1 /* $OpenBSD: vfs_subr.c,v 1.65 2001/08/02 08:16:45 assar Exp $ */ 2 /* $NetBSD: vfs_subr.c,v 1.53 1996/04/22 01:39:13 christos Exp $ */ 3 4 /* 5 * Copyright (c) 1989, 1993 6 * The Regents of the University of California. All rights reserved. 7 * (c) UNIX System Laboratories, Inc. 8 * All or some portions of this file are derived from material licensed 9 * to the University of California by American Telephone and Telegraph 10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 11 * the permission of UNIX System Laboratories, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. All advertising materials mentioning features or use of this software 22 * must display the following acknowledgement: 23 * This product includes software developed by the University of 24 * California, Berkeley and its contributors. 25 * 4. Neither the name of the University nor the names of its contributors 26 * may be used to endorse or promote products derived from this software 27 * without specific prior written permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 39 * SUCH DAMAGE. 40 * 41 * @(#)vfs_subr.c 8.13 (Berkeley) 4/18/94 42 */ 43 44 /* 45 * External virtual filesystem routines 46 */ 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/proc.h> 51 #include <sys/mount.h> 52 #include <sys/time.h> 53 #include <sys/fcntl.h> 54 #include <sys/kernel.h> 55 #include <sys/vnode.h> 56 #include <sys/stat.h> 57 #include <sys/namei.h> 58 #include <sys/ucred.h> 59 #include <sys/buf.h> 60 #include <sys/errno.h> 61 #include <sys/malloc.h> 62 #include <sys/domain.h> 63 #include <sys/mbuf.h> 64 #include <sys/syscallargs.h> 65 #include <sys/pool.h> 66 67 #include <vm/vm.h> 68 #include <sys/sysctl.h> 69 70 #include <miscfs/specfs/specdev.h> 71 72 #include <uvm/uvm_extern.h> 73 74 enum vtype iftovt_tab[16] = { 75 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, 76 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, 77 }; 78 int vttoif_tab[9] = { 79 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, 80 S_IFSOCK, S_IFIFO, S_IFMT, 81 }; 82 83 int doforce = 1; /* 1 => permit forcible unmounting */ 84 int prtactive = 0; /* 1 => print out reclaim of active vnodes */ 85 int suid_clear = 1; /* 1 => clear SUID / SGID on owner change */ 86 87 /* 88 * Insq/Remq for the vnode usage lists. 89 */ 90 #define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) 91 #define bufremvn(bp) { \ 92 LIST_REMOVE(bp, b_vnbufs); \ 93 (bp)->b_vnbufs.le_next = NOLIST; \ 94 } 95 96 struct freelst vnode_hold_list; /* list of vnodes referencing buffers */ 97 struct freelst vnode_free_list; /* vnode free list */ 98 99 struct mntlist mountlist; /* mounted filesystem list */ 100 struct simplelock mountlist_slock; 101 static struct simplelock mntid_slock; 102 struct simplelock mntvnode_slock; 103 struct simplelock vnode_free_list_slock; 104 struct simplelock spechash_slock; 105 106 void vclean __P((struct vnode *, int, struct proc *)); 107 108 void insmntque __P((struct vnode *, struct mount *)); 109 int getdevvp __P((dev_t, struct vnode **, enum vtype)); 110 111 int vfs_hang_addrlist __P((struct mount *, struct netexport *, 112 struct export_args *)); 113 int vfs_free_netcred __P((struct radix_node *, void *)); 114 void vfs_free_addrlist __P((struct netexport *)); 115 static __inline__ void vputonfreelist __P((struct vnode *)); 116 117 #ifdef DEBUG 118 void printlockedvnodes __P((void)); 119 #endif 120 121 #define VN_KNOTE(vp, b) \ 122 KNOTE((struct klist *)&vp->v_selectinfo.vsi_selinfo.si_note, (b)) 123 124 struct pool vnode_pool; 125 126 /* 127 * Initialize the vnode management data structures. 128 */ 129 void 130 vntblinit() 131 { 132 133 pool_init(&vnode_pool, sizeof(struct vnode), 0, 0, 0, "vnodes", 134 0, pool_page_alloc_nointr, pool_page_free_nointr, M_VNODE); 135 simple_lock_init(&mntvnode_slock); 136 simple_lock_init(&mntid_slock); 137 simple_lock_init(&spechash_slock); 138 TAILQ_INIT(&vnode_hold_list); 139 TAILQ_INIT(&vnode_free_list); 140 simple_lock_init(&vnode_free_list_slock); 141 CIRCLEQ_INIT(&mountlist); 142 simple_lock_init(&mountlist_slock); 143 /* 144 * Initialize the filesystem syncer. 145 */ 146 vn_initialize_syncerd(); 147 } 148 149 150 /* 151 * Mark a mount point as busy. Used to synchronize access and to delay 152 * unmounting. Interlock is not released on failure. 153 */ 154 155 int 156 vfs_busy(mp, flags, interlkp, p) 157 struct mount *mp; 158 int flags; 159 struct simplelock *interlkp; 160 struct proc *p; 161 { 162 int lkflags; 163 164 if (mp->mnt_flag & MNT_UNMOUNT) { 165 if (flags & LK_NOWAIT) 166 return (ENOENT); 167 mp->mnt_flag |= MNT_MWAIT; 168 if (interlkp) 169 simple_unlock(interlkp); 170 /* 171 * Since all busy locks are shared except the exclusive 172 * lock granted when unmounting, the only place that a 173 * wakeup needs to be done is at the release of the 174 * exclusive lock at the end of dounmount. 175 */ 176 sleep((caddr_t)mp, PVFS); 177 if (interlkp) 178 simple_lock(interlkp); 179 return (ENOENT); 180 } 181 lkflags = LK_SHARED; 182 if (interlkp) 183 lkflags |= LK_INTERLOCK; 184 if (lockmgr(&mp->mnt_lock, lkflags, interlkp, p)) 185 panic("vfs_busy: unexpected lock failure"); 186 return (0); 187 } 188 189 190 /* 191 * Free a busy file system 192 */ 193 void 194 vfs_unbusy(mp, p) 195 struct mount *mp; 196 struct proc *p; 197 { 198 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, p); 199 } 200 201 /* 202 * Lookup a filesystem type, and if found allocate and initialize 203 * a mount structure for it. 204 * 205 * Devname is usually updated by mount(8) after booting. 206 */ 207 208 int 209 vfs_rootmountalloc(fstypename, devname, mpp) 210 char *fstypename; 211 char *devname; 212 struct mount **mpp; 213 { 214 struct proc *p = curproc; /* XXX */ 215 struct vfsconf *vfsp; 216 struct mount *mp; 217 218 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 219 if (!strcmp(vfsp->vfc_name, fstypename)) 220 break; 221 if (vfsp == NULL) 222 return (ENODEV); 223 mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 224 bzero((char *)mp, (u_long)sizeof(struct mount)); 225 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 226 (void)vfs_busy(mp, LK_NOWAIT, 0, p); 227 LIST_INIT(&mp->mnt_vnodelist); 228 mp->mnt_vfc = vfsp; 229 mp->mnt_op = vfsp->vfc_vfsops; 230 mp->mnt_flag = MNT_RDONLY; 231 mp->mnt_vnodecovered = NULLVP; 232 vfsp->vfc_refcount++; 233 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 234 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 235 mp->mnt_stat.f_mntonname[0] = '/'; 236 (void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); 237 *mpp = mp; 238 return (0); 239 } 240 241 /* 242 * Find an appropriate filesystem to use for the root. If a filesystem 243 * has not been preselected, walk through the list of known filesystems 244 * trying those that have mountroot routines, and try them until one 245 * works or we have tried them all. 246 */ 247 int 248 vfs_mountroot() 249 { 250 struct vfsconf *vfsp; 251 extern int (*mountroot)(void); 252 int error; 253 254 if (mountroot != NULL) 255 return ((*mountroot)()); 256 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) { 257 if (vfsp->vfc_mountroot == NULL) 258 continue; 259 if ((error = (*vfsp->vfc_mountroot)()) == 0) 260 return (0); 261 printf("%s_mountroot failed: %d\n", vfsp->vfc_name, error); 262 } 263 return (ENODEV); 264 } 265 266 /* 267 * Lookup a mount point by filesystem identifier. 