vfs_default.c revision 66615
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 66615 2000-10-04 01:29:17Z jasone $ 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/unistd.h> 52#include <sys/vnode.h> 53#include <sys/poll.h> 54 55#include <machine/limits.h> 56#include <machine/mutex.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_nostrategy __P((struct vop_strategy_args *)); 69 70/* 71 * This vnode table stores what we want to do if the filesystem doesn't 72 * implement a particular VOP. 73 * 74 * If there is no specific entry here, we will return EOPNOTSUPP. 75 * 76 */ 77 78vop_t **default_vnodeop_p; 79static struct vnodeopv_entry_desc default_vnodeop_entries[] = { 80 { &vop_default_desc, (vop_t *) vop_eopnotsupp }, 81 { &vop_advlock_desc, (vop_t *) vop_einval }, 82 { &vop_bwrite_desc, (vop_t *) vop_stdbwrite }, 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_mmap_desc, (vop_t *) vop_einval }, 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 printf("vop_panic[%s]\n", ap->a_desc->vdesc_name); 159 panic("Filesystem goof"); 160 return (0); 161} 162 163/* 164 * vop_nostrategy: 165 * 166 * Strategy routine for VFS devices that have none. 167 * 168 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy 169 * routine. Typically this is done for a BIO_READ strategy call. 170 * Typically B_INVAL is assumed to already be clear prior to a write 171 * and should not be cleared manually unless you just made the buffer 172 * invalid. BIO_ERROR should be cleared either way. 173 */ 174 175static int 176vop_nostrategy (struct vop_strategy_args *ap) 177{ 178 printf("No strategy for buffer at %p\n", ap->a_bp); 179 vprint("", ap->a_vp); 180 vprint("", ap->a_bp->b_vp); 181 ap->a_bp->b_ioflags |= BIO_ERROR; 182 ap->a_bp->b_error = EOPNOTSUPP; 183 bufdone(ap->a_bp); 184 return (EOPNOTSUPP); 185} 186 187int 188vop_stdpathconf(ap) 189 struct vop_pathconf_args /* { 190 struct vnode *a_vp; 191 int a_name; 192 int *a_retval; 193 } */ *ap; 194{ 195 196 switch (ap->a_name) { 197 case _PC_LINK_MAX: 198 *ap->a_retval = LINK_MAX; 199 return (0); 200 case _PC_MAX_CANON: 201 *ap->a_retval = MAX_CANON; 202 return (0); 203 case _PC_MAX_INPUT: 204 *ap->a_retval = MAX_INPUT; 205 return (0); 206 case _PC_PIPE_BUF: 207 *ap->a_retval = PIPE_BUF; 208 return (0); 209 case _PC_CHOWN_RESTRICTED: 210 *ap->a_retval = 1; 211 return (0); 212 case _PC_VDISABLE: 213 *ap->a_retval = _POSIX_VDISABLE; 214 return (0); 215 default: 216 return (EINVAL); 217 } 218 /* NOTREACHED */ 219} 220 221/* 222 * Standard lock, unlock and islocked functions. 223 * 224 * These depend on the lock structure being the first element in the 225 * inode, ie: vp->v_data points to the the lock! 226 */ 227int 228vop_stdlock(ap) 229 struct vop_lock_args /* { 230 struct vnode *a_vp; 231 int a_flags; 232 struct proc *a_p; 233 } */ *ap; 234{ 235 struct vnode *vp = ap->a_vp; 236 237#ifndef DEBUG_LOCKS 238 return (lockmgr(&vp->v_lock, ap->a_flags, &vp->v_interlock, ap->a_p)); 239#else 240 return (debuglockmgr(&vp->v_lock, ap->a_flags, &vp->v_interlock, 241 ap->a_p, "vop_stdlock", vp->filename, vp->line)); 242#endif 243} 244 245int 246vop_stdunlock(ap) 247 struct vop_unlock_args /* { 248 struct vnode *a_vp; 249 int a_flags; 250 struct proc *a_p; 251 } */ *ap; 252{ 253 struct vnode *vp = ap->a_vp; 254 255 return (lockmgr(&vp->v_lock, ap->a_flags | LK_RELEASE, &vp->v_interlock, 256 ap->a_p)); 257} 258 259int 260vop_stdislocked(ap) 