vfs_subr.c revision 1.93
1/* $OpenBSD: vfs_subr.c,v 1.93 2003/05/13 09:31:06 naddy 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 <uvm/uvm_extern.h> 68#include <sys/sysctl.h> 69 70#include <miscfs/specfs/specdev.h> 71 72enum vtype iftovt_tab[16] = { 73 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, 74 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, 75}; 76int vttoif_tab[9] = { 77 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, 78 S_IFSOCK, S_IFIFO, S_IFMT, 79}; 80 81int doforce = 1; /* 1 => permit forcible unmounting */ 82int prtactive = 0; /* 1 => print out reclaim of active vnodes */ 83int suid_clear = 1; /* 1 => clear SUID / SGID on owner change */ 84 85/* 86 * Insq/Remq for the vnode usage lists. 87 */ 88#define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) 89#define bufremvn(bp) { \ 90 LIST_REMOVE(bp, b_vnbufs); \ 91 (bp)->b_vnbufs.le_next = NOLIST; \ 92} 93 94struct freelst vnode_hold_list; /* list of vnodes referencing buffers */ 95struct freelst vnode_free_list; /* vnode free list */ 96 97struct mntlist mountlist; /* mounted filesystem list */ 98struct simplelock mountlist_slock; 99static struct simplelock mntid_slock; 100struct simplelock mntvnode_slock; 101struct simplelock vnode_free_list_slock; 102struct simplelock spechash_slock; 103 104void vclean(struct vnode *, int, struct proc *); 105 106void insmntque(struct vnode *, struct mount *); 107int getdevvp(dev_t, struct vnode **, enum vtype); 108 109int vfs_hang_addrlist(struct mount *, struct netexport *, 110 struct export_args *); 111int vfs_free_netcred(struct radix_node *, void *); 112void vfs_free_addrlist(struct netexport *); 113static __inline__ void vputonfreelist(struct vnode *); 114 115int vflush_vnode(struct vnode *, void *); 116 117#ifdef DEBUG 118void printlockedvnodes(void); 119#endif 120 121#define VN_KNOTE(vp, b) \ 122 KNOTE((struct klist *)&vp->v_selectinfo.vsi_selinfo.si_note, (b)) 123 124struct pool vnode_pool; 125 126/* 127 * Initialize the vnode management data structures. 128 */ 129void 130vntblinit() 131{ 132 133 pool_init(&vnode_pool, sizeof(struct vnode), 0, 0, 0, "vnodes", 134 &pool_allocator_nointr); 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 * Mark a mount point as busy. Used to synchronize access and to delay 151 * unmounting. Interlock is not released on failure. 152 * 153 * historical behavior: 154 * - LK_NOWAIT means that we should just ignore the mount point if it's 155 * being unmounted. 156 * - no flags means that we should sleep on the mountpoint and then 157 * fail. 158 */ 159 160int 161vfs_busy(struct mount *mp, int flags, struct simplelock *interlkp, 162 struct proc *p) 163{ 164 int lkflags; 165 166 switch (flags) { 167 case LK_NOWAIT: 168 lkflags = LK_SHARED|LK_NOWAIT; 169 break; 170 case 0: 171 lkflags = LK_SHARED; 172 break; 173 default: 174 lkflags = flags; 175 } 176 177 /* 178 * Always sleepfail. We will only sleep for an exclusive lock 179 * and the exclusive lock will only be acquired when unmounting. 180 */ 181 lkflags |= LK_SLEEPFAIL; 182 183 if (interlkp) 184 lkflags |= LK_INTERLOCK; 185 if (lockmgr(&mp->mnt_lock, lkflags, interlkp, p)) 186 return (ENOENT); 187 return (0); 188} 189 190 191/* 192 * Free a busy file system 193 */ 194void 195vfs_unbusy(struct mount *mp, struct proc *p) 196{ 197 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, p); 198} 199 200int 201vfs_isbusy(struct mount *mp) 202{ 203 return (lockstatus(&mp->mnt_lock)); 204} 205 206/* 207 * Lookup a filesystem type, and if found allocate and initialize 208 * a mount structure for it. 209 * 210 * Devname is usually updated by mount(8) after booting. 211 */ 212 213int 214vfs_rootmountalloc(fstypename, devname, mpp) 215 char *fstypename; 216 char *devname; 217 struct mount **mpp; 218{ 219 struct proc *p = curproc; /* XXX */ 220 struct vfsconf *vfsp; 221 struct mount *mp; 222 223 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 224 if (!strcmp(vfsp->vfc_name, fstypename)) 225 break; 226 if (vfsp == NULL) 227 return (ENODEV); 228 mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 229 bzero((char *)mp, (u_long)sizeof(struct mount)); 230 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 231 (void)vfs_busy(mp, LK_NOWAIT, 0, p); 232 LIST_INIT(&mp->mnt_vnodelist); 233 mp->mnt_vfc = vfsp; 234 mp->mnt_op = vfsp->vfc_vfsops; 235 mp->mnt_flag = MNT_RDONLY; 236 mp->mnt_vnodecovered = NULLVP; 237 vfsp->vfc_refcount++; 238 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 239 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 240 mp->mnt_stat.f_mntonname[0] = '/'; 241 (void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); 242 *mpp = mp; 243 return (0); 244 } 245 246/* 247 * Find an appropriate filesystem to use for the root. If a filesystem 248 * has not been preselected, walk through the list of known filesystems 249 * trying those that have mountroot routines, and try them until one 250 * works or we have tried them all. 251 */ 252int 253vfs_mountroot() 254{ 255 struct vfsconf *vfsp; 256 int error; 257 258 if (mountroot != NULL) 259 return ((*mountroot)()); 260 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) { 261 if (vfsp->vfc_mountroot == NULL) 262 continue; 263 if ((error = (*vfsp->vfc_mountroot)()) == 0) 264 return (0); 265 printf("%s_mountroot failed: %d\n", vfsp->vfc_name, error); 266 } 267 return (ENODEV); 268} 269 270/* 271 * Lookup a mount point by filesystem identifier. 272 */ 273struct mount * 274vfs_getvfs(fsid) 275 fsid_t *fsid; 276{ 277 register struct mount *mp; 278 279 simple_lock(&mountlist_slock); 280 CIRCLEQ_FOREACH(mp, &mountlist, mnt_list) { 281 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 282 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 283 simple_unlock(&mountlist_slock); 284 return (mp); 285 } 286 } 287 simple_unlock(&mountlist_slock); 288 return ((struct mount *)0); 289} 290 291 292/* 293 * Get a new unique fsid 294 */ 295void 296vfs_getnewfsid(mp) 297 struct mount *mp; 298{ 299 static u_short xxxfs_mntid; 300 301 fsid_t tfsid; 302 int mtype; 303 304 simple_lock(&mntid_slock); 305 mtype = mp->mnt_vfc->vfc_typenum; 306 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 307 mp->mnt_stat.f_fsid.val[1] = mtype; 308 if (xxxfs_mntid == 0) 309 ++xxxfs_mntid; 310 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 311 tfsid.val[1] = mtype; 312 if (!CIRCLEQ_EMPTY(&mountlist)) { 313 while (vfs_getvfs(&tfsid)) { 314 tfsid.val[0]++; 315 xxxfs_mntid++; 316 } 317 } 318 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 319 simple_unlock(&mntid_slock); 320} 321 322/* 323 * Make a 'unique' number from a mount type name. 324 * Note that this is no longer used for ffs which 325 * now has an on-disk filesystem id. 326 */ 327long 328makefstype(type) 329 char *type; 330{ 331 long rv; 332 333 for (rv = 0; *type; type++) { 334 rv <<= 2; 335 rv ^= *type; 336 } 337 return rv; 338} 339 340/* 341 * Set vnode attributes to VNOVAL 342 */ 343void 344vattr_null(vap) 345 register struct vattr *vap; 346{ 347 348 vap->va_type = VNON; 349 /* XXX These next two used to be one line, but for a GCC bug. */ 350 vap->va_size = VNOVAL; 351 vap->va_bytes = VNOVAL; 352 vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = 353 vap->va_fsid = vap->va_fileid = 354 vap->va_blocksize = vap->va_rdev = 355 vap->va_atime.tv_sec = vap->va_atime.tv_nsec = 356 vap->va_mtime.tv_sec = vap->va_mtime.tv_nsec = 357 vap->va_ctime.tv_sec = vap->va_ctime.tv_nsec = 358 vap->va_flags = vap->va_gen = VNOVAL; 359 vap->va_vaflags = 0; 360} 361 362/* 363 * Routines having to do with the management of the vnode table. 