vfs_default.c revision 166774
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 166774 2007-02-15 22:08:35Z pjd $"); 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/malloc.h> 48#include <sys/mount.h> 49#include <sys/mutex.h> 50#include <sys/unistd.h> 51#include <sys/vnode.h> 52#include <sys/poll.h> 53 54#include <vm/vm.h> 55#include <vm/vm_object.h> 56#include <vm/vm_extern.h> 57#include <vm/pmap.h> 58#include <vm/vm_map.h> 59#include <vm/vm_page.h> 60#include <vm/vm_pager.h> 61#include <vm/vnode_pager.h> 62 63static int vop_nolookup(struct vop_lookup_args *); 64static int vop_nostrategy(struct vop_strategy_args *); 65 66/* 67 * This vnode table stores what we want to do if the filesystem doesn't 68 * implement a particular VOP. 69 * 70 * If there is no specific entry here, we will return EOPNOTSUPP. 71 * 72 */ 73 74struct vop_vector default_vnodeops = { 75 .vop_default = NULL, 76 .vop_bypass = VOP_EOPNOTSUPP, 77 78 .vop_advlock = VOP_EINVAL, 79 .vop_bmap = vop_stdbmap, 80 .vop_close = VOP_NULL, 81 .vop_fsync = VOP_NULL, 82 .vop_getpages = vop_stdgetpages, 83 .vop_getwritemount = vop_stdgetwritemount, 84 .vop_inactive = VOP_NULL, 85 .vop_ioctl = VOP_ENOTTY, 86 .vop_kqfilter = vop_stdkqfilter, 87 .vop_islocked = vop_stdislocked, 88 .vop_lease = VOP_NULL, 89 ._vop_lock = vop_stdlock, 90 .vop_lookup = vop_nolookup, 91 .vop_open = VOP_NULL, 92 .vop_pathconf = VOP_EINVAL, 93 .vop_poll = vop_nopoll, 94 .vop_putpages = vop_stdputpages, 95 .vop_readlink = VOP_EINVAL, 96 .vop_revoke = VOP_PANIC, 97 .vop_strategy = vop_nostrategy, 98 .vop_unlock = vop_stdunlock, 99 .vop_vptofh = vop_stdvptofh, 100}; 101 102/* 103 * Series of placeholder functions for various error returns for 104 * VOPs. 105 */ 106 107int 108vop_eopnotsupp(struct vop_generic_args *ap) 109{ 110 /* 111 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 112 */ 113 114 return (EOPNOTSUPP); 115} 116 117int 118vop_ebadf(struct vop_generic_args *ap) 119{ 120 121 return (EBADF); 122} 123 124int 125vop_enotty(struct vop_generic_args *ap) 126{ 127 128 return (ENOTTY); 129} 130 131int 132vop_einval(struct vop_generic_args *ap) 133{ 134 135 return (EINVAL); 136} 137 138int 139vop_null(struct vop_generic_args *ap) 140{ 141 142 return (0); 143} 144 145/* 146 * Helper function to panic on some bad VOPs in some filesystems. 147 */ 148int 149vop_panic(struct vop_generic_args *ap) 150{ 151 152 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 153} 154 155/* 156 * vop_std<something> and vop_no<something> are default functions for use by 157 * filesystems that need the "default reasonable" implementation for a 158 * particular operation. 159 * 160 * The documentation for the operations they implement exists (if it exists) 161 * in the VOP_<SOMETHING>(9) manpage (all uppercase). 162 */ 163 164/* 165 * Default vop for filesystems that do not support name lookup 166 */ 167static int 168vop_nolookup(ap) 169 struct vop_lookup_args /* { 170 struct vnode *a_dvp; 171 struct vnode **a_vpp; 172 struct componentname *a_cnp; 173 } */ *ap; 174{ 175 176 *ap->a_vpp = NULL; 177 return (ENOTDIR); 178} 179 180/* 181 * vop_nostrategy: 182 * 183 * Strategy routine for VFS devices that have none. 184 * 185 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 186 * routine. Typically this is done for a BIO_READ strategy call. 187 * Typically B_INVAL is assumed to already be clear prior to a write 188 * and should not be cleared manually unless you just made the buffer 189 * invalid. BIO_ERROR should be cleared either way. 190 */ 191 192static int 193vop_nostrategy (struct vop_strategy_args *ap) 194{ 195 printf("No strategy for buffer at %p\n", ap->a_bp); 196 vprint("vnode", ap->a_vp); 197 ap->a_bp->b_ioflags |= BIO_ERROR; 198 ap->a_bp->b_error = EOPNOTSUPP; 199 bufdone(ap->a_bp); 200 return (EOPNOTSUPP); 201} 202 203/* 204 * vop_stdpathconf: 205 * 206 * Standard implementation of POSIX pathconf, to get information about limits 207 * for a filesystem. 