vfs_default.c revision 206094
11590Srgrimes/*- 21590Srgrimes * Copyright (c) 1989, 1993 31590Srgrimes * The Regents of the University of California. All rights reserved. 41590Srgrimes * 51590Srgrimes * This code is derived from software contributed 61590Srgrimes * to Berkeley by John Heidemann of the UCLA Ficus project. 71590Srgrimes * 81590Srgrimes * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project 91590Srgrimes * 101590Srgrimes * Redistribution and use in source and binary forms, with or without 111590Srgrimes * modification, are permitted provided that the following conditions 121590Srgrimes * are met: 131590Srgrimes * 1. Redistributions of source code must retain the above copyright 141590Srgrimes * notice, this list of conditions and the following disclaimer. 151590Srgrimes * 2. Redistributions in binary form must reproduce the above copyright 161590Srgrimes * notice, this list of conditions and the following disclaimer in the 171590Srgrimes * documentation and/or other materials provided with the distribution. 181590Srgrimes * 4. Neither the name of the University nor the names of its contributors 191590Srgrimes * may be used to endorse or promote products derived from this software 201590Srgrimes * without specific prior written permission. 211590Srgrimes * 221590Srgrimes * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 231590Srgrimes * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 241590Srgrimes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 251590Srgrimes * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 261590Srgrimes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 271590Srgrimes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 281590Srgrimes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 291590Srgrimes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 3060583Sphk * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 3160583Sphk * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 3260583Sphk * SUCH DAMAGE. 331590Srgrimes */ 341590Srgrimes 351590Srgrimes#include <sys/cdefs.h> 361590Srgrimes__FBSDID("$FreeBSD: head/sys/kern/vfs_default.c 206094 2010-04-02 14:03:43Z kib $"); 371590Srgrimes 381590Srgrimes#include <sys/param.h> 39149388Sbrian#include <sys/systm.h> 4092920Simp#include <sys/bio.h> 4192920Simp#include <sys/buf.h> 4292920Simp#include <sys/conf.h> 4392920Simp#include <sys/event.h> 441590Srgrimes#include <sys/kernel.h> 45227156Sed#include <sys/limits.h> 46#include <sys/lock.h> 47#include <sys/lockf.h> 48#include <sys/malloc.h> 49#include <sys/mount.h> 50#include <sys/mutex.h> 51#include <sys/namei.h> 52#include <sys/fcntl.h> 53#include <sys/unistd.h> 54#include <sys/vnode.h> 55#include <sys/dirent.h> 56#include <sys/poll.h> 57 58#include <security/mac/mac_framework.h> 59 60#include <vm/vm.h> 61#include <vm/vm_object.h> 62#include <vm/vm_extern.h> 63#include <vm/pmap.h> 64#include <vm/vm_map.h> 65#include <vm/vm_page.h> 66#include <vm/vm_pager.h> 67#include <vm/vnode_pager.h> 68 69static int vop_nolookup(struct vop_lookup_args *); 70static int vop_norename(struct vop_rename_args *); 71static int vop_nostrategy(struct vop_strategy_args *); 72static int get_next_dirent(struct vnode *vp, struct dirent **dpp, 73 char *dirbuf, int dirbuflen, off_t *off, 74 char **cpos, int *len, int *eofflag, 75 struct thread *td); 76static int dirent_exists(struct vnode *vp, const char *dirname, 77 struct thread *td); 78 79#define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4) 80 81/* 82 * This vnode table stores what we want to do if the filesystem doesn't 83 * implement a particular VOP. 84 * 85 * If there is no specific entry here, we will return EOPNOTSUPP. 