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 1.12 2004/08/17 18:57:32 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 (struct vop_lookup_args *); 64 static 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 74 struct vop_ops *default_vnode_vops; 75 static struct vnodeopv_entry_desc default_vnodeop_entries[] = { 76 { &vop_default_desc, vop_eopnotsupp }, 77 { &vop_advlock_desc, vop_einval }, 78 { &vop_bwrite_desc, (void *) vop_stdbwrite }, 79 { &vop_close_desc, vop_null }, 80 { &vop_createvobject_desc, (void *) vop_stdcreatevobject }, 81 { &vop_destroyvobject_desc, (void *) vop_stddestroyvobject }, 82 { &vop_fsync_desc, vop_null }, 83 { &vop_getvobject_desc, (void *) vop_stdgetvobject }, 84 { &vop_ioctl_desc, vop_enotty }, 85 { &vop_islocked_desc, (void *) vop_noislocked }, 86 { &vop_lease_desc, vop_null }, 87 { &vop_lock_desc, (void *) vop_nolock }, 88 { &vop_mmap_desc, vop_einval }, 89 { &vop_lookup_desc, (void *) vop_nolookup }, 90 { &vop_open_desc, vop_null }, 91 { &vop_pathconf_desc, vop_einval }, 92 { &vop_poll_desc, (void *) vop_nopoll }, 93 { &vop_readlink_desc, vop_einval }, 94 { &vop_reallocblks_desc, vop_eopnotsupp }, 95 { &vop_revoke_desc, (void *) vop_stdrevoke }, 96 { &vop_strategy_desc, (void *) vop_nostrategy }, 97 { &vop_unlock_desc, (void *) vop_nounlock }, 98 { &vop_getacl_desc, vop_eopnotsupp }, 99 { &vop_setacl_desc, vop_eopnotsupp }, 100 { &vop_aclcheck_desc, vop_eopnotsupp }, 101 { &vop_getextattr_desc, vop_eopnotsupp }, 102 { &vop_setextattr_desc, vop_eopnotsupp }, 103 { NULL, NULL } 104 }; 105 106 static struct vnodeopv_desc default_vnodeop_opv_desc = 107 { &default_vnode_vops, 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 return (EOPNOTSUPP); 118 } 119 120 int 121 vop_ebadf(struct vop_generic_args *ap) 122 { 123 return (EBADF); 124 } 125 126 int 127 vop_enotty(struct vop_generic_args *ap) 128 { 129 return (ENOTTY); 130 } 131 132 int 133 vop_einval(struct vop_generic_args *ap) 134 { 135 return (EINVAL); 136 } 137 138 int 139 vop_null(struct vop_generic_args *ap) 140 { 141 return (0); 142 } 143 144 int 145 vop_defaultop(struct vop_generic_args *ap) 146 { 147 return (VOCALL(default_vnode_vops, ap)); 148 } 149 150 int 151 vop_panic(struct vop_generic_args *ap) 152 { 153 154 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 155 } 156 157 static int 158 vop_nolookup(ap) 159 struct vop_lookup_args /* { 160 struct vnode *a_dvp; 161 struct vnode **a_vpp; 162 struct componentname *a_cnp; 163 } */ *ap; 164 { 165 166 *ap->a_vpp = NULL; 167 return (ENOTDIR); 168 } 169 170 /* 171 * vop_nostrategy: 172 * 173 * Strategy routine for VFS devices that have none. 174 * 175 * B_ERROR and B_INVAL must be cleared prior to calling any strategy 176 * routine. Typically this is done for a B_READ strategy call. Typically 177 * B_INVAL is assumed to already be clear prior to a write and should not 178 * be cleared manually unless you just made the buffer invalid. B_ERROR 179 * should be cleared either way. 180 */ 181 182 static int 183 vop_nostrategy (struct vop_strategy_args *ap) 184 { 185 printf("No strategy for buffer at %p\n", ap->a_bp); 186 