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: src/sys/kern/vfs_default.c,v 1.28.2.7 2003/01/10 18:23:26 bde Exp $ 40 * $DragonFly: src/sys/kern/vfs_default.c,v 1.7 2003/07/22 17:03:33 dillon Exp $ 41 */ 42 43 #include <sys/param.h> 44 #include <sys/systm.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/unistd.h> 52 #include <sys/vnode.h> 53 #include <sys/poll.h> 54 55 #include <machine/limits.h> 56 57 #include <vm/vm.h> 58 #include <vm/vm_object.h> 59 #include <vm/vm_page.h> 60 #include <vm/vm_pager.h> 61 #include <vm/vnode_pager.h> 62 63 static int vop_nolookup __P((struct vop_lookup_args *)); 64 static int vop_nostrategy __P((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 74 vop_t **default_vnodeop_p; 75 static struct vnodeopv_entry_desc default_vnodeop_entries[] = { 76 { &vop_default_desc, (vop_t *) vop_eopnotsupp }, 77 { &vop_advlock_desc, (vop_t *) vop_einval }, 78 { &vop_bwrite_desc, (vop_t *) vop_stdbwrite }, 79 { &vop_close_desc, (vop_t *) vop_null }, 80 { &vop_createvobject_desc, (vop_t *) vop_stdcreatevobject }, 81 { &vop_destroyvobject_desc, (vop_t *) vop_stddestroyvobject }, 82 { &vop_fsync_desc, (vop_t *) vop_null }, 83 { &vop_getvobject_desc, (vop_t *) vop_stdgetvobject }, 84 { &vop_ioctl_desc, (vop_t *) vop_enotty }, 85 { &vop_islocked_desc, (vop_t *) vop_noislocked }, 86 { &vop_lease_desc, (vop_t *) vop_null }, 87 { &vop_lock_desc, (vop_t *) vop_nolock }, 88 { &vop_mmap_desc, (vop_t *) vop_einval }, 89 { &vop_lookup_desc, (vop_t *) vop_nolookup }, 90 { &vop_open_desc, (vop_t *) vop_null }, 91 { &vop_pathconf_desc, (vop_t *) vop_einval }, 92 { &vop_poll_desc, (vop_t *) vop_nopoll }, 93 { &vop_readlink_desc, (vop_t *) vop_einval }, 94 { &vop_reallocblks_desc, (vop_t *) vop_eopnotsupp }, 95 { &vop_revoke_desc, (vop_t *) vop_revoke }, 96 { &vop_strategy_desc, (vop_t *) vop_nostrategy }, 97 { &vop_unlock_desc, (vop_t *) vop_nounlock }, 98 { &vop_getacl_desc, (vop_t *) vop_eopnotsupp }, 99 { &vop_setacl_desc, (vop_t *) vop_eopnotsupp }, 100 { &vop_aclcheck_desc, (vop_t *) vop_eopnotsupp }, 101 { &vop_getextattr_desc, (vop_t *) vop_eopnotsupp }, 102 { &vop_setextattr_desc, (vop_t *) vop_eopnotsupp }, 103 { NULL, NULL } 104 }; 105 106 static struct vnodeopv_desc default_vnodeop_opv_desc = 107 { &default_vnodeop_p, default_vnodeop_entries }; 108 109 VNODEOP_SET(default_vnodeop_opv_desc); 110 111 int 112 vop_eopnotsupp(struct vop_generic_args *ap) 113 { 114 /* 115 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 116 */ 117 118 return (EOPNOTSUPP); 119 } 120 121 int 122 vop_ebadf(struct vop_generic_args *ap) 123 { 124 125 return (EBADF); 126 } 127 128 int 129 vop_enotty(struct vop_generic_args *ap) 130 { 131 132 return (ENOTTY); 133 } 134 135 int 136 vop_einval(struct vop_generic_args *ap) 137 { 138 139 return (EINVAL); 140 } 141 142 int 143 vop_null(struct vop_generic_args *ap) 144 { 145 146 return (0); 147 } 148 149 int 150 vop_defaultop(struct vop_generic_args *ap) 151 { 152 153 return (VOCALL(default_vnodeop_p, ap->a_desc->vdesc_offset, ap)); 154 } 155 156 int 157 vop_panic(struct vop_generic_args *ap) 158 { 159 160 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 161 } 162 163 static int 164 vop_nolookup(ap) 165 struct vop_lookup_args /* { 166 struct vnode *a_dvp; 167 struct vnode **a_vpp; 168 struct componentname *a_cnp; 169 } */ *ap; 170 { 171 172 *ap->a_vpp = NULL; 173 return (ENOTDIR); 174 } 175 176 /* 