vfs_default.c revision 76131
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 76131 2001-04-29 11:48:41Z phk $ 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_bmap_desc, (vop_t *) vop_stdbmap }, 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 105static struct vnodeopv_desc default_vnodeop_opv_desc = 106 { &default_vnodeop_p, default_vnodeop_entries }; 107 108VNODEOP_SET(default_vnodeop_opv_desc); 109 110int 111vop_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 120int 121vop_ebadf(struct vop_generic_args *ap) 122{ 123 124 return (EBADF); 125} 126 127int 128vop_enotty(struct vop_generic_args *ap) 129{ 130 131 return (ENOTTY); 132} 133 134int 135vop_einval(struct vop_generic_args *ap) 136{ 137 138 return (EINVAL); 139} 140 141int 142vop_null(struct vop_generic_args *ap) 143{ 144 145 return (0); 146} 147 148int 149vop_defaultop(struct vop_generic_args *ap) 150{ 151 152 return (VOCALL(default_vnodeop_p, ap->a_desc->vdesc_offset, ap)); 153} 154 155int 156vop_panic(struct vop_generic_args *ap) 157{ 158 159 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 160} 161 162static int 163vop_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 187static int 188vop_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 199int 200vop_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 */ 239int 240vop_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 257int 258vop_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 271int 272vop_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 282int 283vop_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 */ 297int 298vop_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 */ 322int 323vop_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/* 337 * Stubs to use when there is no locking to be done on the underlying object. 338 * A minimal shared lock is necessary to ensure that the underlying object 339 * is not revoked while an operation is in progress. So, an active shared 340 * count is maintained in an auxillary vnode lock structure. 341 */ 342int 343vop_sharedlock(ap) 344 struct vop_lock_args /* { 345 struct vnode *a_vp; 346 int a_flags; 347 struct proc *a_p; 348 } */ *ap; 349{ 350 /* 351 * This code cannot be used until all the non-locking filesystems 352 * (notably NFS) are converted to properly lock and release nodes. 353 * Also, certain vnode operations change the locking state within 354 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 355 * and symlink). Ideally these operations should not change the 356 * lock state, but should be changed to let the caller of the 357 * function unlock them. Otherwise all intermediate vnode layers 358 * (such as union, umapfs, etc) must catch these functions to do 359 * the necessary locking at their layer. Note that the inactive 360 * and lookup operations also change their lock state, but this 361 * cannot be avoided, so these two operations will always need 362 * to be handled in intermediate layers. 363 */ 364 struct vnode *vp = ap->a_vp; 365 int vnflags, flags = ap->a_flags; 366 367 switch (flags & LK_TYPE_MASK) { 368 case LK_DRAIN: 369 vnflags = LK_DRAIN; 370 break; 371 case LK_EXCLUSIVE: 372#ifdef DEBUG_VFS_LOCKS 373 /* 374 * Normally, we use shared locks here, but that confuses 375 * the locking assertions. 376 */ 377 vnflags = LK_EXCLUSIVE; 378 break; 379#endif 380 case LK_SHARED: 381 vnflags = LK_SHARED; 382 break; 383 case LK_UPGRADE: 384 case LK_EXCLUPGRADE: 385 case LK_DOWNGRADE: 386 return (0); 387 case LK_RELEASE: 388 default: 389 panic("vop_sharedlock: bad operation %d", flags & LK_TYPE_MASK); 390 } 391 if (flags & LK_INTERLOCK) 392 vnflags |= LK_INTERLOCK; 393#ifndef DEBUG_LOCKS 394 return (lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p)); 395#else 396 return (debuglockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p, 397 "vop_sharedlock", vp->filename, vp->line)); 398#endif 399} 400 401/* 402 * Stubs to use when there is no locking to be done on the underlying object. 