vfs_subr.c revision 1.131
1/* $OpenBSD: vfs_subr.c,v 1.131 2006/07/08 20:01:13 thib 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. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 * 37 * @(#)vfs_subr.c 8.13 (Berkeley) 4/18/94 38 */ 39 40/* 41 * External virtual filesystem routines 42 */ 43 44#include <sys/param.h> 45#include <sys/systm.h> 46#include <sys/proc.h> 47#include <sys/mount.h> 48#include <sys/time.h> 49#include <sys/fcntl.h> 50#include <sys/kernel.h> 51#include <sys/vnode.h> 52#include <sys/stat.h> 53#include <sys/namei.h> 54#include <sys/ucred.h> 55#include <sys/buf.h> 56#include <sys/errno.h> 57#include <sys/malloc.h> 58#include <sys/domain.h> 59#include <sys/mbuf.h> 60#include <sys/syscallargs.h> 61#include <sys/pool.h> 62 63#include <uvm/uvm_extern.h> 64#include <sys/sysctl.h> 65 66#include <miscfs/specfs/specdev.h> 67 68enum vtype iftovt_tab[16] = { 69 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, 70 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, 71}; 72 73int vttoif_tab[9] = { 74 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, 75 S_IFSOCK, S_IFIFO, S_IFMT, 76}; 77 78int doforce = 1; /* 1 => permit forcible unmounting */ 79int prtactive = 0; /* 1 => print out reclaim of active vnodes */ 80int suid_clear = 1; /* 1 => clear SUID / SGID on owner change */ 81 82/* 83 * Insq/Remq for the vnode usage lists. 84 */ 85#define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) 86#define bufremvn(bp) { \ 87 LIST_REMOVE(bp, b_vnbufs); \ 88 LIST_NEXT(bp, b_vnbufs) = NOLIST; \ 89} 90 91struct freelst vnode_hold_list; /* list of vnodes referencing buffers */ 92struct freelst vnode_free_list; /* vnode free list */ 93 94struct mntlist mountlist; /* mounted filesystem list */ 95static struct simplelock mntid_slock; 96struct simplelock mntvnode_slock; 97struct simplelock vnode_free_list_slock; 98struct simplelock spechash_slock; 99 100void vclean(struct vnode *, int, struct proc *); 101 102void insmntque(struct vnode *, struct mount *); 103int getdevvp(dev_t, struct vnode **, enum vtype); 104 105int vfs_hang_addrlist(struct mount *, struct netexport *, 106 struct export_args *); 107int vfs_free_netcred(struct radix_node *, void *); 108void vfs_free_addrlist(struct netexport *); 109void vputonfreelist(struct vnode *); 110 111int vflush_vnode(struct vnode *, void *); 112 113#ifdef DEBUG 114void printlockedvnodes(void); 115#endif 116 117#define VN_KNOTE(vp, b) \ 118 KNOTE((struct klist *)&vp->v_selectinfo.vsi_selinfo.si_note, (b)) 119 120struct pool vnode_pool; 121 122/* 123 * Initialize the vnode management data structures. 124 */ 125void 126vntblinit(void) 127{ 128 129 pool_init(&vnode_pool, sizeof(struct vnode), 0, 0, 0, "vnodes", 130 &pool_allocator_nointr); 131 simple_lock_init(&mntvnode_slock); 132 simple_lock_init(&mntid_slock); 133 simple_lock_init(&spechash_slock); 134 TAILQ_INIT(&vnode_hold_list); 135 TAILQ_INIT(&vnode_free_list); 136 simple_lock_init(&vnode_free_list_slock); 137 CIRCLEQ_INIT(&mountlist); 138 /* 139 * Initialize the filesystem syncer. 140 */ 141 vn_initialize_syncerd(); 142} 143 144/* 145 * Mark a mount point as busy. Used to synchronize access and to delay 146 * unmounting. 147 * 148 * Default behaviour is to attempt getting a READ lock and in case of an 149 * ongoing unmount, to wait for it to finish and then return failure. 150 */ 151int 152vfs_busy(struct mount *mp, int flags) 153{ 154 int rwflags = 0; 155 156 /* new mountpoints need their lock initialised */ 157 if (mp->mnt_lock.rwl_name == NULL) 158 rw_init(&mp->mnt_lock, "vfslock"); 159 160 if (flags & VB_WRITE) 161 rwflags |= RW_WRITE; 162 else 163 rwflags |= RW_READ; 164 165 if (flags & VB_WAIT) 166 rwflags |= RW_SLEEPFAIL; 167 else 168 rwflags |= RW_NOSLEEP; 169 170 if (rw_enter(&mp->mnt_lock, rwflags)) 171 return (EBUSY); 172 173 return (0); 174} 175 176/* 177 * Free a busy file system 178 */ 179void 180vfs_unbusy(struct mount *mp) 181{ 182 rw_exit(&mp->mnt_lock); 183} 184 185int 186vfs_isbusy(struct mount *mp) 187{ 188 if (RWLOCK_OWNER(&mp->mnt_lock) > 0) 189 return (1); 190 else 191 return (0); 192} 193 194/* 195 * Lookup a filesystem type, and if found allocate and initialize 196 * a mount structure for it. 197 * 198 * Devname is usually updated by mount(8) after booting. 199 */ 200int 201vfs_rootmountalloc(char *fstypename, char *devname, struct mount **mpp) 202{ 203 struct vfsconf *vfsp; 204 struct mount *mp; 205 206 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 207 if (!strcmp(vfsp->vfc_name, fstypename)) 208 break; 209 if (vfsp == NULL) 210 return (ENODEV); 211 mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 212 bzero((char *)mp, (u_long)sizeof(struct mount)); 213 (void) vfs_busy(mp, VB_READ|VB_NOWAIT); 214 LIST_INIT(&mp->mnt_vnodelist); 215 mp->mnt_vfc = vfsp; 216 mp->mnt_op = vfsp->vfc_vfsops; 217 mp->mnt_flag = MNT_RDONLY; 218 mp->mnt_vnodecovered = NULLVP; 219 vfsp->vfc_refcount++; 220 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 221 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 222 mp->mnt_stat.f_mntonname[0] = '/'; 223 (void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); 224 *mpp = mp; 225 return (0); 226 } 227 228/* 229 * Find an appropriate filesystem to use for the root. If a filesystem 230 * has not been preselected, walk through the list of known filesystems 231 * trying those that have mountroot routines, and try them until one 232 * works or we have tried them all. 233 */ 234int 235vfs_mountroot(void) 236{ 237 struct vfsconf *vfsp; 238 int error; 239 240 if (mountroot != NULL) 241 return ((*mountroot)()); 242 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) { 243 if (vfsp->vfc_mountroot == NULL) 244 continue; 245 if ((error = (*vfsp->vfc_mountroot)()) == 0) 246 return (0); 247 printf("%s_mountroot failed: %d\n", vfsp->vfc_name, error); 248 } 249 return (ENODEV); 250} 251 252/* 253 * Lookup a mount point by filesystem identifier. 