177 * vop_nostrategy: 178 * 179 * Strategy routine for VFS devices that have none. 180 * 181 * B_ERROR and B_INVAL must be cleared prior to calling any strategy 182 * routine. Typically this is done for a B_READ strategy call. Typically 183 * B_INVAL is assumed to already be clear prior to a write and should not 184 * be cleared manually unless you just made the buffer invalid. B_ERROR 185 * should be cleared either way. 186 */ 187 188 static int 189 vop_nostrategy (struct vop_strategy_args *ap) 190 { 191 printf("No strategy for buffer at %p\n", ap->a_bp); 192 vprint("", ap->a_vp); 193 vprint("", ap->a_bp->b_vp); 194 ap->a_bp->b_flags |= B_ERROR; 195 ap->a_bp->b_error = EOPNOTSUPP; 196 biodone(ap->a_bp); 197 return (EOPNOTSUPP); 198 } 199 200 int 201 vop_stdpathconf(ap) 202 struct vop_pathconf_args /* { 203 struct vnode *a_vp; 204 int a_name; 205 int *a_retval; 206 } */ *ap; 207 { 208 209 switch (ap->a_name) { 210 case _PC_LINK_MAX: 211 *ap->a_retval = LINK_MAX; 212 return (0); 213 case _PC_MAX_CANON: 214 *ap->a_retval = MAX_CANON; 215 return (0); 216 case _PC_MAX_INPUT: 217 *ap->a_retval = MAX_INPUT; 218 return (0); 219 case _PC_PIPE_BUF: 220 *ap->a_retval = PIPE_BUF; 221 return (0); 222 case _PC_CHOWN_RESTRICTED: 223 *ap->a_retval = 1; 224 return (0); 225 case _PC_VDISABLE: 226 *ap->a_retval = _POSIX_VDISABLE; 227 return (0); 228 default: 229 return (EINVAL); 230 } 231 /* NOTREACHED */ 232 } 233 234 /* 235 * Standard lock, unlock and islocked functions. 236 * 237 * These depend on the lock structure being the first element in the 238 * inode, ie: vp->v_data points to the the lock! 239 */ 240 int 241 vop_stdlock(ap) 242 struct vop_lock_args /* { 243 struct vnode *a_vp; 244 int a_flags; 245 struct proc *a_p; 246 } */ *ap; 247 { 248 struct lock *l; 249 250 if ((l = (struct lock *)ap->a_vp->v_data) == NULL) { 251 if (ap->a_flags & LK_INTERLOCK) 252 lwkt_reltoken(&ap->a_vp->v_interlock); 253 return 0; 254 } 255 256 #ifndef DEBUG_LOCKS 257 return (lockmgr(l, ap->a_flags, &ap->a_vp->v_interlock, ap->a_td)); 258 #else 259 return (debuglockmgr(l, ap->a_flags, &ap->a_vp->v_interlock, ap->a_td, 260 "vop_stdlock", ap->a_vp->filename, ap->a_vp->line)); 261 #endif 262 } 263 264 int 265 vop_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 lock *l; 273 274 if ((l = (struct lock *)ap->a_vp->v_data) == NULL) { 275 if (ap->a_flags & LK_INTERLOCK) 276 lwkt_reltoken(&ap->a_vp->v_interlock); 277 return 0; 278 } 279 280 return (lockmgr(l, ap->a_flags | LK_RELEASE, 281 &ap->a_vp->v_interlock, ap->a_td)); 282 } 283 284 int 285 vop_stdislocked(ap) 286 struct vop_islocked_args /* { 287 struct vnode *a_vp; 288 struct thread *a_td; 289 } */ *ap; 290 { 291 struct lock *l; 292 293 if ((l = (struct lock *)ap->a_vp->v_data) == NULL) 294 return 0; 295 296 return (lockstatus(l, ap->a_td)); 297 } 298 299 /* 300 * Return true for select/poll. 301 */ 302 int 303 vop_nopoll(ap) 304 struct vop_poll_args /* { 305 struct vnode *a_vp; 306 int a_events; 307 struct ucred *a_cred; 308 struct proc *a_p; 309 } */ *ap; 310 { 311 /* 312 * Return true for read/write. If the user asked for something 313 * special, return POLLNVAL, so that clients have a way of 314 * determining reliably whether or not the extended 315 * functionality is present without hard-coding knowledge 316 * of specific filesystem implementations. 