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 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 */ 36 37 #include <sys/cdefs.h> 38 __FBSDID("$FreeBSD$"); 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/bio.h> 43 #include <sys/buf.h> 44 #include <sys/conf.h> 45 #include <sys/event.h> 46 #include <sys/filio.h> 47 #include <sys/kernel.h> 48 #include <sys/limits.h> 49 #include <sys/lock.h> 50 #include <sys/lockf.h> 51 #include <sys/malloc.h> 52 #include <sys/mount.h> 53 #include <sys/namei.h> 54 #include <sys/rwlock.h> 55 #include <sys/fcntl.h> 56 #include <sys/unistd.h> 57 #include <sys/vnode.h> 58 #include <sys/dirent.h> 59 #include <sys/poll.h> 60 #include <sys/stat.h> 61 #include <security/audit/audit.h> 62 #include <sys/priv.h> 63 64 #include <security/mac/mac_framework.h> 65 66 #include <vm/vm.h> 67 #include <vm/vm_object.h> 68 #include <vm/vm_extern.h> 69 #include <vm/pmap.h> 70 #include <vm/vm_map.h> 71 #include <vm/vm_page.h> 72 #include <vm/vm_pager.h> 73 #include <vm/vnode_pager.h> 74 75 static int vop_nolookup(struct vop_lookup_args *); 76 static int vop_norename(struct vop_rename_args *); 77 static int vop_nostrategy(struct vop_strategy_args *); 78 static int get_next_dirent(struct vnode *vp, struct dirent **dpp, 79 char *dirbuf, int dirbuflen, off_t *off, 80 char **cpos, int *len, int *eofflag, 81 struct thread *td); 82 static int dirent_exists(struct vnode *vp, const char *dirname, 83 struct thread *td); 84 85 #define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4) 86 87 static int vop_stdis_text(struct vop_is_text_args *ap); 88 static int vop_stdunset_text(struct vop_unset_text_args *ap); 89 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap); 90 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap); 91 static int vop_stdfdatasync(struct vop_fdatasync_args *ap); 92 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap); 93 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap); 94 static int vop_stdstat(struct vop_stat_args *ap); 95 static int vop_stdvput_pair(struct vop_vput_pair_args *ap); 96 97 /* 98 * This vnode table stores what we want to do if the filesystem doesn't 99 * implement a particular VOP. 100 * 101 * If there is no specific entry here, we will return EOPNOTSUPP. 102 * 103 * Note that every filesystem has to implement either vop_access 104 * or vop_accessx; failing to do so will result in immediate crash 105 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(), 106 * which calls vop_stdaccess() etc. 107 */ 108 109 struct vop_vector default_vnodeops = { 110 .vop_default = NULL, 111 .vop_bypass = VOP_EOPNOTSUPP, 112 113 .vop_access = vop_stdaccess, 114 .vop_accessx = vop_stdaccessx, 115 .vop_advise = vop_stdadvise, 116 .vop_advlock = vop_stdadvlock, 117 .vop_advlockasync = vop_stdadvlockasync, 118 .vop_advlockpurge = vop_stdadvlockpurge, 119 .vop_allocate = vop_stdallocate, 120 .vop_bmap = vop_stdbmap, 121 .vop_close = VOP_NULL, 122 .vop_fsync = VOP_NULL, 123 .vop_stat = vop_stdstat, 124 .vop_fdatasync = vop_stdfdatasync, 125 .vop_getpages = vop_stdgetpages, 126 .vop_getpages_async = vop_stdgetpages_async, 127 .vop_getwritemount = vop_stdgetwritemount, 128 .vop_inactive = VOP_NULL, 129 .vop_need_inactive = vop_stdneed_inactive, 130 .vop_ioctl = vop_stdioctl, 131 .vop_kqfilter = vop_stdkqfilter, 132 .vop_islocked = vop_stdislocked, 133 .vop_lock1 = vop_stdlock, 134 .vop_lookup = vop_nolookup, 135 .vop_open = VOP_NULL, 136 .vop_pathconf = VOP_EINVAL, 137 .vop_poll = vop_nopoll, 138 .vop_putpages = vop_stdputpages, 139 .vop_readlink = VOP_EINVAL, 140 .vop_read_pgcache = vop_stdread_pgcache, 141 .vop_rename = vop_norename, 142 .vop_revoke = VOP_PANIC, 143 .vop_strategy = vop_nostrategy, 144 .vop_unlock = vop_stdunlock, 145 .vop_vptocnp = vop_stdvptocnp, 146 .vop_vptofh = vop_stdvptofh, 147 .vop_unp_bind = vop_stdunp_bind, 148 .vop_unp_connect = vop_stdunp_connect, 149 .vop_unp_detach = vop_stdunp_detach, 150 .vop_is_text = vop_stdis_text, 151 .vop_set_text = vop_stdset_text, 152 .vop_unset_text = vop_stdunset_text, 153 .vop_add_writecount = vop_stdadd_writecount, 154 .vop_copy_file_range = vop_stdcopy_file_range, 155 .vop_vput_pair = vop_stdvput_pair, 156 }; 157 VFS_VOP_VECTOR_REGISTER(default_vnodeops); 158 159 /* 160 * Series of placeholder functions for various error returns for 161 * VOPs. 162 */ 163 164 int 165 vop_eopnotsupp(struct vop_generic_args *ap) 166 { 167 /* 168 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 169 */ 170 171 return (EOPNOTSUPP); 172 } 173 174 int 175 vop_ebadf(struct vop_generic_args *ap) 176 { 177 178 return (EBADF); 179 } 180 181 int 182 vop_enotty(struct vop_generic_args *ap) 183 { 184 185 return (ENOTTY); 186 } 187 188 int 189 vop_einval(struct vop_generic_args *ap) 190 { 191 192 return (EINVAL); 193 } 194 195 int 196 vop_enoent(struct vop_generic_args *ap) 197 { 198 199 return (ENOENT); 200 } 201 202 int 203 vop_eagain(struct vop_generic_args *ap) 204 { 205 206 return (EAGAIN); 207 } 208 209 int 210 vop_null(struct vop_generic_args *ap) 211 { 212 213 return (0); 214 } 215 216 /* 217 * Helper function to panic on some bad VOPs in some filesystems. 