vfs_subr.c revision 1.92
1/* $OpenBSD: vfs_subr.c,v 1.92 2003/05/13 02:30:01 tedu 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 (VOP_ISLOCKED(vp) == 0) 419 break; 420 else 421 simple_unlock(&vp->v_interlock); 422 } 423 } 424 /* 425 * Unless this is a bad time of the month, at most 426 * the first NCPUS items on the free list are 427 * locked, so this is close enough to being empty. 428 */ 429 if (vp == NULL) { 430 splx(s); 431 simple_unlock(&vnode_free_list_slock); 432 tablefull("vnode"); 433 *vpp = 0; 434 return (ENFILE); 435 } 436 if (vp->v_usecount) { 437 vprint("free vnode", vp); 438 panic("free vnode isn't"); 439 } 440 441 TAILQ_REMOVE(listhd, vp, v_freelist); 442 vp->v_bioflag &= ~VBIOONFREELIST; 443 splx(s); 444 445 simple_unlock(&vnode_free_list_slock); 446 if (vp->v_type != VBAD) 447 vgonel(vp, p); 448 else 449 simple_unlock(&vp->v_interlock); 450#ifdef DIAGNOSTIC 451 if (vp->v_data) { 452 vprint("cleaned vnode", vp); 453 panic("cleaned vnode isn't"); 454 } 455 s = splbio(); 456 if (vp->v_numoutput) 457 panic("Clean vnode has pending I/O's"); 458 splx(s); 459#endif 460 vp->v_flag = 0; 461 vp->v_socket = 0; 462 } 463 vp->v_type = VNON; 464 cache_purge(vp); 465 vp->v_vnlock = NULL; 466 lockinit(&vp->v_lock, PVFS, "v_lock", 0, 0); 467 vp->v_tag = tag; 468 vp->v_op = vops; 469 insmntque(vp, mp); 470 *vpp = vp; 471 vp->v_usecount = 1; 472 vp->v_data = 0; 473 simple_lock_init(&vp->v_uvm.u_obj.vmobjlock); 474 return (0); 475} 476 477/* 478 * Move a vnode from one mount queue to another. 479 */ 480void 481insmntque(vp, mp) 482 register struct vnode *vp; 483 register struct mount *mp; 484{ 485 simple_lock(&mntvnode_slock); 486 /* 487 * Delete from old mount point vnode list, if on one. 488 */ 489 490 if (vp->v_mount != NULL) 491 LIST_REMOVE(vp, v_mntvnodes); 492 /* 493 * Insert into list of vnodes for the new mount point, if available. 494 */ 495 if ((vp->v_mount = mp) != NULL) 496 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 497 simple_unlock(&mntvnode_slock); 498} 499 500 501/* 502 * Create a vnode for a block device. 503 * Used for root filesystem, argdev, and swap areas. 504 * Also used for memory file system special devices. 505 */ 506int 507bdevvp(dev, vpp) 508 dev_t dev; 509 struct vnode **vpp; 510{ 511 512 return (getdevvp(dev, vpp, VBLK)); 513} 514 515/* 516 * Create a vnode for a character device. 517 * Used for kernfs and some console handling. 518 */ 519int 520cdevvp(dev, vpp) 521 dev_t dev; 522 struct vnode **vpp; 523{ 524 525 return (getdevvp(dev, vpp, VCHR)); 526} 527 528/* 529 * Create a vnode for a device. 530 * Used by bdevvp (block device) for root file system etc., 531 * and by cdevvp (character device) for console and kernfs. 532 */ 533int 534getdevvp(dev, vpp, type) 535 dev_t dev; 536 struct vnode **vpp; 537 enum vtype type; 538{ 539 register struct vnode *vp; 540 struct vnode *nvp; 541 int error; 542 543 if (dev == NODEV) { 544 *vpp = NULLVP; 545 return (0); 546 } 547 error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); 548 if (error) { 549 *vpp = NULLVP; 550 return (error); 551 } 552 vp = nvp; 553 vp->v_type = type; 554 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 555 vput(vp); 556 vp = nvp; 557 } 558 *vpp = vp; 559 return (0); 560} 561 562/* 563 * Check to see if the new vnode represents a special device 564 * for which we already have a vnode (either because of 565 * bdevvp() or because of a different vnode representing 566 * the same block device). If such an alias exists, deallocate 567 * the existing contents and return the aliased vnode. The 568 * caller is responsible for filling it with its new contents. 569 */ 570struct vnode * 571checkalias(nvp, nvp_rdev, mp) 572 register struct vnode *nvp; 573 dev_t nvp_rdev; 574 struct mount *mp; 575{ 576 struct proc *p = curproc; 577 register struct vnode *vp; 578 struct vnode **vpp; 579 580 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 581 return (NULLVP); 582 583 vpp = &speclisth[SPECHASH(nvp_rdev)]; 584loop: 585 simple_lock(&spechash_slock); 586 for (vp = *vpp; vp; vp = vp->v_specnext) { 587 simple_lock(&vp->v_interlock); 588 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 589 simple_unlock(&vp->v_interlock); 590 continue; 591 } 592 /* 593 * Alias, but not in use, so flush it out. 594 */ 595 if (vp->v_usecount == 0) { 596 simple_unlock(&spechash_slock); 597 vgonel(vp, p); 598 goto loop; 599 } 600 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 601 simple_unlock(&spechash_slock); 602 goto loop; 603 } 604 break; 605 } 606 607 /* 608 * Common case is actually in the if statement 609 */ 610 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 611 MALLOC(nvp->v_specinfo, struct specinfo *, 612 sizeof(struct specinfo), M_VNODE, M_WAITOK); 613 nvp->v_rdev = nvp_rdev; 614 nvp->v_hashchain = vpp; 615 nvp->v_specnext = *vpp; 616 nvp->v_specmountpoint = NULL; 617 nvp->v_speclockf = NULL; 618 simple_unlock(&spechash_slock); 619 *vpp = nvp; 620 if (vp != NULLVP) { 621 nvp->v_flag |= VALIASED; 622 vp->v_flag |= VALIASED; 623 vput(vp); 624 } 625 return (NULLVP); 626 } 627 628 /* 629 * This code is the uncommon case. It is called in case 630 * we found an alias that was VT_NON && vtype of VBLK 631 * This means we found a block device that was created 632 * using bdevvp. 633 * An example of such a vnode is the root partition device vnode 634 * created in ffs_mountroot. 635 * 636 * The vnodes created by bdevvp should not be aliased (why?). 637 */ 638 639 simple_unlock(&spechash_slock); 640 VOP_UNLOCK(vp, 0, p); 641 simple_lock(&vp->v_interlock); 642 vclean(vp, 0, p); 643 vp->v_vnlock = NULL; 644 lockinit(&vp->v_lock, PVFS, "v_lock", 0, 0); 645 vp->v_op = nvp->v_op; 646 vp->v_tag = nvp->v_tag; 647 nvp->v_type = VNON; 648 insmntque(vp, mp); 649 return (vp); 650} 651 652/* 653 * Grab a particular vnode from the free list, increment its 654 * reference count and lock it. The vnode lock bit is set the 655 * vnode is being eliminated in vgone. The process is awakened 656 * when the transition is completed, and an error returned to 657 * indicate that the vnode is no longer usable (possibly having 658 * been changed to a new file system type). 