vprint("", ap->a_vp); 187 vprint("", ap->a_bp->b_vp); 188 ap->a_bp->b_flags |= B_ERROR; 189 ap->a_bp->b_error = EOPNOTSUPP; 190 biodone(ap->a_bp); 191 return (EOPNOTSUPP); 192 } 193 194 int 195 vop_stdpathconf(ap) 196 struct vop_pathconf_args /* { 197 struct vnode *a_vp; 198 int a_name; 199 int *a_retval; 200 } */ *ap; 201 { 202 203 switch (ap->a_name) { 204 case _PC_LINK_MAX: 205 *ap->a_retval = LINK_MAX; 206 return (0); 207 case _PC_MAX_CANON: 208 *ap->a_retval = MAX_CANON; 209 return (0); 210 case _PC_MAX_INPUT: 211 *ap->a_retval = MAX_INPUT; 212 return (0); 213 case _PC_PIPE_BUF: 214 *ap->a_retval = PIPE_BUF; 215 return (0); 216 case _PC_CHOWN_RESTRICTED: 217 *ap->a_retval = 1; 218 return (0); 219 case _PC_VDISABLE: 220 *ap->a_retval = _POSIX_VDISABLE; 221 return (0); 222 default: 223 return (EINVAL); 224 } 225 /* NOTREACHED */ 226 } 227 228 /* 229 * Standard lock, unlock and islocked functions. 230 * 231 * These depend on the lock structure being the first element in the 232 * inode, ie: vp->v_data points to the the lock! 233 */ 234 int 235 vop_stdlock(ap) 236 struct vop_lock_args /* { 237 struct vnode *a_vp; 238 lwkt_tokref_t a_vlock; 239 int a_flags; 240 struct proc *a_p; 241 } */ *ap; 242 { 243 struct lock *l; 244 245 if ((l = (struct lock *)ap->a_vp->v_data) == NULL) { 246 if (ap->a_flags & LK_INTERLOCK) 247 lwkt_reltoken(ap->a_vlock); 248 return 0; 249 } 250 251 #ifndef DEBUG_LOCKS 252 return (lockmgr(l, ap->a_flags, ap->a_vlock, ap->a_td)); 253 #else 254 return (debuglockmgr(l, ap->a_flags, ap->a_vlock, ap->a_td, 255 "vop_stdlock", ap->a_vp->filename, ap->a_vp->line)); 256 #endif 257 } 258 259 int 260 vop_stdunlock(ap) 261 struct vop_unlock_args /* { 262 struct vnode *a_vp; 263 lwkt_tokref_t a_vlock; 264 int a_flags; 265 struct thread *a_td; 266 } */ *ap; 267 { 268 struct lock *l; 269 270 if ((l = (struct lock *)ap->a_vp->v_data) == NULL) { 271 if (ap->a_flags & LK_INTERLOCK) 272 lwkt_reltoken(ap->a_vlock); 273 return 0; 274 } 275 276 return (lockmgr(l, ap->a_flags | LK_RELEASE, ap->a_vlock, ap->a_td)); 277 } 278 279 int 280 vop_stdislocked(ap) 281 struct vop_islocked_args /* { 282 struct vnode *a_vp; 283 struct thread *a_td; 284 } */ *ap; 285 { 286 struct lock *l; 287 288 if ((l = (struct lock *)ap->a_vp->v_data) == NULL) 289 return 0; 290 291 return (lockstatus(l, ap->a_td)); 292 } 293 294 /* 295 * Return true for select/poll. 296 */ 297 int 298 vop_nopoll(ap) 299 struct vop_poll_args /* { 300 struct vnode *a_vp; 301 int a_events; 302 struct ucred *a_cred; 303 struct proc *a_p; 304 } */ *ap; 305 { 306 /* 307 * Return true for read/write. If the user asked for something 308 * special, return POLLNVAL, so that clients have a way of 309 * determining reliably whether or not the extended 310 * functionality is present without hard-coding knowledge 311 * of specific filesystem implementations. 