86 * 87 * Note that every filesystem has to implement either vop_access 88 * or vop_accessx; failing to do so will result in immediate crash 89 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(), 90 * which calls vop_stdaccess() etc. 91 */ 92 93struct vop_vector default_vnodeops = { 94 .vop_default = NULL, 95 .vop_bypass = VOP_EOPNOTSUPP, 96 97 .vop_access = vop_stdaccess, 98 .vop_accessx = vop_stdaccessx, 99 .vop_advlock = vop_stdadvlock, 100 .vop_advlockasync = vop_stdadvlockasync, 101 .vop_bmap = vop_stdbmap, 102 .vop_close = VOP_NULL, 103 .vop_fsync = VOP_NULL, 104 .vop_getpages = vop_stdgetpages, 105 .vop_getwritemount = vop_stdgetwritemount, 106 .vop_inactive = VOP_NULL, 107 .vop_ioctl = VOP_ENOTTY, 108 .vop_kqfilter = vop_stdkqfilter, 109 .vop_islocked = vop_stdislocked, 110 .vop_lock1 = vop_stdlock, 111 .vop_lookup = vop_nolookup, 112 .vop_open = VOP_NULL, 113 .vop_pathconf = VOP_EINVAL, 114 .vop_poll = vop_nopoll, 115 .vop_putpages = vop_stdputpages, 116 .vop_readlink = VOP_EINVAL, 117 .vop_rename = vop_norename, 118 .vop_revoke = VOP_PANIC, 119 .vop_strategy = vop_nostrategy, 120 .vop_unlock = vop_stdunlock, 121 .vop_vptocnp = vop_stdvptocnp, 122 .vop_vptofh = vop_stdvptofh, 123}; 124 125/* 126 * Series of placeholder functions for various error returns for 127 * VOPs. 128 */ 129 130int 131vop_eopnotsupp(struct vop_generic_args *ap) 132{ 133 /* 134 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 135 */ 136 137 return (EOPNOTSUPP); 138} 139 140int 141vop_ebadf(struct vop_generic_args *ap) 142{ 143 144 return (EBADF); 145} 146 147int 148vop_enotty(struct vop_generic_args *ap) 149{ 150 151 return (ENOTTY); 152} 153 154int 155vop_einval(struct vop_generic_args *ap) 156{ 157 158 return (EINVAL); 159} 160 161int 162vop_enoent(struct vop_generic_args *ap) 163{ 164 165 return (ENOENT); 166} 167 168int 169vop_null(struct vop_generic_args *ap) 170{ 171 172 return (0); 173} 174 175/* 176 * Helper function to panic on some bad VOPs in some filesystems. 177 */ 178int 179vop_panic(struct vop_generic_args *ap) 180{ 181 182 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 183} 184 185/* 186 * vop_std<something> and vop_no<something> are default functions for use by 187 * filesystems that need the "default reasonable" implementation for a 188 * particular operation. 189 * 190 * The documentation for the operations they implement exists (if it exists) 191 * in the VOP_<SOMETHING>(9) manpage (all uppercase). 192 */ 193 194/* 195 * Default vop for filesystems that do not support name lookup 196 */ 197static int 198vop_nolookup(ap) 199 struct vop_lookup_args /* { 200 struct vnode *a_dvp; 201 struct vnode **a_vpp; 202 struct componentname *a_cnp; 203 } */ *ap; 204{ 205 206 *ap->a_vpp = NULL; 207 return (ENOTDIR); 208} 209 210/* 211 * vop_norename: 212 * 213 * Handle unlock and reference counting for arguments of vop_rename 214 * for filesystems that do not implement rename operation. 