659 */ 660int 661vget(vp, flags, p) 662 struct vnode *vp; 663 int flags; 664 struct proc *p; 665{ 666 int error; 667 int s; 668 /* 669 * If the vnode is in the process of being cleaned out for 670 * another use, we wait for the cleaning to finish and then 671 * return failure. Cleaning is determined by checking that 672 * the VXLOCK flag is set. 673 */ 674 if ((flags & LK_INTERLOCK) == 0) { 675 simple_lock(&vp->v_interlock); 676 flags |= LK_INTERLOCK; 677 } 678 if (vp->v_flag & VXLOCK) { 679 vp->v_flag |= VXWANT; 680 simple_unlock(&vp->v_interlock); 681 tsleep((caddr_t)vp, PINOD, "vget", 0); 682 return (ENOENT); 683 } 684 if (vp->v_usecount == 0 && 685 (vp->v_bioflag & VBIOONFREELIST)) { 686 s = splbio(); 687 simple_lock(&vnode_free_list_slock); 688 if (vp->v_holdcnt > 0) 689 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 690 else 691 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 692 simple_unlock(&vnode_free_list_slock); 693 vp->v_bioflag &= ~VBIOONFREELIST; 694 splx(s); 695 } 696 vp->v_usecount++; 697 if (flags & LK_TYPE_MASK) { 698 if ((error = vn_lock(vp, flags, p)) != 0) { 699 vp->v_usecount--; 700 if (vp->v_usecount == 0) 701 vputonfreelist(vp); 702 703 simple_unlock(&vp->v_interlock); 704 } 705 return (error); 706 } 707 simple_unlock(&vp->v_interlock); 708 return (0); 709} 710 711 712#ifdef DIAGNOSTIC 713/* 714 * Vnode reference. 715 */ 716void 717vref(vp) 718 struct vnode *vp; 719{ 720 simple_lock(&vp->v_interlock); 721 if (vp->v_usecount == 0) 722 panic("vref used where vget required"); 723 vp->v_usecount++; 724 simple_unlock(&vp->v_interlock); 725} 726#endif /* DIAGNOSTIC */ 727 728static __inline__ void 729vputonfreelist(vp) 730 struct vnode *vp; 731{ 732 int s; 733 struct freelst *lst; 734 735 s = splbio(); 736#ifdef DIAGNOSTIC 737 if (vp->v_usecount != 0) 738 panic("Use count is not zero!"); 739 740 if (vp->v_bioflag & VBIOONFREELIST) { 741 vprint("vnode already on free list: ", vp); 742 panic("vnode already on free list"); 743 } 744#endif 745 746 vp->v_bioflag |= VBIOONFREELIST; 747 748 if (vp->v_holdcnt > 0) 749 lst = &vnode_hold_list; 750 else 751 lst = &vnode_free_list; 752 753 if (vp->v_type == VBAD) 754 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 755 else 756 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 757 758 splx(s); 759} 760 761/* 762 * vput(), just unlock and vrele() 763 */ 764void 765vput(vp) 766 register struct vnode *vp; 767{ 768 struct proc *p = curproc; /* XXX */ 769 770#ifdef DIAGNOSTIC 771 if (vp == NULL) 772 panic("vput: null vp"); 773#endif 774 simple_lock(&vp->v_interlock); 775 776#ifdef DIAGNOSTIC 777 if (vp->v_usecount == 0) { 778 vprint("vput: bad ref count", vp); 779 panic("vput: ref cnt"); 780 } 781#endif 782 vp->v_usecount--; 783 if (vp->v_usecount > 0) { 784 simple_unlock(&vp->v_interlock); 785 VOP_UNLOCK(vp, 0, p); 786 return; 787 } 788 789#ifdef DIAGNOSTIC 790 if (vp->v_writecount != 0) { 791 vprint("vput: bad writecount", vp); 792 panic("vput: v_writecount != 0"); 793 } 794#endif 795 vputonfreelist(vp); 796 797 simple_unlock(&vp->v_interlock); 798 799 VOP_INACTIVE(vp, p); 800} 801 802/* 803 * Vnode release - use for active VNODES. 804 * If count drops to zero, call inactive routine and return to freelist. 805 */ 806void 807vrele(vp) 808 register struct vnode *vp; 809{ 810 struct proc *p = curproc; /* XXX */ 811 812#ifdef DIAGNOSTIC 813 if (vp == NULL) 814 panic("vrele: null vp"); 815#endif 816 simple_lock(&vp->v_interlock); 817#ifdef DIAGNOSTIC 818 if (vp->v_usecount == 0) { 819 vprint("vrele: bad ref count", vp); 820 panic("vrele: ref cnt"); 821 } 822#endif 823 vp->v_usecount--; 824 if (vp->v_usecount > 0) { 825 simple_unlock(&vp->v_interlock); 826 return; 827 } 828 829#ifdef DIAGNOSTIC 830 if (vp->v_writecount != 0) { 831 vprint("vrele: bad writecount", vp); 832 panic("vrele: v_writecount != 0"); 833 } 834#endif 835 vputonfreelist(vp); 836 837 if (vn_lock(vp, LK_EXCLUSIVE|LK_INTERLOCK, p) == 0) 838 VOP_INACTIVE(vp, p); 839} 840 841void vhold(struct vnode *vp); 842 843/* 844 * Page or buffer structure gets a reference. 845 */ 846void 847vhold(vp) 848 register struct vnode *vp; 849{ 850 851 /* 852 * If it is on the freelist and the hold count is currently 853 * zero, move it to the hold list. 854 */ 855 simple_lock(&vp->v_interlock); 856 if ((vp->v_bioflag & VBIOONFREELIST) && 857 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 858 simple_lock(&vnode_free_list_slock); 859 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 860 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 861 simple_unlock(&vnode_free_list_slock); 862 } 863 vp->v_holdcnt++; 864 simple_unlock(&vp->v_interlock); 865} 866 867/* 868 * Remove any vnodes in the vnode table belonging to mount point mp. 869 * 870 * If MNT_NOFORCE is specified, there should not be any active ones, 871 * return error if any are found (nb: this is a user error, not a 872 * system error). If MNT_FORCE is specified, detach any active vnodes 873 * that are found. 