vfs_default.c revision 76117
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 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * 39 * $FreeBSD: head/sys/kern/vfs_default.c 76117 2001-04-29 02:45:39Z grog $ 40 */ 41 42#include <sys/param.h> 43#include <sys/systm.h> 44#include <sys/bio.h> 45#include <sys/buf.h> 46#include <sys/conf.h> 47#include <sys/kernel.h> 48#include <sys/lock.h> 49#include <sys/malloc.h> 50#include <sys/mount.h> 51#include <sys/mutex.h> 52#include <sys/unistd.h> 53#include <sys/vnode.h> 54#include <sys/poll.h> 55 56#include <machine/limits.h> 57 58#include <vm/vm.h> 59#include <vm/vm_object.h> 60#include <vm/vm_extern.h> 61#include <vm/pmap.h> 62#include <vm/vm_map.h> 63#include <vm/vm_page.h> 64#include <vm/vm_pager.h> 65#include <vm/vnode_pager.h> 66#include <vm/vm_zone.h> 67 68static int vop_nolookup __P((struct vop_lookup_args *)); 69static int vop_nostrategy __P((struct vop_strategy_args *)); 70 71/* 72 * This vnode table stores what we want to do if the filesystem doesn't 73 * implement a particular VOP. 74 * 75 * If there is no specific entry here, we will return EOPNOTSUPP. 76 * 77 */ 78 79vop_t **default_vnodeop_p; 80static struct vnodeopv_entry_desc default_vnodeop_entries[] = { 81 { &vop_default_desc, (vop_t *) vop_eopnotsupp }, 82 { &vop_advlock_desc, (vop_t *) vop_einval }, 83 { &vop_close_desc, (vop_t *) vop_null }, 84 { &vop_createvobject_desc, (vop_t *) vop_stdcreatevobject }, 85 { &vop_destroyvobject_desc, (vop_t *) vop_stddestroyvobject }, 86 { &vop_fsync_desc, (vop_t *) vop_null }, 87 { &vop_getvobject_desc, (vop_t *) vop_stdgetvobject }, 88 { &vop_inactive_desc, (vop_t *) vop_stdinactive }, 89 { &vop_ioctl_desc, (vop_t *) vop_enotty }, 90 { &vop_islocked_desc, (vop_t *) vop_noislocked }, 91 { &vop_lease_desc, (vop_t *) vop_null }, 92 { &vop_lock_desc, (vop_t *) vop_nolock }, 93 { &vop_lookup_desc, (vop_t *) vop_nolookup }, 94 { &vop_open_desc, (vop_t *) vop_null }, 95 { &vop_pathconf_desc, (vop_t *) vop_einval }, 96 { &vop_poll_desc, (vop_t *) vop_nopoll }, 97 { &vop_readlink_desc, (vop_t *) vop_einval }, 98 { &vop_revoke_desc, (vop_t *) vop_revoke }, 99 { &vop_strategy_desc, (vop_t *) vop_nostrategy }, 100 { &vop_unlock_desc, (vop_t *) vop_nounlock }, 101 { NULL, NULL } 102}; 103 104static struct vnodeopv_desc default_vnodeop_opv_desc = 105 { &default_vnodeop_p, default_vnodeop_entries }; 106 107VNODEOP_SET(default_vnodeop_opv_desc); 108 109int 110vop_eopnotsupp(struct vop_generic_args *ap) 111{ 112 /* 113 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 114 */ 115 116 return (EOPNOTSUPP); 117} 118 119int 120vop_ebadf(struct vop_generic_args *ap) 121{ 122 123 return (EBADF); 124} 125 126int 127vop_enotty(struct vop_generic_args *ap) 128{ 129 130 return (ENOTTY); 131} 132 133int 134vop_einval(struct vop_generic_args *ap) 135{ 136 137 return (EINVAL); 138} 139 140int 141vop_null(struct vop_generic_args *ap) 142{ 143 144 return (0); 145} 146 147int 148vop_defaultop(struct vop_generic_args *ap) 149{ 150 151 return (VOCALL(default_vnodeop_p, ap->a_desc->vdesc_offset, ap)); 152} 153 154int 155vop_panic(struct vop_generic_args *ap) 156{ 157 158 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 159} 160 161static int 162vop_nolookup(ap) 163 struct vop_lookup_args /* { 164 struct vnode *a_dvp; 165 struct vnode **a_vpp; 166 struct componentname *a_cnp; 167 } */ *ap; 168{ 169 170 *ap->a_vpp = NULL; 171 return (ENOTDIR); 172} 173 174/* 175 * vop_nostrategy: 176 * 177 * Strategy routine for VFS devices that have none. 178 * 179 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 180 * routine. Typically this is done for a BIO_READ strategy call. 181 * Typically B_INVAL is assumed to already be clear prior to a write 182 * and should not be cleared manually unless you just made the buffer 183 * invalid. BIO_ERROR should be cleared either way. 184 */ 185 186static int 187vop_nostrategy (struct vop_strategy_args *ap) 188{ 189 printf("No strategy for buffer at %p\n", ap->a_bp); 190 vprint("", ap->a_vp); 191 vprint("", ap->a_bp->b_vp); 192 ap->a_bp->b_ioflags |= BIO_ERROR; 193 ap->a_bp->b_error = EOPNOTSUPP; 194 bufdone(ap->a_bp); 195 return (EOPNOTSUPP); 196} 197 198int 199vop_stdpathconf(ap) 200 struct vop_pathconf_args /* { 201 struct vnode *a_vp; 202 int a_name; 203 int *a_retval; 204 } */ *ap; 205{ 206 207 switch (ap->a_name) { 208 case _PC_LINK_MAX: 209 *ap->a_retval = LINK_MAX; 210 return (0); 211 case _PC_MAX_CANON: 212 *ap->a_retval = MAX_CANON; 213 return (0); 214 case _PC_MAX_INPUT: 215 *ap->a_retval = MAX_INPUT; 216 return (0); 217 case _PC_PIPE_BUF: 218 *ap->a_retval = PIPE_BUF; 219 return (0); 220 case _PC_CHOWN_RESTRICTED: 221 *ap->a_retval = 1; 222 return (0); 223 case _PC_VDISABLE: 224 *ap->a_retval = _POSIX_VDISABLE; 225 return (0); 226 default: 227 return (EINVAL); 228 } 229 /* NOTREACHED */ 230} 231 232/* 233 * Standard lock, unlock and islocked functions. 234 * 235 * These depend on the lock structure being the first element in the 236 * inode, ie: vp->v_data points to the the lock! 237 */ 238int 239vop_stdlock(ap) 240 struct vop_lock_args /* { 241 struct vnode *a_vp; 242 int a_flags; 243 struct proc *a_p; 244 } */ *ap; 245{ 246 struct vnode *vp = ap->a_vp; 247 248#ifndef DEBUG_LOCKS 249 return (lockmgr(&vp->v_lock, ap->a_flags, &vp->v_interlock, ap->a_p)); 250#else 251 return (debuglockmgr(&vp->v_lock, ap->a_flags, &vp->v_interlock, 252 ap->a_p, "vop_stdlock", vp->filename, vp->line)); 253#endif 254} 255 256int 257vop_stdunlock(ap) 258 struct vop_unlock_args /* { 259 struct vnode *a_vp; 260 int a_flags; 261 struct proc *a_p; 262 } */ *ap; 263{ 264 struct vnode *vp = ap->a_vp; 265 266 return (lockmgr(&vp->v_lock, ap->a_flags | LK_RELEASE, &vp->v_interlock, 267 ap->a_p)); 268} 269 270int 271vop_stdislocked(ap) 272 struct vop_islocked_args /* { 273 struct vnode *a_vp; 274 struct proc *a_p; 275 } */ *ap; 276{ 277 278 return (lockstatus(&ap->a_vp->v_lock, ap->a_p)); 279} 280 281int 282vop_stdinactive(ap) 283 struct vop_inactive_args /* { 284 struct vnode *a_vp; 285 struct proc *a_p; 286 } */ *ap; 287{ 288 289 VOP_UNLOCK(ap->a_vp, 0, ap->a_p); 290 return (0); 291} 292 293/* 294 * Return true for select/poll. 295 */ 296int 297vop_nopoll(ap) 298 struct vop_poll_args /* { 299 struct vnode *a_vp; 300 int a_events; 301 struct ucred *a_cred; 302 struct proc *a_p; 303 } */ *ap; 304{ 305 /* 306 * Return true for read/write. If the user asked for something 307 * special, return POLLNVAL, so that clients have a way of 308 * determining reliably whether or not the extended 309 * functionality is present without hard-coding knowledge 310 * of specific filesystem implementations. 