268 */ 269 struct mount * 270 vfs_getvfs(fsid) 271 fsid_t *fsid; 272 { 273 register struct mount *mp; 274 275 simple_lock(&mountlist_slock); 276 CIRCLEQ_FOREACH(mp, &mountlist, mnt_list) { 277 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 278 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 279 simple_unlock(&mountlist_slock); 280 return (mp); 281 } 282 } 283 simple_unlock(&mountlist_slock); 284 return ((struct mount *)0); 285 } 286 287 288 /* 289 * Get a new unique fsid 290 */ 291 void 292 vfs_getnewfsid(mp) 293 struct mount *mp; 294 { 295 static u_short xxxfs_mntid; 296 297 fsid_t tfsid; 298 int mtype; 299 300 simple_lock(&mntid_slock); 301 mtype = mp->mnt_vfc->vfc_typenum; 302 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 303 mp->mnt_stat.f_fsid.val[1] = mtype; 304 if (xxxfs_mntid == 0) 305 ++xxxfs_mntid; 306 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 307 tfsid.val[1] = mtype; 308 if (!CIRCLEQ_EMPTY(&mountlist)) { 309 while (vfs_getvfs(&tfsid)) { 310 tfsid.val[0]++; 311 xxxfs_mntid++; 312 } 313 } 314 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 315 simple_unlock(&mntid_slock); 316 } 317 318 /* 319 * Make a 'unique' number from a mount type name. 320 * Note that this is no longer used for ffs which 321 * now has an on-disk filesystem id. 322 */ 323 long 324 makefstype(type) 325 char *type; 326 { 327 long rv; 328 329 for (rv = 0; *type; type++) { 330 rv <<= 2; 331 rv ^= *type; 332 } 333 return rv; 334 } 335 336 /* 337 * Set vnode attributes to VNOVAL 338 */ 339 void 340 vattr_null(vap) 341 register struct vattr *vap; 342 { 343 344 vap->va_type = VNON; 345 /* XXX These next two used to be one line, but for a GCC bug. */ 346 vap->va_size = VNOVAL; 347 vap->va_bytes = VNOVAL; 348 vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = 349 vap->va_fsid = vap->va_fileid = 350 vap->va_blocksize = vap->va_rdev = 351 vap->va_atime.tv_sec = vap->va_atime.tv_nsec = 352 vap->va_mtime.tv_sec = vap->va_mtime.tv_nsec = 353 vap->va_ctime.tv_sec = vap->va_ctime.tv_nsec = 354 vap->va_flags = vap->va_gen = VNOVAL; 355 vap->va_vaflags = 0; 356 } 357 358 /* 359 * Routines having to do with the management of the vnode table. 360 */ 361 extern int (**dead_vnodeop_p) __P((void *)); 362 long numvnodes; 363 364 /* 365 * Return the next vnode from the free list. 366 */ 367 int 368 getnewvnode(tag, mp, vops, vpp) 369 enum vtagtype tag; 370 struct mount *mp; 371 int (**vops) __P((void *)); 372 struct vnode **vpp; 373 { 374 struct proc *p = curproc; /* XXX */ 375 struct freelst *listhd; 376 static int toggle; 377 struct vnode *vp; 378 int s; 379 380 /* 381 * We must choose whether to allocate a new vnode or recycle an 382 * existing one. The criterion for allocating a new one is that 383 * the total number of vnodes is less than the number desired or 384 * there are no vnodes on either free list. Generally we only 385 * want to recycle vnodes that have no buffers associated with 386 * them, so we look first on the vnode_free_list. If it is empty, 387 * we next consider vnodes with referencing buffers on the 388 * vnode_hold_list. The toggle ensures that half the time we 389 * will use a buffer from the vnode_hold_list, and half the time 390 * we will allocate a new one unless the list has grown to twice 391 * the desired size. We are reticent to recycle vnodes from the 392 * vnode_hold_list because we will lose the identity of all its 393 * referencing buffers. 394 */ 395 toggle ^= 1; 396 if (numvnodes > 2 * desiredvnodes) 397 toggle = 0; 398 399 simple_lock(&vnode_free_list_slock); 400 s = splbio(); 401 if ((numvnodes < desiredvnodes) || 402 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 403 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 404 splx(s); 405 simple_unlock(&vnode_free_list_slock); 406 vp = pool_get(&vnode_pool, PR_WAITOK); 407 bzero((char *)vp, sizeof *vp); 408 numvnodes++; 409 } else { 410 for (vp = TAILQ_FIRST(listhd); vp != NULLVP; 411 vp = TAILQ_NEXT(vp, v_freelist)) { 412 if (simple_lock_try(&vp->v_interlock)) 413 break; 414 } 415 /* 416 * Unless this is a bad time of the month, at most 417 * the first NCPUS items on the free list are 418 * locked, so this is close enough to being empty. 419 */ 420 if (vp == NULLVP) { 421 splx(s); 422 simple_unlock(&vnode_free_list_slock); 423 tablefull("vnode"); 424 *vpp = 0; 425 return (ENFILE); 426 } 427 if (vp->v_usecount) { 428 vprint("free vnode", vp); 429 panic("free vnode isn't"); 430 } 431 432 TAILQ_REMOVE(listhd, vp, v_freelist); 433 vp->v_bioflag &= ~VBIOONFREELIST; 434 splx(s); 435 436 simple_unlock(&vnode_free_list_slock); 437 if (vp->v_type != VBAD) 438 vgonel(vp, p); 439 else 440 simple_unlock(&vp->v_interlock); 441 #ifdef DIAGNOSTIC 442 if (vp->v_data) { 443 vprint("cleaned vnode", vp); 444 panic("cleaned vnode isn't"); 445 } 446 s = splbio(); 447 if (vp->v_numoutput) 448 panic("Clean vnode has pending I/O's"); 449 splx(s); 450 #endif 451 vp->v_flag = 0; 452 vp->v_bioflag = 0; 453 vp->v_socket = 0; 454 } 455 vp->v_type = VNON; 456 cache_purge(vp); 457 vp->v_tag = tag; 458 vp->v_op = vops; 459 insmntque(vp, mp); 460 *vpp = vp; 461 vp->v_usecount = 1; 462 vp->v_data = 0; 463 simple_lock_init(&vp->v_uvm.u_obj.vmobjlock); 464 return (0); 465 } 466 467 /* 468 * Move a vnode from one mount queue to another. 469 */ 470 void 471 insmntque(vp, mp) 472 register struct vnode *vp; 473 register struct mount *mp; 474 { 475 simple_lock(&mntvnode_slock); 476 /* 477 * Delete from old mount point vnode list, if on one. 478 */ 479 480 if (vp->v_mount != NULL) 481 LIST_REMOVE(vp, v_mntvnodes); 482 /* 483 * Insert into list of vnodes for the new mount point, if available. 484 */ 485 if ((vp->v_mount = mp) != NULL) 486 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 487 simple_unlock(&mntvnode_slock); 488 } 489 490 491 /* 492 * Create a vnode for a block device. 493 * Used for root filesystem, argdev, and swap areas. 494 * Also used for memory file system special devices. 495 */ 496 int 497 bdevvp(dev, vpp) 498 dev_t dev; 499 struct vnode **vpp; 500 { 501 502 return (getdevvp(dev, vpp, VBLK)); 503 } 504 505 /* 506 * Create a vnode for a character device. 507 * Used for kernfs and some console handling. 508 */ 509 int 510 cdevvp(dev, vpp) 511 dev_t dev; 512 struct vnode **vpp; 513 { 514 515 return (getdevvp(dev, vpp, VCHR)); 516 } 517 518 /* 519 * Create a vnode for a device. 520 * Used by bdevvp (block device) for root file system etc., 521 * and by cdevvp (character device) for console and kernfs. 522 */ 523 int 524 getdevvp(dev, vpp, type) 525 dev_t dev; 526 struct vnode **vpp; 527 enum vtype type; 528 { 529 register struct vnode *vp; 530 struct vnode *nvp; 531 int error; 532 533 if (dev == NODEV) { 534 *vpp = NULLVP; 535 return (0); 536 } 537 error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); 538 if (error) { 539 *vpp = NULLVP; 540 return (error); 541 } 542 vp = nvp; 543 vp->v_type = type; 544 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 545 vput(vp); 546 vp = nvp; 547 } 548 *vpp = vp; 549 return (0); 550 } 551 552 /* 553 * Check to see if the new vnode represents a special device 554 * for which we already have a vnode (either because of 555 * bdevvp() or because of a different vnode representing 556 * the same block device). If such an alias exists, deallocate 557 * the existing contents and return the aliased vnode. The 558 * caller is responsible for filling it with its new contents. 