215 */ 216static int 217vop_norename(struct vop_rename_args *ap) 218{ 219 220 vop_rename_fail(ap); 221 return (EOPNOTSUPP); 222} 223 224/* 225 * vop_nostrategy: 226 * 227 * Strategy routine for VFS devices that have none. 228 * 229 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 230 * routine. Typically this is done for a BIO_READ strategy call. 231 * Typically B_INVAL is assumed to already be clear prior to a write 232 * and should not be cleared manually unless you just made the buffer 233 * invalid. BIO_ERROR should be cleared either way. 234 */ 235 236static int 237vop_nostrategy (struct vop_strategy_args *ap) 238{ 239 printf("No strategy for buffer at %p\n", ap->a_bp); 240 vprint("vnode", ap->a_vp); 241 ap->a_bp->b_ioflags |= BIO_ERROR; 242 ap->a_bp->b_error = EOPNOTSUPP; 243 bufdone(ap->a_bp); 244 return (EOPNOTSUPP); 245} 246 247static int 248get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf, 249 int dirbuflen, off_t *off, char **cpos, int *len, 250 int *eofflag, struct thread *td) 251{ 252 int error, reclen; 253 struct uio uio; 254 struct iovec iov; 255 struct dirent *dp; 256 257 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp)); 258 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp)); 259 260 if (*len == 0) { 261 iov.iov_base = dirbuf; 262 iov.iov_len = dirbuflen; 263 264 uio.uio_iov = &iov; 265 uio.uio_iovcnt = 1; 266 uio.uio_offset = *off; 267 uio.uio_resid = dirbuflen; 268 uio.uio_segflg = UIO_SYSSPACE; 269 uio.uio_rw = UIO_READ; 270 uio.uio_td = td; 271 272 *eofflag = 0; 273 274#ifdef MAC 275 error = mac_vnode_check_readdir(td->td_ucred, vp); 276 if (error == 0) 277#endif 278 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag, 279 NULL, NULL); 280 if (error) 281 return (error); 282 283 *off = uio.uio_offset; 284 285 *cpos = dirbuf; 286 *len = (dirbuflen - uio.uio_resid); 287 } 288 289 dp = (struct dirent *)(*cpos); 290 reclen = dp->d_reclen; 291 *dpp = dp; 292 293 /* check for malformed directory.. */ 294 if (reclen < DIRENT_MINSIZE) 295 return (EINVAL); 296 297 *cpos += reclen; 298 *len -= reclen; 299 300 return (0); 301} 302 303/* 304 * Check if a named file exists in a given directory vnode. 305 */ 306static int 307dirent_exists(struct vnode *vp, const char *dirname, struct thread *td) 308{ 309 char *dirbuf, *cpos; 310 int error, eofflag, dirbuflen, len, found; 311 off_t off; 312 struct dirent *dp; 313 struct vattr va; 314 315 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp)); 316 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp)); 317 318 found = 0; 319 320 error = VOP_GETATTR(vp, &va, td->td_ucred); 321 if (error) 322 return (found); 323 324 dirbuflen = DEV_BSIZE; 325 if (dirbuflen < va.va_blocksize) 326 dirbuflen = va.va_blocksize; 327 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK); 328 329 off = 0; 330 len = 0; 331 do { 332 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off, 333 &cpos, &len, &eofflag, td); 334 if (error) 335 goto out; 336 337 if ((dp->d_type != DT_WHT) && 338 !strcmp(dp->d_name, dirname)) { 339 found = 1; 340 goto out; 341 } 342 } while (len > 0 || !eofflag); 343 344out: 345 free(dirbuf, M_TEMP); 346 return (found); 347} 348 349int 350vop_stdaccess(struct vop_access_args *ap) 351{ 352 353 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN | 354 VAPPEND)) == 0, ("invalid bit in accmode")); 355 356 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td)); 357} 358 359int 360vop_stdaccessx(struct