403 * A minimal shared lock is necessary to ensure that the underlying object 404 * is not revoked while an operation is in progress. So, an active shared 405 * count is maintained in an auxillary vnode lock structure. 406 */ 407int 408vop_nolock(ap) 409 struct vop_lock_args /* { 410 struct vnode *a_vp; 411 int a_flags; 412 struct proc *a_p; 413 } */ *ap; 414{ 415#ifdef notyet 416 /* 417 * This code cannot be used until all the non-locking filesystems 418 * (notably NFS) are converted to properly lock and release nodes. 419 * Also, certain vnode operations change the locking state within 420 * the operation (create, mknod, remove, link, rename, mkdir, rmdir, 421 * and symlink). Ideally these operations should not change the 422 * lock state, but should be changed to let the caller of the 423 * function unlock them. Otherwise all intermediate vnode layers 424 * (such as union, umapfs, etc) must catch these functions to do 425 * the necessary locking at their layer. Note that the inactive 426 * and lookup operations also change their lock state, but this 427 * cannot be avoided, so these two operations will always need 428 * to be handled in intermediate layers. 429 */ 430 struct vnode *vp = ap->a_vp; 431 int vnflags, flags = ap->a_flags; 432 433 switch (flags & LK_TYPE_MASK) { 434 case LK_DRAIN: 435 vnflags = LK_DRAIN; 436 break; 437 case LK_EXCLUSIVE: 438 case LK_SHARED: 439 vnflags = LK_SHARED; 440 break; 441 case LK_UPGRADE: 442 case LK_EXCLUPGRADE: 443 case LK_DOWNGRADE: 444 return (0); 445 case LK_RELEASE: 446 default: 447 panic("vop_nolock: bad operation %d", flags & LK_TYPE_MASK); 448 } 449 if (flags & LK_INTERLOCK) 450 vnflags |= LK_INTERLOCK; 451 return(lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_p)); 452#else /* for now */ 453 /* 454 * Since we are not using the lock manager, we must clear 455 * the interlock here. 456 */ 457 if (ap->a_flags & LK_INTERLOCK) 458 mtx_unlock(&ap->a_vp->v_interlock); 459 return (0); 460#endif 461} 462 463/* 464 * Do the inverse of vop_nolock, handling the interlock in a compatible way. 465 */ 466int 467vop_nounlock(ap) 468 struct vop_unlock_args /* { 469 struct vnode *a_vp; 470 int a_flags; 471 struct proc *a_p; 472 } */ *ap; 473{ 474 475 /* 476 * Since we are not using the lock manager, we must clear 477 * the interlock here. 478 */ 479 if (ap->a_flags & LK_INTERLOCK) 480 mtx_unlock(&ap->a_vp->v_interlock); 481 return (0); 482} 483 484/* 485 * Return whether or not the node is in use. 486 */ 487int 488vop_noislocked(ap) 489 struct vop_islocked_args /* { 490 struct vnode *a_vp; 491 struct proc *a_p; 492 } */ *ap; 493{ 494 495 return (0); 496} 497 498/* 499 * Return our mount point, as we will take charge of the writes. 500 */ 501int 502vop_stdgetwritemount(ap) 503 struct vop_getwritemount_args /* { 504 struct vnode *a_vp; 505 struct mount **a_mpp; 506 } */ *ap; 507{ 508 509 *(ap->a_mpp) = ap->a_vp->v_mount; 510 return (0); 511} 512 513int 514vop_stdcreatevobject(ap) 515 struct vop_createvobject_args /* { 516 struct vnode *vp; 517 struct ucred *cred; 518 struct proc *p; 519 } */ *ap; 520{ 521 struct vnode *vp = ap->a_vp; 522 struct ucred *cred = ap->a_cred; 523 struct proc *p = ap->a_p; 524 struct vattr vat; 525 vm_object_t object; 526 int error = 0; 527 528 if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE) 529 return (0); 530 531retry: 532 if ((object = vp->v_object) == NULL) { 533 if (vp->v_type == VREG || vp->v_type == VDIR) { 534 if ((error = VOP_GETATTR(vp, &vat, cred, p)) != 0) 535 goto retn; 536 object = vnode_pager_alloc(vp, vat.va_size, 0, 0); 537 } else if (devsw(vp->v_rdev) != NULL) { 538 /* 539 * This simply allocates the biggest object possible 540 * for a disk vnode. This should be fixed, but doesn't 541 * cause any problems (yet). 