311 */ 312 if (ap->a_events & ~POLLSTANDARD) 313 return (POLLNVAL); 314 315 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 316} 317 318/* 319 * Implement poll for local filesystems that support it. 320 */ 321int 322vop_stdpoll(ap) 323 struct vop_poll_args /* { 324 struct vnode *a_vp; 325 int a_events; 326 struct ucred *a_cred; 327 struct proc *a_p; 328 } */ *ap; 329{ 330 if ((ap->a_events & ~POLLSTANDARD) == 0) 331 return (ap->a_events & (POLLRDNORM|POLLWRNORM)); 332 return (vn_pollrecord(ap->a_vp, ap->a_p, ap->a_events)); 333} 334 335/* 336 * Stubs to use when there is no locking to be done on the underlying object. 337 * A minimal shared lock is necessary to ensure that the underlying object 338 * is not revoked while an operation is in progress. So, an active shared 339 * count is maintained in an auxillary vnode lock structure. 340 */ 341int 342vop_sharedlock(ap) 343 struct vop_lock_args /* { 344 struct vnode *a_vp; 345 int a_flags; 346 struct proc *a_p; 347 } */ *ap; 348{ 349 /* 350 * This code cannot be used until all the non-locking filesystems 351 * (notably NFS) are converted to properly lock and release nodes. 352 * Also, certain vnode operations change the locking state within 353 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 354 * and symlink). Ideally these operations should not change the 355 * lock state, but should be changed to let the caller of the 356 * function unlock them. Otherwise all intermediate vnode layers 357 * (such as union, umapfs, etc) must catch these functions to do 358 * the necessary locking at their layer. Note that the inactive 359 * and lookup operations also change their lock state, but this 360 * cannot be avoided, so these two operations will always need 361 * to be handled in intermediate layers. 362 */ 363 struct vnode *vp = ap->a_vp; 364 int vnflags, flags = ap->a_flags; 365 366 switch (flags & LK_TYPE_MASK) { 367 case LK_DRAIN: 368 vnflags = LK_DRAIN; 369 break; 370 case LK_EXCLUSIVE: 371#ifdef DEBUG_VFS_LOCKS 372 /* 373 * Normally, we use shared locks here, but that confuses 374 * the locking assertions. 375 */ 376 vnflags = LK_EXCLUSIVE; 377 break; 378#endif 379 case LK_SHARED: 380 vnflags = LK_SHARED; 381 break; 382 case LK_UPGRADE: 383 case LK_EXCLUPGRADE: 384 case LK_DOWNGRADE: 385 return (0); 386 case LK_RELEASE: 387 default: 388 panic("vop_sharedlock: bad operation %d", flags & LK_TYPE_MASK); 389 } 390 if (flags & LK_INTERLOCK) 391 vnflags |= LK_INTERLOCK; 392#ifndef DEBUG_LOCKS 393 return (lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p)); 394#else 395 return (debuglockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p, 396 "vop_sharedlock", vp->filename, vp->line)); 397#endif 398} 399 400/* 401 * Stubs to use when there is no locking to be done on the underlying object. 402 * A minimal shared lock is necessary to ensure that the underlying object 403 * is not revoked while an operation is in progress. So, an active shared 404 * count is maintained in an auxillary vnode lock structure. 405 */ 406int 407vop_nolock(ap) 408 struct vop_lock_args /* { 409 struct vnode *a_vp; 410 int a_flags; 411 struct proc *a_p; 412 } */ *ap; 413{ 414#ifdef notyet 415 /* 416 * This code cannot be used until all the non-locking filesystems 417 * (notably NFS) are converted to properly lock and release nodes. 