vop_accessx_args *ap) 361{ 362 int error; 363 accmode_t accmode = ap->a_accmode; 364 365 error = vfs_unixify_accmode(&accmode); 366 if (error != 0) 367 return (error); 368 369 if (accmode == 0) 370 return (0); 371 372 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td)); 373} 374 375/* 376 * Advisory record locking support 377 */ 378int 379vop_stdadvlock(struct vop_advlock_args *ap) 380{ 381 struct vnode *vp; 382 struct ucred *cred; 383 struct vattr vattr; 384 int error; 385 386 vp = ap->a_vp; 387 cred = curthread->td_ucred; 388 vn_lock(vp, LK_SHARED | LK_RETRY); 389 error = VOP_GETATTR(vp, &vattr, cred); 390 VOP_UNLOCK(vp, 0); 391 if (error) 392 return (error); 393 394 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size)); 395} 396 397int 398vop_stdadvlockasync(struct vop_advlockasync_args *ap) 399{ 400 struct vnode *vp; 401 struct ucred *cred; 402 struct vattr vattr; 403 int error; 404 405 vp = ap->a_vp; 406 cred = curthread->td_ucred; 407 vn_lock(vp, LK_SHARED | LK_RETRY); 408 error = VOP_GETATTR(vp, &vattr, cred); 409 VOP_UNLOCK(vp, 0); 410 if (error) 411 return (error); 412 413 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size)); 414} 415 416/* 417 * vop_stdpathconf: 418 * 419 * Standard implementation of POSIX pathconf, to get information about limits 420 * for a filesystem. 421 * Override per filesystem for the case where the filesystem has smaller 422 * limits. 423 */ 424int 425vop_stdpathconf(ap) 426 struct vop_pathconf_args /* { 427 struct vnode *a_vp; 428 int a_name; 429 int *a_retval; 430 } */ *ap; 431{ 432 433 switch (ap->a_name) { 434 case _PC_NAME_MAX: 435 *ap->a_retval = NAME_MAX; 436 return (0); 437 case _PC_PATH_MAX: 438 *ap->a_retval = PATH_MAX; 439 return (0); 440 case _PC_LINK_MAX: 441 *ap->a_retval = LINK_MAX; 442 return (0); 443 case _PC_MAX_CANON: 444 *ap->a_retval = MAX_CANON; 445 return (0); 446 case _PC_MAX_INPUT: 447 *ap->a_retval = MAX_INPUT; 448 return (0); 449 case _PC_PIPE_BUF: 450 *ap->a_retval = PIPE_BUF; 451 return (0); 452 case _PC_CHOWN_RESTRICTED: 453 *ap->a_retval = 1; 454 return (0); 455 case _PC_VDISABLE: 456 *ap->a_retval = _POSIX_VDISABLE; 457 return (0); 458 default: 459 return (EINVAL); 460 } 461 /* NOTREACHED */ 462} 463 464/* 465 * Standard lock, unlock and islocked functions. 466 */ 467int 468vop_stdlock(ap) 469 struct vop_lock1_args /* { 470 struct vnode *a_vp; 471 int a_flags; 472 char *file; 473 int line; 474 } */ *ap; 475{ 476 struct vnode *vp = ap->a_vp; 477 478 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp), 479 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file, 480 ap->a_line)); 481} 482 483/* See above. */ 484int 485vop_stdunlock(ap) 486 struct vop_unlock_args /* { 487 struct vnode *a_vp; 488 int a_flags; 489 } */ *ap; 490{ 491 struct vnode *vp = ap->a_vp; 492 493 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp))); 494} 495 496/* See above. */ 497int 498vop_stdislocked(ap) 499 struct vop_islocked_args /* { 500 struct vnode *a_vp; 501 } */ *ap; 502{ 503 504 return (lockstatus(ap->a_vp->v_vnlock)); 505} 506 507/* 508 * Return true for select/poll. 509 */ 510int 511vop_nopoll(ap) 512 struct vop_poll_args /* { 513 struct vnode *a_vp; 514 int a_events; 515 struct ucred *a_cred; 516 struct thread *a_td; 517 } */ *ap; 518{ 519 520 return (poll_no_poll(ap->a_events)); 521} 522 523/* 524 * Implement poll for local filesystems that support it. 525 */ 526int 527vop_stdpoll(ap) 528 struct vop_poll_args /* { 529 struct vnode *a_vp; 530 int a_events; 531 struct ucred *a_cred; 532 struct thread *a_td; 533 } */ *ap; 534{ 535 if (ap->a_events & ~POLLSTANDARD) 536 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 537 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 538} 539 540/* 541 * Return our mount point, as we will take charge of the writes. 