vfs_subr.c revision 1.79
1/* $OpenBSD: vfs_subr.c,v 1.79 2001/12/10 18:47:16 art 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 __P((struct vnode *, int, struct proc *)); 105 106void insmntque __P((struct vnode *, struct mount *)); 107int getdevvp __P((dev_t, struct vnode **, enum vtype)); 108 109int vfs_hang_addrlist __P((struct mount *, struct netexport *, 110 struct export_args *)); 111int vfs_free_netcred __P((struct radix_node *, void *)); 112void vfs_free_addrlist __P((struct netexport *)); 113static __inline__ void vputonfreelist __P((struct vnode *)); 114 115int vflush_vnode(struct vnode *, void *); 116 117#ifdef DEBUG 118void printlockedvnodes __P((void)); 119#endif 120 121#define VN_KNOTE(vp, b) \ 122 KNOTE((struct klist *)&vp->v_selectinfo.vsi_selinfo.si_note, (b)) 123 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 0, pool_page_alloc_nointr, pool_page_free_nointr, M_VNODE); 135 simple_lock_init(&mntvnode_slock); 136 simple_lock_init(&mntid_slock); 137 simple_lock_init(&spechash_slock); 138 TAILQ_INIT(&vnode_hold_list); 139 TAILQ_INIT(&vnode_free_list); 140 simple_lock_init(&vnode_free_list_slock); 141 CIRCLEQ_INIT(&mountlist); 142 simple_lock_init(&mountlist_slock); 143 /* 144 * Initialize the filesystem syncer. 145 */ 146 vn_initialize_syncerd(); 147} 148 149 150/* 151 * Mark a mount point as busy. Used to synchronize access and to delay 152 * unmounting. Interlock is not released on failure. 153 */ 154 155int 156vfs_busy(mp, flags, interlkp, p) 157 struct mount *mp; 158 int flags; 159 struct simplelock *interlkp; 160 struct proc *p; 161{ 162 int lkflags; 163 164 if (mp->mnt_flag & MNT_UNMOUNT) { 165 if (flags & LK_NOWAIT) 166 return (ENOENT); 167 mp->mnt_flag |= MNT_MWAIT; 168 if (interlkp) 169 simple_unlock(interlkp); 170 /* 171 * Since all busy locks are shared except the exclusive 172 * lock granted when unmounting, the only place that a 173 * wakeup needs to be done is at the release of the 174 * exclusive lock at the end of dounmount. 175 */ 176 sleep((caddr_t)mp, PVFS); 177 if (interlkp) 178 simple_lock(interlkp); 179 return (ENOENT); 180 } 181 lkflags = LK_SHARED; 182 if (interlkp) 183 lkflags |= LK_INTERLOCK; 184 if (lockmgr(&mp->mnt_lock, lkflags, interlkp, p)) 185 panic("vfs_busy: unexpected lock failure"); 186 return (0); 187} 188 189 190/* 191 * Free a busy file system 192 */ 193void 194vfs_unbusy(mp, p) 195 struct mount *mp; 196 struct proc *p; 197{ 198 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, p); 199} 200 201int 202vfs_isbusy(struct mount *mp) 203{ 204 return (lockstatus(&mp->mnt_lock)); 205} 206 207/* 208 * Lookup a filesystem type, and if found allocate and initialize 209 * a mount structure for it. 210 * 211 * Devname is usually updated by mount(8) after booting. 212 */ 213 214int 215vfs_rootmountalloc(fstypename, devname, mpp) 216 char *fstypename; 217 char *devname; 218 struct mount **mpp; 219{ 220 struct proc *p = curproc; /* XXX */ 221 struct vfsconf *vfsp; 222 struct mount *mp; 223 224 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 225 if (!strcmp(vfsp->vfc_name, fstypename)) 226 break; 227 if (vfsp == NULL) 228 return (ENODEV); 229 mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 230 bzero((char *)mp, (u_long)sizeof(struct mount)); 231 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 232 (void)vfs_busy(mp, LK_NOWAIT, 0, p); 233 LIST_INIT(&mp->mnt_vnodelist); 234 mp->mnt_vfc = vfsp; 235 mp->mnt_op = vfsp->vfc_vfsops; 236 mp->mnt_flag = MNT_RDONLY; 237 mp->mnt_vnodecovered = NULLVP; 238 vfsp->vfc_refcount++; 239 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 240 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 241 mp->mnt_stat.f_mntonname[0] = '/'; 242 (void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); 243 *mpp = mp; 244 return (0); 245 } 246 247/* 248 * Find an appropriate filesystem to use for the root. If a filesystem 249 * has not been preselected, walk through the list of known filesystems 250 * trying those that have mountroot routines, and try them until one 251 * works or we have tried them all. 252 */ 253int 254vfs_mountroot() 255{ 256 struct vfsconf *vfsp; 257 extern int (*mountroot)(void); 258 int error; 259 260 if (mountroot != NULL) 261 return ((*mountroot)()); 262 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) { 263 if (vfsp->vfc_mountroot == NULL) 264 continue; 265 if ((error = (*vfsp->vfc_mountroot)()) == 0) 266 return (0); 267 printf("%s_mountroot failed: %d\n", vfsp->vfc_name, error); 268 } 269 return (ENODEV); 270} 271 272/* 273 * Lookup a mount point by filesystem identifier. 274 */ 275struct mount * 276vfs_getvfs(fsid) 277 fsid_t *fsid; 278{ 279 register struct mount *mp; 280 281 simple_lock(&mountlist_slock); 282 CIRCLEQ_FOREACH(mp, &mountlist, mnt_list) { 283 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 284 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 285 simple_unlock(&mountlist_slock); 286 return (mp); 287 } 288 } 289 simple_unlock(&mountlist_slock); 290 return ((struct mount *)0); 291} 292 293 294/* 295 * Get a new unique fsid 296 */ 297void 298vfs_getnewfsid(mp) 299 struct mount *mp; 300{ 301 static u_short xxxfs_mntid; 302 303 fsid_t tfsid; 304 int mtype; 305 306 simple_lock(&mntid_slock); 307 mtype = mp->mnt_vfc->vfc_typenum; 308 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 309 mp->mnt_stat.f_fsid.val[1] = mtype; 310 if (xxxfs_mntid == 0) 311 ++xxxfs_mntid; 312 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 313 tfsid.val[1] = mtype; 314 if (!CIRCLEQ_EMPTY(&mountlist)) { 315 while (vfs_getvfs(&tfsid)) { 316 tfsid.val[0]++; 317 xxxfs_mntid++; 318 } 319 } 320 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 321 simple_unlock(&mntid_slock); 322} 323 324/* 325 * Make a 'unique' number from a mount type name. 326 * Note that this is no longer used for ffs which 327 * now has an on-disk filesystem id. 328 */ 329long 330makefstype(type) 331 char *type; 332{ 333 long rv; 334 335 for (rv = 0; *type; type++) { 336 rv <<= 2; 337 rv ^= *type; 338 } 339 return rv; 340} 341 342/* 343 * Set vnode attributes to VNOVAL 344 */ 345void 346vattr_null(vap) 347 register struct vattr *vap; 348{ 349 350 vap->va_type = VNON; 351 /* XXX These next two used to be one line, but for a GCC bug. */ 352 vap->va_size = VNOVAL; 353 vap->va_bytes = VNOVAL; 354 vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = 355 vap->va_fsid = vap->va_fileid = 356 vap->va_blocksize = vap->va_rdev = 357 vap->va_atime.tv_sec = vap->va_atime.tv_nsec = 358 vap->va_mtime.tv_sec = vap->va_mtime.tv_nsec = 359 vap->va_ctime.tv_sec = vap->va_ctime.tv_nsec = 360 vap->va_flags = vap->va_gen = VNOVAL; 361 vap->va_vaflags = 0; 362} 363 364/* 365 * Routines having to do with the management of the vnode table. 