319 } 320 321 error = ENOENT; 322 323 out: 324 free(dirbuf, M_TEMP); 325 return (error); 326 } 327 328 int 329 vop_stdaccess(struct vop_access_args *ap) 330 { 331 332 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN | 333 VAPPEND)) == 0, ("invalid bit in accmode")); 334 335 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td)); 336 } 337 338 int 339 vop_stdaccessx(struct vop_accessx_args *ap) 340 { 341 int error; 342 accmode_t accmode = ap->a_accmode; 343 344 error = vfs_unixify_accmode(&accmode); 345 if (error != 0) 346 return (error); 347 348 if (accmode == 0) 349 return (0); 350 351 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td)); 352 } 353 354 /* 355 * Advisory record locking support 356 */ 357 int 358 vop_stdadvlock(struct vop_advlock_args *ap) 359 { 360 struct vnode *vp; 361 struct mount *mp; 362 struct vattr vattr; 363 int error; 364 365 vp = ap->a_vp; 366 367 /* 368 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for 369 * local filesystems. See vn_open_cred() for reciprocal part. 370 */ 371 mp = vp->v_mount; 372 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 && 373 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) { 374 VI_LOCK(vp); 375 while ((vp->v_iflag & VI_FOPENING) != 0) 376 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0); 377 VI_UNLOCK(vp); 378 } 379 380 if (ap->a_fl->l_whence == SEEK_END) { 381 /* 382 * The NFSv4 server must avoid doing a vn_lock() here, since it 383 * can deadlock the nfsd threads, due to a LOR. Fortunately 384 * the NFSv4 server always uses SEEK_SET and this code is 385 * only required for the SEEK_END case. 386 */ 387 vn_lock(vp, LK_SHARED | LK_RETRY); 388 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred); 389 VOP_UNLOCK(vp); 390 if (error) 391 return (error); 392 } else 393 vattr.va_size = 0; 394 395 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size)); 396 } 397 398 int 399 vop_stdadvlockasync(struct vop_advlockasync_args *ap) 400 { 401 struct vnode *vp; 402 struct vattr vattr; 403 int error; 404 405 vp = ap->a_vp; 406 if (ap->a_fl->l_whence == SEEK_END) { 407 /* The size argument is only needed for SEEK_END. */ 408 vn_lock(vp, LK_SHARED | LK_RETRY); 409 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred); 410 VOP_UNLOCK(vp); 411 if (error) 412 return (error); 413 } else 414 vattr.va_size = 0; 415 416 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size)); 417 } 418 419 int 420 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap) 421 { 422 struct vnode *vp; 423 424 vp = ap->a_vp; 425 lf_purgelocks(vp, &vp->v_lockf); 426 return (0); 427 } 428 429 /* 430 * vop_stdpathconf: 431 * 432 * Standard implementation of POSIX pathconf, to get information about limits 433 * for a filesystem. 434 * Override per filesystem for the case where the filesystem has smaller 435 * limits. 436 */ 437 int 438 vop_stdpathconf(struct vop_pathconf_args *ap) 439 { 440 441 switch (ap->a_name) { 442 case _PC_ASYNC_IO: 443 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO; 444 return (0); 445 case _PC_PATH_MAX: 446 *ap->a_retval = PATH_MAX; 447 return (0); 448 case _PC_ACL_EXTENDED: 449 case _PC_ACL_NFS4: 450 case _PC_CAP_PRESENT: 451 case _PC_DEALLOC_PRESENT: 452 case _PC_INF_PRESENT: 453 case _PC_MAC_PRESENT: 454 case _PC_NAMEDATTR_ENABLED: 455 *ap->a_retval = 0; 456 return (0); 457 default: 458 return (EINVAL); 459 } 460 /* NOTREACHED */ 461 } 462 463 /* 464 * Standard lock, unlock and islocked functions. 