Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/kern/vfs_default.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
1 /*- 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed 6 * to Berkeley by John Heidemann of the UCLA Ficus project. 7 * 8 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 */ 34 35 #include <sys/cdefs.h> 36 __FBSDID("$FreeBSD: releng/8.3/sys/kern/vfs_default.c 229725 2012-01-06 19:32:39Z jhb $"); 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/bio.h> 41 #include <sys/buf.h> 42 #include <sys/conf.h> 43 #include <sys/event.h> 44 #include <sys/kernel.h> 45 #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 69 static int vop_nolookup(struct vop_lookup_args *); 70 static int vop_norename(struct vop_rename_args *); 71 static int vop_nostrategy(struct vop_strategy_args *); 72 static 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); 76 static 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 93 struct 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_advise = vop_stdadvise, 100 .vop_advlock = vop_stdadvlock, 101 .vop_advlockasync = vop_stdadvlockasync, 102 .vop_allocate = vop_stdallocate, 103 .vop_bmap = vop_stdbmap, 104 .vop_close = VOP_NULL, 105 .vop_fsync = VOP_NULL, 106 .vop_getpages = vop_stdgetpages, 107 .vop_getwritemount = vop_stdgetwritemount, 108 .vop_inactive = VOP_NULL, 109 .vop_ioctl = VOP_ENOTTY, 110 .vop_kqfilter = vop_stdkqfilter, 111 .vop_islocked = vop_stdislocked, 112 .vop_lock1 = vop_stdlock, 113 .vop_lookup = vop_nolookup, 114 .vop_open = VOP_NULL, 115 .vop_pathconf = VOP_EINVAL, 116 .vop_poll = vop_nopoll, 117 .vop_putpages = vop_stdputpages, 118 .vop_readlink = VOP_EINVAL, 119 .vop_rename = vop_norename, 120 .vop_revoke = VOP_PANIC, 121 .vop_strategy = vop_nostrategy, 122 .vop_unlock = vop_stdunlock, 123 .vop_vptocnp = vop_stdvptocnp, 124 .vop_vptofh = vop_stdvptofh, 125 }; 126 127 /* 128 * Series of placeholder functions for various error returns for 129 * VOPs. 130 */ 131 132 int 133 vop_eopnotsupp(struct vop_generic_args *ap) 134 { 135 /* 136 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 137 */ 138 139 return (EOPNOTSUPP); 140 } 141 142 int 143 vop_ebadf(struct vop_generic_args *ap) 144 { 145 146 return (EBADF); 147 } 148 149 int 150 vop_enotty(struct vop_generic_args *ap) 151 { 152 153 return (ENOTTY); 154 } 155 156 int 157 vop_einval(struct vop_generic_args *ap) 158 { 159 160 return (EINVAL); 161 } 162 163 int 164 vop_enoent(struct vop_generic_args *ap) 165 { 166 167 return (ENOENT); 168 } 169 170 int 171 vop_null(struct vop_generic_args *ap) 172 { 173 174 return (0); 175 } 176 177 /* 178 * Helper function to panic on some bad VOPs in some filesystems. 179 */ 180 int 181 vop_panic(struct vop_generic_args *ap) 182 { 183 184 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 185 } 186 187 /* 188 * vop_std<something> and vop_no<something> are default functions for use by 189 * filesystems that need the "default reasonable" implementation for a 190 * particular operation. 191 * 192 * The documentation for the operations they implement exists (if it exists) 193 * in the VOP_<SOMETHING>(9) manpage (all uppercase). 