218 */ 219 int 220 vop_panic(struct vop_generic_args *ap) 221 { 222 223 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 224 } 225 226 /* 227 * vop_std<something> and vop_no<something> are default functions for use by 228 * filesystems that need the "default reasonable" implementation for a 229 * particular operation. 230 * 231 * The documentation for the operations they implement exists (if it exists) 232 * in the VOP_<SOMETHING>(9) manpage (all uppercase). 233 */ 234 235 /* 236 * Default vop for filesystems that do not support name lookup 237 */ 238 static int 239 vop_nolookup(ap) 240 struct vop_lookup_args /* { 241 struct vnode *a_dvp; 242 struct vnode **a_vpp; 243 struct componentname *a_cnp; 244 } */ *ap; 245 { 246 247 *ap->a_vpp = NULL; 248 return (ENOTDIR); 249 } 250 251 /* 252 * vop_norename: 253 * 254 * Handle unlock and reference counting for arguments of vop_rename 255 * for filesystems that do not implement rename operation. 256 */ 257 static int 258 vop_norename(struct vop_rename_args *ap) 259 { 260 261 vop_rename_fail(ap); 262 return (EOPNOTSUPP); 263 } 264 265 /* 266 * vop_nostrategy: 267 * 268 * Strategy routine for VFS devices that have none. 269 * 270 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 271 * routine. Typically this is done for a BIO_READ strategy call. 272 * Typically B_INVAL is assumed to already be clear prior to a write 273 * and should not be cleared manually unless you just made the buffer 274 * invalid. BIO_ERROR should be cleared either way. 275 */ 276 277 static int 278 vop_nostrategy (struct vop_strategy_args *ap) 279 { 280 printf("No strategy for buffer at %p\n", ap->a_bp); 281 vn_printf(ap->a_vp, "vnode "); 282 ap->a_bp->b_ioflags |= BIO_ERROR; 283 ap->a_bp->b_error = EOPNOTSUPP; 284 bufdone(ap->a_bp); 285 return (EOPNOTSUPP); 286 } 287 288 static int 289 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf, 290 int dirbuflen, off_t *off, char **cpos, int *len, 291 int *eofflag, struct thread *td) 292 { 293 int error, reclen; 294 struct uio uio; 295 struct iovec iov; 296 struct dirent *dp; 297 298 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp)); 299 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp)); 300 301 if (*len == 0) { 302 iov.iov_base = dirbuf; 303 iov.iov_len = dirbuflen; 304 305 uio.uio_iov = &iov; 306 uio.uio_iovcnt = 1; 307 uio.uio_offset = *off; 308 uio.uio_resid = dirbuflen; 309 uio.uio_segflg = UIO_SYSSPACE; 310 uio.uio_rw = UIO_READ; 311 uio.uio_td = td; 312 313 *eofflag = 0; 314 315 #ifdef MAC 316 error = mac_vnode_check_readdir(td->td_ucred, vp); 317 if (error == 0) 318 #endif 319 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag, 320 NULL, NULL); 321 if (error) 322 return (error); 323 324 *off = uio.uio_offset; 325 326 *cpos = dirbuf; 327 *len = (dirbuflen - uio.uio_resid); 328 329 if (*len == 0) 330 return (ENOENT); 331 } 332 333 dp = (struct dirent *)(*cpos); 334 reclen = dp->d_reclen; 335 *dpp = dp; 336 337 /* check for malformed directory.. */ 338 if (reclen < DIRENT_MINSIZE) 339 return (EINVAL); 340 341 *cpos += reclen; 342 *len -= reclen; 343 344 return (0); 345 } 346 347 /* 348 * Check if a named file exists in a given directory vnode. 349 */ 350 static int 351 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td) 352 { 353 char *dirbuf, *cpos; 354 int error, eofflag, dirbuflen, len, found; 355 off_t off; 356 struct dirent *dp; 357 struct vattr va; 358 359 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp)); 360 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp)); 361 362 found = 0; 363 364 error = VOP_GETATTR(vp, &va, td->td_ucred); 365 if (error) 366 return (found); 367 368 dirbuflen = DEV_BSIZE; 369 if (dirbuflen < va.va_blocksize) 370 dirbuflen = va.va_blocksize; 371 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK); 372 373 off = 0; 374 len = 0; 375 do { 376 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off, 377 &cpos, &len, &eofflag, td); 378 if (error) 379 goto out; 380 381 if (dp->d_type != DT_WHT && dp->d_fileno != 0 && 382 strcmp(dp->d_name, dirname) == 0) { 383 found = 1; 384 goto out; 385 } 386 } while (len > 0 || !eofflag); 387 388 out: 389 free(dirbuf, M_TEMP); 390 return (found); 391 } 392 393 int 394 vop_stdaccess(struct vop_access_args *ap) 395 { 396 397 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN | 398 VAPPEND)) == 0, ("invalid bit in accmode")); 399 400 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td)); 401 } 402 403 int 404 vop_stdaccessx(struct vop_accessx_args *ap) 405 { 406 int error; 407 accmode_t accmode = ap->a_accmode; 408 409 error = vfs_unixify_accmode(&accmode); 410 if (error != 0) 411 return (error); 412 413 if (accmode == 0) 414 return (0); 415 416 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td)); 417 } 418 419 /* 420 * Advisory record locking support 421 */ 422 int 423 vop_stdadvlock(struct vop_advlock_args *ap) 424 { 425 struct vnode *vp; 426 struct mount *mp; 427 struct vattr vattr; 428 int error; 429 430 vp = ap->a_vp; 431 432 /* 433 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for 434 * local filesystems. See vn_open_cred() for reciprocal part. 435 */ 436 mp = vp->v_mount; 437 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 && 438 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) { 439 VI_LOCK(vp); 440 while ((vp->v_iflag & VI_FOPENING) != 0) 441 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0); 442 VI_UNLOCK(vp); 443 } 444 445 if (ap->a_fl->l_whence == SEEK_END) { 446 /* 447 * The NFSv4 server must avoid doing a vn_lock() here, since it 448 * can deadlock the nfsd threads, due to a LOR. Fortunately 449 * the NFSv4 server always uses SEEK_SET and this code is 450 * only required for the SEEK_END case. 451 */ 452 vn_lock(vp, LK_SHARED | LK_RETRY); 453 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred); 454 VOP_UNLOCK(vp); 455 if (error) 456 return (error); 457 } else 458 vattr.va_size = 0; 459 460 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size)); 461 } 462 463 int 464 vop_stdadvlockasync(struct vop_advlockasync_args *ap) 465 { 466 struct vnode *vp; 467 struct vattr vattr; 468 int error; 469 470 vp = ap->a_vp; 471 if (ap->a_fl->l_whence == SEEK_END) { 472 /* The size argument is only needed for SEEK_END. */ 473 vn_lock(vp, LK_SHARED | LK_RETRY); 474 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred); 475 VOP_UNLOCK(vp); 476 if (error) 477 return (error); 478 } else 479 vattr.va_size = 0; 480 481 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size)); 482 } 483 484 int 485 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap) 486 { 487 struct vnode *vp; 488 489 vp = ap->a_vp; 490 lf_purgelocks(vp, &vp->v_lockf); 491 return (0); 492 } 493 494 /* 495 * vop_stdpathconf: 496 * 497 * Standard implementation of POSIX pathconf, to get information about limits 498 * for a filesystem. 499 * Override per filesystem for the case where the filesystem has smaller 500 * limits. 501 */ 502 int 503 vop_stdpathconf(ap) 504 struct vop_pathconf_args /* { 505 struct vnode *a_vp; 506 int a_name; 507 int *a_retval; 508 } */ *ap; 509 { 510 511 switch (ap->a_name) { 512 case _PC_ASYNC_IO: 513 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO; 514 return (0); 515 case _PC_PATH_MAX: 516 *ap->a_retval = PATH_MAX; 517 return (0); 518 case _PC_ACL_EXTENDED: 519 case _PC_ACL_NFS4: 520 case _PC_CAP_PRESENT: 521 case _PC_INF_PRESENT: 522 case _PC_MAC_PRESENT: 523 *ap->a_retval = 0; 524 return (0); 525 default: 526 return (EINVAL); 527 } 528 /* NOTREACHED */ 529 } 530 531 /* 532 * Standard lock, unlock and islocked functions. 533 */ 534 int 535 vop_stdlock(ap) 536 struct vop_lock1_args /* { 537 struct vnode *a_vp; 538 int a_flags; 539 char *file; 540 int line; 541 } */ *ap; 542 { 543 struct vnode *vp = ap->a_vp; 544 struct mtx *ilk; 545 546 ilk = VI_MTX(vp); 547 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags, 548 &ilk->lock_object, ap->a_file, ap->a_line)); 549 } 550 551 /* See above. */ 552 int 553 vop_stdunlock(ap) 554 struct vop_unlock_args /* { 555 struct vnode *a_vp; 556 } */ *ap; 557 { 558 struct vnode *vp = ap->a_vp; 559 560 return (lockmgr_unlock(vp->v_vnlock)); 561 } 562 563 /* See above. */ 564 int 565 vop_stdislocked(ap) 566 struct vop_islocked_args /* { 567 struct vnode *a_vp; 568 } */ *ap; 569 { 570 571 return (lockstatus(ap->a_vp->v_vnlock)); 572 } 573 574 /* 575 * Variants of the above set. 576 * 577 * Differences are: 578 * - shared locking disablement is not supported 579 * - v_vnlock pointer is not honored 580 */ 581 int 582 vop_lock(ap) 583 struct vop_lock1_args /* { 584 struct vnode *a_vp; 585 int a_flags; 586 char *file; 587 int line; 588 } */ *ap; 589 { 590 struct vnode *vp = ap->a_vp; 591 int flags = ap->a_flags; 592 struct mtx *ilk; 593 594 MPASS(vp->v_vnlock == &vp->v_lock); 595 596 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0)) 597 goto other; 598 599 switch (flags & LK_TYPE_MASK) { 600 case LK_SHARED: 601 