261 struct vop_islocked_args /* { 262 struct vnode *a_vp; 263 struct proc *a_p; 264 } */ *ap; 265{ 266 267 return (lockstatus(&ap->a_vp->v_lock, ap->a_p)); 268} 269 270int 271vop_stdinactive(ap) 272 struct vop_inactive_args /* { 273 struct vnode *a_vp; 274 struct proc *a_p; 275 } */ *ap; 276{ 277 278 VOP_UNLOCK(ap->a_vp, 0, ap->a_p); 279 return (0); 280} 281 282/* 283 * Return true for select/poll. 284 */ 285int 286vop_nopoll(ap) 287 struct vop_poll_args /* { 288 struct vnode *a_vp; 289 int a_events; 290 struct ucred *a_cred; 291 struct proc *a_p; 292 } */ *ap; 293{ 294 /* 295 * Return true for read/write. If the user asked for something 296 * special, return POLLNVAL, so that clients have a way of 297 * determining reliably whether or not the extended 298 * functionality is present without hard-coding knowledge 299 * of specific filesystem implementations. 300 */ 301 if (ap->a_events & ~POLLSTANDARD) 302 return (POLLNVAL); 303 304 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 305} 306 307/* 308 * Implement poll for local filesystems that support it. 309 */ 310int 311vop_stdpoll(ap) 312 struct vop_poll_args /* { 313 struct vnode *a_vp; 314 int a_events; 315 struct ucred *a_cred; 316 struct proc *a_p; 317 } */ *ap; 318{ 319 if ((ap->a_events & ~POLLSTANDARD) == 0) 320 return (ap->a_events & (POLLRDNORM|POLLWRNORM)); 321 return (vn_pollrecord(ap->a_vp, ap->a_p, ap->a_events)); 322} 323 324int 325vop_stdbwrite(ap) 326 struct vop_bwrite_args *ap; 327{ 328 return (bwrite(ap->a_bp)); 329} 330 331/* 332 * Stubs to use when there is no locking to be done on the underlying object. 333 * A minimal shared lock is necessary to ensure that the underlying object 334 * is not revoked while an operation is in progress. So, an active shared 335 * count is maintained in an auxillary vnode lock structure. 336 */ 337int 338vop_sharedlock(ap) 339 struct vop_lock_args /* { 340 struct vnode *a_vp; 341 int a_flags; 342 struct proc *a_p; 343 } */ *ap; 344{ 345 /* 346 * This code cannot be used until all the non-locking filesystems 347 * (notably NFS) are converted to properly lock and release nodes. 348 * Also, certain vnode operations change the locking state within 349 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 350 * and symlink). Ideally these operations should not change the 351 * lock state, but should be changed to let the caller of the 352 * function unlock them. Otherwise all intermediate vnode layers 353 * (such as union, umapfs, etc) must catch these functions to do 354 * the necessary locking at their layer. Note that the inactive 355 * and lookup operations also change their lock state, but this 356 * cannot be avoided, so these two operations will always need 357 * to be handled in intermediate layers. 358 */ 359 struct vnode *vp = ap->a_vp; 360 int vnflags, flags = ap->a_flags; 361 362 switch (flags & LK_TYPE_MASK) { 363 case LK_DRAIN: 364 vnflags = LK_DRAIN; 365 break; 366 case LK_EXCLUSIVE: 367#ifdef DEBUG_VFS_LOCKS 368 /* 369 * Normally, we use shared locks here, but that confuses 370 * the locking assertions. 371 */ 372 vnflags = LK_EXCLUSIVE; 373 break; 374#endif 375 case LK_SHARED: 376 vnflags = LK_SHARED; 377 break; 378 case LK_UPGRADE: 379 case LK_EXCLUPGRADE: 380 case LK_DOWNGRADE: 381 return (0); 382 case LK_RELEASE: 383 default: 384 panic("vop_sharedlock: bad operation %d", flags & LK_TYPE_MASK); 385 } 386 if (flags & LK_INTERLOCK) 387 vnflags |= LK_INTERLOCK; 388#ifndef DEBUG_LOCKS 389 return (lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p)); 390#else 391 return (debuglockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p, 392 "vop_sharedlock", vp->filename, vp->line)); 393#endif 394} 395 396/* 397 * Stubs to use when there is no locking to be done on the underlying object. 