366 */ 367extern int (**dead_vnodeop_p) __P((void *)); 368long numvnodes; 369 370/* 371 * Return the next vnode from the free list. 372 */ 373int 374getnewvnode(tag, mp, vops, vpp) 375 enum vtagtype tag; 376 struct mount *mp; 377 int (**vops) __P((void *)); 378 struct vnode **vpp; 379{ 380 extern struct uvm_pagerops uvm_vnodeops; 381 struct uvm_object *uobj; 382 struct proc *p = curproc; /* XXX */ 383 struct freelst *listhd; 384 static int toggle; 385 struct vnode *vp; 386 int s; 387 388 /* 389 * We must choose whether to allocate a new vnode or recycle an 390 * existing one. The criterion for allocating a new one is that 391 * the total number of vnodes is less than the number desired or 392 * there are no vnodes on either free list. Generally we only 393 * want to recycle vnodes that have no buffers associated with 394 * them, so we look first on the vnode_free_list. If it is empty, 395 * we next consider vnodes with referencing buffers on the 396 * vnode_hold_list. The toggle ensures that half the time we 397 * will use a buffer from the vnode_hold_list, and half the time 398 * we will allocate a new one unless the list has grown to twice 399 * the desired size. We are reticent to recycle vnodes from the 400 * vnode_hold_list because we will lose the identity of all its 401 * referencing buffers. 402 */ 403 toggle ^= 1; 404 if (numvnodes > 2 * desiredvnodes) 405 toggle = 0; 406 407 simple_lock(&vnode_free_list_slock); 408 s = splbio(); 409 if ((numvnodes < desiredvnodes) || 410 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 411 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 412 splx(s); 413 simple_unlock(&vnode_free_list_slock); 414 vp = pool_get(&vnode_pool, PR_WAITOK); 415 bzero(vp, sizeof *vp); 416 /* 417 * initialize uvm_object within vnode. 418 */ 419 uobj = &vp->v_uobj; 420 uobj->pgops = &uvm_vnodeops; 421 uobj->uo_npages = 0; 422 TAILQ_INIT(&uobj->memq); 423 numvnodes++; 424 } else { 425 TAILQ_FOREACH(vp, listhd, v_freelist) { 426 if (simple_lock_try(&vp->v_interlock)) 427 break; 428 } 429 /* 430 * Unless this is a bad time of the month, at most 431 * the first NCPUS items on the free list are 432 * locked, so this is close enough to being empty. 433 */ 434 if (vp == NULL) { 435 splx(s); 436 simple_unlock(&vnode_free_list_slock); 437 tablefull("vnode"); 438 *vpp = 0; 439 return (ENFILE); 440 } 441 if (vp->v_usecount) { 442 vprint("free vnode", vp); 443 panic("free vnode isn't"); 444 } 445 446 TAILQ_REMOVE(listhd, vp, v_freelist); 447 vp->v_bioflag &= ~VBIOONFREELIST; 448 splx(s); 449 450 simple_unlock(&vnode_free_list_slock); 451 if (vp->v_type != VBAD) 452 vgonel(vp, p); 453 else 454 simple_unlock(&vp->v_interlock); 455#ifdef DIAGNOSTIC 456 if (vp->v_data || vp->v_uobj.uo_npages || 457 TAILQ_FIRST(&vp->v_uobj.memq)) { 458 vprint("cleaned vnode", vp); 459 panic("cleaned vnode isn't"); 460 } 461 if (vp->v_numoutput) 462 panic("Clean vnode has pending I/O's"); 463#endif 464 vp->v_flag = 0; 465 vp->v_socket = 0; 466 } 467 vp->v_type = VNON; 468 cache_purge(vp); 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_uobj.vmobjlock); 476 477 vp->v_size = VSIZENOTSET; 478 479 return (0); 480} 481 482/* 483 * Move a vnode from one mount queue to another. 484 */ 485void 486insmntque(vp, mp) 487 register struct vnode *vp; 488 register struct mount *mp; 489{ 490 simple_lock(&mntvnode_slock); 491 /* 492 * Delete from old mount point vnode list, if on one. 493 */ 494 495 if (vp->v_mount != NULL) 496 LIST_REMOVE(vp, v_mntvnodes); 497 /* 498 * Insert into list of vnodes for the new mount point, if available. 499 */ 500 if ((vp->v_mount = mp) != NULL) 501 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 502 simple_unlock(&mntvnode_slock); 503} 504 505 506/* 507 * Create a vnode for a block device. 508 * Used for root filesystem, argdev, and swap areas. 509 * Also used for memory file system special devices. 510 */ 511int 512bdevvp(dev, vpp) 513 dev_t dev; 514 struct vnode **vpp; 515{ 516 517 return (getdevvp(dev, vpp, VBLK)); 518} 519 520/* 521 * Create a vnode for a character device. 522 * Used for kernfs and some console handling. 523 */ 524int 525cdevvp(dev, vpp) 526 dev_t dev; 527 struct vnode **vpp; 528{ 529 530 return (getdevvp(dev, vpp, VCHR)); 531} 532 533/* 534 * Create a vnode for a device. 535 * Used by bdevvp (block device) for root file system etc., 536 * and by cdevvp (character device) for console and kernfs. 537 */ 538int 539getdevvp(dev, vpp, type) 540 dev_t dev; 541 struct vnode **vpp; 542 enum vtype type; 543{ 544 register struct vnode *vp; 545 struct vnode *nvp; 546 int error; 547 548 if (dev == NODEV) { 549 *vpp = NULLVP; 550 return (0); 551 } 552 error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); 553 if (error) { 554 *vpp = NULLVP; 555 return (error); 556 } 557 vp = nvp; 558 vp->v_type = type; 559 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 560 vput(vp); 561 vp = nvp; 562 } 563 *vpp = vp; 564 return (0); 565} 566 567/* 568 * Check to see if the new vnode represents a special device 569 * for which we already have a vnode (either because of 570 * bdevvp() or because of a different vnode representing 571 * the same block device). If such an alias exists, deallocate 572 * the existing contents and return the aliased vnode. The 573 * caller is responsible for filling it with its new contents. 574 */ 575struct vnode * 576checkalias(nvp, nvp_rdev, mp) 577 register struct vnode *nvp; 578 dev_t nvp_rdev; 579 struct mount *mp; 580{ 581 struct proc *p = curproc; 582 register struct vnode *vp; 583 struct vnode **vpp; 584 585 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 586 return (NULLVP); 587 588 vpp = &speclisth[SPECHASH(nvp_rdev)]; 589loop: 590 simple_lock(&spechash_slock); 591 for (vp = *vpp; vp; vp = vp->v_specnext) { 592 simple_lock(&vp->v_interlock); 593 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 594 simple_unlock(&vp->v_interlock); 595 continue; 596 } 597 /* 598 * Alias, but not in use, so flush it out. 599 */ 600 if (vp->v_usecount == 0) { 601 simple_unlock(&spechash_slock); 602 vgonel(vp, p); 603 goto loop; 604 } 605 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 606 simple_unlock(&spechash_slock); 607 goto loop; 608 } 609 break; 610 } 611 612 /* 613 * Common case is actually in the if statement 614 */ 615 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 616 MALLOC(nvp->v_specinfo, struct specinfo *, 617 sizeof(struct specinfo), M_VNODE, M_WAITOK); 618 nvp->v_rdev = nvp_rdev; 619 nvp->v_hashchain = vpp; 620 nvp->v_specnext = *vpp; 621 nvp->v_specmountpoint = NULL; 622 nvp->v_speclockf = NULL; 623 simple_unlock(&spechash_slock); 624 *vpp = nvp; 625 if (vp != NULLVP) { 626 nvp->v_flag |= VALIASED; 627 vp->v_flag |= VALIASED; 628 vput(vp); 629 } 630 return (NULLVP); 631 } 632 633 /* 634 * This code is the uncommon case. It is called in case 635 * we found an alias that was VT_NON && vtype of VBLK 636 * This means we found a block device that was created 637 * using bdevvp. 638 * An example of such a vnode is the root partition device vnode 639 * created in ffs_mountroot. 640 * 641 * The vnodes created by bdevvp should not be aliased (why?). 