874 */ 875#ifdef DEBUG 876int busyprt = 0; /* print out busy vnodes */ 877struct ctldebug debug1 = { "busyprt", &busyprt }; 878#endif 879 880int 881vfs_mount_foreach_vnode(struct mount *mp, 882 int (*func)(struct vnode *, void *), void *arg) { 883 struct vnode *vp, *nvp; 884 int error = 0; 885 886 simple_lock(&mntvnode_slock); 887loop: 888 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { 889 if (vp->v_mount != mp) 890 goto loop; 891 nvp = vp->v_mntvnodes.le_next; 892 simple_lock(&vp->v_interlock); 893 simple_unlock(&mntvnode_slock); 894 895 error = func(vp, arg); 896 897 simple_lock(&mntvnode_slock); 898 899 if (error != 0) 900 break; 901 } 902 simple_unlock(&mntvnode_slock); 903 904 return (error); 905} 906 907 908struct vflush_args { 909 struct vnode *skipvp; 910 int busy; 911 int flags; 912}; 913 914int 915vflush_vnode(struct vnode *vp, void *arg) { 916 struct vflush_args *va = arg; 917 struct proc *p = curproc; 918 919 if (vp == va->skipvp) { 920 simple_unlock(&vp->v_interlock); 921 return (0); 922 } 923 924 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 925 simple_unlock(&vp->v_interlock); 926 return (0); 927 } 928 929 /* 930 * If WRITECLOSE is set, only flush out regular file 931 * vnodes open for writing. 932 */ 933 if ((va->flags & WRITECLOSE) && 934 (vp->v_writecount == 0 || vp->v_type != VREG)) { 935 simple_unlock(&vp->v_interlock); 936 return (0); 937 } 938 939 /* 940 * With v_usecount == 0, all we need to do is clear 941 * out the vnode data structures and we are done. 942 */ 943 if (vp->v_usecount == 0) { 944 vgonel(vp, p); 945 return (0); 946 } 947 948 /* 949 * If FORCECLOSE is set, forcibly close the vnode. 950 * For block or character devices, revert to an 951 * anonymous device. For all other files, just kill them. 952 */ 953 if (va->flags & FORCECLOSE) { 954 if (vp->v_type != VBLK && vp->v_type != VCHR) { 955 vgonel(vp, p); 956 } else { 957 vclean(vp, 0, p); 958 vp->v_op = spec_vnodeop_p; 959 insmntque(vp, (struct mount *)0); 960 } 961 return (0); 962 } 963 964#ifdef DEBUG 965 if (busyprt) 966 vprint("vflush: busy vnode", vp); 967#endif 968 simple_unlock(&vp->v_interlock); 969 va->busy++; 970 return (0); 971} 972 973int 974vflush(mp, skipvp, flags) 975 struct mount *mp; 976 struct vnode *skipvp; 977 int flags; 978{ 979 struct vflush_args va; 980 va.skipvp = skipvp; 981 va.busy = 0; 982 va.flags = flags; 983 984 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 985 986 if (va.busy) 987 return (EBUSY); 988 return (0); 989} 990 991/* 992 * Disassociate the underlying file system from a vnode. 993 * The vnode interlock is held on entry. 994 */ 995void 996vclean(vp, flags, p) 997 register struct vnode *vp; 998 int flags; 999 struct proc *p; 1000{ 1001 int active; 1002 1003 /* 1004 * Check to see if the vnode is in use. 1005 * If so we have to reference it before we clean it out 1006 * so that its count cannot fall to zero and generate a 1007 * race against ourselves to recycle it. 1008 */ 1009 if ((active = vp->v_usecount) != 0) 1010 vp->v_usecount++; 1011 1012 /* 1013 * Prevent the vnode from being recycled or 1014 * brought into use while we clean it out. 1015 */ 1016 if (vp->v_flag & VXLOCK) 1017 panic("vclean: deadlock"); 1018 vp->v_flag |= VXLOCK; 1019 /* 1020 * Even if the count is zero, the VOP_INACTIVE routine may still 1021 * have the object locked while it cleans it out. The VOP_LOCK 1022 * ensures that the VOP_INACTIVE routine is done with its work. 1023 * For active vnodes, it ensures that no other activity can 1024 * occur while the underlying object is being cleaned out. 1025 */ 1026 VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK, p); 1027 1028 /* 1029 * clean out any VM data associated with the vnode. 1030 */ 1031 uvm_vnp_terminate(vp); 1032 /* 1033 * Clean out any buffers associated with the vnode. 1034 */ 1035 if (flags & DOCLOSE) 1036 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 1037 /* 1038 * If purging an active vnode, it must be closed and 1039 * deactivated before being reclaimed. Note that the 1040 * VOP_INACTIVE will unlock the vnode 1041 */ 1042 if (active) { 1043 if (flags & DOCLOSE) 1044 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 1045 VOP_INACTIVE(vp, p); 1046 } else { 1047 /* 1048 * Any other processes trying to obtain this lock must first 1049 * wait for VXLOCK to clear, then call the new lock operation. 1050 */ 1051 VOP_UNLOCK(vp, 0, p); 1052 } 1053 1054 /* 1055 * Reclaim the vnode. 1056 */ 1057 if (VOP_RECLAIM(vp, p)) 1058 panic("vclean: cannot reclaim"); 1059 if (active) { 1060 simple_lock(&vp->v_interlock); 1061 1062 vp->v_usecount--; 1063 if (vp->v_usecount == 0) { 1064 if (vp->v_holdcnt > 0) 1065 panic("vclean: not clean"); 1066 vputonfreelist(vp); 1067 } 1068 1069 simple_unlock(&vp->v_interlock); 1070 } 1071 cache_purge(vp); 1072 1073 /* 1074 * Done with purge, notify sleepers of the grim news. 1075 */ 1076 vp->v_op = dead_vnodeop_p; 1077 simple_lock(&vp->v_selectinfo.vsi_lock); 1078 VN_KNOTE(vp, NOTE_REVOKE); 1079 simple_unlock(&vp->v_selectinfo.vsi_lock); 1080 vp->v_tag = VT_NON; 1081 vp->v_flag &= ~VXLOCK; 1082#ifdef DIAGNOSTIC 1083 vp->v_flag &= ~VLOCKSWORK; 1084#endif 1085 if (vp->v_flag & VXWANT) { 1086 vp->v_flag &= ~VXWANT; 1087 wakeup((caddr_t)vp); 1088 } 1089} 1090 1091 1092 1093/* 1094 * Recycle an unused vnode to the front of the free list. 1095 * Release the passed interlock if the vnode will be recycled. 1096 */ 1097int 1098vrecycle(vp, inter_lkp, p) 1099 struct vnode *vp; 1100 struct simplelock *inter_lkp; 1101 struct proc *p; 1102{ 1103 1104 simple_lock(&vp->v_interlock); 1105 if (vp->v_usecount == 0) { 1106 if (inter_lkp) 1107 simple_unlock(inter_lkp); 1108 vgonel(vp, p); 1109 return (1); 1110 } 1111 simple_unlock(&vp->v_interlock); 1112 return (0); 1113} 1114 1115 1116/* 1117 * Eliminate all activity associated with a vnode 1118 * in preparation for reuse. 1119 */ 1120void 1121vgone(vp) 1122 register struct vnode *vp; 1123{ 1124 struct proc *p = curproc; 1125 1126 simple_lock (&vp->v_interlock); 1127 vgonel(vp, p); 1128} 1129 1130/* 1131 * vgone, with the vp interlock held. 1132 */ 1133void 1134vgonel(vp, p) 1135 struct vnode *vp; 1136 struct proc *p; 1137{ 1138 register struct vnode *vq; 1139 struct vnode *vx; 1140 1141 /* 1142 * If a vgone (or vclean) is already in progress, 1143 * wait until it is done and return. 