398 * A minimal shared lock is necessary to ensure that the underlying object 399 * is not revoked while an operation is in progress. So, an active shared 400 * count is maintained in an auxillary vnode lock structure. 401 */ 402int 403vop_nolock(ap) 404 struct vop_lock_args /* { 405 struct vnode *a_vp; 406 int a_flags; 407 struct proc *a_p; 408 } */ *ap; 409{ 410#ifdef notyet 411 /* 412 * This code cannot be used until all the non-locking filesystems 413 * (notably NFS) are converted to properly lock and release nodes. 414 * Also, certain vnode operations change the locking state within 415 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 416 * and symlink). Ideally these operations should not change the 417 * lock state, but should be changed to let the caller of the 418 * function unlock them. Otherwise all intermediate vnode layers 419 * (such as union, umapfs, etc) must catch these functions to do 420 * the necessary locking at their layer. Note that the inactive 421 * and lookup operations also change their lock state, but this 422 * cannot be avoided, so these two operations will always need 423 * to be handled in intermediate layers. 424 */ 425 struct vnode *vp = ap->a_vp; 426 int vnflags, flags = ap->a_flags; 427 428 switch (flags & LK_TYPE_MASK) { 429 case LK_DRAIN: 430 vnflags = LK_DRAIN; 431 break; 432 case LK_EXCLUSIVE: 433 case LK_SHARED: 434 vnflags = LK_SHARED; 435 break; 436 case LK_UPGRADE: 437 case LK_EXCLUPGRADE: 438 case LK_DOWNGRADE: 439 return (0); 440 case LK_RELEASE: 441 default: 442 panic("vop_nolock: bad operation %d", flags & LK_TYPE_MASK); 443 } 444 if (flags & LK_INTERLOCK) 445 vnflags |= LK_INTERLOCK; 446 return(lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p)); 447#else /* for now */ 448 /* 449 * Since we are not using the lock manager, we must clear 450 * the interlock here. 451 */ 452 if (ap->a_flags & LK_INTERLOCK) 453 mtx_exit(&ap->a_vp->v_interlock, MTX_DEF); 454 return (0); 455#endif 456} 457 458/* 459 * Do the inverse of vop_nolock, handling the interlock in a compatible way. 460 */ 461int 462vop_nounlock(ap) 463 struct vop_unlock_args /* { 464 struct vnode *a_vp; 465 int a_flags; 466 struct proc *a_p; 467 } */ *ap; 468{ 469 470 /* 471 * Since we are not using the lock manager, we must clear 472 * the interlock here. 473 */ 474 if (ap->a_flags & LK_INTERLOCK) 475 mtx_exit(&ap->a_vp->v_interlock, MTX_DEF); 476 return (0); 477} 478 479/* 480 * Return whether or not the node is in use. 481 */ 482int 483vop_noislocked(ap) 484 struct vop_islocked_args /* { 485 struct vnode *a_vp; 486 struct proc *a_p; 487 } */ *ap; 488{ 489 490 return (0); 491} 492 493/* 494 * Return our mount point, as we will take charge of the writes. 