642 */ 643 644 simple_unlock(&spechash_slock); 645 VOP_UNLOCK(vp, 0, p); 646 simple_lock(&vp->v_interlock); 647 vclean(vp, 0, p); 648 vp->v_op = nvp->v_op; 649 vp->v_tag = nvp->v_tag; 650 nvp->v_type = VNON; 651 insmntque(vp, mp); 652 return (vp); 653} 654 655/* 656 * Grab a particular vnode from the free list, increment its 657 * reference count and lock it. The vnode lock bit is set the 658 * vnode is being eliminated in vgone. The process is awakened 659 * when the transition is completed, and an error returned to 660 * indicate that the vnode is no longer usable (possibly having 661 * been changed to a new file system type). 662 */ 663int 664vget(vp, flags, p) 665 struct vnode *vp; 666 int flags; 667 struct proc *p; 668{ 669 int error; 670 int s; 671 /* 672 * If the vnode is in the process of being cleaned out for 673 * another use, we wait for the cleaning to finish and then 674 * return failure. Cleaning is determined by checking that 675 * the VXLOCK flag is set. 676 */ 677 if ((flags & LK_INTERLOCK) == 0) { 678 simple_lock(&vp->v_interlock); 679 flags |= LK_INTERLOCK; 680 } 681 if (vp->v_flag & VXLOCK) { 682 if (flags & LK_NOWAIT) { 683 simple_unlock(&vp->v_interlock); 684 return (EBUSY); 685 } 686 vp->v_flag |= VXWANT; 687 simple_unlock(&vp->v_interlock); 688 tsleep((caddr_t)vp, PINOD, "vget", 0); 689 return (ENOENT); 690 } 691 if (vp->v_usecount == 0 && 692 (vp->v_bioflag & VBIOONFREELIST)) { 693 s = splbio(); 694 simple_lock(&vnode_free_list_slock); 695 if (vp->v_holdcnt > 0) 696 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 697 else 698 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 699 simple_unlock(&vnode_free_list_slock); 700 vp->v_bioflag &= ~VBIOONFREELIST; 701 splx(s); 702 } 703 vp->v_usecount++; 704 if (flags & LK_TYPE_MASK) { 705 if ((error = vn_lock(vp, flags, p)) != 0) { 706 vp->v_usecount--; 707 if (vp->v_usecount == 0) 708 vputonfreelist(vp); 709 710 simple_unlock(&vp->v_interlock); 711 } 712 return (error); 713 } 714 simple_unlock(&vp->v_interlock); 715 return (0); 716} 717 718 719#ifdef DIAGNOSTIC 720/* 721 * Vnode reference. 722 */ 723void 724vref(vp) 725 struct vnode *vp; 726{ 727 simple_lock(&vp->v_interlock); 728 if (vp->v_usecount == 0) 729 panic("vref used where vget required"); 730 vp->v_usecount++; 731 simple_unlock(&vp->v_interlock); 732} 733#endif /* DIAGNOSTIC */ 734 735static __inline__ void 736vputonfreelist(vp) 737 struct vnode *vp; 738{ 739 int s; 740 struct freelst *lst; 741 742 s = splbio(); 743#ifdef DIAGNOSTIC 744 if (vp->v_usecount != 0) 745 panic("Use count is not zero!"); 746 747 if (vp->v_bioflag & VBIOONFREELIST) { 748 vprint("vnode already on free list: ", vp); 749 panic("vnode already on free list"); 750 } 751#endif 752 753 vp->v_bioflag |= VBIOONFREELIST; 754 755 if (vp->v_holdcnt > 0) 756 lst = &vnode_hold_list; 757 else 758 lst = &vnode_free_list; 759 760 if (vp->v_type == VBAD) 761 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 762 else 763 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 764 765 splx(s); 766} 767 768/* 769 * vput(), just unlock and vrele() 770 */ 771void 772vput(vp) 773 register struct vnode *vp; 774{ 775 struct proc *p = curproc; /* XXX */ 776 777#ifdef DIAGNOSTIC 778 if (vp == NULL) 779 panic("vput: null vp"); 780#endif 781 simple_lock(&vp->v_interlock); 782 783#ifdef DIAGNOSTIC 784 if (vp->v_usecount == 0) { 785 vprint("vput: bad ref count", vp); 786 panic("vput: ref cnt"); 787 } 788#endif 789 vp->v_usecount--; 790 if (vp->v_usecount > 0) { 791 simple_unlock(&vp->v_interlock); 792 VOP_UNLOCK(vp, 0, p); 793 return; 794 } 795 796#ifdef DIAGNOSTIC 797 if (vp->v_writecount != 0) { 798 vprint("vput: bad writecount", vp); 799 panic("vput: v_writecount != 0"); 800 } 801#endif 802 vputonfreelist(vp); 803 804 if (vp->v_flag & VTEXT) { 805 uvmexp.vtextpages -= vp->v_uobj.uo_npages; 806 uvmexp.vnodepages += vp->v_uobj.uo_npages; 807 } 808 vp->v_flag &= ~VTEXT; 809 simple_unlock(&vp->v_interlock); 810 811 VOP_INACTIVE(vp, p); 812} 813 814/* 815 * Vnode release - use for active VNODES. 816 * If count drops to zero, call inactive routine and return to freelist. 817 */ 818void 819vrele(vp) 820 register struct vnode *vp; 821{ 822 struct proc *p = curproc; /* XXX */ 823 824#ifdef DIAGNOSTIC 825 if (vp == NULL) 826 panic("vrele: null vp"); 827#endif 828 simple_lock(&vp->v_interlock); 829#ifdef DIAGNOSTIC 830 if (vp->v_usecount == 0) { 831 vprint("vrele: bad ref count", vp); 832 panic("vrele: ref cnt"); 833 } 834#endif 835 vp->v_usecount--; 836 if (vp->v_usecount > 0) { 837 simple_unlock(&vp->v_interlock); 838 return; 839 } 840 841#ifdef DIAGNOSTIC 842 if (vp->v_writecount != 0) { 843 vprint("vrele: bad writecount", vp); 844 panic("vrele: v_writecount != 0"); 845 } 846#endif 847 vputonfreelist(vp); 848 849 if (vp->v_flag & VTEXT) { 850 uvmexp.vtextpages -= vp->v_uobj.uo_npages; 851 uvmexp.vnodepages += vp->v_uobj.uo_npages; 852 } 853 vp->v_flag &= ~VTEXT; 854 if (vn_lock(vp, LK_EXCLUSIVE|LK_INTERLOCK, p) == 0) 855 VOP_INACTIVE(vp, p); 856} 857 858/* 859 * Page or buffer structure gets a reference. 860 * Must be called at splbio(); 861 */ 862void 863vhold(struct vnode *vp) 864{ 865 866 /* 867 * If it is on the freelist and the hold count is currently 868 * zero, move it to the hold list. 869 */ 870 simple_lock(&vp->v_interlock); 871 if ((vp->v_bioflag & VBIOONFREELIST) && 872 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 873 simple_lock(&vnode_free_list_slock); 874 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 875 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 876 simple_unlock(&vnode_free_list_slock); 877 } 878 vp->v_holdcnt++; 879 simple_unlock(&vp->v_interlock); 880} 881 882/* 883 * Release a vhold reference. 884 * Must be called at splbio(); 885 */ 886void 887vholdrele(struct vnode *vp) 888{ 889 simple_lock(&vp->v_interlock); 890#ifdef DIAGNOSTIC 891 if (vp->v_holdcnt == 0) 892 panic("vholdrele: holdcnt"); 893#endif 894 vp->v_holdcnt--; 895 896 /* 897 * If it is on the holdlist and the hold count drops to 898 * zero, move it to the free list. 899 */ 900 if ((vp->v_bioflag & VBIOONFREELIST) && 901 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 902 simple_lock(&vnode_free_list_slock); 903 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 904 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 905 simple_unlock(&vnode_free_list_slock); 906 } 907 simple_unlock(&vp->v_interlock); 908} 909 910/* 911 * Remove any vnodes in the vnode table belonging to mount point mp. 912 * 913 * If MNT_NOFORCE is specified, there should not be any active ones, 914 * return error if any are found (nb: this is a user error, not a 915 * system error). If MNT_FORCE is specified, detach any active vnodes 916 * that are found. 