208 * Override per filesystem for the case where the filesystem has smaller 209 * limits. 210 */ 211int 212vop_stdpathconf(ap) 213 struct vop_pathconf_args /* { 214 struct vnode *a_vp; 215 int a_name; 216 int *a_retval; 217 } */ *ap; 218{ 219 220 switch (ap->a_name) { 221 case _PC_NAME_MAX: 222 *ap->a_retval = NAME_MAX; 223 return (0); 224 case _PC_PATH_MAX: 225 *ap->a_retval = PATH_MAX; 226 return (0); 227 case _PC_LINK_MAX: 228 *ap->a_retval = LINK_MAX; 229 return (0); 230 case _PC_MAX_CANON: 231 *ap->a_retval = MAX_CANON; 232 return (0); 233 case _PC_MAX_INPUT: 234 *ap->a_retval = MAX_INPUT; 235 return (0); 236 case _PC_PIPE_BUF: 237 *ap->a_retval = PIPE_BUF; 238 return (0); 239 case _PC_CHOWN_RESTRICTED: 240 *ap->a_retval = 1; 241 return (0); 242 case _PC_VDISABLE: 243 *ap->a_retval = _POSIX_VDISABLE; 244 return (0); 245 default: 246 return (EINVAL); 247 } 248 /* NOTREACHED */ 249} 250 251/* 252 * Standard lock, unlock and islocked functions. 253 */ 254int 255vop_stdlock(ap) 256 struct _vop_lock_args /* { 257 struct vnode *a_vp; 258 int a_flags; 259 struct thread *a_td; 260 char *file; 261 int line; 262 } */ *ap; 263{ 264 struct vnode *vp = ap->a_vp; 265 266 return (_lockmgr(vp->v_vnlock, ap->a_flags, VI_MTX(vp), ap->a_td, ap->a_file, ap->a_line)); 267} 268 269/* See above. */ 270int 271vop_stdunlock(ap) 272 struct vop_unlock_args /* { 273 struct vnode *a_vp; 274 int a_flags; 275 struct thread *a_td; 276 } */ *ap; 277{ 278 struct vnode *vp = ap->a_vp; 279 280 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp), 281 ap->a_td)); 282} 283 284/* See above. */ 285int 286vop_stdislocked(ap) 287 struct vop_islocked_args /* { 288 struct vnode *a_vp; 289 struct thread *a_td; 290 } */ *ap; 291{ 292 293 return (lockstatus(ap->a_vp->v_vnlock, ap->a_td)); 294} 295 296/* 297 * Return true for select/poll. 298 */ 299int 300vop_nopoll(ap) 301 struct vop_poll_args /* { 302 struct vnode *a_vp; 303 int a_events; 304 struct ucred *a_cred; 305 struct thread *a_td; 306 } */ *ap; 307{ 308 /* 309 * Return true for read/write. If the user asked for something 310 * special, return POLLNVAL, so that clients have a way of 311 * determining reliably whether or not the extended 312 * functionality is present without hard-coding knowledge 313 * of specific filesystem implementations. 314 * Stay in sync with kern_conf.c::no_poll(). 315 */ 316 if (ap->a_events & ~POLLSTANDARD) 317 return (POLLNVAL); 318 319 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 320} 321 322/* 323 * Implement poll for local filesystems that support it. 324 */ 325int 326vop_stdpoll(ap) 327 struct vop_poll_args /* { 328 struct vnode *a_vp; 329 int a_events; 330 struct ucred *a_cred; 331 struct thread *a_td; 332 } */ *ap; 333{ 334 if (ap->a_events & ~POLLSTANDARD) 335 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 336 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 337} 338 339/* 340 * Return our mount point, as we will take charge of the writes. 