194 */ 195 196 /* 197 * Default vop for filesystems that do not support name lookup 198 */ 199 static int 200 vop_nolookup(ap) 201 struct vop_lookup_args /* { 202 struct vnode *a_dvp; 203 struct vnode **a_vpp; 204 struct componentname *a_cnp; 205 } */ *ap; 206 { 207 208 *ap->a_vpp = NULL; 209 return (ENOTDIR); 210 } 211 212 /* 213 * vop_norename: 214 * 215 * Handle unlock and reference counting for arguments of vop_rename 216 * for filesystems that do not implement rename operation. 217 */ 218 static int 219 vop_norename(struct vop_rename_args *ap) 220 { 221 222 vop_rename_fail(ap); 223 return (EOPNOTSUPP); 224 } 225 226 /* 227 * vop_nostrategy: 228 * 229 * Strategy routine for VFS devices that have none. 230 * 231 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 232 * routine. Typically this is done for a BIO_READ strategy call. 233 * Typically B_INVAL is assumed to already be clear prior to a write 234 * and should not be cleared manually unless you just made the buffer 235 * invalid. BIO_ERROR should be cleared either way. 236 */ 237 238 static int 239 vop_nostrategy (struct vop_strategy_args *ap) 240 { 241 printf("No strategy for buffer at %p\n", ap->a_bp); 242 vprint("vnode", ap->a_vp); 243 ap->a_bp->b_ioflags |= BIO_ERROR; 244 ap->a_bp->b_error = EOPNOTSUPP; 245 bufdone(ap->a_bp); 246 return (EOPNOTSUPP); 247 } 248 249 static int 250 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf, 251 int dirbuflen, off_t *off, char **cpos, int *len, 252 int *eofflag, struct thread *td) 253 { 254 int error, reclen; 255 struct uio uio; 256 struct iovec iov; 257 struct dirent *dp; 258 259 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp)); 260 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp)); 261 262 if (*len == 0) { 263 iov.iov_base = dirbuf; 264 iov.iov_len = dirbuflen; 265 266 uio.uio_iov = &iov; 267 uio.uio_iovcnt = 1; 268 uio.uio_offset = *off; 269 uio.uio_resid = dirbuflen; 270 uio.uio_segflg = UIO_SYSSPACE; 271 uio.uio_rw = UIO_READ; 272 uio.uio_td = td; 273 274 *eofflag = 0; 275 276 #ifdef MAC 277 error = mac_vnode_check_readdir(td->td_ucred, vp); 278 if (error == 0) 279 #endif 280 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag, 281 NULL, NULL); 282 if (error) 283 return (error); 284 285 *off = uio.uio_offset; 286 287 *cpos = dirbuf; 288 *len = (dirbuflen - uio.uio_resid); 289 290 if (*len == 0) 291 return (ENOENT); 292 } 293 294 dp = (struct dirent *)(*cpos); 295 reclen = dp->d_reclen; 296 *dpp = dp; 297 298 /* check for malformed directory.. */ 299 if (reclen < DIRENT_MINSIZE) 300 return (EINVAL); 301 302 *cpos += reclen; 303 *len -= reclen; 304 305 return (0); 306 } 307 308 /* 309 * Check if a named file exists in a given directory vnode. 