vfs_default.c revision 142153
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 142153 2005-02-20 23:02:20Z das $"); 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_stdinactive, 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/* Mark the vnode inactive */ 291int 292vop_stdinactive(ap) 293 struct vop_inactive_args /* { 294 struct vnode *a_vp; 295 struct thread *a_td; 296 } */ *ap; 297{ 298 299 VOP_UNLOCK(ap->a_vp, 0, ap->a_td); 300 return (0); 301} 302 303/* 304 * Return true for select/poll. 305 */ 306int 307vop_nopoll(ap) 308 struct vop_poll_args /* { 309 struct vnode *a_vp; 310 int a_events; 311 struct ucred *a_cred; 312 struct thread *a_td; 313 } */ *ap; 314{ 315 /* 316 * Return true for read/write. If the user asked for something 317 * special, return POLLNVAL, so that clients have a way of 318 * determining reliably whether or not the extended 319 * functionality is present without hard-coding knowledge 320 * of specific filesystem implementations. 321 * Stay in sync with kern_conf.c::no_poll(). 322 */ 323 if (ap->a_events & ~POLLSTANDARD) 324 return (POLLNVAL); 325 326 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 327} 328 329/* 330 * Implement poll for local filesystems that support it. 331 */ 332int 333vop_stdpoll(ap) 334 struct vop_poll_args /* { 335 struct vnode *a_vp; 336 int a_events; 337 struct ucred *a_cred; 338 struct thread *a_td; 339 } */ *ap; 340{ 341 if (ap->a_events & ~POLLSTANDARD) 342 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 343 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 344} 345 346/* 347 * Return our mount point, as we will take charge of the writes. 348 */ 349int 350vop_stdgetwritemount(ap) 351 struct vop_getwritemount_args /* { 352 struct vnode *a_vp; 353 struct mount **a_mpp; 354 } */ *ap; 355{ 356 357 *(ap->a_mpp) = ap->a_vp->v_mount; 358 return (0); 359} 360 361/* XXX Needs good comment and VOP_BMAP(9) manpage */ 362int 363vop_stdbmap(ap) 364 struct vop_bmap_args /* { 365 struct vnode *a_vp; 366 daddr_t a_bn; 367 struct bufobj **a_bop; 368 daddr_t *a_bnp; 369 int *a_runp; 370 int *a_runb; 371 } */ *ap; 372{ 373 374 if (ap->a_bop != NULL) 375 *ap->a_bop = &ap->a_vp->v_bufobj; 376 if (ap->a_bnp != NULL) 377 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 378 if (ap->a_runp != NULL) 379 *ap->a_runp = 0; 380 if (ap->a_runb != NULL) 381 *ap->a_runb = 0; 382 return (0); 383} 384 385int 386vop_stdfsync(ap) 387 struct vop_fsync_args /* { 388 struct vnode *a_vp; 389 struct ucred *a_cred; 390 int a_waitfor; 391 struct thread *a_td; 392 } */ *ap; 393{ 394 struct vnode *vp = ap->a_vp; 395 struct buf *bp; 396 struct bufobj *bo; 397 struct buf *nbp; 398 int s, error = 0; 399 int maxretry = 100; /* large, arbitrarily chosen */ 400 401 VI_LOCK(vp); 402loop1: 403 /* 404 * MARK/SCAN initialization to avoid infinite loops. 405 */ 406 s = splbio(); 407 TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs) { 408 bp->b_vflags &= ~BV_SCANNED; 409 bp->b_error = 0; 410 } 411 splx(s); 412 413 /* 414 * Flush all dirty buffers associated with a block device. 415 */ 416loop2: 417 s = splbio(); 418 TAILQ_FOREACH_SAFE(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs, nbp) { 419 if ((bp->b_vflags & BV_SCANNED) != 0) 420 continue; 421 bp->b_vflags |= BV_SCANNED; 422 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) 423 continue; 424 VI_UNLOCK(vp); 425 if ((bp->b_flags & B_DELWRI) == 0) 426 panic("fsync: not dirty"); 427 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) { 428 vfs_bio_awrite(bp); 429 splx(s); 430 } else { 431 bremfree(bp); 432 splx(s); 433 bawrite(bp); 434 } 435 VI_LOCK(vp); 436 goto loop2; 437 } 438 439 /* 440 * If synchronous the caller expects us to completely resolve all 441 * dirty buffers in the system. Wait for in-progress I/O to 442 * complete (which could include background bitmap writes), then 443 * retry if dirty blocks still exist. 