364 */ 365extern int (**dead_vnodeop_p)(void *); 366long numvnodes; 367 368/* 369 * Return the next vnode from the free list. 370 */ 371int 372getnewvnode(tag, mp, vops, vpp) 373 enum vtagtype tag; 374 struct mount *mp; 375 int (**vops)(void *); 376 struct vnode **vpp; 377{ 378 struct proc *p = curproc; /* XXX */ 379 struct freelst *listhd; 380 static int toggle; 381 struct vnode *vp; 382 int s; 383 384 /* 385 * We must choose whether to allocate a new vnode or recycle an 386 * existing one. The criterion for allocating a new one is that 387 * the total number of vnodes is less than the number desired or 388 * there are no vnodes on either free list. Generally we only 389 * want to recycle vnodes that have no buffers associated with 390 * them, so we look first on the vnode_free_list. If it is empty, 391 * we next consider vnodes with referencing buffers on the 392 * vnode_hold_list. The toggle ensures that half the time we 393 * will use a buffer from the vnode_hold_list, and half the time 394 * we will allocate a new one unless the list has grown to twice 395 * the desired size. We are reticent to recycle vnodes from the 396 * vnode_hold_list because we will lose the identity of all its 397 * referencing buffers. 398 */ 399 toggle ^= 1; 400 if (numvnodes > 2 * desiredvnodes) 401 toggle = 0; 402 403 simple_lock(&vnode_free_list_slock); 404 s = splbio(); 405 if ((numvnodes < desiredvnodes) || 406 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 407 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 408 splx(s); 409 simple_unlock(&vnode_free_list_slock); 410 vp = pool_get(&vnode_pool, PR_WAITOK); 411 bzero((char *)vp, sizeof *vp); 412 simple_lock_init(&vp->v_interlock); 413 numvnodes++; 414 } else { 415 for (vp = TAILQ_FIRST(listhd); vp != NULLVP; 416 vp = TAILQ_NEXT(vp, v_freelist)) { 417 if (simple_lock_try(&vp->v_interlock)) { 418 if ((vp->v_flag & VLAYER) == 0) 419 break; 420 if (VOP_ISLOCKED(vp) == 0) 421 break; 422 else 423 simple_unlock(&vp->v_interlock); 424 } 425 } 426 /* 427 * Unless this is a bad time of the month, at most 428 * the first NCPUS items on the free list are 429 * locked, so this is close enough to being empty. 430 */ 431 if (vp == NULL) { 432 splx(s); 433 simple_unlock(&vnode_free_list_slock); 434 tablefull("vnode"); 435 *vpp = 0; 436 return (ENFILE); 437 } 438 if (vp->v_usecount) { 439 vprint("free vnode", vp); 440 panic("free vnode isn't"); 441 } 442 443 TAILQ_REMOVE(listhd, vp, v_freelist); 444 vp->v_bioflag &= ~VBIOONFREELIST; 445 splx(s); 446 447 simple_unlock(&vnode_free_list_slock); 448 if (vp->v_type != VBAD) 449 vgonel(vp, p); 450 else 451 simple_unlock(&vp->v_interlock); 452#ifdef DIAGNOSTIC 453 if (vp->v_data) { 454 vprint("cleaned vnode", vp); 455 panic("cleaned vnode isn't"); 456 } 457 s = splbio(); 458 if (vp->v_numoutput) 459 panic("Clean vnode has pending I/O's"); 460 splx(s); 461#endif 462 vp->v_flag = 0; 463 vp->v_socket = 0; 464 } 465 vp->v_type = VNON; 466 cache_purge(vp); 467 vp->v_vnlock = NULL; 468 lockinit(&vp->v_lock, PVFS, "v_lock", 0, 0); 469 vp->v_tag = tag; 470 vp->v_op = vops; 471 insmntque(vp, mp); 472 *vpp = vp; 473 vp->v_usecount = 1; 474 vp->v_data = 0; 475 simple_lock_init(&vp->v_uvm.u_obj.vmobjlock); 476 return (0); 477} 478 479/* 480 * Move a vnode from one mount queue to another. 481 */ 482void 483insmntque(vp, mp) 484 register struct vnode *vp; 485 register struct mount *mp; 486{ 487 simple_lock(&mntvnode_slock); 488 /* 489 * Delete from old mount point vnode list, if on one. 490 */ 491 492 if (vp->v_mount != NULL) 493 LIST_REMOVE(vp, v_mntvnodes); 494 /* 495 * Insert into list of vnodes for the new mount point, if available. 496 */ 497 if ((vp->v_mount = mp) != NULL) 498 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 499 simple_unlock(&mntvnode_slock); 500} 501 502 503/* 504 * Create a vnode for a block device. 505 * Used for root filesystem, argdev, and swap areas. 506 * Also used for memory file system special devices. 507 */ 508int 509bdevvp(dev, vpp) 510 dev_t dev; 511 struct vnode **vpp; 512{ 513 514 return (getdevvp(dev, vpp, VBLK)); 515} 516 517/* 518 * Create a vnode for a character device. 519 * Used for kernfs and some console handling. 520 */ 521int 522cdevvp(dev, vpp) 523 dev_t dev; 524 struct vnode **vpp; 525{ 526 527 return (getdevvp(dev, vpp, VCHR)); 528} 529 530/* 531 * Create a vnode for a device. 532 * Used by bdevvp (block device) for root file system etc., 533 * and by cdevvp (character device) for console and kernfs. 534 */ 535int 536getdevvp(dev, vpp, type) 537 dev_t dev; 538 struct vnode **vpp; 539 enum vtype type; 540{ 541 register struct vnode *vp; 542 struct vnode *nvp; 543 int error; 544 545 if (dev == NODEV) { 546 *vpp = NULLVP; 547 return (0); 548 } 549 error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); 550 if (error) { 551 *vpp = NULLVP; 552 return (error); 553 } 554 vp = nvp; 555 vp->v_type = type; 556 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 557 vput(vp); 558 vp = nvp; 559 } 560 *vpp = vp; 561 return (0); 562} 563 564/* 565 * Check to see if the new vnode represents a special device 566 * for which we already have a vnode (either because of 567 * bdevvp() or because of a different vnode representing 568 * the same block device). If such an alias exists, deallocate 569 * the existing contents and return the aliased vnode. The 570 * caller is responsible for filling it with its new contents. 571 */ 572struct vnode * 573checkalias(nvp, nvp_rdev, mp) 574 register struct vnode *nvp; 575 dev_t nvp_rdev; 576 struct mount *mp; 577{ 578 struct proc *p = curproc; 579 register struct vnode *vp; 580 struct vnode **vpp; 581 582 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 583 return (NULLVP); 584 585 vpp = &speclisth[SPECHASH(nvp_rdev)]; 586loop: 587 simple_lock(&spechash_slock); 588 for (vp = *vpp; vp; vp = vp->v_specnext) { 589 simple_lock(&vp->v_interlock); 590 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 591 simple_unlock(&vp->v_interlock); 592 continue; 593 } 594 /* 595 * Alias, but not in use, so flush it out. 596 */ 597 if (vp->v_usecount == 0) { 598 simple_unlock(&spechash_slock); 599 vgonel(vp, p); 600 goto loop; 601 } 602 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 603 simple_unlock(&spechash_slock); 604 goto loop; 605 } 606 break; 607 } 608 609 /* 610 * Common case is actually in the if statement 611 */ 612 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 613 MALLOC(nvp->v_specinfo, struct specinfo *, 614 sizeof(struct specinfo), M_VNODE, M_WAITOK); 615 nvp->v_rdev = nvp_rdev; 616 nvp->v_hashchain = vpp; 617 nvp->v_specnext = *vpp; 618 nvp->v_specmountpoint = NULL; 619 nvp->v_speclockf = NULL; 620 simple_unlock(&spechash_slock); 621 *vpp = nvp; 622 if (vp != NULLVP) { 623 nvp->v_flag |= VALIASED; 624 vp->v_flag |= VALIASED; 625 vput(vp); 626 } 627 return (NULLVP); 628 } 629 630 /* 631 * This code is the uncommon case. It is called in case 632 * we found an alias that was VT_NON && vtype of VBLK 633 * This means we found a block device that was created 634 * using bdevvp. 