310 */ 311 static int 312 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td) 313 { 314 char *dirbuf, *cpos; 315 int error, eofflag, dirbuflen, len, found; 316 off_t off; 317 struct dirent *dp; 318 struct vattr va; 319 320 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp)); 321 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp)); 322 323 found = 0; 324 325 error = VOP_GETATTR(vp, &va, td->td_ucred); 326 if (error) 327 return (found); 328 329 dirbuflen = DEV_BSIZE; 330 if (dirbuflen < va.va_blocksize) 331 dirbuflen = va.va_blocksize; 332 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK); 333 334 off = 0; 335 len = 0; 336 do { 337 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off, 338 &cpos, &len, &eofflag, td); 339 if (error) 340 goto out; 341 342 if ((dp->d_type != DT_WHT) && 343 !strcmp(dp->d_name, dirname)) { 344 found = 1; 345 goto out; 346 } 347 } while (len > 0 || !eofflag); 348 349 out: 350 free(dirbuf, M_TEMP); 351 return (found); 352 } 353 354 int 355 vop_stdaccess(struct vop_access_args *ap) 356 { 357 358 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN | 359 VAPPEND)) == 0, ("invalid bit in accmode")); 360 361 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td)); 362 } 363 364 int 365 vop_stdaccessx(struct vop_accessx_args *ap) 366 { 367 int error; 368 accmode_t accmode = ap->a_accmode; 369 370 error = vfs_unixify_accmode(&accmode); 371 if (error != 0) 372 return (error); 373 374 if (accmode == 0) 375 return (0); 376 377 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td)); 378 } 379 380 /* 381 * Advisory record locking support 382 */ 383 int 384 vop_stdadvlock(struct vop_advlock_args *ap) 385 { 386 struct vnode *vp; 387 struct ucred *cred; 388 struct vattr vattr; 389 int error; 390 391 vp = ap->a_vp; 392 cred = curthread->td_ucred; 393 vn_lock(vp, LK_SHARED | LK_RETRY); 394 error = VOP_GETATTR(vp, &vattr, cred); 395 VOP_UNLOCK(vp, 0); 396 if (error) 397 return (error); 398 399 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size)); 400 } 401 402 int 403 vop_stdadvlockasync(struct vop_advlockasync_args *ap) 404 { 405 struct vnode *vp; 406 struct ucred *cred; 407 struct vattr vattr; 408 int error; 409 410 vp = ap->a_vp; 411 cred = curthread->td_ucred; 412 vn_lock(vp, LK_SHARED | LK_RETRY); 413 error = VOP_GETATTR(vp, &vattr, cred); 414 VOP_UNLOCK(vp, 0); 415 if (error) 416 return (error); 417 418 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size)); 419 } 420 421 /* 422 * vop_stdpathconf: 423 * 424 * Standard implementation of POSIX pathconf, to get information about limits 425 * for a filesystem. 426 * Override per filesystem for the case where the filesystem has smaller 427 * limits. 428 */ 429 int 430 vop_stdpathconf(ap) 431 struct vop_pathconf_args /* { 432 struct vnode *a_vp; 433 int a_name; 434 int *a_retval; 435 } */ *ap; 436 { 437 438 switch (ap->a_name) { 439 case _PC_NAME_MAX: 440 *ap->a_retval = NAME_MAX; 441 return (0); 442 case _PC_PATH_MAX: 443 *ap->a_retval = PATH_MAX; 444 return (0); 445 case _PC_LINK_MAX: 446 *ap->a_retval = LINK_MAX; 447 return (0); 448 case _PC_MAX_CANON: 449 *ap->a_retval = MAX_CANON; 450 return (0); 451 case _PC_MAX_INPUT: 452 *ap->a_retval = MAX_INPUT; 453 return (0); 454 case _PC_PIPE_BUF: 455 *ap->a_retval = PIPE_BUF; 456 return (0); 457 case _PC_CHOWN_RESTRICTED: 458 *ap->a_retval = 1; 459 return (0); 460 case _PC_VDISABLE: 461 *ap->a_retval = _POSIX_VDISABLE; 462 return (0); 463 default: 464 return (EINVAL); 465 } 466 /* NOTREACHED */ 467 } 468 469 /* 470 * Standard lock, unlock and islocked functions. 