vfs_subr.c revision 1.184
1/* $OpenBSD: vfs_subr.c,v 1.184 2009/12/17 16:44:12 oga Exp $ */ 2/* $NetBSD: vfs_subr.c,v 1.53 1996/04/22 01:39:13 christos Exp $ */ 3 4/* 5 * Copyright (c) 1989, 1993 6 * The Regents of the University of California. All rights reserved. 7 * (c) UNIX System Laboratories, Inc. 8 * All or some portions of this file are derived from material licensed 9 * to the University of California by American Telephone and Telegraph 10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 11 * the permission of UNIX System Laboratories, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 * 37 * @(#)vfs_subr.c 8.13 (Berkeley) 4/18/94 38 */ 39 40/* 41 * External virtual filesystem routines 42 */ 43 44#include <sys/param.h> 45#include <sys/systm.h> 46#include <sys/proc.h> 47#include <sys/mount.h> 48#include <sys/time.h> 49#include <sys/fcntl.h> 50#include <sys/kernel.h> 51#include <sys/vnode.h> 52#include <sys/stat.h> 53#include <sys/namei.h> 54#include <sys/ucred.h> 55#include <sys/buf.h> 56#include <sys/errno.h> 57#include <sys/malloc.h> 58#include <sys/domain.h> 59#include <sys/mbuf.h> 60#include <sys/syscallargs.h> 61#include <sys/pool.h> 62#include <sys/tree.h> 63 64#include <uvm/uvm_extern.h> 65#include <sys/sysctl.h> 66 67#include <miscfs/specfs/specdev.h> 68 69enum vtype iftovt_tab[16] = { 70 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, 71 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, 72}; 73 74int vttoif_tab[9] = { 75 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, 76 S_IFSOCK, S_IFIFO, S_IFMT, 77}; 78 79int doforce = 1; /* 1 => permit forcible unmounting */ 80int prtactive = 0; /* 1 => print out reclaim of active vnodes */ 81int suid_clear = 1; /* 1 => clear SUID / SGID on owner change */ 82 83/* 84 * Insq/Remq for the vnode usage lists. 85 */ 86#define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) 87#define bufremvn(bp) { \ 88 LIST_REMOVE(bp, b_vnbufs); \ 89 LIST_NEXT(bp, b_vnbufs) = NOLIST; \ 90} 91 92struct freelst vnode_hold_list; /* list of vnodes referencing buffers */ 93struct freelst vnode_free_list; /* vnode free list */ 94 95struct mntlist mountlist; /* mounted filesystem list */ 96 97void vclean(struct vnode *, int, struct proc *); 98void vhold(struct vnode *); 99void vdrop(struct vnode *); 100 101void insmntque(struct vnode *, struct mount *); 102int getdevvp(dev_t, struct vnode **, enum vtype); 103 104int vfs_hang_addrlist(struct mount *, struct netexport *, 105 struct export_args *); 106int vfs_free_netcred(struct radix_node *, void *); 107void vfs_free_addrlist(struct netexport *); 108void vputonfreelist(struct vnode *); 109 110int vflush_vnode(struct vnode *, void *); 111int maxvnodes; 112 113#ifdef DEBUG 114void printlockedvnodes(void); 115#endif 116 117struct pool vnode_pool; 118 119static int rb_buf_compare(struct buf *b1, struct buf *b2); 120RB_GENERATE(buf_rb_bufs, buf, b_rbbufs, rb_buf_compare); 121 122static int 123rb_buf_compare(struct buf *b1, struct buf *b2) 124{ 125 if (b1->b_lblkno < b2->b_lblkno) 126 return(-1); 127 if (b1->b_lblkno > b2->b_lblkno) 128 return(1); 129 return(0); 130} 131 132/* 133 * Initialize the vnode management data structures. 134 */ 135void 136vntblinit(void) 137{ 138 /* buffer cache may need a vnode for each buffer */ 139 maxvnodes = desiredvnodes; 140 pool_init(&vnode_pool, sizeof(struct vnode), 0, 0, 0, "vnodes", 141 &pool_allocator_nointr); 142 TAILQ_INIT(&vnode_hold_list); 143 TAILQ_INIT(&vnode_free_list); 144 CIRCLEQ_INIT(&mountlist); 145 /* 146 * Initialize the filesystem syncer. 147 */ 148 vn_initialize_syncerd(); 149} 150 151/* 152 * Mark a mount point as busy. Used to synchronize access and to delay 153 * unmounting. 154 * 155 * Default behaviour is to attempt getting a READ lock and in case of an 156 * ongoing unmount, to wait for it to finish and then return failure. 157 */ 158int 159vfs_busy(struct mount *mp, int flags) 160{ 161 int rwflags = 0; 162 163 /* new mountpoints need their lock initialised */ 164 if (mp->mnt_lock.rwl_name == NULL) 165 rw_init(&mp->mnt_lock, "vfslock"); 166 167 if (flags & VB_WRITE) 168 rwflags |= RW_WRITE; 169 else 170 rwflags |= RW_READ; 171 172 if (flags & VB_WAIT) 173 rwflags |= RW_SLEEPFAIL; 174 else 175 rwflags |= RW_NOSLEEP; 176 177 if (rw_enter(&mp->mnt_lock, rwflags)) 178 return (EBUSY); 179 180 return (0); 181} 182 183/* 184 * Free a busy file system 185 */ 186void 187vfs_unbusy(struct mount *mp) 188{ 189 rw_exit(&mp->mnt_lock); 190} 191 192int 193vfs_isbusy(struct mount *mp) 194{ 195 if (RWLOCK_OWNER(&mp->mnt_lock) > 0) 196 return (1); 197 else 198 return (0); 199} 200 201/* 202 * Lookup a filesystem type, and if found allocate and initialize 203 * a mount structure for it. 204 * 205 * Devname is usually updated by mount(8) after booting. 