1144 */ 1145 if (vp->v_flag & VXLOCK) { 1146 vp->v_flag |= VXWANT; 1147 simple_unlock(&vp->v_interlock); 1148 tsleep((caddr_t)vp, PINOD, "vgone", 0); 1149 return; 1150 } 1151 /* 1152 * Clean out the filesystem specific data. 1153 */ 1154 vclean(vp, DOCLOSE, p); 1155 /* 1156 * Delete from old mount point vnode list, if on one. 1157 */ 1158 if (vp->v_mount != NULL) 1159 insmntque(vp, (struct mount *)0); 1160 /* 1161 * If special device, remove it from special device alias list 1162 * if it is on one. 1163 */ 1164 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1165 simple_lock(&spechash_slock); 1166 if (*vp->v_hashchain == vp) { 1167 *vp->v_hashchain = vp->v_specnext; 1168 } else { 1169 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1170 if (vq->v_specnext != vp) 1171 continue; 1172 vq->v_specnext = vp->v_specnext; 1173 break; 1174 } 1175 if (vq == NULL) 1176 panic("missing bdev"); 1177 } 1178 if (vp->v_flag & VALIASED) { 1179 vx = NULL; 1180 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1181 if (vq->v_rdev != vp->v_rdev || 1182 vq->v_type != vp->v_type) 1183 continue; 1184 if (vx) 1185 break; 1186 vx = vq; 1187 } 1188 if (vx == NULL) 1189 panic("missing alias"); 1190 if (vq == NULL) 1191 vx->v_flag &= ~VALIASED; 1192 vp->v_flag &= ~VALIASED; 1193 } 1194 simple_unlock(&spechash_slock); 1195 FREE(vp->v_specinfo, M_VNODE); 1196 vp->v_specinfo = NULL; 1197 } 1198 /* 1199 * If it is on the freelist and not already at the head, 1200 * move it to the head of the list. 1201 */ 1202 vp->v_type = VBAD; 1203 1204 /* 1205 * Move onto the free list, unless we were called from 1206 * getnewvnode and we're not on any free list 1207 */ 1208 if (vp->v_usecount == 0 && 1209 (vp->v_bioflag & VBIOONFREELIST)) { 1210 int s; 1211 1212 simple_lock(&vnode_free_list_slock); 1213 s = splbio(); 1214 1215 if (vp->v_holdcnt > 0) 1216 panic("vgonel: not clean"); 1217 1218 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1219 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1220 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1221 } 1222 splx(s); 1223 simple_unlock(&vnode_free_list_slock); 1224 } 1225} 1226 1227/* 1228 * Lookup a vnode by device number. 1229 */ 1230int 1231vfinddev(dev, type, vpp) 1232 dev_t dev; 1233 enum vtype type; 1234 struct vnode **vpp; 1235{ 1236 register struct vnode *vp; 1237 int rc =0; 1238 1239 simple_lock(&spechash_slock); 1240 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1241 if (dev != vp->v_rdev || type != vp->v_type) 1242 continue; 1243 *vpp = vp; 1244 rc = 1; 1245 break; 1246 } 1247 simple_unlock(&spechash_slock); 1248 return (rc); 1249} 1250 1251/* 1252 * Revoke all the vnodes corresponding to the specified minor number 1253 * range (endpoints inclusive) of the specified major. 1254 */ 1255void 1256vdevgone(maj, minl, minh, type) 1257 int maj, minl, minh; 1258 enum vtype type; 1259{ 1260 struct vnode *vp; 1261 int mn; 1262 1263 for (mn = minl; mn <= minh; mn++) 1264 if (vfinddev(makedev(maj, mn), type, &vp)) 1265 VOP_REVOKE(vp, REVOKEALL); 1266} 1267 1268/* 1269 * Calculate the total number of references to a special device. 1270 */ 1271int 1272vcount(vp) 1273 struct vnode *vp; 1274{ 1275 struct vnode *vq, *vnext; 1276 int count; 1277 1278loop: 1279 if ((vp->v_flag & VALIASED) == 0) 1280 return (vp->v_usecount); 1281 simple_lock(&spechash_slock); 1282 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1283 vnext = vq->v_specnext; 1284 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1285 continue; 1286 /* 1287 * Alias, but not in use, so flush it out. 1288 */ 1289 if (vq->v_usecount == 0 && vq != vp) { 1290 simple_unlock(&spechash_slock); 1291 vgone(vq); 1292 goto loop; 1293 } 1294 count += vq->v_usecount; 1295 } 1296 simple_unlock(&spechash_slock); 1297 return (count); 1298} 1299 1300/* 1301 * Print out a description of a vnode. 1302 */ 1303static char *typename[] = 1304 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1305 1306void 1307vprint(label, vp) 1308 char *label; 1309 register struct vnode *vp; 1310{ 1311 char buf[64]; 1312 1313 if (label != NULL) 1314 printf("%s: ", label); 1315 printf("type %s, usecount %u, writecount %u, holdcount %u,", 1316 typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1317 vp->v_holdcnt); 1318 buf[0] = '\0'; 1319 if (vp->v_flag & VROOT) 1320 strlcat(buf, "|VROOT", sizeof buf); 1321 if (vp->v_flag & VTEXT) 1322 strlcat(buf, "|VTEXT", sizeof buf); 1323 if (vp->v_flag & VSYSTEM) 1324 strlcat(buf, "|VSYSTEM", sizeof buf); 1325 if (vp->v_flag & VXLOCK) 1326 strlcat(buf, "|VXLOCK", sizeof buf); 1327 if (vp->v_flag & VXWANT) 1328 strlcat(buf, "|VXWANT", sizeof buf); 1329 if (vp->v_bioflag & VBIOWAIT) 1330 strlcat(buf, "| VBIOWAIT", sizeof buf); 1331 if (vp->v_flag & VALIASED) 1332 strlcat(buf, "|VALIASED", sizeof buf); 1333 if (buf[0] != '\0') 1334 printf(" flags (%s)", &buf[1]); 1335 if (vp->v_data == NULL) { 1336 printf("\n"); 1337 } else { 1338 printf("\n\t"); 1339 VOP_PRINT(vp); 1340 } 1341} 1342 1343#ifdef DEBUG 1344/* 1345 * List all of the locked vnodes in the system. 1346 * Called when debugging the kernel. 1347 */ 1348void 1349printlockedvnodes() 1350{ 1351 struct proc *p = curproc; 1352 register struct mount *mp, *nmp; 1353 register struct vnode *vp; 1354 1355 printf("Locked vnodes\n"); 1356 simple_lock(&mountlist_slock); 1357 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1358 mp = nmp) { 1359 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { 1360 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1361 continue; 1362 } 1363 for (vp = mp->mnt_vnodelist.lh_first; vp; 1364 vp = vp->v_mntvnodes.le_next) { 1365 if (VOP_ISLOCKED(vp)) 1366 vprint((char *)0, vp); 1367 } 1368 simple_lock(&mountlist_slock); 1369 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1370 vfs_unbusy(mp, p); 1371 } 1372 simple_unlock(&mountlist_slock); 1373 1374} 1375#endif 1376 1377/* 1378 * Top level filesystem related information gathering. 