471 */ 472 int 473 vop_stdlock(ap) 474 struct vop_lock1_args /* { 475 struct vnode *a_vp; 476 int a_flags; 477 char *file; 478 int line; 479 } */ *ap; 480 { 481 struct vnode *vp = ap->a_vp; 482 483 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp), 484 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file, 485 ap->a_line)); 486 } 487 488 /* See above. */ 489 int 490 vop_stdunlock(ap) 491 struct vop_unlock_args /* { 492 struct vnode *a_vp; 493 int a_flags; 494 } */ *ap; 495 { 496 struct vnode *vp = ap->a_vp; 497 498 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp))); 499 } 500 501 /* See above. */ 502 int 503 vop_stdislocked(ap) 504 struct vop_islocked_args /* { 505 struct vnode *a_vp; 506 } */ *ap; 507 { 508 509 return (lockstatus(ap->a_vp->v_vnlock)); 510 } 511 512 /* 513 * Return true for select/poll. 514 */ 515 int 516 vop_nopoll(ap) 517 struct vop_poll_args /* { 518 struct vnode *a_vp; 519 int a_events; 520 struct ucred *a_cred; 521 struct thread *a_td; 522 } */ *ap; 523 { 524 525 return (poll_no_poll(ap->a_events)); 526 } 527 528 /* 529 * Implement poll for local filesystems that support it. 530 */ 531 int 532 vop_stdpoll(ap) 533 struct vop_poll_args /* { 534 struct vnode *a_vp; 535 int a_events; 536 struct ucred *a_cred; 537 struct thread *a_td; 538 } */ *ap; 539 { 540 if (ap->a_events & ~POLLSTANDARD) 541 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 542 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 543 } 544 545 /* 546 * Return our mount point, as we will take charge of the writes. 547 */ 548 int 549 vop_stdgetwritemount(ap) 550 struct vop_getwritemount_args /* { 551 struct vnode *a_vp; 552 struct mount **a_mpp; 553 } */ *ap; 554 { 555 struct mount *mp; 556 557 /* 558 * XXX Since this is called unlocked we may be recycled while 559 * attempting to ref the mount. If this is the case or mountpoint 560 * will be set to NULL. We only have to prevent this call from 561 * returning with a ref to an incorrect mountpoint. It is not 562 * harmful to return with a ref to our previous mountpoint. 563 */ 564 mp = ap->a_vp->v_mount; 565 if (mp != NULL) { 566 vfs_ref(mp); 567 if (mp != ap->a_vp->v_mount) { 568 vfs_rel(mp); 569 mp = NULL; 570 } 571 } 572 *(ap->a_mpp) = mp; 573 return (0); 574 } 575 576 /* XXX Needs good comment and VOP_BMAP(9) manpage */ 577 int 578 vop_stdbmap(ap) 579 struct vop_bmap_args /* { 580 struct vnode *a_vp; 581 daddr_t a_bn; 582 struct bufobj **a_bop; 583 daddr_t *a_bnp; 584 int *a_runp; 585 int *a_runb; 586 } */ *ap; 587 { 588 589 if (ap->a_bop != NULL) 590 *ap->a_bop = &ap->a_vp->v_bufobj; 591 if (ap->a_bnp != NULL) 592 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 593 if (ap->a_runp != NULL) 594 *ap->a_runp = 0; 595 if (ap->a_runb != NULL) 596 *ap->a_runb = 0; 597 return (0); 598 } 599 600 int 601 vop_stdfsync(ap) 602 struct vop_fsync_args /* { 603 struct vnode *a_vp; 604 struct ucred *a_cred; 605 int a_waitfor; 606 struct thread *a_td; 607 } */ *ap; 608 { 609 struct vnode *vp = ap->a_vp; 610 struct buf *bp; 611 struct bufobj *bo; 612 struct buf *nbp; 613 int error = 0; 614 int maxretry = 1000; /* large, arbitrarily chosen */ 615 616 bo = &vp->v_bufobj; 617 BO_LOCK(bo); 618 loop1: 619 /* 620 * MARK/SCAN initialization to avoid infinite loops. 