317 */ 318 if (ap->a_events & ~POLLSTANDARD) 319 return (POLLNVAL); 320 321 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 322 } 323 324 /* 325 * Implement poll for local filesystems that support it. 326 */ 327 int 328 vop_stdpoll(ap) 329 struct vop_poll_args /* { 330 struct vnode *a_vp; 331 int a_events; 332 struct ucred *a_cred; 333 struct thread *a_td; 334 } */ *ap; 335 { 336 if (ap->a_events & ~POLLSTANDARD) 337 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 338 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 339 } 340 341 int 342 vop_stdbwrite(ap) 343 struct vop_bwrite_args *ap; 344 { 345 return (bwrite(ap->a_bp)); 346 } 347 348 /* 349 * Stubs to use when there is no locking to be done on the underlying object. 350 * A minimal shared lock is necessary to ensure that the underlying object 351 * is not revoked while an operation is in progress. So, an active shared 352 * count is maintained in an auxillary vnode lock structure. 353 */ 354 int 355 vop_sharedlock(ap) 356 struct vop_lock_args /* { 357 struct vnode *a_vp; 358 int a_flags; 359 struct proc *a_p; 360 } */ *ap; 361 { 362 /* 363 * This code cannot be used until all the non-locking filesystems 364 * (notably NFS) are converted to properly lock and release nodes. 365 * Also, certain vnode operations change the locking state within 366 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 367 * and symlink). Ideally these operations should not change the 368 * lock state, but should be changed to let the caller of the 369 * function unlock them. Otherwise all intermediate vnode layers 370 * (such as union, umapfs, etc) must catch these functions to do 371 * the necessary locking at their layer. Note that the inactive 372 * and lookup operations also change their lock state, but this 373 * cannot be avoided, so these two operations will always need 374 * to be handled in intermediate layers. 375 */ 376 struct vnode *vp = ap->a_vp; 377 struct lock *l = (struct lock *)vp->v_data; 378 int vnflags, flags = ap->a_flags; 379 380 if (l == NULL) { 381 if (ap->a_flags & LK_INTERLOCK) 382 lwkt_reltoken(&ap->a_vp->v_interlock); 383 return 0; 384 } 385 switch (flags & LK_TYPE_MASK) { 386 case LK_DRAIN: 387 vnflags = LK_DRAIN; 388 break; 389 case LK_EXCLUSIVE: 390 #ifdef DEBUG_VFS_LOCKS 391 /* 392 * Normally, we use shared locks here, but that confuses 393 * the locking assertions. 394 */ 395 vnflags = LK_EXCLUSIVE; 396 break; 397 #endif 398 case LK_SHARED: 399 vnflags = LK_SHARED; 400 break; 401 case LK_UPGRADE: 402 case LK_EXCLUPGRADE: 403 case LK_DOWNGRADE: 404 return (0); 405 case LK_RELEASE: 406 default: 407 panic("vop_sharedlock: bad operation %d", flags & LK_TYPE_MASK); 408 } 409 if (flags & LK_INTERLOCK) 410 vnflags |= LK_INTERLOCK; 411 #ifndef DEBUG_LOCKS 412 return (lockmgr(l, vnflags, &vp->v_interlock, ap->a_td)); 413 #else 414 return (debuglockmgr(l, vnflags, &vp->v_interlock, ap->a_td, 415 "vop_sharedlock", vp->filename, vp->line)); 416 #endif 417 } 418 419 /* 420 * Stubs to use when there is no locking to be done on the underlying object. 421 * A minimal shared lock is necessary to ensure that the underlying object 422 * is not revoked while an operation is in progress. So, an active shared 423 * count is maintained in an auxillary vnode lock structure. 424 */ 425 int 426 vop_nolock(ap) 427 struct vop_lock_args /* { 428 struct vnode *a_vp; 429 int a_flags; 430 struct proc *a_p; 431 } */ *ap; 432 { 433 #ifdef notyet 434 /* 435 * This code cannot be used until all the non-locking filesystems 436 * (notably NFS) are converted to properly lock and release nodes. 