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("$FreeBSD$"); 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/namei.h> 51 #include <sys/rwlock.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 static int vop_stdis_text(struct vop_is_text_args *ap); 82 static int vop_stdset_text(struct vop_set_text_args *ap); 83 static int vop_stdunset_text(struct vop_unset_text_args *ap); 84 static int vop_stdget_writecount(struct vop_get_writecount_args *ap); 85 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap); 86 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap); 87 88 /* 89 * This vnode table stores what we want to do if the filesystem doesn't 90 * implement a particular VOP. 91 * 92 * If there is no specific entry here, we will return EOPNOTSUPP. 93 * 94 * Note that every filesystem has to implement either vop_access 95 * or vop_accessx; failing to do so will result in immediate crash 96 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(), 97 * which calls vop_stdaccess() etc. 98 */ 99 100 struct vop_vector default_vnodeops = { 101 .vop_default = NULL, 102 .vop_bypass = VOP_EOPNOTSUPP, 103 104 .vop_access = vop_stdaccess, 105 .vop_accessx = vop_stdaccessx, 106 .vop_advise = vop_stdadvise, 107 .vop_advlock = vop_stdadvlock, 108 .vop_advlockasync = vop_stdadvlockasync, 109 .vop_advlockpurge = vop_stdadvlockpurge, 110 .vop_allocate = vop_stdallocate, 111 .vop_bmap = vop_stdbmap, 112 .vop_close = VOP_NULL, 113 .vop_fsync = VOP_NULL, 114 .vop_getpages = vop_stdgetpages, 115 .vop_getpages_async = vop_stdgetpages_async, 116 .vop_getwritemount = vop_stdgetwritemount, 117 .vop_inactive = VOP_NULL, 118 .vop_ioctl = VOP_ENOTTY, 119 .vop_kqfilter = vop_stdkqfilter, 120 .vop_islocked = vop_stdislocked, 121 .vop_lock1 = vop_stdlock, 122 .vop_lookup = vop_nolookup, 123 .vop_open = VOP_NULL, 124 .vop_pathconf = VOP_EINVAL, 125 .vop_poll = vop_nopoll, 126 .vop_putpages = vop_stdputpages, 127 .vop_readlink = VOP_EINVAL, 128 .vop_rename = vop_norename, 129 .vop_revoke = VOP_PANIC, 130 .vop_strategy = vop_nostrategy, 131 .vop_unlock = vop_stdunlock, 132 .vop_vptocnp = vop_stdvptocnp, 133 .vop_vptofh = vop_stdvptofh, 134 .vop_unp_bind = vop_stdunp_bind, 135 .vop_unp_connect = vop_stdunp_connect, 136 .vop_unp_detach = vop_stdunp_detach, 137 .vop_is_text = vop_stdis_text, 138 .vop_set_text = vop_stdset_text, 139 .vop_unset_text = vop_stdunset_text, 140 .vop_get_writecount = vop_stdget_writecount, 141 .vop_add_writecount = vop_stdadd_writecount, 142 }; 143 144 /* 145 * Series of placeholder functions for various error returns for 146 * VOPs. 147 */ 148 149 int 150 vop_eopnotsupp(struct vop_generic_args *ap) 151 { 152 /* 153 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 154 */ 155 156 return (EOPNOTSUPP); 157 } 158 159 int 160 vop_ebadf(struct vop_generic_args *ap) 161 { 162 163 return (EBADF); 164 } 165 166 int 167 vop_enotty(struct vop_generic_args *ap) 168 { 169 170 return (ENOTTY); 171 } 172 173 int 174 vop_einval(struct vop_generic_args *ap) 175 { 176 177 return (EINVAL); 178 } 179 180 int 181 vop_enoent(struct vop_generic_args *ap) 182 { 183 184 return (ENOENT); 185 } 186 187 int 188 vop_null(struct vop_generic_args *ap) 189 { 190 191 return (0); 192 } 193 194 /* 195 * Helper function to panic on some bad VOPs in some filesystems. 