635 * An example of such a vnode is the root partition device vnode 636 * created in ffs_mountroot. 637 * 638 * The vnodes created by bdevvp should not be aliased (why?). 639 */ 640 641 simple_unlock(&spechash_slock); 642 VOP_UNLOCK(vp, 0, p); 643 simple_lock(&vp->v_interlock); 644 vclean(vp, 0, p); 645 vp->v_vnlock = NULL; 646 lockinit(&vp->v_lock, PVFS, "v_lock", 0, 0); 647 vp->v_op = nvp->v_op; 648 vp->v_tag = nvp->v_tag; 649 nvp->v_type = VNON; 650 insmntque(vp, mp); 651 return (vp); 652} 653 654/* 655 * Grab a particular vnode from the free list, increment its 656 * reference count and lock it. The vnode lock bit is set the 657 * vnode is being eliminated in vgone. The process is awakened 658 * when the transition is completed, and an error returned to 659 * indicate that the vnode is no longer usable (possibly having 660 * been changed to a new file system type). 661 */ 662int 663vget(vp, flags, p) 664 struct vnode *vp; 665 int flags; 666 struct proc *p; 667{ 668 int error; 669 int s; 670 /* 671 * If the vnode is in the process of being cleaned out for 672 * another use, we wait for the cleaning to finish and then 673 * return failure. Cleaning is determined by checking that 674 * the VXLOCK flag is set. 675 */ 676 if ((flags & LK_INTERLOCK) == 0) { 677 simple_lock(&vp->v_interlock); 678 flags |= LK_INTERLOCK; 679 } 680 if (vp->v_flag & VXLOCK) { 681 vp->v_flag |= VXWANT; 682 simple_unlock(&vp->v_interlock); 683 tsleep((caddr_t)vp, PINOD, "vget", 0); 684 return (ENOENT); 685 } 686 if (vp->v_usecount == 0 && 687 (vp->v_bioflag & VBIOONFREELIST)) { 688 s = splbio(); 689 simple_lock(&vnode_free_list_slock); 690 if (vp->v_holdcnt > 0) 691 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 692 else 693 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 694 simple_unlock(&vnode_free_list_slock); 695 vp->v_bioflag &= ~VBIOONFREELIST; 696 splx(s); 697 } 698 vp->v_usecount++; 699 if (flags & LK_TYPE_MASK) { 700 if ((error = vn_lock(vp, flags, p)) != 0) { 701 vp->v_usecount--; 702 if (vp->v_usecount == 0) 703 vputonfreelist(vp); 704 705 simple_unlock(&vp->v_interlock); 706 } 707 return (error); 708 } 709 simple_unlock(&vp->v_interlock); 710 return (0); 711} 712 713 714#ifdef DIAGNOSTIC 715/* 716 * Vnode reference. 717 */ 718void 719vref(vp) 720 struct vnode *vp; 721{ 722 simple_lock(&vp->v_interlock); 723 if (vp->v_usecount == 0) 724 panic("vref used where vget required"); 725 vp->v_usecount++; 726 simple_unlock(&vp->v_interlock); 727} 728#endif /* DIAGNOSTIC */ 729 730static __inline__ void 731vputonfreelist(vp) 732 struct vnode *vp; 733{ 734 int s; 735 struct freelst *lst; 736 737 s = splbio(); 738#ifdef DIAGNOSTIC 739 if (vp->v_usecount != 0) 740 panic("Use count is not zero!"); 741 742 if (vp->v_bioflag & VBIOONFREELIST) { 743 vprint("vnode already on free list: ", vp); 744 panic("vnode already on free list"); 745 } 746#endif 747 748 vp->v_bioflag |= VBIOONFREELIST; 749 750 if (vp->v_holdcnt > 0) 751 lst = &vnode_hold_list; 752 else 753 lst = &vnode_free_list; 754 755 if (vp->v_type == VBAD) 756 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 757 else 758 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 759 760 splx(s); 761} 762 763/* 764 * vput(), just unlock and vrele() 765 */ 766void 767vput(vp) 768 register struct vnode *vp; 769{ 770 struct proc *p = curproc; /* XXX */ 771 772#ifdef DIAGNOSTIC 773 if (vp == NULL) 774 panic("vput: null vp"); 775#endif 776 simple_lock(&vp->v_interlock); 777 778#ifdef DIAGNOSTIC 779 if (vp->v_usecount == 0) { 780 vprint("vput: bad ref count", vp); 781 panic("vput: ref cnt"); 782 } 783#endif 784 vp->v_usecount--; 785 if (vp->v_usecount > 0) { 786 simple_unlock(&vp->v_interlock); 787 VOP_UNLOCK(vp, 0, p); 788 return; 789 } 790 791#ifdef DIAGNOSTIC 792 if (vp->v_writecount != 0) { 793 vprint("vput: bad writecount", vp); 794 panic("vput: v_writecount != 0"); 795 } 796#endif 797 vputonfreelist(vp); 798 799 simple_unlock(&vp->v_interlock); 800 801 VOP_INACTIVE(vp, p); 802} 803 804/* 805 * Vnode release - use for active VNODES. 806 * If count drops to zero, call inactive routine and return to freelist. 807 */ 808void 809vrele(vp) 810 register struct vnode *vp; 811{ 812 struct proc *p = curproc; /* XXX */ 813 814#ifdef DIAGNOSTIC 815 if (vp == NULL) 816 panic("vrele: null vp"); 817#endif 818 simple_lock(&vp->v_interlock); 819#ifdef DIAGNOSTIC 820 if (vp->v_usecount == 0) { 821 vprint("vrele: bad ref count", vp); 822 panic("vrele: ref cnt"); 823 } 824#endif 825 vp->v_usecount--; 826 if (vp->v_usecount > 0) { 827 simple_unlock(&vp->v_interlock); 828 return; 829 } 830 831#ifdef DIAGNOSTIC 832 if (vp->v_writecount != 0) { 833 vprint("vrele: bad writecount", vp); 834 panic("vrele: v_writecount != 0"); 835 } 836#endif 837 vputonfreelist(vp); 838 839 if (vn_lock(vp, LK_EXCLUSIVE|LK_INTERLOCK, p) == 0) 840 VOP_INACTIVE(vp, p); 841} 842 843void vhold(struct vnode *vp); 844 845/* 846 * Page or buffer structure gets a reference. 847 */ 848void 849vhold(vp) 850 register struct vnode *vp; 851{ 852 853 /* 854 * If it is on the freelist and the hold count is currently 855 * zero, move it to the hold list. 856 */ 857 simple_lock(&vp->v_interlock); 858 if ((vp->v_bioflag & VBIOONFREELIST) && 859 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 860 simple_lock(&vnode_free_list_slock); 861 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 862 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 863 simple_unlock(&vnode_free_list_slock); 864 } 865 vp->v_holdcnt++; 866 simple_unlock(&vp->v_interlock); 867} 868 869/* 870 * Remove any vnodes in the vnode table belonging to mount point mp. 871 * 872 * If MNT_NOFORCE is specified, there should not be any active ones, 873 * return error if any are found (nb: this is a user error, not a 874 * system error). If MNT_FORCE is specified, detach any active vnodes 875 * that are found. 876 */ 877#ifdef DEBUG 878int busyprt = 0; /* print out busy vnodes */ 879struct ctldebug debug1 = { "busyprt", &busyprt }; 880#endif 881 882int 883vfs_mount_foreach_vnode(struct mount *mp, 884 int (*func)(struct vnode *, void *), void *arg) { 885 struct vnode *vp, *nvp; 886 int error = 0; 887 888 simple_lock(&mntvnode_slock); 889loop: 890 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { 891 if (vp->v_mount != mp) 892 goto loop; 893 nvp = vp->v_mntvnodes.le_next; 894 simple_lock(&vp->v_interlock); 895 simple_unlock(&mntvnode_slock); 896 897 error = func(vp, arg); 898 899 simple_lock(&mntvnode_slock); 900 901 if (error != 0) 902 break; 903 } 904 simple_unlock(&mntvnode_slock); 905 906 return (error); 907} 908 909 910struct vflush_args { 911 struct vnode *skipvp; 912 int busy; 913 int flags; 914}; 915 916int 917vflush_vnode(struct vnode *vp, void *arg) { 918 struct vflush_args *va = arg; 919 struct proc *p = curproc; 920 921 if (vp == va->skipvp) { 922 simple_unlock(&vp->v_interlock); 923 return (0); 924 } 925 926 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 927 simple_unlock(&vp->v_interlock); 928 return (0); 929 } 930 931 /* 932 * If WRITECLOSE is set, only flush out regular file 933 * vnodes open for writing. 