254 */ 255struct mount * 256vfs_getvfs(fsid_t *fsid) 257{ 258 struct mount *mp; 259 260 CIRCLEQ_FOREACH(mp, &mountlist, mnt_list) { 261 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 262 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 263 return (mp); 264 } 265 } 266 267 return ((struct mount *)0); 268} 269 270 271/* 272 * Get a new unique fsid 273 */ 274void 275vfs_getnewfsid(struct mount *mp) 276{ 277 static u_short xxxfs_mntid; 278 279 fsid_t tfsid; 280 int mtype; 281 282 simple_lock(&mntid_slock); 283 mtype = mp->mnt_vfc->vfc_typenum; 284 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 285 mp->mnt_stat.f_fsid.val[1] = mtype; 286 if (xxxfs_mntid == 0) 287 ++xxxfs_mntid; 288 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 289 tfsid.val[1] = mtype; 290 if (!CIRCLEQ_EMPTY(&mountlist)) { 291 while (vfs_getvfs(&tfsid)) { 292 tfsid.val[0]++; 293 xxxfs_mntid++; 294 } 295 } 296 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 297 simple_unlock(&mntid_slock); 298} 299 300/* 301 * Make a 'unique' number from a mount type name. 302 * Note that this is no longer used for ffs which 303 * now has an on-disk filesystem id. 304 */ 305long 306makefstype(char *type) 307{ 308 long rv; 309 310 for (rv = 0; *type; type++) { 311 rv <<= 2; 312 rv ^= *type; 313 } 314 return rv; 315} 316 317/* 318 * Set vnode attributes to VNOVAL 319 */ 320void 321vattr_null(struct vattr *vap) 322{ 323 324 vap->va_type = VNON; 325 /* XXX These next two used to be one line, but for a GCC bug. */ 326 vap->va_size = VNOVAL; 327 vap->va_bytes = VNOVAL; 328 vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = 329 vap->va_fsid = vap->va_fileid = 330 vap->va_blocksize = vap->va_rdev = 331 vap->va_atime.tv_sec = vap->va_atime.tv_nsec = 332 vap->va_mtime.tv_sec = vap->va_mtime.tv_nsec = 333 vap->va_ctime.tv_sec = vap->va_ctime.tv_nsec = 334 vap->va_flags = vap->va_gen = VNOVAL; 335 vap->va_vaflags = 0; 336} 337 338/* 339 * Routines having to do with the management of the vnode table. 340 */ 341extern int (**dead_vnodeop_p)(void *); 342long numvnodes; 343 344/* 345 * Return the next vnode from the free list. 346 */ 347int 348getnewvnode(enum vtagtype tag, struct mount *mp, int (**vops)(void *), 349 struct vnode **vpp) 350{ 351 struct proc *p = curproc; 352 struct freelst *listhd; 353 static int toggle; 354 struct vnode *vp; 355 int s; 356 357 /* 358 * We must choose whether to allocate a new vnode or recycle an 359 * existing one. The criterion for allocating a new one is that 360 * the total number of vnodes is less than the number desired or 361 * there are no vnodes on either free list. Generally we only 362 * want to recycle vnodes that have no buffers associated with 363 * them, so we look first on the vnode_free_list. If it is empty, 364 * we next consider vnodes with referencing buffers on the 365 * vnode_hold_list. The toggle ensures that half the time we 366 * will use a buffer from the vnode_hold_list, and half the time 367 * we will allocate a new one unless the list has grown to twice 368 * the desired size. We are reticent to recycle vnodes from the 369 * vnode_hold_list because we will lose the identity of all its 370 * referencing buffers. 371 */ 372 toggle ^= 1; 373 if (numvnodes > 2 * desiredvnodes) 374 toggle = 0; 375 376 simple_lock(&vnode_free_list_slock); 377 s = splbio(); 378 if ((numvnodes < desiredvnodes) || 379 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 380 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 381 splx(s); 382 simple_unlock(&vnode_free_list_slock); 383 vp = pool_get(&vnode_pool, PR_WAITOK); 384 bzero((char *)vp, sizeof *vp); 385 simple_lock_init(&vp->v_interlock); 386 numvnodes++; 387 } else { 388 for (vp = TAILQ_FIRST(listhd); vp != NULLVP; 389 vp = TAILQ_NEXT(vp, v_freelist)) { 390 if (simple_lock_try(&vp->v_interlock)) { 391 if (VOP_ISLOCKED(vp) == 0) 392 break; 393 else 394 simple_unlock(&vp->v_interlock); 395 } 396 } 397 /* 398 * Unless this is a bad time of the month, at most 399 * the first NCPUS items on the free list are 400 * locked, so this is close enough to being empty. 401 */ 402 if (vp == NULL) { 403 splx(s); 404 simple_unlock(&vnode_free_list_slock); 405 tablefull("vnode"); 406 *vpp = 0; 407 return (ENFILE); 408 } 409 410#ifdef DIAGNOSTIC 411 if (vp->v_usecount) { 412 vprint("free vnode", vp); 413 panic("free vnode isn't"); 414 } 415#endif 416 417 TAILQ_REMOVE(listhd, vp, v_freelist); 418 vp->v_bioflag &= ~VBIOONFREELIST; 419 splx(s); 420 421 simple_unlock(&vnode_free_list_slock); 422 if (vp->v_type != VBAD) 423 vgonel(vp, p); 424 else 425 simple_unlock(&vp->v_interlock); 426#ifdef DIAGNOSTIC 427 if (vp->v_data) { 428 vprint("cleaned vnode", vp); 429 panic("cleaned vnode isn't"); 430 } 431 s = splbio(); 432 if (vp->v_numoutput) 433 panic("Clean vnode has pending I/O's"); 434 splx(s); 435#endif 436 vp->v_flag = 0; 437 vp->v_socket = 0; 438 } 439 vp->v_type = VNON; 440 cache_purge(vp); 441 vp->v_tag = tag; 442 vp->v_op = vops; 443 insmntque(vp, mp); 444 *vpp = vp; 445 vp->v_usecount = 1; 446 vp->v_data = 0; 447 simple_lock_init(&vp->v_uvm.u_obj.vmobjlock); 448 return (0); 449} 450 451/* 452 * Move a vnode from one mount queue to another. 453 */ 454void 455insmntque(struct vnode *vp, struct mount *mp) 456{ 457 simple_lock(&mntvnode_slock); 458 459 /* 460 * Delete from old mount point vnode list, if on one. 461 */ 462 if (vp->v_mount != NULL) 463 LIST_REMOVE(vp, v_mntvnodes); 464 /* 465 * Insert into list of vnodes for the new mount point, if available. 466 */ 467 if ((vp->v_mount = mp) != NULL) 468 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 469 470 simple_unlock(&mntvnode_slock); 471} 472 473/* 474 * Create a vnode for a block device. 475 * Used for root filesystem, argdev, and swap areas. 476 * Also used for memory file system special devices. 477 */ 478int 479bdevvp(dev_t dev, struct vnode **vpp) 480{ 481 return (getdevvp(dev, vpp, VBLK)); 482} 483 484/* 485 * Create a vnode for a character device. 486 * Used for console handling. 487 */ 488int 489cdevvp(dev_t dev, struct vnode **vpp) 490{ 491 return (getdevvp(dev, vpp, VCHR)); 492} 493 494/* 495 * Create a vnode for a device. 496 * Used by bdevvp (block device) for root file system etc., 497 * and by cdevvp (character device) for console. 498 */ 499int 500getdevvp(dev_t dev, struct vnode **vpp, enum vtype type) 501{ 502 struct vnode *vp; 503 struct vnode *nvp; 504 int error; 505 506 if (dev == NODEV) { 507 *vpp = NULLVP; 508 return (0); 509 } 510 error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); 511 if (error) { 512 *vpp = NULLVP; 513 return (error); 514 } 515 vp = nvp; 516 vp->v_type = type; 517 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 518 vput(vp); 519 vp = nvp; 520 } 521 *vpp = vp; 522 return (0); 523} 524 525/* 526 * Check to see if the new vnode represents a special device 527 * for which we already have a vnode (either because of 528 * bdevvp() or because of a different vnode representing 529 * the same block device). If such an alias exists, deallocate 530 * the existing contents and return the aliased vnode. The 531 * caller is responsible for filling it with its new contents. 