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 * $FreeBSD$ 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 68 static int vop_nolookup __P((struct vop_lookup_args *)); 69 static 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 79 vop_t **default_vnodeop_p; 80 static 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_bwrite_desc, (vop_t *) vop_stdbwrite }, 84 { &vop_close_desc, (vop_t *) vop_null }, 85 { &vop_createvobject_desc, (vop_t *) vop_stdcreatevobject }, 86 { &vop_destroyvobject_desc, (vop_t *) vop_stddestroyvobject }, 87 { &vop_fsync_desc, (vop_t *) vop_null }, 88 { &vop_getvobject_desc, (vop_t *) vop_stdgetvobject }, 89 { &vop_inactive_desc, (vop_t *) vop_stdinactive }, 90 { &vop_ioctl_desc, (vop_t *) vop_enotty }, 91 { &vop_islocked_desc, (vop_t *) vop_noislocked }, 92 { &vop_lease_desc, (vop_t *) vop_null }, 93 { &vop_lock_desc, (vop_t *) vop_nolock }, 94 { &vop_lookup_desc, (vop_t *) vop_nolookup }, 95 { &vop_open_desc, (vop_t *) vop_null }, 96 { &vop_pathconf_desc, (vop_t *) vop_einval }, 97 { &vop_poll_desc, (vop_t *) vop_nopoll }, 98 { &vop_readlink_desc, (vop_t *) vop_einval }, 99 { &vop_revoke_desc, (vop_t *) vop_revoke }, 100 { &vop_strategy_desc, (vop_t *) vop_nostrategy }, 101 { &vop_unlock_desc, (vop_t *) vop_nounlock }, 102 { NULL, NULL } 103 }; 104 105 static struct vnodeopv_desc default_vnodeop_opv_desc = 106 { &default_vnodeop_p, default_vnodeop_entries }; 107 108 VNODEOP_SET(default_vnodeop_opv_desc); 109 110 int 111 vop_eopnotsupp(struct vop_generic_args *ap) 112 { 113 /* 114 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name); 115 */ 116 117 return (EOPNOTSUPP); 118 } 119 120 int 121 vop_ebadf(struct vop_generic_args *ap) 122 { 123 124 return (EBADF); 125 } 126 127 int 128 vop_enotty(struct vop_generic_args *ap) 129 { 130 131 return (ENOTTY); 132 } 133 134 int 135 vop_einval(struct vop_generic_args *ap) 136 { 137 138 return (EINVAL); 139 } 140 141 int 142 vop_null(struct vop_generic_args *ap) 143 { 144 145 return (0); 146 } 147 148 int 149 vop_defaultop(struct vop_generic_args *ap) 150 { 151 152 return (VOCALL(default_vnodeop_p, ap->a_desc->vdesc_offset, ap)); 153 } 154 155 int 156 vop_panic(struct vop_generic_args *ap) 157 { 158 159 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 160 } 161 162 static int 163 vop_nolookup(ap) 164 struct vop_lookup_args /* { 165 struct vnode *a_dvp; 166 struct vnode **a_vpp; 167 struct componentname *a_cnp; 168 } */ *ap; 169 { 170 171 *ap->a_vpp = NULL; 172 return (ENOTDIR); 173 } 174 175 /* 176 * vop_nostrategy: 177 * 178 * Strategy routine for VFS devices that have none. 179 * 180 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 181 * routine. Typically this is done for a BIO_READ strategy call. 182 * Typically B_INVAL is assumed to already be clear prior to a write 183 * and should not be cleared manually unless you just made the buffer 184 * invalid. BIO_ERROR should be cleared either way. 185 */ 186 187 static int 188 vop_nostrategy (struct vop_strategy_args *ap) 189 { 190 printf("No strategy for buffer at %p\n", ap->a_bp); 191 vprint("", ap->a_vp); 192 vprint("", ap->a_bp->b_vp); 193 ap->a_bp->b_ioflags |= BIO_ERROR; 194 ap->a_bp->b_error = EOPNOTSUPP; 195 bufdone(ap->a_bp); 196 return (EOPNOTSUPP); 197 } 198 199 int 200 vop_stdpathconf(ap) 201 struct vop_pathconf_args /* { 202 struct vnode *a_vp; 203 int a_name; 204 int *a_retval; 205 } */ *ap; 206 { 207 208 switch (ap->a_name) { 209 case _PC_LINK_MAX: 210 *ap->a_retval = LINK_MAX; 211 return (0); 212 case _PC_MAX_CANON: 213 *ap->a_retval = MAX_CANON; 214 return (0); 215 case _PC_MAX_INPUT: 216 *ap->a_retval = MAX_INPUT; 217 return (0); 218 case _PC_PIPE_BUF: 219 *ap->a_retval = PIPE_BUF; 220 return (0); 221 case _PC_CHOWN_RESTRICTED: 222 *ap->a_retval = 1; 223 return (0); 224 case _PC_VDISABLE: 225 *ap->a_retval = _POSIX_VDISABLE; 226 return (0); 227 default: 228 return (EINVAL); 229 } 230 /* NOTREACHED */ 231 } 232 233 /* 234 * Standard lock, unlock and islocked functions. 