934 */ 935 if ((va->flags & WRITECLOSE) && 936 (vp->v_writecount == 0 || vp->v_type != VREG)) { 937 simple_unlock(&vp->v_interlock); 938 return (0); 939 } 940 941 /* 942 * With v_usecount == 0, all we need to do is clear 943 * out the vnode data structures and we are done. 944 */ 945 if (vp->v_usecount == 0) { 946 vgonel(vp, p); 947 return (0); 948 } 949 950 /* 951 * If FORCECLOSE is set, forcibly close the vnode. 952 * For block or character devices, revert to an 953 * anonymous device. For all other files, just kill them. 954 */ 955 if (va->flags & FORCECLOSE) { 956 if (vp->v_type != VBLK && vp->v_type != VCHR) { 957 vgonel(vp, p); 958 } else { 959 vclean(vp, 0, p); 960 vp->v_op = spec_vnodeop_p; 961 insmntque(vp, (struct mount *)0); 962 } 963 return (0); 964 } 965 966#ifdef DEBUG 967 if (busyprt) 968 vprint("vflush: busy vnode", vp); 969#endif 970 simple_unlock(&vp->v_interlock); 971 va->busy++; 972 return (0); 973} 974 975int 976vflush(mp, skipvp, flags) 977 struct mount *mp; 978 struct vnode *skipvp; 979 int flags; 980{ 981 struct vflush_args va; 982 va.skipvp = skipvp; 983 va.busy = 0; 984 va.flags = flags; 985 986 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 987 988 if (va.busy) 989 return (EBUSY); 990 return (0); 991} 992 993/* 994 * Disassociate the underlying file system from a vnode. 995 * The vnode interlock is held on entry. 996 */ 997void 998vclean(vp, flags, p) 999 register struct vnode *vp; 1000 int flags; 1001 struct proc *p; 1002{ 1003 int active; 1004 1005 /* 1006 * Check to see if the vnode is in use. 1007 * If so we have to reference it before we clean it out 1008 * so that its count cannot fall to zero and generate a 1009 * race against ourselves to recycle it. 1010 */ 1011 if ((active = vp->v_usecount) != 0) 1012 vp->v_usecount++; 1013 1014 /* 1015 * Prevent the vnode from being recycled or 1016 * brought into use while we clean it out. 1017 */ 1018 if (vp->v_flag & VXLOCK) 1019 panic("vclean: deadlock"); 1020 vp->v_flag |= VXLOCK; 1021 /* 1022 * Even if the count is zero, the VOP_INACTIVE routine may still 1023 * have the object locked while it cleans it out. The VOP_LOCK 1024 * ensures that the VOP_INACTIVE routine is done with its work. 1025 * For active vnodes, it ensures that no other activity can 1026 * occur while the underlying object is being cleaned out. 1027 */ 1028 VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK, p); 1029 1030 /* 1031 * clean out any VM data associated with the vnode. 1032 */ 1033 uvm_vnp_terminate(vp); 1034 /* 1035 * Clean out any buffers associated with the vnode. 1036 */ 1037 if (flags & DOCLOSE) 1038 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 1039 /* 1040 * If purging an active vnode, it must be closed and 1041 * deactivated before being reclaimed. Note that the 1042 * VOP_INACTIVE will unlock the vnode 1043 */ 1044 if (active) { 1045 if (flags & DOCLOSE) 1046 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 1047 VOP_INACTIVE(vp, p); 1048 } else { 1049 /* 1050 * Any other processes trying to obtain this lock must first 1051 * wait for VXLOCK to clear, then call the new lock operation. 1052 */ 1053 VOP_UNLOCK(vp, 0, p); 1054 } 1055 1056 /* 1057 * Reclaim the vnode. 1058 */ 1059 if (VOP_RECLAIM(vp, p)) 1060 panic("vclean: cannot reclaim"); 1061 if (active) { 1062 simple_lock(&vp->v_interlock); 1063 1064 vp->v_usecount--; 1065 if (vp->v_usecount == 0) { 1066 if (vp->v_holdcnt > 0) 1067 panic("vclean: not clean"); 1068 vputonfreelist(vp); 1069 } 1070 1071 simple_unlock(&vp->v_interlock); 1072 } 1073 cache_purge(vp); 1074 1075 /* 1076 * Done with purge, notify sleepers of the grim news. 1077 */ 1078 vp->v_op = dead_vnodeop_p; 1079 simple_lock(&vp->v_selectinfo.vsi_lock); 1080 VN_KNOTE(vp, NOTE_REVOKE); 1081 simple_unlock(&vp->v_selectinfo.vsi_lock); 1082 vp->v_tag = VT_NON; 1083 vp->v_flag &= ~VXLOCK; 1084#ifdef DIAGNOSTIC 1085 vp->v_flag &= ~VLOCKSWORK; 1086#endif 1087 if (vp->v_flag & VXWANT) { 1088 vp->v_flag &= ~VXWANT; 1089 wakeup((caddr_t)vp); 1090 } 1091} 1092 1093 1094 1095/* 1096 * Recycle an unused vnode to the front of the free list. 1097 * Release the passed interlock if the vnode will be recycled. 1098 */ 1099int 1100vrecycle(vp, inter_lkp, p) 1101 struct vnode *vp; 1102 struct simplelock *inter_lkp; 1103 struct proc *p; 1104{ 1105 1106 simple_lock(&vp->v_interlock); 1107 if (vp->v_usecount == 0) { 1108 if (inter_lkp) 1109 simple_unlock(inter_lkp); 1110 vgonel(vp, p); 1111 return (1); 1112 } 1113 simple_unlock(&vp->v_interlock); 1114 return (0); 1115} 1116 1117 1118/* 1119 * Eliminate all activity associated with a vnode 1120 * in preparation for reuse. 1121 */ 1122void 1123vgone(vp) 1124 register struct vnode *vp; 1125{ 1126 struct proc *p = curproc; 1127 1128 simple_lock (&vp->v_interlock); 1129 vgonel(vp, p); 1130} 1131 1132/* 1133 * vgone, with the vp interlock held. 1134 */ 1135void 1136vgonel(vp, p) 1137 struct vnode *vp; 1138 struct proc *p; 1139{ 1140 register struct vnode *vq; 1141 struct vnode *vx; 1142 1143 /* 1144 * If a vgone (or vclean) is already in progress, 1145 * wait until it is done and return. 1146 */ 1147 if (vp->v_flag & VXLOCK) { 1148 vp->v_flag |= VXWANT; 1149 simple_unlock(&vp->v_interlock); 1150 tsleep((caddr_t)vp, PINOD, "vgone", 0); 1151 return; 1152 } 1153 /* 1154 * Clean out the filesystem specific data. 1155 */ 1156 vclean(vp, DOCLOSE, p); 1157 /* 1158 * Delete from old mount point vnode list, if on one. 1159 */ 1160 if (vp->v_mount != NULL) 1161 insmntque(vp, (struct mount *)0); 1162 /* 1163 * If special device, remove it from special device alias list 1164 * if it is on one. 1165 */ 1166 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1167 simple_lock(&spechash_slock); 1168 if (*vp->v_hashchain == vp) { 1169 *vp->v_hashchain = vp->v_specnext; 1170 } else { 1171 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1172 if (vq->v_specnext != vp) 1173 continue; 1174 vq->v_specnext = vp->v_specnext; 1175 break; 1176 } 1177 if (vq == NULL) 1178 panic("missing bdev"); 1179 } 1180 if (vp->v_flag & VALIASED) { 1181 vx = NULL; 1182 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1183 if (vq->v_rdev != vp->v_rdev || 1184 vq->v_type != vp->v_type) 1185 continue; 1186 if (vx) 1187 break; 1188 vx = vq; 1189 } 1190 if (vx == NULL) 1191 panic("missing alias"); 1192 if (vq == NULL) 1193 vx->v_flag &= ~VALIASED; 1194 vp->v_flag &= ~VALIASED; 1195 } 1196 simple_unlock(&spechash_slock); 1197 FREE(vp->v_specinfo, M_VNODE); 1198 vp->v_specinfo = NULL; 1199 } 1200 /* 1201 * If it is on the freelist and not already at the head, 1202 * move it to the head of the list. 1203 */ 1204 vp->v_type = VBAD; 1205 1206 /* 1207 * Move onto the free list, unless we were called from 1208 * getnewvnode and we're not on any free list 1209 */ 1210 if (vp->v_usecount == 0 && 1211 (vp->v_bioflag & VBIOONFREELIST)) { 1212 int s; 1213 1214 simple_lock(&vnode_free_list_slock); 1215 s = splbio(); 1216 1217 if (vp->v_holdcnt > 0) 1218 panic("vgonel: not clean"); 1219 1220 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1221 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1222 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1223 } 1224 splx(s); 1225 simple_unlock(&vnode_free_list_slock); 1226 } 1227} 1228 1229/* 1230 * Lookup a vnode by device number. 