532 */ 533struct vnode * 534checkalias(struct vnode *nvp, dev_t nvp_rdev, struct mount *mp) 535{ 536 struct proc *p = curproc; 537 struct vnode *vp; 538 struct vnode **vpp; 539 540 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 541 return (NULLVP); 542 543 vpp = &speclisth[SPECHASH(nvp_rdev)]; 544loop: 545 simple_lock(&spechash_slock); 546 for (vp = *vpp; vp; vp = vp->v_specnext) { 547 simple_lock(&vp->v_interlock); 548 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 549 simple_unlock(&vp->v_interlock); 550 continue; 551 } 552 /* 553 * Alias, but not in use, so flush it out. 554 */ 555 if (vp->v_usecount == 0) { 556 simple_unlock(&spechash_slock); 557 vgonel(vp, p); 558 goto loop; 559 } 560 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 561 simple_unlock(&spechash_slock); 562 goto loop; 563 } 564 break; 565 } 566 567 /* 568 * Common case is actually in the if statement 569 */ 570 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 571 MALLOC(nvp->v_specinfo, struct specinfo *, 572 sizeof(struct specinfo), M_VNODE, M_WAITOK); 573 nvp->v_rdev = nvp_rdev; 574 nvp->v_hashchain = vpp; 575 nvp->v_specnext = *vpp; 576 nvp->v_specmountpoint = NULL; 577 nvp->v_speclockf = NULL; 578 bzero(nvp->v_specbitmap, sizeof(nvp->v_specbitmap)); 579 simple_unlock(&spechash_slock); 580 *vpp = nvp; 581 if (vp != NULLVP) { 582 nvp->v_flag |= VALIASED; 583 vp->v_flag |= VALIASED; 584 vput(vp); 585 } 586 return (NULLVP); 587 } 588 589 /* 590 * This code is the uncommon case. It is called in case 591 * we found an alias that was VT_NON && vtype of VBLK 592 * This means we found a block device that was created 593 * using bdevvp. 594 * An example of such a vnode is the root partition device vnode 595 * created in ffs_mountroot. 596 * 597 * The vnodes created by bdevvp should not be aliased (why?). 598 */ 599 600 simple_unlock(&spechash_slock); 601 VOP_UNLOCK(vp, 0, p); 602 simple_lock(&vp->v_interlock); 603 vclean(vp, 0, p); 604 vp->v_op = nvp->v_op; 605 vp->v_tag = nvp->v_tag; 606 nvp->v_type = VNON; 607 insmntque(vp, mp); 608 return (vp); 609} 610 611/* 612 * Grab a particular vnode from the free list, increment its 613 * reference count and lock it. If the vnode lock bit is set, 614 * the vnode is being eliminated in vgone. In that case, we 615 * cannot grab it, so the process is awakened when the 616 * transition is completed, and an error code is returned to 617 * indicate that the vnode is no longer usable, possibly 618 * having been changed to a new file system type. 619 */ 620int 621vget(struct vnode *vp, int flags, struct proc *p) 622{ 623 int error, s, onfreelist; 624 625 /* 626 * If the vnode is in the process of being cleaned out for 627 * another use, we wait for the cleaning to finish and then 628 * return failure. Cleaning is determined by checking that 629 * the VXLOCK flag is set. 630 */ 631 if ((flags & LK_INTERLOCK) == 0) { 632 simple_lock(&vp->v_interlock); 633 flags |= LK_INTERLOCK; 634 } 635 636 if (vp->v_flag & VXLOCK) { 637 if (flags & LK_NOWAIT) { 638 simple_unlock(&vp->v_interlock); 639 return (EBUSY); 640 } 641 642 vp->v_flag |= VXWANT; 643 ltsleep(vp, PINOD | PNORELOCK, "vget", 0, &vp->v_interlock); 644 return (ENOENT); 645 } 646 647 onfreelist = vp->v_bioflag & VBIOONFREELIST; 648 if (vp->v_usecount == 0 && onfreelist) { 649 s = splbio(); 650 simple_lock(&vnode_free_list_slock); 651 if (vp->v_holdcnt > 0) 652 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 653 else 654 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 655 simple_unlock(&vnode_free_list_slock); 656 vp->v_bioflag &= ~VBIOONFREELIST; 657 splx(s); 658 } 659 660 vp->v_usecount++; 661 if (flags & LK_TYPE_MASK) { 662 if ((error = vn_lock(vp, flags, p)) != 0) { 663 vp->v_usecount--; 664 if (vp->v_usecount == 0 && onfreelist) 665 vputonfreelist(vp); 666 667 simple_unlock(&vp->v_interlock); 668 } 669 return (error); 670 } 671 672 simple_unlock(&vp->v_interlock); 673 674 return (0); 675} 676 677 678#ifdef DIAGNOSTIC 679/* 680 * Vnode reference. 681 */ 682void 683vref(struct vnode *vp) 684{ 685 simple_lock(&vp->v_interlock); 686 if (vp->v_usecount == 0) 687 panic("vref used where vget required"); 688 vp->v_usecount++; 689 simple_unlock(&vp->v_interlock); 690} 691#endif /* DIAGNOSTIC */ 692 693void 694vputonfreelist(struct vnode *vp) 695{ 696 int s; 697 struct freelst *lst; 698 699 s = splbio(); 700#ifdef DIAGNOSTIC 701 if (vp->v_usecount != 0) 702 panic("Use count is not zero!"); 703 704 if (vp->v_bioflag & VBIOONFREELIST) { 705 vprint("vnode already on free list: ", vp); 706 panic("vnode already on free list"); 707 } 708#endif 709 710 vp->v_bioflag |= VBIOONFREELIST; 711 712 if (vp->v_holdcnt > 0) 713 lst = &vnode_hold_list; 714 else 715 lst = &vnode_free_list; 716 717 if (vp->v_type == VBAD) 718 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 719 else 720 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 721 722 splx(s); 723} 724 725/* 726 * vput(), just unlock and vrele() 727 */ 728void 729vput(struct vnode *vp) 730{ 731 struct proc *p = curproc; 732 733#ifdef DIAGNOSTIC 734 if (vp == NULL) 735 panic("vput: null vp"); 736#endif 737 simple_lock(&vp->v_interlock); 738 739#ifdef DIAGNOSTIC 740 if (vp->v_usecount == 0) { 741 vprint("vput: bad ref count", vp); 742 panic("vput: ref cnt"); 743 } 744#endif 745 vp->v_usecount--; 746 if (vp->v_usecount > 0) { 747 simple_unlock(&vp->v_interlock); 748 VOP_UNLOCK(vp, 0, p); 749 return; 750 } 751 752#ifdef DIAGNOSTIC 753 if (vp->v_writecount != 0) { 754 vprint("vput: bad writecount", vp); 755 panic("vput: v_writecount != 0"); 756 } 757#endif 758 simple_unlock(&vp->v_interlock); 759 760 VOP_INACTIVE(vp, p); 761 762 simple_lock(&vp->v_interlock); 763 764 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 765 vputonfreelist(vp); 766 767 simple_unlock(&vp->v_interlock); 768} 769 770/* 771 * Vnode release - use for active VNODES. 772 * If count drops to zero, call inactive routine and return to freelist. 773 */ 774void 775vrele(struct vnode *vp) 776{ 777 struct proc *p = curproc; 778 779#ifdef DIAGNOSTIC 780 if (vp == NULL) 781 panic("vrele: null vp"); 782#endif 783 simple_lock(&vp->v_interlock); 784#ifdef DIAGNOSTIC 785 if (vp->v_usecount == 0) { 786 vprint("vrele: bad ref count", vp); 787 panic("vrele: ref cnt"); 788 } 789#endif 790 vp->v_usecount--; 791 if (vp->v_usecount > 0) { 792 simple_unlock(&vp->v_interlock); 793 return; 794 } 795 796#ifdef DIAGNOSTIC 797 if (vp->v_writecount != 0) { 798 vprint("vrele: bad writecount", vp); 799 panic("vrele: v_writecount != 0"); 800 } 801#endif 802 803 if (vn_lock(vp, LK_EXCLUSIVE|LK_INTERLOCK, p)) { 804#ifdef DIAGNOSTIC 805 vprint("vrele: cannot lock", vp); 806#endif 807 return; 808 } 809 810 VOP_INACTIVE(vp, p); 811 812 simple_lock(&vp->v_interlock); 813 814 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 815 vputonfreelist(vp); 816 817 simple_unlock(&vp->v_interlock); 818} 819 820void vhold(struct vnode *vp); 821 822/* 823 * Page or buffer structure gets a reference. 