vfs_default.c revision 144584
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: head/sys/kern/vfs_default.c 144584 2005-04-03 10:24:03Z jeff $"); 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/kernel.h> 44#include <sys/limits.h> 45#include <sys/lock.h> 46#include <sys/malloc.h> 47#include <sys/mount.h> 48#include <sys/mutex.h> 49#include <sys/unistd.h> 50#include <sys/vnode.h> 51#include <sys/poll.h> 52 53#include <vm/vm.h> 54#include <vm/vm_object.h> 55#include <vm/vm_extern.h> 56#include <vm/pmap.h> 57#include <vm/vm_map.h> 58#include <vm/vm_page.h> 59#include <vm/vm_pager.h> 60#include <vm/vnode_pager.h> 61 62static int vop_nolookup(struct vop_lookup_args *); 63static int vop_nostrategy(struct vop_strategy_args *); 64 65/* 66 * This vnode table stores what we want to do if the filesystem doesn't 67 * implement a particular VOP. 68 * 69 * If there is no specific entry here, we will return EOPNOTSUPP. 70 * 71 */ 72 73struct vop_vector default_vnodeops = { 74 .vop_default = NULL, 75 .vop_bypass = VOP_EOPNOTSUPP, 76 77 .vop_advlock = VOP_EINVAL, 78 .vop_bmap = vop_stdbmap, 79 .vop_close = VOP_NULL, 80 .vop_fsync = VOP_NULL, 81 .vop_getpages = vop_stdgetpages, 82 .vop_getwritemount = vop_stdgetwritemount, 83 .vop_inactive = VOP_NULL, 84 .vop_ioctl = VOP_ENOTTY, 85 .vop_islocked = vop_stdislocked, 86 .vop_lease = VOP_NULL, 87 .vop_lock = vop_stdlock, 88 .vop_lookup = vop_nolookup, 89 .vop_open = VOP_NULL, 90 .vop_pathconf = VOP_EINVAL, 91 .vop_poll = vop_nopoll, 92 .vop_putpages = vop_stdputpages, 93 .vop_readlink = VOP_EINVAL, 94 .vop_revoke = VOP_PANIC, 95 .vop_strategy = vop_nostrategy, 96 .vop_unlock = vop_stdunlock, 97}; 98 99/* 100 * Series of placeholder functions for various error returns for 101 * VOPs. 102 */ 103 104int 105vop_eopnotsupp(struct vop_generic_args *ap) 106{ 107 /* 108 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 109 */ 110 111 return (EOPNOTSUPP); 112} 113 114int 115vop_ebadf(struct vop_generic_args *ap) 116{ 117 118 return (EBADF); 119} 120 121int 122vop_enotty(struct vop_generic_args *ap) 123{ 124 125 return (ENOTTY); 126} 127 128int 129vop_einval(struct vop_generic_args *ap) 130{ 131 132 return (EINVAL); 133} 134 135int 136vop_null(struct vop_generic_args *ap) 137{ 138 139 return (0); 140} 141 142/* 143 * Helper function to panic on some bad VOPs in some filesystems. 144 */ 145int 146vop_panic(struct vop_generic_args *ap) 147{ 148 149 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 150} 151 152/* 153 * vop_std<something> and vop_no<something> are default functions for use by 154 * filesystems that need the "default reasonable" implementation for a 155 * particular operation. 156 * 157 * The documentation for the operations they implement exists (if it exists) 158 * in the VOP_<SOMETHING>(9) manpage (all uppercase). 159 */ 160 161/* 162 * Default vop for filesystems that do not support name lookup 163 */ 164static int 165vop_nolookup(ap) 166 struct vop_lookup_args /* { 167 struct vnode *a_dvp; 168 struct vnode **a_vpp; 169 struct componentname *a_cnp; 170 } */ *ap; 171{ 172 173 *ap->a_vpp = NULL; 174 return (ENOTDIR); 175} 176 177/* 178 * vop_nostrategy: 179 * 180 * Strategy routine for VFS devices that have none. 181 * 182 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 183 * routine. Typically this is