196 */ 197 int 198 vop_panic(struct vop_generic_args *ap) 199 { 200 201 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 202 } 203 204 /* 205 * vop_std<something> and vop_no<something> are default functions for use by 206 * filesystems that need the "default reasonable" implementation for a 207 * particular operation. 208 * 209 * The documentation for the operations they implement exists (if it exists) 210 * in the VOP_<SOMETHING>(9) manpage (all uppercase). 211 */ 212 213 /* 214 * Default vop for filesystems that do not support name lookup 215 */ 216 static int 217 vop_nolookup(ap) 218 struct vop_lookup_args /* { 219 struct vnode *a_dvp; 220 struct vnode **a_vpp; 221 struct componentname *a_cnp; 222 } */ *ap; 223 { 224 225 *ap->a_vpp = NULL; 226 return (ENOTDIR); 227 } 228 229 /* 230 * vop_norename: 231 * 232 * Handle unlock and reference counting for arguments of vop_rename 233 * for filesystems that do not implement rename operation. 234 */ 235 static int 236 vop_norename(struct vop_rename_args *ap) 237 { 238 239 vop_rename_fail(ap); 240 return (EOPNOTSUPP); 241 } 242 243 /* 244 * vop_nostrategy: 245 * 246 * Strategy routine for VFS devices that have none. 247 * 248 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 249 * routine. Typically this is done for a BIO_READ strategy call. 250 * Typically B_INVAL is assumed to already be clear prior to a write 251 * and should not be cleared manually unless you just made the buffer 252 * invalid. BIO_ERROR should be cleared either way. 253 */ 254 255 static int 256 vop_nostrategy (struct vop_strategy_args *ap) 257 { 258 printf("No strategy for buffer at %p\n", ap->a_bp); 259 vprint("vnode", ap->a_vp); 260 ap->a_bp->b_ioflags |= BIO_ERROR; 261 ap->a_bp->b_error = EOPNOTSUPP; 262 bufdone(ap->a_bp); 263 return (EOPNOTSUPP); 264 } 265 266 static int 267 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf, 268 int dirbuflen, off_t *off, char **cpos, int *len, 269 int *eofflag, struct thread *td) 270 { 271 int error, reclen; 272 struct uio uio; 273 struct iovec iov; 274 struct dirent *dp; 275 276 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp)); 277 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp)); 278 279 if (*len == 0) { 280 iov.iov_base = dirbuf; 281 iov.iov_len = dirbuflen; 282 283 uio.uio_iov = &iov; 284 uio.uio_iovcnt = 1; 285 uio.uio_offset = *off; 286 uio.uio_resid = dirbuflen; 287 uio.uio_segflg = UIO_SYSSPACE; 288 uio.uio_rw = UIO_READ; 289 uio.uio_td = td; 290 291 *eofflag = 0; 292 293 #ifdef MAC 294 error = mac_vnode_check_readdir(td->td_ucred, vp); 295 if (error == 0) 296 #endif 297 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag, 298 NULL, NULL); 299 if (error) 300 return (error); 301 302 *off = uio.uio_offset; 303 304 *cpos = dirbuf; 305 *len = (dirbuflen - uio.uio_resid); 306 307 if (*len == 0) 308 return (ENOENT); 309 } 310 311 dp = (struct dirent *)(*cpos); 312 reclen = dp->d_reclen; 313 *dpp = dp; 314 315 /* check for malformed directory.. */ 316 if (reclen < DIRENT_MINSIZE) 317 return (EINVAL); 318 319 *cpos += reclen; 320 *len -= reclen; 321 322 return (0); 323 } 324 325 /* 326 * Check if a named file exists in a given directory vnode. 