341 */ 342int 343vop_stdgetwritemount(ap) 344 struct vop_getwritemount_args /* { 345 struct vnode *a_vp; 346 struct mount **a_mpp; 347 } */ *ap; 348{ 349 struct mount *mp; 350 351 /* 352 * XXX Since this is called unlocked we may be recycled while 353 * attempting to ref the mount. If this is the case or mountpoint 354 * will be set to NULL. We only have to prevent this call from 355 * returning with a ref to an incorrect mountpoint. It is not 356 * harmful to return with a ref to our previous mountpoint. 357 */ 358 mp = ap->a_vp->v_mount; 359 if (mp != NULL) { 360 vfs_ref(mp); 361 if (mp != ap->a_vp->v_mount) { 362 vfs_rel(mp); 363 mp = NULL; 364 } 365 } 366 *(ap->a_mpp) = mp; 367 return (0); 368} 369 370/* XXX Needs good comment and VOP_BMAP(9) manpage */ 371int 372vop_stdbmap(ap) 373 struct vop_bmap_args /* { 374 struct vnode *a_vp; 375 daddr_t a_bn; 376 struct bufobj **a_bop; 377 daddr_t *a_bnp; 378 int *a_runp; 379 int *a_runb; 380 } */ *ap; 381{ 382 383 if (ap->a_bop != NULL) 384 *ap->a_bop = &ap->a_vp->v_bufobj; 385 if (ap->a_bnp != NULL) 386 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 387 if (ap->a_runp != NULL) 388 *ap->a_runp = 0; 389 if (ap->a_runb != NULL) 390 *ap->a_runb = 0; 391 return (0); 392} 393 394int 395vop_stdfsync(ap) 396 struct vop_fsync_args /* { 397 struct vnode *a_vp; 398 struct ucred *a_cred; 399 int a_waitfor; 400 struct thread *a_td; 401 } */ *ap; 402{ 403 struct vnode *vp = ap->a_vp; 404 struct buf *bp; 405 struct bufobj *bo; 406 struct buf *nbp; 407 int error = 0; 408 int maxretry = 1000; /* large, arbitrarily chosen */ 409 410 VI_LOCK(vp); 411loop1: 412 /* 413 * MARK/SCAN initialization to avoid infinite loops. 414 */ 415 TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs) { 416 bp->b_vflags &= ~BV_SCANNED; 417 bp->b_error = 0; 418 } 419 420 /* 421 * Flush all dirty buffers associated with a vnode. 422 */ 423loop2: 424 TAILQ_FOREACH_SAFE(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs, nbp) { 425 if ((bp->b_vflags & BV_SCANNED) != 0) 426 continue; 427 bp->b_vflags |= BV_SCANNED; 428 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) 429 continue; 430 VI_UNLOCK(vp); 431 KASSERT(bp->b_bufobj == &vp->v_bufobj, 432 ("bp %p wrong b_bufobj %p should be %p", 433 bp, bp->b_bufobj, &vp->v_bufobj)); 434 if ((bp->b_flags & B_DELWRI) == 0) 435 panic("fsync: not dirty"); 436 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) { 437 vfs_bio_awrite(bp); 438 } else { 439 bremfree(bp); 440 bawrite(bp); 441 } 442 VI_LOCK(vp); 443 goto loop2; 444 } 445 446 /* 447 * If synchronous the caller expects us to completely resolve all 448 * dirty buffers in the system. Wait for in-progress I/O to 449 * complete (which could include background bitmap writes), then 450 * retry if dirty blocks still exist. 451 */ 452 if (ap->a_waitfor == MNT_WAIT) { 453 bo = &vp->v_bufobj; 454 bufobj_wwait(bo, 0, 0); 455 if (bo->bo_dirty.bv_cnt > 0) { 456 /* 457 * If we are unable to write any of these buffers 458 * then we fail now rather than trying endlessly 459 * to write them out. 460 */ 461 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) 462 if ((error = bp->b_error) == 0) 463 continue; 464 if (error == 0 && --maxretry >= 0) 465 goto loop1; 466 error = EAGAIN; 467 } 468 } 469 VI_UNLOCK(vp); 470 if (error == EAGAIN) 471 vprint("fsync: giving up on dirty", vp); 472 473 return (error); 474} 475 476/* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */ 477int 478vop_stdgetpages(ap) 479 struct vop_getpages_args /* { 480 struct vnode *a_vp; 481 vm_page_t *a_m; 482 int a_count; 483 int a_reqpage; 484 vm_ooffset_t a_offset; 485 } */ *ap; 486{ 487 488 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, 489 ap->a_count, ap->a_reqpage); 490} 491 492int 493vop_stdkqfilter(struct