495 */ 496int 497vop_stdgetwritemount(ap) 498 struct vop_getwritemount_args /* { 499 struct vnode *a_vp; 500 struct mount **a_mpp; 501 } */ *ap; 502{ 503 504 *(ap->a_mpp) = ap->a_vp->v_mount; 505 return (0); 506} 507 508int 509vop_stdcreatevobject(ap) 510 struct vop_createvobject_args /* { 511 struct vnode *vp; 512 struct ucred *cred; 513 struct proc *p; 514 } */ *ap; 515{ 516 struct vnode *vp = ap->a_vp; 517 struct ucred *cred = ap->a_cred; 518 struct proc *p = ap->a_p; 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 526retry: 527 if ((object = vp->v_object) == NULL) { 528 if (vp->v_type == VREG || vp->v_type == VDIR) { 529 if ((error = VOP_GETATTR(vp, &vat, cred, p)) != 0) 530 goto retn; 531 object = vnode_pager_alloc(vp, vat.va_size, 0, 0); 532 } else if (devsw(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, p); 551 tsleep(object, PVM, "vodead", 0); 552 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); 553 goto retry; 554 } 555 } 556 557 KASSERT(vp->v_object != NULL, ("vfs_object_create: NULL object")); 558 vp->v_flag |= VOBJBUF; 559 560retn: 561 return (error); 562} 563 564int 565vop_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 vm_object_terminate(obj); 584 } else { 585 /* 586 * Woe to the process that tries to page now :-). 587 */ 588 vm_pager_deallocate(obj); 589 } 590 return (0); 591} 592 593int 594vop_stdgetvobject(ap) 595 struct vop_getvobject_args /* { 596 struct vnode *vp; 597 struct vm_object **objpp; 598 } */ *ap; 599{ 600 struct vnode *vp = ap->a_vp; 601 struct vm_object **objpp = ap->a_objpp; 602 603 if (objpp) 604 *objpp = vp->v_object; 605 return (vp->v_object ? 0 : EINVAL); 606} 607 608/* 609 * vfs default ops 610 * used to fill the vfs fucntion table to get reasonable default return values. 611 */ 612int 613vfs_stdmount (mp, path, data, ndp, p) 614 struct mount *mp; 615 char *path; 616 caddr_t data; 617 struct nameidata *ndp; 618 struct proc *p; 619{ 620 return (0); 621} 622 623int 624vfs_stdunmount (mp, mntflags, p) 625 struct mount *mp; 626 int mntflags; 627 struct proc *p; 628{ 629 return (0); 630} 631 632int 633vfs_stdroot (mp, vpp) 634 struct mount *mp; 635 struct vnode **vpp; 636{ 637 return (EOPNOTSUPP); 638} 639 640int 641vfs_stdstatfs (mp, sbp, p) 642 struct mount *mp; 643 struct statfs *sbp; 644 struct proc *p; 645{ 646 return (EOPNOTSUPP); 647} 648 649int 650vfs_stdvptofh (vp, fhp) 651 struct vnode *vp; 652 struct fid *fhp; 653{ 654 return (EOPNOTSUPP); 655} 656 657int 658vfs_stdstart (mp, flags, p) 659 struct mount *mp; 660 int flags; 661 struct proc *p; 662{ 663 return (0); 664} 665 666int 667vfs_stdquotactl (mp, cmds, uid, arg, p) 668 struct mount *mp; 669 int cmds; 670 uid_t uid; 671 caddr_t arg; 672 struct proc *p; 673{ 674 return (EOPNOTSUPP); 675} 676 677int 678vfs_stdsync (mp, waitfor, cred, p) 679 struct mount *mp; 680 int waitfor; 681 struct ucred *cred; 682 struct proc *p; 683{ 684 return (0); 685} 686 687int 688vfs_stdvget (mp, ino, vpp) 689 struct mount *mp; 690 ino_t ino; 691 struct vnode **vpp; 692{ 693 return (EOPNOTSUPP); 694} 695 696int 697vfs_stdfhtovp (mp, fhp, vpp) 698 struct mount *mp; 699 struct fid *fhp; 700 struct vnode **vpp; 701{ 702 return (EOPNOTSUPP); 703} 704 705int 706vfs_stdcheckexp (mp, nam, extflagsp, credanonp) 707 struct mount *mp; 708 struct sockaddr *nam; 709 int *extflagsp; 710 struct ucred **credanonp; 711{ 712 return (EOPNOTSUPP); 713} 714 715int 716vfs_stdinit (vfsp) 717 struct vfsconf *vfsp; 718{ 719 return (0); 720} 721 722int 723vfs_stduninit (vfsp) 724 struct vfsconf *vfsp; 725{ 726 return(0); 727} 728 729int 730vfs_stdextattrctl(mp, cmd, attrname, arg, p) 731 struct mount *mp; 732 int cmd; 733 const char *attrname; 734 caddr_t arg; 735 struct proc *p; 736{ 737 return(EOPNOTSUPP); 738} 739 740/* end of vfs default ops */ 741