542 */ 543int 544vop_stdgetwritemount(ap) 545 struct vop_getwritemount_args /* { 546 struct vnode *a_vp; 547 struct mount **a_mpp; 548 } */ *ap; 549{ 550 struct mount *mp; 551 552 /* 553 * XXX Since this is called unlocked we may be recycled while 554 * attempting to ref the mount. If this is the case or mountpoint 555 * will be set to NULL. We only have to prevent this call from 556 * returning with a ref to an incorrect mountpoint. It is not 557 * harmful to return with a ref to our previous mountpoint. 558 */ 559 mp = ap->a_vp->v_mount; 560 if (mp != NULL) { 561 vfs_ref(mp); 562 if (mp != ap->a_vp->v_mount) { 563 vfs_rel(mp); 564 mp = NULL; 565 } 566 } 567 *(ap->a_mpp) = mp; 568 return (0); 569} 570 571/* XXX Needs good comment and VOP_BMAP(9) manpage */ 572int 573vop_stdbmap(ap) 574 struct vop_bmap_args /* { 575 struct vnode *a_vp; 576 daddr_t a_bn; 577 struct bufobj **a_bop; 578 daddr_t *a_bnp; 579 int *a_runp; 580 int *a_runb; 581 } */ *ap; 582{ 583 584 if (ap->a_bop != NULL) 585 *ap->a_bop = &ap->a_vp->v_bufobj; 586 if (ap->a_bnp != NULL) 587 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 588 if (ap->a_runp != NULL) 589 *ap->a_runp = 0; 590 if (ap->a_runb != NULL) 591 *ap->a_runb = 0; 592 return (0); 593} 594 595int 596vop_stdfsync(ap) 597 struct vop_fsync_args /* { 598 struct vnode *a_vp; 599 struct ucred *a_cred; 600 int a_waitfor; 601 struct thread *a_td; 602 } */ *ap; 603{ 604 struct vnode *vp = ap->a_vp; 605 struct buf *bp; 606 struct bufobj *bo; 607 struct buf *nbp; 608 int error = 0; 609 int maxretry = 1000; /* large, arbitrarily chosen */ 610 611 bo = &vp->v_bufobj; 612 BO_LOCK(bo); 613loop1: 614 /* 615 * MARK/SCAN initialization to avoid infinite loops. 616 */ 617 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) { 618 bp->b_vflags &= ~BV_SCANNED; 619 bp->b_error = 0; 620 } 621 622 /* 623 * Flush all dirty buffers associated with a vnode. 624 */ 625loop2: 626 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) { 627 if ((bp->b_vflags & BV_SCANNED) != 0) 628 continue; 629 bp->b_vflags |= BV_SCANNED; 630 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) 631 continue; 632 BO_UNLOCK(bo); 633 KASSERT(bp->b_bufobj == bo, 634 ("bp %p wrong b_bufobj %p should be %p", 635 bp, bp->b_bufobj, bo)); 636 if ((bp->b_flags & B_DELWRI) == 0) 637 panic("fsync: not dirty"); 638 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) { 639 vfs_bio_awrite(bp); 640 } else { 641 bremfree(bp); 642 bawrite(bp); 643 } 644 BO_LOCK(bo); 645 goto loop2; 646 } 647 648 /* 649 * If synchronous the caller expects us to completely resolve all 650 * dirty buffers in the system. Wait for in-progress I/O to 651 * complete (which could include background bitmap writes), then 652 * retry if dirty blocks still exist. 