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 1.20 2004/10/07 10:03:02 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/namei.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_page.h> 61 #include <vm/vm_pager.h> 62 #include <vm/vnode_pager.h> 63 64 static int vop_nolookup (struct vop_lookup_args *); 65 static int vop_nostrategy (struct vop_strategy_args *); 66 67 /* 68 * This vnode table stores what we want to do if the filesystem doesn't 69 * implement a particular VOP. 70 * 71 * If there is no specific entry here, we will return EOPNOTSUPP. 72 */ 73 struct vop_ops *default_vnode_vops; 74 static struct vnodeopv_entry_desc default_vnodeop_entries[] = { 75 { &vop_default_desc, vop_eopnotsupp }, 76 { &vop_advlock_desc, vop_einval }, 77 { &vop_bwrite_desc, (void *) vop_stdbwrite }, 78 { &vop_close_desc, vop_null }, 79 { &vop_createvobject_desc, (void *) vop_stdcreatevobject }, 80 { &vop_destroyvobject_desc, (void *) vop_stddestroyvobject }, 81 { &vop_fsync_desc, vop_null }, 82 { &vop_getvobject_desc, (void *) vop_stdgetvobject }, 83 { &vop_ioctl_desc, vop_enotty }, 84 { &vop_islocked_desc, (void *) vop_stdislocked }, 85 { &vop_lease_desc, vop_null }, 86 { &vop_lock_desc, (void *) vop_stdlock }, 87 { &vop_mmap_desc, vop_einval }, 88 { &vop_resolve_desc, (void *) vop_noresolve }, 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_stdunlock }, 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 return (EOPNOTSUPP); 115 } 116 117 int 118 vop_ebadf(struct vop_generic_args *ap) 119 { 120 return (EBADF); 121 } 122 123 int 124 vop_enotty(struct vop_generic_args *ap) 125 { 126 return (ENOTTY); 127 } 128 129 int 130 vop_einval(struct vop_generic_args *ap) 131 { 132 return (EINVAL); 133 } 134 135 int 136 vop_null(struct vop_generic_args *ap) 137 { 138 return (0); 139 } 140 141 int 142 vop_defaultop(struct vop_generic_args *ap) 143 { 144 return (VOCALL(default_vnode_vops, ap)); 145 } 146 147 int 148 vop_panic(struct vop_generic_args *ap) 149 { 150 151 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name); 152 } 153 154 /* 155 * vop_noresolve { struct namecache *a_ncp } XXX STOPGAP FUNCTION 156 * 157 * XXX OLD API ROUTINE! WHEN ALL VFSs HAVE BEEN CLEANED UP THIS PROCEDURE 158 * WILL BE REMOVED. This procedure exists for all VFSs which have not 159 * yet implemented vop_resolve(). It converts vop_resolve() into a 160 * vop_lookup() and does appropriate translations. 161 * 162 * Resolve a ncp for VFSs which do not support the VOP. Eventually all 163 * VFSs will support this VOP and this routine can be removed, since 164 * vop_resolve() is far less complex then the older LOOKUP/CACHEDLOOKUP 165 * API. 166 * 167 * A locked ncp is passed in to be resolved. The NCP is resolved by 168 * figuring out the vnode (if any) and calling cache_setvp() to attach the 169 * vnode to the entry. If the entry represents a non-existant node then 170 * cache_setvp() is called with a NULL vnode to resolve the entry into a 171 * negative cache entry. No vnode locks are retained and the 172 * ncp is left locked on return. 