206 */ 207int 208vfs_rootmountalloc(char *fstypename, char *devname, struct mount **mpp) 209{ 210 struct vfsconf *vfsp; 211 struct mount *mp; 212 213 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 214 if (!strcmp(vfsp->vfc_name, fstypename)) 215 break; 216 if (vfsp == NULL) 217 return (ENODEV); 218 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK|M_ZERO); 219 (void)vfs_busy(mp, VB_READ|VB_NOWAIT); 220 LIST_INIT(&mp->mnt_vnodelist); 221 mp->mnt_vfc = vfsp; 222 mp->mnt_op = vfsp->vfc_vfsops; 223 mp->mnt_flag = MNT_RDONLY; 224 mp->mnt_vnodecovered = NULLVP; 225 vfsp->vfc_refcount++; 226 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 227 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 228 mp->mnt_stat.f_mntonname[0] = '/'; 229 (void)copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); 230 *mpp = mp; 231 return (0); 232 } 233 234/* 235 * Lookup a mount point by filesystem identifier. 236 */ 237struct mount * 238vfs_getvfs(fsid_t *fsid) 239{ 240 struct mount *mp; 241 242 CIRCLEQ_FOREACH(mp, &mountlist, mnt_list) { 243 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 244 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 245 return (mp); 246 } 247 } 248 249 return (NULL); 250} 251 252 253/* 254 * Get a new unique fsid 255 */ 256void 257vfs_getnewfsid(struct mount *mp) 258{ 259 static u_short xxxfs_mntid; 260 261 fsid_t tfsid; 262 int mtype; 263 264 mtype = mp->mnt_vfc->vfc_typenum; 265 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 266 mp->mnt_stat.f_fsid.val[1] = mtype; 267 if (xxxfs_mntid == 0) 268 ++xxxfs_mntid; 269 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 270 tfsid.val[1] = mtype; 271 if (!CIRCLEQ_EMPTY(&mountlist)) { 272 while (vfs_getvfs(&tfsid)) { 273 tfsid.val[0]++; 274 xxxfs_mntid++; 275 } 276 } 277 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 278} 279 280/* 281 * Make a 'unique' number from a mount type name. 282 * Note that this is no longer used for ffs which 283 * now has an on-disk filesystem id. 284 */ 285long 286makefstype(char *type) 287{ 288 long rv; 289 290 for (rv = 0; *type; type++) { 291 rv <<= 2; 292 rv ^= *type; 293 } 294 return rv; 295} 296 297/* 298 * Set vnode attributes to VNOVAL 299 */ 300void 301vattr_null(struct vattr *vap) 302{ 303 304 vap->va_type = VNON; 305 /* XXX These next two used to be one line, but for a GCC bug. */ 306 vap->va_size = VNOVAL; 307 vap->va_bytes = VNOVAL; 308 vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = 309 vap->va_fsid = vap->va_fileid = 310 vap->va_blocksize = vap->va_rdev = 311 vap->va_atime.tv_sec = vap->va_atime.tv_nsec = 312 vap->va_mtime.tv_sec = vap->va_mtime.tv_nsec = 313 vap->va_ctime.tv_sec = vap->va_ctime.tv_nsec = 314 vap->va_flags = vap->va_gen = VNOVAL; 315 vap->va_vaflags = 0; 316} 317 318/* 319 * Routines having to do with the management of the vnode table. 320 */ 321extern int (**dead_vnodeop_p)(void *); 322long numvnodes; 323 324/* 325 * Return the next vnode from the free list. 326 */ 327int 328getnewvnode(enum vtagtype tag, struct mount *mp, int (**vops)(void *), 329 struct vnode **vpp) 330{ 331 struct proc *p = curproc; 332 struct freelst *listhd; 333 static int toggle; 334 struct vnode *vp; 335 int s; 336 337 /* 338 * We must choose whether to allocate a new vnode or recycle an 339 * existing one. The criterion for allocating a new one is that 340 * the total number of vnodes is less than the number desired or 341 * there are no vnodes on either free list. Generally we only 342 * want to recycle vnodes that have no buffers associated with 343 * them, so we look first on the vnode_free_list. If it is empty, 344 * we next consider vnodes with referencing buffers on the 345 * vnode_hold_list. The toggle ensures that half the time we 346 * will use a buffer from the vnode_hold_list, and half the time 347 * we will allocate a new one unless the list has grown to twice 348 * the desired size. We are reticent to recycle vnodes from the 349 * vnode_hold_list because we will lose the identity of all its 350 * referencing buffers. 