917 */ 918#ifdef DEBUG 919int busyprt = 0; /* print out busy vnodes */ 920struct ctldebug debug1 = { "busyprt", &busyprt }; 921#endif 922 923int 924vfs_mount_foreach_vnode(struct mount *mp, 925 int (*func)(struct vnode *, void *), void *arg) { 926 struct vnode *vp, *nvp; 927 int error = 0; 928 929 simple_lock(&mntvnode_slock); 930loop: 931 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { 932 if (vp->v_mount != mp) 933 goto loop; 934 nvp = vp->v_mntvnodes.le_next; 935 simple_lock(&vp->v_interlock); 936 simple_unlock(&mntvnode_slock); 937 938 error = func(vp, arg); 939 940 simple_lock(&mntvnode_slock); 941 942 if (error != 0) 943 break; 944 } 945 simple_unlock(&mntvnode_slock); 946 947 return (error); 948} 949 950 951struct vflush_args { 952 struct vnode *skipvp; 953 int busy; 954 int flags; 955}; 956 957int 958vflush_vnode(struct vnode *vp, void *arg) { 959 struct vflush_args *va = arg; 960 struct proc *p = curproc; 961 962 if (vp == va->skipvp) { 963 simple_unlock(&vp->v_interlock); 964 return (0); 965 } 966 967 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 968 simple_unlock(&vp->v_interlock); 969 return (0); 970 } 971 972 /* 973 * If WRITECLOSE is set, only flush out regular file 974 * vnodes open for writing. 975 */ 976 if ((va->flags & WRITECLOSE) && 977 (vp->v_writecount == 0 || vp->v_type != VREG)) { 978 simple_unlock(&vp->v_interlock); 979 return (0); 980 } 981 982 /* 983 * With v_usecount == 0, all we need to do is clear 984 * out the vnode data structures and we are done. 985 */ 986 if (vp->v_usecount == 0) { 987 vgonel(vp, p); 988 return (0); 989 } 990 991 /* 992 * If FORCECLOSE is set, forcibly close the vnode. 993 * For block or character devices, revert to an 994 * anonymous device. For all other files, just kill them. 995 */ 996 if (va->flags & FORCECLOSE) { 997 if (vp->v_type != VBLK && vp->v_type != VCHR) { 998 vgonel(vp, p); 999 } else { 1000 vclean(vp, 0, p); 1001 vp->v_op = spec_vnodeop_p; 1002 insmntque(vp, (struct mount *)0); 1003 } 1004 return (0); 1005 } 1006 1007#ifdef DEBUG 1008 if (busyprt) 1009 vprint("vflush: busy vnode", vp); 1010#endif 1011 simple_unlock(&vp->v_interlock); 1012 va->busy++; 1013 return (0); 1014} 1015 1016int 1017vflush(mp, skipvp, flags) 1018 struct mount *mp; 1019 struct vnode *skipvp; 1020 int flags; 1021{ 1022 struct vflush_args va; 1023 va.skipvp = skipvp; 1024 va.busy = 0; 1025 va.flags = flags; 1026 1027 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 1028 1029 if (va.busy) 1030 return (EBUSY); 1031 return (0); 1032} 1033 1034/* 1035 * Disassociate the underlying file system from a vnode. 1036 * The vnode interlock is held on entry. 1037 */ 1038void 1039vclean(vp, flags, p) 1040 register struct vnode *vp; 1041 int flags; 1042 struct proc *p; 1043{ 1044 int active; 1045 1046 /* 1047 * Check to see if the vnode is in use. 1048 * If so we have to reference it before we clean it out 1049 * so that its count cannot fall to zero and generate a 1050 * race against ourselves to recycle it. 1051 */ 1052 if ((active = vp->v_usecount) != 0) 1053 vp->v_usecount++; 1054 1055 /* 1056 * Prevent the vnode from being recycled or 1057 * brought into use while we clean it out. 1058 */ 1059 if (vp->v_flag & VXLOCK) 1060 panic("vclean: deadlock"); 1061 vp->v_flag |= VXLOCK; 1062 if (vp->v_flag & VTEXT) { 1063 uvmexp.vtextpages -= vp->v_uobj.uo_npages; 1064 uvmexp.vnodepages += vp->v_uobj.uo_npages; 1065 } 1066 vp->v_flag &= ~VTEXT; 1067 1068 /* 1069 * Even if the count is zero, the VOP_INACTIVE routine may still 1070 * have the object locked while it cleans it out. The VOP_LOCK 1071 * ensures that the VOP_INACTIVE routine is done with its work. 1072 * For active vnodes, it ensures that no other activity can 1073 * occur while the underlying object is being cleaned out. 1074 */ 1075 VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK, p); 1076 1077 /* 1078 * Clean out any cached data associated with the vnode. 1079 */ 1080 if (flags & DOCLOSE) 1081 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 1082 /* 1083 * If purging an active vnode, it must be closed and 1084 * deactivated before being reclaimed. Note that the 1085 * VOP_INACTIVE will unlock the vnode 1086 */ 1087 if (active) { 1088 if (flags & DOCLOSE) 1089 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 1090 VOP_INACTIVE(vp, p); 1091 } else { 1092 /* 1093 * Any other processes trying to obtain this lock must first 1094 * wait for VXLOCK to clear, then call the new lock operation. 1095 */ 1096 VOP_UNLOCK(vp, 0, p); 1097 } 1098 1099 /* 1100 * Reclaim the vnode. 1101 */ 1102 if (VOP_RECLAIM(vp, p)) 1103 panic("vclean: cannot reclaim"); 1104 if (active) { 1105 simple_lock(&vp->v_interlock); 1106 1107 vp->v_usecount--; 1108 if (vp->v_usecount == 0) { 1109 if (vp->v_holdcnt > 0) 1110 panic("vclean: not clean"); 1111 vputonfreelist(vp); 1112 } 1113 1114 simple_unlock(&vp->v_interlock); 1115 } 1116 cache_purge(vp); 1117 if (vp->v_vnlock) { 1118 if ((vp->v_vnlock->lk_flags & LK_DRAINED) == 0) 1119 vprint("vclean: lock not drained", vp); 1120 FREE(vp->v_vnlock, M_VNODE); 1121 vp->v_vnlock = NULL; 1122 } 1123 1124 /* 1125 * Done with purge, notify sleepers of the grim news. 1126 */ 1127 vp->v_op = dead_vnodeop_p; 1128 simple_lock(&vp->v_selectinfo.vsi_lock); 1129 VN_KNOTE(vp, NOTE_REVOKE); 1130 simple_unlock(&vp->v_selectinfo.vsi_lock); 1131 vp->v_tag = VT_NON; 1132 vp->v_flag &= ~VXLOCK; 1133#ifdef DIAGNOSTIC 1134 vp->v_flag &= ~VLOCKSWORK; 1135#endif 1136 if (vp->v_flag & VXWANT) { 1137 vp->v_flag &= ~VXWANT; 1138 wakeup((caddr_t)vp); 1139 } 1140} 1141 1142 1143 1144/* 1145 * Recycle an unused vnode to the front of the free list. 1146 * Release the passed interlock if the vnode will be recycled. 1147 */ 1148int 1149vrecycle(vp, inter_lkp, p) 1150 struct vnode *vp; 1151 struct simplelock *inter_lkp; 1152 struct proc *p; 1153{ 1154 1155 simple_lock(&vp->v_interlock); 1156 if (vp->v_usecount == 0) { 1157 if (inter_lkp) 1158 simple_unlock(inter_lkp); 1159 vgonel(vp, p); 1160 return (1); 1161 } 1162 simple_unlock(&vp->v_interlock); 1163 return (0); 1164} 1165 1166 1167/* 1168 * Eliminate all activity associated with a vnode 1169 * in preparation for reuse. 1170 */ 1171void 1172vgone(vp) 1173 register struct vnode *vp; 1174{ 1175 struct proc *p = curproc; 1176 1177 simple_lock (&vp->v_interlock); 1178 vgonel(vp, p); 1179} 1180 1181/* 1182 * vgone, with the vp interlock held. 1183 */ 1184void 1185vgonel(vp, p) 1186 struct vnode *vp; 1187 struct proc *p; 1188{ 1189 register struct vnode *vq; 1190 struct vnode *vx; 1191 1192 /* 1193 * If a vgone (or vclean) is already in progress, 1194 * wait until it is done and return. 