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line)); 602 case LK_EXCLUSIVE: 603 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line)); 604 } 605 other: 606 ilk = VI_MTX(vp); 607 return (lockmgr_lock_flags(&vp->v_lock, flags, 608 &ilk->lock_object, ap->a_file, ap->a_line)); 609 } 610 611 int 612 vop_unlock(ap) 613 struct vop_unlock_args /* { 614 struct vnode *a_vp; 615 } */ *ap; 616 { 617 struct vnode *vp = ap->a_vp; 618 619 MPASS(vp->v_vnlock == &vp->v_lock); 620 621 return (lockmgr_unlock(&vp->v_lock)); 622 } 623 624 int 625 vop_islocked(ap) 626 struct vop_islocked_args /* { 627 struct vnode *a_vp; 628 } */ *ap; 629 { 630 struct vnode *vp = ap->a_vp; 631 632 MPASS(vp->v_vnlock == &vp->v_lock); 633 634 return (lockstatus(&vp->v_lock)); 635 } 636 637 /* 638 * Return true for select/poll. 639 */ 640 int 641 vop_nopoll(ap) 642 struct vop_poll_args /* { 643 struct vnode *a_vp; 644 int a_events; 645 struct ucred *a_cred; 646 struct thread *a_td; 647 } */ *ap; 648 { 649 650 if (ap->a_events & ~POLLSTANDARD) 651 return (POLLNVAL); 652 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 653 } 654 655 /* 656 * Implement poll for local filesystems that support it. 657 */ 658 int 659 vop_stdpoll(ap) 660 struct vop_poll_args /* { 661 struct vnode *a_vp; 662 int a_events; 663 struct ucred *a_cred; 664 struct thread *a_td; 665 } */ *ap; 666 { 667 if (ap->a_events & ~POLLSTANDARD) 668 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 669 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 670 } 671 672 /* 673 * Return our mount point, as we will take charge of the writes. 674 */ 675 int 676 vop_stdgetwritemount(ap) 677 struct vop_getwritemount_args /* { 678 struct vnode *a_vp; 679 struct mount **a_mpp; 680 } */ *ap; 681 { 682 struct mount *mp; 683 struct vnode *vp; 684 685 /* 686 * Note that having a reference does not prevent forced unmount from 687 * setting ->v_mount to NULL after the lock gets released. This is of 688 * no consequence for typical consumers (most notably vn_start_write) 689 * since in this case the vnode is VIRF_DOOMED. Unmount might have 690 * progressed far enough that its completion is only delayed by the 691 * reference obtained here. The consumer only needs to concern itself 692 * with releasing it. 693 */ 694 vp = ap->a_vp; 695 mp = vfs_ref_from_vp(vp); 696 *(ap->a_mpp) = mp; 697 return (0); 698 } 699 700 /* 701 * If the file system doesn't implement VOP_BMAP, then return sensible defaults: 702 * - Return the vnode's bufobj instead of any underlying device's bufobj 703 * - Calculate the physical block number as if there were equal size 704 * consecutive blocks, but 705 * - Report no contiguous runs of blocks. 706 */ 707 int 708 vop_stdbmap(ap) 709 struct vop_bmap_args /* { 710 struct vnode *a_vp; 711 daddr_t a_bn; 712 struct bufobj **a_bop; 713 daddr_t *a_bnp; 714 int *a_runp; 715 int *a_runb; 716 } */ *ap; 717 { 718 719 if (ap->a_bop != NULL) 720 *ap->a_bop = &ap->a_vp->v_bufobj; 721 if (ap->a_bnp != NULL) 722 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 723 if (ap->a_runp != NULL) 724 *ap->a_runp = 0; 725 if (ap->a_runb != NULL) 726 *ap->a_runb = 0; 727 return (0); 728 } 729 730 int 731 vop_stdfsync(ap) 732 struct vop_fsync_args /* { 733 struct vnode *a_vp; 734 int a_waitfor; 735 struct thread *a_td; 736 } */ *ap; 737 { 738 739 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor)); 740 } 741 742 static int 743 vop_stdfdatasync(struct vop_fdatasync_args *ap) 744 { 745 746 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td)); 747 } 748 749 int 750 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap) 751 { 752 753 return (vn_fsync_buf(ap->a_vp, MNT_WAIT)); 754 } 755 756 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */ 757 int 758 vop_stdgetpages(ap) 759 struct vop_getpages_args /* { 760 struct vnode *a_vp; 761 vm_page_t *a_m; 762 int a_count; 763 int *a_rbehind; 764 int *a_rahead; 765 } */ *ap; 766 { 767 768 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, 769 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL); 770 } 771 772 static int 773 vop_stdgetpages_async(struct vop_getpages_async_args *ap) 774 { 775 int error; 776 777 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind, 778 ap->a_rahead); 779 if (ap->a_iodone != NULL) 780 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error); 781 return (error); 782 } 783 784 int 785 vop_stdkqfilter(struct vop_kqfilter_args *ap) 786 { 787 return vfs_kqfilter(ap); 788 } 789 790 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */ 791 int 792 vop_stdputpages(ap) 793 struct vop_putpages_args /* { 794 struct vnode *a_vp; 795 vm_page_t *a_m; 796 int a_count; 797 int a_sync; 798 int *a_rtvals; 799 } */ *ap; 800 { 