351 */ 352 toggle ^= 1; 353 if (numvnodes > 2 * maxvnodes) 354 toggle = 0; 355 356 s = splbio(); 357 if ((numvnodes < maxvnodes) || 358 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 359 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 360 splx(s); 361 vp = pool_get(&vnode_pool, PR_WAITOK | PR_ZERO); 362 RB_INIT(&vp->v_bufs_tree); 363 RB_INIT(&vp->v_nc_tree); 364 TAILQ_INIT(&vp->v_cache_dst); 365 numvnodes++; 366 } else { 367 for (vp = TAILQ_FIRST(listhd); vp != NULLVP; 368 vp = TAILQ_NEXT(vp, v_freelist)) { 369 if (VOP_ISLOCKED(vp) == 0) 370 break; 371 } 372 /* 373 * Unless this is a bad time of the month, at most 374 * the first NCPUS items on the free list are 375 * locked, so this is close enough to being empty. 376 */ 377 if (vp == NULL) { 378 splx(s); 379 tablefull("vnode"); 380 *vpp = 0; 381 return (ENFILE); 382 } 383 384#ifdef DIAGNOSTIC 385 if (vp->v_usecount) { 386 vprint("free vnode", vp); 387 panic("free vnode isn't"); 388 } 389#endif 390 391 TAILQ_REMOVE(listhd, vp, v_freelist); 392 vp->v_bioflag &= ~VBIOONFREELIST; 393 splx(s); 394 395 if (vp->v_type != VBAD) 396 vgonel(vp, p); 397#ifdef DIAGNOSTIC 398 if (vp->v_data) { 399 vprint("cleaned vnode", vp); 400 panic("cleaned vnode isn't"); 401 } 402 s = splbio(); 403 if (vp->v_numoutput) 404 panic("Clean vnode has pending I/O's"); 405 splx(s); 406#endif 407 vp->v_flag = 0; 408 vp->v_socket = 0; 409 } 410 vp->v_type = VNON; 411 cache_purge(vp); 412 vp->v_tag = tag; 413 vp->v_op = vops; 414 insmntque(vp, mp); 415 *vpp = vp; 416 vp->v_usecount = 1; 417 vp->v_data = 0; 418 simple_lock_init(&vp->v_uvm.u_obj.vmobjlock); 419 return (0); 420} 421 422/* 423 * Move a vnode from one mount queue to another. 424 */ 425void 426insmntque(struct vnode *vp, struct mount *mp) 427{ 428 /* 429 * Delete from old mount point vnode list, if on one. 430 */ 431 if (vp->v_mount != NULL) 432 LIST_REMOVE(vp, v_mntvnodes); 433 /* 434 * Insert into list of vnodes for the new mount point, if available. 435 */ 436 if ((vp->v_mount = mp) != NULL) 437 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 438} 439 440/* 441 * Create a vnode for a block device. 442 * Used for root filesystem, argdev, and swap areas. 443 * Also used for memory file system special devices. 444 */ 445int 446bdevvp(dev_t dev, struct vnode **vpp) 447{ 448 return (getdevvp(dev, vpp, VBLK)); 449} 450 451/* 452 * Create a vnode for a character device. 453 * Used for console handling. 454 */ 455int 456cdevvp(dev_t dev, struct vnode **vpp) 457{ 458 return (getdevvp(dev, vpp, VCHR)); 459} 460 461/* 462 * Create a vnode for a device. 463 * Used by bdevvp (block device) for root file system etc., 464 * and by cdevvp (character device) for console. 465 */ 466int 467getdevvp(dev_t dev, struct vnode **vpp, enum vtype type) 468{ 469 struct vnode *vp; 470 struct vnode *nvp; 471 int error; 472 473 if (dev == NODEV) { 474 *vpp = NULLVP; 475 return (0); 476 } 477 error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); 478 if (error) { 479 *vpp = NULLVP; 480 return (error); 481 } 482 vp = nvp; 483 vp->v_type = type; 484 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 485 vput(vp); 486 vp = nvp; 487 } 488 *vpp = vp; 489 return (0); 490} 491 492/* 493 * Check to see if the new vnode represents a special device 494 * for which we already have a vnode (either because of 495 * bdevvp() or because of a different vnode representing 496 * the same block device). If such an alias exists, deallocate 497 * the existing contents and return the aliased vnode. The 498 * caller is responsible for filling it with its new contents. 499 */ 500struct vnode * 501checkalias(struct vnode *nvp, dev_t nvp_rdev, struct mount *mp) 502{ 503 struct proc *p = curproc; 504 struct vnode *vp; 505 struct vnode **vpp; 506 507 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 508 return (NULLVP); 509 510 vpp = &speclisth[SPECHASH(nvp_rdev)]; 511loop: 512 for (vp = *vpp; vp; vp = vp->v_specnext) { 513 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 514 continue; 515 } 516 /* 517 * Alias, but not in use, so flush it out. 518 */ 519 if (vp->v_usecount == 0) { 520 vgonel(vp, p); 521 goto loop; 522 } 523 if (vget(vp, LK_EXCLUSIVE, p)) { 524 goto loop; 525 } 526 break; 527 } 528 529 /* 530 * Common case is actually in the if statement 531 */ 532 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 533 nvp->v_specinfo = malloc(sizeof(struct specinfo), M_VNODE, 534 M_WAITOK); 535 nvp->v_rdev = nvp_rdev; 536 nvp->v_hashchain = vpp; 537 nvp->v_specnext = *vpp; 538 nvp->v_specmountpoint = NULL; 539 nvp->v_speclockf = NULL; 540 bzero(nvp->v_specbitmap, sizeof(nvp->v_specbitmap)); 541 *vpp = nvp; 542 if (vp != NULLVP) { 543 nvp->v_flag |= VALIASED; 544 vp->v_flag |= VALIASED; 545 vput(vp); 546 } 547 return (NULLVP); 548 } 549 550 /* 551 * This code is the uncommon case. It is called in case 552 * we found an alias that was VT_NON && vtype of VBLK 553 * This means we found a block device that was created 554 * using bdevvp. 