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 static int vop_stdioctl(struct vop_ioctl_args *ap); 91 92 /* 93 * This vnode table stores what we want to do if the filesystem doesn't 94 * implement a particular VOP. 95 * 96 * If there is no specific entry here, we will return EOPNOTSUPP. 97 * 98 * Note that every filesystem has to implement either vop_access 99 * or vop_accessx; failing to do so will result in immediate crash 100 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(), 101 * which calls vop_stdaccess() etc. 102 */ 103 104 struct vop_vector default_vnodeops = { 105 .vop_default = NULL, 106 .vop_bypass = VOP_EOPNOTSUPP, 107 108 .vop_access = vop_stdaccess, 109 .vop_accessx = vop_stdaccessx, 110 .vop_advise = vop_stdadvise, 111 .vop_advlock = vop_stdadvlock, 112 .vop_advlockasync = vop_stdadvlockasync, 113 .vop_advlockpurge = vop_stdadvlockpurge, 114 .vop_allocate = vop_stdallocate, 115 .vop_bmap = vop_stdbmap, 116 .vop_close = VOP_NULL, 117 .vop_fsync = VOP_NULL, 118 .vop_fdatasync = vop_stdfdatasync, 119 .vop_getpages = vop_stdgetpages, 120 .vop_getpages_async = vop_stdgetpages_async, 121 .vop_getwritemount = vop_stdgetwritemount, 122 .vop_inactive = VOP_NULL, 123 .vop_need_inactive = vop_stdneed_inactive, 124 .vop_ioctl = vop_stdioctl, 125 .vop_kqfilter = vop_stdkqfilter, 126 .vop_islocked = vop_stdislocked, 127 .vop_lock1 = vop_stdlock, 128 .vop_lookup = vop_nolookup, 129 .vop_open = VOP_NULL, 130 .vop_pathconf = VOP_EINVAL, 131 .vop_poll = vop_nopoll, 132 .vop_putpages = vop_stdputpages, 133 .vop_readlink = VOP_EINVAL, 134 .vop_rename = vop_norename, 135 .vop_revoke = VOP_PANIC, 136 .vop_strategy = vop_nostrategy, 137 .vop_unlock = vop_stdunlock, 138 .vop_vptocnp = vop_stdvptocnp, 139 .vop_vptofh = vop_stdvptofh, 140 .vop_unp_bind = vop_stdunp_bind, 141 .vop_unp_connect = vop_stdunp_connect, 142 .vop_unp_detach = vop_stdunp_detach, 143 .vop_is_text = vop_stdis_text, 144 .vop_set_text = vop_stdset_text, 145 .vop_unset_text = vop_stdunset_text, 146 .vop_add_writecount = vop_stdadd_writecount, 147 .vop_copy_file_range = vop_stdcopy_file_range, 148 }; 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, 0); 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, 0); 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 int a_flags; 526 } */ *ap; 527 { 528 struct vnode *vp = ap->a_vp; 529 struct mtx *ilk; 530 531 ilk = VI_MTX(vp); 532 return (lockmgr_unlock_fast_path(vp->v_vnlock, ap->a_flags, 533 &ilk->lock_object)); 534 } 535 536 /* See above. */ 537 int 538 vop_stdislocked(ap) 539 struct vop_islocked_args /* { 540 struct vnode *a_vp; 541 } */ *ap; 542 { 543 544 return (lockstatus(ap->a_vp->v_vnlock)); 545 } 546 547 /* 548 * Return true for select/poll. 549 */ 550 int 551 vop_nopoll(ap) 552 struct vop_poll_args /* { 553 struct vnode *a_vp; 554 int a_events; 555 struct ucred *a_cred; 556 struct thread *a_td; 557 } */ *ap; 558 { 559 560 return (poll_no_poll(ap->a_events)); 561 } 562 563 /* 564 * Implement poll for local filesystems that support it. 565 */ 566 int 567 vop_stdpoll(ap) 568 struct vop_poll_args /* { 569 struct vnode *a_vp; 570 int a_events; 571 struct ucred *a_cred; 572 struct thread *a_td; 573 } */ *ap; 574 { 575 if (ap->a_events & ~POLLSTANDARD) 576 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 577 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 578 } 579 580 /* 581 * Return our mount point, as we will take charge of the writes. 