465 */ 466 int 467 vop_stdlock(struct vop_lock1_args *ap) 468 { 469 struct vnode *vp = ap->a_vp; 470 struct mtx *ilk; 471 472 ilk = VI_MTX(vp); 473 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags, 474 &ilk->lock_object, ap->a_file, ap->a_line)); 475 } 476 477 /* See above. */ 478 int 479 vop_stdunlock(struct vop_unlock_args *ap) 480 { 481 struct vnode *vp = ap->a_vp; 482 483 return (lockmgr_unlock(vp->v_vnlock)); 484 } 485 486 /* See above. */ 487 int 488 vop_stdislocked(struct vop_islocked_args *ap) 489 { 490 491 return (lockstatus(ap->a_vp->v_vnlock)); 492 } 493 494 /* 495 * Variants of the above set. 496 * 497 * Differences are: 498 * - shared locking disablement is not supported 499 * - v_vnlock pointer is not honored 500 */ 501 int 502 vop_lock(struct vop_lock1_args *ap) 503 { 504 struct vnode *vp = ap->a_vp; 505 int flags = ap->a_flags; 506 struct mtx *ilk; 507 508 MPASS(vp->v_vnlock == &vp->v_lock); 509 510 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0)) 511 goto other; 512 513 switch (flags & LK_TYPE_MASK) { 514 case LK_SHARED: 515 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line)); 516 case LK_EXCLUSIVE: 517 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line)); 518 } 519 other: 520 ilk = VI_MTX(vp); 521 return (lockmgr_lock_flags(&vp->v_lock, flags, 522 &ilk->lock_object, ap->a_file, ap->a_line)); 523 } 524 525 int 526 vop_unlock(struct vop_unlock_args *ap) 527 { 528 struct vnode *vp = ap->a_vp; 529 530 MPASS(vp->v_vnlock == &vp->v_lock); 531 532 return (lockmgr_unlock(&vp->v_lock)); 533 } 534 535 int 536 vop_islocked(struct vop_islocked_args *ap) 537 { 538 struct vnode *vp = ap->a_vp; 539 540 MPASS(vp->v_vnlock == &vp->v_lock); 541 542 return (lockstatus(&vp->v_lock)); 543 } 544 545 /* 546 * Return true for select/poll. 547 */ 548 int 549 vop_nopoll(struct vop_poll_args *ap) 550 { 551 552 if (ap->a_events & ~POLLSTANDARD) 553 return (POLLNVAL); 554 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 555 } 556 557 /* 558 * Implement poll for local filesystems that support it. 559 */ 560 int 561 vop_stdpoll(struct vop_poll_args *ap) 562 { 563 if (ap->a_events & ~POLLSTANDARD) 564 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 565 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 566 } 567 568 /* 569 * Return our mount point, as we will take charge of the writes. 570 */ 571 int 572 vop_stdgetwritemount(struct vop_getwritemount_args *ap) 573 { 574 struct mount *mp; 575 struct vnode *vp; 576 577 /* 578 * Note that having a reference does not prevent forced unmount from 579 * setting ->v_mount to NULL after the lock gets released. This is of 580 * no consequence for typical consumers (most notably vn_start_write) 581 * since in this case the vnode is VIRF_DOOMED. Unmount might have 582 * progressed far enough that its completion is only delayed by the 583 * reference obtained here. The consumer only needs to concern itself 584 * with releasing it. 585 */ 586 vp = ap->a_vp; 587 mp = vfs_ref_from_vp(vp); 588 *(ap->a_mpp) = mp; 589 return (0); 590 } 591 592 /* 593 * If the file system doesn't implement VOP_BMAP, then return sensible defaults: 594 * - Return the vnode's bufobj instead of any underlying device's bufobj 595 * - Calculate the physical block number as if there were equal size 596 * consecutive blocks, but 597 * - Report no contiguous runs of blocks. 598 */ 599 int 600 vop_stdbmap(struct vop_bmap_args *ap) 601 { 602 603 if (ap->a_bop != NULL) 604 *ap->a_bop = &ap->a_vp->v_bufobj; 605 if (ap->a_bnp != NULL) 606 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 607 if (ap->a_runp != NULL) 608 *ap->a_runp = 0; 609 if (ap->a_runb != NULL) 610 *ap->a_runb = 0; 611 return (0); 612 } 613 614 int 615 vop_stdfsync(struct vop_fsync_args *ap) 616 { 617 618 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor)); 619 } 620 621 static int 622 vop_stdfdatasync(struct vop_fdatasync_args *ap) 623 { 624 625 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td)); 626 } 627 628 int 629 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap) 630 { 631 632 return (vn_fsync_buf(ap->a_vp, MNT_WAIT)); 633 } 634 635 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */ 636 int 637 vop_stdgetpages(struct vop_getpages_args *ap) 638 { 639 640 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, 641 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL); 642 } 643 644 static int 645 vop_stdgetpages_async(struct vop_getpages_async_args *ap) 646 { 647 int error; 648 649 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind, 650 ap->a_rahead); 651 if (ap->a_iodone != NULL) 652 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error); 653 return (error); 654 } 655 656 int 657 vop_stdkqfilter(struct vop_kqfilter_args *ap) 658 { 659 return vfs_kqfilter(ap); 660 } 661 662 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */ 663 int 664 vop_stdputpages(struct vop_putpages_args *ap) 665 { 666 667 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count, 668 ap->a_sync, ap->a_rtvals); 669 } 670 671 int 672 vop_stdvptofh(struct vop_vptofh_args *ap) 673 { 674 return (EOPNOTSUPP); 675 } 676 677 int 678 vop_stdvptocnp(struct vop_vptocnp_args *ap) 679 { 680 struct vnode *const vp = ap->a_vp; 681 struct vnode **const dvp = ap->a_vpp; 682 char *buf = ap->a_buf; 683 size_t *buflen = ap->a_buflen; 684 char *dirbuf; 685 int i = *buflen; 686 int error = 0, covered = 0; 687 int eofflag, flags, locked; 688 size_t dirbuflen, len; 689 off_t off; 690 ino_t fileno; 691 struct vattr va; 692 struct nameidata nd; 693 struct thread *const td = curthread; 694 struct ucred *const cred = td->td_ucred; 695 struct dirent *dp; 696 struct vnode *mvp; 697 698 if (vp->v_type != VDIR) 699 return (ENOENT); 700 701 error = VOP_GETATTR(vp, &va, cred); 702 if (error) 703 return (error); 704 705 VREF(vp); 706 locked = VOP_ISLOCKED(vp); 707 VOP_UNLOCK(vp); 708 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE, 709 "..", vp); 710 flags = FREAD; 711 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL); 712 if (error) { 713 vn_lock(vp, locked | LK_RETRY); 714 return (error); 715 } 716 NDFREE_PNBUF(&nd); 717 718 mvp = *dvp = nd.ni_vp; 719 720 if (vp->v_mount != (*dvp)->v_mount && 721 ((*dvp)->v_vflag & VV_ROOT) && 722 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) { 723 *dvp = (*dvp)->v_mount->mnt_vnodecovered; 724 VREF(mvp); 725 VOP_UNLOCK(mvp); 726 vn_close(mvp, FREAD, cred, td); 727 VREF(*dvp); 728 vn_lock(*dvp, LK_SHARED | LK_RETRY); 729 covered = 1; 730 } 731 732 fileno = va.va_fileid; 733 734 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ); 735 if (dirbuflen < va.va_blocksize) 736 dirbuflen = va.va_blocksize; 737 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK); 738 739 if ((*dvp)->v_type != VDIR) { 740 error = ENOENT; 741 goto out; 742 } 743 744 len = 0; 745 off = 0; 746 eofflag = 0; 747 748 for (;;) { 749 /* call VOP_READDIR of parent */ 750 error = vn_dir_next_dirent(*dvp, td, 751 dirbuf, dirbuflen, &dp, &len, &off, &eofflag); 752 if (error != 0) 753 goto out; 754 755 if (len == 0) { 756 error = ENOENT; 757 goto out; 758 } 759 760 if ((dp->d_type != DT_WHT) && 761 (dp->d_fileno == fileno)) { 762 if (covered) { 763 VOP_UNLOCK(*dvp); 764 vn_lock(mvp, LK_SHARED | LK_RETRY); 765 if (dirent_exists(mvp, dp->d_name, td) == 0) { 766 error = ENOENT; 767 VOP_UNLOCK(mvp); 768 vn_lock(*dvp, LK_SHARED | LK_RETRY); 769 goto out; 770 } 771 VOP_UNLOCK(mvp); 772 vn_lock(*dvp, LK_SHARED | LK_RETRY); 773 } 774 i -= dp->d_namlen; 775 776 if (i < 0) { 777 error = ENOMEM; 778 goto out; 779 } 780 if (dp->d_namlen == 1 && dp->d_name[0] == '.') { 781 error = ENOENT; 782 } else { 783 bcopy(dp->d_name, buf + i, dp->d_namlen); 784 error = 0; 785 } 786 goto out; 787 } 788 } 789 790 out: 791 free(dirbuf, M_TEMP); 792 if (!error) { 793 *buflen = i; 794 vref(*dvp); 795 } 796 if (covered) { 797 vput(*dvp); 798 vrele(mvp); 