1379 */ 1380int 1381vfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1382 int *name; 1383 u_int namelen; 1384 void *oldp; 1385 size_t *oldlenp; 1386 void *newp; 1387 size_t newlen; 1388 struct proc *p; 1389{ 1390 struct vfsconf *vfsp; 1391 1392 /* all sysctl names at this level are at least name and field */ 1393 if (namelen < 2) 1394 return (ENOTDIR); /* overloaded */ 1395 if (name[0] != VFS_GENERIC) { 1396 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1397 if (vfsp->vfc_typenum == name[0]) 1398 break; 1399 if (vfsp == NULL) 1400 return (EOPNOTSUPP); 1401 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1402 oldp, oldlenp, newp, newlen, p)); 1403 } 1404 switch (name[1]) { 1405 case VFS_MAXTYPENUM: 1406 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1407 case VFS_CONF: 1408 if (namelen < 3) 1409 return (ENOTDIR); /* overloaded */ 1410 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1411 if (vfsp->vfc_typenum == name[2]) 1412 break; 1413 if (vfsp == NULL) 1414 return (EOPNOTSUPP); 1415 return (sysctl_rdstruct(oldp, oldlenp, newp, vfsp, 1416 sizeof(struct vfsconf))); 1417 } 1418 return (EOPNOTSUPP); 1419} 1420 1421 1422int kinfo_vdebug = 1; 1423int kinfo_vgetfailed; 1424#define KINFO_VNODESLOP 10 1425/* 1426 * Dump vnode list (via sysctl). 1427 * Copyout address of vnode followed by vnode. 1428 */ 1429/* ARGSUSED */ 1430int 1431sysctl_vnode(where, sizep, p) 1432 char *where; 1433 size_t *sizep; 1434 struct proc *p; 1435{ 1436 register struct mount *mp, *nmp; 1437 struct vnode *vp, *nvp; 1438 register char *bp = where, *savebp; 1439 char *ewhere; 1440 int error; 1441 1442 if (where == NULL) { 1443 *sizep = (numvnodes + KINFO_VNODESLOP) * sizeof(struct e_vnode); 1444 return (0); 1445 } 1446 ewhere = where + *sizep; 1447 1448 simple_lock(&mountlist_slock); 1449 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1450 mp = nmp) { 1451 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { 1452 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1453 continue; 1454 } 1455 savebp = bp; 1456again: 1457 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; 1458 vp = nvp) { 1459 /* 1460 * Check that the vp is still associated with 1461 * this filesystem. RACE: could have been 1462 * recycled onto the same filesystem. 1463 */ 1464 if (vp->v_mount != mp) { 1465 simple_unlock(&mntvnode_slock); 1466 if (kinfo_vdebug) 1467 printf("kinfo: vp changed\n"); 1468 bp = savebp; 1469 goto again; 1470 } 1471 nvp = vp->v_mntvnodes.le_next; 1472 if (bp + sizeof(struct e_vnode) > ewhere) { 1473 simple_unlock(&mntvnode_slock); 1474 *sizep = bp - where; 1475 vfs_unbusy(mp, p); 1476 return (ENOMEM); 1477 } 1478 if ((error = copyout((caddr_t)&vp, 1479 &((struct e_vnode *)bp)->vptr, 1480 sizeof(struct vnode *))) || 1481 (error = copyout((caddr_t)vp, 1482 &((struct e_vnode *)bp)->vnode, 1483 sizeof(struct vnode)))) { 1484 vfs_unbusy(mp, p); 1485 return (error); 1486 } 1487 bp += sizeof(struct e_vnode); 1488 simple_lock(&mntvnode_slock); 1489 } 1490 1491 simple_unlock(&mntvnode_slock); 1492 simple_lock(&mountlist_slock); 1493 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1494 vfs_unbusy(mp, p); 1495 } 1496 1497 simple_unlock(&mountlist_slock); 1498 1499 *sizep = bp - where; 1500 return (0); 1501} 1502 1503/* 1504 * Check to see if a filesystem is mounted on a block device. 1505 */ 1506int 1507vfs_mountedon(vp) 1508 register struct vnode *vp; 1509{ 1510 register struct vnode *vq; 1511 int error = 0; 1512 1513 if (vp->v_specmountpoint != NULL) 1514 return (EBUSY); 1515 if (vp->v_flag & VALIASED) { 1516 simple_lock(&spechash_slock); 1517 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1518 if (vq->v_rdev != vp->v_rdev || 1519 vq->v_type != vp->v_type) 1520 continue; 1521 if (vq->v_specmountpoint != NULL) { 1522 error = EBUSY; 1523 break; 1524 } 1525 } 1526 simple_unlock(&spechash_slock); 1527 } 1528 return (error); 1529} 1530 1531/* 1532 * Build hash lists of net addresses and hang them off the mount point. 1533 * Called by ufs_mount() to set up the lists of export addresses. 1534 */ 1535int 1536vfs_hang_addrlist(mp, nep, argp) 1537 struct mount *mp; 1538 struct netexport *nep; 1539 struct export_args *argp; 1540{ 1541 register struct netcred *np; 1542 register struct radix_node_head *rnh; 1543 register int i; 1544 struct radix_node *rn; 1545 struct sockaddr *saddr, *smask = 0; 1546 struct domain *dom; 1547 int error; 1548 1549 if (argp->ex_addrlen == 0) { 1550 if (mp->mnt_flag & MNT_DEFEXPORTED) 1551 return (EPERM); 1552 np = &nep->ne_defexported; 1553 np->netc_exflags = argp->ex_flags; 1554 np->netc_anon = argp->ex_anon; 1555 np->netc_anon.cr_ref = 1; 1556 mp->mnt_flag |= MNT_DEFEXPORTED; 1557 return (0); 1558 } 1559 if (argp->ex_addrlen > MLEN) 1560 return (EINVAL); 1561 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1562 np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK); 1563 bzero((caddr_t)np, i); 1564 saddr = (struct sockaddr *)(np + 1); 1565 error = copyin(argp->ex_addr, (caddr_t)saddr, argp->ex_addrlen); 1566 if (error) 1567 goto out; 1568 if (saddr->sa_len > argp->ex_addrlen) 1569 saddr->sa_len = argp->ex_addrlen; 1570 if (argp->ex_masklen) { 1571 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1572 error = copyin(argp->ex_mask, (caddr_t)smask, argp->ex_masklen); 1573 if (error) 1574 goto out; 1575 if (smask->sa_len > argp->ex_masklen) 1576 smask->sa_len = argp->ex_masklen; 1577 } 1578 i = saddr->sa_family; 1579 if (i < 0 || i > AF_MAX) { 1580 error = EINVAL; 1581 goto out; 1582 } 1583 if ((rnh = nep->ne_rtable[i]) == 0) { 1584 /* 1585 * Seems silly to initialize every AF when most are not 1586 * used, do so on demand here 1587 */ 1588 for (dom = domains; dom; dom = dom->dom_next) 1589 if (dom->dom_family == i && dom->dom_rtattach) { 1590 dom->dom_rtattach((void **)&nep->ne_rtable[i], 1591 dom->dom_rtoffset); 1592 break; 1593 } 1594 if ((rnh = nep->ne_rtable[i]) == 0) { 1595 error = ENOBUFS; 1596 goto out; 1597 } 1598 } 1599 rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, 1600 np->netc_rnodes); 1601 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1602 error = EPERM; 1603 goto out; 1604 } 1605 np->netc_exflags = argp->ex_flags; 1606 np->netc_anon = argp->ex_anon; 1607 np->netc_anon.cr_ref = 1; 1608 return (0); 1609out: 1610 free(np, M_NETADDR); 1611 return (error); 1612} 1613 1614/* ARGSUSED */ 1615int 1616vfs_free_netcred(rn, w) 1617 struct radix_node *rn; 1618 void *w; 1619{ 1620 register struct radix_node_head *rnh = (struct radix_node_head *)w; 1621 1622 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh); 1623 free((caddr_t)rn, M_NETADDR); 1624 return (0); 1625} 1626 1627/* 1628 * Free the net address hash lists that are hanging off the mount points. 