312 */ 313 if (ap->a_events & ~POLLSTANDARD) 314 return (POLLNVAL); 315 316 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 317 } 318 319 /* 320 * Implement poll for local filesystems that support it. 321 */ 322 int 323 vop_stdpoll(ap) 324 struct vop_poll_args /* { 325 struct vnode *a_vp; 326 int a_events; 327 struct ucred *a_cred; 328 struct thread *a_td; 329 } */ *ap; 330 { 331 if (ap->a_events & ~POLLSTANDARD) 332 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 333 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 334 } 335 336 int 337 vop_stdbwrite(ap) 338 struct vop_bwrite_args *ap; 339 { 340 return (bwrite(ap->a_bp)); 341 } 342 343 /* 344 * Stubs to use when there is no locking to be done on the underlying object. 345 * A minimal shared lock is necessary to ensure that the underlying object 346 * is not revoked while an operation is in progress. So, an active shared 347 * count is maintained in an auxillary vnode lock structure. 348 */ 349 int 350 vop_sharedlock(ap) 351 struct vop_lock_args /* { 352 struct vnode *a_vp; 353 lwkt_tokref_t a_vlock; 354 int a_flags; 355 struct proc *a_p; 356 } */ *ap; 357 { 358 /* 359 * This code cannot be used until all the non-locking filesystems 360 * (notably NFS) are converted to properly lock and release nodes. 361 * Also, certain vnode operations change the locking state within 362 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 363 * and symlink). Ideally these operations should not change the 364 * lock state, but should be changed to let the caller of the 365 * function unlock them. Otherwise all intermediate vnode layers 366 * (such as union, umapfs, etc) must catch these functions to do 367 * the necessary locking at their layer. Note that the inactive 368 * and lookup operations also change their lock state, but this 369 * cannot be avoided, so these two operations will always need 370 * to be handled in intermediate layers. 371 */ 372 struct vnode *vp = ap->a_vp; 373 struct lock *l = (struct lock *)vp->v_data; 374 int vnflags, flags = ap->a_flags; 375 376 if (l == NULL) { 377 if (ap->a_flags & LK_INTERLOCK) 378 lwkt_reltoken(ap->a_vlock); 379 return 0; 380 } 381 switch (flags & LK_TYPE_MASK) { 382 case LK_DRAIN: 383 vnflags = LK_DRAIN; 384 break; 385 case LK_EXCLUSIVE: 386 #ifdef DEBUG_VFS_LOCKS 387 /* 388 * Normally, we use shared locks here, but that confuses 389 * the locking assertions. 390 */ 391 vnflags = LK_EXCLUSIVE; 392 break; 393 #endif 394 case LK_SHARED: 395 vnflags = LK_SHARED; 396 break; 397 case LK_UPGRADE: 398 case LK_EXCLUPGRADE: 399 case LK_DOWNGRADE: 400 return (0); 401 case LK_RELEASE: 402 default: 403 panic("vop_sharedlock: bad operation %d", flags & LK_TYPE_MASK); 404 } 405 if (flags & LK_INTERLOCK) 406 vnflags |= LK_INTERLOCK; 407 #ifndef DEBUG_LOCKS 408 return (lockmgr(l, vnflags, ap->a_vlock, ap->a_td)); 409 #else 410 return (debuglockmgr(l, vnflags, ap->a_vlock, ap->a_td, 411 "vop_sharedlock", vp->filename, vp->line)); 412 #endif 413 } 414 415 /* 416 * Stubs to use when there is no locking to be done on the underlying object. 417 * A minimal shared lock is necessary to ensure that the underlying object 418 * is not revoked while an operation is in progress. So, an active shared 419 * count is maintained in an auxillary vnode lock structure. 