799 } else { 800 VOP_UNLOCK(mvp); 801 vn_close(mvp, FREAD, cred, td); 802 } 803 vn_lock(vp, locked | LK_RETRY); 804 return (error); 805 } 806 807 int 808 vop_stdallocate(struct vop_allocate_args *ap) 809 { 810 #ifdef __notyet__ 811 struct statfs *sfs; 812 off_t maxfilesize = 0; 813 #endif 814 struct iovec aiov; 815 struct vattr vattr, *vap; 816 struct uio auio; 817 off_t fsize, len, cur, offset; 818 uint8_t *buf; 819 struct thread *td; 820 struct vnode *vp; 821 size_t iosize; 822 int error; 823 824 buf = NULL; 825 error = 0; 826 td = curthread; 827 vap = &vattr; 828 vp = ap->a_vp; 829 len = *ap->a_len; 830 offset = *ap->a_offset; 831 832 error = VOP_GETATTR(vp, vap, ap->a_cred); 833 if (error != 0) 834 goto out; 835 fsize = vap->va_size; 836 iosize = vap->va_blocksize; 837 if (iosize == 0) 838 iosize = BLKDEV_IOSIZE; 839 if (iosize > maxphys) 840 iosize = maxphys; 841 buf = malloc(iosize, M_TEMP, M_WAITOK); 842 843 #ifdef __notyet__ 844 /* 845 * Check if the filesystem sets f_maxfilesize; if not use 846 * VOP_SETATTR to perform the check. 847 */ 848 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 849 error = VFS_STATFS(vp->v_mount, sfs, td); 850 if (error == 0) 851 maxfilesize = sfs->f_maxfilesize; 852 free(sfs, M_STATFS); 853 if (error != 0) 854 goto out; 855 if (maxfilesize) { 856 if (offset > maxfilesize || len > maxfilesize || 857 offset + len > maxfilesize) { 858 error = EFBIG; 859 goto out; 860 } 861 } else 862 #endif 863 if (offset + len > vap->va_size) { 864 /* 865 * Test offset + len against the filesystem's maxfilesize. 866 */ 867 VATTR_NULL(vap); 868 vap->va_size = offset + len; 869 error = VOP_SETATTR(vp, vap, ap->a_cred); 870 if (error != 0) 871 goto out; 872 VATTR_NULL(vap); 873 vap->va_size = fsize; 874 error = VOP_SETATTR(vp, vap, ap->a_cred); 875 if (error != 0) 876 goto out; 877 } 878 879 for (;;) { 880 /* 881 * Read and write back anything below the nominal file 882 * size. There's currently no way outside the filesystem 883 * to know whether this area is sparse or not. 884 */ 885 cur = iosize; 886 if ((offset % iosize) != 0) 887 cur -= (offset % iosize); 888 if (cur > len) 889 cur = len; 890 if (offset < fsize) { 891 aiov.iov_base = buf; 892 aiov.iov_len = cur; 893 auio.uio_iov = &aiov; 894 auio.uio_iovcnt = 1; 895 auio.uio_offset = offset; 896 auio.uio_resid = cur; 897 auio.uio_segflg = UIO_SYSSPACE; 898 auio.uio_rw = UIO_READ; 899 auio.uio_td = td; 900 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred); 901 if (error != 0) 902 break; 903 if (auio.uio_resid > 0) { 904 bzero(buf + cur - auio.uio_resid, 905 auio.uio_resid); 906 } 907 } else { 908 bzero(buf, cur); 909 } 910 911 aiov.iov_base = buf; 912 aiov.iov_len = cur; 913 auio.uio_iov = &aiov; 914 auio.uio_iovcnt = 1; 915 auio.uio_offset = offset; 916 auio.uio_resid = cur; 917 auio.uio_segflg = UIO_SYSSPACE; 918 auio.uio_rw = UIO_WRITE; 919 auio.uio_td = td; 920 921 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred); 922 if (error != 0) 923 break; 924 925 len -= cur; 926 offset += cur; 927 if (len == 0) 928 break; 929 if (should_yield()) 930 break; 931 } 932 933 out: 934 *ap->a_len = len; 935 *ap->a_offset = offset; 936 free(buf, M_TEMP); 937 return (error); 938 } 939 940 static int 941 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp, 942 int ioflag, struct ucred *cred) 943 { 944 int iosize; 945 int error = 0; 946 struct iovec aiov; 947 struct uio auio; 948 struct thread *td; 949 off_t offset, len; 950 951 iosize = vap->va_blocksize; 952 td = curthread; 953 offset = *offsetp; 954 len = *lenp; 955 956 if (iosize == 0) 957 iosize = BLKDEV_IOSIZE; 958 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */ 959 iosize = min(iosize * 8, ZERO_REGION_SIZE); 960 961 while (len > 0) { 962 int xfersize = iosize; 963 if (offset % iosize != 0) 964 xfersize -= offset % iosize; 965 if (xfersize > len) 966 xfersize = len; 967 968 aiov.iov_base = __DECONST(void *, zero_region); 969 aiov.iov_len = xfersize; 970 auio.uio_iov = &aiov; 971 auio.uio_iovcnt = 1; 972 auio.uio_offset = offset; 973 auio.uio_resid = xfersize; 974 auio.uio_segflg = UIO_SYSSPACE; 975 auio.uio_rw = UIO_WRITE; 976 auio.uio_td = td; 977 978 error = VOP_WRITE(vp, &auio, ioflag, cred); 979 if (error != 0) { 980 len -= xfersize - auio.uio_resid; 981 offset += xfersize - auio.uio_resid; 982 break; 983 } 984 985 len -= xfersize; 986 offset += xfersize; 987 } 988 989 *offsetp = offset; 990 *lenp = len; 991 return (error); 992 } 993 994 int 995 vop_stddeallocate(struct vop_deallocate_args *ap) 996 { 997 struct vnode *vp; 998 off_t offset, len; 999 struct ucred *cred; 1000 int error; 1001 struct vattr va; 1002 off_t noff, xfersize, rem; 1003 1004 vp = ap->a_vp; 1005 offset = *ap->a_offset; 1006 cred = ap->a_cred; 1007 1008 error = VOP_GETATTR(vp, &va, cred); 1009 if (error) 1010 return (error); 1011 1012 len = omin((off_t)va.va_size - offset, *ap->a_len); 1013 while (len > 0) { 1014 noff = offset; 1015 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred); 1016 if (error) { 1017 if (error != ENXIO) 1018 /* XXX: Is it okay to fallback further? */ 1019 goto out; 1020 1021 /* 1022 * No more data region to be filled 1023 */ 1024 offset += len; 1025 len = 0; 1026 error = 0; 1027 break; 1028 } 1029 KASSERT(noff >= offset, ("FIOSEEKDATA going backward")); 1030 if (noff != offset) { 1031 xfersize = omin(noff - offset, len); 1032 len -= xfersize; 1033 offset += xfersize; 1034 if (len == 0) 1035 break; 1036 } 1037 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred); 1038 if (error) 1039 goto out; 1040 1041 /* Fill zeroes */ 1042 xfersize = rem = omin(noff - offset, len); 1043 error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred); 1044 if (error) { 1045 len -= xfersize - rem; 1046 goto out; 1047 } 1048 1049 len -= xfersize; 1050 if (should_yield()) 1051 break; 1052 } 1053 /* Handle the case when offset is beyond EOF */ 1054 if (len < 0) 1055 len = 0; 1056 out: 1057 *ap->a_offset = offset; 1058 *ap->a_len = len; 1059 return (error); 1060 } 1061 1062 int 1063 vop_stdadvise(struct vop_advise_args *ap) 1064 { 1065 struct vnode *vp; 1066 struct bufobj *bo; 1067 uintmax_t bstart, bend; 1068 daddr_t startn, endn; 1069 int bsize, error; 1070 1071 vp = ap->a_vp; 1072 switch (ap->a_advice) { 1073 case POSIX_FADV_WILLNEED: 1074 /* 1075 * Do nothing for now. Filesystems should provide a 1076 * custom method which starts an asynchronous read of 1077 * the requested region. 1078 */ 1079 error = 0; 1080 break; 1081 case POSIX_FADV_DONTNEED: 1082 error = 0; 1083 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1084 if (VN_IS_DOOMED(vp)) { 1085 VOP_UNLOCK(vp); 1086 break; 1087 } 1088 1089 /* 1090 * Round to block boundaries (and later possibly further to 1091 * page boundaries). Applications cannot reasonably be aware 1092 * of the boundaries, and the rounding must be to expand at 1093 * both extremities to cover enough. It still doesn't cover 1094 * read-ahead. For partial blocks, this gives unnecessary 1095 * discarding of buffers but is efficient enough since the 1096 * pages usually remain in VMIO for some time. 1097 */ 1098 bsize = vp->v_bufobj.bo_bsize; 1099 bstart = rounddown(ap->a_start, bsize); 1100 bend = ap->a_end; 1101 bend = roundup(bend, bsize); 1102 1103 /* 1104 * Deactivate pages in the specified range from the backing VM 1105 * object. Pages that are resident in the buffer cache will 1106 * remain wired until their corresponding buffers are released 1107 * below. 