vop_kqfilter_args *ap) 494{ 495 return vfs_kqfilter(ap); 496} 497 498/* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */ 499int 500vop_stdputpages(ap) 501 struct vop_putpages_args /* { 502 struct vnode *a_vp; 503 vm_page_t *a_m; 504 int a_count; 505 int a_sync; 506 int *a_rtvals; 507 vm_ooffset_t a_offset; 508 } */ *ap; 509{ 510 511 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count, 512 ap->a_sync, ap->a_rtvals); 513} 514 515int 516vop_stdvptofh(struct vop_vptofh_args *ap) 517{ 518 return VFS_VPTOFH(ap->a_vp, ap->a_fhp); 519} 520 521/* 522 * vfs default ops 523 * used to fill the vfs function table to get reasonable default return values. 524 */ 525int 526vfs_stdroot (mp, flags, vpp, td) 527 struct mount *mp; 528 int flags; 529 struct vnode **vpp; 530 struct thread *td; 531{ 532 533 return (EOPNOTSUPP); 534} 535 536int 537vfs_stdstatfs (mp, sbp, td) 538 struct mount *mp; 539 struct statfs *sbp; 540 struct thread *td; 541{ 542 543 return (EOPNOTSUPP); 544} 545 546#if __FreeBSD_version < 800000 547int 548vfs_stdvptofh (vp, fhp) 549 struct vnode *vp; 550 struct fid *fhp; 551{ 552 553 return (EOPNOTSUPP); 554} 555#else 556#error Remove this code, vfs_vptofh was replaced with vop_vptofh. 557#endif 558 559int 560vfs_stdquotactl (mp, cmds, uid, arg, td) 561 struct mount *mp; 562 int cmds; 563 uid_t uid; 564 void *arg; 565 struct thread *td; 566{ 567 568 return (EOPNOTSUPP); 569} 570 571int 572vfs_stdsync(mp, waitfor, td) 573 struct mount *mp; 574 int waitfor; 575 struct thread *td; 576{ 577 struct vnode *vp, *mvp; 578 int error, lockreq, allerror = 0; 579 580 lockreq = LK_EXCLUSIVE | LK_INTERLOCK; 581 if (waitfor != MNT_WAIT) 582 lockreq |= LK_NOWAIT; 583 /* 584 * Force stale buffer cache information to be flushed. 585 */ 586 MNT_ILOCK(mp); 587loop: 588 MNT_VNODE_FOREACH(vp, mp, mvp) { 589 590 VI_LOCK(vp); 591 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) { 592 VI_UNLOCK(vp); 593 continue; 594 } 595 MNT_IUNLOCK(mp); 596 597 if ((error = vget(vp, lockreq, td)) != 0) { 598 MNT_ILOCK(mp); 599 if (error == ENOENT) { 600 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp); 601 goto loop; 602 } 603 continue; 604 } 605 error = VOP_FSYNC(vp, waitfor, td); 606 if (error) 607 allerror = error; 608 609 /* Do not turn this into vput. td is not always curthread. */ 610 VOP_UNLOCK(vp, 0, td); 611 vrele(vp); 612 MNT_ILOCK(mp); 613 } 614 MNT_IUNLOCK(mp); 615 return (allerror); 616} 617 618int 619vfs_stdnosync (mp, waitfor, td) 620 struct mount *mp; 621 int waitfor; 622 struct thread *td; 623{ 624 625 return (0); 626} 627 628int 629vfs_stdvget (mp, ino, flags, vpp) 630 struct mount *mp; 631 ino_t ino; 632 int flags; 633 struct vnode **vpp; 634{ 635 636 return (EOPNOTSUPP); 637} 638 639int 640vfs_stdfhtovp (mp, fhp, vpp) 641 struct mount *mp; 642 struct fid *fhp; 643 struct vnode **vpp; 644{ 645 646 return (EOPNOTSUPP); 647} 648 649int 650vfs_stdinit (vfsp) 651 struct vfsconf *vfsp; 652{ 653 654 return (0); 655} 656 657int 658vfs_stduninit (vfsp) 659 struct vfsconf *vfsp; 660{ 661 662 return(0); 663} 664 665int 666vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td) 667 struct mount *mp; 668 int cmd; 669 struct vnode *filename_vp; 670 int attrnamespace; 671 const char *attrname; 672 struct thread *td; 673{ 674 675 if (filename_vp != NULL) 676 VOP_UNLOCK(filename_vp, 0, td); 677 return (EOPNOTSUPP); 678} 679 680int 681vfs_stdsysctl(mp, op, req) 682 struct mount *mp; 683 fsctlop_t op; 684 struct sysctl_req *req; 685{ 686 687 return (EOPNOTSUPP); 688} 689 690/* end of vfs default ops */ 691