801 802 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count, 803 ap->a_sync, ap->a_rtvals); 804 } 805 806 int 807 vop_stdvptofh(struct vop_vptofh_args *ap) 808 { 809 return (EOPNOTSUPP); 810 } 811 812 int 813 vop_stdvptocnp(struct vop_vptocnp_args *ap) 814 { 815 struct vnode *vp = ap->a_vp; 816 struct vnode **dvp = ap->a_vpp; 817 struct ucred *cred; 818 char *buf = ap->a_buf; 819 size_t *buflen = ap->a_buflen; 820 char *dirbuf, *cpos; 821 int i, error, eofflag, dirbuflen, flags, locked, len, covered; 822 off_t off; 823 ino_t fileno; 824 struct vattr va; 825 struct nameidata nd; 826 struct thread *td; 827 struct dirent *dp; 828 struct vnode *mvp; 829 830 i = *buflen; 831 error = 0; 832 covered = 0; 833 td = curthread; 834 cred = td->td_ucred; 835 836 if (vp->v_type != VDIR) 837 return (ENOENT); 838 839 error = VOP_GETATTR(vp, &va, cred); 840 if (error) 841 return (error); 842 843 VREF(vp); 844 locked = VOP_ISLOCKED(vp); 845 VOP_UNLOCK(vp); 846 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE, 847 "..", vp, td); 848 flags = FREAD; 849 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL); 850 if (error) { 851 vn_lock(vp, locked | LK_RETRY); 852 return (error); 853 } 854 NDFREE(&nd, NDF_ONLY_PNBUF); 855 856 mvp = *dvp = nd.ni_vp; 857 858 if (vp->v_mount != (*dvp)->v_mount && 859 ((*dvp)->v_vflag & VV_ROOT) && 860 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) { 861 *dvp = (*dvp)->v_mount->mnt_vnodecovered; 862 VREF(mvp); 863 VOP_UNLOCK(mvp); 864 vn_close(mvp, FREAD, cred, td); 865 VREF(*dvp); 866 vn_lock(*dvp, LK_SHARED | LK_RETRY); 867 covered = 1; 868 } 869 870 fileno = va.va_fileid; 871 872 dirbuflen = DEV_BSIZE; 873 if (dirbuflen < va.va_blocksize) 874 dirbuflen = va.va_blocksize; 875 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK); 876 877 if ((*dvp)->v_type != VDIR) { 878 error = ENOENT; 879 goto out; 880 } 881 882 off = 0; 883 len = 0; 884 do { 885 /* call VOP_READDIR of parent */ 886 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off, 887 &cpos, &len, &eofflag, td); 888 if (error) 889 goto out; 890 891 if ((dp->d_type != DT_WHT) && 892 (dp->d_fileno == fileno)) { 893 if (covered) { 894 VOP_UNLOCK(*dvp); 895 vn_lock(mvp, LK_SHARED | LK_RETRY); 896 if (dirent_exists(mvp, dp->d_name, td)) { 897 error = ENOENT; 898 VOP_UNLOCK(mvp); 899 vn_lock(*dvp, LK_SHARED | LK_RETRY); 900 goto out; 901 } 902 VOP_UNLOCK(mvp); 903 vn_lock(*dvp, LK_SHARED | LK_RETRY); 904 } 905 i -= dp->d_namlen; 906 907 if (i < 0) { 908 error = ENOMEM; 909 goto out; 910 } 911 if (dp->d_namlen == 1 && dp->d_name[0] == '.') { 912 error = ENOENT; 913 } else { 914 bcopy(dp->d_name, buf + i, dp->d_namlen); 915 error = 0; 916 } 917 goto out; 918 } 919 } while (len > 0 || !eofflag); 920 error = ENOENT; 921 922 out: 923 free(dirbuf, M_TEMP); 924 if (!error) { 925 *buflen = i; 926 vref(*dvp); 927 } 928 if (covered) { 929 vput(*dvp); 930 vrele(mvp); 931 } else { 932 VOP_UNLOCK(mvp); 933 vn_close(mvp, FREAD, cred, td); 934 } 935 vn_lock(vp, locked | LK_RETRY); 936 return (error); 937 } 938 939 int 940 vop_stdallocate(struct vop_allocate_args *ap) 941 { 942 #ifdef __notyet__ 943 struct statfs *sfs; 944 off_t maxfilesize = 0; 945 #endif 946 struct iovec aiov; 947 struct vattr vattr, *vap; 948 struct uio auio; 949 off_t fsize, len, cur, offset; 950 uint8_t *buf; 951 struct thread *td; 952 struct vnode *vp; 953 size_t iosize; 954 int error; 955 956 buf = NULL; 957 error = 0; 958 td = curthread; 959 vap = &vattr; 960 vp = ap->a_vp; 961 len = *ap->a_len; 962 offset = *ap->a_offset; 963 964 error = VOP_GETATTR(vp, vap, td->td_ucred); 965 if (error != 0) 966 goto out; 967 fsize = vap->va_size; 968 iosize = vap->va_blocksize; 969 if (iosize == 0) 970 iosize = BLKDEV_IOSIZE; 971 if (iosize > maxphys) 972 iosize = maxphys; 973 buf = malloc(iosize, M_TEMP, M_WAITOK); 974 975 #ifdef __notyet__ 976 /* 977 * Check if the filesystem sets f_maxfilesize; if not use 978 * VOP_SETATTR to perform the check. 979 */ 980 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 981 error = VFS_STATFS(vp->v_mount, sfs, td); 982 if (error == 0) 983 maxfilesize = sfs->f_maxfilesize; 984 free(sfs, M_STATFS); 985 if (error != 0) 986 goto out; 987 if (maxfilesize) { 988 if (offset > maxfilesize || len > maxfilesize || 989 offset + len > maxfilesize) { 990 error = EFBIG; 991 goto out; 992 } 993 } else 994 #endif 995 if (offset + len > vap->va_size) { 996 /* 997 * Test offset + len against the filesystem's maxfilesize. 