1108 */ 1109 if (vp->v_object != NULL) { 1110 VM_OBJECT_RLOCK(vp->v_object); 1111 vm_object_page_noreuse(vp->v_object, 1112 OFF_TO_IDX(trunc_page(bstart)), 1113 OFF_TO_IDX(round_page(bend))); 1114 VM_OBJECT_RUNLOCK(vp->v_object); 1115 } 1116 1117 bo = &vp->v_bufobj; 1118 startn = bstart / bsize; 1119 endn = bend / bsize; 1120 BO_RLOCK(bo); 1121 error = bnoreuselist(&bo->bo_clean, bo, startn, endn); 1122 if (error == 0) 1123 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn); 1124 BO_RUNLOCK(bo); 1125 VOP_UNLOCK(vp); 1126 break; 1127 default: 1128 error = EINVAL; 1129 break; 1130 } 1131 return (error); 1132 } 1133 1134 int 1135 vop_stdunp_bind(struct vop_unp_bind_args *ap) 1136 { 1137 1138 ap->a_vp->v_unpcb = ap->a_unpcb; 1139 return (0); 1140 } 1141 1142 int 1143 vop_stdunp_connect(struct vop_unp_connect_args *ap) 1144 { 1145 1146 *ap->a_unpcb = ap->a_vp->v_unpcb; 1147 return (0); 1148 } 1149 1150 int 1151 vop_stdunp_detach(struct vop_unp_detach_args *ap) 1152 { 1153 1154 ap->a_vp->v_unpcb = NULL; 1155 return (0); 1156 } 1157 1158 static int 1159 vop_stdis_text(struct vop_is_text_args *ap) 1160 { 1161 1162 return ((int)atomic_load_int(&ap->a_vp->v_writecount) < 0); 1163 } 1164 1165 int 1166 vop_stdset_text(struct vop_set_text_args *ap) 1167 { 1168 struct vnode *vp; 1169 int n; 1170 bool gotref; 1171 1172 vp = ap->a_vp; 1173 1174 n = atomic_load_int(&vp->v_writecount); 1175 for (;;) { 1176 if (__predict_false(n > 0)) { 1177 return (ETXTBSY); 1178 } 1179 1180 /* 1181 * Transition point, we may need to grab a reference on the vnode. 1182 * 1183 * Take the ref early As a safety measure against bogus calls 1184 * to vop_stdunset_text. 1185 */ 1186 if (n == 0) { 1187 gotref = false; 1188 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) { 1189 vref(vp); 1190 gotref = true; 1191 } 1192 if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) { 1193 return (0); 1194 } 1195 if (gotref) { 1196 vunref(vp); 1197 } 1198 continue; 1199 } 1200 1201 MPASS(n < 0); 1202 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) { 1203 return (0); 1204 } 1205 } 1206 __assert_unreachable(); 1207 } 1208 1209 static int 1210 vop_stdunset_text(struct vop_unset_text_args *ap) 1211 { 1212 struct vnode *vp; 1213 int n; 1214 1215 vp = ap->a_vp; 1216 1217 n = atomic_load_int(&vp->v_writecount); 1218 for (;;) { 1219 if (__predict_false(n >= 0)) { 1220 return (EINVAL); 1221 } 1222 1223 /* 1224 * Transition point, we may need to release a reference on the vnode. 1225 */ 1226 if (n == -1) { 1227 if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) { 1228 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) { 1229 vunref(vp); 1230 } 1231 return (0); 1232 } 1233 continue; 1234 } 1235 1236 MPASS(n < -1); 1237 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) { 1238 return (0); 1239 } 1240 } 1241 __assert_unreachable(); 1242 } 1243 1244 static __always_inline int 1245 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync) 1246 { 1247 struct vnode *vp; 1248 struct mount *mp __diagused; 1249 int n; 1250 1251 vp = ap->a_vp; 1252 1253 #ifdef INVARIANTS 1254 mp = vp->v_mount; 1255 if (mp != NULL) { 1256 if (handle_msync) { 1257 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp); 1258 } else { 1259 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp); 1260 } 1261 } 1262 #endif 1263 1264 n = atomic_load_int(&vp->v_writecount); 1265 for (;;) { 1266 if (__predict_false(n < 0)) { 1267 return (ETXTBSY); 1268 } 1269 1270 VNASSERT(n + ap->a_inc >= 0, vp, 1271 ("neg writecount increment %d + %d = %d", n, ap->a_inc, 1272 n + ap->a_inc)); 1273 if (n == 0) { 1274 if (handle_msync) { 1275 vlazy(vp); 1276 } 1277 } 1278 1279 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) { 1280 return (0); 1281 } 1282 } 1283 __assert_unreachable(); 1284 } 1285 1286 int 1287 vop_stdadd_writecount(struct vop_add_writecount_args *ap) 1288 { 1289 