1231 */ 1232int 1233vfinddev(dev, type, vpp) 1234 dev_t dev; 1235 enum vtype type; 1236 struct vnode **vpp; 1237{ 1238 register struct vnode *vp; 1239 int rc =0; 1240 1241 simple_lock(&spechash_slock); 1242 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1243 if (dev != vp->v_rdev || type != vp->v_type) 1244 continue; 1245 *vpp = vp; 1246 rc = 1; 1247 break; 1248 } 1249 simple_unlock(&spechash_slock); 1250 return (rc); 1251} 1252 1253/* 1254 * Revoke all the vnodes corresponding to the specified minor number 1255 * range (endpoints inclusive) of the specified major. 1256 */ 1257void 1258vdevgone(maj, minl, minh, type) 1259 int maj, minl, minh; 1260 enum vtype type; 1261{ 1262 struct vnode *vp; 1263 int mn; 1264 1265 for (mn = minl; mn <= minh; mn++) 1266 if (vfinddev(makedev(maj, mn), type, &vp)) 1267 VOP_REVOKE(vp, REVOKEALL); 1268} 1269 1270/* 1271 * Calculate the total number of references to a special device. 1272 */ 1273int 1274vcount(vp) 1275 struct vnode *vp; 1276{ 1277 struct vnode *vq, *vnext; 1278 int count; 1279 1280loop: 1281 if ((vp->v_flag & VALIASED) == 0) 1282 return (vp->v_usecount); 1283 simple_lock(&spechash_slock); 1284 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1285 vnext = vq->v_specnext; 1286 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1287 continue; 1288 /* 1289 * Alias, but not in use, so flush it out. 1290 */ 1291 if (vq->v_usecount == 0 && vq != vp) { 1292 simple_unlock(&spechash_slock); 1293 vgone(vq); 1294 goto loop; 1295 } 1296 count += vq->v_usecount; 1297 } 1298 simple_unlock(&spechash_slock); 1299 return (count); 1300} 1301 1302/* 1303 * Print out a description of a vnode. 1304 */ 1305static char *typename[] = 1306 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1307 1308void 1309vprint(label, vp) 1310 char *label; 1311 register struct vnode *vp; 1312{ 1313 char buf[64]; 1314 1315 if (label != NULL) 1316 printf("%s: ", label); 1317 printf("type %s, usecount %u, writecount %u, holdcount %u,", 1318 typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1319 vp->v_holdcnt); 1320 buf[0] = '\0'; 1321 if (vp->v_flag & VROOT) 1322 strlcat(buf, "|VROOT", sizeof buf); 1323 if (vp->v_flag & VTEXT) 1324 strlcat(buf, "|VTEXT", sizeof buf); 1325 if (vp->v_flag & VSYSTEM) 1326 strlcat(buf, "|VSYSTEM", sizeof buf); 1327 if (vp->v_flag & VXLOCK) 1328 strlcat(buf, "|VXLOCK", sizeof buf); 1329 if (vp->v_flag & VXWANT) 1330 strlcat(buf, "|VXWANT", sizeof buf); 1331 if (vp->v_bioflag & VBIOWAIT) 1332 strlcat(buf, "| VBIOWAIT", sizeof buf); 1333 if (vp->v_flag & VALIASED) 1334 strlcat(buf, "|VALIASED", sizeof buf); 1335 if (buf[0] != '\0') 1336 printf(" flags (%s)", &buf[1]); 1337 if (vp->v_data == NULL) { 1338 printf("\n"); 1339 } else { 1340 printf("\n\t"); 1341 VOP_PRINT(vp); 1342 } 1343} 1344 1345#ifdef DEBUG 1346/* 1347 * List all of the locked vnodes in the system. 1348 * Called when debugging the kernel. 1349 */ 1350void 1351printlockedvnodes() 1352{ 1353 struct proc *p = curproc; 1354 register struct mount *mp, *nmp; 1355 register struct vnode *vp; 1356 1357 printf("Locked vnodes\n"); 1358 simple_lock(&mountlist_slock); 1359 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1360 mp = nmp) { 1361 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { 1362 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1363 continue; 1364 } 1365 for (vp = mp->mnt_vnodelist.lh_first; vp; 1366 vp = vp->v_mntvnodes.le_next) { 1367 if (VOP_ISLOCKED(vp)) 1368 vprint((char *)0, vp); 1369 } 1370 simple_lock(&mountlist_slock); 1371 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1372 vfs_unbusy(mp, p); 1373 } 1374 simple_unlock(&mountlist_slock); 1375 1376} 1377#endif 1378 1379/* 1380 * Top level filesystem related information gathering. 1381 */ 1382int 1383vfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1384 int *name; 1385 u_int namelen; 1386 void *oldp; 1387 size_t *oldlenp; 1388 void *newp; 1389 size_t newlen; 1390 struct proc *p; 1391{ 1392 struct vfsconf *vfsp; 1393 1394 /* all sysctl names at this level are at least name and field */ 1395 if (namelen < 2) 1396 return (ENOTDIR); /* overloaded */ 1397 if (name[0] != VFS_GENERIC) { 1398 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1399 if (vfsp->vfc_typenum == name[0]) 1400 break; 1401 if (vfsp == NULL) 1402 return (EOPNOTSUPP); 1403 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1404 oldp, oldlenp, newp, newlen, p)); 1405 } 1406 switch (name[1]) { 1407 case VFS_MAXTYPENUM: 1408 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1409 case VFS_CONF: 1410 if (namelen < 3) 1411 return (ENOTDIR); /* overloaded */ 1412 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1413 if (vfsp->vfc_typenum == name[2]) 1414 break; 1415 if (vfsp == NULL) 1416 return (EOPNOTSUPP); 1417 return (sysctl_rdstruct(oldp, oldlenp, newp, vfsp, 1418 sizeof(struct vfsconf))); 1419 } 1420 return (EOPNOTSUPP); 1421} 1422 1423 1424int kinfo_vdebug = 1; 1425int kinfo_vgetfailed; 1426#define KINFO_VNODESLOP 10 1427/* 1428 * Dump vnode list (via sysctl). 1429 * Copyout address of vnode followed by vnode. 1430 */ 1431/* ARGSUSED */ 1432int 1433sysctl_vnode(where, sizep, p) 1434 char *where; 1435 size_t *sizep; 1436 struct proc *p; 1437{ 1438 register struct mount *mp, *nmp; 1439 struct vnode *vp, *nvp; 1440 register char *bp = where, *savebp; 1441 char *ewhere; 1442 int error; 1443 1444 if (where == NULL) { 1445 *sizep = (numvnodes + KINFO_VNODESLOP) * sizeof(struct e_vnode); 1446 return (0); 1447 } 1448 ewhere = where + *sizep; 1449 1450 simple_lock(&mountlist_slock); 1451 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1452 mp = nmp) { 1453 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { 1454 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1455 continue; 1456 } 1457 savebp = bp; 1458again: 1459 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; 1460 vp = nvp) { 1461 /* 1462 * Check that the vp is still associated with 1463 * this filesystem. RACE: could have been 1464 * recycled onto the same filesystem. 