621 */ 622 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) { 623 bp->b_vflags &= ~BV_SCANNED; 624 bp->b_error = 0; 625 } 626 627 /* 628 * Flush all dirty buffers associated with a vnode. 629 */ 630 loop2: 631 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) { 632 if ((bp->b_vflags & BV_SCANNED) != 0) 633 continue; 634 bp->b_vflags |= BV_SCANNED; 635 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) 636 continue; 637 BO_UNLOCK(bo); 638 KASSERT(bp->b_bufobj == bo, 639 ("bp %p wrong b_bufobj %p should be %p", 640 bp, bp->b_bufobj, bo)); 641 if ((bp->b_flags & B_DELWRI) == 0) 642 panic("fsync: not dirty"); 643 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) { 644 vfs_bio_awrite(bp); 645 } else { 646 bremfree(bp); 647 bawrite(bp); 648 } 649 BO_LOCK(bo); 650 goto loop2; 651 } 652 653 /* 654 * If synchronous the caller expects us to completely resolve all 655 * dirty buffers in the system. Wait for in-progress I/O to 656 * complete (which could include background bitmap writes), then 657 * retry if dirty blocks still exist. 658 */ 659 if (ap->a_waitfor == MNT_WAIT) { 660 bufobj_wwait(bo, 0, 0); 661 if (bo->bo_dirty.bv_cnt > 0) { 662 /* 663 * If we are unable to write any of these buffers 664 * then we fail now rather than trying endlessly 665 * to write them out. 666 */ 667 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) 668 if ((error = bp->b_error) == 0) 669 continue; 670 if (error == 0 && --maxretry >= 0) 671 goto loop1; 672 error = EAGAIN; 673 } 674 } 675 BO_UNLOCK(bo); 676 if (error == EAGAIN) 677 vprint("fsync: giving up on dirty", vp); 678 679 return (error); 680 } 681 682 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */ 683 int 684 vop_stdgetpages(ap) 685 struct vop_getpages_args /* { 686 struct vnode *a_vp; 687 vm_page_t *a_m; 688 int a_count; 689 int a_reqpage; 690 vm_ooffset_t a_offset; 691 } */ *ap; 692 { 693 694 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, 695 ap->a_count, ap->a_reqpage); 696 } 697 698 int 699 vop_stdkqfilter(struct vop_kqfilter_args *ap) 700 { 701 return vfs_kqfilter(ap); 702 } 703 704 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */ 705 int 706 vop_stdputpages(ap) 707 struct vop_putpages_args /* { 708 struct vnode *a_vp; 709 vm_page_t *a_m; 710 int a_count; 711 int a_sync; 712 int *a_rtvals; 713 vm_ooffset_t a_offset; 714 } */ *ap; 715 { 716 717 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count, 718 ap->a_sync, ap->a_rtvals); 719 } 720 721 int 722 vop_stdvptofh(struct vop_vptofh_args *ap) 723 { 724 return (EOPNOTSUPP); 725 } 726 727 int 728 vop_stdvptocnp(struct vop_vptocnp_args *ap) 729 { 730 struct vnode *vp = ap->a_vp; 731 struct vnode **dvp = ap->a_vpp; 732 struct ucred *cred = ap->a_cred; 733 char *buf = ap->a_buf; 734 int *buflen = ap->a_buflen; 735 char *dirbuf, *cpos; 736 int i, error, eofflag, dirbuflen, flags, locked, len, covered; 737 off_t off; 738 ino_t fileno; 739 struct vattr va; 740 struct nameidata nd; 741 struct thread *td; 742 struct dirent *dp; 743 struct vnode *mvp; 744 745 i = *buflen; 746 error = 0; 747 covered = 0; 748 td = curthread; 749 750 if (vp->v_type != VDIR) 751 return (ENOENT); 752 753 error = VOP_GETATTR(vp, &va, cred); 754 if (error) 755 return (error); 756 757 VREF(vp); 758 locked = VOP_ISLOCKED(vp); 759 VOP_UNLOCK(vp, 0); 760 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 761 "..", vp, td); 762 flags = FREAD; 763 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL); 764 if (error) { 765 vn_lock(vp, locked | LK_RETRY); 766 return (error); 767 } 768 NDFREE(&nd, NDF_ONLY_PNBUF); 769 770 mvp = *dvp = nd.ni_vp; 771 772 if (vp->v_mount != (*dvp)->v_mount && 773 ((*dvp)->v_vflag & VV_ROOT) && 774 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) { 775 *dvp = (*dvp)->v_mount->mnt_vnodecovered; 776 VREF(mvp); 777 VOP_UNLOCK(mvp, 0); 778 vn_close(mvp, FREAD, cred, td); 779 VREF(*dvp); 780 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY); 781 covered = 1; 782 } 783 784 fileno = va.va_fileid; 785 786 dirbuflen = DEV_BSIZE; 787 if (dirbuflen < va.va_blocksize) 788 dirbuflen = va.va_blocksize; 789 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK); 790 791 if ((*dvp)->v_type != VDIR) { 792 error = ENOENT; 793 goto out; 794 } 795 796 off = 0; 797 len = 0; 798 do { 799 /* call VOP_READDIR of parent */ 800 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off, 801 &cpos, &len, &eofflag, td); 802 if (error) 803 goto out; 804 805 if ((dp->d_type != DT_WHT) && 806 (dp->d_fileno == fileno)) { 807 if (covered) { 808 VOP_UNLOCK(*dvp, 0); 809 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY); 810 if (dirent_exists(mvp, dp->d_name, td)) { 811 error = ENOENT; 812 VOP_UNLOCK(mvp, 0); 813 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY); 814 goto out; 815 } 816 VOP_UNLOCK(mvp, 0); 817 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY); 818 } 819 i -= dp->d_namlen; 820 821 if (i < 0) { 822 error = ENOMEM; 823 goto out; 824 } 825 bcopy(dp->d_name, buf + i, dp->d_namlen); 826 error = 0; 827 goto out; 828 } 829 } while (len > 0 || !eofflag); 830 error = ENOENT; 831 832 out: 833 free(dirbuf, M_TEMP); 834 if (!error) { 835 *buflen = i; 836 vhold(*dvp); 837 } 838 if (covered) { 839 vput(*dvp); 840 vrele(mvp); 841 } else { 842 VOP_UNLOCK(mvp, 0); 843 vn_close(mvp, FREAD, cred, td); 844 } 845 vn_lock(vp, locked | LK_RETRY); 846 return (error); 847 } 848 849 int 850 vop_stdallocate(struct vop_allocate_args *ap) 851 { 852 #ifdef __notyet__ 853 struct statfs sfs; 854 #endif 855 struct iovec aiov; 856 struct vattr vattr, *vap; 857 struct uio auio; 858 off_t fsize, len, cur, offset; 859 uint8_t *buf; 860 struct thread *td; 861 struct vnode *vp; 862 size_t iosize; 863 int error; 864 865 buf = NULL; 866 error = 0; 867 td = curthread; 868 vap = &vattr; 869 vp = ap->a_vp; 870 len = *ap->a_len; 871 offset = *ap->a_offset; 872 873 error = VOP_GETATTR(vp, vap, td->td_ucred); 874 if (error != 0) 875 goto out; 876 fsize = vap->va_size; 877 iosize = vap->va_blocksize; 878 if (iosize == 0) 879 iosize = BLKDEV_IOSIZE; 880 if (iosize > MAXPHYS) 881 iosize = MAXPHYS; 882 buf = malloc(iosize, M_TEMP, M_WAITOK); 883 884 #ifdef __notyet__ 885 /* 886 * Check if the filesystem sets f_maxfilesize; if not use 887 * VOP_SETATTR to perform the check. 