235 * 236 * These depend on the lock structure being the first element in the 237 * inode, ie: vp->v_data points to the the lock! 238 */ 239 int 240 vop_stdlock(ap) 241 struct vop_lock_args /* { 242 struct vnode *a_vp; 243 int a_flags; 244 struct proc *a_p; 245 } */ *ap; 246 { 247 struct vnode *vp = ap->a_vp; 248 249 #ifndef DEBUG_LOCKS 250 return (lockmgr(&vp->v_lock, ap->a_flags, &vp->v_interlock, ap->a_p)); 251 #else 252 return (debuglockmgr(&vp->v_lock, ap->a_flags, &vp->v_interlock, 253 ap->a_p, "vop_stdlock", vp->filename, vp->line)); 254 #endif 255 } 256 257 int 258 vop_stdunlock(ap) 259 struct vop_unlock_args /* { 260 struct vnode *a_vp; 261 int a_flags; 262 struct proc *a_p; 263 } */ *ap; 264 { 265 struct vnode *vp = ap->a_vp; 266 267 return (lockmgr(&vp->v_lock, ap->a_flags | LK_RELEASE, &vp->v_interlock, 268 ap->a_p)); 269 } 270 271 int 272 vop_stdislocked(ap) 273 struct vop_islocked_args /* { 274 struct vnode *a_vp; 275 struct proc *a_p; 276 } */ *ap; 277 { 278 279 return (lockstatus(&ap->a_vp->v_lock, ap->a_p)); 280 } 281 282 int 283 vop_stdinactive(ap) 284 struct vop_inactive_args /* { 285 struct vnode *a_vp; 286 struct proc *a_p; 287 } */ *ap; 288 { 289 290 VOP_UNLOCK(ap->a_vp, 0, ap->a_p); 291 return (0); 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 proc *a_p; 329 } */ *ap; 330 { 331 if ((ap->a_events & ~POLLSTANDARD) == 0) 332 return (ap->a_events & (POLLRDNORM|POLLWRNORM)); 333 return (vn_pollrecord(ap->a_vp, ap->a_p, ap->a_events)); 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 int a_flags; 354 struct proc *a_p; 355 } */ *ap; 356 { 357 /* 358 * This code cannot be used until all the non-locking filesystems 359 * (notably NFS) are converted to properly lock and release nodes. 360 * Also, certain vnode operations change the locking state within 361 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 362 * and symlink). Ideally these operations should not change the 363 * lock state, but should be changed to let the caller of the 364 * function unlock them. Otherwise all intermediate vnode layers 365 * (such as union, umapfs, etc) must catch these functions to do 366 * the necessary locking at their layer. Note that the inactive 367 * and lookup operations also change their lock state, but this 368 * cannot be avoided, so these two operations will always need 369 * to be handled in intermediate layers. 370 */ 371 struct vnode *vp = ap->a_vp; 372 int vnflags, flags = ap->a_flags; 373 374 switch (flags & LK_TYPE_MASK) { 375 case LK_DRAIN: 376 vnflags = LK_DRAIN; 377 break; 378 case LK_EXCLUSIVE: 379 #ifdef DEBUG_VFS_LOCKS 380 /* 381 * Normally, we use shared locks here, but that confuses 382 * the locking assertions. 383 */ 384 vnflags = LK_EXCLUSIVE; 385 break; 386 #endif 387 case LK_SHARED: 388 vnflags = LK_SHARED; 389 break; 390 case LK_UPGRADE: 391 case LK_EXCLUPGRADE: 392 case LK_DOWNGRADE: 393 return (0); 394 case LK_RELEASE: 395 default: 396 panic("vop_sharedlock: bad operation %d", flags & LK_TYPE_MASK); 397 } 398 if (flags & LK_INTERLOCK) 399 vnflags |= LK_INTERLOCK; 400 #ifndef DEBUG_LOCKS 401 return (lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p)); 402 #else 403 return (debuglockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p, 404 "vop_sharedlock", vp->filename, vp->line)); 405 #endif 406 } 407 408 /* 409 * Stubs to use when there is no locking to be done on the underlying object. 