1629 */ 1630void 1631vfs_free_addrlist(nep) 1632 struct netexport *nep; 1633{ 1634 register int i; 1635 register struct radix_node_head *rnh; 1636 1637 for (i = 0; i <= AF_MAX; i++) 1638 if ((rnh = nep->ne_rtable[i]) != NULL) { 1639 (*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh); 1640 free((caddr_t)rnh, M_RTABLE); 1641 nep->ne_rtable[i] = 0; 1642 } 1643} 1644 1645int 1646vfs_export(mp, nep, argp) 1647 struct mount *mp; 1648 struct netexport *nep; 1649 struct export_args *argp; 1650{ 1651 int error; 1652 1653 if (argp->ex_flags & MNT_DELEXPORT) { 1654 vfs_free_addrlist(nep); 1655 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1656 } 1657 if (argp->ex_flags & MNT_EXPORTED) { 1658 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1659 return (error); 1660 mp->mnt_flag |= MNT_EXPORTED; 1661 } 1662 return (0); 1663} 1664 1665struct netcred * 1666vfs_export_lookup(mp, nep, nam) 1667 register struct mount *mp; 1668 struct netexport *nep; 1669 struct mbuf *nam; 1670{ 1671 register struct netcred *np; 1672 register struct radix_node_head *rnh; 1673 struct sockaddr *saddr; 1674 1675 np = NULL; 1676 if (mp->mnt_flag & MNT_EXPORTED) { 1677 /* 1678 * Lookup in the export list first. 1679 */ 1680 if (nam != NULL) { 1681 saddr = mtod(nam, struct sockaddr *); 1682 rnh = nep->ne_rtable[saddr->sa_family]; 1683 if (rnh != NULL) { 1684 np = (struct netcred *) 1685 (*rnh->rnh_matchaddr)((caddr_t)saddr, 1686 rnh); 1687 if (np && np->netc_rnodes->rn_flags & RNF_ROOT) 1688 np = NULL; 1689 } 1690 } 1691 /* 1692 * If no address match, use the default if it exists. 1693 */ 1694 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1695 np = &nep->ne_defexported; 1696 } 1697 return (np); 1698} 1699 1700/* 1701 * Do the usual access checking. 1702 * file_mode, uid and gid are from the vnode in question, 1703 * while acc_mode and cred are from the VOP_ACCESS parameter list 1704 */ 1705int 1706vaccess(file_mode, uid, gid, acc_mode, cred) 1707 mode_t file_mode; 1708 uid_t uid; 1709 gid_t gid; 1710 mode_t acc_mode; 1711 struct ucred *cred; 1712{ 1713 mode_t mask; 1714 1715 /* User id 0 always gets access. */ 1716 if (cred->cr_uid == 0) 1717 return 0; 1718 1719 mask = 0; 1720 1721 /* Otherwise, check the owner. */ 1722 if (cred->cr_uid == uid) { 1723 if (acc_mode & VEXEC) 1724 mask |= S_IXUSR; 1725 if (acc_mode & VREAD) 1726 mask |= S_IRUSR; 1727 if (acc_mode & VWRITE) 1728 mask |= S_IWUSR; 1729 return (file_mode & mask) == mask ? 0 : EACCES; 1730 } 1731 1732 /* Otherwise, check the groups. */ 1733 if (cred->cr_gid == gid || groupmember(gid, cred)) { 1734 if (acc_mode & VEXEC) 1735 mask |= S_IXGRP; 1736 if (acc_mode & VREAD) 1737 mask |= S_IRGRP; 1738 if (acc_mode & VWRITE) 1739 mask |= S_IWGRP; 1740 return (file_mode & mask) == mask ? 0 : EACCES; 1741 } 1742 1743 /* Otherwise, check everyone else. */ 1744 if (acc_mode & VEXEC) 1745 mask |= S_IXOTH; 1746 if (acc_mode & VREAD) 1747 mask |= S_IROTH; 1748 if (acc_mode & VWRITE) 1749 mask |= S_IWOTH; 1750 return (file_mode & mask) == mask ? 0 : EACCES; 1751} 1752 1753/* 1754 * Unmount all file systems. 1755 * We traverse the list in reverse order under the assumption that doing so 1756 * will avoid needing to worry about dependencies. 1757 */ 1758void 1759vfs_unmountall(void) 1760{ 1761 struct mount *mp, *nmp; 1762 int allerror, error, again = 1; 1763 struct proc *p = curproc; 1764 1765 retry: 1766 allerror = 0; 1767 for (mp = CIRCLEQ_LAST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1768 mp = nmp) { 1769 nmp = CIRCLEQ_PREV(mp, mnt_list); 1770 if ((vfs_busy(mp, LK_EXCLUSIVE|LK_NOWAIT, NULL, p)) != 0) 1771 continue; 1772 if ((error = dounmount(mp, MNT_FORCE, curproc, NULL)) != 0) { 1773 printf("unmount of %s failed with error %d\n", 1774 mp->mnt_stat.f_mntonname, error); 1775 allerror = 1; 1776 } 1777 } 1778 1779 if (allerror) { 1780 printf("WARNING: some file systems would not unmount\n"); 1781 if (again) { 1782 printf("retrying\n"); 1783 again = 0; 1784 goto retry; 1785 } 1786 } 1787} 1788 1789/* 1790 * Sync and unmount file systems before shutting down. 1791 */ 1792void 1793vfs_shutdown() 1794{ 1795 /* XXX Should suspend scheduling. */ 1796 (void) spl0(); 1797 1798 printf("syncing disks... "); 1799 1800 if (panicstr == 0) { 1801 /* Sync before unmount, in case we hang on something. */ 1802 sys_sync(&proc0, (void *)0, (register_t *)0); 1803 1804 /* Unmount file systems. */ 1805 vfs_unmountall(); 1806 } 1807 1808 if (vfs_syncwait(1)) 1809 printf("giving up\n"); 1810 else 1811 printf("done\n"); 1812} 1813 1814/* 1815 * perform sync() operation and wait for buffers to flush. 1816 * assumtions: called w/ scheduler disabled and physical io enabled 1817 * for now called at spl0() XXX 1818 */ 1819int 1820vfs_syncwait(verbose) 1821 int verbose; 1822{ 1823 register struct buf *bp; 1824 int iter, nbusy, dcount, s; 1825 struct proc *p; 1826 1827 p = curproc? curproc : &proc0; 1828 sys_sync(p, (void *)0, (register_t *)0); 1829 1830 /* Wait for sync to finish. */ 1831 dcount = 10000; 1832 for (iter = 0; iter < 20; iter++) { 1833 nbusy = 0; 1834 for (bp = &buf[nbuf]; --bp >= buf; ) { 1835 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1836 nbusy++; 1837 /* 1838 * With soft updates, some buffers that are 1839 * written will be remarked as dirty until other 1840 * buffers are written. 