1195 */ 1196 if (vp->v_flag & VXLOCK) { 1197 vp->v_flag |= VXWANT; 1198 simple_unlock(&vp->v_interlock); 1199 tsleep((caddr_t)vp, PINOD, "vgone", 0); 1200 return; 1201 } 1202 /* 1203 * Clean out the filesystem specific data. 1204 */ 1205 vclean(vp, DOCLOSE, p); 1206 /* 1207 * Delete from old mount point vnode list, if on one. 1208 */ 1209 if (vp->v_mount != NULL) 1210 insmntque(vp, (struct mount *)0); 1211 /* 1212 * If special device, remove it from special device alias list 1213 * if it is on one. 1214 */ 1215 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1216 simple_lock(&spechash_slock); 1217 if (*vp->v_hashchain == vp) { 1218 *vp->v_hashchain = vp->v_specnext; 1219 } else { 1220 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1221 if (vq->v_specnext != vp) 1222 continue; 1223 vq->v_specnext = vp->v_specnext; 1224 break; 1225 } 1226 if (vq == NULL) 1227 panic("missing bdev"); 1228 } 1229 if (vp->v_flag & VALIASED) { 1230 vx = NULL; 1231 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1232 if (vq->v_rdev != vp->v_rdev || 1233 vq->v_type != vp->v_type) 1234 continue; 1235 if (vx) 1236 break; 1237 vx = vq; 1238 } 1239 if (vx == NULL) 1240 panic("missing alias"); 1241 if (vq == NULL) 1242 vx->v_flag &= ~VALIASED; 1243 vp->v_flag &= ~VALIASED; 1244 } 1245 simple_unlock(&spechash_slock); 1246 FREE(vp->v_specinfo, M_VNODE); 1247 vp->v_specinfo = NULL; 1248 } 1249 /* 1250 * If it is on the freelist and not already at the head, 1251 * move it to the head of the list. 1252 */ 1253 vp->v_type = VBAD; 1254 1255 /* 1256 * Move onto the free list, unless we were called from 1257 * getnewvnode and we're not on any free list 1258 */ 1259 if (vp->v_usecount == 0 && 1260 (vp->v_bioflag & VBIOONFREELIST)) { 1261 int s; 1262 1263 simple_lock(&vnode_free_list_slock); 1264 s = splbio(); 1265 1266 if (vp->v_holdcnt > 0) 1267 panic("vgonel: not clean"); 1268 1269 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1270 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1271 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1272 } 1273 splx(s); 1274 simple_unlock(&vnode_free_list_slock); 1275 } 1276} 1277 1278/* 1279 * Lookup a vnode by device number. 1280 */ 1281int 1282vfinddev(dev, type, vpp) 1283 dev_t dev; 1284 enum vtype type; 1285 struct vnode **vpp; 1286{ 1287 register struct vnode *vp; 1288 int rc =0; 1289 1290 simple_lock(&spechash_slock); 1291 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1292 if (dev != vp->v_rdev || type != vp->v_type) 1293 continue; 1294 *vpp = vp; 1295 rc = 1; 1296 break; 1297 } 1298 simple_unlock(&spechash_slock); 1299 return (rc); 1300} 1301 1302/* 1303 * Revoke all the vnodes corresponding to the specified minor number 1304 * range (endpoints inclusive) of the specified major. 1305 */ 1306void 1307vdevgone(maj, minl, minh, type) 1308 int maj, minl, minh; 1309 enum vtype type; 1310{ 1311 struct vnode *vp; 1312 int mn; 1313 1314 for (mn = minl; mn <= minh; mn++) 1315 if (vfinddev(makedev(maj, mn), type, &vp)) 1316 VOP_REVOKE(vp, REVOKEALL); 1317} 1318 1319/* 1320 * Calculate the total number of references to a special device. 1321 */ 1322int 1323vcount(vp) 1324 struct vnode *vp; 1325{ 1326 struct vnode *vq, *vnext; 1327 int count; 1328 1329loop: 1330 if ((vp->v_flag & VALIASED) == 0) 1331 return (vp->v_usecount); 1332 simple_lock(&spechash_slock); 1333 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1334 vnext = vq->v_specnext; 1335 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1336 continue; 1337 /* 1338 * Alias, but not in use, so flush it out. 1339 */ 1340 if (vq->v_usecount == 0 && vq != vp) { 1341 simple_unlock(&spechash_slock); 1342 vgone(vq); 1343 goto loop; 1344 } 1345 count += vq->v_usecount; 1346 } 1347 simple_unlock(&spechash_slock); 1348 return (count); 1349} 1350 1351/* 1352 * Print out a description of a vnode. 1353 */ 1354static char *typename[] = 1355 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1356 1357void 1358vprint(label, vp) 1359 char *label; 1360 register struct vnode *vp; 1361{ 1362 char buf[64]; 1363 1364 if (label != NULL) 1365 printf("%s: ", label); 1366 printf("type %s, usecount %u, writecount %u, holdcount %u,", 1367 typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1368 vp->v_holdcnt); 1369 buf[0] = '\0'; 1370 if (vp->v_flag & VROOT) 1371 strcat(buf, "|VROOT"); 1372 if (vp->v_flag & VTEXT) 1373 strcat(buf, "|VTEXT"); 1374 if (vp->v_flag & VSYSTEM) 1375 strcat(buf, "|VSYSTEM"); 1376 if (vp->v_flag & VXLOCK) 1377 strcat(buf, "|VXLOCK"); 1378 if (vp->v_flag & VXWANT) 1379 strcat(buf, "|VXWANT"); 1380 if (vp->v_bioflag & VBIOWAIT) 1381 strcat(buf, "| VBIOWAIT"); 1382 if (vp->v_flag & VALIASED) 1383 strcat(buf, "|VALIASED"); 1384 if (buf[0] != '\0') 1385 printf(" flags (%s)", &buf[1]); 1386 if (vp->v_data == NULL) { 1387 printf("\n"); 1388 } else { 1389 printf("\n\t"); 1390 VOP_PRINT(vp); 1391 } 1392} 1393 1394#ifdef DEBUG 1395/* 1396 * List all of the locked vnodes in the system. 1397 * Called when debugging the kernel. 1398 */ 1399void 1400printlockedvnodes() 1401{ 1402 struct proc *p = curproc; 1403 register struct mount *mp, *nmp; 1404 register struct vnode *vp; 1405 1406 printf("Locked vnodes\n"); 1407 simple_lock(&mountlist_slock); 1408 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1409 mp = nmp) { 1410 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { 1411 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1412 continue; 1413 } 1414 for (vp = mp->mnt_vnodelist.lh_first; vp; 1415 vp = vp->v_mntvnodes.le_next) { 1416 if (VOP_ISLOCKED(vp)) 1417 vprint((char *)0, vp); 1418 } 1419 simple_lock(&mountlist_slock); 1420 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1421 vfs_unbusy(mp, p); 1422 } 1423 simple_unlock(&mountlist_slock); 1424 1425} 1426#endif 1427 1428/* 1429 * Top level filesystem related information gathering. 1430 */ 1431int 1432vfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1433 int *name; 1434 u_int namelen; 1435 void *oldp; 1436 size_t *oldlenp; 1437 void *newp; 1438 size_t newlen; 1439 struct proc *p; 1440{ 1441 struct vfsconf *vfsp; 1442 1443 /* all sysctl names at this level are at least name and field */ 1444 if (namelen < 2) 1445 return (ENOTDIR); /* overloaded */ 1446 if (name[0] != VFS_GENERIC) { 1447 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1448 if (vfsp->vfc_typenum == name[0]) 1449 break; 1450 if (vfsp == NULL) 1451 return (EOPNOTSUPP); 1452 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1453 oldp, oldlenp, newp, newlen, p)); 1454 } 1455 switch (name[1]) { 1456 case VFS_MAXTYPENUM: 1457 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1458 case VFS_CONF: 1459 if (namelen < 3) 1460 return (ENOTDIR); /* overloaded */ 1461 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1462 if (vfsp->vfc_typenum == name[2]) 1463 break; 1464 if (vfsp == NULL) 1465 return (EOPNOTSUPP); 1466 return (sysctl_rdstruct(oldp, oldlenp, newp, vfsp, 1467 sizeof(struct vfsconf))); 1468 } 1469 return (EOPNOTSUPP); 1470} 1471 1472 1473int kinfo_vdebug = 1; 1474int kinfo_vgetfailed; 1475#define KINFO_VNODESLOP 10 1476/* 1477 * Dump vnode list (via sysctl). 1478 * Copyout address of vnode followed by vnode. 1479 */ 1480/* ARGSUSED */ 1481int 1482sysctl_vnode(where, sizep, p) 1483 char *where; 1484 size_t *sizep; 1485 struct proc *p; 1486{ 1487 register struct mount *mp, *nmp; 1488 struct vnode *vp, *nvp; 1489 register char *bp = where, *savebp; 1490 char *ewhere; 1491 int error; 1492 1493#define VPTRSZ sizeof (struct vnode *) 1494#define VNODESZ sizeof (struct vnode) 1495 if (where == NULL) { 1496 *sizep = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ); 1497 return (0); 1498 } 1499 ewhere = where + *sizep; 1500 1501 simple_lock(&mountlist_slock); 1502 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1503 mp = nmp) { 1504 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { 1505 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1506 continue; 1507 } 1508 savebp = bp; 1509again: 1510 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; 1511 vp = nvp) { 1512 /* 1513 * Check that the vp is still associated with 1514 * this filesystem. RACE: could have been 1515 * recycled onto the same filesystem. 