824 */ 825void 826vhold(struct vnode *vp) 827{ 828 /* 829 * If it is on the freelist and the hold count is currently 830 * zero, move it to the hold list. 831 */ 832 simple_lock(&vp->v_interlock); 833 if ((vp->v_bioflag & VBIOONFREELIST) && 834 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 835 simple_lock(&vnode_free_list_slock); 836 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 837 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 838 simple_unlock(&vnode_free_list_slock); 839 } 840 vp->v_holdcnt++; 841 simple_unlock(&vp->v_interlock); 842} 843 844/* 845 * Remove any vnodes in the vnode table belonging to mount point mp. 846 * 847 * If MNT_NOFORCE is specified, there should not be any active ones, 848 * return error if any are found (nb: this is a user error, not a 849 * system error). If MNT_FORCE is specified, detach any active vnodes 850 * that are found. 851 */ 852#ifdef DEBUG 853int busyprt = 0; /* print out busy vnodes */ 854struct ctldebug debug1 = { "busyprt", &busyprt }; 855#endif 856 857int 858vfs_mount_foreach_vnode(struct mount *mp, 859 int (*func)(struct vnode *, void *), void *arg) { 860 struct vnode *vp, *nvp; 861 int error = 0; 862 863 simple_lock(&mntvnode_slock); 864loop: 865 for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; vp = nvp) { 866 if (vp->v_mount != mp) 867 goto loop; 868 nvp = LIST_NEXT(vp, v_mntvnodes); 869 simple_lock(&vp->v_interlock); 870 simple_unlock(&mntvnode_slock); 871 872 error = func(vp, arg); 873 874 simple_lock(&mntvnode_slock); 875 876 if (error != 0) 877 break; 878 } 879 simple_unlock(&mntvnode_slock); 880 881 return (error); 882} 883 884struct vflush_args { 885 struct vnode *skipvp; 886 int busy; 887 int flags; 888}; 889 890int 891vflush_vnode(struct vnode *vp, void *arg) { 892 struct vflush_args *va = arg; 893 struct proc *p = curproc; 894 895 if (vp == va->skipvp) { 896 simple_unlock(&vp->v_interlock); 897 return (0); 898 } 899 900 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 901 simple_unlock(&vp->v_interlock); 902 return (0); 903 } 904 905 /* 906 * If WRITECLOSE is set, only flush out regular file 907 * vnodes open for writing. 908 */ 909 if ((va->flags & WRITECLOSE) && 910 (vp->v_writecount == 0 || vp->v_type != VREG)) { 911 simple_unlock(&vp->v_interlock); 912 return (0); 913 } 914 915 /* 916 * With v_usecount == 0, all we need to do is clear 917 * out the vnode data structures and we are done. 918 */ 919 if (vp->v_usecount == 0) { 920 vgonel(vp, p); 921 return (0); 922 } 923 924 /* 925 * If FORCECLOSE is set, forcibly close the vnode. 926 * For block or character devices, revert to an 927 * anonymous device. For all other files, just kill them. 928 */ 929 if (va->flags & FORCECLOSE) { 930 if (vp->v_type != VBLK && vp->v_type != VCHR) { 931 vgonel(vp, p); 932 } else { 933 vclean(vp, 0, p); 934 vp->v_op = spec_vnodeop_p; 935 insmntque(vp, (struct mount *)0); 936 } 937 return (0); 938 } 939 940#ifdef DEBUG 941 if (busyprt) 942 vprint("vflush: busy vnode", vp); 943#endif 944 simple_unlock(&vp->v_interlock); 945 va->busy++; 946 return (0); 947} 948 949int 950vflush(struct mount *mp, struct vnode *skipvp, int flags) 951{ 952 struct vflush_args va; 953 va.skipvp = skipvp; 954 va.busy = 0; 955 va.flags = flags; 956 957 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 958 959 if (va.busy) 960 return (EBUSY); 961 return (0); 962} 963 964/* 965 * Disassociate the underlying file system from a vnode. 966 * The vnode interlock is held on entry. 967 */ 968void 969vclean(struct vnode *vp, int flags, struct proc *p) 970{ 971 int active; 972 973 /* 974 * Check to see if the vnode is in use. 975 * If so we have to reference it before we clean it out 976 * so that its count cannot fall to zero and generate a 977 * race against ourselves to recycle it. 978 */ 979 if ((active = vp->v_usecount) != 0) 980 vp->v_usecount++; 981 982 /* 983 * Prevent the vnode from being recycled or 984 * brought into use while we clean it out. 985 */ 986 if (vp->v_flag & VXLOCK) 987 panic("vclean: deadlock"); 988 vp->v_flag |= VXLOCK; 989 /* 990 * Even if the count is zero, the VOP_INACTIVE routine may still 991 * have the object locked while it cleans it out. The VOP_LOCK 992 * ensures that the VOP_INACTIVE routine is done with its work. 993 * For active vnodes, it ensures that no other activity can 994 * occur while the underlying object is being cleaned out. 995 */ 996 VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK, p); 997 998 /* 999 * Clean out any VM data associated with the vnode. 1000 */ 1001 uvm_vnp_terminate(vp); 1002 /* 1003 * Clean out any buffers associated with the vnode. 1004 */ 1005 if (flags & DOCLOSE) 1006 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 1007 /* 1008 * If purging an active vnode, it must be closed and 1009 * deactivated before being reclaimed. Note that the 1010 * VOP_INACTIVE will unlock the vnode 1011 */ 1012 if (active) { 1013 if (flags & DOCLOSE) 1014 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 1015 VOP_INACTIVE(vp, p); 1016 } else { 1017 /* 1018 * Any other processes trying to obtain this lock must first 1019 * wait for VXLOCK to clear, then call the new lock operation. 1020 */ 1021 VOP_UNLOCK(vp, 0, p); 1022 } 1023 1024 /* 1025 * Reclaim the vnode. 1026 */ 1027 if (VOP_RECLAIM(vp, p)) 1028 panic("vclean: cannot reclaim"); 1029 if (active) { 1030 simple_lock(&vp->v_interlock); 1031 1032 vp->v_usecount--; 1033 if (vp->v_usecount == 0) { 1034 if (vp->v_holdcnt > 0) 1035 panic("vclean: not clean"); 1036 vputonfreelist(vp); 1037 } 1038 1039 simple_unlock(&vp->v_interlock); 1040 } 1041 cache_purge(vp); 1042 1043 /* 1044 * Done with purge, notify sleepers of the grim news. 1045 */ 1046 vp->v_op = dead_vnodeop_p; 1047 simple_lock(&vp->v_selectinfo.vsi_lock); 1048 VN_KNOTE(vp, NOTE_REVOKE); 1049 simple_unlock(&vp->v_selectinfo.vsi_lock); 1050 vp->v_tag = VT_NON; 1051 vp->v_flag &= ~VXLOCK; 1052#ifdef VFSDEBUG 1053 vp->v_flag &= ~VLOCKSWORK; 1054#endif 1055 if (vp->v_flag & VXWANT) { 1056 vp->v_flag &= ~VXWANT; 1057 wakeup(vp); 1058 } 1059} 1060 1061/* 1062 * Recycle an unused vnode to the front of the free list. 1063 * Release the passed interlock if the vnode will be recycled. 1064 */ 1065int 1066vrecycle(struct vnode *vp, struct simplelock *inter_lkp, struct proc *p) 1067{ 1068 simple_lock(&vp->v_interlock); 1069 if (vp->v_usecount == 0) { 1070 if (inter_lkp) 1071 simple_unlock(inter_lkp); 1072 vgonel(vp, p); 1073 return (1); 1074 } 1075 simple_unlock(&vp->v_interlock); 1076 return (0); 1077} 1078 1079/* 1080 * Eliminate all activity associated with a vnode 1081 * in preparation for reuse. 