173 * 174 * There is a potential directory and vnode interlock. The lock order 175 * requirement is: namecache, governing directory, resolved vnode. 176 */ 177 int 178 vop_noresolve(struct vop_resolve_args *ap) 179 { 180 int error; 181 struct vnode *dvp; 182 struct vnode *vp; 183 struct namecache *ncp; 184 struct componentname cnp; 185 186 ncp = ap->a_ncp; /* locked namecache node */ 187 if (ncp->nc_flag & NCF_MOUNTPT) /* can't cross a mount point! */ 188 return(EPERM); 189 if (ncp->nc_parent == NULL) 190 return(EPERM); 191 if ((dvp = ncp->nc_parent->nc_vp) == NULL) 192 return(EPERM); 193 194 if ((error = vget(dvp, NULL, LK_EXCLUSIVE, curthread)) != 0) { 195 printf("[diagnostic] vop_noresolve: EAGAIN on ncp %p %*.*s\n", 196 ncp, ncp->nc_nlen, ncp->nc_nlen, ncp->nc_name); 197 return(EAGAIN); 198 } 199 200 bzero(&cnp, sizeof(cnp)); 201 cnp.cn_nameiop = NAMEI_LOOKUP; 202 cnp.cn_flags = CNP_ISLASTCN; 203 cnp.cn_nameptr = ncp->nc_name; 204 cnp.cn_namelen = ncp->nc_nlen; 205 cnp.cn_cred = ap->a_cred; 206 cnp.cn_td = curthread; /* XXX */ 207 208 /* 209 * vop_lookup() always returns vp locked. dvp may or may not be 210 * left locked depending on CNP_PDIRUNLOCK. 211 */ 212 error = vop_lookup(ap->a_head.a_ops, dvp, &vp, &cnp); 213 if (error == 0) 214 VOP_UNLOCK(vp, NULL, 0, curthread); 215 if ((cnp.cn_flags & CNP_PDIRUNLOCK) == 0) 216 VOP_UNLOCK(dvp, NULL, 0, curthread); 217 if ((ncp->nc_flag & NCF_UNRESOLVED) == 0) { 218 /* was resolved by another process while we were unlocked */ 219 if (error == 0) 220 vrele(vp); 221 } else if (error == 0) { 222 KKASSERT(vp != NULL); 223 cache_setvp(ncp, vp); 224 vrele(vp); 225 } else if (error == ENOENT) { 226 KKASSERT(vp == NULL); 227 if (cnp.cn_flags & CNP_ISWHITEOUT) 228 ncp->nc_flag |= NCF_WHITEOUT; 229 cache_setvp(ncp, NULL); 230 } 231 vrele(dvp); 232 return (error); 233 } 234 235 #if 0 236 237 /* 238 * vop_noremove { struct namecache *a_ncp } XXX STOPGAP FUNCTION 239 * 240 * Remove the file/dir represented by a_ncp. 241 * 242 * XXX ultra difficult. A number of existing filesystems, including UFS, 243 * assume that the directory will remain locked and the lookup will 244 * store the directory offset and other things in the directory inode 245 * for the later VOP_REMOVE to use. We have to move all that 246 * functionality into e.g. UFS's VOP_REMOVE itself. 247 */ 248 static int 249 vop_nonremove(struct vop_nremove_args *ap) 250 { 251 struct namecache *ncfile; 252 struct namecache *ncdir; 253 struct componentname cnd; 254 struct vnode *vp; 255 struct vnode *vpd; 256 thread_t td; 257 int error; 258 259 td = curthread; 260 ncfile = ap->a_ncp; 261 ncdir = ncfile->nc_parent; 262 263 if ((error = cache_vget(ncdir, ap->a_cred, LK_EXCLUSIVE, &vpd)) != 0) 264 return (error); 265 if ((error = cache_vget(ncfile, ap->a_cred, LK_EXCLUSIVE, &vp)) != 0) { 266 vput(vpd); 267 return (error); 268 } 269 bzero(&cnd, sizeof(cnd)); 270 cnd.cn_nameiop = NAMEI_DELETE; 271 cnd.cn_td = td; 272 cnd.cn_cred = ap->a_cred; 273 cnd.cn_nameptr = ncfile->nc_name; 274 cnd.cn_namelen = ncfile->nc_nlen; 275 error = VOP_REMOVE(vpd, NCPNULL, vp, &cnd); 276 if (error == 0) 277 cache_purge(vp); 278 vput(vp); 279 vput(vpd); 280 281 /* 282 * Re-resolve the ncp to match the fact that the file has been 283 * deleted from the namespace. If an error occured leave the ncp 284 * unresolved (meaning that we have no idea what the correct state 285 * is). 