559 */ 560 struct vnode * 561 checkalias(nvp, nvp_rdev, mp) 562 register struct vnode *nvp; 563 dev_t nvp_rdev; 564 struct mount *mp; 565 { 566 struct proc *p = curproc; 567 register struct vnode *vp; 568 struct vnode **vpp; 569 570 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 571 return (NULLVP); 572 573 vpp = &speclisth[SPECHASH(nvp_rdev)]; 574 loop: 575 simple_lock(&spechash_slock); 576 for (vp = *vpp; vp; vp = vp->v_specnext) { 577 simple_lock(&vp->v_interlock); 578 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 579 simple_unlock(&vp->v_interlock); 580 continue; 581 } 582 /* 583 * Alias, but not in use, so flush it out. 584 */ 585 if (vp->v_usecount == 0) { 586 simple_unlock(&spechash_slock); 587 vgonel(vp, p); 588 goto loop; 589 } 590 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 591 simple_unlock(&spechash_slock); 592 goto loop; 593 } 594 break; 595 } 596 597 /* 598 * Common case is actually in the if statement 599 */ 600 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 601 MALLOC(nvp->v_specinfo, struct specinfo *, 602 sizeof(struct specinfo), M_VNODE, M_WAITOK); 603 nvp->v_rdev = nvp_rdev; 604 nvp->v_hashchain = vpp; 605 nvp->v_specnext = *vpp; 606 nvp->v_specmountpoint = NULL; 607 nvp->v_speclockf = NULL; 608 simple_unlock(&spechash_slock); 609 *vpp = nvp; 610 if (vp != NULLVP) { 611 nvp->v_flag |= VALIASED; 612 vp->v_flag |= VALIASED; 613 vput(vp); 614 } 615 return (NULLVP); 616 } 617 618 /* 619 * This code is the uncommon case. It is called in case 620 * we found an alias that was VT_NON && vtype of VBLK 621 * This means we found a block device that was created 622 * using bdevvp. 623 * An example of such a vnode is the root partition device vnode 624 * created in ffs_mountroot. 625 * 626 * The vnodes created by bdevvp should not be aliased (why?). 627 */ 628 629 simple_unlock(&spechash_slock); 630 VOP_UNLOCK(vp, 0, p); 631 simple_lock(&vp->v_interlock); 632 vclean(vp, 0, p); 633 vp->v_op = nvp->v_op; 634 vp->v_tag = nvp->v_tag; 635 nvp->v_type = VNON; 636 insmntque(vp, mp); 637 return (vp); 638 } 639 640 /* 641 * Grab a particular vnode from the free list, increment its 642 * reference count and lock it. The vnode lock bit is set the 643 * vnode is being eliminated in vgone. The process is awakened 644 * when the transition is completed, and an error returned to 645 * indicate that the vnode is no longer usable (possibly having 646 * been changed to a new file system type). 647 */ 648 int 649 vget(vp, flags, p) 650 struct vnode *vp; 651 int flags; 652 struct proc *p; 653 { 654 int error; 655 int s; 656 /* 657 * If the vnode is in the process of being cleaned out for 658 * another use, we wait for the cleaning to finish and then 659 * return failure. Cleaning is determined by checking that 660 * the VXLOCK flag is set. 661 */ 662 if ((flags & LK_INTERLOCK) == 0) { 663 simple_lock(&vp->v_interlock); 664 flags |= LK_INTERLOCK; 665 } 666 if (vp->v_flag & VXLOCK) { 667 vp->v_flag |= VXWANT; 668 simple_unlock(&vp->v_interlock); 669 tsleep((caddr_t)vp, PINOD, "vget", 0); 670 return (ENOENT); 671 } 672 if (vp->v_usecount == 0 && 673 (vp->v_bioflag & VBIOONFREELIST)) { 674 s = splbio(); 675 simple_lock(&vnode_free_list_slock); 676 if (vp->v_holdcnt > 0) 677 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 678 else 679 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 680 simple_unlock(&vnode_free_list_slock); 681 vp->v_bioflag &= ~VBIOONFREELIST; 682 splx(s); 683 } 684 vp->v_usecount++; 685 if (flags & LK_TYPE_MASK) { 686 if ((error = vn_lock(vp, flags, p)) != 0) { 687 vp->v_usecount--; 688 if (vp->v_usecount == 0) 689 vputonfreelist(vp); 690 691 simple_unlock(&vp->v_interlock); 692 } 693 return (error); 694 } 695 simple_unlock(&vp->v_interlock); 696 return (0); 697 } 698 699 700 #ifdef DIAGNOSTIC 701 /* 702 * Vnode reference. 703 */ 704 void 705 vref(vp) 706 struct vnode *vp; 707 { 708 simple_lock(&vp->v_interlock); 709 if (vp->v_usecount == 0) 710 panic("vref used where vget required"); 711 vp->v_usecount++; 712 simple_unlock(&vp->v_interlock); 713 } 714 #endif /* DIAGNOSTIC */ 715 716 static __inline__ void 717 vputonfreelist(vp) 718 struct vnode *vp; 719 { 720 int s; 721 struct freelst *lst; 722 723 s = splbio(); 724 #ifdef DIAGNOSTIC 725 if (vp->v_usecount != 0) 726 panic("Use count is not zero!"); 727 728 if (vp->v_bioflag & VBIOONFREELIST) { 729 vprint("vnode already on free list: ", vp); 730 panic("vnode already on free list"); 731 } 732 #endif 733 734 vp->v_bioflag |= VBIOONFREELIST; 735 736 if (vp->v_holdcnt > 0) 737 lst = &vnode_hold_list; 738 else 739 lst = &vnode_free_list; 740 741 if (vp->v_type == VBAD) 742 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 743 else 744 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 745 746 splx(s); 747 } 748 749 /* 750 * vput(), just unlock and vrele() 751 */ 752 void 753 vput(vp) 754 register struct vnode *vp; 755 { 756 struct proc *p = curproc; /* XXX */ 757 758 #ifdef DIAGNOSTIC 759 if (vp == NULL) 760 panic("vput: null vp"); 761 #endif 762 simple_lock(&vp->v_interlock); 763 764 #ifdef DIAGNOSTIC 765 if (vp->v_usecount == 0) { 766 vprint("vput: bad ref count", vp); 767 panic("vput: ref cnt"); 768 } 769 #endif 770 vp->v_usecount--; 771 if (vp->v_usecount > 0) { 772 simple_unlock(&vp->v_interlock); 773 VOP_UNLOCK(vp, 0, p); 774 return; 775 } 776 777 #ifdef DIAGNOSTIC 778 if (vp->v_writecount != 0) { 779 vprint("vput: bad writecount", vp); 780 panic("vput: v_writecount != 0"); 781 } 782 #endif 783 vputonfreelist(vp); 784 785 simple_unlock(&vp->v_interlock); 786 787 VOP_INACTIVE(vp, p); 788 } 789 790 /* 791 * Vnode release - use for active VNODES. 792 * If count drops to zero, call inactive routine and return to freelist. 793 */ 794 void 795 vrele(vp) 796 register struct vnode *vp; 797 { 798 struct proc *p = curproc; /* XXX */ 799 800 #ifdef DIAGNOSTIC 801 if (vp == NULL) 802 panic("vrele: null vp"); 803 #endif 804 simple_lock(&vp->v_interlock); 805 #ifdef DIAGNOSTIC 806 if (vp->v_usecount == 0) { 807 vprint("vrele: bad ref count", vp); 808 panic("vrele: ref cnt"); 809 } 810 #endif 811 vp->v_usecount--; 812 if (vp->v_usecount > 0) { 813 simple_unlock(&vp->v_interlock); 814 return; 815 } 816 817 #ifdef DIAGNOSTIC 818 if (vp->v_writecount != 0) { 819 vprint("vrele: bad writecount", vp); 820 panic("vrele: v_writecount != 0"); 821 } 822 #endif 823 vputonfreelist(vp); 824 825 if (vn_lock(vp, LK_EXCLUSIVE|LK_INTERLOCK, p) == 0) 826 VOP_INACTIVE(vp, p); 827 } 828 829 void vhold __P((struct vnode *vp)); 830 831 /* 832 * Page or buffer structure gets a reference. 