1465 */ 1466 if (vp->v_mount != mp) { 1467 simple_unlock(&mntvnode_slock); 1468 if (kinfo_vdebug) 1469 printf("kinfo: vp changed\n"); 1470 bp = savebp; 1471 goto again; 1472 } 1473 nvp = vp->v_mntvnodes.le_next; 1474 if (bp + sizeof(struct e_vnode) > ewhere) { 1475 simple_unlock(&mntvnode_slock); 1476 *sizep = bp - where; 1477 vfs_unbusy(mp, p); 1478 return (ENOMEM); 1479 } 1480 if ((error = copyout((caddr_t)&vp, 1481 &((struct e_vnode *)bp)->vptr, 1482 sizeof(struct vnode *))) || 1483 (error = copyout((caddr_t)vp, 1484 &((struct e_vnode *)bp)->vnode, 1485 sizeof(struct vnode)))) { 1486 vfs_unbusy(mp, p); 1487 return (error); 1488 } 1489 bp += sizeof(struct e_vnode); 1490 simple_lock(&mntvnode_slock); 1491 } 1492 1493 simple_unlock(&mntvnode_slock); 1494 simple_lock(&mountlist_slock); 1495 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1496 vfs_unbusy(mp, p); 1497 } 1498 1499 simple_unlock(&mountlist_slock); 1500 1501 *sizep = bp - where; 1502 return (0); 1503} 1504 1505/* 1506 * Check to see if a filesystem is mounted on a block device. 1507 */ 1508int 1509vfs_mountedon(vp) 1510 register struct vnode *vp; 1511{ 1512 register struct vnode *vq; 1513 int error = 0; 1514 1515 if (vp->v_specmountpoint != NULL) 1516 return (EBUSY); 1517 if (vp->v_flag & VALIASED) { 1518 simple_lock(&spechash_slock); 1519 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1520 if (vq->v_rdev != vp->v_rdev || 1521 vq->v_type != vp->v_type) 1522 continue; 1523 if (vq->v_specmountpoint != NULL) { 1524 error = EBUSY; 1525 break; 1526 } 1527 } 1528 simple_unlock(&spechash_slock); 1529 } 1530 return (error); 1531} 1532 1533/* 1534 * Build hash lists of net addresses and hang them off the mount point. 1535 * Called by ufs_mount() to set up the lists of export addresses. 1536 */ 1537int 1538vfs_hang_addrlist(mp, nep, argp) 1539 struct mount *mp; 1540 struct netexport *nep; 1541 struct export_args *argp; 1542{ 1543 register struct netcred *np; 1544 register struct radix_node_head *rnh; 1545 register int i; 1546 struct radix_node *rn; 1547 struct sockaddr *saddr, *smask = 0; 1548 struct domain *dom; 1549 int error; 1550 1551 if (argp->ex_addrlen == 0) { 1552 if (mp->mnt_flag & MNT_DEFEXPORTED) 1553 return (EPERM); 1554 np = &nep->ne_defexported; 1555 np->netc_exflags = argp->ex_flags; 1556 np->netc_anon = argp->ex_anon; 1557 np->netc_anon.cr_ref = 1; 1558 mp->mnt_flag |= MNT_DEFEXPORTED; 1559 return (0); 1560 } 1561 if (argp->ex_addrlen > MLEN) 1562 return (EINVAL); 1563 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1564 np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK); 1565 bzero((caddr_t)np, i); 1566 saddr = (struct sockaddr *)(np + 1); 1567 error = copyin(argp->ex_addr, (caddr_t)saddr, argp->ex_addrlen); 1568 if (error) 1569 goto out; 1570 if (saddr->sa_len > argp->ex_addrlen) 1571 saddr->sa_len = argp->ex_addrlen; 1572 if (argp->ex_masklen) { 1573 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1574 error = copyin(argp->ex_mask, (caddr_t)smask, argp->ex_masklen); 1575 if (error) 1576 goto out; 1577 if (smask->sa_len > argp->ex_masklen) 1578 smask->sa_len = argp->ex_masklen; 1579 } 1580 i = saddr->sa_family; 1581 if (i < 0 || i > AF_MAX) { 1582 error = EINVAL; 1583 goto out; 1584 } 1585 if ((rnh = nep->ne_rtable[i]) == 0) { 1586 /* 1587 * Seems silly to initialize every AF when most are not 1588 * used, do so on demand here 1589 */ 1590 for (dom = domains; dom; dom = dom->dom_next) 1591 if (dom->dom_family == i && dom->dom_rtattach) { 1592 dom->dom_rtattach((void **)&nep->ne_rtable[i], 1593 dom->dom_rtoffset); 1594 break; 1595 } 1596 if ((rnh = nep->ne_rtable[i]) == 0) { 1597 error = ENOBUFS; 1598 goto out; 1599 } 1600 } 1601 rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, 1602 np->netc_rnodes); 1603 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1604 error = EPERM; 1605 goto out; 1606 } 1607 np->netc_exflags = argp->ex_flags; 1608 np->netc_anon = argp->ex_anon; 1609 np->netc_anon.cr_ref = 1; 1610 return (0); 1611out: 1612 free(np, M_NETADDR); 1613 return (error); 1614} 1615 1616/* ARGSUSED */ 1617int 1618vfs_free_netcred(rn, w) 1619 struct radix_node *rn; 1620 void *w; 1621{ 1622 register struct radix_node_head *rnh = (struct radix_node_head *)w; 1623 1624 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh); 1625 free((caddr_t)rn, M_NETADDR); 1626 return (0); 1627} 1628 1629/* 1630 * Free the net address hash lists that are hanging off the mount points. 1631 */ 1632void 1633vfs_free_addrlist(nep) 1634 struct netexport *nep; 1635{ 1636 register int i; 1637 register struct radix_node_head *rnh; 1638 1639 for (i = 0; i <= AF_MAX; i++) 1640 if ((rnh = nep->ne_rtable[i]) != NULL) { 1641 (*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh); 1642 free((caddr_t)rnh, M_RTABLE); 1643 nep->ne_rtable[i] = 0; 1644 } 1645} 1646 1647int 1648vfs_export(mp, nep, argp) 1649 struct mount *mp; 1650 struct netexport *nep; 1651 struct export_args *argp; 1652{ 1653 int error; 1654 1655 if (argp->ex_flags & MNT_DELEXPORT) { 1656 vfs_free_addrlist(nep); 1657 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1658 } 1659 if (argp->ex_flags & MNT_EXPORTED) { 1660 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1661 return (error); 1662 mp->mnt_flag |= MNT_EXPORTED; 1663 } 1664 return (0); 1665} 1666 1667struct netcred * 1668vfs_export_lookup(mp, nep, nam) 1669 register struct mount *mp; 1670 struct netexport *nep; 1671 struct mbuf *nam; 1672{ 1673 register struct netcred *np; 1674 register struct radix_node_head *rnh; 1675 struct sockaddr *saddr; 1676 1677 np = NULL; 1678 if (mp->mnt_flag & MNT_EXPORTED) { 1679 /* 1680 * Lookup in the export list first. 1681 */ 1682 if (nam != NULL) { 1683 saddr = mtod(nam, struct sockaddr *); 1684 rnh = nep->ne_rtable[saddr->sa_family]; 1685 if (rnh != NULL) { 1686 np = (struct netcred *) 1687 (*rnh->rnh_matchaddr)((caddr_t)saddr, 1688 rnh); 1689 if (np && np->netc_rnodes->rn_flags & RNF_ROOT) 1690 np = NULL; 1691 } 1692 } 1693 /* 1694 * If no address match, use the default if it exists. 1695 */ 1696 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1697 np = &nep->ne_defexported; 1698 } 1699 return (np); 1700} 1701 1702/* 1703 * Do the usual access checking. 1704 * file_mode, uid and gid are from the vnode in question, 1705 * while acc_mode and cred are from the VOP_ACCESS parameter list 1706 */ 1707int 1708vaccess(file_mode, uid, gid, acc_mode, cred) 1709 mode_t file_mode; 1710 uid_t uid; 1711 gid_t gid; 1712 mode_t acc_mode; 1713 struct ucred *cred; 1714{ 1715 mode_t mask; 1716 1717 /* User id 0 always gets access. */ 1718 if (cred->cr_uid == 0) 1719 return 0; 1720 1721 mask = 0; 1722 1723 /* Otherwise, check the owner. */ 1724 if (cred->cr_uid == uid) { 1725 if (acc_mode & VEXEC) 1726 mask |= S_IXUSR; 1727 if (acc_mode & VREAD) 1728 mask |= S_IRUSR; 1729 if (acc_mode & VWRITE) 1730 mask |= S_IWUSR; 1731 return (file_mode & mask) == mask ? 0 : EACCES; 1732 } 1733 1734 /* Otherwise, check the groups. */ 1735 if (cred->cr_gid == gid || groupmember(gid, cred)) { 1736 if (acc_mode & VEXEC) 1737 mask |= S_IXGRP; 1738 if (acc_mode & VREAD) 1739 mask |= S_IRGRP; 1740 if (acc_mode & VWRITE) 1741 mask |= S_IWGRP; 1742 return (file_mode & mask) == mask ? 