286 */ 287 if (error == 0) { 288 cache_setunresolved(ncfile); 289 cache_setvp(ncfile, NULL); 290 } 291 return (error); 292 } 293 294 #endif 295 296 297 static int 298 vop_nolookup(ap) 299 struct vop_lookup_args /* { 300 struct vnode *a_dvp; 301 struct vnode **a_vpp; 302 struct componentname *a_cnp; 303 } */ *ap; 304 { 305 306 *ap->a_vpp = NULL; 307 return (ENOTDIR); 308 } 309 310 /* 311 * vop_nostrategy: 312 * 313 * Strategy routine for VFS devices that have none. 314 * 315 * B_ERROR and B_INVAL must be cleared prior to calling any strategy 316 * routine. Typically this is done for a B_READ strategy call. Typically 317 * B_INVAL is assumed to already be clear prior to a write and should not 318 * be cleared manually unless you just made the buffer invalid. B_ERROR 319 * should be cleared either way. 320 */ 321 322 static int 323 vop_nostrategy (struct vop_strategy_args *ap) 324 { 325 printf("No strategy for buffer at %p\n", ap->a_bp); 326 vprint("", ap->a_vp); 327 vprint("", ap->a_bp->b_vp); 328 ap->a_bp->b_flags |= B_ERROR; 329 ap->a_bp->b_error = EOPNOTSUPP; 330 biodone(ap->a_bp); 331 return (EOPNOTSUPP); 332 } 333 334 int 335 vop_stdpathconf(ap) 336 struct vop_pathconf_args /* { 337 struct vnode *a_vp; 338 int a_name; 339 int *a_retval; 340 } */ *ap; 341 { 342 343 switch (ap->a_name) { 344 case _PC_LINK_MAX: 345 *ap->a_retval = LINK_MAX; 346 return (0); 347 case _PC_MAX_CANON: 348 *ap->a_retval = MAX_CANON; 349 return (0); 350 case _PC_MAX_INPUT: 351 *ap->a_retval = MAX_INPUT; 352 return (0); 353 case _PC_PIPE_BUF: 354 *ap->a_retval = PIPE_BUF; 355 return (0); 356 case _PC_CHOWN_RESTRICTED: 357 *ap->a_retval = 1; 358 return (0); 359 case _PC_VDISABLE: 360 *ap->a_retval = _POSIX_VDISABLE; 361 return (0); 362 default: 363 return (EINVAL); 364 } 365 /* NOTREACHED */ 366 } 367 368 /* 369 * Standard lock. The lock is recursive-capable only if the lock was 370 * initialized with LK_CANRECURSE or that flag is passed in a_flags. 371 */ 372 int 373 vop_stdlock(ap) 374 struct vop_lock_args /* { 375 struct vnode *a_vp; 376 lwkt_tokref_t a_vlock; 377 int a_flags; 378 struct proc *a_p; 379 } */ *ap; 380 { 381 int error; 382 383 #ifndef DEBUG_LOCKS 384 error = lockmgr(&ap->a_vp->v_lock, ap->a_flags, 385 ap->a_vlock, ap->a_td); 386 #else 387 error = debuglockmgr(&ap->a_vp->v_lock, ap->a_flags, 388 ap->a_vlock, ap->a_td, 389 "vop_stdlock", ap->a_vp->filename, ap->a_vp->line); 390 #endif 391 return(error); 392 } 393 394 int 395 vop_stdunlock(ap) 396 struct vop_unlock_args /* { 397 struct vnode *a_vp; 398 lwkt_tokref_t a_vlock; 399 int a_flags; 400 struct thread *a_td; 401 } */ *ap; 402 { 403 int error; 404 405 error = lockmgr(&ap->a_vp->v_lock, ap->a_flags | LK_RELEASE, 406 ap->a_vlock, ap->a_td); 407 return(error); 