444 */ 445 if (ap->a_waitfor == MNT_WAIT) { 446 bo = &vp->v_bufobj; 447 bufobj_wwait(bo, 0, 0); 448 if (bo->bo_dirty.bv_cnt > 0) { 449 /* 450 * If we are unable to write any of these buffers 451 * then we fail now rather than trying endlessly 452 * to write them out. 453 */ 454 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) 455 if ((error = bp->b_error) == 0) 456 continue; 457 if (error == 0 && --maxretry >= 0) { 458 splx(s); 459 goto loop1; 460 } 461 vprint("fsync: giving up on dirty", vp); 462 error = EAGAIN; 463 } 464 } 465 VI_UNLOCK(vp); 466 splx(s); 467 468 return (error); 469} 470 471/* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */ 472int 473vop_stdgetpages(ap) 474 struct vop_getpages_args /* { 475 struct vnode *a_vp; 476 vm_page_t *a_m; 477 int a_count; 478 int a_reqpage; 479 vm_ooffset_t a_offset; 480 } */ *ap; 481{ 482 483 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, 484 ap->a_count, ap->a_reqpage); 485} 486 487/* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */ 488int 489vop_stdputpages(ap) 490 struct vop_putpages_args /* { 491 struct vnode *a_vp; 492 vm_page_t *a_m; 493 int a_count; 494 int a_sync; 495 int *a_rtvals; 496 vm_ooffset_t a_offset; 497 } */ *ap; 498{ 499 500 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count, 501 ap->a_sync, ap->a_rtvals); 502} 503 504/* 505 * vfs default ops 506 * used to fill the vfs function table to get reasonable default return values. 507 */ 508int 509vfs_stdroot (mp, vpp, td) 510 struct mount *mp; 511 struct vnode **vpp; 512 struct thread *td; 513{ 514 515 return (EOPNOTSUPP); 516} 517 518int 519vfs_stdstatfs (mp, sbp, td) 520 struct mount *mp; 521 struct statfs *sbp; 522 struct thread *td; 523{ 524 525 return (EOPNOTSUPP); 526} 527 528int 529vfs_stdvptofh (vp, fhp) 530 struct vnode *vp; 531 struct fid *fhp; 532{ 533 534 return (EOPNOTSUPP); 535} 536 537int 538vfs_stdquotactl (mp, cmds, uid, arg, td) 539 struct mount *mp; 540 int cmds; 541 uid_t uid; 542 caddr_t arg; 543 struct thread *td; 544{ 545 546 return (EOPNOTSUPP); 547} 548 549int 550vfs_stdsync(mp, waitfor, td) 551 struct mount *mp; 552 int waitfor; 553 struct thread *td; 554{ 555 struct vnode *vp, *nvp; 556 int error, lockreq, allerror = 0; 557 558 lockreq = LK_EXCLUSIVE | LK_INTERLOCK; 559 if (waitfor != MNT_WAIT) 560 lockreq |= LK_NOWAIT; 561 /* 562 * Force stale buffer cache information to be flushed. 563 */ 564 MNT_ILOCK(mp); 565loop: 566 MNT_VNODE_FOREACH(vp, mp, nvp) { 567 568 VI_LOCK(vp); 569 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) { 570 VI_UNLOCK(vp); 571 continue; 572 } 573 MNT_IUNLOCK(mp); 574 575 if ((error = vget(vp, lockreq, td)) != 0) { 576 MNT_ILOCK(mp); 577 if (error == ENOENT) 578 goto loop; 579 continue; 580 } 581 error = VOP_FSYNC(vp, waitfor, td); 582 if (error) 583 allerror = error; 584 585 VOP_UNLOCK(vp, 0, td); 586 vrele(vp); 587 MNT_ILOCK(mp); 588 } 589 MNT_IUNLOCK(mp); 590 return (allerror); 591} 592 593int 594vfs_stdnosync (mp, waitfor, td) 595 struct mount *mp; 596 int waitfor; 597 struct thread *td; 598{ 599 600 return (0); 601} 602 603int 604vfs_stdvget (mp, ino, flags, vpp) 605 struct mount *mp; 606 ino_t ino; 607 int flags; 608 struct vnode **vpp; 609{ 610 611 return (EOPNOTSUPP); 612} 613 614int 615vfs_stdfhtovp (mp, fhp, vpp) 616 struct mount *mp; 617 struct fid *fhp; 618 struct vnode **vpp; 619{ 620 621 return (EOPNOTSUPP); 622} 623 624int 625vfs_stdinit (vfsp) 626 struct vfsconf *vfsp; 627{ 628 629 return (0); 630} 631 632int 633vfs_stduninit (vfsp) 634 struct vfsconf *vfsp; 635{ 636 637 return(0); 638} 639 640int 641vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td) 642 struct mount *mp; 643 int cmd; 644 struct vnode *filename_vp; 645 int attrnamespace; 646 const char *attrname; 647 struct thread *td; 648{ 649 650 if (filename_vp != NULL) 651 VOP_UNLOCK(filename_vp, 0, td); 652 return (EOPNOTSUPP); 653} 654 655int 656vfs_stdsysctl(mp, op, req) 657 struct mount *mp; 658 fsctlop_t op; 659 struct sysctl_req *req; 660{ 661 662 return (EOPNOTSUPP); 663} 664 665/* end of vfs default ops */ 666