done for a BIO_READ strategy call. 184 * Typically B_INVAL is assumed to already be clear prior to a write 185 * and should not be cleared manually unless you just made the buffer 186 * invalid. BIO_ERROR should be cleared either way. 187 */ 188 189static int 190vop_nostrategy (struct vop_strategy_args *ap) 191{ 192 printf("No strategy for buffer at %p\n", ap->a_bp); 193 vprint("vnode", ap->a_vp); 194 ap->a_bp->b_ioflags |= BIO_ERROR; 195 ap->a_bp->b_error = EOPNOTSUPP; 196 bufdone(ap->a_bp); 197 return (EOPNOTSUPP); 198} 199 200/* 201 * vop_stdpathconf: 202 * 203 * Standard implementation of POSIX pathconf, to get information about limits 204 * for a filesystem. 205 * Override per filesystem for the case where the filesystem has smaller 206 * limits. 207 */ 208int 209vop_stdpathconf(ap) 210 struct vop_pathconf_args /* { 211 struct vnode *a_vp; 212 int a_name; 213 int *a_retval; 214 } */ *ap; 215{ 216 217 switch (ap->a_name) { 218 case _PC_LINK_MAX: 219 *ap->a_retval = LINK_MAX; 220 return (0); 221 case _PC_MAX_CANON: 222 *ap->a_retval = MAX_CANON; 223 return (0); 224 case _PC_MAX_INPUT: 225 *ap->a_retval = MAX_INPUT; 226 return (0); 227 case _PC_PIPE_BUF: 228 *ap->a_retval = PIPE_BUF; 229 return (0); 230 case _PC_CHOWN_RESTRICTED: 231 *ap->a_retval = 1; 232 return (0); 233 case _PC_VDISABLE: 234 *ap->a_retval = _POSIX_VDISABLE; 235 return (0); 236 default: 237 return (EINVAL); 238 } 239 /* NOTREACHED */ 240} 241 242/* 243 * Standard lock, unlock and islocked functions. 244 */ 245int 246vop_stdlock(ap) 247 struct vop_lock_args /* { 248 struct vnode *a_vp; 249 int a_flags; 250 struct thread *a_td; 251 } */ *ap; 252{ 253 struct vnode *vp = ap->a_vp; 254 255#ifndef DEBUG_LOCKS 256 return (lockmgr(vp->v_vnlock, ap->a_flags, VI_MTX(vp), ap->a_td)); 257#else 258 return (debuglockmgr(vp->v_vnlock, ap->a_flags, VI_MTX(vp), 259 ap->a_td, "vop_stdlock", vp->filename, vp->line)); 260#endif 261} 262 263/* See above. */ 264int 265vop_stdunlock(ap) 266 struct vop_unlock_args /* { 267 struct vnode *a_vp; 268 int a_flags; 269 struct thread *a_td; 270 } */ *ap; 271{ 272 struct vnode *vp = ap->a_vp; 273 274 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp), 275 ap->a_td)); 276} 277 278/* See above. */ 279int 280vop_stdislocked(ap) 281 struct vop_islocked_args /* { 282 struct vnode *a_vp; 283 struct thread *a_td; 284 } */ *ap; 285{ 286 287 return (lockstatus(ap->a_vp->v_vnlock, ap->a_td)); 288} 289 290/* 291 * Return true for select/poll. 292 */ 293int 294vop_nopoll(ap) 295 struct vop_poll_args /* { 296 struct vnode *a_vp; 297 int a_events; 298 struct ucred *a_cred; 299 struct thread *a_td; 300 } */ *ap; 301{ 302 /* 303 * Return true for read/write. If the user asked for something 304 * special, return POLLNVAL, so that clients have a way of 305 * determining reliably whether or not the extended 306 * functionality is present without hard-coding knowledge 307 * of specific filesystem implementations. 308 * Stay in sync with kern_conf.c::no_poll(). 309 */ 310 if (ap->a_events & ~POLLSTANDARD) 311 return (POLLNVAL); 312 313 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 314} 315 316/* 317 * Implement poll for local filesystems that support it. 318 */ 319int 320vop_stdpoll(ap) 321 struct vop_poll_args /* { 322 struct vnode *a_vp; 323 int a_events; 324 struct ucred *a_cred; 325 struct thread *a_td; 326 } */ *ap; 327{ 328 if (ap->a_events & ~POLLSTANDARD) 329 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 330 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 331} 332 333/* 334 * Return our mount point, as we will take charge of the writes. 