418 * Also, certain vnode operations change the locking state within 419 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 420 * and symlink). Ideally these operations should not change the 421 * lock state, but should be changed to let the caller of the 422 * function unlock them. Otherwise all intermediate vnode layers 423 * (such as union, umapfs, etc) must catch these functions to do 424 * the necessary locking at their layer. Note that the inactive 425 * and lookup operations also change their lock state, but this 426 * cannot be avoided, so these two operations will always need 427 * to be handled in intermediate layers. 428 */ 429 struct vnode *vp = ap->a_vp; 430 int vnflags, flags = ap->a_flags; 431 432 switch (flags & LK_TYPE_MASK) { 433 case LK_DRAIN: 434 vnflags = LK_DRAIN; 435 break; 436 case LK_EXCLUSIVE: 437 case LK_SHARED: 438 vnflags = LK_SHARED; 439 break; 440 case LK_UPGRADE: 441 case LK_EXCLUPGRADE: 442 case LK_DOWNGRADE: 443 return (0); 444 case LK_RELEASE: 445 default: 446 panic("vop_nolock: bad operation %d", flags & LK_TYPE_MASK); 447 } 448 if (flags & LK_INTERLOCK) 449 vnflags |= LK_INTERLOCK; 450 return(lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p)); 451#else /* for now */ 452 /* 453 * Since we are not using the lock manager, we must clear 454 * the interlock here. 455 */ 456 if (ap->a_flags & LK_INTERLOCK) 457 mtx_unlock(&ap->a_vp->v_interlock); 458 return (0); 459#endif 460} 461 462/* 463 * Do the inverse of vop_nolock, handling the interlock in a compatible way. 464 */ 465int 466vop_nounlock(ap) 467 struct vop_unlock_args /* { 468 struct vnode *a_vp; 469 int a_flags; 470 struct proc *a_p; 471 } */ *ap; 472{ 473 474 /* 475 * Since we are not using the lock manager, we must clear 476 * the interlock here. 477 */ 478 if (ap->a_flags & LK_INTERLOCK) 479 mtx_unlock(&ap->a_vp->v_interlock); 480 return (0); 481} 482 483/* 484 * Return whether or not the node is in use. 485 */ 486int 487vop_noislocked(ap) 488 struct vop_islocked_args /* { 489 struct vnode *a_vp; 490 struct proc *a_p; 491 } */ *ap; 492{ 493 494 return (0); 495} 496 497/* 498 * Return our mount point, as we will take charge of the writes. 499 */ 500int 501vop_stdgetwritemount(ap) 502 struct vop_getwritemount_args /* { 503 struct vnode *a_vp; 504 struct mount **a_mpp; 505 } */ *ap; 506{ 507 508 *(ap->a_mpp) = ap->a_vp->v_mount; 509 return (0); 510} 511 512int 513vop_stdcreatevobject(ap) 514 struct vop_createvobject_args /* { 515 struct vnode *vp; 516 struct ucred *cred; 517 struct proc *p; 518 } */ *ap; 519{ 520 struct vnode *vp = ap->a_vp; 521 struct ucred *cred = ap->a_cred; 522 struct proc *p = ap->a_p; 523 struct vattr vat; 524 vm_object_t object; 525 int error = 0; 526 527 if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE) 528 return (0); 529 530retry: 531 if ((object = vp->v_object) == NULL) { 532 if (vp->v_type == VREG || vp->v_type == VDIR) { 533 if ((error = VOP_GETATTR(vp, &vat, cred, p)) != 0) 534 goto retn; 535 object = vnode_pager_alloc(vp, vat.va_size, 0, 0); 536 } else if (devsw(vp->v_rdev) != NULL) { 537 /* 538 * This simply allocates the biggest object possible 539 * for a disk vnode. This should be fixed, but doesn't 540 * cause any problems (yet). 