410 * A minimal shared lock is necessary to ensure that the underlying object 411 * is not revoked while an operation is in progress. So, an active shared 412 * count is maintained in an auxillary vnode lock structure. 413 */ 414 int 415 vop_nolock(ap) 416 struct vop_lock_args /* { 417 struct vnode *a_vp; 418 int a_flags; 419 struct proc *a_p; 420 } */ *ap; 421 { 422 #ifdef notyet 423 /* 424 * This code cannot be used until all the non-locking filesystems 425 * (notably NFS) are converted to properly lock and release nodes. 426 * Also, certain vnode operations change the locking state within 427 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 428 * and symlink). Ideally these operations should not change the 429 * lock state, but should be changed to let the caller of the 430 * function unlock them. Otherwise all intermediate vnode layers 431 * (such as union, umapfs, etc) must catch these functions to do 432 * the necessary locking at their layer. Note that the inactive 433 * and lookup operations also change their lock state, but this 434 * cannot be avoided, so these two operations will always need 435 * to be handled in intermediate layers. 436 */ 437 struct vnode *vp = ap->a_vp; 438 int vnflags, flags = ap->a_flags; 439 440 switch (flags & LK_TYPE_MASK) { 441 case LK_DRAIN: 442 vnflags = LK_DRAIN; 443 break; 444 case LK_EXCLUSIVE: 445 case LK_SHARED: 446 vnflags = LK_SHARED; 447 break; 448 case LK_UPGRADE: 449 case LK_EXCLUPGRADE: 450 case LK_DOWNGRADE: 451 return (0); 452 case LK_RELEASE: 453 default: 454 panic("vop_nolock: bad operation %d", flags & LK_TYPE_MASK); 455 } 456 if (flags & LK_INTERLOCK) 457 vnflags |= LK_INTERLOCK; 458 return(lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p)); 459 #else /* for now */ 460 /* 461 * Since we are not using the lock manager, we must clear 462 * the interlock here. 463 */ 464 if (ap->a_flags & LK_INTERLOCK) 465 mtx_unlock(&ap->a_vp->v_interlock); 466 return (0); 467 #endif 468 } 469 470 /* 471 * Do the inverse of vop_nolock, handling the interlock in a compatible way. 472 */ 473 int 474 vop_nounlock(ap) 475 struct vop_unlock_args /* { 476 struct vnode *a_vp; 477 int a_flags; 478 struct proc *a_p; 479 } */ *ap; 480 { 481 482 /* 483 * Since we are not using the lock manager, we must clear 484 * the interlock here. 485 */ 486 if (ap->a_flags & LK_INTERLOCK) 487 mtx_unlock(&ap->a_vp->v_interlock); 488 return (0); 489 } 490 491 /* 492 * Return whether or not the node is in use. 493 */ 494 int 495 vop_noislocked(ap) 496 struct vop_islocked_args /* { 497 struct vnode *a_vp; 498 struct proc *a_p; 499 } */ *ap; 500 { 501 502 return (0); 503 } 504 505 /* 506 * Return our mount point, as we will take charge of the writes. 507 */ 508 int 509 vop_stdgetwritemount(ap) 510 struct vop_getwritemount_args /* { 511 struct vnode *a_vp; 512 struct mount **a_mpp; 513 } */ *ap; 514 { 515 516 *(ap->a_mpp) = ap->a_vp->v_mount; 517 return (0); 518 } 519 520 int 521 vop_stdcreatevobject(ap) 522 struct vop_createvobject_args /* { 523 struct vnode *vp; 524 struct ucred *cred; 525 struct proc *p; 526 } */ *ap; 527 { 528 struct vnode *vp = ap->a_vp; 529 struct ucred *cred = ap->a_cred; 530 struct proc *p = ap->a_p; 531 struct vattr vat; 532 vm_object_t object; 533 int error = 0; 534 535 if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE) 536 return (0); 537 538 retry: 539 if ((object = vp->v_object) == NULL) { 540 if (vp->v_type == VREG || vp->v_type == VDIR) { 541 if ((error = VOP_GETATTR(vp, &vat, cred, p)) != 0) 542 goto retn; 543 object = vnode_pager_alloc(vp, vat.va_size, 0, 0); 544 } else if (devsw(vp->v_rdev) != NULL) { 545 /* 546 * This simply allocates the biggest object possible 547 * for a disk vnode. This should be fixed, but doesn't 548 * cause any problems (yet). 