0 : EACCES; 1743 } 1744 1745 /* Otherwise, check everyone else. */ 1746 if (acc_mode & VEXEC) 1747 mask |= S_IXOTH; 1748 if (acc_mode & VREAD) 1749 mask |= S_IROTH; 1750 if (acc_mode & VWRITE) 1751 mask |= S_IWOTH; 1752 return (file_mode & mask) == mask ? 0 : EACCES; 1753} 1754 1755/* 1756 * Unmount all file systems. 1757 * We traverse the list in reverse order under the assumption that doing so 1758 * will avoid needing to worry about dependencies. 1759 */ 1760void 1761vfs_unmountall(void) 1762{ 1763 struct mount *mp, *nmp; 1764 int allerror, error, again = 1; 1765 struct proc *p = curproc; 1766 1767 retry: 1768 allerror = 0; 1769 for (mp = CIRCLEQ_LAST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1770 mp = nmp) { 1771 nmp = CIRCLEQ_PREV(mp, mnt_list); 1772 if ((vfs_busy(mp, LK_EXCLUSIVE|LK_NOWAIT, NULL, p)) != 0) 1773 continue; 1774 if ((error = dounmount(mp, MNT_FORCE, curproc, NULL)) != 0) { 1775 printf("unmount of %s failed with error %d\n", 1776 mp->mnt_stat.f_mntonname, error); 1777 allerror = 1; 1778 } 1779 } 1780 1781 if (allerror) { 1782 printf("WARNING: some file systems would not unmount\n"); 1783 if (again) { 1784 printf("retrying\n"); 1785 again = 0; 1786 goto retry; 1787 } 1788 } 1789} 1790 1791/* 1792 * Sync and unmount file systems before shutting down. 1793 */ 1794void 1795vfs_shutdown() 1796{ 1797 /* XXX Should suspend scheduling. */ 1798 (void) spl0(); 1799 1800 printf("syncing disks... "); 1801 1802 if (panicstr == 0) { 1803 /* Sync before unmount, in case we hang on something. */ 1804 sys_sync(&proc0, (void *)0, (register_t *)0); 1805 1806 /* Unmount file systems. */ 1807 vfs_unmountall(); 1808 } 1809 1810 if (vfs_syncwait(1)) 1811 printf("giving up\n"); 1812 else 1813 printf("done\n"); 1814} 1815 1816/* 1817 * perform sync() operation and wait for buffers to flush. 1818 * assumtions: called w/ scheduler disabled and physical io enabled 1819 * for now called at spl0() XXX 1820 */ 1821int 1822vfs_syncwait(verbose) 1823 int verbose; 1824{ 1825 register struct buf *bp; 1826 int iter, nbusy, dcount, s; 1827 struct proc *p; 1828 1829 p = curproc? curproc : &proc0; 1830 sys_sync(p, (void *)0, (register_t *)0); 1831 1832 /* Wait for sync to finish. */ 1833 dcount = 10000; 1834 for (iter = 0; iter < 20; iter++) { 1835 nbusy = 0; 1836 for (bp = &buf[nbuf]; --bp >= buf; ) { 1837 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1838 nbusy++; 1839 /* 1840 * With soft updates, some buffers that are 1841 * written will be remarked as dirty until other 1842 * buffers are written. 1843 */ 1844 if (bp->b_flags & B_DELWRI) { 1845 s = splbio(); 1846 bremfree(bp); 1847 bp->b_flags |= B_BUSY; 1848 splx(s); 1849 nbusy++; 1850 bawrite(bp); 1851 if (dcount-- <= 0) { 1852 if (verbose) 1853 printf("softdep "); 1854 return 1; 1855 } 1856 } 1857 } 1858 if (nbusy == 0) 1859 break; 1860 if (verbose) 1861 printf("%d ", nbusy); 1862 DELAY(40000 * iter); 1863 } 1864 1865 return nbusy; 1866} 1867 1868/* 1869 * posix file system related system variables. 1870 */ 1871int 1872fs_posix_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1873 int *name; 1874 u_int namelen; 1875 void *oldp; 1876 size_t *oldlenp; 1877 void *newp; 1878 size_t newlen; 1879 struct proc *p; 1880{ 1881 /* all sysctl names at this level are terminal */ 1882 if (namelen != 1) 1883 return (ENOTDIR); 1884 1885 switch (name[0]) { 1886 case FS_POSIX_SETUID: 1887 if (newp && securelevel > 0) 1888 return (EPERM); 1889 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1890 default: 1891 return (EOPNOTSUPP); 1892 } 1893 /* NOTREACHED */ 1894} 1895 1896/* 1897 * file system related system variables. 1898 */ 1899int 1900fs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1901 int *name; 1902 u_int namelen; 1903 void *oldp; 1904 size_t *oldlenp; 1905 void *newp; 1906 size_t newlen; 1907 struct proc *p; 1908{ 1909 sysctlfn *fn; 1910 1911 switch (name[0]) { 1912 case FS_POSIX: 1913 fn = fs_posix_sysctl; 1914 break; 1915 default: 1916 return (EOPNOTSUPP); 1917 } 1918 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1919} 1920 1921 1922/* 1923 * Routines dealing with vnodes and buffers 1924 */ 1925 1926/* 1927 * Wait for all outstanding I/Os to complete 1928 * 1929 * Manipulates v_numoutput. Must be called at splbio() 1930 */ 1931int 1932vwaitforio(vp, slpflag, wmesg, timeo) 1933 struct vnode *vp; 1934 int slpflag, timeo; 1935 char *wmesg; 1936{ 1937 int error = 0; 1938 1939 splassert(IPL_BIO); 1940 1941 while (vp->v_numoutput) { 1942 vp->v_bioflag |= VBIOWAIT; 1943 error = tsleep((caddr_t)&vp->v_numoutput, 1944 slpflag | (PRIBIO + 1), wmesg, timeo); 1945 if (error) 1946 break; 1947 } 1948 1949 return (error); 1950} 1951 1952 1953/* 1954 * Update outstanding I/O count and do wakeup if requested. 1955 * 1956 * Manipulates v_numoutput. Must be called at splbio() 1957 */ 1958void 1959vwakeup(vp) 1960 struct vnode *vp; 1961{ 1962 splassert(IPL_BIO); 1963 1964 if (vp != NULL) { 1965 if (vp->v_numoutput-- == 0) 1966 panic("vwakeup: neg numoutput"); 1967 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1968 vp->v_bioflag &= ~VBIOWAIT; 1969 wakeup((caddr_t)&vp->v_numoutput); 1970 } 1971 } 1972} 1973 1974/* 1975 * Flush out and invalidate all buffers associated with a vnode. 1976 * Called with the underlying object locked. 1977 */ 1978int 1979vinvalbuf(vp, flags, cred, p, slpflag, slptimeo) 1980 register struct vnode *vp; 1981 int flags; 1982 struct ucred *cred; 1983 struct proc *p; 1984 int slpflag, slptimeo; 1985{ 1986 register struct buf *bp; 1987 struct buf *nbp, *blist; 1988 int s, error; 1989 1990 if (flags & V_SAVE) { 1991 s = splbio(); 1992 vwaitforio(vp, 0, "vinvalbuf", 0); 1993 if (vp->v_dirtyblkhd.lh_first != NULL) { 1994 splx(s); 1995 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1996 return (error); 1997 s = splbio(); 1998 if (vp->v_numoutput > 0 || 1999 vp->v_dirtyblkhd.lh_first != NULL) 2000 panic("vinvalbuf: dirty bufs"); 2001 } 2002 splx(s); 2003 } 2004loop: 2005 s = splbio(); 2006 for (;;) { 2007 if ((blist = vp->v_cleanblkhd.lh_first) && 2008 (flags & V_SAVEMETA)) 2009 while (blist && blist->b_lblkno < 0) 2010 blist = blist->b_vnbufs.le_next; 2011 if (!blist && (blist = vp->v_dirtyblkhd.lh_first) && 2012 (flags & V_SAVEMETA)) 2013 while (blist && blist->b_lblkno < 0) 2014 blist = blist->b_vnbufs.le_next; 2015 if (!blist) 2016 break; 2017 2018 for (bp = blist; bp; bp = nbp) { 2019 nbp = bp->b_vnbufs.le_next; 2020 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 2021 continue; 2022 if (bp->b_flags & B_BUSY) { 2023 bp->b_flags |= B_WANTED; 2024 error = tsleep((caddr_t)bp, 2025 slpflag | (PRIBIO + 1), "vinvalbuf", 2026 slptimeo); 2027 if (error) { 2028 splx(s); 2029 return (error); 2030 } 2031 break; 2032 } 2033 bremfree(bp); 2034 bp->b_flags |= B_BUSY; 2035 /* 2036 * XXX Since there are no node locks for NFS, I believe 2037 * there is a slight chance that a delayed write will 2038 * occur while sleeping just above, so check for it. 2039 */ 2040 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 2041 splx(s); 2042 (void) VOP_BWRITE(bp); 2043 goto loop; 2044 } 2045 bp->b_flags |= B_INVAL; 2046 brelse(bp); 2047 } 2048 } 2049 if (!