998 */ 999 VATTR_NULL(vap); 1000 vap->va_size = offset + len; 1001 error = VOP_SETATTR(vp, vap, td->td_ucred); 1002 if (error != 0) 1003 goto out; 1004 VATTR_NULL(vap); 1005 vap->va_size = fsize; 1006 error = VOP_SETATTR(vp, vap, td->td_ucred); 1007 if (error != 0) 1008 goto out; 1009 } 1010 1011 for (;;) { 1012 /* 1013 * Read and write back anything below the nominal file 1014 * size. There's currently no way outside the filesystem 1015 * to know whether this area is sparse or not. 1016 */ 1017 cur = iosize; 1018 if ((offset % iosize) != 0) 1019 cur -= (offset % iosize); 1020 if (cur > len) 1021 cur = len; 1022 if (offset < fsize) { 1023 aiov.iov_base = buf; 1024 aiov.iov_len = cur; 1025 auio.uio_iov = &aiov; 1026 auio.uio_iovcnt = 1; 1027 auio.uio_offset = offset; 1028 auio.uio_resid = cur; 1029 auio.uio_segflg = UIO_SYSSPACE; 1030 auio.uio_rw = UIO_READ; 1031 auio.uio_td = td; 1032 error = VOP_READ(vp, &auio, 0, td->td_ucred); 1033 if (error != 0) 1034 break; 1035 if (auio.uio_resid > 0) { 1036 bzero(buf + cur - auio.uio_resid, 1037 auio.uio_resid); 1038 } 1039 } else { 1040 bzero(buf, cur); 1041 } 1042 1043 aiov.iov_base = buf; 1044 aiov.iov_len = cur; 1045 auio.uio_iov = &aiov; 1046 auio.uio_iovcnt = 1; 1047 auio.uio_offset = offset; 1048 auio.uio_resid = cur; 1049 auio.uio_segflg = UIO_SYSSPACE; 1050 auio.uio_rw = UIO_WRITE; 1051 auio.uio_td = td; 1052 1053 error = VOP_WRITE(vp, &auio, 0, td->td_ucred); 1054 if (error != 0) 1055 break; 1056 1057 len -= cur; 1058 offset += cur; 1059 if (len == 0) 1060 break; 1061 if (should_yield()) 1062 break; 1063 } 1064 1065 out: 1066 *ap->a_len = len; 1067 *ap->a_offset = offset; 1068 free(buf, M_TEMP); 1069 return (error); 1070 } 1071 1072 int 1073 vop_stdadvise(struct vop_advise_args *ap) 1074 { 1075 struct vnode *vp; 1076 struct bufobj *bo; 1077 daddr_t startn, endn; 1078 off_t bstart, bend, start, end; 1079 int bsize, error; 1080 1081 vp = ap->a_vp; 1082 switch (ap->a_advice) { 1083 case POSIX_FADV_WILLNEED: 1084 /* 1085 * Do nothing for now. Filesystems should provide a 1086 * custom method which starts an asynchronous read of 1087 * the requested region. 1088 */ 1089 error = 0; 1090 break; 1091 case POSIX_FADV_DONTNEED: 1092 error = 0; 1093 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1094 if (VN_IS_DOOMED(vp)) { 1095 VOP_UNLOCK(vp); 1096 break; 1097 } 1098 1099 /* 1100 * Round to block boundaries (and later possibly further to 1101 * page boundaries). Applications cannot reasonably be aware 1102 * of the boundaries, and the rounding must be to expand at 1103 * both extremities to cover enough. It still doesn't cover 1104 * read-ahead. For partial blocks, this gives unnecessary 1105 * discarding of buffers but is efficient enough since the 1106 * pages usually remain in VMIO for some time. 1107 */ 1108 bsize = vp->v_bufobj.bo_bsize; 1109 bstart = rounddown(ap->a_start, bsize); 1110 bend = roundup(ap->a_end, bsize); 1111 1112 /* 1113 * Deactivate pages in the specified range from the backing VM 1114 * object. Pages that are resident in the buffer cache will 1115 * remain wired until their corresponding buffers are released 1116 * below. 1117 */ 1118 if (vp->v_object != NULL) { 1119 start = trunc_page(bstart); 1120 end = round_page(bend); 1121 VM_OBJECT_RLOCK(vp->v_object); 1122 vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start), 1123 OFF_TO_IDX(end)); 1124 VM_OBJECT_RUNLOCK(vp->v_object); 1125 } 1126 1127 bo = &vp->v_bufobj; 1128 BO_RLOCK(bo); 1129 startn = bstart / bsize; 1130 endn = bend / bsize; 1131 error = bnoreuselist(&bo->bo_clean, bo, startn, endn); 1132 if (error == 0) 1133 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn); 1134 BO_RUNLOCK(bo); 1135 VOP_UNLOCK(vp); 1136 break; 1137 default: 1138 error = EINVAL; 1139 break; 1140 } 1141 return (error); 1142 } 1143 1144 int 1145 vop_stdunp_bind(struct vop_unp_bind_args *ap) 1146 { 1147 1148 ap->a_vp->v_unpcb = ap->a_unpcb; 1149 return (0); 1150 } 1151 1152 int 1153 vop_stdunp_connect(struct vop_unp_connect_args *ap) 1154 { 1155 1156 *ap->a_unpcb = ap->a_vp->v_unpcb; 1157 return (0); 1158 } 1159 1160 int 1161 vop_stdunp_detach(struct vop_unp_detach_args *ap) 1162 { 1163 1164 ap->a_vp->v_unpcb = NULL; 1165 return (0); 1166 } 1167 1168 static int 1169 vop_stdis_text(struct vop_is_text_args *ap) 1170 { 1171 1172 return (ap->a_vp->v_writecount < 0); 1173 } 1174 1175 int 1176 vop_stdset_text(struct vop_set_text_args *ap) 1177 { 1178 struct vnode *vp; 1179 struct mount *mp; 1180 int error, n; 1181 1182 vp = ap->a_vp; 1183 1184 /* 1185 * Avoid the interlock if execs are already present. 1186 */ 1187 n = atomic_load_int(&vp->v_writecount); 1188 for (;;) { 1189 if (n > -1) { 1190 break; 1191 } 1192 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) { 1193 return (0); 1194 } 1195 } 1196 1197 VI_LOCK(vp); 1198 if (vp->v_writecount > 0) { 1199 error = ETXTBSY; 1200 } else { 1201 /* 1202 * If requested by fs, keep a use reference to the 1203 * vnode until the last text reference is released. 