549 */ 550 object = vnode_pager_alloc(vp, IDX_TO_OFF(INT_MAX), 0, 0); 551 } else { 552 goto retn; 553 } 554 /* 555 * Dereference the reference we just created. This assumes 556 * that the object is associated with the vp. 557 */ 558 object->ref_count--; 559 vp->v_usecount--; 560 } else { 561 if (object->flags & OBJ_DEAD) { 562 VOP_UNLOCK(vp, 0, p); 563 tsleep(object, PVM, "vodead", 0); 564 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); 565 goto retry; 566 } 567 } 568 569 KASSERT(vp->v_object != NULL, ("vfs_object_create: NULL object")); 570 vp->v_flag |= VOBJBUF; 571 572 retn: 573 return (error); 574 } 575 576 int 577 vop_stddestroyvobject(ap) 578 struct vop_destroyvobject_args /* { 579 struct vnode *vp; 580 } */ *ap; 581 { 582 struct vnode *vp = ap->a_vp; 583 vm_object_t obj = vp->v_object; 584 585 if (vp->v_object == NULL) 586 return (0); 587 588 if (obj->ref_count == 0) { 589 /* 590 * vclean() may be called twice. The first time 591 * removes the primary reference to the object, 592 * the second time goes one further and is a 593 * special-case to terminate the object. 594 */ 595 vm_object_terminate(obj); 596 } else { 597 /* 598 * Woe to the process that tries to page now :-). 599 */ 600 vm_pager_deallocate(obj); 601 } 602 return (0); 603 } 604 605 int 606 vop_stdgetvobject(ap) 607 struct vop_getvobject_args /* { 608 struct vnode *vp; 609 struct vm_object **objpp; 610 } */ *ap; 611 { 612 struct vnode *vp = ap->a_vp; 613 struct vm_object **objpp = ap->a_objpp; 614 615 if (objpp) 616 *objpp = vp->v_object; 617 return (vp->v_object ? 0 : EINVAL); 618 } 619 620 /* 621 * vfs default ops 622 * used to fill the vfs fucntion table to get reasonable default return values. 623 */ 624 int 625 vfs_stdmount (mp, path, data, ndp, p) 626 struct mount *mp; 627 char *path; 628 caddr_t data; 629 struct nameidata *ndp; 630 struct proc *p; 631 { 632 return (0); 633 } 634 635 int 636 vfs_stdunmount (mp, mntflags, p) 637 struct mount *mp; 638 int mntflags; 639 struct proc *p; 640 { 641 return (0); 642 } 643 644 int 645 vfs_stdroot (mp, vpp) 646 struct mount *mp; 647 struct vnode **vpp; 648 { 649 return (EOPNOTSUPP); 650 } 651 652 int 653 vfs_stdstatfs (mp, sbp, p) 654 struct mount *mp; 655 struct statfs *sbp; 656 struct proc *p; 657 { 658 return (EOPNOTSUPP); 659 } 660 661 int 662 vfs_stdvptofh (vp, fhp) 663 struct vnode *vp; 664 struct fid *fhp; 665 { 666 return (EOPNOTSUPP); 667 } 668 669 int 670 vfs_stdstart (mp, flags, p) 671 struct mount *mp; 672 int flags; 673 struct proc *p; 674 { 675 return (0); 676 } 677 678 int 679 vfs_stdquotactl (mp, cmds, uid, arg, p) 680 struct mount *mp; 681 int cmds; 682 uid_t uid; 683 caddr_t arg; 684 struct proc *p; 685 { 686 return (EOPNOTSUPP); 687 } 688 689 int 690 vfs_stdsync (mp, waitfor, cred, p) 691 struct mount *mp; 692 int waitfor; 693 struct ucred *cred; 694 struct proc *p; 695 { 696 return (0); 697 } 698 699 int 700 vfs_stdvget (mp, ino, vpp) 701 struct mount *mp; 702 ino_t ino; 703 struct vnode **vpp; 704 { 705 return (EOPNOTSUPP); 706 } 707 708 int 709 vfs_stdfhtovp (mp, fhp, vpp) 710 struct mount *mp; 711 struct fid *fhp; 712 struct vnode **vpp; 713 { 714 return (EOPNOTSUPP); 715 } 716 717 int 718 vfs_stdcheckexp (mp, nam, extflagsp, credanonp) 719 struct mount *mp; 720 struct sockaddr *nam; 721 int *extflagsp; 722 struct ucred **credanonp; 723 { 724 return (EOPNOTSUPP); 725 } 726 727 int 728 vfs_stdinit (vfsp) 729 struct vfsconf *vfsp; 730 { 731 return (0); 732 } 733 734 int 735 vfs_stduninit (vfsp) 736 struct vfsconf *vfsp; 737 { 738 return(0); 739 } 740 741 int 742 vfs_stdextattrctl(mp, cmd, attrname, arg, p) 743 struct mount *mp; 744 int cmd; 745 const char *attrname; 746 caddr_t arg; 747 struct proc *p; 748 { 749 return(EOPNOTSUPP); 750 } 751 752 /* end of vfs default ops */ 753