582 */ 583 int 584 vop_stdgetwritemount(ap) 585 struct vop_getwritemount_args /* { 586 struct vnode *a_vp; 587 struct mount **a_mpp; 588 } */ *ap; 589 { 590 struct mount *mp; 591 struct vnode *vp; 592 593 /* 594 * Note that having a reference does not prevent forced unmount from 595 * setting ->v_mount to NULL after the lock gets released. This is of 596 * no consequence for typical consumers (most notably vn_start_write) 597 * since in this case the vnode is VI_DOOMED. Unmount might have 598 * progressed far enough that its completion is only delayed by the 599 * reference obtained here. The consumer only needs to concern itself 600 * with releasing it. 601 */ 602 vp = ap->a_vp; 603 mp = vp->v_mount; 604 MNT_ILOCK(mp); 605 if (mp != vp->v_mount) { 606 MNT_IUNLOCK(mp); 607 mp = NULL; 608 goto out; 609 } 610 MNT_REF(mp); 611 MNT_IUNLOCK(mp); 612 out: 613 *(ap->a_mpp) = mp; 614 return (0); 615 } 616 617 /* 618 * If the file system doesn't implement VOP_BMAP, then return sensible defaults: 619 * - Return the vnode's bufobj instead of any underlying device's bufobj 620 * - Calculate the physical block number as if there were equal size 621 * consecutive blocks, but 622 * - Report no contiguous runs of blocks. 623 */ 624 int 625 vop_stdbmap(ap) 626 struct vop_bmap_args /* { 627 struct vnode *a_vp; 628 daddr_t a_bn; 629 struct bufobj **a_bop; 630 daddr_t *a_bnp; 631 int *a_runp; 632 int *a_runb; 633 } */ *ap; 634 { 635 636 if (ap->a_bop != NULL) 637 *ap->a_bop = &ap->a_vp->v_bufobj; 638 if (ap->a_bnp != NULL) 639 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 640 if (ap->a_runp != NULL) 641 *ap->a_runp = 0; 642 if (ap->a_runb != NULL) 643 *ap->a_runb = 0; 644 return (0); 645 } 646 647 int 648 vop_stdfsync(ap) 649 struct vop_fsync_args /* { 650 struct vnode *a_vp; 651 int a_waitfor; 652 struct thread *a_td; 653 } */ *ap; 654 { 655 656 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor)); 657 } 658 659 static int 660 vop_stdfdatasync(struct vop_fdatasync_args *ap) 661 { 662 663 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td)); 664 } 665 666 int 667 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap) 668 { 669 670 return (vn_fsync_buf(ap->a_vp, MNT_WAIT)); 671 } 672 673 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */ 674 int 675 vop_stdgetpages(ap) 676 struct vop_getpages_args /* { 677 struct vnode *a_vp; 678 vm_page_t *a_m; 679 int a_count; 680 int *a_rbehind; 681 int *a_rahead; 682 } */ *ap; 683 { 684 685 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, 686 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL); 687 } 688 689 static int 690 vop_stdgetpages_async(struct vop_getpages_async_args *ap) 691 { 692 int error; 693 694 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind, 