542 */ 543 object = vnode_pager_alloc(vp, IDX_TO_OFF(INT_MAX), 0, 0); 544 } else { 545 goto retn; 546 } 547 /* 548 * Dereference the reference we just created. This assumes 549 * that the object is associated with the vp. 550 */ 551 object->ref_count--; 552 vp->v_usecount--; 553 } else { 554 if (object->flags & OBJ_DEAD) { 555 VOP_UNLOCK(vp, 0, p); 556 tsleep(object, PVM, "vodead", 0); 557 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); 558 goto retry; 559 } 560 } 561 562 KASSERT(vp->v_object != NULL, ("vfs_object_create: NULL object")); 563 vp->v_flag |= VOBJBUF; 564 565retn: 566 return (error); 567} 568 569int 570vop_stddestroyvobject(ap) 571 struct vop_destroyvobject_args /* { 572 struct vnode *vp; 573 } */ *ap; 574{ 575 struct vnode *vp = ap->a_vp; 576 vm_object_t obj = vp->v_object; 577 578 if (vp->v_object == NULL) 579 return (0); 580 581 if (obj->ref_count == 0) { 582 /* 583 * vclean() may be called twice. The first time 584 * removes the primary reference to the object, 585 * the second time goes one further and is a 586 * special-case to terminate the object. 587 */ 588 vm_object_terminate(obj); 589 } else { 590 /* 591 * Woe to the process that tries to page now :-). 592 */ 593 vm_pager_deallocate(obj); 594 } 595 return (0); 596} 597 598int 599vop_stdgetvobject(ap) 600 struct vop_getvobject_args /* { 601 struct vnode *vp; 602 struct vm_object **objpp; 603 } */ *ap; 604{ 605 struct vnode *vp = ap->a_vp; 606 struct vm_object **objpp = ap->a_objpp; 607 608 if (objpp) 609 *objpp = vp->v_object; 610 return (vp->v_object ? 0 : EINVAL); 611} 612 613int 614vop_stdbmap(ap) 615 struct vop_bmap_args /* { 616 struct vnode *a_vp; 617 daddr_t a_bn; 618 struct vnode **a_vpp; 619 daddr_t *a_bnp; 620 int *a_runp; 621 int *a_runb; 622 } */ *ap; 623{ 624 625 if (ap->a_vpp != NULL) 626 *ap->a_vpp = ap->a_vp; 627 if (ap->a_bnp != NULL) 628 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize); 629 if (ap->a_runp != NULL) 630 *ap->a_runp = 0; 631 if (ap->a_runb != NULL) 632 *ap->a_runb = 0; 633 return (0); 634} 635 636 637/* 638 * vfs default ops 639 * used to fill the vfs fucntion table to get reasonable default return values. 640 */ 641int 642vfs_stdmount (mp, path, data, ndp, p) 643 struct mount *mp; 644 char *path; 645 caddr_t data; 646 struct nameidata *ndp; 647 struct proc *p; 648{ 649 return (0); 650} 651 652int 653vfs_stdunmount (mp, mntflags, p) 654 struct mount *mp; 655 int mntflags; 656 struct proc *p; 657{ 658 return (0); 659} 660 661int 662vfs_stdroot (mp, vpp) 663 struct mount *mp; 664 struct vnode **vpp; 665{ 666 return (EOPNOTSUPP); 667} 668 669int 670vfs_stdstatfs (mp, sbp, p) 671 struct mount *mp; 672 struct statfs *sbp; 673 struct proc *p; 674{ 675 return (EOPNOTSUPP); 676} 677 678int 679vfs_stdvptofh (vp, fhp) 680 struct vnode *vp; 681 struct fid *fhp; 682{ 683 return (EOPNOTSUPP); 684} 685 686int 687vfs_stdstart (mp, flags, p) 688 struct mount *mp; 689 int flags; 690 struct proc *p; 691{ 692 return (0); 693} 694 695int 696vfs_stdquotactl (mp, cmds, uid, arg, p) 697 struct mount *mp; 698 int cmds; 699 uid_t uid; 700 caddr_t arg; 701 struct proc *p; 702{ 703 return (EOPNOTSUPP); 704} 705 706int 707vfs_stdsync (mp, waitfor, cred, p) 708 struct mount *mp; 709 int waitfor; 710 struct ucred *cred; 711 struct proc *p; 712{ 713 return (0); 714} 715 716int 717vfs_stdvget (mp, ino, vpp) 718 struct mount *mp; 719 ino_t ino; 720 struct vnode **vpp; 721{ 722 return (EOPNOTSUPP); 723} 724 725int 726vfs_stdfhtovp (mp, fhp, vpp) 727 struct mount *mp; 728 struct fid *fhp; 729 struct vnode **vpp; 730{ 731 return (EOPNOTSUPP); 732} 733 734int 735vfs_stdinit (vfsp) 736 struct vfsconf *vfsp; 737{ 738 return (0); 739} 740 741int 742vfs_stduninit (vfsp) 743 struct vfsconf *vfsp; 744{ 745 return(0); 746} 747 748int 749vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, p) 750 struct mount *mp; 751 int cmd; 752 struct vnode *filename_vp; 753 int attrnamespace; 754 const char *attrname; 755 struct proc *p; 756{ 757 return(EOPNOTSUPP); 758} 759 760/* end of vfs default ops */ 761