437 * Also, certain vnode operations change the locking state within 438 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 439 * and symlink). Ideally these operations should not change the 440 * lock state, but should be changed to let the caller of the 441 * function unlock them. Otherwise all intermediate vnode layers 442 * (such as union, umapfs, etc) must catch these functions to do 443 * the necessary locking at their layer. Note that the inactive 444 * and lookup operations also change their lock state, but this 445 * cannot be avoided, so these two operations will always need 446 * to be handled in intermediate layers. 447 */ 448 struct vnode *vp = ap->a_vp; 449 int vnflags, flags = ap->a_flags; 450 451 switch (flags & LK_TYPE_MASK) { 452 case LK_DRAIN: 453 vnflags = LK_DRAIN; 454 break; 455 case LK_EXCLUSIVE: 456 case LK_SHARED: 457 vnflags = LK_SHARED; 458 break; 459 case LK_UPGRADE: 460 case LK_EXCLUPGRADE: 461 case LK_DOWNGRADE: 462 return (0); 463 case LK_RELEASE: 464 default: 465 panic("vop_nolock: bad operation %d", flags & LK_TYPE_MASK); 466 } 467 if (flags & LK_INTERLOCK) 468 vnflags |= LK_INTERLOCK; 469 return(lockmgr(vp->v_vnlock, vnflags, &vp->v_interlock, ap->a_p)); 470 #else /* for now */ 471 /* 472 * Since we are not using the lock manager, we must clear 473 * the interlock here. 474 */ 475 if (ap->a_flags & LK_INTERLOCK) 476 lwkt_reltoken(&ap->a_vp->v_interlock); 477 return (0); 478 #endif 479 } 480 481 /* 482 * Do the inverse of vop_nolock, handling the interlock in a compatible way. 483 */ 484 int 485 vop_nounlock(ap) 486 struct vop_unlock_args /* { 487 struct vnode *a_vp; 488 int a_flags; 489 struct proc *a_p; 490 } */ *ap; 491 { 492 if (ap->a_flags & LK_INTERLOCK) 493 lwkt_reltoken(&ap->a_vp->v_interlock); 494 return (0); 495 } 496 497 /* 498 * Return whether or not the node is in use. 499 */ 500 int 501 vop_noislocked(ap) 502 struct vop_islocked_args /* { 503 struct vnode *a_vp; 504 struct proc *a_p; 505 } */ *ap; 506 { 507 return (0); 508 } 509 510 int 511 vop_stdcreatevobject(ap) 512 struct vop_createvobject_args /* { 513 struct vnode *a_vp; 514 struct proc *a_td; 515 } */ *ap; 516 { 517 struct vnode *vp = ap->a_vp; 518 struct thread *td = ap->a_td; 519 struct vattr vat; 520 vm_object_t object; 521 int error = 0; 522 523 if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE) 524 return (0); 525 526 retry: 527 if ((object = vp->v_object) == NULL) { 528 if (vp->v_type == VREG || vp->v_type == VDIR) { 529 if ((error = VOP_GETATTR(vp, &vat, td)) != 0) 530 goto retn; 531 object = vnode_pager_alloc(vp, vat.va_size, 0, 0); 532 } else if (dev_dport(vp->v_rdev) != NULL) { 533 /* 534 * This simply allocates the biggest object possible 535 * for a disk vnode. This should be fixed, but doesn't 536 * cause any problems (yet). 