695 ap->a_rahead); 696 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error); 697 return (error); 698 } 699 700 int 701 vop_stdkqfilter(struct vop_kqfilter_args *ap) 702 { 703 return vfs_kqfilter(ap); 704 } 705 706 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */ 707 int 708 vop_stdputpages(ap) 709 struct vop_putpages_args /* { 710 struct vnode *a_vp; 711 vm_page_t *a_m; 712 int a_count; 713 int a_sync; 714 int *a_rtvals; 715 } */ *ap; 716 { 717 718 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count, 719 ap->a_sync, ap->a_rtvals); 720 } 721 722 int 723 vop_stdvptofh(struct vop_vptofh_args *ap) 724 { 725 return (EOPNOTSUPP); 726 } 727 728 int 729 vop_stdvptocnp(struct vop_vptocnp_args *ap) 730 { 731 struct vnode *vp = ap->a_vp; 732 struct vnode **dvp = ap->a_vpp; 733 struct ucred *cred = ap->a_cred; 734 char *buf = ap->a_buf; 735 int *buflen = ap->a_buflen; 736 char *dirbuf, *cpos; 737 int i, error, eofflag, dirbuflen, flags, locked, len, covered; 738 off_t off; 739 ino_t fileno; 740 struct vattr va; 741 struct nameidata nd; 742 struct thread *td; 743 struct dirent *dp; 744 struct vnode *mvp; 745 746 i = *buflen; 747 error = 0; 748 covered = 0; 749 td = curthread; 750 751 if (vp->v_type != VDIR) 752 return (ENOENT); 753 754 error = VOP_GETATTR(vp, &va, cred); 755 if (error) 756 return (error); 757 758 VREF(vp); 759 locked = VOP_ISLOCKED(vp); 760 VOP_UNLOCK(vp, 0); 761 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE, 762 "..", vp, td); 763 flags = FREAD; 764 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL); 765 if (error) { 766 vn_lock(vp, locked | LK_RETRY); 767 return (error); 768 } 769 NDFREE(&nd, NDF_ONLY_PNBUF); 770 771 mvp = *dvp = nd.ni_vp; 772 773 if (vp->v_mount != (*dvp)->v_mount && 774 ((*dvp)->v_vflag & VV_ROOT) && 775 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) { 776 *dvp = (*dvp)->v_mount->mnt_vnodecovered; 777 VREF(mvp); 778 VOP_UNLOCK(mvp, 0); 779 vn_close(mvp, FREAD, cred, td); 780 VREF(*dvp); 781 vn_lock(*dvp, LK_SHARED | LK_RETRY); 782 covered = 1; 783 } 784 785 fileno = va.va_fileid; 786 787 dirbuflen = DEV_BSIZE; 788 if (dirbuflen < va.va_blocksize) 789 dirbuflen = va.va_blocksize; 790 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK); 791 792 if ((*dvp)->v_type != VDIR) { 793 error = ENOENT; 794 goto out; 795 } 796 797 off = 0; 798 len = 0; 799 do { 800 /* call VOP_READDIR of parent */ 801 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off, 802 &cpos, &len, &eofflag, td); 803 if (error) 804 goto out; 805 806 if ((dp->d_type != DT_WHT) && 807 (dp->d_fileno == fileno)) { 808 if (covered) { 809 VOP_UNLOCK(*dvp, 0); 810 vn_lock(mvp, LK_SHARED | LK_RETRY); 811 if (dirent_exists(mvp, dp->d_name, td)) { 812 error = ENOENT; 813 VOP_UNLOCK(mvp, 0); 814 vn_lock(*dvp, LK_SHARED | LK_RETRY); 815 goto out; 816 } 817 VOP_UNLOCK(mvp, 0); 818 vn_lock(*dvp, LK_SHARED | LK_RETRY); 819 } 820 i -= dp->d_namlen; 821 822 if (i < 0) { 823 error = ENOMEM; 824 goto out; 825 } 826 if (dp->d_namlen == 1 && dp->d_name[0] == '.') { 827 error = ENOENT; 828 } else { 