555 * An example of such a vnode is the root partition device vnode 556 * created in ffs_mountroot. 557 * 558 * The vnodes created by bdevvp should not be aliased (why?). 559 */ 560 561 VOP_UNLOCK(vp, 0, p); 562 vclean(vp, 0, p); 563 vp->v_op = nvp->v_op; 564 vp->v_tag = nvp->v_tag; 565 nvp->v_type = VNON; 566 insmntque(vp, mp); 567 return (vp); 568} 569 570/* 571 * Grab a particular vnode from the free list, increment its 572 * reference count and lock it. If the vnode lock bit is set, 573 * the vnode is being eliminated in vgone. In that case, we 574 * cannot grab it, so the process is awakened when the 575 * transition is completed, and an error code is returned to 576 * indicate that the vnode is no longer usable, possibly 577 * having been changed to a new file system type. 578 */ 579int 580vget(struct vnode *vp, int flags, struct proc *p) 581{ 582 int error, s, onfreelist; 583 584 /* 585 * If the vnode is in the process of being cleaned out for 586 * another use, we wait for the cleaning to finish and then 587 * return failure. Cleaning is determined by checking that 588 * the VXLOCK flag is set. 589 */ 590 591 if (vp->v_flag & VXLOCK) { 592 if (flags & LK_NOWAIT) { 593 return (EBUSY); 594 } 595 596 vp->v_flag |= VXWANT; 597 tsleep(vp, PINOD, "vget", 0); 598 return (ENOENT); 599 } 600 601 onfreelist = vp->v_bioflag & VBIOONFREELIST; 602 if (vp->v_usecount == 0 && onfreelist) { 603 s = splbio(); 604 if (vp->v_holdcnt > 0) 605 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 606 else 607 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 608 vp->v_bioflag &= ~VBIOONFREELIST; 609 splx(s); 610 } 611 612 vp->v_usecount++; 613 if (flags & LK_TYPE_MASK) { 614 if ((error = vn_lock(vp, flags, p)) != 0) { 615 vp->v_usecount--; 616 if (vp->v_usecount == 0 && onfreelist) 617 vputonfreelist(vp); 618 } 619 return (error); 620 } 621 622 return (0); 623} 624 625 626/* Vnode reference. */ 627void 628vref(struct vnode *vp) 629{ 630#ifdef DIAGNOSTIC 631 if (vp->v_usecount == 0) 632 panic("vref used where vget required"); 633 if (vp->v_type == VNON) 634 panic("vref on a VNON vnode"); 635#endif 636 vp->v_usecount++; 637} 638 639void 640vputonfreelist(struct vnode *vp) 641{ 642 int s; 643 struct freelst *lst; 644 645 s = splbio(); 646#ifdef DIAGNOSTIC 647 if (vp->v_usecount != 0) 648 panic("Use count is not zero!"); 649 650 if (vp->v_bioflag & VBIOONFREELIST) { 651 vprint("vnode already on free list: ", vp); 652 panic("vnode already on free list"); 653 } 654#endif 655 656 vp->v_bioflag |= VBIOONFREELIST; 657 658 if (vp->v_holdcnt > 0) 659 lst = &vnode_hold_list; 660 else 661 lst = &vnode_free_list; 662 663 if (vp->v_type == VBAD) 664 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 665 else 666 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 667 668 splx(s); 669} 670 671/* 672 * vput(), just unlock and vrele() 673 */ 674void 675vput(struct vnode *vp) 676{ 677 struct proc *p = curproc; 678 679#ifdef DIAGNOSTIC 680 if (vp == NULL) 681 panic("vput: null vp"); 682#endif 683 684#ifdef DIAGNOSTIC 685 if (vp->v_usecount == 0) { 686 vprint("vput: bad ref count", vp); 687 panic("vput: ref cnt"); 688 } 689#endif 690 vp->v_usecount--; 691 if (vp->v_usecount > 0) { 692 VOP_UNLOCK(vp, 0, p); 693 return; 694 } 695 696#ifdef DIAGNOSTIC 697 if (vp->v_writecount != 0) { 698 vprint("vput: bad writecount", vp); 699 panic("vput: v_writecount != 0"); 700 } 701#endif 702 703 VOP_INACTIVE(vp, p); 704 705 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 706 vputonfreelist(vp); 707} 708 709/* 710 * Vnode release - use for active VNODES. 711 * If count drops to zero, call inactive routine and return to freelist. 712 * Returns 0 if it did not sleep. 