888 */ 889 error = VFS_STATFS(vp->v_mount, &sfs, td); 890 if (error != 0) 891 goto out; 892 if (sfs.f_maxfilesize) { 893 if (offset > sfs.f_maxfilesize || len > sfs.f_maxfilesize || 894 offset + len > sfs.f_maxfilesize) { 895 error = EFBIG; 896 goto out; 897 } 898 } else 899 #endif 900 if (offset + len > vap->va_size) { 901 /* 902 * Test offset + len against the filesystem's maxfilesize. 903 */ 904 VATTR_NULL(vap); 905 vap->va_size = offset + len; 906 error = VOP_SETATTR(vp, vap, td->td_ucred); 907 if (error != 0) 908 goto out; 909 VATTR_NULL(vap); 910 vap->va_size = fsize; 911 error = VOP_SETATTR(vp, vap, td->td_ucred); 912 if (error != 0) 913 goto out; 914 } 915 916 for (;;) { 917 /* 918 * Read and write back anything below the nominal file 919 * size. There's currently no way outside the filesystem 920 * to know whether this area is sparse or not. 921 */ 922 cur = iosize; 923 if ((offset % iosize) != 0) 924 cur -= (offset % iosize); 925 if (cur > len) 926 cur = len; 927 if (offset < fsize) { 928 aiov.iov_base = buf; 929 aiov.iov_len = cur; 930 auio.uio_iov = &aiov; 931 auio.uio_iovcnt = 1; 932 auio.uio_offset = offset; 933 auio.uio_resid = cur; 934 auio.uio_segflg = UIO_SYSSPACE; 935 auio.uio_rw = UIO_READ; 936 auio.uio_td = td; 937 error = VOP_READ(vp, &auio, 0, td->td_ucred); 938 if (error != 0) 939 break; 940 if (auio.uio_resid > 0) { 941 bzero(buf + cur - auio.uio_resid, 942 auio.uio_resid); 943 } 944 } else { 945 bzero(buf, cur); 946 } 947 948 aiov.iov_base = buf; 949 aiov.iov_len = cur; 950 auio.uio_iov = &aiov; 951 auio.uio_iovcnt = 1; 952 auio.uio_offset = offset; 953 auio.uio_resid = cur; 954 auio.uio_segflg = UIO_SYSSPACE; 955 auio.uio_rw = UIO_WRITE; 956 auio.uio_td = td; 957 958 error = VOP_WRITE(vp, &auio, 0, td->td_ucred); 959 if (error != 0) 960 break; 961 962 len -= cur; 963 offset += cur; 964 if (len == 0) 965 break; 966 if (should_yield()) 967 break; 968 } 969 970 out: 971 *ap->a_len = len; 972 *ap->a_offset = offset; 973 free(buf, M_TEMP); 974 return (error); 975 } 976 977 int 978 vop_stdadvise(struct vop_advise_args *ap) 979 { 980 struct vnode *vp; 981 off_t start, end; 982 int error, vfslocked; 983 984 vp = ap->a_vp; 985 switch (ap->a_advice) { 986 case POSIX_FADV_WILLNEED: 987 /* 988 * Do nothing for now. Filesystems should provide a 989 * custom method which starts an asynchronous read of 990 * the requested region. 991 */ 992 error = 0; 993 break; 994 case POSIX_FADV_DONTNEED: 995 /* 996 * Flush any open FS buffers and then remove pages 997 * from the backing VM object. Using vinvalbuf() here 998 * is a bit heavy-handed as it flushes all buffers for 999 * the given vnode, not just the buffers covering the 1000 * requested range. 1001 */ 1002 error = 0; 1003 vfslocked = VFS_LOCK_GIANT(vp->v_mount); 1004 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1005 if (vp->v_iflag & VI_DOOMED) { 1006 VOP_UNLOCK(vp, 0); 1007 VFS_UNLOCK_GIANT(vfslocked); 1008 break; 1009 } 1010 vinvalbuf(vp, V_CLEANONLY, 0, 0); 1011 if (vp->v_object != NULL) { 1012 start = trunc_page(ap->a_start); 1013 end = round_page(ap->a_end); 1014 VM_OBJECT_LOCK(vp->v_object); 1015 vm_object_page_cache(vp->v_object, OFF_TO_IDX(start), 1016 OFF_TO_IDX(end)); 1017 VM_OBJECT_UNLOCK(vp->v_object); 1018 } 1019 VOP_UNLOCK(vp, 0); 1020 VFS_UNLOCK_GIANT(vfslocked); 1021 break; 1022 default: 1023 error = EINVAL; 1024 break; 1025 } 1026 return (error); 1027 } 1028 1029 /* 1030 * vfs default ops 1031 * used to fill the vfs function table to get reasonable default return values. 