1290 return (vop_stdadd_writecount_impl(ap, true)); 1291 } 1292 1293 int 1294 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap) 1295 { 1296 1297 return (vop_stdadd_writecount_impl(ap, false)); 1298 } 1299 1300 int 1301 vop_stdneed_inactive(struct vop_need_inactive_args *ap) 1302 { 1303 1304 return (1); 1305 } 1306 1307 int 1308 vop_stdioctl(struct vop_ioctl_args *ap) 1309 { 1310 struct vnode *vp; 1311 struct vattr va; 1312 off_t *offp; 1313 int error; 1314 1315 switch (ap->a_command) { 1316 case FIOSEEKDATA: 1317 case FIOSEEKHOLE: 1318 vp = ap->a_vp; 1319 error = vn_lock(vp, LK_SHARED); 1320 if (error != 0) 1321 return (EBADF); 1322 if (vp->v_type == VREG) 1323 error = VOP_GETATTR(vp, &va, ap->a_cred); 1324 else 1325 error = ENOTTY; 1326 if (error == 0) { 1327 offp = ap->a_data; 1328 if (*offp < 0 || *offp >= va.va_size) 1329 error = ENXIO; 1330 else if (ap->a_command == FIOSEEKHOLE) 1331 *offp = va.va_size; 1332 } 1333 VOP_UNLOCK(vp); 1334 break; 1335 default: 1336 error = ENOTTY; 1337 break; 1338 } 1339 return (error); 1340 } 1341 1342 /* 1343 * vfs default ops 1344 * used to fill the vfs function table to get reasonable default return values. 1345 */ 1346 int 1347 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp) 1348 { 1349 1350 return (EOPNOTSUPP); 1351 } 1352 1353 int 1354 vfs_stdstatfs(struct mount *mp, struct statfs *sbp) 1355 { 1356 1357 return (EOPNOTSUPP); 1358 } 1359 1360 int 1361 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy) 1362 { 1363 return (EOPNOTSUPP); 1364 } 1365 1366 int 1367 vfs_stdsync(struct mount *mp, int waitfor) 1368 { 1369 struct vnode *vp, *mvp; 1370 struct thread *td; 1371 int error, lockreq, allerror = 0; 1372 1373 td = curthread; 1374 lockreq = LK_EXCLUSIVE | LK_INTERLOCK; 1375 if (waitfor != MNT_WAIT) 1376 lockreq |= LK_NOWAIT; 1377 /* 1378 * Force stale buffer cache information to be flushed. 1379 */ 1380 loop: 1381 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 1382 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) { 1383 VI_UNLOCK(vp); 1384 continue; 1385 } 1386 if ((error = vget(vp, lockreq)) != 0) { 1387 if (error == ENOENT) { 1388 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 1389 goto loop; 1390 } 1391 continue; 1392 } 1393 error = VOP_FSYNC(vp, waitfor, td); 1394 if (error) 1395 allerror = error; 1396 vput(vp); 1397 } 1398 return (allerror); 1399 } 1400 1401 int 1402 vfs_stdnosync(struct mount *mp, int waitfor) 1403 { 1404 1405 return (0); 1406 } 1407 1408 static int 1409 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap) 1410 { 1411 int error; 1412 1413 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp, 1414 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred, 1415 ap->a_outcred, ap->a_fsizetd); 1416 return (error); 1417 } 1418 1419 int 1420 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp) 1421 { 1422 1423 return (EOPNOTSUPP); 1424 } 1425 1426 int 1427 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp) 1428 { 1429 1430 return (EOPNOTSUPP); 1431 } 1432 1433 int 1434 vfs_stdinit(struct vfsconf *vfsp) 1435 { 1436 1437 return (0); 1438 } 1439 1440 int 1441 vfs_stduninit(struct vfsconf *vfsp) 1442 { 1443 1444 return(0); 1445 } 1446 1447 int 1448 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp, 1449 int attrnamespace, const char *attrname) 1450 { 1451 1452 if (filename_vp != NULL) 1453 VOP_UNLOCK(filename_vp); 1454 return (EOPNOTSUPP); 1455 } 1456 1457 int 1458 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req) 1459 { 1460 1461 return (EOPNOTSUPP); 1462 } 1463 1464 static vop_bypass_t * 1465 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a) 1466 { 1467 1468 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset)); 1469 } 1470 1471 int 1472 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a) 