833 */ 834 void 835 vhold(vp) 836 register struct vnode *vp; 837 { 838 839 /* 840 * If it is on the freelist and the hold count is currently 841 * zero, move it to the hold list. 842 */ 843 simple_lock(&vp->v_interlock); 844 if ((vp->v_bioflag & VBIOONFREELIST) && 845 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 846 simple_lock(&vnode_free_list_slock); 847 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 848 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 849 simple_unlock(&vnode_free_list_slock); 850 } 851 vp->v_holdcnt++; 852 simple_unlock(&vp->v_interlock); 853 } 854 855 /* 856 * Remove any vnodes in the vnode table belonging to mount point mp. 857 * 858 * If MNT_NOFORCE is specified, there should not be any active ones, 859 * return error if any are found (nb: this is a user error, not a 860 * system error). If MNT_FORCE is specified, detach any active vnodes 861 * that are found. 862 */ 863 #ifdef DEBUG 864 int busyprt = 0; /* print out busy vnodes */ 865 struct ctldebug debug1 = { "busyprt", &busyprt }; 866 #endif 867 868 int 869 vflush(mp, skipvp, flags) 870 struct mount *mp; 871 struct vnode *skipvp; 872 int flags; 873 { 874 struct proc *p = curproc; 875 register struct vnode *vp, *nvp; 876 int busy = 0; 877 878 simple_lock(&mntvnode_slock); 879 loop: 880 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { 881 if (vp->v_mount != mp) 882 goto loop; 883 nvp = vp->v_mntvnodes.le_next; 884 /* 885 * Skip over a selected vnode. 886 */ 887 if (vp == skipvp) 888 continue; 889 890 simple_lock(&vp->v_interlock); 891 /* 892 * Skip over a vnodes marked VSYSTEM. 893 */ 894 if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 895 simple_unlock(&vp->v_interlock); 896 continue; 897 } 898 /* 899 * If WRITECLOSE is set, only flush out regular file 900 * vnodes open for writing. 901 */ 902 if ((flags & WRITECLOSE) && 903 (vp->v_writecount == 0 || vp->v_type != VREG)) { 904 simple_unlock(&vp->v_interlock); 905 continue; 906 } 907 /* 908 * With v_usecount == 0, all we need to do is clear 909 * out the vnode data structures and we are done. 910 */ 911 if (vp->v_usecount == 0) { 912 simple_unlock(&mntvnode_slock); 913 vgonel(vp, p); 914 simple_lock(&mntvnode_slock); 915 continue; 916 } 917 /* 918 * If FORCECLOSE is set, forcibly close the vnode. 919 * For block or character devices, revert to an 920 * anonymous device. For all other files, just kill them. 921 */ 922 if (flags & FORCECLOSE) { 923 simple_unlock(&mntvnode_slock); 924 if (vp->v_type != VBLK && vp->v_type != VCHR) { 925 vgonel(vp, p); 926 } else { 927 vclean(vp, 0, p); 928 vp->v_op = spec_vnodeop_p; 929 insmntque(vp, (struct mount *)0); 930 } 931 simple_lock(&mntvnode_slock); 932 continue; 933 } 934 #ifdef DEBUG 935 if (busyprt) 936 vprint("vflush: busy vnode", vp); 937 #endif 938 simple_unlock(&vp->v_interlock); 939 busy++; 940 } 941 simple_unlock(&mntvnode_slock); 942 if (busy) 943 return (EBUSY); 944 return (0); 945 } 946 947 /* 948 * Disassociate the underlying file system from a vnode. 949 * The vnode interlock is held on entry. 950 */ 951 void 952 vclean(vp, flags, p) 953 register struct vnode *vp; 954 int flags; 955 struct proc *p; 956 { 957 int active; 958 959 /* 960 * Check to see if the vnode is in use. 961 * If so we have to reference it before we clean it out 962 * so that its count cannot fall to zero and generate a 963 * race against ourselves to recycle it. 964 */ 965 if ((active = vp->v_usecount) != 0) 966 vp->v_usecount++; 967 968 /* 969 * Prevent the vnode from being recycled or 970 * brought into use while we clean it out. 971 */ 972 if (vp->v_flag & VXLOCK) 973 panic("vclean: deadlock"); 974 vp->v_flag |= VXLOCK; 975 /* 976 * Even if the count is zero, the VOP_INACTIVE routine may still 977 * have the object locked while it cleans it out. The VOP_LOCK 978 * ensures that the VOP_INACTIVE routine is done with its work. 979 * For active vnodes, it ensures that no other activity can 980 * occur while the underlying object is being cleaned out. 981 */ 982 VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK, p); 983 984 /* 985 * clean out any VM data associated with the vnode. 986 */ 987 uvm_vnp_terminate(vp); 988 /* 989 * Clean out any buffers associated with the vnode. 990 */ 991 if (flags & DOCLOSE) 992 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 993 /* 994 * If purging an active vnode, it must be closed and 995 * deactivated before being reclaimed. Note that the 996 * VOP_INACTIVE will unlock the vnode 997 */ 998 if (active) { 999 if (flags & DOCLOSE) 1000 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 1001 VOP_INACTIVE(vp, p); 1002 } else { 1003 /* 1004 * Any other processes trying to obtain this lock must first 1005 * wait for VXLOCK to clear, then call the new lock operation. 1006 */ 1007 VOP_UNLOCK(vp, 0, p); 1008 } 1009 1010 /* 1011 * Reclaim the vnode. 1012 */ 1013 if (VOP_RECLAIM(vp, p)) 1014 panic("vclean: cannot reclaim"); 1015 if (active) { 1016 simple_lock(&vp->v_interlock); 1017 1018 vp->v_usecount--; 1019 if (vp->v_usecount == 0) { 1020 if (vp->v_holdcnt > 0) 1021 panic("vclean: not clean"); 1022 vputonfreelist(vp); 1023 } 1024 1025 simple_unlock(&vp->v_interlock); 1026 } 1027 cache_purge(vp); 1028 if (vp->v_vnlock) { 1029 if ((vp->v_vnlock->lk_flags & LK_DRAINED) == 0) 1030 vprint("vclean: lock not drained", vp); 1031 FREE(vp->v_vnlock, M_VNODE); 1032 vp->v_vnlock = NULL; 1033 } 1034 1035 /* 1036 * Done with purge, notify sleepers of the grim news. 1037 */ 1038 vp->v_op = dead_vnodeop_p; 1039 simple_lock(&vp->v_selectinfo.vsi_lock); 1040 VN_KNOTE(vp, NOTE_REVOKE); 1041 simple_unlock(&vp->v_selectinfo.vsi_lock); 1042 vp->v_tag = VT_NON; 1043 vp->v_flag &= ~VXLOCK; 1044 #ifdef DIAGNOSTIC 1045 vp->v_flag &= ~VLOCKSWORK; 1046 #endif 1047 if (vp->v_flag & VXWANT) { 1048 vp->v_flag &= ~VXWANT; 1049 wakeup((caddr_t)vp); 1050 } 1051 } 1052 1053 1054 1055 /* 1056 * Recycle an unused vnode to the front of the free list. 1057 * Release the passed interlock if the vnode will be recycled. 1058 */ 1059 int 1060 vrecycle(vp, inter_lkp, p) 1061 struct vnode *vp; 1062 struct simplelock *inter_lkp; 1063 struct proc *p; 1064 { 1065 1066 simple_lock(&vp->v_interlock); 1067 if (vp->v_usecount == 0) { 1068 if (inter_lkp) 1069 simple_unlock(inter_lkp); 1070 vgonel(vp, p); 1071 return (1); 1072 } 1073 simple_unlock(&vp->v_interlock); 1074 return (0); 1075 } 1076 1077 1078 /* 1079 * Eliminate all activity associated with a vnode 1080 * in preparation for reuse. 1081 */ 1082 void 1083 vgone(vp) 1084 register struct vnode *vp; 1085 { 1086 struct proc *p = curproc; 1087 1088 simple_lock (&vp->v_interlock); 1089 vgonel(vp, p); 1090 } 1091 1092 /* 1093 * vgone, with the vp interlock held. 1094 */ 1095 void 1096 vgonel(vp, p) 1097 struct vnode *vp; 1098 struct proc *p; 1099 { 1100 register struct vnode *vq; 1101 struct vnode *vx; 1102 1103 /* 1104 * If a vgone (or vclean) is already in progress, 1105 * wait until it is done and return. 