335 */ 336int 337vop_stdgetwritemount(ap) 338 struct vop_getwritemount_args /* { 339 struct vnode *a_vp; 340 struct mount **a_mpp; 341 } */ *ap; 342{ 343 344 *(ap->a_mpp) = ap->a_vp->v_mount; 345 return (0); 346} 347 348/* XXX Needs good comment and VOP_BMAP(9) manpage */ 349int 350vop_stdbmap(ap) 351 struct vop_bmap_args /* { 352 struct vnode *a_vp; 353 daddr_t a_bn; 354 struct bufobj **a_bop; 355 daddr_t *a_bnp; 356 int *a_runp; 357 int *a_runb; 358 } */ *ap; 359{ 360 361 if (ap->a_bop != NULL) 362 *ap->a_bop = &ap->a_vp->v_bufobj; 363 if (ap->a_bnp != NULL) 364 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 365 if (ap->a_runp != NULL) 366 *ap->a_runp = 0; 367 if (ap->a_runb != NULL) 368 *ap->a_runb = 0; 369 return (0); 370} 371 372int 373vop_stdfsync(ap) 374 struct vop_fsync_args /* { 375 struct vnode *a_vp; 376 struct ucred *a_cred; 377 int a_waitfor; 378 struct thread *a_td; 379 } */ *ap; 380{ 381 struct vnode *vp = ap->a_vp; 382 struct buf *bp; 383 struct bufobj *bo; 384 struct buf *nbp; 385 int s, error = 0; 386 int maxretry = 1000; /* large, arbitrarily chosen */ 387 388 VI_LOCK(vp); 389loop1: 390 /* 391 * MARK/SCAN initialization to avoid infinite loops. 392 */ 393 s = splbio(); 394 TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs) { 395 bp->b_vflags &= ~BV_SCANNED; 396 bp->b_error = 0; 397 } 398 splx(s); 399 400 /* 401 * Flush all dirty buffers associated with a vnode. 402 */ 403loop2: 404 s = splbio(); 405 TAILQ_FOREACH_SAFE(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs, nbp) { 406 if ((bp->b_vflags & BV_SCANNED) != 0) 407 continue; 408 bp->b_vflags |= BV_SCANNED; 409 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) 410 continue; 411 VI_UNLOCK(vp); 412 if ((bp->b_flags & B_DELWRI) == 0) 413 panic("fsync: not dirty"); 414 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) { 415 vfs_bio_awrite(bp); 416 splx(s); 417 } else { 418 bremfree(bp); 419 splx(s); 420 bawrite(bp); 421 } 422 VI_LOCK(vp); 423 goto loop2; 424 } 425 426 /* 427 * If synchronous the caller expects us to completely resolve all 428 * dirty buffers in the system. Wait for in-progress I/O to 429 * complete (which could include background bitmap writes), then 430 * retry if dirty blocks still exist. 431 */ 432 if (ap->a_waitfor == MNT_WAIT) { 433 bo = &vp->v_bufobj; 434 bufobj_wwait(bo, 0, 0); 435 if (bo->bo_dirty.bv_cnt > 0) { 436 /* 437 * If we are unable to write any of these buffers 438 * then we fail now rather than trying endlessly 439 * to write them out. 440 */ 441 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) 442 if ((error = bp->b_error) == 0) 443 continue; 444 if (error == 0 && --maxretry >= 0) { 445 splx(s); 446 goto loop1; 447 } 448 error = EAGAIN; 449 } 450 } 451 VI_UNLOCK(vp); 452 if (error == EAGAIN) 453 vprint("fsync: giving up on dirty", vp); 454 splx(s); 455 456 return (error); 457} 458 459/* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */ 460int 461vop_stdgetpages(ap) 462 struct vop_getpages_args /* { 463 struct vnode *a_vp; 464 vm_page_t *a_m; 465 int a_count; 466 int a_reqpage; 467 vm_ooffset_t a_offset; 468 } */ *ap; 469{ 470 471 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, 472 ap->a_count, ap->a_reqpage); 473} 474 475/* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */ 476int 477vop_stdputpages(ap) 478 struct vop_putpages_args /* { 479 struct vnode *a_vp; 480 vm_page_t *a_m; 481 int a_count; 482 int a_sync; 483 int *a_rtvals; 484 vm_ooffset_t a_offset; 485 } */ *ap; 486{ 487 488 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count, 489 ap->a_sync, ap->a_rtvals); 490} 491 492/* 493 * vfs default ops 494 * used to fill the vfs function table to get reasonable default return values. 495 */ 496int 497vfs_stdroot (mp, flags, vpp, td) 498 struct mount *mp; 499 int flags; 500 struct vnode **vpp; 501 struct thread *td; 502{ 503 504 return (EOPNOTSUPP); 505} 506 507int 508vfs_stdstatfs (mp, sbp, td) 509 struct mount *mp; 510 struct statfs *sbp; 511 struct thread *td; 512{ 513 514 return (EOPNOTSUPP); 515} 516 517int 518vfs_stdvptofh (vp, fhp) 519 struct vnode *vp; 520 struct fid *fhp; 521{ 522 523 return (EOPNOTSUPP); 524} 525 526int 527vfs_stdquotactl (mp, cmds, uid, arg, td) 528 struct mount *mp; 529 int cmds; 530 uid_t uid; 531 caddr_t arg; 532 struct thread *td; 533{ 534 535 return (EOPNOTSUPP); 536} 537 538int 539vfs_stdsync(mp, waitfor, td) 540 struct mount *mp; 541 int waitfor; 542 struct thread *td; 543{ 544 struct vnode *vp, *nvp; 545 int error, lockreq, allerror = 0; 546 547 lockreq = LK_EXCLUSIVE | LK_INTERLOCK; 548 if (waitfor != MNT_WAIT) 549 lockreq |= LK_NOWAIT; 550 /* 551 * Force stale buffer cache information to be flushed. 552 */ 553 MNT_ILOCK(mp); 554loop: 555 MNT_VNODE_FOREACH(vp, mp, nvp) { 556 557 VI_LOCK(vp); 558 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) { 559 VI_UNLOCK(vp); 560 continue; 561 } 562 MNT_IUNLOCK(mp); 563 564 if ((error = vget(vp, lockreq, td)) != 0) { 565 MNT_ILOCK(mp); 566 if (error == ENOENT) 567 goto loop; 568 continue; 569 } 570 error = VOP_FSYNC(vp, waitfor, td); 571 if (error) 572 allerror = error; 573 574 VOP_UNLOCK(vp, 0, td); 575 vrele(vp); 576 MNT_ILOCK(mp); 577 } 578 MNT_IUNLOCK(mp); 579 return (allerror); 580} 581 582int 583vfs_stdnosync (mp, waitfor, td) 584 struct mount *mp; 585 int waitfor; 586 struct thread *td; 587{ 588 589 return (0); 590} 591 592int 593vfs_stdvget (mp, ino, flags, vpp) 594 struct mount *mp; 595 ino_t ino; 596 int flags; 597 struct vnode **vpp; 598{ 599 600 return (EOPNOTSUPP); 601} 602 603int 604vfs_stdfhtovp (mp, fhp, vpp) 605 struct mount *mp; 606 struct fid *fhp; 607 struct vnode **vpp; 608{ 609 610 return (EOPNOTSUPP); 611} 612 613int 614vfs_stdinit (vfsp) 615 struct vfsconf *vfsp; 616{ 617 618 return (0); 619} 620 621int 622vfs_stduninit (vfsp) 623 struct vfsconf *vfsp; 624{ 625 626 return(0); 627} 628 629int 630vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td) 631 struct mount *mp; 632 int cmd; 633 struct vnode *filename_vp; 634 int attrnamespace; 635 const char *attrname; 636 struct thread *td; 637{ 638 639 if (filename_vp != NULL) 640 VOP_UNLOCK(filename_vp, 0, td); 641 return (EOPNOTSUPP); 642} 643 644int 645vfs_stdsysctl(mp, op, req) 646 struct mount *mp; 647 fsctlop_t op; 648 struct sysctl_req *req; 649{ 650 651 return (EOPNOTSUPP); 652} 653 654/* end of vfs default ops */ 655