653 */ 654 if (ap->a_waitfor == MNT_WAIT) { 655 bufobj_wwait(bo, 0, 0); 656 if (bo->bo_dirty.bv_cnt > 0) { 657 /* 658 * If we are unable to write any of these buffers 659 * then we fail now rather than trying endlessly 660 * to write them out. 661 */ 662 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) 663 if ((error = bp->b_error) == 0) 664 continue; 665 if (error == 0 && --maxretry >= 0) 666 goto loop1; 667 error = EAGAIN; 668 } 669 } 670 BO_UNLOCK(bo); 671 if (error == EAGAIN) 672 vprint("fsync: giving up on dirty", vp); 673 674 return (error); 675} 676 677/* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */ 678int 679vop_stdgetpages(ap) 680 struct vop_getpages_args /* { 681 struct vnode *a_vp; 682 vm_page_t *a_m; 683 int a_count; 684 int a_reqpage; 685 vm_ooffset_t a_offset; 686 } */ *ap; 687{ 688 689 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, 690 ap->a_count, ap->a_reqpage); 691} 692 693int 694vop_stdkqfilter(struct vop_kqfilter_args *ap) 695{ 696 return vfs_kqfilter(ap); 697} 698 699/* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */ 700int 701vop_stdputpages(ap) 702 struct vop_putpages_args /* { 703 struct vnode *a_vp; 704 vm_page_t *a_m; 705 int a_count; 706 int a_sync; 707 int *a_rtvals; 708 vm_ooffset_t a_offset; 709 } */ *ap; 710{ 711 712 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count, 713 ap->a_sync, ap->a_rtvals); 714} 715 716int 717vop_stdvptofh(struct vop_vptofh_args *ap) 718{ 719 return (EOPNOTSUPP); 720} 721 722int 723vop_stdvptocnp(struct vop_vptocnp_args *ap) 724{ 725 struct vnode *vp = ap->a_vp; 726 struct vnode **dvp = ap->a_vpp; 727 struct ucred *cred = ap->a_cred; 728 char *buf = ap->a_buf; 729 int *buflen = ap->a_buflen; 730 char *dirbuf, *cpos; 731 int i, error, eofflag, dirbuflen, flags, locked, len, covered; 732 off_t off; 733 ino_t fileno; 734 struct vattr va; 735 struct nameidata nd; 736 struct thread *td; 737 struct dirent *dp; 738 struct vnode *mvp; 739 740 i = *buflen; 741 error = 0; 742 covered = 0; 743 td = curthread; 744 745 if (vp->v_type != VDIR) 746 return (ENOENT); 747 748 error = VOP_GETATTR(vp, &va, cred); 749 if (error) 750 return (error); 751 752 VREF(vp); 753 locked = VOP_ISLOCKED(vp); 754 VOP_UNLOCK(vp, 0); 755 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 756 "..", vp, td); 757 flags = FREAD; 758 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL); 759 if (error) { 760 vn_lock(vp, locked | LK_RETRY); 761 return (error); 762 } 763 NDFREE(&nd, NDF_ONLY_PNBUF); 764 765 mvp = *dvp = nd.ni_vp; 766 767 if (vp->v_mount != (*dvp)->v_mount && 768 ((*dvp)->v_vflag & VV_ROOT) && 769 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) { 770 *dvp = (*dvp)->v_mount->mnt_vnodecovered; 771 VREF(mvp); 772 VOP_UNLOCK(mvp, 0); 773 vn_close(mvp, FREAD, cred, td); 774 VREF(*dvp); 775 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY); 776 covered = 1; 777 } 778 779 fileno = va.va_fileid; 780 781 dirbuflen = DEV_BSIZE; 782 if (dirbuflen < va.va_blocksize) 783 dirbuflen = va.va_blocksize; 784 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK); 785 786 if ((*dvp)->v_type != VDIR) { 787 error = ENOENT; 788 goto out; 789 } 790 791 off = 0; 792 len = 0; 793 do { 794 /* call VOP_READDIR of parent */ 795 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off, 796 &cpos, &len, &eofflag, td); 797 if (error) 798 goto out; 799 800 if ((dp->d_type != DT_WHT) && 801 (dp->d_fileno == fileno)) { 802 if (covered) { 803 VOP_UNLOCK(*dvp, 0); 804 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY); 