829 bcopy(dp->d_name, buf + i, dp->d_namlen); 830 error = 0; 831 } 832 goto out; 833 } 834 } while (len > 0 || !eofflag); 835 error = ENOENT; 836 837 out: 838 free(dirbuf, M_TEMP); 839 if (!error) { 840 *buflen = i; 841 vref(*dvp); 842 } 843 if (covered) { 844 vput(*dvp); 845 vrele(mvp); 846 } else { 847 VOP_UNLOCK(mvp, 0); 848 vn_close(mvp, FREAD, cred, td); 849 } 850 vn_lock(vp, locked | LK_RETRY); 851 return (error); 852 } 853 854 int 855 vop_stdallocate(struct vop_allocate_args *ap) 856 { 857 #ifdef __notyet__ 858 struct statfs *sfs; 859 off_t maxfilesize = 0; 860 #endif 861 struct iovec aiov; 862 struct vattr vattr, *vap; 863 struct uio auio; 864 off_t fsize, len, cur, offset; 865 uint8_t *buf; 866 struct thread *td; 867 struct vnode *vp; 868 size_t iosize; 869 int error; 870 871 buf = NULL; 872 error = 0; 873 td = curthread; 874 vap = &vattr; 875 vp = ap->a_vp; 876 len = *ap->a_len; 877 offset = *ap->a_offset; 878 879 error = VOP_GETATTR(vp, vap, td->td_ucred); 880 if (error != 0) 881 goto out; 882 fsize = vap->va_size; 883 iosize = vap->va_blocksize; 884 if (iosize == 0) 885 iosize = BLKDEV_IOSIZE; 886 if (iosize > MAXPHYS) 887 iosize = MAXPHYS; 888 buf = malloc(iosize, M_TEMP, M_WAITOK); 889 890 #ifdef __notyet__ 891 /* 892 * Check if the filesystem sets f_maxfilesize; if not use 893 * VOP_SETATTR to perform the check. 894 */ 895 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK); 896 error = VFS_STATFS(vp->v_mount, sfs, td); 897 if (error == 0) 898 maxfilesize = sfs->f_maxfilesize; 899 free(sfs, M_STATFS); 900 if (error != 0) 901 goto out; 902 if (maxfilesize) { 903 if (offset > maxfilesize || len > maxfilesize || 904 offset + len > maxfilesize) { 905 error = EFBIG; 906 goto out; 907 } 908 } else 909 #endif 910 if (offset + len > vap->va_size) { 911 /* 912 * Test offset + len against the filesystem's maxfilesize. 913 */ 914 VATTR_NULL(vap); 915 vap->va_size = offset + len; 916 error = VOP_SETATTR(vp, vap, td->td_ucred); 917 if (error != 0) 918 goto out; 919 VATTR_NULL(vap); 920 vap->va_size = fsize; 921 error = VOP_SETATTR(vp, vap, td->td_ucred); 922 if (error != 0) 923 goto out; 924 } 925 926 for (;;) { 927 /* 928 * Read and write back anything below the nominal file 929 * size. There's currently no way outside the filesystem 930 * to know whether this area is sparse or not. 931 */ 932 cur = iosize; 933 if ((offset % iosize) != 0) 934 cur -= (offset % iosize); 935 if (cur > len) 936 cur = len; 937 if (offset < fsize) { 938 aiov.iov_base = buf; 939 aiov.iov_len = cur; 940 auio.uio_iov = &aiov; 941 auio.uio_iovcnt = 1; 942 auio.uio_offset = offset; 943 auio.uio_resid = cur; 944 auio.uio_segflg = UIO_SYSSPACE; 945 auio.uio_rw = UIO_READ; 946 auio.uio_td = td; 947 error = VOP_READ(vp, &auio, 0, td->td_ucred); 948 if (error != 0) 949 break; 950 if (auio.uio_resid > 0) { 951 bzero(buf + cur - auio.uio_resid, 952 auio.uio_resid); 953 } 954 } else { 955 bzero(buf, cur); 956 } 957 958 aiov.iov_base = buf; 959 aiov.iov_len = cur; 960 auio.uio_iov = &aiov; 961 auio.uio_iovcnt = 1; 962 auio.uio_offset = offset; 963 auio.uio_resid = cur; 