713 */ 714int 715vrele(struct vnode *vp) 716{ 717 struct proc *p = curproc; 718 719#ifdef DIAGNOSTIC 720 if (vp == NULL) 721 panic("vrele: null vp"); 722#endif 723#ifdef DIAGNOSTIC 724 if (vp->v_usecount == 0) { 725 vprint("vrele: bad ref count", vp); 726 panic("vrele: ref cnt"); 727 } 728#endif 729 vp->v_usecount--; 730 if (vp->v_usecount > 0) { 731 return (0); 732 } 733 734#ifdef DIAGNOSTIC 735 if (vp->v_writecount != 0) { 736 vprint("vrele: bad writecount", vp); 737 panic("vrele: v_writecount != 0"); 738 } 739#endif 740 741 if (vn_lock(vp, LK_EXCLUSIVE, p)) { 742#ifdef DIAGNOSTIC 743 vprint("vrele: cannot lock", vp); 744#endif 745 return (1); 746 } 747 748 VOP_INACTIVE(vp, p); 749 750 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 751 vputonfreelist(vp); 752 return (1); 753} 754 755/* Page or buffer structure gets a reference. */ 756void 757vhold(struct vnode *vp) 758{ 759 /* 760 * If it is on the freelist and the hold count is currently 761 * zero, move it to the hold list. 762 */ 763 if ((vp->v_bioflag & VBIOONFREELIST) && 764 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 765 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 766 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 767 } 768 vp->v_holdcnt++; 769} 770 771/* Lose interest in a vnode. */ 772void 773vdrop(struct vnode *vp) 774{ 775#ifdef DIAGNOSTIC 776 if (vp->v_holdcnt == 0) 777 panic("vdrop: zero holdcnt"); 778#endif 779 780 vp->v_holdcnt--; 781 782 /* 783 * If it is on the holdlist and the hold count drops to 784 * zero, move it to the free list. 785 */ 786 if ((vp->v_bioflag & VBIOONFREELIST) && 787 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 788 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 789 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 790 } 791} 792 793/* 794 * Remove any vnodes in the vnode table belonging to mount point mp. 795 * 796 * If MNT_NOFORCE is specified, there should not be any active ones, 797 * return error if any are found (nb: this is a user error, not a 798 * system error). If MNT_FORCE is specified, detach any active vnodes 799 * that are found. 800 */ 801#ifdef DEBUG 802int busyprt = 0; /* print out busy vnodes */ 803struct ctldebug debug1 = { "busyprt", &busyprt }; 804#endif 805 806int 807vfs_mount_foreach_vnode(struct mount *mp, 808 int (*func)(struct vnode *, void *), void *arg) { 809 struct vnode *vp, *nvp; 810 int error = 0; 811 812loop: 813 for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; vp = nvp) { 814 if (vp->v_mount != mp) 815 goto loop; 816 nvp = LIST_NEXT(vp, v_mntvnodes); 817 818 error = func(vp, arg); 819 820 if (error != 0) 821 break; 822 } 823 824 return (error); 825} 826 827struct vflush_args { 828 struct vnode *skipvp; 829 int busy; 830 int flags; 831}; 832 833int 834vflush_vnode(struct vnode *vp, void *arg) { 835 struct vflush_args *va = arg; 836 struct proc *p = curproc; 837 838 if (vp == va->skipvp) { 839 return (0); 840 } 841 842 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 843 return (0); 844 } 845 846 /* 847 * If WRITECLOSE is set, only flush out regular file 848 * vnodes open for writing. 849 */ 850 if ((va->flags & WRITECLOSE) && 851 (vp->v_writecount == 0 || vp->v_type != VREG)) { 852 return (0); 853 } 854 855 /* 856 * With v_usecount == 0, all we need to do is clear 857 * out the vnode data structures and we are done. 858 */ 859 if (vp->v_usecount == 0) { 860 vgonel(vp, p); 861 return (0); 862 } 863 864 /* 865 * If FORCECLOSE is set, forcibly close the vnode. 866 * For block or character devices, revert to an 867 * anonymous device. For all other files, just kill them. 868 */ 869 if (va->flags & FORCECLOSE) { 870 if (vp->v_type != VBLK && vp->v_type != VCHR) { 871 vgonel(vp, p); 872 } else { 873 vclean(vp, 0, p); 874 vp->v_op = spec_vnodeop_p; 875 insmntque(vp, (struct mount *)0); 876 } 877 return (0); 878 } 879 880#ifdef DEBUG 881 if (busyprt) 882 vprint("vflush: busy vnode", vp); 883#endif 884 va->busy++; 885 return (0); 886} 887 888int 889vflush(struct mount *mp, struct vnode *skipvp, int flags) 890{ 891 struct vflush_args va; 892 va.skipvp = skipvp; 893 va.busy = 0; 894 va.flags = flags; 895 896 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 897 898 if (va.busy) 899 return (EBUSY); 900 return (0); 901} 902 903/* 904 * Disassociate the underlying file system from a vnode. 