1516 */ 1517 if (vp->v_mount != mp) { 1518 simple_unlock(&mntvnode_slock); 1519 if (kinfo_vdebug) 1520 printf("kinfo: vp changed\n"); 1521 bp = savebp; 1522 goto again; 1523 } 1524 nvp = vp->v_mntvnodes.le_next; 1525 if (bp + VPTRSZ + VNODESZ > ewhere) { 1526 simple_unlock(&mntvnode_slock); 1527 *sizep = bp - where; 1528 return (ENOMEM); 1529 } 1530 if ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) || 1531 (error = copyout((caddr_t)vp, bp + VPTRSZ, VNODESZ))) 1532 return (error); 1533 bp += VPTRSZ + VNODESZ; 1534 simple_lock(&mntvnode_slock); 1535 } 1536 1537 simple_unlock(&mntvnode_slock); 1538 simple_lock(&mountlist_slock); 1539 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1540 vfs_unbusy(mp, p); 1541 } 1542 1543 simple_unlock(&mountlist_slock); 1544 1545 *sizep = bp - where; 1546 return (0); 1547} 1548 1549/* 1550 * Check to see if a filesystem is mounted on a block device. 1551 */ 1552int 1553vfs_mountedon(vp) 1554 register struct vnode *vp; 1555{ 1556 register struct vnode *vq; 1557 int error = 0; 1558 1559 if (vp->v_specmountpoint != NULL) 1560 return (EBUSY); 1561 if (vp->v_flag & VALIASED) { 1562 simple_lock(&spechash_slock); 1563 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1564 if (vq->v_rdev != vp->v_rdev || 1565 vq->v_type != vp->v_type) 1566 continue; 1567 if (vq->v_specmountpoint != NULL) { 1568 error = EBUSY; 1569 break; 1570 } 1571 } 1572 simple_unlock(&spechash_slock); 1573 } 1574 return (error); 1575} 1576 1577/* 1578 * Build hash lists of net addresses and hang them off the mount point. 1579 * Called by ufs_mount() to set up the lists of export addresses. 1580 */ 1581int 1582vfs_hang_addrlist(mp, nep, argp) 1583 struct mount *mp; 1584 struct netexport *nep; 1585 struct export_args *argp; 1586{ 1587 register struct netcred *np; 1588 register struct radix_node_head *rnh; 1589 register int i; 1590 struct radix_node *rn; 1591 struct sockaddr *saddr, *smask = 0; 1592 struct domain *dom; 1593 int error; 1594 1595 if (argp->ex_addrlen == 0) { 1596 if (mp->mnt_flag & MNT_DEFEXPORTED) 1597 return (EPERM); 1598 np = &nep->ne_defexported; 1599 np->netc_exflags = argp->ex_flags; 1600 np->netc_anon = argp->ex_anon; 1601 np->netc_anon.cr_ref = 1; 1602 mp->mnt_flag |= MNT_DEFEXPORTED; 1603 return (0); 1604 } 1605 if (argp->ex_addrlen > MLEN) 1606 return (EINVAL); 1607 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1608 np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK); 1609 bzero((caddr_t)np, i); 1610 saddr = (struct sockaddr *)(np + 1); 1611 error = copyin(argp->ex_addr, (caddr_t)saddr, argp->ex_addrlen); 1612 if (error) 1613 goto out; 1614 if (saddr->sa_len > argp->ex_addrlen) 1615 saddr->sa_len = argp->ex_addrlen; 1616 if (argp->ex_masklen) { 1617 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1618 error = copyin(argp->ex_mask, (caddr_t)smask, argp->ex_masklen); 1619 if (error) 1620 goto out; 1621 if (smask->sa_len > argp->ex_masklen) 1622 smask->sa_len = argp->ex_masklen; 1623 } 1624 i = saddr->sa_family; 1625 if (i < 0 || i > AF_MAX) { 1626 error = EINVAL; 1627 goto out; 1628 } 1629 if ((rnh = nep->ne_rtable[i]) == 0) { 1630 /* 1631 * Seems silly to initialize every AF when most are not 1632 * used, do so on demand here 1633 */ 1634 for (dom = domains; dom; dom = dom->dom_next) 1635 if (dom->dom_family == i && dom->dom_rtattach) { 1636 dom->dom_rtattach((void **)&nep->ne_rtable[i], 1637 dom->dom_rtoffset); 1638 break; 1639 } 1640 if ((rnh = nep->ne_rtable[i]) == 0) { 1641 error = ENOBUFS; 1642 goto out; 1643 } 1644 } 1645 rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, 1646 np->netc_rnodes); 1647 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1648 error = EPERM; 1649 goto out; 1650 } 1651 np->netc_exflags = argp->ex_flags; 1652 np->netc_anon = argp->ex_anon; 1653 np->netc_anon.cr_ref = 1; 1654 return (0); 1655out: 1656 free(np, M_NETADDR); 1657 return (error); 1658} 1659 1660/* ARGSUSED */ 1661int 1662vfs_free_netcred(rn, w) 1663 struct radix_node *rn; 1664 void *w; 1665{ 1666 register struct radix_node_head *rnh = (struct radix_node_head *)w; 1667 1668 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh); 1669 free((caddr_t)rn, M_NETADDR); 1670 return (0); 1671} 1672 1673/* 1674 * Free the net address hash lists that are hanging off the mount points. 1675 */ 1676void 1677vfs_free_addrlist(nep) 1678 struct netexport *nep; 1679{ 1680 register int i; 1681 register struct radix_node_head *rnh; 1682 1683 for (i = 0; i <= AF_MAX; i++) 1684 if ((rnh = nep->ne_rtable[i]) != NULL) { 1685 (*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh); 1686 free((caddr_t)rnh, M_RTABLE); 1687 nep->ne_rtable[i] = 0; 1688 } 1689} 1690 1691int 1692vfs_export(mp, nep, argp) 1693 struct mount *mp; 1694 struct netexport *nep; 1695 struct export_args *argp; 1696{ 1697 int error; 1698 1699 if (argp->ex_flags & MNT_DELEXPORT) { 1700 vfs_free_addrlist(nep); 1701 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1702 } 1703 if (argp->ex_flags & MNT_EXPORTED) { 1704 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1705 return (error); 1706 mp->mnt_flag |= MNT_EXPORTED; 1707 } 1708 return (0); 1709} 1710 1711struct netcred * 1712vfs_export_lookup(mp, nep, nam) 1713 register struct mount *mp; 1714 struct netexport *nep; 1715 struct mbuf *nam; 1716{ 1717 register struct netcred *np; 1718 register struct radix_node_head *rnh; 1719 struct sockaddr *saddr; 1720 1721 np = NULL; 1722 if (mp->mnt_flag & MNT_EXPORTED) { 1723 /* 1724 * Lookup in the export list first. 1725 */ 1726 if (nam != NULL) { 1727 saddr = mtod(nam, struct sockaddr *); 1728 rnh = nep->ne_rtable[saddr->sa_family]; 1729 if (rnh != NULL) { 1730 np = (struct netcred *) 1731 (*rnh->rnh_matchaddr)((caddr_t)saddr, 1732 rnh); 1733 if (np && np->netc_rnodes->rn_flags & RNF_ROOT) 1734 np = NULL; 1735 } 1736 } 1737 /* 1738 * If no address match, use the default if it exists. 1739 */ 1740 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1741 np = &nep->ne_defexported; 1742 } 1743 return (np); 1744} 1745 1746/* 1747 * Do the usual access checking. 1748 * file_mode, uid and gid are from the vnode in question, 1749 * while acc_mode and cred are from the VOP_ACCESS parameter list 1750 */ 1751int 1752vaccess(file_mode, uid, gid, acc_mode, cred) 1753 mode_t file_mode; 1754 uid_t uid; 1755 gid_t gid; 1756 mode_t acc_mode; 1757 struct ucred *cred; 1758{ 1759 mode_t mask; 1760 1761 /* User id 0 always gets access. */ 1762 if (cred->cr_uid == 0) 1763 return 0; 1764 1765 mask = 0; 1766 1767 /* Otherwise, check the owner. */ 1768 if (cred->cr_uid == uid) { 1769 if (acc_mode & VEXEC) 1770 mask |= S_IXUSR; 1771 if (acc_mode & VREAD) 1772 mask |= S_IRUSR; 1773 if (acc_mode & VWRITE) 1774 mask |= S_IWUSR; 1775 return (file_mode & mask) == mask ? 