964 auio.uio_segflg = UIO_SYSSPACE; 965 auio.uio_rw = UIO_WRITE; 966 auio.uio_td = td; 967 968 error = VOP_WRITE(vp, &auio, 0, td->td_ucred); 969 if (error != 0) 970 break; 971 972 len -= cur; 973 offset += cur; 974 if (len == 0) 975 break; 976 if (should_yield()) 977 break; 978 } 979 980 out: 981 *ap->a_len = len; 982 *ap->a_offset = offset; 983 free(buf, M_TEMP); 984 return (error); 985 } 986 987 int 988 vop_stdadvise(struct vop_advise_args *ap) 989 { 990 struct vnode *vp; 991 struct bufobj *bo; 992 daddr_t startn, endn; 993 off_t bstart, bend, start, end; 994 int bsize, error; 995 996 vp = ap->a_vp; 997 switch (ap->a_advice) { 998 case POSIX_FADV_WILLNEED: 999 /* 1000 * Do nothing for now. Filesystems should provide a 1001 * custom method which starts an asynchronous read of 1002 * the requested region. 1003 */ 1004 error = 0; 1005 break; 1006 case POSIX_FADV_DONTNEED: 1007 error = 0; 1008 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1009 if (vp->v_iflag & VI_DOOMED) { 1010 VOP_UNLOCK(vp, 0); 1011 break; 1012 } 1013 1014 /* 1015 * Round to block boundaries (and later possibly further to 1016 * page boundaries). Applications cannot reasonably be aware 1017 * of the boundaries, and the rounding must be to expand at 1018 * both extremities to cover enough. It still doesn't cover 1019 * read-ahead. For partial blocks, this gives unnecessary 1020 * discarding of buffers but is efficient enough since the 1021 * pages usually remain in VMIO for some time. 1022 */ 1023 bsize = vp->v_bufobj.bo_bsize; 1024 bstart = rounddown(ap->a_start, bsize); 1025 bend = roundup(ap->a_end, bsize); 1026 1027 /* 1028 * Deactivate pages in the specified range from the backing VM 1029 * object. Pages that are resident in the buffer cache will 1030 * remain wired until their corresponding buffers are released 1031 * below. 1032 */ 1033 if (vp->v_object != NULL) { 1034 start = trunc_page(bstart); 1035 end = round_page(bend); 1036 VM_OBJECT_RLOCK(vp->v_object); 1037 vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start), 1038 OFF_TO_IDX(end)); 1039 VM_OBJECT_RUNLOCK(vp->v_object); 1040 } 1041 1042 bo = &vp->v_bufobj; 1043 BO_RLOCK(bo); 1044 startn = bstart / bsize; 1045 endn = bend / bsize; 1046 error = bnoreuselist(&bo->bo_clean, bo, startn, endn); 1047 if (error == 0) 1048 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn); 1049 BO_RUNLOCK(bo); 1050 VOP_UNLOCK(vp, 0); 1051 break; 1052 default: 1053 error = EINVAL; 1054 break; 1055 } 1056 return (error); 1057 } 1058 1059 int 1060 vop_stdunp_bind(struct vop_unp_bind_args *ap) 1061 { 1062 1063 ap->a_vp->v_unpcb = ap->a_unpcb; 1064 return (0); 1065 } 1066 1067 int 1068 vop_stdunp_connect(struct vop_unp_connect_args *ap) 1069 { 1070 1071 *ap->a_unpcb = ap->a_vp->v_unpcb; 1072 return (0); 1073 } 1074 1075 int 1076 vop_stdunp_detach(struct vop_unp_detach_args *ap) 1077 { 1078 1079 ap->a_vp->v_unpcb = NULL; 1080 return (0); 1081 } 1082 1083 static int 1084 vop_stdis_text(struct vop_is_text_args *ap) 1085 { 1086 1087 return (ap->a_vp->v_writecount < 0); 1088 } 1089 1090 int 1091 vop_stdset_text(struct vop_set_text_args *ap) 1092 { 1093 struct vnode *vp; 1094 struct mount *mp; 1095 int error; 1096 1097 vp = ap->a_vp; 1098 VI_LOCK(vp); 1099 if (vp->v_writecount > 0) { 1100 error = ETXTBSY; 1101 } else { 1102 /* 1103 * If requested by fs, keep a use reference to the 1104 * vnode until the last text reference is released. 