1106 */ 1107 if (vp->v_flag & VXLOCK) { 1108 vp->v_flag |= VXWANT; 1109 simple_unlock(&vp->v_interlock); 1110 tsleep((caddr_t)vp, PINOD, "vgone", 0); 1111 return; 1112 } 1113 /* 1114 * Clean out the filesystem specific data. 1115 */ 1116 vclean(vp, DOCLOSE, p); 1117 /* 1118 * Delete from old mount point vnode list, if on one. 1119 */ 1120 if (vp->v_mount != NULL) 1121 insmntque(vp, (struct mount *)0); 1122 /* 1123 * If special device, remove it from special device alias list 1124 * if it is on one. 1125 */ 1126 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1127 simple_lock(&spechash_slock); 1128 if (*vp->v_hashchain == vp) { 1129 *vp->v_hashchain = vp->v_specnext; 1130 } else { 1131 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1132 if (vq->v_specnext != vp) 1133 continue; 1134 vq->v_specnext = vp->v_specnext; 1135 break; 1136 } 1137 if (vq == NULL) 1138 panic("missing bdev"); 1139 } 1140 if (vp->v_flag & VALIASED) { 1141 vx = NULL; 1142 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1143 if (vq->v_rdev != vp->v_rdev || 1144 vq->v_type != vp->v_type) 1145 continue; 1146 if (vx) 1147 break; 1148 vx = vq; 1149 } 1150 if (vx == NULL) 1151 panic("missing alias"); 1152 if (vq == NULL) 1153 vx->v_flag &= ~VALIASED; 1154 vp->v_flag &= ~VALIASED; 1155 } 1156 simple_unlock(&spechash_slock); 1157 FREE(vp->v_specinfo, M_VNODE); 1158 vp->v_specinfo = NULL; 1159 } 1160 /* 1161 * If it is on the freelist and not already at the head, 1162 * move it to the head of the list. 1163 */ 1164 vp->v_type = VBAD; 1165 1166 /* 1167 * Move onto the free list, unless we were called from 1168 * getnewvnode and we're not on any free list 1169 */ 1170 if (vp->v_usecount == 0 && 1171 (vp->v_bioflag & VBIOONFREELIST)) { 1172 int s; 1173 1174 simple_lock(&vnode_free_list_slock); 1175 s = splbio(); 1176 1177 if (vp->v_holdcnt > 0) 1178 panic("vgonel: not clean"); 1179 1180 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1181 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1182 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1183 } 1184 splx(s); 1185 simple_unlock(&vnode_free_list_slock); 1186 } 1187 } 1188 1189 /* 1190 * Lookup a vnode by device number. 1191 */ 1192 int 1193 vfinddev(dev, type, vpp) 1194 dev_t dev; 1195 enum vtype type; 1196 struct vnode **vpp; 1197 { 1198 register struct vnode *vp; 1199 int rc =0; 1200 1201 simple_lock(&spechash_slock); 1202 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1203 if (dev != vp->v_rdev || type != vp->v_type) 1204 continue; 1205 *vpp = vp; 1206 rc = 1; 1207 break; 1208 } 1209 simple_unlock(&spechash_slock); 1210 return (rc); 1211 } 1212 1213 /* 1214 * Revoke all the vnodes corresponding to the specified minor number 1215 * range (endpoints inclusive) of the specified major. 1216 */ 1217 void 1218 vdevgone(maj, minl, minh, type) 1219 int maj, minl, minh; 1220 enum vtype type; 1221 { 1222 struct vnode *vp; 1223 int mn; 1224 1225 for (mn = minl; mn <= minh; mn++) 1226 if (vfinddev(makedev(maj, mn), type, &vp)) 1227 VOP_REVOKE(vp, REVOKEALL); 1228 } 1229 1230 /* 1231 * Calculate the total number of references to a special device. 1232 */ 1233 int 1234 vcount(vp) 1235 struct vnode *vp; 1236 { 1237 struct vnode *vq, *vnext; 1238 int count; 1239 1240 loop: 1241 if ((vp->v_flag & VALIASED) == 0) 1242 return (vp->v_usecount); 1243 simple_lock(&spechash_slock); 1244 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1245 vnext = vq->v_specnext; 1246 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1247 continue; 1248 /* 1249 * Alias, but not in use, so flush it out. 1250 */ 1251 if (vq->v_usecount == 0 && vq != vp) { 1252 simple_unlock(&spechash_slock); 1253 vgone(vq); 1254 goto loop; 1255 } 1256 count += vq->v_usecount; 1257 } 1258 simple_unlock(&spechash_slock); 1259 return (count); 1260 } 1261 1262 /* 1263 * Print out a description of a vnode. 1264 */ 1265 static char *typename[] = 1266 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1267 1268 void 1269 vprint(label, vp) 1270 char *label; 1271 register struct vnode *vp; 1272 { 1273 char buf[64]; 1274 1275 if (label != NULL) 1276 printf("%s: ", label); 1277 printf("type %s, usecount %u, writecount %u, holdcount %u,", 1278 typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1279 vp->v_holdcnt); 1280 buf[0] = '\0'; 1281 if (vp->v_flag & VROOT) 1282 strcat(buf, "|VROOT"); 1283 if (vp->v_flag & VTEXT) 1284 strcat(buf, "|VTEXT"); 1285 if (vp->v_flag & VSYSTEM) 1286 strcat(buf, "|VSYSTEM"); 1287 if (vp->v_flag & VXLOCK) 1288 strcat(buf, "|VXLOCK"); 1289 if (vp->v_flag & VXWANT) 1290 strcat(buf, "|VXWANT"); 1291 if (vp->v_bioflag & VBIOWAIT) 1292 strcat(buf, "| VBIOWAIT"); 1293 if (vp->v_flag & VALIASED) 1294 strcat(buf, "|VALIASED"); 1295 if (buf[0] != '\0') 1296 printf(" flags (%s)", &buf[1]); 1297 if (vp->v_data == NULL) { 1298 printf("\n"); 1299 } else { 1300 printf("\n\t"); 1301 VOP_PRINT(vp); 1302 } 1303 } 1304 1305 #ifdef DEBUG 1306 /* 1307 * List all of the locked vnodes in the system. 1308 * Called when debugging the kernel. 1309 */ 1310 void 1311 printlockedvnodes() 1312 { 1313 struct proc *p = curproc; 1314 register struct mount *mp, *nmp; 1315 register struct vnode *vp; 1316 1317 printf("Locked vnodes\n"); 1318 simple_lock(&mountlist_slock); 1319 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1320 mp = nmp) { 1321 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { 1322 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1323 continue; 1324 } 1325 for (vp = mp->mnt_vnodelist.lh_first; vp; 1326 vp = vp->v_mntvnodes.le_next) { 1327 if (VOP_ISLOCKED(vp)) 1328 vprint((char *)0, vp); 1329 } 1330 simple_lock(&mountlist_slock); 1331 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1332 vfs_unbusy(mp, p); 1333 } 1334 simple_unlock(&mountlist_slock); 1335 1336 } 1337 #endif 1338 1339 /* 1340 * Top level filesystem related information gathering. 1341 */ 1342 int 1343 vfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1344 int *name; 1345 u_int namelen; 1346 void *oldp; 1347 size_t *oldlenp; 1348 void *newp; 1349 size_t newlen; 1350 struct proc *p; 1351 { 1352 struct vfsconf *vfsp; 1353 1354 /* all sysctl names at this level are at least name and field */ 1355 if (namelen < 2) 1356 return (ENOTDIR); /* overloaded */ 1357 if (name[0] != VFS_GENERIC) { 1358 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1359 if (vfsp->vfc_typenum == name[0]) 1360 break; 1361 if (vfsp == NULL) 1362 return (EOPNOTSUPP); 1363 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1364 oldp, oldlenp, newp, newlen, p)); 1365 } 1366 switch (name[1]) { 1367 case VFS_MAXTYPENUM: 1368 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1369 case VFS_CONF: 1370 if (namelen < 3) 1371 return (ENOTDIR); /* overloaded */ 1372 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1373 if (vfsp->vfc_typenum == name[2]) 1374 break; 1375 if (vfsp == NULL) 1376 return (EOPNOTSUPP); 1377 return (sysctl_rdstruct(oldp, oldlenp, newp, vfsp, 1378 sizeof(struct vfsconf))); 1379 } 1380 return (EOPNOTSUPP); 1381 } 1382 1383 1384 int kinfo_vdebug = 1; 1385 int kinfo_vgetfailed; 1386 #define KINFO_VNODESLOP 10 1387 /* 1388 * Dump vnode list (via sysctl). 1389 * Copyout address of vnode followed by vnode. 