905 */ 906void 907vclean(struct vnode *vp, int flags, struct proc *p) 908{ 909 int active; 910 911 /* 912 * Check to see if the vnode is in use. 913 * If so we have to reference it before we clean it out 914 * so that its count cannot fall to zero and generate a 915 * race against ourselves to recycle it. 916 */ 917 if ((active = vp->v_usecount) != 0) 918 vp->v_usecount++; 919 920 /* 921 * Prevent the vnode from being recycled or 922 * brought into use while we clean it out. 923 */ 924 if (vp->v_flag & VXLOCK) 925 panic("vclean: deadlock"); 926 vp->v_flag |= VXLOCK; 927 /* 928 * Even if the count is zero, the VOP_INACTIVE routine may still 929 * have the object locked while it cleans it out. The VOP_LOCK 930 * ensures that the VOP_INACTIVE routine is done with its work. 931 * For active vnodes, it ensures that no other activity can 932 * occur while the underlying object is being cleaned out. 933 */ 934 VOP_LOCK(vp, LK_DRAIN, p); 935 936 /* 937 * Clean out any VM data associated with the vnode. 938 */ 939 uvm_vnp_terminate(vp); 940 /* 941 * Clean out any buffers associated with the vnode. 942 */ 943 if (flags & DOCLOSE) 944 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 945 /* 946 * If purging an active vnode, it must be closed and 947 * deactivated before being reclaimed. Note that the 948 * VOP_INACTIVE will unlock the vnode 949 */ 950 if (active) { 951 if (flags & DOCLOSE) 952 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 953 VOP_INACTIVE(vp, p); 954 } else { 955 /* 956 * Any other processes trying to obtain this lock must first 957 * wait for VXLOCK to clear, then call the new lock operation. 958 */ 959 VOP_UNLOCK(vp, 0, p); 960 } 961 962 /* 963 * Reclaim the vnode. 964 */ 965 if (VOP_RECLAIM(vp, p)) 966 panic("vclean: cannot reclaim"); 967 if (active) { 968 vp->v_usecount--; 969 if (vp->v_usecount == 0) { 970 if (vp->v_holdcnt > 0) 971 panic("vclean: not clean"); 972 vputonfreelist(vp); 973 } 974 } 975 cache_purge(vp); 976 977 /* 978 * Done with purge, notify sleepers of the grim news. 979 */ 980 vp->v_op = dead_vnodeop_p; 981 VN_KNOTE(vp, NOTE_REVOKE); 982 vp->v_tag = VT_NON; 983 vp->v_flag &= ~VXLOCK; 984#ifdef VFSDEBUG 985 vp->v_flag &= ~VLOCKSWORK; 986#endif 987 if (vp->v_flag & VXWANT) { 988 vp->v_flag &= ~VXWANT; 989 wakeup(vp); 990 } 991} 992 993/* 994 * Recycle an unused vnode to the front of the free list. 995 */ 996int 997vrecycle(struct vnode *vp, struct proc *p) 998{ 999 if (vp->v_usecount == 0) { 1000 vgonel(vp, p); 1001 return (1); 1002 } 1003 return (0); 1004} 1005 1006/* 1007 * Eliminate all activity associated with a vnode 1008 * in preparation for reuse. 1009 */ 1010void 1011vgone(struct vnode *vp) 1012{ 1013 struct proc *p = curproc; 1014 vgonel(vp, p); 1015} 1016 1017/* 1018 * vgone, with struct proc. 1019 */ 1020void 1021vgonel(struct vnode *vp, struct proc *p) 1022{ 1023 struct vnode *vq; 1024 struct vnode *vx; 1025 1026 /* 1027 * If a vgone (or vclean) is already in progress, 1028 * wait until it is done and return. 1029 */ 1030 if (vp->v_flag & VXLOCK) { 1031 vp->v_flag |= VXWANT; 1032 tsleep(vp, PINOD, "vgone", 0); 1033 return; 1034 } 1035 1036 /* 1037 * Clean out the filesystem specific data. 1038 */ 1039 vclean(vp, DOCLOSE, p); 1040 /* 1041 * Delete from old mount point vnode list, if on one. 1042 */ 1043 if (vp->v_mount != NULL) 1044 insmntque(vp, (struct mount *)0); 1045 /* 1046 * If special device, remove it from special device alias list 1047 * if it is on one. 1048 */ 1049 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1050 if (*vp->v_hashchain == vp) { 1051 *vp->v_hashchain = vp->v_specnext; 1052 } else { 1053 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1054 if (vq->v_specnext != vp) 1055 continue; 1056 vq->v_specnext = vp->v_specnext; 1057 break; 1058 } 1059 if (vq == NULL) 1060 panic("missing bdev"); 1061 } 1062 if (vp->v_flag & VALIASED) { 1063 vx = NULL; 1064 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1065 if (vq->v_rdev != vp->v_rdev || 1066 vq->v_type != vp->v_type) 1067 continue; 1068 if (vx) 1069 break; 1070 vx = vq; 1071 } 1072 if (vx == NULL) 1073 panic("missing alias"); 1074 if (vq == NULL) 1075 vx->v_flag &= ~VALIASED; 1076 vp->v_flag &= ~VALIASED; 1077 } 1078 free(vp->v_specinfo, M_VNODE); 1079 vp->v_specinfo = NULL; 1080 } 1081 /* 1082 * If it is on the freelist and not already at the head, 1083 * move it to the head of the list. 1084 */ 1085 vp->v_type = VBAD; 1086 1087 /* 1088 * Move onto the free list, unless we were called from 1089 * getnewvnode and we're not on any free list 1090 */ 1091 if (vp->v_usecount == 0 && 1092 (vp->v_bioflag & VBIOONFREELIST)) { 1093 int s; 1094 1095 s = splbio(); 1096 1097 if (vp->v_holdcnt > 0) 1098 panic("vgonel: not clean"); 1099 1100 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1101 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1102 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1103 } 1104 splx(s); 1105 } 1106} 1107 1108/* 1109 * Lookup a vnode by device number. 