1082 */ 1083void 1084vgone(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 */ 1095void 1096vgonel(struct vnode *vp, struct proc *p) 1097{ 1098 struct vnode *vq; 1099 struct vnode *vx; 1100 struct mount *mp; 1101 int flags; 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 ltsleep(vp, PINOD | PNORELOCK, "vgone", 0, &vp->v_interlock); 1110 return; 1111 } 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 1158 /* 1159 * If we have a mount point associated with the vnode, we must 1160 * flush it out now, as to not leave a dangling zombie mount 1161 * point laying around in VFS. 1162 */ 1163 mp = vp->v_specmountpoint; 1164 if (mp != NULL) { 1165 if (!vfs_busy(mp, VB_WRITE|VB_WAIT)) { 1166 flags = MNT_FORCE | MNT_DOOMED; 1167 dounmount(mp, flags, p, NULL); 1168 } 1169 } 1170 1171 FREE(vp->v_specinfo, M_VNODE); 1172 vp->v_specinfo = NULL; 1173 } 1174 /* 1175 * If it is on the freelist and not already at the head, 1176 * move it to the head of the list. 1177 */ 1178 vp->v_type = VBAD; 1179 1180 /* 1181 * Move onto the free list, unless we were called from 1182 * getnewvnode and we're not on any free list 1183 */ 1184 if (vp->v_usecount == 0 && 1185 (vp->v_bioflag & VBIOONFREELIST)) { 1186 int s; 1187 1188 simple_lock(&vnode_free_list_slock); 1189 s = splbio(); 1190 1191 if (vp->v_holdcnt > 0) 1192 panic("vgonel: not clean"); 1193 1194 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1195 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1196 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1197 } 1198 splx(s); 1199 simple_unlock(&vnode_free_list_slock); 1200 } 1201} 1202 1203/* 1204 * Lookup a vnode by device number. 1205 */ 1206int 1207vfinddev(dev_t dev, enum vtype type, struct vnode **vpp) 1208{ 1209 struct vnode *vp; 1210 int rc =0; 1211 1212 simple_lock(&spechash_slock); 1213 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1214 if (dev != vp->v_rdev || type != vp->v_type) 1215 continue; 1216 *vpp = vp; 1217 rc = 1; 1218 break; 1219 } 1220 simple_unlock(&spechash_slock); 1221 return (rc); 1222} 1223 1224/* 1225 * Revoke all the vnodes corresponding to the specified minor number 1226 * range (endpoints inclusive) of the specified major. 1227 */ 1228void 1229vdevgone(int maj, int minl, int minh, enum vtype type) 1230{ 1231 struct vnode *vp; 1232 int mn; 1233 1234 for (mn = minl; mn <= minh; mn++) 1235 if (vfinddev(makedev(maj, mn), type, &vp)) 1236 VOP_REVOKE(vp, REVOKEALL); 1237} 1238 1239/* 1240 * Calculate the total number of references to a special device. 1241 */ 1242int 1243vcount(struct vnode *vp) 1244{ 1245 struct vnode *vq, *vnext; 1246 int count; 1247 1248loop: 1249 if ((vp->v_flag & VALIASED) == 0) 1250 return (vp->v_usecount); 1251 simple_lock(&spechash_slock); 1252 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1253 vnext = vq->v_specnext; 1254 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1255 continue; 1256 /* 1257 * Alias, but not in use, so flush it out. 1258 */ 1259 if (vq->v_usecount == 0 && vq != vp) { 1260 simple_unlock(&spechash_slock); 1261 vgone(vq); 1262 goto loop; 1263 } 1264 count += vq->v_usecount; 1265 } 1266 simple_unlock(&spechash_slock); 1267 return (count); 1268} 1269 1270#ifdef DIAGNOSTIC 1271/* 1272 * Print out a description of a vnode. 1273 */ 1274static char *typename[] = 1275 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1276 1277void 1278vprint(char *label, struct vnode *vp) 1279{ 1280 char buf[64]; 1281 1282 if (label != NULL) 1283 printf("%s: ", label); 1284 printf("%p, type %s, use %u, write %u, hold %u,", 1285 vp, typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1286 vp->v_holdcnt); 1287 buf[0] = '\0'; 1288 if (vp->v_flag & VROOT) 1289 strlcat(buf, "|VROOT", sizeof buf); 1290 if (vp->v_flag & VTEXT) 1291 strlcat(buf, "|VTEXT", sizeof buf); 1292 if (vp->v_flag & VSYSTEM) 1293 strlcat(buf, "|VSYSTEM", sizeof buf); 1294 if (vp->v_flag & VXLOCK) 1295 strlcat(buf, "|VXLOCK", sizeof buf); 1296 if (vp->v_flag & VXWANT) 1297 strlcat(buf, "|VXWANT", sizeof buf); 1298 if (vp->v_bioflag & VBIOWAIT) 1299 strlcat(buf, "|VBIOWAIT", sizeof buf); 1300 if (vp->v_bioflag & VBIOONFREELIST) 1301 strlcat(buf, "|VBIOONFREELIST", sizeof buf); 1302 if (vp->v_bioflag & VBIOONSYNCLIST) 1303 strlcat(buf, "|VBIOONSYNCLIST", sizeof buf); 1304 if (vp->v_flag & VALIASED) 1305 strlcat(buf, "|VALIASED", sizeof buf); 1306 if (buf[0] != '\0') 1307 printf(" flags (%s)", &buf[1]); 1308 if (vp->v_data == NULL) { 1309 printf("\n"); 1310 } else { 1311 printf("\n\t"); 1312 VOP_PRINT(vp); 1313 } 1314} 1315#endif /* DIAGNOSTIC */ 1316 1317#ifdef DEBUG 1318/* 1319 * List all of the locked vnodes in the system. 1320 * Called when debugging the kernel. 1321 */ 1322void 1323printlockedvnodes(void) 1324{ 1325 struct mount *mp, *nmp; 1326 struct vnode *vp; 1327 1328 printf("Locked vnodes\n"); 1329 1330 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1331 mp = nmp) { 1332 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) { 1333 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1334 continue; 1335 } 1336 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 1337 if (VOP_ISLOCKED(vp)) 1338 vprint((char *)0, vp); 1339 } 1340 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1341 vfs_unbusy(mp); 1342 } 1343 1344} 1345#endif 1346 1347/* 1348 * Top level filesystem related information gathering. 1349 */ 1350int 1351vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1352 size_t newlen, struct proc *p) 1353{ 1354 struct vfsconf *vfsp; 1355 1356 /* all sysctl names at this level are at least name and field */ 1357 if (namelen < 2) 1358 return (ENOTDIR); /* overloaded */ 1359 1360 if (name[0] != VFS_GENERIC) { 1361 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1362 if (vfsp->vfc_typenum == name[0]) 1363 break; 1364 1365 if (vfsp == NULL) 1366 return (EOPNOTSUPP); 1367 1368 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1369 oldp, oldlenp, newp, newlen, p)); 1370 } 1371 1372 switch (name[1]) { 1373 case VFS_MAXTYPENUM: 1374 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1375 1376 case VFS_CONF: 1377 if (namelen < 3) 1378 return (ENOTDIR); /* overloaded */ 1379 1380 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1381 if (vfsp->vfc_typenum == name[2]) 1382 break; 1383 1384 if (vfsp == NULL) 1385 return (EOPNOTSUPP); 1386 1387 return (sysctl_rdstruct(oldp, oldlenp, newp, vfsp, 1388 sizeof(struct vfsconf))); 1389 } 1390 1391 return (EOPNOTSUPP); 1392} 1393 1394int kinfo_vdebug = 1; 1395int kinfo_vgetfailed; 1396#define KINFO_VNODESLOP 10 1397/* 1398 * Dump vnode list (via sysctl). 1399 * Copyout address of vnode followed by vnode. 