1390 */ 1391 /* ARGSUSED */ 1392 int 1393 sysctl_vnode(where, sizep, p) 1394 char *where; 1395 size_t *sizep; 1396 struct proc *p; 1397 { 1398 register struct mount *mp, *nmp; 1399 struct vnode *vp, *nvp; 1400 register char *bp = where, *savebp; 1401 char *ewhere; 1402 int error; 1403 1404 #define VPTRSZ sizeof (struct vnode *) 1405 #define VNODESZ sizeof (struct vnode) 1406 if (where == NULL) { 1407 *sizep = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ); 1408 return (0); 1409 } 1410 ewhere = where + *sizep; 1411 1412 simple_lock(&mountlist_slock); 1413 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1414 mp = nmp) { 1415 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { 1416 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1417 continue; 1418 } 1419 savebp = bp; 1420 again: 1421 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; 1422 vp = nvp) { 1423 /* 1424 * Check that the vp is still associated with 1425 * this filesystem. RACE: could have been 1426 * recycled onto the same filesystem. 1427 */ 1428 if (vp->v_mount != mp) { 1429 simple_unlock(&mntvnode_slock); 1430 if (kinfo_vdebug) 1431 printf("kinfo: vp changed\n"); 1432 bp = savebp; 1433 goto again; 1434 } 1435 nvp = vp->v_mntvnodes.le_next; 1436 if (bp + VPTRSZ + VNODESZ > ewhere) { 1437 simple_unlock(&mntvnode_slock); 1438 *sizep = bp - where; 1439 return (ENOMEM); 1440 } 1441 if ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) || 1442 (error = copyout((caddr_t)vp, bp + VPTRSZ, VNODESZ))) 1443 return (error); 1444 bp += VPTRSZ + VNODESZ; 1445 simple_lock(&mntvnode_slock); 1446 } 1447 1448 simple_unlock(&mntvnode_slock); 1449 simple_lock(&mountlist_slock); 1450 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1451 vfs_unbusy(mp, p); 1452 } 1453 1454 simple_unlock(&mountlist_slock); 1455 1456 *sizep = bp - where; 1457 return (0); 1458 } 1459 1460 /* 1461 * Check to see if a filesystem is mounted on a block device. 1462 */ 1463 int 1464 vfs_mountedon(vp) 1465 register struct vnode *vp; 1466 { 1467 register struct vnode *vq; 1468 int error = 0; 1469 1470 if (vp->v_specmountpoint != NULL) 1471 return (EBUSY); 1472 if (vp->v_flag & VALIASED) { 1473 simple_lock(&spechash_slock); 1474 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1475 if (vq->v_rdev != vp->v_rdev || 1476 vq->v_type != vp->v_type) 1477 continue; 1478 if (vq->v_specmountpoint != NULL) { 1479 error = EBUSY; 1480 break; 1481 } 1482 } 1483 simple_unlock(&spechash_slock); 1484 } 1485 return (error); 1486 } 1487 1488 /* 1489 * Build hash lists of net addresses and hang them off the mount point. 1490 * Called by ufs_mount() to set up the lists of export addresses. 1491 */ 1492 int 1493 vfs_hang_addrlist(mp, nep, argp) 1494 struct mount *mp; 1495 struct netexport *nep; 1496 struct export_args *argp; 1497 { 1498 register struct netcred *np; 1499 register struct radix_node_head *rnh; 1500 register int i; 1501 struct radix_node *rn; 1502 struct sockaddr *saddr, *smask = 0; 1503 struct domain *dom; 1504 int error; 1505 1506 if (argp->ex_addrlen == 0) { 1507 if (mp->mnt_flag & MNT_DEFEXPORTED) 1508 return (EPERM); 1509 np = &nep->ne_defexported; 1510 np->netc_exflags = argp->ex_flags; 1511 np->netc_anon = argp->ex_anon; 1512 np->netc_anon.cr_ref = 1; 1513 mp->mnt_flag |= MNT_DEFEXPORTED; 1514 return (0); 1515 } 1516 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1517 np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK); 1518 bzero((caddr_t)np, i); 1519 saddr = (struct sockaddr *)(np + 1); 1520 error = copyin(argp->ex_addr, (caddr_t)saddr, argp->ex_addrlen); 1521 if (error) 1522 goto out; 1523 if (saddr->sa_len > argp->ex_addrlen) 1524 saddr->sa_len = argp->ex_addrlen; 1525 if (argp->ex_masklen) { 1526 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1527 error = copyin(argp->ex_mask, (caddr_t)smask, argp->ex_masklen); 1528 if (error) 1529 goto out; 1530 if (smask->sa_len > argp->ex_masklen) 1531 smask->sa_len = argp->ex_masklen; 1532 } 1533 i = saddr->sa_family; 1534 if ((rnh = nep->ne_rtable[i]) == 0) { 1535 /* 1536 * Seems silly to initialize every AF when most are not 1537 * used, do so on demand here 1538 */ 1539 for (dom = domains; dom; dom = dom->dom_next) 1540 if (dom->dom_family == i && dom->dom_rtattach) { 1541 dom->dom_rtattach((void **)&nep->ne_rtable[i], 1542 dom->dom_rtoffset); 1543 break; 1544 } 1545 if ((rnh = nep->ne_rtable[i]) == 0) { 1546 error = ENOBUFS; 1547 goto out; 1548 } 1549 } 1550 rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, 1551 np->netc_rnodes); 1552 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1553 error = EPERM; 1554 goto out; 1555 } 1556 np->netc_exflags = argp->ex_flags; 1557 np->netc_anon = argp->ex_anon; 1558 np->netc_anon.cr_ref = 1; 1559 return (0); 1560 out: 1561 free(np, M_NETADDR); 1562 return (error); 1563 } 1564 1565 /* ARGSUSED */ 1566 int 1567 vfs_free_netcred(rn, w) 1568 struct radix_node *rn; 1569 void *w; 1570 { 1571 register struct radix_node_head *rnh = (struct radix_node_head *)w; 1572 1573 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh); 1574 free((caddr_t)rn, M_NETADDR); 1575 return (0); 1576 } 1577 1578 /* 1579 * Free the net address hash lists that are hanging off the mount points. 1580 */ 1581 void 1582 vfs_free_addrlist(nep) 1583 struct netexport *nep; 1584 { 1585 register int i; 1586 register struct radix_node_head *rnh; 1587 1588 for (i = 0; i <= AF_MAX; i++) 1589 if ((rnh = nep->ne_rtable[i]) != NULL) { 1590 (*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh); 1591 free((caddr_t)rnh, M_RTABLE); 1592 nep->ne_rtable[i] = 0; 1593 } 1594 } 1595 1596 int 1597 vfs_export(mp, nep, argp) 1598 struct mount *mp; 1599 struct netexport *nep; 1600 struct export_args *argp; 1601 { 1602 int error; 1603 1604 if (argp->ex_flags & MNT_DELEXPORT) { 1605 vfs_free_addrlist(nep); 1606 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1607 } 1608 if (argp->ex_flags & MNT_EXPORTED) { 1609 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1610 return (error); 1611 mp->mnt_flag |= MNT_EXPORTED; 1612 } 1613 return (0); 1614 } 1615 1616 struct netcred * 1617 vfs_export_lookup(mp, nep, nam) 1618 register struct mount *mp; 1619 struct netexport *nep; 1620 struct mbuf *nam; 1621 { 1622 register struct netcred *np; 1623 register struct radix_node_head *rnh; 1624 struct sockaddr *saddr; 1625 1626 np = NULL; 1627 if (mp->mnt_flag & MNT_EXPORTED) { 1628 /* 1629 * Lookup in the export list first. 1630 */ 1631 if (nam != NULL) { 1632 saddr = mtod(nam, struct sockaddr *); 1633 rnh = nep->ne_rtable[saddr->sa_family]; 1634 if (rnh != NULL) { 1635 np = (struct netcred *) 1636 (*rnh->rnh_matchaddr)((caddr_t)saddr, 1637 rnh); 1638 if (np && np->netc_rnodes->rn_flags & RNF_ROOT) 1639 np = NULL; 1640 } 1641 } 1642 /* 1643 * If no address match, use the default if it exists. 1644 */ 1645 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1646 np = &nep->ne_defexported; 1647 } 1648 return (np); 1649 } 1650 1651 /* 1652 * Do the usual access checking. 1653 * file_mode, uid and gid are from the vnode in question, 1654 * while acc_mode and cred are from the VOP_ACCESS parameter list 1655 */ 1656 int 1657 vaccess(file_mode, uid, gid, acc_mode, cred) 1658 mode_t file_mode; 1659 uid_t uid; 1660 gid_t gid; 1661 mode_t acc_mode; 1662 struct ucred *cred; 1663 { 1664 mode_t mask; 1665 1666 /* User id 0 always gets access. */ 1667 if (cred->cr_uid == 0) 1668 return 0; 1669 1670 mask = 0; 1671 1672 /* Otherwise, check the owner. */ 1673 if (cred->cr_uid == uid) { 1674 if (acc_mode & VEXEC) 1675 mask |= S_IXUSR; 1676 if (acc_mode & VREAD) 1677 mask |= S_IRUSR; 1678 if (acc_mode & VWRITE) 1679 mask |= S_IWUSR; 1680 return (file_mode & mask) == mask ? 