327 */ 328 static int 329 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td) 330 { 331 char *dirbuf, *cpos; 332 int error, eofflag, dirbuflen, len, found; 333 off_t off; 334 struct dirent *dp; 335 struct vattr va; 336 337 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp)); 338 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp)); 339 340 found = 0; 341 342 error = VOP_GETATTR(vp, &va, td->td_ucred); 343 if (error) 344 return (found); 345 346 dirbuflen = DEV_BSIZE; 347 if (dirbuflen < va.va_blocksize) 348 dirbuflen = va.va_blocksize; 349 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK); 350 351 off = 0; 352 len = 0; 353 do { 354 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off, 355 &cpos, &len, &eofflag, td); 356 if (error) 357 goto out; 358 359 if (dp->d_type != DT_WHT && dp->d_fileno != 0 && 360 strcmp(dp->d_name, dirname) == 0) { 361 found = 1; 362 goto out; 363 } 364 } while (len > 0 || !eofflag); 365 366 out: 367 free(dirbuf, M_TEMP); 368 return (found); 369 } 370 371 int 372 vop_stdaccess(struct vop_access_args *ap) 373 { 374 375 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN | 376 VAPPEND)) == 0, ("invalid bit in accmode")); 377 378 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td)); 379 } 380 381 int 382 vop_stdaccessx(struct vop_accessx_args *ap) 383 { 384 int error; 385 accmode_t accmode = ap->a_accmode; 386 387 error = vfs_unixify_accmode(&accmode); 388 if (error != 0) 389 return (error); 390 391 if (accmode == 0) 392 return (0); 393 394 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td)); 395 } 396 397 /* 398 * Advisory record locking support 399 */ 400 int 401 vop_stdadvlock(struct vop_advlock_args *ap) 402 { 403 struct vnode *vp; 404 struct ucred *cred; 405 struct vattr vattr; 406 int error; 407 408 vp = ap->a_vp; 409 cred = curthread->td_ucred; 410 vn_lock(vp, LK_SHARED | LK_RETRY); 411 error = VOP_GETATTR(vp, &vattr, cred); 412 VOP_UNLOCK(vp, 0); 413 if (error) 414 return (error); 415 416 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size)); 417 } 418 419 int 420 vop_stdadvlockasync(struct vop_advlockasync_args *ap) 421 { 422 struct vnode *vp; 423 struct ucred *cred; 424 struct vattr vattr; 425 int error; 426 427 vp = ap->a_vp; 428 cred = curthread->td_ucred; 429 vn_lock(vp, LK_SHARED | LK_RETRY); 430 error = VOP_GETATTR(vp, &vattr, cred); 431 VOP_UNLOCK(vp, 0); 432 if (error) 433 return (error); 434 435 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size)); 436 } 437 438 int 439 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap) 440 { 441 struct vnode *vp; 442 443 vp = ap->a_vp; 444 lf_purgelocks(vp, &vp->v_lockf); 445 return (0); 446 } 447 448 /* 449 * vop_stdpathconf: 450 * 451 * Standard implementation of POSIX pathconf, to get information about limits 452 * for a filesystem. 453 * Override per filesystem for the case where the filesystem has smaller 454 * limits. 455 */ 456 int 457 vop_stdpathconf(ap) 458 struct vop_pathconf_args /* { 459 struct vnode *a_vp; 460 int a_name; 461 int *a_retval; 462 } */ *ap; 463 { 464 465 switch (ap->a_name) { 466 case _PC_NAME_MAX: 467 *ap->a_retval = NAME_MAX; 468 return (0); 469 case _PC_PATH_MAX: 470 *ap->a_retval = PATH_MAX; 471 return (0); 472 case _PC_LINK_MAX: 473 *ap->a_retval = LINK_MAX; 474 return (0); 475 case _PC_MAX_CANON: 476 *ap->a_retval = MAX_CANON; 477 return (0); 478 case _PC_MAX_INPUT: 479 *ap->a_retval = MAX_INPUT; 480 return (0); 481 case _PC_PIPE_BUF: 482 *ap->a_retval = PIPE_BUF; 483 return (0); 484 case _PC_CHOWN_RESTRICTED: 485 *ap->a_retval = 1; 486 return (0); 487 case _PC_VDISABLE: 488 *ap->a_retval = _POSIX_VDISABLE; 489 return (0); 490 default: 491 return (EINVAL); 492 } 493 /* NOTREACHED */ 494 } 495 496 /* 497 * Standard lock, unlock and islocked functions. 