1110 */ 1111int 1112vfinddev(dev_t dev, enum vtype type, struct vnode **vpp) 1113{ 1114 struct vnode *vp; 1115 int rc =0; 1116 1117 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1118 if (dev != vp->v_rdev || type != vp->v_type) 1119 continue; 1120 *vpp = vp; 1121 rc = 1; 1122 break; 1123 } 1124 return (rc); 1125} 1126 1127/* 1128 * Revoke all the vnodes corresponding to the specified minor number 1129 * range (endpoints inclusive) of the specified major. 1130 */ 1131void 1132vdevgone(int maj, int minl, int minh, enum vtype type) 1133{ 1134 struct vnode *vp; 1135 int mn; 1136 1137 for (mn = minl; mn <= minh; mn++) 1138 if (vfinddev(makedev(maj, mn), type, &vp)) 1139 VOP_REVOKE(vp, REVOKEALL); 1140} 1141 1142/* 1143 * Calculate the total number of references to a special device. 1144 */ 1145int 1146vcount(struct vnode *vp) 1147{ 1148 struct vnode *vq, *vnext; 1149 int count; 1150 1151loop: 1152 if ((vp->v_flag & VALIASED) == 0) 1153 return (vp->v_usecount); 1154 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1155 vnext = vq->v_specnext; 1156 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1157 continue; 1158 /* 1159 * Alias, but not in use, so flush it out. 1160 */ 1161 if (vq->v_usecount == 0 && vq != vp) { 1162 vgone(vq); 1163 goto loop; 1164 } 1165 count += vq->v_usecount; 1166 } 1167 return (count); 1168} 1169 1170#if defined(DEBUG) || defined(DIAGNOSTIC) 1171/* 1172 * Print out a description of a vnode. 1173 */ 1174static char *typename[] = 1175 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1176 1177void 1178vprint(char *label, struct vnode *vp) 1179{ 1180 char buf[64]; 1181 1182 if (label != NULL) 1183 printf("%s: ", label); 1184 printf("%p, type %s, use %u, write %u, hold %u,", 1185 vp, typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1186 vp->v_holdcnt); 1187 buf[0] = '\0'; 1188 if (vp->v_flag & VROOT) 1189 strlcat(buf, "|VROOT", sizeof buf); 1190 if (vp->v_flag & VTEXT) 1191 strlcat(buf, "|VTEXT", sizeof buf); 1192 if (vp->v_flag & VSYSTEM) 1193 strlcat(buf, "|VSYSTEM", sizeof buf); 1194 if (vp->v_flag & VXLOCK) 1195 strlcat(buf, "|VXLOCK", sizeof buf); 1196 if (vp->v_flag & VXWANT) 1197 strlcat(buf, "|VXWANT", sizeof buf); 1198 if (vp->v_bioflag & VBIOWAIT) 1199 strlcat(buf, "|VBIOWAIT", sizeof buf); 1200 if (vp->v_bioflag & VBIOONFREELIST) 1201 strlcat(buf, "|VBIOONFREELIST", sizeof buf); 1202 if (vp->v_bioflag & VBIOONSYNCLIST) 1203 strlcat(buf, "|VBIOONSYNCLIST", sizeof buf); 1204 if (vp->v_flag & VALIASED) 1205 strlcat(buf, "|VALIASED", sizeof buf); 1206 if (buf[0] != '\0') 1207 printf(" flags (%s)", &buf[1]); 1208 if (vp->v_data == NULL) { 1209 printf("\n"); 1210 } else { 1211 printf("\n\t"); 1212 VOP_PRINT(vp); 1213 } 1214} 1215#endif /* DEBUG || DIAGNOSTIC */ 1216 1217#ifdef DEBUG 1218/* 1219 * List all of the locked vnodes in the system. 1220 * Called when debugging the kernel. 1221 */ 1222void 1223printlockedvnodes(void) 1224{ 1225 struct mount *mp, *nmp; 1226 struct vnode *vp; 1227 1228 printf("Locked vnodes\n"); 1229 1230 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1231 mp = nmp) { 1232 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) { 1233 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1234 continue; 1235 } 1236 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 1237 if (VOP_ISLOCKED(vp)) 1238 vprint((char *)0, vp); 1239 } 1240 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1241 vfs_unbusy(mp); 1242 } 1243 1244} 1245#endif 1246 1247/* 1248 * Top level filesystem related information gathering. 