1841 */ 1842 if (bp->b_flags & B_DELWRI) { 1843 s = splbio(); 1844 bremfree(bp); 1845 bp->b_flags |= B_BUSY; 1846 splx(s); 1847 nbusy++; 1848 bawrite(bp); 1849 if (dcount-- <= 0) { 1850 if (verbose) 1851 printf("softdep "); 1852 return 1; 1853 } 1854 } 1855 } 1856 if (nbusy == 0) 1857 break; 1858 if (verbose) 1859 printf("%d ", nbusy); 1860 DELAY(40000 * iter); 1861 } 1862 1863 return nbusy; 1864} 1865 1866/* 1867 * posix file system related system variables. 1868 */ 1869int 1870fs_posix_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1871 int *name; 1872 u_int namelen; 1873 void *oldp; 1874 size_t *oldlenp; 1875 void *newp; 1876 size_t newlen; 1877 struct proc *p; 1878{ 1879 /* all sysctl names at this level are terminal */ 1880 if (namelen != 1) 1881 return (ENOTDIR); 1882 1883 switch (name[0]) { 1884 case FS_POSIX_SETUID: 1885 if (newp && securelevel > 0) 1886 return (EPERM); 1887 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1888 default: 1889 return (EOPNOTSUPP); 1890 } 1891 /* NOTREACHED */ 1892} 1893 1894/* 1895 * file system related system variables. 1896 */ 1897int 1898fs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1899 int *name; 1900 u_int namelen; 1901 void *oldp; 1902 size_t *oldlenp; 1903 void *newp; 1904 size_t newlen; 1905 struct proc *p; 1906{ 1907 sysctlfn *fn; 1908 1909 switch (name[0]) { 1910 case FS_POSIX: 1911 fn = fs_posix_sysctl; 1912 break; 1913 default: 1914 return (EOPNOTSUPP); 1915 } 1916 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1917} 1918 1919 1920/* 1921 * Routines dealing with vnodes and buffers 1922 */ 1923 1924/* 1925 * Wait for all outstanding I/Os to complete 1926 * 1927 * Manipulates v_numoutput. Must be called at splbio() 1928 */ 1929int 1930vwaitforio(vp, slpflag, wmesg, timeo) 1931 struct vnode *vp; 1932 int slpflag, timeo; 1933 char *wmesg; 1934{ 1935 int error = 0; 1936 1937 splassert(IPL_BIO); 1938 1939 while (vp->v_numoutput) { 1940 vp->v_bioflag |= VBIOWAIT; 1941 error = tsleep((caddr_t)&vp->v_numoutput, 1942 slpflag | (PRIBIO + 1), wmesg, timeo); 1943 if (error) 1944 break; 1945 } 1946 1947 return (error); 1948} 1949 1950 1951/* 1952 * Update outstanding I/O count and do wakeup if requested. 1953 * 1954 * Manipulates v_numoutput. Must be called at splbio() 1955 */ 1956void 1957vwakeup(vp) 1958 struct vnode *vp; 1959{ 1960 splassert(IPL_BIO); 1961 1962 if (vp != NULL) { 1963 if (vp->v_numoutput-- == 0) 1964 panic("vwakeup: neg numoutput"); 1965 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1966 vp->v_bioflag &= ~VBIOWAIT; 1967 wakeup((caddr_t)&vp->v_numoutput); 1968 } 1969 } 1970} 1971 1972/* 1973 * Flush out and invalidate all buffers associated with a vnode. 1974 * Called with the underlying object locked. 1975 */ 1976int 1977vinvalbuf(vp, flags, cred, p, slpflag, slptimeo) 1978 register struct vnode *vp; 1979 int flags; 1980 struct ucred *cred; 1981 struct proc *p; 1982 int slpflag, slptimeo; 1983{ 1984 register struct buf *bp; 1985 struct buf *nbp, *blist; 1986 int s, error; 1987 1988 if (flags & V_SAVE) { 1989 s = splbio(); 1990 vwaitforio(vp, 0, "vinvalbuf", 0); 1991 if (vp->v_dirtyblkhd.lh_first != NULL) { 1992 splx(s); 1993 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1994 return (error); 1995 s = splbio(); 1996 if (vp->v_numoutput > 0 || 1997 vp->v_dirtyblkhd.lh_first != NULL) 1998 panic("vinvalbuf: dirty bufs"); 1999 } 2000 splx(s); 2001 } 2002loop: 2003 s = splbio(); 2004 for (;;) { 2005 if ((blist = vp->v_cleanblkhd.lh_first) && 2006 (flags & V_SAVEMETA)) 2007 while (blist && blist->b_lblkno < 0) 2008 blist = blist->b_vnbufs.le_next; 2009 if (!blist && (blist = vp->v_dirtyblkhd.lh_first) && 2010 (flags & V_SAVEMETA)) 2011 while (blist && blist->b_lblkno < 0) 2012 blist = blist->b_vnbufs.le_next; 2013 if (!blist) 2014 break; 2015 2016 for (bp = blist; bp; bp = nbp) { 2017 nbp = bp->b_vnbufs.le_next; 2018 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 2019 continue; 2020 if (bp->b_flags & B_BUSY) { 2021 bp->b_flags |= B_WANTED; 2022 error = tsleep((caddr_t)bp, 2023 slpflag | (PRIBIO + 1), "vinvalbuf", 2024 slptimeo); 2025 if (error) { 2026 splx(s); 2027 return (error); 2028 } 2029 break; 2030 } 2031 bremfree(bp); 2032 bp->b_flags |= B_BUSY; 2033 /* 2034 * XXX Since there are no node locks for NFS, I believe 2035 * there is a slight chance that a delayed write will 2036 * occur while sleeping just above, so check for it. 2037 */ 2038 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 2039 splx(s); 2040 (void) VOP_BWRITE(bp); 2041 goto loop; 2042 } 2043 bp->b_flags |= B_INVAL; 2044 brelse(bp); 2045 } 2046 } 2047 if (!(flags & V_SAVEMETA) && 2048 (vp->v_dirtyblkhd.lh_first || vp->v_cleanblkhd.lh_first)) 2049 panic("vinvalbuf: flush failed"); 2050 splx(s); 2051 return (0); 2052} 2053 2054void 2055vflushbuf(vp, sync) 2056 register struct vnode *vp; 2057 int sync; 2058{ 2059 register struct buf *bp, *nbp; 2060 int s; 2061 2062loop: 2063 s = splbio(); 2064 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { 2065 nbp = bp->b_vnbufs.le_next; 2066 if ((bp->b_flags & B_BUSY)) 2067 continue; 2068 if ((bp->b_flags & B_DELWRI) == 0) 2069 panic("vflushbuf: not dirty"); 2070 bremfree(bp); 2071 bp->b_flags |= B_BUSY; 2072 splx(s); 2073 /* 2074 * Wait for I/O associated with indirect blocks to complete, 2075 * since there is no way to quickly wait for them below. 2076 */ 2077 if (bp->b_vp == vp || sync == 0) 2078 (void) bawrite(bp); 2079 else 2080 (void) bwrite(bp); 2081 goto loop; 2082 } 2083 if (sync == 0) { 2084 splx(s); 2085 return; 2086 } 2087 vwaitforio(vp, 0, "vflushbuf", 0); 2088 if (vp->v_dirtyblkhd.lh_first != NULL) { 2089 splx(s); 2090 vprint("vflushbuf: dirty", vp); 2091 goto loop; 2092 } 2093 splx(s); 2094} 2095 2096/* 2097 * Associate a buffer with a vnode. 