1204 */ 1205 mp = vp->v_mount; 1206 if (mp != NULL && (mp->mnt_kern_flag & MNTK_TEXT_REFS) != 0 && 1207 vp->v_writecount == 0) { 1208 VNPASS((vp->v_iflag & VI_TEXT_REF) == 0, vp); 1209 vp->v_iflag |= VI_TEXT_REF; 1210 vrefl(vp); 1211 } 1212 1213 atomic_subtract_int(&vp->v_writecount, 1); 1214 error = 0; 1215 } 1216 VI_UNLOCK(vp); 1217 return (error); 1218 } 1219 1220 static int 1221 vop_stdunset_text(struct vop_unset_text_args *ap) 1222 { 1223 struct vnode *vp; 1224 int error, n; 1225 bool last; 1226 1227 vp = ap->a_vp; 1228 1229 /* 1230 * Avoid the interlock if this is not the last exec. 1231 */ 1232 n = atomic_load_int(&vp->v_writecount); 1233 for (;;) { 1234 if (n >= -1) { 1235 break; 1236 } 1237 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) { 1238 return (0); 1239 } 1240 } 1241 1242 last = false; 1243 VI_LOCK(vp); 1244 if (vp->v_writecount < 0) { 1245 if ((vp->v_iflag & VI_TEXT_REF) != 0 && 1246 vp->v_writecount == -1) { 1247 last = true; 1248 vp->v_iflag &= ~VI_TEXT_REF; 1249 } 1250 atomic_add_int(&vp->v_writecount, 1); 1251 error = 0; 1252 } else { 1253 error = EINVAL; 1254 } 1255 VI_UNLOCK(vp); 1256 if (last) 1257 vunref(vp); 1258 return (error); 1259 } 1260 1261 static int 1262 vop_stdadd_writecount(struct vop_add_writecount_args *ap) 1263 { 1264 struct vnode *vp; 1265 struct mount *mp; 1266 int error; 1267 1268 vp = ap->a_vp; 1269 VI_LOCK_FLAGS(vp, MTX_DUPOK); 1270 if (vp->v_writecount < 0) { 1271 error = ETXTBSY; 1272 } else { 1273 VNASSERT(vp->v_writecount + ap->a_inc >= 0, vp, 1274 ("neg writecount increment %d", ap->a_inc)); 1275 if (vp->v_writecount == 0) { 1276 mp = vp->v_mount; 1277 if (mp != NULL && (mp->mnt_kern_flag & MNTK_NOMSYNC) == 0) 1278 vlazy(vp); 1279 } 1280 vp->v_writecount += ap->a_inc; 1281 error = 0; 1282 } 1283 VI_UNLOCK(vp); 1284 return (error); 1285 } 1286 1287 int 1288 vop_stdneed_inactive(struct vop_need_inactive_args *ap) 1289 { 1290 1291 return (1); 1292 } 1293 1294 int 1295 vop_stdioctl(struct vop_ioctl_args *ap) 1296 { 1297 struct vnode *vp; 1298 struct vattr va; 1299 off_t *offp; 1300 int error; 1301 1302 switch (ap->a_command) { 1303 case FIOSEEKDATA: 1304 case FIOSEEKHOLE: 1305 vp = ap->a_vp; 1306 error = vn_lock(vp, LK_SHARED); 1307 if (error != 0) 1308 return (EBADF); 1309 if (vp->v_type == VREG) 1310 error = VOP_GETATTR(vp, &va, ap->a_cred); 1311 else 1312 error = ENOTTY; 1313 if (error == 0) { 1314 offp = ap->a_data; 1315 if (*offp < 0 || *offp >= va.va_size) 1316 error = ENXIO; 1317 else if (ap->a_command == FIOSEEKHOLE) 1318 *offp = va.va_size; 1319 } 1320 VOP_UNLOCK(vp); 1321 break; 1322 default: 1323 error = ENOTTY; 1324 break; 1325 } 1326 return (error); 1327 } 1328 1329 /* 1330 * vfs default ops 1331 * used to fill the vfs function table to get reasonable default return values. 1332 */ 1333 int 1334 vfs_stdroot (mp, flags, vpp) 1335 struct mount *mp; 1336 int flags; 1337 struct vnode **vpp; 1338 { 1339 1340 return (EOPNOTSUPP); 1341 } 1342 1343 int 1344 vfs_stdstatfs (mp, sbp) 1345 struct mount *mp; 1346 struct statfs *sbp; 1347 { 1348 1349 return (EOPNOTSUPP); 1350 } 1351 1352 int 1353 vfs_stdquotactl (mp, cmds, uid, arg, mp_busy) 1354 struct mount *mp; 1355 int cmds; 1356 uid_t uid; 1357 void *arg; 1358 bool *mp_busy; 1359 { 1360 return (EOPNOTSUPP); 1361 } 1362 1363 int 1364 vfs_stdsync(mp, waitfor) 1365 struct mount *mp; 1366 int waitfor; 1367 { 1368 struct vnode *vp, *mvp; 1369 struct thread *td; 1370 int error, lockreq, allerror = 0; 1371 1372 td = curthread; 1373 lockreq = LK_EXCLUSIVE | LK_INTERLOCK; 1374 if (waitfor != MNT_WAIT) 1375 lockreq |= LK_NOWAIT; 1376 /* 1377 * Force stale buffer cache information to be flushed. 1378 */ 1379 loop: 1380 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 1381 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) { 1382 VI_UNLOCK(vp); 1383 continue; 1384 } 1385 if ((error = vget(vp, lockreq)) != 0) { 1386 if (error == ENOENT) { 1387 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 1388 goto loop; 1389 } 1390 continue; 1391 } 1392 error = VOP_FSYNC(vp, waitfor, td); 1393 if (error) 1394 allerror = error; 1395 vput(vp); 1396 } 1397 return (allerror); 1398 } 1399 1400 int 1401 vfs_stdnosync (mp, waitfor) 1402 struct mount *mp; 1403 int waitfor; 1404 { 1405 1406 return (0); 1407 } 1408 1409 static int 1410 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap) 1411 { 1412 int error; 1413 1414 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp, 1415 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred, 1416 ap->a_outcred, ap->a_fsizetd); 1417 return (error); 1418 } 1419 1420 int 1421 vfs_stdvget (mp, ino, flags, vpp) 1422 struct