0 : EACCES; 1681 } 1682 1683 /* Otherwise, check the groups. */ 1684 if (cred->cr_gid == gid || groupmember(gid, cred)) { 1685 if (acc_mode & VEXEC) 1686 mask |= S_IXGRP; 1687 if (acc_mode & VREAD) 1688 mask |= S_IRGRP; 1689 if (acc_mode & VWRITE) 1690 mask |= S_IWGRP; 1691 return (file_mode & mask) == mask ? 0 : EACCES; 1692 } 1693 1694 /* Otherwise, check everyone else. */ 1695 if (acc_mode & VEXEC) 1696 mask |= S_IXOTH; 1697 if (acc_mode & VREAD) 1698 mask |= S_IROTH; 1699 if (acc_mode & VWRITE) 1700 mask |= S_IWOTH; 1701 return (file_mode & mask) == mask ? 0 : EACCES; 1702 } 1703 1704 /* 1705 * Unmount all file systems. 1706 * We traverse the list in reverse order under the assumption that doing so 1707 * will avoid needing to worry about dependencies. 1708 */ 1709 void 1710 vfs_unmountall() 1711 { 1712 register struct mount *mp, *nmp; 1713 int allerror, error, again = 1; 1714 1715 retry: 1716 allerror = 0; 1717 for (mp = CIRCLEQ_LAST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1718 mp = nmp) { 1719 nmp = CIRCLEQ_PREV(mp, mnt_list); 1720 if ((error = dounmount(mp, MNT_FORCE, curproc)) != 0) { 1721 printf("unmount of %s failed with error %d\n", 1722 mp->mnt_stat.f_mntonname, error); 1723 allerror = 1; 1724 } 1725 } 1726 1727 if (allerror) { 1728 printf("WARNING: some file systems would not unmount\n"); 1729 if (again) { 1730 printf("retrying\n"); 1731 again = 0; 1732 goto retry; 1733 } 1734 } 1735 } 1736 1737 /* 1738 * Sync and unmount file systems before shutting down. 1739 */ 1740 void 1741 vfs_shutdown() 1742 { 1743 /* XXX Should suspend scheduling. */ 1744 (void) spl0(); 1745 1746 printf("syncing disks... "); 1747 1748 if (panicstr == 0) { 1749 /* Sync before unmount, in case we hang on something. */ 1750 sys_sync(&proc0, (void *)0, (register_t *)0); 1751 1752 /* Unmount file systems. */ 1753 vfs_unmountall(); 1754 } 1755 1756 if (vfs_syncwait(1)) 1757 printf("giving up\n"); 1758 else 1759 printf("done\n"); 1760 } 1761 1762 /* 1763 * perform sync() operation and wait for buffers to flush. 1764 * assumtions: called w/ scheduler disabled and physical io enabled 1765 * for now called at spl0() XXX 1766 */ 1767 int 1768 vfs_syncwait(verbose) 1769 int verbose; 1770 { 1771 register struct buf *bp; 1772 int iter, nbusy, dcount, s; 1773 struct proc *p; 1774 1775 p = curproc? curproc : &proc0; 1776 sys_sync(p, (void *)0, (register_t *)0); 1777 1778 /* Wait for sync to finish. */ 1779 dcount = 10000; 1780 for (iter = 0; iter < 20; iter++) { 1781 nbusy = 0; 1782 for (bp = &buf[nbuf]; --bp >= buf; ) { 1783 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1784 nbusy++; 1785 /* 1786 * With soft updates, some buffers that are 1787 * written will be remarked as dirty until other 1788 * buffers are written. 1789 */ 1790 if (bp->b_flags & B_DELWRI) { 1791 s = splbio(); 1792 bremfree(bp); 1793 bp->b_flags |= B_BUSY; 1794 splx(s); 1795 nbusy++; 1796 bawrite(bp); 1797 if (dcount-- <= 0) { 1798 if (verbose) 1799 printf("softdep "); 1800 return 1; 1801 } 1802 } 1803 } 1804 if (nbusy == 0) 1805 break; 1806 if (verbose) 1807 printf("%d ", nbusy); 1808 DELAY(40000 * iter); 1809 } 1810 1811 return nbusy; 1812 } 1813 1814 /* 1815 * posix file system related system variables. 1816 */ 1817 int 1818 fs_posix_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1819 int *name; 1820 u_int namelen; 1821 void *oldp; 1822 size_t *oldlenp; 1823 void *newp; 1824 size_t newlen; 1825 struct proc *p; 1826 { 1827 /* all sysctl names at this level are terminal */ 1828 if (namelen != 1) 1829 return (ENOTDIR); 1830 1831 switch (name[0]) { 1832 case FS_POSIX_SETUID: 1833 if (newp && securelevel > 0) 1834 return (EPERM); 1835 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1836 default: 1837 return (EOPNOTSUPP); 1838 } 1839 /* NOTREACHED */ 1840 } 1841 1842 /* 1843 * file system related system variables. 1844 */ 1845 int 1846 fs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1847 int *name; 1848 u_int namelen; 1849 void *oldp; 1850 size_t *oldlenp; 1851 void *newp; 1852 size_t newlen; 1853 struct proc *p; 1854 { 1855 sysctlfn *fn; 1856 1857 switch (name[0]) { 1858 case FS_POSIX: 1859 fn = fs_posix_sysctl; 1860 break; 1861 default: 1862 return (EOPNOTSUPP); 1863 } 1864 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1865 } 1866 1867 1868 /* 1869 * Routines dealing with vnodes and buffers 1870 */ 1871 1872 /* 1873 * Wait for all outstanding I/Os to complete 1874 * 1875 * Manipulates v_numoutput. Must be called at splbio() 1876 */ 1877 int 1878 vwaitforio(vp, slpflag, wmesg, timeo) 1879 struct vnode *vp; 1880 int slpflag, timeo; 1881 char *wmesg; 1882 { 1883 int error = 0; 1884 1885 while (vp->v_numoutput) { 1886 vp->v_bioflag |= VBIOWAIT; 1887 error = tsleep((caddr_t)&vp->v_numoutput, 1888 slpflag | (PRIBIO + 1), wmesg, timeo); 1889 if (error) 1890 break; 1891 } 1892 1893 return (error); 1894 } 1895 1896 1897 /* 1898 * Update outstanding I/O count and do wakeup if requested. 1899 * 1900 * Manipulates v_numoutput. Must be called at splbio() 1901 */ 1902 void 1903 vwakeup(vp) 1904 struct vnode *vp; 1905 { 1906 if (vp != NULL) { 1907 if (vp->v_numoutput-- == 0) 1908 panic("vwakeup: neg numoutput"); 1909 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1910 vp->v_bioflag &= ~VBIOWAIT; 1911 wakeup((caddr_t)&vp->v_numoutput); 1912 } 1913 } 1914 } 1915 1916 /* 1917 * Flush out and invalidate all buffers associated with a vnode. 1918 * Called with the underlying object locked. 1919 */ 1920 int 1921 vinvalbuf(vp, flags, cred, p, slpflag, slptimeo) 1922 register struct vnode *vp; 1923 int flags; 1924 struct ucred *cred; 1925 struct proc *p; 1926 int slpflag, slptimeo; 1927 { 1928 register struct buf *bp; 1929 struct buf *nbp, *blist; 1930 int s, error; 1931 1932 if (flags & V_SAVE) { 1933 s = splbio(); 1934 vwaitforio(vp, 0, "vinvalbuf", 0); 1935 if (vp->v_dirtyblkhd.lh_first != NULL) { 1936 splx(s); 1937 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1938 return (error); 1939 s = splbio(); 1940 if (vp->v_numoutput > 0 || 1941 vp->v_dirtyblkhd.lh_first != NULL) 1942 panic("vinvalbuf: dirty bufs"); 1943 } 1944 splx(s); 1945 } 1946 loop: 1947 s = splbio(); 1948 for (;;) { 1949 if ((blist = vp->v_cleanblkhd.lh_first) && 1950 (flags & V_SAVEMETA)) 1951 while (blist && blist->b_lblkno < 0) 1952 blist = blist->b_vnbufs.le_next; 1953 if (!blist && (blist = vp->v_dirtyblkhd.lh_first) && 1954 (flags & V_SAVEMETA)) 1955 while (blist && blist->b_lblkno < 0) 1956 blist = blist->b_vnbufs.le_next; 1957 if (!blist) 1958 break; 1959 1960 for (bp = blist; bp; bp = nbp) { 1961 nbp = bp->b_vnbufs.le_next; 1962 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1963 continue; 1964 if (bp->b_flags & B_BUSY) { 1965 bp->b_flags |= B_WANTED; 1966 error = tsleep((caddr_t)bp, 1967 slpflag | (PRIBIO + 1), "vinvalbuf", 1968 slptimeo); 1969 if (error) { 1970 splx(s); 1971 return (error); 1972 } 1973 break; 1974 } 1975 bp->b_flags |= B_BUSY | B_VFLUSH; 1976 /* 1977 * XXX Since there are no node locks for NFS, I believe 1978 * there is a slight chance that a delayed write will 1979 * occur while sleeping just above, so check for it. 1980 */ 1981 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1982 splx(s); 1983 (void) VOP_BWRITE(bp); 1984 goto loop; 1985 } 1986 bp->b_flags |= B_INVAL; 1987 brelse(bp); 1988 } 1989 } 1990 if (!