1400 */ 1401/* ARGSUSED */ 1402int 1403sysctl_vnode(char *where, size_t *sizep, struct proc *p) 1404{ 1405 struct mount *mp, *nmp; 1406 struct vnode *vp, *nvp; 1407 char *bp = where, *savebp; 1408 char *ewhere; 1409 int error; 1410 1411 if (where == NULL) { 1412 *sizep = (numvnodes + KINFO_VNODESLOP) * sizeof(struct e_vnode); 1413 return (0); 1414 } 1415 ewhere = where + *sizep; 1416 1417 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1418 mp = nmp) { 1419 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) { 1420 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1421 continue; 1422 } 1423 savebp = bp; 1424again: 1425 for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; 1426 vp = nvp) { 1427 /* 1428 * Check that the vp is still associated with 1429 * this filesystem. RACE: could have been 1430 * recycled onto the same filesystem. 1431 */ 1432 if (vp->v_mount != mp) { 1433 simple_unlock(&mntvnode_slock); 1434 if (kinfo_vdebug) 1435 printf("kinfo: vp changed\n"); 1436 bp = savebp; 1437 goto again; 1438 } 1439 nvp = LIST_NEXT(vp, v_mntvnodes); 1440 if (bp + sizeof(struct e_vnode) > ewhere) { 1441 simple_unlock(&mntvnode_slock); 1442 *sizep = bp - where; 1443 vfs_unbusy(mp); 1444 return (ENOMEM); 1445 } 1446 if ((error = copyout(&vp, 1447 &((struct e_vnode *)bp)->vptr, 1448 sizeof(struct vnode *))) || 1449 (error = copyout(vp, 1450 &((struct e_vnode *)bp)->vnode, 1451 sizeof(struct vnode)))) { 1452 vfs_unbusy(mp); 1453 return (error); 1454 } 1455 bp += sizeof(struct e_vnode); 1456 simple_lock(&mntvnode_slock); 1457 } 1458 1459 simple_unlock(&mntvnode_slock); 1460 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1461 vfs_unbusy(mp); 1462 } 1463 1464 *sizep = bp - where; 1465 1466 return (0); 1467} 1468 1469/* 1470 * Check to see if a filesystem is mounted on a block device. 1471 */ 1472int 1473vfs_mountedon(struct vnode *vp) 1474{ 1475 struct vnode *vq; 1476 int error = 0; 1477 1478 if (vp->v_specmountpoint != NULL) 1479 return (EBUSY); 1480 if (vp->v_flag & VALIASED) { 1481 simple_lock(&spechash_slock); 1482 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1483 if (vq->v_rdev != vp->v_rdev || 1484 vq->v_type != vp->v_type) 1485 continue; 1486 if (vq->v_specmountpoint != NULL) { 1487 error = EBUSY; 1488 break; 1489 } 1490 } 1491 simple_unlock(&spechash_slock); 1492 } 1493 return (error); 1494} 1495 1496/* 1497 * Build hash lists of net addresses and hang them off the mount point. 1498 * Called by ufs_mount() to set up the lists of export addresses. 1499 */ 1500int 1501vfs_hang_addrlist(struct mount *mp, struct netexport *nep, 1502 struct export_args *argp) 1503{ 1504 struct netcred *np; 1505 struct radix_node_head *rnh; 1506 int i; 1507 struct radix_node *rn; 1508 struct sockaddr *saddr, *smask = 0; 1509 struct domain *dom; 1510 int error; 1511 1512 if (argp->ex_addrlen == 0) { 1513 if (mp->mnt_flag & MNT_DEFEXPORTED) 1514 return (EPERM); 1515 np = &nep->ne_defexported; 1516 np->netc_exflags = argp->ex_flags; 1517 np->netc_anon = argp->ex_anon; 1518 np->netc_anon.cr_ref = 1; 1519 mp->mnt_flag |= MNT_DEFEXPORTED; 1520 return (0); 1521 } 1522 if (argp->ex_addrlen > MLEN || argp->ex_masklen > MLEN || 1523 argp->ex_addrlen < 0 || argp->ex_masklen < 0) 1524 return (EINVAL); 1525 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1526 np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK); 1527 bzero(np, i); 1528 saddr = (struct sockaddr *)(np + 1); 1529 error = copyin(argp->ex_addr, saddr, argp->ex_addrlen); 1530 if (error) 1531 goto out; 1532 if (saddr->sa_len > argp->ex_addrlen) 1533 saddr->sa_len = argp->ex_addrlen; 1534 if (argp->ex_masklen) { 1535 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1536 error = copyin(argp->ex_mask, smask, argp->ex_masklen); 1537 if (error) 1538 goto out; 1539 if (smask->sa_len > argp->ex_masklen) 1540 smask->sa_len = argp->ex_masklen; 1541 } 1542 i = saddr->sa_family; 1543 if (i < 0 || i > AF_MAX) { 1544 error = EINVAL; 1545 goto out; 1546 } 1547 if ((rnh = nep->ne_rtable[i]) == 0) { 1548 /* 1549 * Seems silly to initialize every AF when most are not 1550 * used, do so on demand here 1551 */ 1552 for (dom = domains; dom; dom = dom->dom_next) 1553 if (dom->dom_family == i && dom->dom_rtattach) { 1554 dom->dom_rtattach((void **)&nep->ne_rtable[i], 1555 dom->dom_rtoffset); 1556 break; 1557 } 1558 if ((rnh = nep->ne_rtable[i]) == 0) { 1559 error = ENOBUFS; 1560 goto out; 1561 } 1562 } 1563 rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, 1564 np->netc_rnodes); 1565 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1566 error = EPERM; 1567 goto out; 1568 } 1569 np->netc_exflags = argp->ex_flags; 1570 np->netc_anon = argp->ex_anon; 1571 np->netc_anon.cr_ref = 1; 1572 return (0); 1573out: 1574 free(np, M_NETADDR); 1575 return (error); 1576} 1577 1578/* ARGSUSED */ 1579int 1580vfs_free_netcred(struct radix_node *rn, void *w) 1581{ 1582 struct radix_node_head *rnh = (struct radix_node_head *)w; 1583 1584 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh, NULL); 1585 free(rn, M_NETADDR); 1586 return (0); 1587} 1588 1589/* 1590 * Free the net address hash lists that are hanging off the mount points. 1591 */ 1592void 1593vfs_free_addrlist(struct netexport *nep) 1594{ 1595 int i; 1596 struct radix_node_head *rnh; 1597 1598 for (i = 0; i <= AF_MAX; i++) 1599 if ((rnh = nep->ne_rtable[i]) != NULL) { 1600 (*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh); 1601 free(rnh, M_RTABLE); 1602 nep->ne_rtable[i] = 0; 1603 } 1604} 1605 1606int 1607vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp) 1608{ 1609 int error; 1610 1611 if (argp->ex_flags & MNT_DELEXPORT) { 1612 vfs_free_addrlist(nep); 1613 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1614 } 1615 if (argp->ex_flags & MNT_EXPORTED) { 1616 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1617 return (error); 1618 mp->mnt_flag |= MNT_EXPORTED; 1619 } 1620 return (0); 1621} 1622 1623struct netcred * 1624vfs_export_lookup(struct mount *mp, struct netexport *nep, struct mbuf *nam) 1625{ 1626 struct netcred *np; 1627 struct radix_node_head *rnh; 1628 struct sockaddr *saddr; 1629 1630 np = NULL; 1631 if (mp->mnt_flag & MNT_EXPORTED) { 1632 /* 1633 * Lookup in the export list first. 1634 */ 1635 if (nam != NULL) { 1636 saddr = mtod(nam, struct sockaddr *); 1637 rnh = nep->ne_rtable[saddr->sa_family]; 1638 if (rnh != NULL) { 1639 np = (struct netcred *) 1640 (*rnh->rnh_matchaddr)((caddr_t)saddr, 1641 rnh); 1642 if (np && np->netc_rnodes->rn_flags & RNF_ROOT) 1643 np = NULL; 1644 } 1645 } 1646 /* 1647 * If no address match, use the default if it exists. 1648 */ 1649 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1650 np = &nep->ne_defexported; 1651 } 1652 return (np); 1653} 1654 1655/* 1656 * Do the usual access checking. 