mount *mp; 1423 ino_t ino; 1424 int flags; 1425 struct vnode **vpp; 1426 { 1427 1428 return (EOPNOTSUPP); 1429 } 1430 1431 int 1432 vfs_stdfhtovp (mp, fhp, flags, vpp) 1433 struct mount *mp; 1434 struct fid *fhp; 1435 int flags; 1436 struct vnode **vpp; 1437 { 1438 1439 return (EOPNOTSUPP); 1440 } 1441 1442 int 1443 vfs_stdinit (vfsp) 1444 struct vfsconf *vfsp; 1445 { 1446 1447 return (0); 1448 } 1449 1450 int 1451 vfs_stduninit (vfsp) 1452 struct vfsconf *vfsp; 1453 { 1454 1455 return(0); 1456 } 1457 1458 int 1459 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname) 1460 struct mount *mp; 1461 int cmd; 1462 struct vnode *filename_vp; 1463 int attrnamespace; 1464 const char *attrname; 1465 { 1466 1467 if (filename_vp != NULL) 1468 VOP_UNLOCK(filename_vp); 1469 return (EOPNOTSUPP); 1470 } 1471 1472 int 1473 vfs_stdsysctl(mp, op, req) 1474 struct mount *mp; 1475 fsctlop_t op; 1476 struct sysctl_req *req; 1477 { 1478 1479 return (EOPNOTSUPP); 1480 } 1481 1482 static vop_bypass_t * 1483 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a) 1484 { 1485 1486 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset)); 1487 } 1488 1489 int 1490 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a) 1491 { 1492 vop_bypass_t *bp; 1493 int prev_stops, rc; 1494 1495 bp = bp_by_off(vop, a); 1496 MPASS(bp != NULL); 1497 1498 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 1499 rc = bp(a); 1500 sigallowstop(prev_stops); 1501 return (rc); 1502 } 1503 1504 static int 1505 vop_stdstat(struct vop_stat_args *a) 1506 { 1507 struct vattr vattr; 1508 struct vattr *vap; 1509 struct vnode *vp; 1510 struct stat *sb; 1511 int error; 1512 u_short mode; 1513 1514 vp = a->a_vp; 1515 sb = a->a_sb; 1516 1517 error = vop_stat_helper_pre(a); 1518 if (error != 0) 1519 return (error); 1520 1521 vap = &vattr; 1522 1523 /* 1524 * Initialize defaults for new and unusual fields, so that file 1525 * systems which don't support these fields don't need to know 1526 * about them. 1527 */ 1528 vap->va_birthtime.tv_sec = -1; 1529 vap->va_birthtime.tv_nsec = 0; 1530 vap->va_fsid = VNOVAL; 1531 vap->va_gen = 0; 1532 vap->va_rdev = NODEV; 1533 1534 error = VOP_GETATTR(vp, vap, a->a_active_cred); 1535 if (error) 1536 goto out; 1537 1538 /* 1539 * Zero the spare stat fields 1540 */ 1541 bzero(sb, sizeof *sb); 1542 1543 /* 1544 * Copy from vattr table 1545 */ 1546 if (vap->va_fsid != VNOVAL) 1547 sb->st_dev = vap->va_fsid; 1548 else 1549 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0]; 1550 sb->st_ino = vap->va_fileid; 1551 mode = vap->va_mode; 1552 switch (vap->va_type) { 1553 case VREG: 1554 mode |= S_IFREG; 1555 break; 1556 case VDIR: 1557 mode |= S_IFDIR; 1558 break; 1559 case VBLK: 1560 mode |= S_IFBLK; 1561 break; 1562 case VCHR: 1563 mode |= S_IFCHR; 1564 break; 1565 case VLNK: 1566 mode |= S_IFLNK; 1567 break; 1568 case VSOCK: 1569 mode |= S_IFSOCK; 1570 break; 1571 case VFIFO: 1572 mode |= S_IFIFO; 1573 break; 1574 default: 1575 error = EBADF; 1576 goto out; 1577 } 1578 sb->st_mode = mode; 1579 sb->st_nlink = vap->va_nlink; 1580 sb->st_uid = vap->va_uid; 1581 sb->st_gid = vap->va_gid; 1582 sb->st_rdev = vap->va_rdev; 1583 if (vap->va_size > OFF_MAX) { 1584 error = EOVERFLOW; 1585 goto out; 1586 } 1587 sb->st_size = vap->va_size; 1588 sb->st_atim.tv_sec = vap->va_atime.tv_sec; 1589 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec; 1590 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec; 1591 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec; 1592 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec; 1593 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec; 1594 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec; 1595 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec; 1596 1597 /* 1598 * According to www.opengroup.org, the meaning of st_blksize is 1599 * "a filesystem-specific preferred I/O block size for this 1600 * object. In some filesystem types, this may vary from file 1601 * to file" 1602 * Use minimum/default of PAGE_SIZE (e.g. for VCHR). 1603 */ 1604 1605 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize); 1606 sb->st_flags = vap->va_flags; 1607 sb->st_blocks = vap->va_bytes / S_BLKSIZE; 1608 sb->st_gen = vap->va_gen; 1609 out: 1610 return (vop_stat_helper_post(a, error)); 1611 } 1612 1613 static int 1614 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused) 1615 { 1616 return (EJUSTRETURN); 1617 } 1618 1619 static int 1620 vop_stdvput_pair(struct vop_vput_pair_args *ap) 1621 { 1622 struct vnode *dvp, *vp, **vpp; 1623 1624 dvp = ap->a_dvp; 1625 vpp = ap->a_vpp; 1626 vput(dvp); 1627 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL) 1628 vput(vp); 1629 return (0); 1630 } 1631