408 } 409 410 int 411 vop_stdislocked(ap) 412 struct vop_islocked_args /* { 413 struct vnode *a_vp; 414 struct thread *a_td; 415 } */ *ap; 416 { 417 return (lockstatus(&ap->a_vp->v_lock, ap->a_td)); 418 } 419 420 /* 421 * Return true for select/poll. 422 */ 423 int 424 vop_nopoll(ap) 425 struct vop_poll_args /* { 426 struct vnode *a_vp; 427 int a_events; 428 struct ucred *a_cred; 429 struct proc *a_p; 430 } */ *ap; 431 { 432 /* 433 * Return true for read/write. If the user asked for something 434 * special, return POLLNVAL, so that clients have a way of 435 * determining reliably whether or not the extended 436 * functionality is present without hard-coding knowledge 437 * of specific filesystem implementations. 438 */ 439 if (ap->a_events & ~POLLSTANDARD) 440 return (POLLNVAL); 441 442 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 443 } 444 445 /* 446 * Implement poll for local filesystems that support it. 447 */ 448 int 449 vop_stdpoll(ap) 450 struct vop_poll_args /* { 451 struct vnode *a_vp; 452 int a_events; 453 struct ucred *a_cred; 454 struct thread *a_td; 455 } */ *ap; 456 { 457 if (ap->a_events & ~POLLSTANDARD) 458 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events)); 459 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 460 } 461 462 int 463 vop_stdbwrite(ap) 464 struct vop_bwrite_args *ap; 465 { 466 return (bwrite(ap->a_bp)); 467 } 468 469 int 470 vop_stdcreatevobject(ap) 471 struct vop_createvobject_args /* { 472 struct vnode *a_vp; 473 struct proc *a_td; 474 } */ *ap; 475 { 476 struct vnode *vp = ap->a_vp; 477 struct thread *td = ap->a_td; 478 struct vattr vat; 479 vm_object_t object; 480 int error = 0; 481 482 if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE) 483 return (0); 484 485 retry: 486 if ((object = vp->v_object) == NULL) { 487 if (vp->v_type == VREG || vp->v_type == VDIR) { 488 if ((error = VOP_GETATTR(vp, &vat, td)) != 0) 489 goto retn; 490 object = vnode_pager_alloc(vp, vat.va_size, 0, 0); 491 } else if (vp->v_rdev && dev_is_good(vp->v_rdev)) { 492 /* 493 * XXX v_rdev uses NULL/non-NULL instead of NODEV 494 * 495 * This simply allocates the biggest object possible 496 * for a disk vnode. This should be fixed, but doesn't 497 * cause any problems (yet). 498 */ 499 object = vnode_pager_alloc(vp, IDX_TO_OFF(INT_MAX), 0, 0); 500 } else { 501 goto retn; 502 } 503 /* 504 * Dereference the reference we just created. This assumes 505 * that the object is associated with the vp. 506 */ 507 object->ref_count--; 508 vp->v_usecount--; 509 } else { 510 if (object->flags & OBJ_DEAD) { 511 VOP_UNLOCK(vp, NULL, 0, td); 512 tsleep(object, 0, "vodead", 0); 513 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 514 goto retry; 515 } 516 } 517 518 KASSERT(vp->v_object != NULL, ("vfs_object_create: NULL object")); 519 vp->v_flag |= VOBJBUF; 520 521 retn: 522 return (error); 523 } 524 525 int 526 vop_stddestroyvobject(ap) 527 struct vop_destroyvobject_args /* { 528 struct vnode *vp; 529 } */ *ap; 530 { 531 struct vnode *vp = ap->a_vp; 532 vm_object_t obj = vp->v_object; 533 534 if (vp->v_object == NULL) 535 return (0); 536 537 if (obj->ref_count == 0) { 538 /* 539 * vclean() may be called twice. The first time 540 * removes the primary reference to the object, 541 * the second time goes one further and is a 542 * special-case to terminate the object. 