805 if (dirent_exists(mvp, dp->d_name, td)) { 806 error = ENOENT; 807 VOP_UNLOCK(mvp, 0); 808 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY); 809 goto out; 810 } 811 VOP_UNLOCK(mvp, 0); 812 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY); 813 } 814 i -= dp->d_namlen; 815 816 if (i < 0) { 817 error = ENOMEM; 818 goto out; 819 } 820 bcopy(dp->d_name, buf + i, dp->d_namlen); 821 error = 0; 822 goto out; 823 } 824 } while (len > 0 || !eofflag); 825 error = ENOENT; 826 827out: 828 free(dirbuf, M_TEMP); 829 if (!error) { 830 *buflen = i; 831 vhold(*dvp); 832 } 833 if (covered) { 834 vput(*dvp); 835 vrele(mvp); 836 } else { 837 VOP_UNLOCK(mvp, 0); 838 vn_close(mvp, FREAD, cred, td); 839 } 840 vn_lock(vp, locked | LK_RETRY); 841 return (error); 842} 843 844/* 845 * vfs default ops 846 * used to fill the vfs function table to get reasonable default return values. 847 */ 848int 849vfs_stdroot (mp, flags, vpp) 850 struct mount *mp; 851 int flags; 852 struct vnode **vpp; 853{ 854 855 return (EOPNOTSUPP); 856} 857 858int 859vfs_stdstatfs (mp, sbp) 860 struct mount *mp; 861 struct statfs *sbp; 862{ 863 864 return (EOPNOTSUPP); 865} 866 867int 868vfs_stdquotactl (mp, cmds, uid, arg) 869 struct mount *mp; 870 int cmds; 871 uid_t uid; 872 void *arg; 873{ 874 875 return (EOPNOTSUPP); 876} 877 878int 879vfs_stdsync(mp, waitfor) 880 struct mount *mp; 881 int waitfor; 882{ 883 struct vnode *vp, *mvp; 884 struct thread *td; 885 int error, lockreq, allerror = 0; 886 887 td = curthread; 888 lockreq = LK_EXCLUSIVE | LK_INTERLOCK; 889 if (waitfor != MNT_WAIT) 890 lockreq |= LK_NOWAIT; 891 /* 892 * Force stale buffer cache information to be flushed. 893 */ 894 MNT_ILOCK(mp); 895loop: 896 MNT_VNODE_FOREACH(vp, mp, mvp) { 897 /* bv_cnt is an acceptable race here. */ 898 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) 899 continue; 900 VI_LOCK(vp); 901 MNT_IUNLOCK(mp); 902 if ((error = vget(vp, lockreq, td)) != 0) { 903 MNT_ILOCK(mp); 904 if (error == ENOENT) { 905 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp); 906 goto loop; 907 } 908 continue; 909 } 910 error = VOP_FSYNC(vp, waitfor, td); 911 if (error) 912 allerror = error; 913 vput(vp); 914 MNT_ILOCK(mp); 915 } 916 MNT_IUNLOCK(mp); 917 return (allerror); 918} 919 920int 921vfs_stdnosync (mp, waitfor) 922 struct mount *mp; 923 int waitfor; 924{ 925 926 return (0); 927} 928 929int 930vfs_stdvget (mp, ino, flags, vpp) 931 struct mount *mp; 932 ino_t ino; 933 int flags; 934 struct vnode **vpp; 935{ 936 937 return (EOPNOTSUPP); 938} 939 940int 941vfs_stdfhtovp (mp, fhp, vpp) 942 struct mount *mp; 943 struct fid *fhp; 944 struct vnode **vpp; 945{ 946 947 return (EOPNOTSUPP); 948} 949 950int 951vfs_stdinit (vfsp) 952 struct vfsconf *vfsp; 953{ 954 955 return (0); 956} 957 958int 959vfs_stduninit (vfsp) 960 struct vfsconf *vfsp; 961{ 962 963 return(0); 964} 965 966int 967vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname) 968 struct mount *mp; 969 int cmd; 970 struct vnode *filename_vp; 971 int attrnamespace; 972 const char *attrname; 973{ 974 975 if (filename_vp != NULL) 976 VOP_UNLOCK(filename_vp, 0); 977 return (EOPNOTSUPP); 978} 979 980int 981vfs_stdsysctl(mp, op, req) 982 struct mount *mp; 983 fsctlop_t op; 984 struct sysctl_req *req; 985{ 986 987 return (EOPNOTSUPP); 988} 989 990/* end of vfs default ops */ 991