(flags & V_SAVEMETA) && 2050 (vp->v_dirtyblkhd.lh_first || vp->v_cleanblkhd.lh_first)) 2051 panic("vinvalbuf: flush failed"); 2052 splx(s); 2053 return (0); 2054} 2055 2056void 2057vflushbuf(vp, sync) 2058 register struct vnode *vp; 2059 int sync; 2060{ 2061 register struct buf *bp, *nbp; 2062 int s; 2063 2064loop: 2065 s = splbio(); 2066 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { 2067 nbp = bp->b_vnbufs.le_next; 2068 if ((bp->b_flags & B_BUSY)) 2069 continue; 2070 if ((bp->b_flags & B_DELWRI) == 0) 2071 panic("vflushbuf: not dirty"); 2072 bremfree(bp); 2073 bp->b_flags |= B_BUSY; 2074 splx(s); 2075 /* 2076 * Wait for I/O associated with indirect blocks to complete, 2077 * since there is no way to quickly wait for them below. 2078 */ 2079 if (bp->b_vp == vp || sync == 0) 2080 (void) bawrite(bp); 2081 else 2082 (void) bwrite(bp); 2083 goto loop; 2084 } 2085 if (sync == 0) { 2086 splx(s); 2087 return; 2088 } 2089 vwaitforio(vp, 0, "vflushbuf", 0); 2090 if (vp->v_dirtyblkhd.lh_first != NULL) { 2091 splx(s); 2092 vprint("vflushbuf: dirty", vp); 2093 goto loop; 2094 } 2095 splx(s); 2096} 2097 2098/* 2099 * Associate a buffer with a vnode. 2100 * 2101 * Manipulates buffer vnode queues. Must be called at splbio(). 2102 */ 2103void 2104bgetvp(vp, bp) 2105 register struct vnode *vp; 2106 register struct buf *bp; 2107{ 2108 splassert(IPL_BIO); 2109 2110 2111 if (bp->b_vp) 2112 panic("bgetvp: not free"); 2113 vhold(vp); 2114 bp->b_vp = vp; 2115 if (vp->v_type == VBLK || vp->v_type == VCHR) 2116 bp->b_dev = vp->v_rdev; 2117 else 2118 bp->b_dev = NODEV; 2119 /* 2120 * Insert onto list for new vnode. 2121 */ 2122 bufinsvn(bp, &vp->v_cleanblkhd); 2123} 2124 2125/* 2126 * Disassociate a buffer from a vnode. 2127 * 2128 * Manipulates vnode buffer queues. Must be called at splbio(). 2129 */ 2130void 2131brelvp(bp) 2132 register struct buf *bp; 2133{ 2134 struct vnode *vp; 2135 2136 splassert(IPL_BIO); 2137 2138 if ((vp = bp->b_vp) == (struct vnode *) 0) 2139 panic("brelvp: NULL"); 2140 /* 2141 * Delete from old vnode list, if on one. 2142 */ 2143 if (bp->b_vnbufs.le_next != NOLIST) 2144 bufremvn(bp); 2145 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2146 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2147 vp->v_bioflag &= ~VBIOONSYNCLIST; 2148 LIST_REMOVE(vp, v_synclist); 2149 } 2150 bp->b_vp = (struct vnode *) 0; 2151 2152 simple_lock(&vp->v_interlock); 2153#ifdef DIAGNOSTIC 2154 if (vp->v_holdcnt == 0) 2155 panic("brelvp: holdcnt"); 2156#endif 2157 vp->v_holdcnt--; 2158 2159 /* 2160 * If it is on the holdlist and the hold count drops to 2161 * zero, move it to the free list. 2162 */ 2163 if ((vp->v_bioflag & VBIOONFREELIST) && 2164 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 2165 simple_lock(&vnode_free_list_slock); 2166 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 2167 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 2168 simple_unlock(&vnode_free_list_slock); 2169 } 2170 simple_unlock(&vp->v_interlock); 2171} 2172 2173/* 2174 * Replaces the current vnode associated with the buffer, if any 2175 * with a new vnode. 2176 * 2177 * If an output I/O is pending on the buffer, the old vnode is 2178 * I/O count is adjusted. 2179 * 2180 * Ignores vnode buffer queues. Must be called at splbio(). 2181 */ 2182void 2183buf_replacevnode(bp, newvp) 2184 struct buf *bp; 2185 struct vnode *newvp; 2186{ 2187 struct vnode *oldvp = bp->b_vp; 2188 2189 splassert(IPL_BIO); 2190 2191 if (oldvp) 2192 brelvp(bp); 2193 2194 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2195 newvp->v_numoutput++; /* put it on swapdev */ 2196 vwakeup(oldvp); 2197 } 2198 2199 bgetvp(newvp, bp); 2200 bufremvn(bp); 2201} 2202 2203/* 2204 * Used to assign buffers to the appropriate clean or dirty list on 2205 * the vnode and to add newly dirty vnodes to the appropriate 2206 * filesystem syncer list. 2207 * 2208 * Manipulates vnode buffer queues. Must be called at splbio(). 2209 */ 2210void 2211reassignbuf(bp) 2212 struct buf *bp; 2213{ 2214 struct buflists *listheadp; 2215 int delay; 2216 struct vnode *vp = bp->b_vp; 2217 2218 splassert(IPL_BIO); 2219 2220 /* 2221 * Delete from old vnode list, if on one. 2222 */ 2223 if (bp->b_vnbufs.le_next != NOLIST) 2224 bufremvn(bp); 2225 /* 2226 * If dirty, put on list of dirty buffers; 2227 * otherwise insert onto list of clean buffers. 2228 */ 2229 if ((bp->b_flags & B_DELWRI) == 0) { 2230 listheadp = &vp->v_cleanblkhd; 2231 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2232 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2233 vp->v_bioflag &= ~VBIOONSYNCLIST; 2234 LIST_REMOVE(vp, v_synclist); 2235 } 2236 } else { 2237 listheadp = &vp->v_dirtyblkhd; 2238 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2239 switch (vp->v_type) { 2240 case VDIR: 2241 delay = syncdelay / 2; 2242 break; 2243 case VBLK: 2244 if (vp->v_specmountpoint != NULL) { 2245 delay = syncdelay / 3; 2246 break; 2247 } 2248 /* fall through */ 2249 default: 2250 delay = syncdelay; 2251 } 2252 vn_syncer_add_to_worklist(vp, delay); 2253 } 2254 } 2255 bufinsvn(bp, listheadp); 2256} 2257 2258int 2259vfs_register(vfs) 2260 struct vfsconf *vfs; 2261{ 2262 struct vfsconf *vfsp; 2263 struct vfsconf **vfspp; 2264 2265#ifdef DIAGNOSTIC 2266 /* Paranoia? */ 2267 if (vfs->vfc_refcount != 0) 2268 printf("vfs_register called with vfc_refcount > 0\n"); 2269#endif 2270 2271 /* Check if filesystem already known */ 2272 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2273 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2274 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2275 return (EEXIST); 2276 2277 if (vfs->vfc_typenum > maxvfsconf) 2278 maxvfsconf = vfs->vfc_typenum; 2279 2280 vfs->vfc_next = NULL; 2281 2282 /* Add to the end of the list */ 2283 *vfspp = vfs; 2284 2285 /* Call vfs_init() */ 2286 if (vfs->vfc_vfsops->vfs_init) 2287 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2288 2289 return 0; 2290} 2291 2292int 2293vfs_unregister(vfs) 2294 struct vfsconf *vfs; 2295{ 2296 struct vfsconf *vfsp; 2297 struct vfsconf **vfspp; 2298 int maxtypenum; 2299 2300 /* Find our vfsconf struct */ 2301 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2302 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2303 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2304 break; 2305 } 2306 2307 if (!vfsp) /* Not found */ 2308 return (ENOENT); 2309 2310 if (vfsp->vfc_refcount) /* In use */ 2311 return (EBUSY); 2312 2313 /* Remove from list and free */ 2314 *vfspp = vfsp->vfc_next; 2315 2316 maxtypenum = 0; 2317 2318 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2319 if (vfsp->vfc_typenum > maxtypenum) 2320 maxtypenum = vfsp->vfc_typenum; 2321 2322 maxvfsconf = maxtypenum; 2323 return 0; 2324} 2325 2326/* 2327 * Check if vnode represents a disk device 2328 */ 2329int 2330vn_isdisk(vp, errp) 2331 struct vnode *vp; 2332 int *errp; 2333{ 2334 if (vp->v_type != VBLK && vp->v_type != VCHR) 2335 return (0); 2336 2337 return (1); 2338} 2339