(flags & V_SAVEMETA) && 1991 (vp->v_dirtyblkhd.lh_first || vp->v_cleanblkhd.lh_first)) 1992 panic("vinvalbuf: flush failed"); 1993 splx(s); 1994 return (0); 1995 } 1996 1997 void 1998 vflushbuf(vp, sync) 1999 register struct vnode *vp; 2000 int sync; 2001 { 2002 register struct buf *bp, *nbp; 2003 int s; 2004 2005 loop: 2006 s = splbio(); 2007 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { 2008 nbp = bp->b_vnbufs.le_next; 2009 if ((bp->b_flags & B_BUSY)) 2010 continue; 2011 if ((bp->b_flags & B_DELWRI) == 0) 2012 panic("vflushbuf: not dirty"); 2013 bp->b_flags |= B_BUSY | B_VFLUSH; 2014 splx(s); 2015 /* 2016 * Wait for I/O associated with indirect blocks to complete, 2017 * since there is no way to quickly wait for them below. 2018 */ 2019 if (bp->b_vp == vp || sync == 0) 2020 (void) bawrite(bp); 2021 else 2022 (void) bwrite(bp); 2023 goto loop; 2024 } 2025 if (sync == 0) { 2026 splx(s); 2027 return; 2028 } 2029 vwaitforio(vp, 0, "vflushbuf", 0); 2030 if (vp->v_dirtyblkhd.lh_first != NULL) { 2031 splx(s); 2032 vprint("vflushbuf: dirty", vp); 2033 goto loop; 2034 } 2035 splx(s); 2036 } 2037 2038 /* 2039 * Associate a buffer with a vnode. 2040 * 2041 * Manipulates buffer vnode queues. Must be called at splbio(). 2042 */ 2043 void 2044 bgetvp(vp, bp) 2045 register struct vnode *vp; 2046 register struct buf *bp; 2047 { 2048 2049 if (bp->b_vp) 2050 panic("bgetvp: not free"); 2051 vhold(vp); 2052 bp->b_vp = vp; 2053 if (vp->v_type == VBLK || vp->v_type == VCHR) 2054 bp->b_dev = vp->v_rdev; 2055 else 2056 bp->b_dev = NODEV; 2057 /* 2058 * Insert onto list for new vnode. 2059 */ 2060 bufinsvn(bp, &vp->v_cleanblkhd); 2061 } 2062 2063 /* 2064 * Disassociate a buffer from a vnode. 2065 * 2066 * Manipulates vnode buffer queues. Must be called at splbio(). 2067 */ 2068 void 2069 brelvp(bp) 2070 register struct buf *bp; 2071 { 2072 struct vnode *vp; 2073 2074 if ((vp = bp->b_vp) == (struct vnode *) 0) 2075 panic("brelvp: NULL"); 2076 /* 2077 * Delete from old vnode list, if on one. 2078 */ 2079 if (bp->b_vnbufs.le_next != NOLIST) 2080 bufremvn(bp); 2081 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2082 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2083 vp->v_bioflag &= ~VBIOONSYNCLIST; 2084 LIST_REMOVE(vp, v_synclist); 2085 } 2086 bp->b_vp = (struct vnode *) 0; 2087 2088 simple_lock(&vp->v_interlock); 2089 #ifdef DIAGNOSTIC 2090 if (vp->v_holdcnt == 0) 2091 panic("brelvp: holdcnt"); 2092 #endif 2093 vp->v_holdcnt--; 2094 2095 /* 2096 * If it is on the holdlist and the hold count drops to 2097 * zero, move it to the free list. 2098 */ 2099 if ((vp->v_bioflag & VBIOONFREELIST) && 2100 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 2101 simple_lock(&vnode_free_list_slock); 2102 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 2103 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 2104 simple_unlock(&vnode_free_list_slock); 2105 } 2106 simple_unlock(&vp->v_interlock); 2107 } 2108 2109 /* 2110 * Replaces the current vnode associated with the buffer, if any 2111 * with a new vnode. 2112 * 2113 * If an output I/O is pending on the buffer, the old vnode is 2114 * I/O count is adjusted. 2115 * 2116 * Ignores vnode buffer queues. Must be called at splbio(). 2117 */ 2118 void 2119 buf_replacevnode(bp, newvp) 2120 struct buf *bp; 2121 struct vnode *newvp; 2122 { 2123 struct vnode *oldvp = bp->b_vp; 2124 2125 if (oldvp) 2126 brelvp(bp); 2127 2128 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2129 newvp->v_numoutput++; /* put it on swapdev */ 2130 vwakeup(oldvp); 2131 } 2132 2133 bgetvp(newvp, bp); 2134 bufremvn(bp); 2135 } 2136 2137 /* 2138 * Used to assign buffers to the appropriate clean or dirty list on 2139 * the vnode and to add newly dirty vnodes to the appropriate 2140 * filesystem syncer list. 2141 * 2142 * Manipulates vnode buffer queues. Must be called at splbio(). 2143 */ 2144 void 2145 reassignbuf(bp) 2146 register struct buf *bp; 2147 { 2148 struct buflists *listheadp; 2149 int delay; 2150 struct vnode *vp = bp->b_vp; 2151 2152 if (vp == NULL) { 2153 printf("reassignbuf: NULL"); 2154 return; 2155 } 2156 /* 2157 * Delete from old vnode list, if on one. 2158 */ 2159 if (bp->b_vnbufs.le_next != NOLIST) 2160 bufremvn(bp); 2161 /* 2162 * If dirty, put on list of dirty buffers; 2163 * otherwise insert onto list of clean buffers. 2164 */ 2165 if ((bp->b_flags & B_DELWRI) == 0) { 2166 listheadp = &vp->v_cleanblkhd; 2167 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2168 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2169 vp->v_bioflag &= ~VBIOONSYNCLIST; 2170 LIST_REMOVE(vp, v_synclist); 2171 } 2172 } else { 2173 listheadp = &vp->v_dirtyblkhd; 2174 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2175 switch (vp->v_type) { 2176 case VDIR: 2177 delay = syncdelay / 2; 2178 break; 2179 case VBLK: 2180 if (vp->v_specmountpoint != NULL) { 2181 delay = syncdelay / 3; 2182 break; 2183 } 2184 /* fall through */ 2185 default: 2186 delay = syncdelay; 2187 } 2188 vn_syncer_add_to_worklist(vp, delay); 2189 } 2190 } 2191 bufinsvn(bp, listheadp); 2192 } 2193 2194 int 2195 vfs_register(vfs) 2196 struct vfsconf *vfs; 2197 { 2198 struct vfsconf *vfsp; 2199 struct vfsconf **vfspp; 2200 2201 #ifdef DIAGNOSTIC 2202 /* Paranoia? */ 2203 if (vfs->vfc_refcount != 0) 2204 printf("vfs_register called with vfc_refcount > 0\n"); 2205 #endif 2206 2207 /* Check if filesystem already known */ 2208 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2209 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2210 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2211 return (EEXIST); 2212 2213 if (vfs->vfc_typenum > maxvfsconf) 2214 maxvfsconf = vfs->vfc_typenum; 2215 2216 vfs->vfc_next = NULL; 2217 2218 /* Add to the end of the list */ 2219 *vfspp = vfs; 2220 2221 /* Call vfs_init() */ 2222 if (vfs->vfc_vfsops->vfs_init) 2223 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2224 2225 return 0; 2226 } 2227 2228 int 2229 vfs_unregister(vfs) 2230 struct vfsconf *vfs; 2231 { 2232 struct vfsconf *vfsp; 2233 struct vfsconf **vfspp; 2234 int maxtypenum; 2235 2236 /* Find our vfsconf struct */ 2237 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2238 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2239 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2240 break; 2241 } 2242 2243 if (!vfsp) /* Not found */ 2244 return (ENOENT); 2245 2246 if (vfsp->vfc_refcount) /* In use */ 2247 return (EBUSY); 2248 2249 /* Remove from list and free */ 2250 *vfspp = vfsp->vfc_next; 2251 2252 maxtypenum = 0; 2253 2254 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2255 if (vfsp->vfc_typenum > maxtypenum) 2256 maxtypenum = vfsp->vfc_typenum; 2257 2258 maxvfsconf = maxtypenum; 2259 return 0; 2260 } 2261 2262 /* 2263 * Check if vnode represents a disk device 2264 */ 2265 int 2266 vn_isdisk(vp, errp) 2267 struct vnode *vp; 2268 int *errp; 2269 { 2270 if (vp->v_type != VBLK && vp->v_type != VCHR) 2271 return (0); 2272 2273 return (1); 2274 } 2275