543 * 544 * don't double-terminate the object. 545 */ 546 if ((obj->flags & OBJ_DEAD) == 0) 547 vm_object_terminate(obj); 548 } else { 549 /* 550 * Woe to the process that tries to page now :-). 551 */ 552 vm_pager_deallocate(obj); 553 } 554 return (0); 555 } 556 557 /* 558 * Return the underlying VM object. This routine may be called with or 559 * without the vnode interlock held. If called without, the returned 560 * object is not guarenteed to be valid. The syncer typically gets the 561 * object without holding the interlock in order to quickly test whether 562 * it might be dirty before going heavy-weight. vm_object's use zalloc 563 * and thus stable-storage, so this is safe. 564 */ 565 int 566 vop_stdgetvobject(ap) 567 struct vop_getvobject_args /* { 568 struct vnode *vp; 569 struct vm_object **objpp; 570 } */ *ap; 571 { 572 struct vnode *vp = ap->a_vp; 573 struct vm_object **objpp = ap->a_objpp; 574 575 if (objpp) 576 *objpp = vp->v_object; 577 return (vp->v_object ? 0 : EINVAL); 578 } 579 580 /* 581 * vfs default ops 582 * used to fill the vfs fucntion table to get reasonable default return values. 583 */ 584 int 585 vfs_stdmount(struct mount *mp, char *path, caddr_t data, 586 struct nameidata *ndp, struct thread *td) 587 { 588 return (0); 589 } 590 591 int 592 vfs_stdunmount(struct mount *mp, int mntflags, struct thread *td) 593 { 594 return (0); 595 } 596 597 int 598 vfs_stdroot(struct mount *mp, struct vnode **vpp) 599 { 600 return (EOPNOTSUPP); 601 } 602 603 int 604 vfs_stdstatfs(struct mount *mp, struct statfs *sbp, struct thread *td) 605 { 606 return (EOPNOTSUPP); 607 } 608 609 int 610 vfs_stdvptofh(struct vnode *vp, struct fid *fhp) 611 { 612 return (EOPNOTSUPP); 613 } 614 615 int 616 vfs_stdstart(struct mount *mp, int flags, struct thread *td) 617 { 618 return (0); 619 } 620 621 int 622 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, 623 caddr_t arg, struct thread *td) 624 { 625 return (EOPNOTSUPP); 626 } 627 628 int 629 vfs_stdsync(struct mount *mp, int waitfor, struct thread *td) 630 { 631 return (0); 632 } 633 634 int 635 vfs_stdvget(struct mount *mp, ino_t ino, struct vnode **vpp) 636 { 637 return (EOPNOTSUPP); 638 } 639 640 int 641 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp) 642 { 643 return (EOPNOTSUPP); 644 } 645 646 int 647 vfs_stdcheckexp(struct mount *mp, struct sockaddr *nam, int *extflagsp, 648 struct ucred **credanonp) 649 { 650 return (EOPNOTSUPP); 651 } 652 653 int 654 vfs_stdinit(struct vfsconf *vfsp) 655 { 656 return (0); 657 } 658 659 int 660 vfs_stduninit(struct vfsconf *vfsp) 661 { 662 return(0); 663 } 664 665 int 666 vfs_stdextattrctl(struct mount *mp, int cmd, const char *attrname, 667 caddr_t arg, struct thread *td) 668 { 669 return(EOPNOTSUPP); 670 } 671 672 /* end of vfs default ops */ 673