537 */ 538 object = vnode_pager_alloc(vp, IDX_TO_OFF(INT_MAX), 0, 0); 539 } else { 540 goto retn; 541 } 542 /* 543 * Dereference the reference we just created. This assumes 544 * that the object is associated with the vp. 545 */ 546 object->ref_count--; 547 vp->v_usecount--; 548 } else { 549 if (object->flags & OBJ_DEAD) { 550 VOP_UNLOCK(vp, 0, td); 551 tsleep(object, 0, "vodead", 0); 552 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 553 goto retry; 554 } 555 } 556 557 KASSERT(vp->v_object != NULL, ("vfs_object_create: NULL object")); 558 vp->v_flag |= VOBJBUF; 559 560 retn: 561 return (error); 562 } 563 564 int 565 vop_stddestroyvobject(ap) 566 struct vop_destroyvobject_args /* { 567 struct vnode *vp; 568 } */ *ap; 569 { 570 struct vnode *vp = ap->a_vp; 571 vm_object_t obj = vp->v_object; 572 573 if (vp->v_object == NULL) 574 return (0); 575 576 if (obj->ref_count == 0) { 577 /* 578 * vclean() may be called twice. The first time 579 * removes the primary reference to the object, 580 * the second time goes one further and is a 581 * special-case to terminate the object. 582 * 583 * don't double-terminate the object. 584 */ 585 if ((obj->flags & OBJ_DEAD) == 0) 586 vm_object_terminate(obj); 587 } else { 588 /* 589 * Woe to the process that tries to page now :-). 590 */ 591 vm_pager_deallocate(obj); 592 } 593 return (0); 594 } 595 596 /* 597 * Return the underlying VM object. This routine may be called with or 598 * without the vnode interlock held. If called without, the returned 599 * object is not guarenteed to be valid. The syncer typically gets the 600 * object without holding the interlock in order to quickly test whether 601 * it might be dirty before going heavy-weight. vm_object's use zalloc 602 * and thus stable-storage, so this is safe. 603 */ 604 int 605 vop_stdgetvobject(ap) 606 struct vop_getvobject_args /* { 607 struct vnode *vp; 608 struct vm_object **objpp; 609 } */ *ap; 610 { 611 struct vnode *vp = ap->a_vp; 612 struct vm_object **objpp = ap->a_objpp; 613 614 if (objpp) 615 *objpp = vp->v_object; 616 return (vp->v_object ? 0 : EINVAL); 617 } 618 619 /* 620 * vfs default ops 621 * used to fill the vfs fucntion table to get reasonable default return values. 622 */ 623 int 624 vfs_stdmount(struct mount *mp, char *path, caddr_t data, 625 struct nameidata *ndp, struct thread *td) 626 { 627 return (0); 628 } 629 630 int 631 vfs_stdunmount(struct mount *mp, int mntflags, struct thread *td) 632 { 633 return (0); 634 } 635 636 int 637 vfs_stdroot(struct mount *mp, struct vnode **vpp) 638 { 639 return (EOPNOTSUPP); 640 } 641 642 int 643 vfs_stdstatfs(struct mount *mp, struct statfs *sbp, struct thread *td) 644 { 645 return (EOPNOTSUPP); 646 } 647 648 int 649 vfs_stdvptofh(struct vnode *vp, struct fid *fhp) 650 { 651 return (EOPNOTSUPP); 652 } 653 654 int 655 vfs_stdstart(struct mount *mp, int flags, struct thread *td) 656 { 657 return (0); 658 } 659 660 int 661 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, 662 caddr_t arg, struct thread *td) 663 { 664 return (EOPNOTSUPP); 665 } 666 667 int 668 vfs_stdsync(struct mount *mp, int waitfor, struct thread *td) 669 { 670 return (0); 671 } 672 673 int 674 vfs_stdvget(struct mount *mp, ino_t ino, struct vnode **vpp) 675 { 676 return (EOPNOTSUPP); 677 } 678 679 int 680 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp) 681 { 682 return (EOPNOTSUPP); 683 } 684 685 int 686 vfs_stdcheckexp(struct mount *mp, struct sockaddr *nam, int *extflagsp, 687 struct ucred **credanonp) 688 { 689 return (EOPNOTSUPP); 690 } 691 692 int 693 vfs_stdinit(struct vfsconf *vfsp) 694 { 695 return (0); 696 } 697 698 int 699 vfs_stduninit(struct vfsconf *vfsp) 700 { 701 return(0); 702 } 703 704 int 705 vfs_stdextattrctl(struct mount *mp, int cmd, const char *attrname, 706 caddr_t arg, struct thread *td) 707 { 708 return(EOPNOTSUPP); 709 } 710 711 /* end of vfs default ops */ 712