1032 */ 1033 int 1034 vfs_stdroot (mp, flags, vpp) 1035 struct mount *mp; 1036 int flags; 1037 struct vnode **vpp; 1038 { 1039 1040 return (EOPNOTSUPP); 1041 } 1042 1043 int 1044 vfs_stdstatfs (mp, sbp) 1045 struct mount *mp; 1046 struct statfs *sbp; 1047 { 1048 1049 return (EOPNOTSUPP); 1050 } 1051 1052 int 1053 vfs_stdquotactl (mp, cmds, uid, arg) 1054 struct mount *mp; 1055 int cmds; 1056 uid_t uid; 1057 void *arg; 1058 { 1059 1060 return (EOPNOTSUPP); 1061 } 1062 1063 int 1064 vfs_stdsync(mp, waitfor) 1065 struct mount *mp; 1066 int waitfor; 1067 { 1068 struct vnode *vp, *mvp; 1069 struct thread *td; 1070 int error, lockreq, allerror = 0; 1071 1072 td = curthread; 1073 lockreq = LK_EXCLUSIVE | LK_INTERLOCK; 1074 if (waitfor != MNT_WAIT) 1075 lockreq |= LK_NOWAIT; 1076 /* 1077 * Force stale buffer cache information to be flushed. 1078 */ 1079 MNT_ILOCK(mp); 1080 loop: 1081 MNT_VNODE_FOREACH(vp, mp, mvp) { 1082 /* bv_cnt is an acceptable race here. */ 1083 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) 1084 continue; 1085 VI_LOCK(vp); 1086 MNT_IUNLOCK(mp); 1087 if ((error = vget(vp, lockreq, td)) != 0) { 1088 MNT_ILOCK(mp); 1089 if (error == ENOENT) { 1090 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp); 1091 goto loop; 1092 } 1093 continue; 1094 } 1095 error = VOP_FSYNC(vp, waitfor, td); 1096 if (error) 1097 allerror = error; 1098 1099 /* Do not turn this into vput. td is not always curthread. */ 1100 VOP_UNLOCK(vp, 0); 1101 vrele(vp); 1102 MNT_ILOCK(mp); 1103 } 1104 MNT_IUNLOCK(mp); 1105 return (allerror); 1106 } 1107 1108 int 1109 vfs_stdnosync (mp, waitfor) 1110 struct mount *mp; 1111 int waitfor; 1112 { 1113 1114 return (0); 1115 } 1116 1117 int 1118 vfs_stdvget (mp, ino, flags, vpp) 1119 struct mount *mp; 1120 ino_t ino; 1121 int flags; 1122 struct vnode **vpp; 1123 { 1124 1125 return (EOPNOTSUPP); 1126 } 1127 1128 int 1129 vfs_stdfhtovp (mp, fhp, vpp) 1130 struct mount *mp; 1131 struct fid *fhp; 1132 struct vnode **vpp; 1133 { 1134 1135 return (EOPNOTSUPP); 1136 } 1137 1138 int 1139 vfs_stdinit (vfsp) 1140 struct vfsconf *vfsp; 1141 { 1142 1143 return (0); 1144 } 1145 1146 int 1147 vfs_stduninit (vfsp) 1148 struct vfsconf *vfsp; 1149 { 1150 1151 return(0); 1152 } 1153 1154 int 1155 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname) 1156 struct mount *mp; 1157 int cmd; 1158 struct vnode *filename_vp; 1159 int attrnamespace; 1160 const char *attrname; 1161 { 1162 1163 if (filename_vp != NULL) 1164 VOP_UNLOCK(filename_vp, 0); 1165 return (EOPNOTSUPP); 1166 } 1167 1168 int 1169 vfs_stdsysctl(mp, op, req) 1170 struct mount *mp; 1171 fsctlop_t op; 1172 struct sysctl_req *req; 1173 { 1174 1175 return (EOPNOTSUPP); 1176 } 1177 1178 /* end of vfs default ops */
Cache object: c361a02d99eeef30a43056e4db19e100
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]