2098 * 2099 * Manipulates buffer vnode queues. Must be called at splbio(). 2100 */ 2101void 2102bgetvp(vp, bp) 2103 register struct vnode *vp; 2104 register struct buf *bp; 2105{ 2106 splassert(IPL_BIO); 2107 2108 2109 if (bp->b_vp) 2110 panic("bgetvp: not free"); 2111 vhold(vp); 2112 bp->b_vp = vp; 2113 if (vp->v_type == VBLK || vp->v_type == VCHR) 2114 bp->b_dev = vp->v_rdev; 2115 else 2116 bp->b_dev = NODEV; 2117 /* 2118 * Insert onto list for new vnode. 2119 */ 2120 bufinsvn(bp, &vp->v_cleanblkhd); 2121} 2122 2123/* 2124 * Disassociate a buffer from a vnode. 2125 * 2126 * Manipulates vnode buffer queues. Must be called at splbio(). 2127 */ 2128void 2129brelvp(bp) 2130 register struct buf *bp; 2131{ 2132 struct vnode *vp; 2133 2134 splassert(IPL_BIO); 2135 2136 if ((vp = bp->b_vp) == (struct vnode *) 0) 2137 panic("brelvp: NULL"); 2138 /* 2139 * Delete from old vnode list, if on one. 2140 */ 2141 if (bp->b_vnbufs.le_next != NOLIST) 2142 bufremvn(bp); 2143 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2144 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2145 vp->v_bioflag &= ~VBIOONSYNCLIST; 2146 LIST_REMOVE(vp, v_synclist); 2147 } 2148 bp->b_vp = (struct vnode *) 0; 2149 2150 simple_lock(&vp->v_interlock); 2151#ifdef DIAGNOSTIC 2152 if (vp->v_holdcnt == 0) 2153 panic("brelvp: holdcnt"); 2154#endif 2155 vp->v_holdcnt--; 2156 2157 /* 2158 * If it is on the holdlist and the hold count drops to 2159 * zero, move it to the free list. 2160 */ 2161 if ((vp->v_bioflag & VBIOONFREELIST) && 2162 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 2163 simple_lock(&vnode_free_list_slock); 2164 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 2165 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 2166 simple_unlock(&vnode_free_list_slock); 2167 } 2168 simple_unlock(&vp->v_interlock); 2169} 2170 2171/* 2172 * Replaces the current vnode associated with the buffer, if any 2173 * with a new vnode. 2174 * 2175 * If an output I/O is pending on the buffer, the old vnode is 2176 * I/O count is adjusted. 2177 * 2178 * Ignores vnode buffer queues. Must be called at splbio(). 2179 */ 2180void 2181buf_replacevnode(bp, newvp) 2182 struct buf *bp; 2183 struct vnode *newvp; 2184{ 2185 struct vnode *oldvp = bp->b_vp; 2186 2187 splassert(IPL_BIO); 2188 2189 if (oldvp) 2190 brelvp(bp); 2191 2192 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2193 newvp->v_numoutput++; /* put it on swapdev */ 2194 vwakeup(oldvp); 2195 } 2196 2197 bgetvp(newvp, bp); 2198 bufremvn(bp); 2199} 2200 2201/* 2202 * Used to assign buffers to the appropriate clean or dirty list on 2203 * the vnode and to add newly dirty vnodes to the appropriate 2204 * filesystem syncer list. 2205 * 2206 * Manipulates vnode buffer queues. Must be called at splbio(). 2207 */ 2208void 2209reassignbuf(bp) 2210 struct buf *bp; 2211{ 2212 struct buflists *listheadp; 2213 int delay; 2214 struct vnode *vp = bp->b_vp; 2215 2216 splassert(IPL_BIO); 2217 2218 /* 2219 * Delete from old vnode list, if on one. 2220 */ 2221 if (bp->b_vnbufs.le_next != NOLIST) 2222 bufremvn(bp); 2223 /* 2224 * If dirty, put on list of dirty buffers; 2225 * otherwise insert onto list of clean buffers. 2226 */ 2227 if ((bp->b_flags & B_DELWRI) == 0) { 2228 listheadp = &vp->v_cleanblkhd; 2229 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2230 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2231 vp->v_bioflag &= ~VBIOONSYNCLIST; 2232 LIST_REMOVE(vp, v_synclist); 2233 } 2234 } else { 2235 listheadp = &vp->v_dirtyblkhd; 2236 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2237 switch (vp->v_type) { 2238 case VDIR: 2239 delay = syncdelay / 2; 2240 break; 2241 case VBLK: 2242 if (vp->v_specmountpoint != NULL) { 2243 delay = syncdelay / 3; 2244 break; 2245 } 2246 /* fall through */ 2247 default: 2248 delay = syncdelay; 2249 } 2250 vn_syncer_add_to_worklist(vp, delay); 2251 } 2252 } 2253 bufinsvn(bp, listheadp); 2254} 2255 2256int 2257vfs_register(vfs) 2258 struct vfsconf *vfs; 2259{ 2260 struct vfsconf *vfsp; 2261 struct vfsconf **vfspp; 2262 2263#ifdef DIAGNOSTIC 2264 /* Paranoia? */ 2265 if (vfs->vfc_refcount != 0) 2266 printf("vfs_register called with vfc_refcount > 0\n"); 2267#endif 2268 2269 /* Check if filesystem already known */ 2270 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2271 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2272 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2273 return (EEXIST); 2274 2275 if (vfs->vfc_typenum > maxvfsconf) 2276 maxvfsconf = vfs->vfc_typenum; 2277 2278 vfs->vfc_next = NULL; 2279 2280 /* Add to the end of the list */ 2281 *vfspp = vfs; 2282 2283 /* Call vfs_init() */ 2284 if (vfs->vfc_vfsops->vfs_init) 2285 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2286 2287 return 0; 2288} 2289 2290int 2291vfs_unregister(vfs) 2292 struct vfsconf *vfs; 2293{ 2294 struct vfsconf *vfsp; 2295 struct vfsconf **vfspp; 2296 int maxtypenum; 2297 2298 /* Find our vfsconf struct */ 2299 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2300 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2301 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2302 break; 2303 } 2304 2305 if (!vfsp) /* Not found */ 2306 return (ENOENT); 2307 2308 if (vfsp->vfc_refcount) /* In use */ 2309 return (EBUSY); 2310 2311 /* Remove from list and free */ 2312 *vfspp = vfsp->vfc_next; 2313 2314 maxtypenum = 0; 2315 2316 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2317 if (vfsp->vfc_typenum > maxtypenum) 2318 maxtypenum = vfsp->vfc_typenum; 2319 2320 maxvfsconf = maxtypenum; 2321 return 0; 2322} 2323 2324/* 2325 * Check if vnode represents a disk device 2326 */ 2327int 2328vn_isdisk(vp, errp) 2329 struct vnode *vp; 2330 int *errp; 2331{ 2332 if (vp->v_type != VBLK && vp->v_type != VCHR) 2333 return (0); 2334 2335 return (1); 2336} 2337