0 : EACCES; 1776 } 1777 1778 /* Otherwise, check the groups. */ 1779 if (cred->cr_gid == gid || groupmember(gid, cred)) { 1780 if (acc_mode & VEXEC) 1781 mask |= S_IXGRP; 1782 if (acc_mode & VREAD) 1783 mask |= S_IRGRP; 1784 if (acc_mode & VWRITE) 1785 mask |= S_IWGRP; 1786 return (file_mode & mask) == mask ? 0 : EACCES; 1787 } 1788 1789 /* Otherwise, check everyone else. */ 1790 if (acc_mode & VEXEC) 1791 mask |= S_IXOTH; 1792 if (acc_mode & VREAD) 1793 mask |= S_IROTH; 1794 if (acc_mode & VWRITE) 1795 mask |= S_IWOTH; 1796 return (file_mode & mask) == mask ? 0 : EACCES; 1797} 1798 1799/* 1800 * Unmount all file systems. 1801 * We traverse the list in reverse order under the assumption that doing so 1802 * will avoid needing to worry about dependencies. 1803 */ 1804void 1805vfs_unmountall() 1806{ 1807 register struct mount *mp, *nmp; 1808 int allerror, error, again = 1; 1809 1810 retry: 1811 allerror = 0; 1812 for (mp = CIRCLEQ_LAST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1813 mp = nmp) { 1814 nmp = CIRCLEQ_PREV(mp, mnt_list); 1815 if ((error = dounmount(mp, MNT_FORCE, curproc)) != 0) { 1816 printf("unmount of %s failed with error %d\n", 1817 mp->mnt_stat.f_mntonname, error); 1818 allerror = 1; 1819 } 1820 } 1821 1822 if (allerror) { 1823 printf("WARNING: some file systems would not unmount\n"); 1824 if (again) { 1825 printf("retrying\n"); 1826 again = 0; 1827 goto retry; 1828 } 1829 } 1830} 1831 1832/* 1833 * Sync and unmount file systems before shutting down. 1834 */ 1835void 1836vfs_shutdown() 1837{ 1838 /* XXX Should suspend scheduling. */ 1839 (void) spl0(); 1840 1841 printf("syncing disks... "); 1842 1843 if (panicstr == 0) { 1844 /* Sync before unmount, in case we hang on something. */ 1845 sys_sync(&proc0, (void *)0, (register_t *)0); 1846 1847 /* Unmount file systems. */ 1848 vfs_unmountall(); 1849 } 1850 1851 if (vfs_syncwait(1)) 1852 printf("giving up\n"); 1853 else 1854 printf("done\n"); 1855} 1856 1857/* 1858 * perform sync() operation and wait for buffers to flush. 1859 * assumtions: called w/ scheduler disabled and physical io enabled 1860 * for now called at spl0() XXX 1861 */ 1862int 1863vfs_syncwait(verbose) 1864 int verbose; 1865{ 1866 register struct buf *bp; 1867 int iter, nbusy, dcount, s; 1868 struct proc *p; 1869 1870 p = curproc? curproc : &proc0; 1871 sys_sync(p, (void *)0, (register_t *)0); 1872 1873 /* Wait for sync to finish. */ 1874 dcount = 10000; 1875 for (iter = 0; iter < 20; iter++) { 1876 nbusy = 0; 1877 for (bp = &buf[nbuf]; --bp >= buf; ) { 1878 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1879 nbusy++; 1880 /* 1881 * With soft updates, some buffers that are 1882 * written will be remarked as dirty until other 1883 * buffers are written. 1884 */ 1885 if (bp->b_flags & B_DELWRI) { 1886 s = splbio(); 1887 bremfree(bp); 1888 bp->b_flags |= B_BUSY; 1889 splx(s); 1890 nbusy++; 1891 bawrite(bp); 1892 if (dcount-- <= 0) { 1893 if (verbose) 1894 printf("softdep "); 1895 return 1; 1896 } 1897 } 1898 } 1899 if (nbusy == 0) 1900 break; 1901 if (verbose) 1902 printf("%d ", nbusy); 1903 DELAY(40000 * iter); 1904 } 1905 1906 return nbusy; 1907} 1908 1909/* 1910 * posix file system related system variables. 1911 */ 1912int 1913fs_posix_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1914 int *name; 1915 u_int namelen; 1916 void *oldp; 1917 size_t *oldlenp; 1918 void *newp; 1919 size_t newlen; 1920 struct proc *p; 1921{ 1922 /* all sysctl names at this level are terminal */ 1923 if (namelen != 1) 1924 return (ENOTDIR); 1925 1926 switch (name[0]) { 1927 case FS_POSIX_SETUID: 1928 if (newp && securelevel > 0) 1929 return (EPERM); 1930 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1931 default: 1932 return (EOPNOTSUPP); 1933 } 1934 /* NOTREACHED */ 1935} 1936 1937/* 1938 * file system related system variables. 1939 */ 1940int 1941fs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) 1942 int *name; 1943 u_int namelen; 1944 void *oldp; 1945 size_t *oldlenp; 1946 void *newp; 1947 size_t newlen; 1948 struct proc *p; 1949{ 1950 sysctlfn *fn; 1951 1952 switch (name[0]) { 1953 case FS_POSIX: 1954 fn = fs_posix_sysctl; 1955 break; 1956 default: 1957 return (EOPNOTSUPP); 1958 } 1959 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1960} 1961 1962 1963/* 1964 * Routines dealing with vnodes and buffers 1965 */ 1966 1967/* 1968 * Wait for all outstanding I/Os to complete 1969 * 1970 * Manipulates v_numoutput. Must be called at splbio() 1971 */ 1972int 1973vwaitforio(vp, slpflag, wmesg, timeo) 1974 struct vnode *vp; 1975 int slpflag, timeo; 1976 char *wmesg; 1977{ 1978 int error = 0; 1979 1980 while (vp->v_numoutput) { 1981 vp->v_bioflag |= VBIOWAIT; 1982 error = tsleep((caddr_t)&vp->v_numoutput, 1983 slpflag | (PRIBIO + 1), wmesg, timeo); 1984 if (error) 1985 break; 1986 } 1987 1988 return (error); 1989} 1990 1991 1992/* 1993 * Update outstanding I/O count and do wakeup if requested. 1994 * 1995 * Manipulates v_numoutput. Must be called at splbio() 1996 */ 1997void 1998vwakeup(vp) 1999 struct vnode *vp; 2000{ 2001 if (vp != NULL) { 2002 if (vp->v_numoutput-- == 0) 2003 panic("vwakeup: neg numoutput"); 2004 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 2005 vp->v_bioflag &= ~VBIOWAIT; 2006 wakeup((caddr_t)&vp->v_numoutput); 2007 } 2008 } 2009} 2010 2011/* 2012 * Flush out and invalidate all buffers associated with a vnode. 2013 * Called with the underlying object locked. 2014 */ 2015int 2016vinvalbuf(vp, flags, cred, p, slpflag, slptimeo) 2017 register struct vnode *vp; 2018 int flags; 2019 struct ucred *cred; 2020 struct proc *p; 2021 int slpflag, slptimeo; 2022{ 2023 struct uvm_object *uobj = &vp->v_uobj; 2024 struct buf *bp; 2025 struct buf *nbp, *blist; 2026 int s, error, rv; 2027 int flushflags = PGO_ALLPAGES|PGO_FREE|PGO_SYNCIO| 2028 (flags & V_SAVE ? PGO_CLEANIT : 0); 2029 2030 /* XXXUBC this doesn't look at flags or slp* */ 2031 if (vp->v_type == VREG) { 2032 simple_lock(&uobj->vmobjlock); 2033 rv = (uobj->pgops->pgo_flush)(uobj, 0, 0, flushflags); 2034 simple_unlock(&uobj->vmobjlock); 2035 if (!rv) { 2036 return EIO; 2037 } 2038 } 2039 2040 if (flags & V_SAVE) { 2041 s = splbio(); 2042 vwaitforio(vp, 0, "vinvalbuf", 0); 2043 if (vp->v_dirtyblkhd.lh_first != NULL) { 2044 splx(s); 2045 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 2046 return (error); 2047 s = splbio(); 2048 if (vp->v_numoutput > 0 || 2049 vp->v_dirtyblkhd.lh_first != NULL) 2050 panic("vinvalbuf: dirty bufs"); 2051 } 2052 splx(s); 2053 } 2054loop: 2055 s = splbio(); 2056 for (;;) { 2057 if ((blist = vp->v_cleanblkhd.lh_first) && 2058 (flags & V_SAVEMETA)) 2059 while (blist && blist->b_lblkno < 0) 2060 blist = blist->b_vnbufs.le_next; 2061 if (!blist && (blist = vp->v_dirtyblkhd.lh_first) && 2062 (flags & V_SAVEMETA)) 2063 while (blist && blist->b_lblkno < 0) 2064 blist = blist->b_vnbufs.le_next; 2065 if (!blist) 2066 break; 2067 2068 for (bp = blist; bp; bp = nbp) { 2069 nbp = bp->b_vnbufs.le_next; 2070 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 2071 continue; 2072 if (bp->b_flags & B_BUSY) { 2073 bp->b_flags |= B_WANTED; 2074 error = tsleep((caddr_t)bp, 2075 slpflag | (PRIBIO + 1), "vinvalbuf", 2076 slptimeo); 2077 if (error) { 2078 splx(s); 2079 return (error); 2080 } 2081 break; 2082 } 2083 bremfree(bp); 2084 bp->b_flags |= B_BUSY; 2085 /* 2086 * XXX Since there are no node locks for NFS, I believe 2087 * there is a slight chance that a delayed write will 2088 * occur while sleeping just above, so check for it. 2089 */ 2090 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 2091 splx(s); 2092 (void) VOP_BWRITE(bp); 2093 goto loop; 2094 } 2095 bp->b_flags |= B_INVAL; 2096 brelse(bp); 2097 } 2098 } 2099 if (!