420 */ 421 int 422 vop_nolock(ap) 423 struct vop_lock_args /* { 424 struct vnode *a_vp; 425 lwkt_tokref_t a_vlock; 426 int a_flags; 427 struct proc *a_p; 428 } */ *ap; 429 { 430 #ifdef notyet 431 /* 432 * This code cannot be used until all the non-locking filesystems 433 * (notably NFS) are converted to properly lock and release nodes. 434 * Also, certain vnode operations change the locking state within 435 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 436 * and symlink). Ideally these operations should not change the 437 * lock state, but should be changed to let the caller of the 438 * function unlock them. Otherwise all intermediate vnode layers 439 * (such as union, umapfs, etc) must catch these functions to do 440 * the necessary locking at their layer. Note that the inactive 441 * and lookup operations also change their lock state, but this 442 * cannot be avoided, so these two operations will always need 443 * to be handled in intermediate layers. 444 */ 445 struct vnode *vp = ap->a_vp; 446 int vnflags, flags = ap->a_flags; 447 448 switch (flags & LK_TYPE_MASK) { 449 case LK_DRAIN: 450 vnflags = LK_DRAIN; 451 break; 452 case LK_EXCLUSIVE: 453 case LK_SHARED: 454 vnflags = LK_SHARED; 455 break; 456 case LK_UPGRADE: 457 case LK_EXCLUPGRADE: 458 case LK_DOWNGRADE: 459 return (0); 460 case LK_RELEASE: 461 default: 462 panic("vop_nolock: bad operation %d", flags & LK_TYPE_MASK); 463 } 464 if (flags & LK_INTERLOCK) 465 vnflags |= LK_INTERLOCK; 466 return(lockmgr(vp->v_vnlock, vnflags, ap->a_vlock, ap->a_p)); 467 #else /* for now */ 468 /* 469 * Since we are not using the lock manager, we must clear 470 * the interlock here. 471 */ 472 if (ap->a_flags & LK_INTERLOCK) 473 lwkt_reltoken(ap->a_vlock); 474 return (0); 475 #endif 476 } 477 478 /* 479 * Do the inverse of vop_nolock, handling the interlock in a compatible way. 480 */ 481 int 482 vop_nounlock(ap) 483 struct vop_unlock_args /* { 484 struct vnode *a_vp; 485 lwkt_tokref_t a_vlock; 486 int a_flags; 487 struct proc *a_p; 488 } */ *ap; 489 { 490 if (ap->a_flags & LK_INTERLOCK) 491 lwkt_reltoken(ap->a_vlock); 492 return (0); 493 } 494 495 /* 496 * Return whether or not the node is in use. 497 */ 498 int 499 vop_noislocked(ap) 500 struct vop_islocked_args /* { 501 struct vnode *a_vp; 502 struct proc *a_p; 503 } */ *ap; 504 { 505 return (0); 506 } 507 508 int 509 vop_stdcreatevobject(ap) 510 struct vop_createvobject_args /* { 511 struct vnode *a_vp; 512 struct proc *a_td; 513 } */ *ap; 514 { 515 struct vnode *vp = ap->a_vp; 516 struct thread *td = ap->a_td; 517 struct vattr vat; 518 vm_object_t object; 519 int error = 0; 520 521 if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE) 522 return (0); 523 524 retry: 525 if ((object = vp->v_object) == NULL) { 526 if (vp->v_type == VREG || vp->v_type == VDIR) { 527 if ((error = VOP_GETATTR(vp, &vat, td)) != 0) 528 goto retn; 529 object = vnode_pager_alloc(vp, vat.va_size, 0, 0); 530 } else if (vp->v_rdev && dev_is_good(vp->v_rdev)) { 531 /* 532 * XXX v_rdev uses NULL/non-NULL instead of NODEV 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, NULL, 0, td); 551 tsleep(object, 0, "vodead", 0); 552 vn_lock(vp, NULL, 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