498 */ 499 int 500 vop_stdlock(ap) 501 struct vop_lock1_args /* { 502 struct vnode *a_vp; 503 int a_flags; 504 char *file; 505 int line; 506 } */ *ap; 507 { 508 struct vnode *vp = ap->a_vp; 509 510 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp), 511 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file, 512 ap->a_line)); 513 } 514 515 /* See above. */ 516 int 517 vop_stdunlock(ap) 518 struct vop_unlock_args /* { 519 struct vnode *a_vp; 520 int a_flags; 521 } */ *ap; 522 { 523 struct vnode *vp = ap->a_vp; 524 525 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp))); 526 } 527 528 /* See above. */ 529 int 530 vop_stdislocked(ap) 531 struct vop_islocked_args /* { 532 struct vnode *a_vp; 533 } */ *ap; 534 { 535 536 return (lockstatus(ap->a_vp->v_vnlock)); 537 } 538 539 /* 540 * Return true for select/poll. 541 */ 542 int 543 vop_nopoll(ap) 544 struct vop_poll_args /* { 545 struct vnode *a_vp; 546 int a_events; 547 struct ucred *a_cred; 548 struct thread *a_td; 549 } */ *ap; 550 { 551 552 return (poll_no_poll(ap->a_events)); 553 } 554 555 /* 556 * Implement poll for local filesystems that support it. 557 */ 558 int 559 vop_stdpoll(ap) 560 struct vop_poll_args /* { 561 struct vnode *a_vp; 562 int a_events; 563 struct ucred *a_cred; 564 struct thread *a_td; 565 } */ *ap; 566 { 567 if (ap->a_events & ~POLLSTANDARD) 568 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 569 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 570 } 571 572 /* 573 * Return our mount point, as we will take charge of the writes. 574 */ 575 int 576 vop_stdgetwritemount(ap) 577 struct vop_getwritemount_args /* { 578 struct vnode *a_vp; 579 struct mount **a_mpp; 580 } */ *ap; 581 { 582 struct mount *mp; 583 584 /* 585 * XXX Since this is called unlocked we may be recycled while 586 * attempting to ref the mount. If this is the case or mountpoint 587 * will be set to NULL. We only have to prevent this call from 588 * returning with a ref to an incorrect mountpoint. It is not 589 * harmful to return with a ref to our previous mountpoint. 590 */ 591 mp = ap->a_vp->v_mount; 592 if (mp != NULL) { 593 vfs_ref(mp); 594 if (mp != ap->a_vp->v_mount) { 595 vfs_rel(mp); 596 mp = NULL; 597 } 598 } 599 *(ap->a_mpp) = mp; 600 return (0); 601 } 602 603 /* XXX Needs good comment and VOP_BMAP(9) manpage */ 604 int 605 vop_stdbmap(ap) 606 struct vop_bmap_args /* { 607 struct vnode *a_vp; 608 daddr_t a_bn; 609 struct bufobj **a_bop; 610 daddr_t *a_bnp; 611 int *a_runp; 612 int *a_runb; 613 } */ *ap; 614 { 615 616 if (ap->a_bop != NULL) 617 *ap->a_bop = &ap->a_vp->v_bufobj; 618 if (ap->a_bnp != NULL) 619 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 620 if (ap->a_runp != NULL) 621 *ap->a_runp = 0; 622 if (ap->a_runb != NULL) 623 *ap->a_runb = 0; 624 return (0); 625 } 626 627 int 628 vop_stdfsync(ap) 629 struct vop_fsync_args /* { 630 struct vnode *a_vp; 631 struct ucred *a_cred; 632 int a_waitfor; 633 struct thread *a_td; 634 } */ *ap; 635 { 636 struct vnode *vp = ap->a_vp; 637 struct buf *bp; 638 struct bufobj *bo; 639 struct buf *nbp; 640 int error = 0; 641 int maxretry = 1000; /* large, arbitrarily chosen */ 642 643 bo = &vp->v_bufobj; 644 BO_LOCK(bo); 645 loop1: 646 /* 647 * MARK/SCAN initialization to avoid infinite loops. 648 */ 649 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) { 650 bp->b_vflags &= ~BV_SCANNED; 651 bp->b_error = 0; 652 } 653 654 /* 655 * Flush all dirty buffers associated with a vnode. 