1657 * file_mode, uid and gid are from the vnode in question, 1658 * while acc_mode and cred are from the VOP_ACCESS parameter list 1659 */ 1660int 1661vaccess(mode_t file_mode, uid_t uid, gid_t gid, 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 */ 1709void 1710vfs_unmountall(void) 1711{ 1712 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 ((vfs_busy(mp, VB_WRITE|VB_NOWAIT)) != 0) 1721 continue; 1722 if ((error = dounmount(mp, MNT_FORCE, curproc, NULL)) != 0) { 1723 printf("unmount of %s failed with error %d\n", 1724 mp->mnt_stat.f_mntonname, error); 1725 allerror = 1; 1726 } 1727 } 1728 1729 if (allerror) { 1730 printf("WARNING: some file systems would not unmount\n"); 1731 if (again) { 1732 printf("retrying\n"); 1733 again = 0; 1734 goto retry; 1735 } 1736 } 1737} 1738 1739/* 1740 * Sync and unmount file systems before shutting down. 1741 */ 1742void 1743vfs_shutdown(void) 1744{ 1745#ifdef ACCOUNTING 1746 extern void acct_shutdown(void); 1747 1748 acct_shutdown(); 1749#endif 1750 1751 /* XXX Should suspend scheduling. */ 1752 (void) spl0(); 1753 1754 printf("syncing disks... "); 1755 1756 if (panicstr == 0) { 1757 /* Sync before unmount, in case we hang on something. */ 1758 sys_sync(&proc0, (void *)0, (register_t *)0); 1759 1760 /* Unmount file systems. */ 1761 vfs_unmountall(); 1762 } 1763 1764 if (vfs_syncwait(1)) 1765 printf("giving up\n"); 1766 else 1767 printf("done\n"); 1768} 1769 1770/* 1771 * perform sync() operation and wait for buffers to flush. 1772 * assumtions: called w/ scheduler disabled and physical io enabled 1773 * for now called at spl0() XXX 1774 */ 1775int 1776vfs_syncwait(int verbose) 1777{ 1778 struct buf *bp; 1779 int iter, nbusy, dcount, s; 1780 struct proc *p; 1781 1782 p = curproc? curproc : &proc0; 1783 sys_sync(p, (void *)0, (register_t *)0); 1784 1785 /* Wait for sync to finish. */ 1786 dcount = 10000; 1787 for (iter = 0; iter < 20; iter++) { 1788 nbusy = 0; 1789 for (bp = &buf[nbuf]; --bp >= buf; ) { 1790 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1791 nbusy++; 1792 /* 1793 * With soft updates, some buffers that are 1794 * written will be remarked as dirty until other 1795 * buffers are written. 1796 */ 1797 if (bp->b_flags & B_DELWRI) { 1798 s = splbio(); 1799 bremfree(bp); 1800 bp->b_flags |= B_BUSY; 1801 splx(s); 1802 nbusy++; 1803 bawrite(bp); 1804 if (dcount-- <= 0) { 1805 if (verbose) 1806 printf("softdep "); 1807 return 1; 1808 } 1809 } 1810 } 1811 if (nbusy == 0) 1812 break; 1813 if (verbose) 1814 printf("%d ", nbusy); 1815 DELAY(40000 * iter); 1816 } 1817 1818 return nbusy; 1819} 1820 1821/* 1822 * posix file system related system variables. 1823 */ 1824int 1825fs_posix_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, 1826 void *newp, size_t newlen, struct proc *p) 1827{ 1828 /* all sysctl names at this level are terminal */ 1829 if (namelen != 1) 1830 return (ENOTDIR); 1831 1832 switch (name[0]) { 1833 case FS_POSIX_SETUID: 1834 if (newp && securelevel > 0) 1835 return (EPERM); 1836 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1837 default: 1838 return (EOPNOTSUPP); 1839 } 1840 /* NOTREACHED */ 1841} 1842 1843/* 1844 * file system related system variables. 1845 */ 1846int 1847fs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1848 size_t newlen, struct proc *p) 1849{ 1850 sysctlfn *fn; 1851 1852 switch (name[0]) { 1853 case FS_POSIX: 1854 fn = fs_posix_sysctl; 1855 break; 1856 default: 1857 return (EOPNOTSUPP); 1858 } 1859 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1860} 1861 1862 1863/* 1864 * Routines dealing with vnodes and buffers 1865 */ 1866 1867/* 1868 * Wait for all outstanding I/Os to complete 1869 * 1870 * Manipulates v_numoutput. Must be called at splbio() 1871 */ 1872int 1873vwaitforio(struct vnode *vp, int slpflag, char *wmesg, int timeo) 1874{ 1875 int error = 0; 1876 1877 splassert(IPL_BIO); 1878 1879 while (vp->v_numoutput) { 1880 vp->v_bioflag |= VBIOWAIT; 1881 error = tsleep(&vp->v_numoutput, 1882 slpflag | (PRIBIO + 1), wmesg, timeo); 1883 if (error) 1884 break; 1885 } 1886 1887 return (error); 1888} 1889 1890/* 1891 * Update outstanding I/O count and do wakeup if requested. 1892 * 1893 * Manipulates v_numoutput. Must be called at splbio() 1894 */ 1895void 1896vwakeup(struct vnode *vp) 1897{ 1898 splassert(IPL_BIO); 1899 1900 if (vp != NULL) { 1901 if (vp->v_numoutput-- == 0) 1902 panic("vwakeup: neg numoutput"); 1903 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1904 vp->v_bioflag &= ~VBIOWAIT; 1905 wakeup(&vp->v_numoutput); 1906 } 1907 } 1908} 1909 1910/* 1911 * Flush out and invalidate all buffers associated with a vnode. 1912 * Called with the underlying object locked. 1913 */ 1914int 1915vinvalbuf(struct vnode *vp, int flags, struct ucred *cred, struct proc *p, 1916 int slpflag, int slptimeo) 1917{ 1918 struct buf *bp; 1919 struct buf *nbp, *blist; 1920 int s, error; 1921 1922#ifdef VFSDEBUG 1923 if ((vp->v_flag & VLOCKSWORK) && !VOP_ISLOCKED(vp)) 1924 panic("vinvalbuf(): vp isn't locked"); 1925#endif 1926 1927 if (flags & V_SAVE) { 1928 s = splbio(); 1929 vwaitforio(vp, 0, "vinvalbuf", 0); 1930 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1931 splx(s); 1932 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1933 return (error); 1934 s = splbio(); 1935 if (vp->v_numoutput > 0 || 1936 !LIST_EMPTY(&vp->v_dirtyblkhd)) 1937 panic("vinvalbuf: dirty bufs"); 1938 } 1939 splx(s); 1940 } 1941loop: 1942 s = splbio(); 1943 for (;;) { 1944 if ((blist = LIST_FIRST(&vp->v_cleanblkhd)) && 1945 (flags & V_SAVEMETA)) 1946 while (blist && blist->b_lblkno < 0) 1947 blist = LIST_NEXT(blist, b_vnbufs); 1948 if (blist == NULL && 1949 (blist = LIST_FIRST(&vp->v_dirtyblkhd)) && 1950 (flags & V_SAVEMETA)) 1951 while (blist && blist->b_lblkno < 0) 1952 blist = LIST_NEXT(blist, b_vnbufs); 1953 if (!blist) 1954 break; 1955 1956 for (bp = blist; bp; bp = nbp) { 1957 nbp = LIST_NEXT(bp, b_vnbufs); 1958 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1959 continue; 1960 if (bp->b_flags & B_BUSY) { 1961 bp->b_flags |= B_WANTED; 1962 error = tsleep(bp, slpflag | (PRIBIO + 1), 1963 "vinvalbuf", slptimeo); 1964 if (error) { 1965 splx(s); 1966 return (error); 1967 } 1968 break; 1969 } 1970 bremfree(bp); 1971 bp->b_flags |= B_BUSY; 1972 /* 1973 * XXX Since there are no node locks for NFS, I believe 1974 * there is a slight chance that a delayed write will 1975 * occur while sleeping just above, so check for it. 1976 */ 1977 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1978 splx(s); 1979 (void) VOP_BWRITE(bp); 1980 goto loop; 1981 } 1982 bp->b_flags |= B_INVAL; 1983 brelse(bp); 1984 } 1985 } 1986 if (!