541 */ 542 object = vnode_pager_alloc(vp, IDX_TO_OFF(INT_MAX), 0, 0); 543 } else { 544 goto retn; 545 } 546 /* 547 * Dereference the reference we just created. This assumes 548 * that the object is associated with the vp. 549 */ 550 object->ref_count--; 551 vp->v_usecount--; 552 } else { 553 if (object->flags & OBJ_DEAD) { 554 VOP_UNLOCK(vp, 0, p); 555 tsleep(object, PVM, "vodead", 0); 556 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); 557 goto retry; 558 } 559 } 560 561 KASSERT(vp->v_object != NULL, ("vfs_object_create: NULL object")); 562 vp->v_flag |= VOBJBUF; 563 564retn: 565 return (error); 566} 567 568int 569vop_stddestroyvobject(ap) 570 struct vop_destroyvobject_args /* { 571 struct vnode *vp; 572 } */ *ap; 573{ 574 struct vnode *vp = ap->a_vp; 575 vm_object_t obj = vp->v_object; 576 577 if (vp->v_object == NULL) 578 return (0); 579 580 if (obj->ref_count == 0) { 581 /* 582 * vclean() may be called twice. The first time 583 * removes the primary reference to the object, 584 * the second time goes one further and is a 585 * special-case to terminate the object. 586 */ 587 vm_object_terminate(obj); 588 } else { 589 /* 590 * Woe to the process that tries to page now :-). 591 */ 592 vm_pager_deallocate(obj); 593 } 594 return (0); 595} 596 597int 598vop_stdgetvobject(ap) 599 struct vop_getvobject_args /* { 600 struct vnode *vp; 601 struct vm_object **objpp; 602 } */ *ap; 603{ 604 struct vnode *vp = ap->a_vp; 605 struct vm_object **objpp = ap->a_objpp; 606 607 if (objpp) 608 *objpp = vp->v_object; 609 return (vp->v_object ? 0 : EINVAL); 610} 611 612/* 613 * vfs default ops 614 * used to fill the vfs fucntion table to get reasonable default return values. 615 */ 616int 617vfs_stdmount (mp, path, data, ndp, p) 618 struct mount *mp; 619 char *path; 620 caddr_t data; 621 struct nameidata *ndp; 622 struct proc *p; 623{ 624 return (0); 625} 626 627int 628vfs_stdunmount (mp, mntflags, p) 629 struct mount *mp; 630 int mntflags; 631 struct proc *p; 632{ 633 return (0); 634} 635 636int 637vfs_stdroot (mp, vpp) 638 struct mount *mp; 639 struct vnode **vpp; 640{ 641 return (EOPNOTSUPP); 642} 643 644int 645vfs_stdstatfs (mp, sbp, p) 646 struct mount *mp; 647 struct statfs *sbp; 648 struct proc *p; 649{ 650 return (EOPNOTSUPP); 651} 652 653int 654vfs_stdvptofh (vp, fhp) 655 struct vnode *vp; 656 struct fid *fhp; 657{ 658 return (EOPNOTSUPP); 659} 660 661int 662vfs_stdstart (mp, flags, p) 663 struct mount *mp; 664 int flags; 665 struct proc *p; 666{ 667 return (0); 668} 669 670int 671vfs_stdquotactl (mp, cmds, uid, arg, p) 672 struct mount *mp; 673 int cmds; 674 uid_t uid; 675 caddr_t arg; 676 struct proc *p; 677{ 678 return (EOPNOTSUPP); 679} 680 681int 682vfs_stdsync (mp, waitfor, cred, p) 683 struct mount *mp; 684 int waitfor; 685 struct ucred *cred; 686 struct proc *p; 687{ 688 return (0); 689} 690 691int 692vfs_stdvget (mp, ino, vpp) 693 struct mount *mp; 694 ino_t ino; 695 struct vnode **vpp; 696{ 697 return (EOPNOTSUPP); 698} 699 700int 701vfs_stdfhtovp (mp, fhp, vpp) 702 struct mount *mp; 703 struct fid *fhp; 704 struct vnode **vpp; 705{ 706 return (EOPNOTSUPP); 707} 708 709int 710vfs_stdinit (vfsp) 711 struct vfsconf *vfsp; 712{ 713 return (0); 714} 715 716int 717vfs_stduninit (vfsp) 718 struct vfsconf *vfsp; 719{ 720 return(0); 721} 722 723int 724vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, p) 725 struct mount *mp; 726 int cmd; 727 struct vnode *filename_vp; 728 int attrnamespace; 729 const char *attrname; 730 struct proc *p; 731{ 732 return(EOPNOTSUPP); 733} 734 735/* end of vfs default ops */ 736