656 */ 657 loop2: 658 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) { 659 if ((bp->b_vflags & BV_SCANNED) != 0) 660 continue; 661 bp->b_vflags |= BV_SCANNED; 662 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) { 663 if (ap->a_waitfor != MNT_WAIT) 664 continue; 665 if (BUF_LOCK(bp, 666 LK_EXCLUSIVE | LK_INTERLOCK | LK_SLEEPFAIL, 667 BO_LOCKPTR(bo)) != 0) { 668 BO_LOCK(bo); 669 goto loop1; 670 } 671 BO_LOCK(bo); 672 } 673 BO_UNLOCK(bo); 674 KASSERT(bp->b_bufobj == bo, 675 ("bp %p wrong b_bufobj %p should be %p", 676 bp, bp->b_bufobj, bo)); 677 if ((bp->b_flags & B_DELWRI) == 0) 678 panic("fsync: not dirty"); 679 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) { 680 vfs_bio_awrite(bp); 681 } else { 682 bremfree(bp); 683 bawrite(bp); 684 } 685 BO_LOCK(bo); 686 goto loop2; 687 } 688 689 /* 690 * If synchronous the caller expects us to completely resolve all 691 * dirty buffers in the system. Wait for in-progress I/O to 692 * complete (which could include background bitmap writes), then 693 * retry if dirty blocks still exist. 694 */ 695 if (ap->a_waitfor == MNT_WAIT) { 696 bufobj_wwait(bo, 0, 0); 697 if (bo->bo_dirty.bv_cnt > 0) { 698 /* 699 * If we are unable to write any of these buffers 700 * then we fail now rather than trying endlessly 701 * to write them out. 702 */ 703 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) 704 if ((error = bp->b_error) == 0) 705 continue; 706 if (error == 0 && --maxretry >= 0) 707 goto loop1; 708 error = EAGAIN; 709 } 710 } 711 BO_UNLOCK(bo); 712 if (error == EAGAIN) 713 vprint("fsync: giving up on dirty", vp); 714 715 return (error); 716 } 717 718 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */ 719 int 720 vop_stdgetpages(ap) 721 struct vop_getpages_args /* { 722 struct vnode *a_vp; 723 vm_page_t *a_m; 724 int a_count; 725 int a_reqpage; 726 } */ *ap; 727 { 728 729 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, 730 ap->a_count, ap->a_reqpage, NULL, NULL); 731 } 732 733 static int 734 vop_stdgetpages_async(struct vop_getpages_async_args *ap) 735 { 736 int error; 737 738 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_reqpage); 739 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_reqpage, error); 740 return (error); 741 } 742 743 int 744 vop_stdkqfilter(struct vop_kqfilter_args *ap) 745 { 746 return vfs_kqfilter(ap); 747 } 748 749 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */ 750 int 751 vop_stdputpages(ap) 752 struct vop_putpages_args /* { 753 struct vnode *a_vp; 754 vm_page_t *a_m; 755 int a_count; 756 int a_sync; 757 int *a_rtvals; 758 } */ *ap; 759 { 760 761 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count, 762 ap->a_sync, ap->a_rtvals); 763 } 764 765 int 766 vop_stdvptofh(struct vop_vptofh_args *ap) 767 { 768 return (EOPNOTSUPP); 769 } 770 771 int 772 vop_stdvptocnp(struct vop_vptocnp_args *ap) 773 { 774 struct vnode *vp = ap->a_vp; 775 struct vnode **dvp = ap->a_vpp; 776 struct ucred *cred = ap->a_cred; 777 char *buf = ap->a_buf; 778 int *buflen = ap->a_buflen; 779 char *dirbuf, *cpos; 780 int i, error, eofflag, dirbuflen, flags, locked, len, covered; 781 off_t off; 782 ino_t fileno; 783 struct vattr va; 784 struct nameidata nd; 785 struct thread *td; 786 struct dirent *dp; 787 struct vnode *mvp; 788 789 i = *buflen; 790 error = 0; 791 covered = 0; 792 td = curthread; 793 794 if (vp->v_type != VDIR) 795 return (ENOENT); 796 797 error = VOP_GETATTR(vp, &va, cred); 798 if (error) 799 return (error); 800 801 VREF(vp); 802 locked = VOP_ISLOCKED(vp); 803 VOP_UNLOCK(vp, 0); 804 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 805 "..", vp, td); 806 flags = FREAD; 807 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL); 808 if (error) { 809 vn_lock(vp, locked | LK_RETRY); 810 return (error); 811 } 812 NDFREE(&nd, NDF_ONLY_PNBUF); 813 814 mvp = *dvp = nd.ni_vp; 815 816 if (vp->v_mount != (*dvp)->v_mount && 817 ((*dvp)->v_vflag & VV_ROOT) && 818 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) { 819 *dvp = (*dvp)->v_mount->mnt_vnodecovered; 820 VREF(mvp); 821 VOP_UNLOCK(mvp, 0); 822 vn_close(mvp, FREAD, cred, td); 823 VREF(*dvp); 824 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY); 825 covered = 1; 826 } 827 828 fileno = va.va_fileid; 829 830 dirbuflen = DEV_BSIZE; 831 if (dirbuflen < va.va_blocksize) 832 dirbuflen = va.va_blocksize; 833 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK); 834 835 if ((*dvp)->v_type != VDIR) { 836 error = ENOENT; 837 goto out; 838 } 839 840 off = 0; 841 len = 0; 842 do { 843 /* call VOP_READDIR of parent */ 844 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off, 845 &cpos, &len, &eofflag, td); 846 if (error) 847 goto out; 848 849 if ((dp->d_type != DT_WHT) && 850 (dp->d_fileno == fileno)) { 851 if (covered) { 852 VOP_UNLOCK(*dvp, 0); 853 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY); 854 if (dirent_exists(mvp, dp->d_name, td)) { 855 error = ENOENT; 856 VOP_UNLOCK(mvp, 0); 857 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY); 858 goto out; 859 } 860 VOP_UNLOCK(mvp, 0); 861 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY); 862 } 863 i -= dp->d_namlen; 864 865 if (i < 0) { 866 error = ENOMEM; 867 goto out; 868 } 869 if (dp->d_namlen == 1 && dp->d_name[0] == '.') { 870 error = ENOENT; 871 } else { 872 bcopy(dp->d_name, buf + i, dp->d_namlen); 873 error = 0; 874 } 875 goto out; 876 } 877 } while (len > 0 || !eofflag); 878 error = ENOENT; 879 880 out: 881 free(dirbuf, M_TEMP); 882 if (!error) { 883 *buflen = i; 884 vref(*dvp); 885 } 886 if (covered) { 887 vput(*dvp); 888 vrele(mvp); 889 } else { 890 VOP_UNLOCK(mvp, 0); 891 vn_close(mvp, FREAD, cred, td); 892 } 893 vn_lock(vp, locked | LK_RETRY); 894 return (error); 895 } 896 897 int 898 vop_stdallocate(struct vop_allocate_args *ap) 899 { 900 #ifdef __notyet__ 901 struct statfs sfs; 902 #endif 903 struct iovec aiov; 904 struct vattr vattr, *vap; 905 struct uio auio; 906 off_t fsize, len, cur, offset; 907 uint8_t *buf; 908 struct thread *td; 909 struct vnode *vp; 910 size_t iosize; 911 int error; 912 913 buf = NULL; 914 error = 0; 915 td = curthread; 916 vap = &vattr; 917 vp = ap->a_vp; 918 len = *ap->a_len; 919 offset = *ap->a_offset; 920 921 error = VOP_GETATTR(vp, vap, td->td_ucred); 922 if (error != 0) 923 goto out; 924 fsize = vap->va_size; 925 iosize = vap->va_blocksize; 926 if (iosize == 0) 927 iosize = BLKDEV_IOSIZE; 928 if (iosize > MAXPHYS) 929 iosize = MAXPHYS; 930 buf = malloc(iosize, M_TEMP, M_WAITOK); 931 932 #ifdef __notyet__ 933 /* 934 * Check if the filesystem sets f_maxfilesize; if not use 935 * VOP_SETATTR to perform the check. 936 */ 937 error = VFS_STATFS(vp->v_mount, &sfs, td); 938 if (error != 0) 939 goto out; 940 if (sfs.f_maxfilesize) { 941 if (offset > sfs.f_maxfilesize || len > sfs.f_maxfilesize || 942 offset + len > sfs.f_maxfilesize) { 943 error = EFBIG; 944 goto out; 945 } 946 } else 947 #endif 948 if (offset + len > vap->va_size) { 949 /* 950 * Test offset + len against the filesystem's maxfilesize. 