1473 { 1474 vop_bypass_t *bp; 1475 int prev_stops, rc; 1476 1477 bp = bp_by_off(vop, a); 1478 MPASS(bp != NULL); 1479 1480 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 1481 rc = bp(a); 1482 sigallowstop(prev_stops); 1483 return (rc); 1484 } 1485 1486 static int 1487 vop_stdstat(struct vop_stat_args *a) 1488 { 1489 struct vattr vattr; 1490 struct vattr *vap; 1491 struct vnode *vp; 1492 struct stat *sb; 1493 int error; 1494 u_short mode; 1495 1496 vp = a->a_vp; 1497 sb = a->a_sb; 1498 1499 error = vop_stat_helper_pre(a); 1500 if (error != 0) 1501 return (error); 1502 1503 vap = &vattr; 1504 1505 /* 1506 * Initialize defaults for new and unusual fields, so that file 1507 * systems which don't support these fields don't need to know 1508 * about them. 1509 */ 1510 vap->va_birthtime.tv_sec = -1; 1511 vap->va_birthtime.tv_nsec = 0; 1512 vap->va_fsid = VNOVAL; 1513 vap->va_gen = 0; 1514 vap->va_rdev = NODEV; 1515 vap->va_filerev = 0; 1516 vap->va_bsdflags = 0; 1517 1518 error = VOP_GETATTR(vp, vap, a->a_active_cred); 1519 if (error) 1520 goto out; 1521 1522 /* 1523 * Zero the spare stat fields 1524 */ 1525 bzero(sb, sizeof *sb); 1526 1527 /* 1528 * Copy from vattr table 1529 */ 1530 if (vap->va_fsid != VNOVAL) 1531 sb->st_dev = vap->va_fsid; 1532 else 1533 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0]; 1534 sb->st_ino = vap->va_fileid; 1535 mode = vap->va_mode; 1536 switch (vap->va_type) { 1537 case VREG: 1538 mode |= S_IFREG; 1539 break; 1540 case VDIR: 1541 mode |= S_IFDIR; 1542 break; 1543 case VBLK: 1544 mode |= S_IFBLK; 1545 break; 1546 case VCHR: 1547 mode |= S_IFCHR; 1548 break; 1549 case VLNK: 1550 mode |= S_IFLNK; 1551 break; 1552 case VSOCK: 1553 mode |= S_IFSOCK; 1554 break; 1555 case VFIFO: 1556 mode |= S_IFIFO; 1557 break; 1558 default: 1559 error = EBADF; 1560 goto out; 1561 } 1562 sb->st_mode = mode; 1563 sb->st_nlink = vap->va_nlink; 1564 sb->st_uid = vap->va_uid; 1565 sb->st_gid = vap->va_gid; 1566 sb->st_rdev = vap->va_rdev; 1567 if (vap->va_size > OFF_MAX) { 1568 error = EOVERFLOW; 1569 goto out; 1570 } 1571 sb->st_size = vap->va_size; 1572 sb->st_atim.tv_sec = vap->va_atime.tv_sec; 1573 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec; 1574 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec; 1575 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec; 1576 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec; 1577 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec; 1578 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec; 1579 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec; 1580 1581 /* 1582 * According to www.opengroup.org, the meaning of st_blksize is 1583 * "a filesystem-specific preferred I/O block size for this 1584 * object. In some filesystem types, this may vary from file 1585 * to file" 1586 * Use minimum/default of PAGE_SIZE (e.g. for VCHR). 1587 */ 1588 1589 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize); 1590 sb->st_flags = vap->va_flags; 1591 sb->st_blocks = vap->va_bytes / S_BLKSIZE; 1592 sb->st_gen = vap->va_gen; 1593 sb->st_filerev = vap->va_filerev; 1594 sb->st_bsdflags = vap->va_bsdflags; 1595 out: 1596 return (vop_stat_helper_post(a, error)); 1597 } 1598 1599 static int 1600 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused) 1601 { 1602 return (EJUSTRETURN); 1603 } 1604 1605 static int 1606 vop_stdvput_pair(struct vop_vput_pair_args *ap) 1607 { 1608 struct vnode *dvp, *vp, **vpp; 1609 1610 dvp = ap->a_dvp; 1611 vpp = ap->a_vpp; 1612 vput(dvp); 1613 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL) 1614 vput(vp); 1615 return (0); 1616 } 1617 1618 static int 1619 vop_stdgetlowvnode(struct vop_getlowvnode_args *ap) 1620 { 1621 vref(ap->a_vp); 1622 *ap->a_vplp = ap->a_vp; 1623 return (0); 1624 } 1625