1105 */ 1106 mp = vp->v_mount; 1107 if (mp != NULL && (mp->mnt_kern_flag & MNTK_TEXT_REFS) != 0 && 1108 vp->v_writecount == 0) { 1109 vp->v_iflag |= VI_TEXT_REF; 1110 vrefl(vp); 1111 } 1112 1113 vp->v_writecount--; 1114 error = 0; 1115 } 1116 VI_UNLOCK(vp); 1117 return (error); 1118 } 1119 1120 static int 1121 vop_stdunset_text(struct vop_unset_text_args *ap) 1122 { 1123 struct vnode *vp; 1124 int error; 1125 bool last; 1126 1127 vp = ap->a_vp; 1128 last = false; 1129 VI_LOCK(vp); 1130 if (vp->v_writecount < 0) { 1131 if ((vp->v_iflag & VI_TEXT_REF) != 0 && 1132 vp->v_writecount == -1) { 1133 last = true; 1134 vp->v_iflag &= ~VI_TEXT_REF; 1135 } 1136 vp->v_writecount++; 1137 error = 0; 1138 } else { 1139 error = EINVAL; 1140 } 1141 VI_UNLOCK(vp); 1142 if (last) 1143 vunref(vp); 1144 return (error); 1145 } 1146 1147 static int 1148 vop_stdadd_writecount(struct vop_add_writecount_args *ap) 1149 { 1150 struct vnode *vp; 1151 int error; 1152 1153 vp = ap->a_vp; 1154 VI_LOCK_FLAGS(vp, MTX_DUPOK); 1155 if (vp->v_writecount < 0) { 1156 error = ETXTBSY; 1157 } else { 1158 VNASSERT(vp->v_writecount + ap->a_inc >= 0, vp, 1159 ("neg writecount increment %d", ap->a_inc)); 1160 vp->v_writecount += ap->a_inc; 1161 error = 0; 1162 } 1163 VI_UNLOCK(vp); 1164 return (error); 1165 } 1166 1167 int 1168 vop_stdneed_inactive(struct vop_need_inactive_args *ap) 1169 { 1170 1171 return (1); 1172 } 1173 1174 static int 1175 vop_stdioctl(struct vop_ioctl_args *ap) 1176 { 1177 struct vnode *vp; 1178 struct vattr va; 1179 off_t *offp; 1180 int error; 1181 1182 switch (ap->a_command) { 1183 case FIOSEEKDATA: 1184 case FIOSEEKHOLE: 1185 vp = ap->a_vp; 1186 error = vn_lock(vp, LK_SHARED); 1187 if (error != 0) 1188 return (EBADF); 1189 if (vp->v_type == VREG) 1190 error = VOP_GETATTR(vp, &va, ap->a_cred); 1191 else 1192 error = ENOTTY; 1193 if (error == 0) { 1194 offp = ap->a_data; 1195 if (*offp < 0 || *offp >= va.va_size) 1196 error = ENXIO; 1197 else if (ap->a_command == FIOSEEKHOLE) 1198 *offp = va.va_size; 1199 } 1200 VOP_UNLOCK(vp, 0); 1201 break; 1202 default: 1203 error = ENOTTY; 1204 break; 1205 } 1206 return (error); 1207 } 1208 1209 /* 1210 * vfs default ops 1211 * used to fill the vfs function table to get reasonable default return values. 1212 */ 1213 int 1214 vfs_stdroot (mp, flags, vpp) 1215 struct mount *mp; 1216 int flags; 1217 struct vnode **vpp; 1218 { 1219 1220 return (EOPNOTSUPP); 1221 } 1222 1223 int 1224 vfs_stdstatfs (mp, sbp) 1225 struct mount *mp; 1226 struct statfs *sbp; 1227 { 1228 1229 return (EOPNOTSUPP); 1230 } 1231 1232 int 1233 vfs_stdquotactl (mp, cmds, uid, arg) 1234 struct mount *mp; 1235 int cmds; 1236 uid_t uid; 1237 void *arg; 1238 { 1239 1240 return (EOPNOTSUPP); 1241 } 1242 1243 int 1244 vfs_stdsync(mp, waitfor) 1245 struct mount *mp; 1246 int waitfor; 1247 { 1248 struct vnode *vp, *mvp; 1249 struct thread *td; 1250 int error, lockreq, allerror = 0; 1251 1252 td = curthread; 1253 lockreq = LK_EXCLUSIVE | LK_INTERLOCK; 1254 if (waitfor != MNT_WAIT) 1255 lockreq |= LK_NOWAIT; 1256 /* 1257 * Force stale buffer cache information to be flushed. 