(flags & V_SAVEMETA) && 2100 (vp->v_dirtyblkhd.lh_first || vp->v_cleanblkhd.lh_first)) 2101 panic("vinvalbuf: flush failed"); 2102 splx(s); 2103 return (0); 2104} 2105 2106void 2107vflushbuf(vp, sync) 2108 struct vnode *vp; 2109 int sync; 2110{ 2111 struct uvm_object *uobj = &vp->v_uobj; 2112 struct buf *bp, *nbp; 2113 int s; 2114 2115 if (vp->v_type == VREG) { 2116 int flags = PGO_CLEANIT|PGO_ALLPAGES| (sync ? PGO_SYNCIO : 0); 2117 2118 simple_lock(&uobj->vmobjlock); 2119 (uobj->pgops->pgo_flush)(uobj, 0, 0, flags); 2120 simple_unlock(&uobj->vmobjlock); 2121 } 2122 2123loop: 2124 s = splbio(); 2125 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { 2126 nbp = bp->b_vnbufs.le_next; 2127 if ((bp->b_flags & B_BUSY)) 2128 continue; 2129 if ((bp->b_flags & B_DELWRI) == 0) 2130 panic("vflushbuf: not dirty"); 2131 bremfree(bp); 2132 bp->b_flags |= B_BUSY; 2133 splx(s); 2134 /* 2135 * Wait for I/O associated with indirect blocks to complete, 2136 * since there is no way to quickly wait for them below. 2137 */ 2138 if (bp->b_vp == vp || sync == 0) 2139 (void) bawrite(bp); 2140 else 2141 (void) bwrite(bp); 2142 goto loop; 2143 } 2144 if (sync == 0) { 2145 splx(s); 2146 return; 2147 } 2148 vwaitforio(vp, 0, "vflushbuf", 0); 2149 if (vp->v_dirtyblkhd.lh_first != NULL) { 2150 splx(s); 2151 vprint("vflushbuf: dirty", vp); 2152 goto loop; 2153 } 2154 splx(s); 2155} 2156 2157/* 2158 * Associate a buffer with a vnode. 2159 * 2160 * Manipulates buffer vnode queues. Must be called at splbio(). 2161 */ 2162void 2163bgetvp(vp, bp) 2164 register struct vnode *vp; 2165 register struct buf *bp; 2166{ 2167 2168 if (bp->b_vp) 2169 panic("bgetvp: not free"); 2170 vhold(vp); 2171 bp->b_vp = vp; 2172 if (vp->v_type == VBLK || vp->v_type == VCHR) 2173 bp->b_dev = vp->v_rdev; 2174 else 2175 bp->b_dev = NODEV; 2176 /* 2177 * Insert onto list for new vnode. 2178 */ 2179 bufinsvn(bp, &vp->v_cleanblkhd); 2180} 2181 2182/* 2183 * Disassociate a buffer from a vnode. 2184 * 2185 * Manipulates vnode buffer queues. Must be called at splbio(). 2186 */ 2187void 2188brelvp(struct buf *bp) 2189{ 2190 struct vnode *vp; 2191 2192 if ((vp = bp->b_vp) == NULL) 2193 panic("brelvp: NULL"); 2194 2195 /* 2196 * Delete from old vnode list, if on one. 2197 */ 2198 if (bp->b_vnbufs.le_next != NOLIST) 2199 bufremvn(bp); 2200 if (TAILQ_EMPTY(&vp->v_uobj.memq) && 2201 (vp->v_bioflag & VBIOONSYNCLIST) && 2202 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2203 vp->v_bioflag &= ~VBIOONSYNCLIST; 2204 LIST_REMOVE(vp, v_synclist); 2205 } 2206 bp->b_vp = NULL; 2207 2208 vholdrele(vp); 2209} 2210 2211/* 2212 * Replaces the current vnode associated with the buffer, if any 2213 * with a new vnode. 2214 * 2215 * If an output I/O is pending on the buffer, the old vnode is 2216 * I/O count is adjusted. 2217 * 2218 * Ignores vnode buffer queues. Must be called at splbio(). 2219 */ 2220void 2221buf_replacevnode(bp, newvp) 2222 struct buf *bp; 2223 struct vnode *newvp; 2224{ 2225 struct vnode *oldvp = bp->b_vp; 2226 2227 if (oldvp) 2228 brelvp(bp); 2229 2230 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2231 newvp->v_numoutput++; /* put it on swapdev */ 2232 vwakeup(oldvp); 2233 } 2234 2235 bgetvp(newvp, bp); 2236 bufremvn(bp); 2237} 2238 2239/* 2240 * Used to assign buffers to the appropriate clean or dirty list on 2241 * the vnode and to add newly dirty vnodes to the appropriate 2242 * filesystem syncer list. 2243 * 2244 * Manipulates vnode buffer queues. Must be called at splbio(). 2245 */ 2246void 2247reassignbuf(bp) 2248 struct buf *bp; 2249{ 2250 struct buflists *listheadp; 2251 int delay; 2252 struct vnode *vp = bp->b_vp; 2253 2254 /* 2255 * Delete from old vnode list, if on one. 2256 */ 2257 if (bp->b_vnbufs.le_next != NOLIST) 2258 bufremvn(bp); 2259 /* 2260 * If dirty, put on list of dirty buffers; 2261 * otherwise insert onto list of clean buffers. 2262 */ 2263 if ((bp->b_flags & B_DELWRI) == 0) { 2264 listheadp = &vp->v_cleanblkhd; 2265 if (TAILQ_EMPTY(&vp->v_uobj.memq) && 2266 (vp->v_bioflag & VBIOONSYNCLIST) && 2267 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2268 vp->v_bioflag &= ~VBIOONSYNCLIST; 2269 LIST_REMOVE(vp, v_synclist); 2270 } 2271 } else { 2272 listheadp = &vp->v_dirtyblkhd; 2273 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2274 switch (vp->v_type) { 2275 case VDIR: 2276 delay = syncdelay / 2; 2277 break; 2278 case VBLK: 2279 if (vp->v_specmountpoint != NULL) { 2280 delay = syncdelay / 3; 2281 break; 2282 } 2283 /* fall through */ 2284 default: 2285 delay = syncdelay; 2286 } 2287 vn_syncer_add_to_worklist(vp, delay); 2288 } 2289 } 2290 bufinsvn(bp, listheadp); 2291} 2292 2293int 2294vfs_register(vfs) 2295 struct vfsconf *vfs; 2296{ 2297 struct vfsconf *vfsp; 2298 struct vfsconf **vfspp; 2299 2300#ifdef DIAGNOSTIC 2301 /* Paranoia? */ 2302 if (vfs->vfc_refcount != 0) 2303 printf("vfs_register called with vfc_refcount > 0\n"); 2304#endif 2305 2306 /* Check if filesystem already known */ 2307 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2308 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2309 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2310 return (EEXIST); 2311 2312 if (vfs->vfc_typenum > maxvfsconf) 2313 maxvfsconf = vfs->vfc_typenum; 2314 2315 vfs->vfc_next = NULL; 2316 2317 /* Add to the end of the list */ 2318 *vfspp = vfs; 2319 2320 /* Call vfs_init() */ 2321 if (vfs->vfc_vfsops->vfs_init) 2322 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2323 2324 return 0; 2325} 2326 2327int 2328vfs_unregister(vfs) 2329 struct vfsconf *vfs; 2330{ 2331 struct vfsconf *vfsp; 2332 struct vfsconf **vfspp; 2333 int maxtypenum; 2334 2335 /* Find our vfsconf struct */ 2336 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2337 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2338 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2339 break; 2340 } 2341 2342 if (!vfsp) /* Not found */ 2343 return (ENOENT); 2344 2345 if (vfsp->vfc_refcount) /* In use */ 2346 return (EBUSY); 2347 2348 /* Remove from list and free */ 2349 *vfspp = vfsp->vfc_next; 2350 2351 maxtypenum = 0; 2352 2353 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2354 if (vfsp->vfc_typenum > maxtypenum) 2355 maxtypenum = vfsp->vfc_typenum; 2356 2357 maxvfsconf = maxtypenum; 2358 return 0; 2359} 2360 2361/* 2362 * Check if vnode represents a disk device 2363 */ 2364int 2365vn_isdisk(vp, errp) 2366 struct vnode *vp; 2367 int *errp; 2368{ 2369 if (vp->v_type != VBLK && vp->v_type != VCHR) 2370 return (0); 2371 2372 return (1); 2373} 2374