951 */ 952 VATTR_NULL(vap); 953 vap->va_size = offset + len; 954 error = VOP_SETATTR(vp, vap, td->td_ucred); 955 if (error != 0) 956 goto out; 957 VATTR_NULL(vap); 958 vap->va_size = fsize; 959 error = VOP_SETATTR(vp, vap, td->td_ucred); 960 if (error != 0) 961 goto out; 962 } 963 964 for (;;) { 965 /* 966 * Read and write back anything below the nominal file 967 * size. There's currently no way outside the filesystem 968 * to know whether this area is sparse or not. 969 */ 970 cur = iosize; 971 if ((offset % iosize) != 0) 972 cur -= (offset % iosize); 973 if (cur > len) 974 cur = len; 975 if (offset < fsize) { 976 aiov.iov_base = buf; 977 aiov.iov_len = cur; 978 auio.uio_iov = &aiov; 979 auio.uio_iovcnt = 1; 980 auio.uio_offset = offset; 981 auio.uio_resid = cur; 982 auio.uio_segflg = UIO_SYSSPACE; 983 auio.uio_rw = UIO_READ; 984 auio.uio_td = td; 985 error = VOP_READ(vp, &auio, 0, td->td_ucred); 986 if (error != 0) 987 break; 988 if (auio.uio_resid > 0) { 989 bzero(buf + cur - auio.uio_resid, 990 auio.uio_resid); 991 } 992 } else { 993 bzero(buf, cur); 994 } 995 996 aiov.iov_base = buf; 997 aiov.iov_len = cur; 998 auio.uio_iov = &aiov; 999 auio.uio_iovcnt = 1; 1000 auio.uio_offset = offset; 1001 auio.uio_resid = cur; 1002 auio.uio_segflg = UIO_SYSSPACE; 1003 auio.uio_rw = UIO_WRITE; 1004 auio.uio_td = td; 1005 1006 error = VOP_WRITE(vp, &auio, 0, td->td_ucred); 1007 if (error != 0) 1008 break; 1009 1010 len -= cur; 1011 offset += cur; 1012 if (len == 0) 1013 break; 1014 if (should_yield()) 1015 break; 1016 } 1017 1018 out: 1019 *ap->a_len = len; 1020 *ap->a_offset = offset; 1021 free(buf, M_TEMP); 1022 return (error); 1023 } 1024 1025 int 1026 vop_stdadvise(struct vop_advise_args *ap) 1027 { 1028 struct vnode *vp; 1029 off_t start, end; 1030 int error; 1031 1032 vp = ap->a_vp; 1033 switch (ap->a_advice) { 1034 case POSIX_FADV_WILLNEED: 1035 /* 1036 * Do nothing for now. Filesystems should provide a 1037 * custom method which starts an asynchronous read of 1038 * the requested region. 1039 */ 1040 error = 0; 1041 break; 1042 case POSIX_FADV_DONTNEED: 1043 /* 1044 * Flush any open FS buffers and then remove pages 1045 * from the backing VM object. Using vinvalbuf() here 1046 * is a bit heavy-handed as it flushes all buffers for 1047 * the given vnode, not just the buffers covering the 1048 * requested range. 1049 */ 1050 error = 0; 1051 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1052 if (vp->v_iflag & VI_DOOMED) { 1053 VOP_UNLOCK(vp, 0); 1054 break; 1055 } 1056 vinvalbuf(vp, V_CLEANONLY, 0, 0); 1057 if (vp->v_object != NULL) { 1058 start = trunc_page(ap->a_start); 1059 end = round_page(ap->a_end); 1060 VM_OBJECT_WLOCK(vp->v_object); 1061 vm_object_page_cache(vp->v_object, OFF_TO_IDX(start), 1062 OFF_TO_IDX(end)); 1063 VM_OBJECT_WUNLOCK(vp->v_object); 1064 } 1065 VOP_UNLOCK(vp, 0); 1066 break; 1067 default: 1068 error = EINVAL; 1069 break; 1070 } 1071 return (error); 1072 } 1073 1074 int 1075 vop_stdunp_bind(struct vop_unp_bind_args *ap) 1076 { 1077 1078 ap->a_vp->v_socket = ap->a_socket; 1079 return (0); 1080 } 1081 1082 int 1083 vop_stdunp_connect(struct vop_unp_connect_args *ap) 1084 { 1085 1086 *ap->a_socket = ap->a_vp->v_socket; 1087 return (0); 1088 } 1089 1090 int 1091 vop_stdunp_detach(struct vop_unp_detach_args *ap) 1092 { 1093 1094 ap->a_vp->v_socket = NULL; 1095 return (0); 1096 } 1097 1098 static int 1099 vop_stdis_text(struct vop_is_text_args *ap) 1100 { 1101 1102 return ((ap->a_vp->v_vflag & VV_TEXT) != 0); 1103 } 1104 1105 static int 1106 vop_stdset_text(struct vop_set_text_args *ap) 1107 { 1108 1109 ap->a_vp->v_vflag |= VV_TEXT; 1110 return (0); 1111 } 1112 1113 static int 1114 vop_stdunset_text(struct vop_unset_text_args *ap) 1115 { 1116 1117 ap->a_vp->v_vflag &= ~VV_TEXT; 1118 return (0); 1119 } 1120 1121 static int 1122 vop_stdget_writecount(struct vop_get_writecount_args *ap) 1123 { 1124 1125 *ap->a_writecount = ap->a_vp->v_writecount; 1126 return (0); 1127 } 1128 1129 static int 1130 vop_stdadd_writecount(struct vop_add_writecount_args *ap) 1131 { 1132 1133 ap->a_vp->v_writecount += ap->a_inc; 1134 return (0); 1135 } 1136 1137 /* 1138 * vfs default ops 1139 * used to fill the vfs function table to get reasonable default return values. 1140 */ 1141 int 1142 vfs_stdroot (mp, flags, vpp) 1143 struct mount *mp; 1144 int flags; 1145 struct vnode **vpp; 1146 { 1147 1148 return (EOPNOTSUPP); 1149 } 1150 1151 int 1152 vfs_stdstatfs (mp, sbp) 1153 struct mount *mp; 1154 struct statfs *sbp; 1155 { 1156 1157 return (EOPNOTSUPP); 1158 } 1159 1160 int 1161 vfs_stdquotactl (mp, cmds, uid, arg) 1162 struct mount *mp; 1163 int cmds; 1164 uid_t uid; 1165 void *arg; 1166 { 1167 1168 return (EOPNOTSUPP); 1169 } 1170 1171 int 1172 vfs_stdsync(mp, waitfor) 1173 struct mount *mp; 1174 int waitfor; 1175 { 1176 struct vnode *vp, *mvp; 1177 struct thread *td; 1178 int error, lockreq, allerror = 0; 1179 1180 td = curthread; 1181 lockreq = LK_EXCLUSIVE | LK_INTERLOCK; 1182 if (waitfor != MNT_WAIT) 1183 lockreq |= LK_NOWAIT; 1184 /* 1185 * Force stale buffer cache information to be flushed. 1186 */ 1187 loop: 1188 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 1189 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) { 1190 VI_UNLOCK(vp); 1191 continue; 1192 } 1193 if ((error = vget(vp, lockreq, td)) != 0) { 1194 if (error == ENOENT) { 1195 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 1196 goto loop; 1197 } 1198 continue; 1199 } 1200 error = VOP_FSYNC(vp, waitfor, td); 1201 if (error) 1202 allerror = error; 1203 vput(vp); 1204 } 1205 return (allerror); 1206 } 1207 1208 int 1209 vfs_stdnosync (mp, waitfor) 1210 struct mount *mp; 1211 int waitfor; 1212 { 1213 1214 return (0); 1215 } 1216 1217 int 1218 vfs_stdvget (mp, ino, flags, vpp) 1219 struct mount *mp; 1220 ino_t ino; 1221 int flags; 1222 struct vnode **vpp; 1223 { 1224 1225 return (EOPNOTSUPP); 1226 } 1227 1228 int 1229 vfs_stdfhtovp (mp, fhp, flags, vpp) 1230 struct mount *mp; 1231 struct fid *fhp; 1232 int flags; 1233 struct vnode **vpp; 1234 { 1235 1236 return (EOPNOTSUPP); 1237 } 1238 1239 int 1240 vfs_stdinit (vfsp) 1241 struct vfsconf *vfsp; 1242 { 1243 1244 return (0); 1245 } 1246 1247 int 1248 vfs_stduninit (vfsp) 1249 struct vfsconf *vfsp; 1250 { 1251 1252 return(0); 1253 } 1254 1255 int 1256 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname) 1257 struct mount *mp; 1258 int cmd; 1259 struct vnode *filename_vp; 1260 int attrnamespace; 1261 const char *attrname; 1262 { 1263 1264 if (filename_vp != NULL) 1265 VOP_UNLOCK(filename_vp, 0); 1266 return (EOPNOTSUPP); 1267 } 1268 1269 int 1270 vfs_stdsysctl(mp, op, req) 1271 struct mount *mp; 1272 fsctlop_t op; 1273 struct sysctl_req *req; 1274 { 1275 1276 return (EOPNOTSUPP); 1277 } 1278 1279 /* end of vfs default ops */ 1280