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