1258 */ 1259 loop: 1260 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 1261 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) { 1262 VI_UNLOCK(vp); 1263 continue; 1264 } 1265 if ((error = vget(vp, lockreq, td)) != 0) { 1266 if (error == ENOENT) { 1267 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 1268 goto loop; 1269 } 1270 continue; 1271 } 1272 error = VOP_FSYNC(vp, waitfor, td); 1273 if (error) 1274 allerror = error; 1275 vput(vp); 1276 } 1277 return (allerror); 1278 } 1279 1280 int 1281 vfs_stdnosync (mp, waitfor) 1282 struct mount *mp; 1283 int waitfor; 1284 { 1285 1286 return (0); 1287 } 1288 1289 static int 1290 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap) 1291 { 1292 int error; 1293 1294 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp, 1295 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred, 1296 ap->a_outcred, ap->a_fsizetd); 1297 return (error); 1298 } 1299 1300 int 1301 vfs_stdvget (mp, ino, flags, vpp) 1302 struct mount *mp; 1303 ino_t ino; 1304 int flags; 1305 struct vnode **vpp; 1306 { 1307 1308 return (EOPNOTSUPP); 1309 } 1310 1311 int 1312 vfs_stdfhtovp (mp, fhp, flags, vpp) 1313 struct mount *mp; 1314 struct fid *fhp; 1315 int flags; 1316 struct vnode **vpp; 1317 { 1318 1319 return (EOPNOTSUPP); 1320 } 1321 1322 int 1323 vfs_stdinit (vfsp) 1324 struct vfsconf *vfsp; 1325 { 1326 1327 return (0); 1328 } 1329 1330 int 1331 vfs_stduninit (vfsp) 1332 struct vfsconf *vfsp; 1333 { 1334 1335 return(0); 1336 } 1337 1338 int 1339 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname) 1340 struct mount *mp; 1341 int cmd; 1342 struct vnode *filename_vp; 1343 int attrnamespace; 1344 const char *attrname; 1345 { 1346 1347 if (filename_vp != NULL) 1348 VOP_UNLOCK(filename_vp, 0); 1349 return (EOPNOTSUPP); 1350 } 1351 1352 int 1353 vfs_stdsysctl(mp, op, req) 1354 struct mount *mp; 1355 fsctlop_t op; 1356 struct sysctl_req *req; 1357 { 1358 1359 return (EOPNOTSUPP); 1360 } 1361 1362 static vop_bypass_t * 1363 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a) 1364 { 1365 1366 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset)); 1367 } 1368 1369 int 1370 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a) 1371 { 1372 vop_bypass_t *bp; 1373 int prev_stops, rc; 1374 1375 for (; vop != NULL; vop = vop->vop_default) { 1376 bp = bp_by_off(vop, a); 1377 if (bp != NULL) 1378 break; 1379 1380 /* 1381 * Bypass is not really supported. It is done for 1382 * fallback to unimplemented vops in the default 1383 * vector. 1384 */ 1385 bp = vop->vop_bypass; 1386 if (bp != NULL) 1387 break; 1388 } 1389 MPASS(bp != NULL); 1390 1391 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 1392 rc = bp(a); 1393 sigallowstop(prev_stops); 1394 return (rc); 1395 } 1396