(flags & V_SAVEMETA) && 1987 (!LIST_EMPTY(&vp->v_dirtyblkhd) || !LIST_EMPTY(&vp->v_cleanblkhd))) 1988 panic("vinvalbuf: flush failed"); 1989 splx(s); 1990 return (0); 1991} 1992 1993void 1994vflushbuf(struct vnode *vp, int sync) 1995{ 1996 struct buf *bp, *nbp; 1997 int s; 1998 1999loop: 2000 s = splbio(); 2001 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); 2002 bp != LIST_END(&vp->v_dirtyblkhd); bp = nbp) { 2003 nbp = LIST_NEXT(bp, b_vnbufs); 2004 if ((bp->b_flags & B_BUSY)) 2005 continue; 2006 if ((bp->b_flags & B_DELWRI) == 0) 2007 panic("vflushbuf: not dirty"); 2008 bremfree(bp); 2009 bp->b_flags |= B_BUSY; 2010 splx(s); 2011 /* 2012 * Wait for I/O associated with indirect blocks to complete, 2013 * since there is no way to quickly wait for them below. 2014 */ 2015 if (bp->b_vp == vp || sync == 0) 2016 (void) bawrite(bp); 2017 else 2018 (void) bwrite(bp); 2019 goto loop; 2020 } 2021 if (sync == 0) { 2022 splx(s); 2023 return; 2024 } 2025 vwaitforio(vp, 0, "vflushbuf", 0); 2026 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 2027 splx(s); 2028#ifdef DIAGNOSTIC 2029 vprint("vflushbuf: dirty", vp); 2030#endif 2031 goto loop; 2032 } 2033 splx(s); 2034} 2035 2036/* 2037 * Associate a buffer with a vnode. 2038 * 2039 * Manipulates buffer vnode queues. Must be called at splbio(). 2040 */ 2041void 2042bgetvp(struct vnode *vp, struct buf *bp) 2043{ 2044 splassert(IPL_BIO); 2045 2046 2047 if (bp->b_vp) 2048 panic("bgetvp: not free"); 2049 vhold(vp); 2050 bp->b_vp = vp; 2051 if (vp->v_type == VBLK || vp->v_type == VCHR) 2052 bp->b_dev = vp->v_rdev; 2053 else 2054 bp->b_dev = NODEV; 2055 /* 2056 * Insert onto list for new vnode. 2057 */ 2058 bufinsvn(bp, &vp->v_cleanblkhd); 2059} 2060 2061/* 2062 * Disassociate a buffer from a vnode. 2063 * 2064 * Manipulates vnode buffer queues. Must be called at splbio(). 2065 */ 2066void 2067brelvp(struct buf *bp) 2068{ 2069 struct vnode *vp; 2070 2071 splassert(IPL_BIO); 2072 2073 if ((vp = bp->b_vp) == (struct vnode *) 0) 2074 panic("brelvp: NULL"); 2075 /* 2076 * Delete from old vnode list, if on one. 2077 */ 2078 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2079 bufremvn(bp); 2080 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2081 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2082 vp->v_bioflag &= ~VBIOONSYNCLIST; 2083 LIST_REMOVE(vp, v_synclist); 2084 } 2085 bp->b_vp = (struct vnode *) 0; 2086 2087 simple_lock(&vp->v_interlock); 2088#ifdef DIAGNOSTIC 2089 if (vp->v_holdcnt == 0) 2090 panic("brelvp: holdcnt"); 2091#endif 2092 vp->v_holdcnt--; 2093 2094 /* 2095 * If it is on the holdlist and the hold count drops to 2096 * zero, move it to the free list. 2097 */ 2098 if ((vp->v_bioflag & VBIOONFREELIST) && 2099 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 2100 simple_lock(&vnode_free_list_slock); 2101 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 2102 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 2103 simple_unlock(&vnode_free_list_slock); 2104 } 2105 simple_unlock(&vp->v_interlock); 2106} 2107 2108/* 2109 * Replaces the current vnode associated with the buffer, if any, 2110 * with a new vnode. 2111 * 2112 * If an output I/O is pending on the buffer, the old vnode 2113 * I/O count is adjusted. 2114 * 2115 * Ignores vnode buffer queues. Must be called at splbio(). 2116 */ 2117void 2118buf_replacevnode(struct buf *bp, struct vnode *newvp) 2119{ 2120 struct vnode *oldvp = bp->b_vp; 2121 2122 splassert(IPL_BIO); 2123 2124 if (oldvp) 2125 brelvp(bp); 2126 2127 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2128 newvp->v_numoutput++; /* put it on swapdev */ 2129 vwakeup(oldvp); 2130 } 2131 2132 bgetvp(newvp, bp); 2133 bufremvn(bp); 2134} 2135 2136/* 2137 * Used to assign buffers to the appropriate clean or dirty list on 2138 * the vnode and to add newly dirty vnodes to the appropriate 2139 * filesystem syncer list. 2140 * 2141 * Manipulates vnode buffer queues. Must be called at splbio(). 2142 */ 2143void 2144reassignbuf(struct buf *bp) 2145{ 2146 struct buflists *listheadp; 2147 int delay; 2148 struct vnode *vp = bp->b_vp; 2149 2150 splassert(IPL_BIO); 2151 2152 /* 2153 * Delete from old vnode list, if on one. 2154 */ 2155 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2156 bufremvn(bp); 2157 2158 /* 2159 * If dirty, put on list of dirty buffers; 2160 * otherwise insert onto list of clean buffers. 2161 */ 2162 if ((bp->b_flags & B_DELWRI) == 0) { 2163 listheadp = &vp->v_cleanblkhd; 2164 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2165 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2166 vp->v_bioflag &= ~VBIOONSYNCLIST; 2167 LIST_REMOVE(vp, v_synclist); 2168 } 2169 } else { 2170 listheadp = &vp->v_dirtyblkhd; 2171 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2172 switch (vp->v_type) { 2173 case VDIR: 2174 delay = syncdelay / 2; 2175 break; 2176 case VBLK: 2177 if (vp->v_specmountpoint != NULL) { 2178 delay = syncdelay / 3; 2179 break; 2180 } 2181 /* fall through */ 2182 default: 2183 delay = syncdelay; 2184 } 2185 vn_syncer_add_to_worklist(vp, delay); 2186 } 2187 } 2188 bufinsvn(bp, listheadp); 2189} 2190 2191int 2192vfs_register(struct vfsconf *vfs) 2193{ 2194 struct vfsconf *vfsp; 2195 struct vfsconf **vfspp; 2196 2197#ifdef DIAGNOSTIC 2198 /* Paranoia? */ 2199 if (vfs->vfc_refcount != 0) 2200 printf("vfs_register called with vfc_refcount > 0\n"); 2201#endif 2202 2203 /* Check if filesystem already known */ 2204 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2205 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2206 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2207 return (EEXIST); 2208 2209 if (vfs->vfc_typenum > maxvfsconf) 2210 maxvfsconf = vfs->vfc_typenum; 2211 2212 vfs->vfc_next = NULL; 2213 2214 /* Add to the end of the list */ 2215 *vfspp = vfs; 2216 2217 /* Call vfs_init() */ 2218 if (vfs->vfc_vfsops->vfs_init) 2219 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2220 2221 return 0; 2222} 2223 2224int 2225vfs_unregister(struct vfsconf *vfs) 2226{ 2227 struct vfsconf *vfsp; 2228 struct vfsconf **vfspp; 2229 int maxtypenum; 2230 2231 /* Find our vfsconf struct */ 2232 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2233 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2234 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2235 break; 2236 } 2237 2238 if (!vfsp) /* Not found */ 2239 return (ENOENT); 2240 2241 if (vfsp->vfc_refcount) /* In use */ 2242 return (EBUSY); 2243 2244 /* Remove from list and free */ 2245 *vfspp = vfsp->vfc_next; 2246 2247 maxtypenum = 0; 2248 2249 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2250 if (vfsp->vfc_typenum > maxtypenum) 2251 maxtypenum = vfsp->vfc_typenum; 2252 2253 maxvfsconf = maxtypenum; 2254 return 0; 2255} 2256 2257/* 2258 * Check if vnode represents a disk device 2259 */ 2260int 2261vn_isdisk(struct vnode *vp, int *errp) 2262{ 2263 if (vp->v_type != VBLK && vp->v_type != VCHR) 2264 return (0); 2265 2266 return (1); 2267} 2268