1249 */ 1250int 1251vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1252 size_t newlen, struct proc *p) 1253{ 1254 struct vfsconf *vfsp, *tmpvfsp; 1255 int ret; 1256 1257 /* all sysctl names at this level are at least name and field */ 1258 if (namelen < 2) 1259 return (ENOTDIR); /* overloaded */ 1260 1261 if (name[0] != VFS_GENERIC) { 1262 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1263 if (vfsp->vfc_typenum == name[0]) 1264 break; 1265 1266 if (vfsp == NULL) 1267 return (EOPNOTSUPP); 1268 1269 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1270 oldp, oldlenp, newp, newlen, p)); 1271 } 1272 1273 switch (name[1]) { 1274 case VFS_MAXTYPENUM: 1275 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1276 1277 case VFS_CONF: 1278 if (namelen < 3) 1279 return (ENOTDIR); /* overloaded */ 1280 1281 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1282 if (vfsp->vfc_typenum == name[2]) 1283 break; 1284 1285 if (vfsp == NULL) 1286 return (EOPNOTSUPP); 1287 1288 /* Make a copy, clear out kernel pointers */ 1289 tmpvfsp = malloc(sizeof(*tmpvfsp), M_TEMP, M_WAITOK); 1290 bcopy(vfsp, tmpvfsp, sizeof(*tmpvfsp)); 1291 tmpvfsp->vfc_vfsops = NULL; 1292 tmpvfsp->vfc_next = NULL; 1293 1294 ret = sysctl_rdstruct(oldp, oldlenp, newp, tmpvfsp, 1295 sizeof(struct vfsconf)); 1296 1297 free(tmpvfsp, M_TEMP); 1298 return (ret); 1299 case VFS_BCACHESTAT: /* buffer cache statistics */ 1300 ret = sysctl_rdstruct(oldp, oldlenp, newp, &bcstats, 1301 sizeof(struct bcachestats)); 1302 return(ret); 1303 } 1304 return (EOPNOTSUPP); 1305} 1306 1307int kinfo_vdebug = 1; 1308#define KINFO_VNODESLOP 10 1309/* 1310 * Dump vnode list (via sysctl). 1311 * Copyout address of vnode followed by vnode. 1312 */ 1313/* ARGSUSED */ 1314int 1315sysctl_vnode(char *where, size_t *sizep, struct proc *p) 1316{ 1317 struct mount *mp, *nmp; 1318 struct vnode *vp, *nvp; 1319 char *bp = where, *savebp; 1320 char *ewhere; 1321 int error; 1322 1323 if (where == NULL) { 1324 *sizep = (numvnodes + KINFO_VNODESLOP) * sizeof(struct e_vnode); 1325 return (0); 1326 } 1327 ewhere = where + *sizep; 1328 1329 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1330 mp = nmp) { 1331 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) { 1332 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1333 continue; 1334 } 1335 savebp = bp; 1336again: 1337 for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; 1338 vp = nvp) { 1339 /* 1340 * Check that the vp is still associated with 1341 * this filesystem. RACE: could have been 1342 * recycled onto the same filesystem. 1343 */ 1344 if (vp->v_mount != mp) { 1345 if (kinfo_vdebug) 1346 printf("kinfo: vp changed\n"); 1347 bp = savebp; 1348 goto again; 1349 } 1350 nvp = LIST_NEXT(vp, v_mntvnodes); 1351 if (bp + sizeof(struct e_vnode) > ewhere) { 1352 *sizep = bp - where; 1353 vfs_unbusy(mp); 1354 return (ENOMEM); 1355 } 1356 if ((error = copyout(&vp, 1357 &((struct e_vnode *)bp)->vptr, 1358 sizeof(struct vnode *))) || 1359 (error = copyout(vp, 1360 &((struct e_vnode *)bp)->vnode, 1361 sizeof(struct vnode)))) { 1362 vfs_unbusy(mp); 1363 return (error); 1364 } 1365 bp += sizeof(struct e_vnode); 1366 } 1367 1368 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1369 vfs_unbusy(mp); 1370 } 1371 1372 *sizep = bp - where; 1373 1374 return (0); 1375} 1376 1377/* 1378 * Check to see if a filesystem is mounted on a block device. 1379 */ 1380int 1381vfs_mountedon(struct vnode *vp) 1382{ 1383 struct vnode *vq; 1384 int error = 0; 1385 1386 if (vp->v_specmountpoint != NULL) 1387 return (EBUSY); 1388 if (vp->v_flag & VALIASED) { 1389 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1390 if (vq->v_rdev != vp->v_rdev || 1391 vq->v_type != vp->v_type) 1392 continue; 1393 if (vq->v_specmountpoint != NULL) { 1394 error = EBUSY; 1395 break; 1396 } 1397 } 1398 } 1399 return (error); 1400} 1401 1402/* 1403 * Build hash lists of net addresses and hang them off the mount point. 1404 * Called by ufs_mount() to set up the lists of export addresses. 1405 */ 1406int 1407vfs_hang_addrlist(struct mount *mp, struct netexport *nep, 1408 struct export_args *argp) 1409{ 1410 struct netcred *np; 1411 struct radix_node_head *rnh; 1412 int i; 1413 struct radix_node *rn; 1414 struct sockaddr *saddr, *smask = 0; 1415 struct domain *dom; 1416 int error; 1417 1418 if (argp->ex_addrlen == 0) { 1419 if (mp->mnt_flag & MNT_DEFEXPORTED) 1420 return (EPERM); 1421 np = &nep->ne_defexported; 1422 np->netc_exflags = argp->ex_flags; 1423 np->netc_anon = argp->ex_anon; 1424 np->netc_anon.cr_ref = 1; 1425 mp->mnt_flag |= MNT_DEFEXPORTED; 1426 return (0); 1427 } 1428 if (argp->ex_addrlen > MLEN || argp->ex_masklen > MLEN || 1429 argp->ex_addrlen < 0 || argp->ex_masklen < 0) 1430 return (EINVAL); 1431 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1432 np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK|M_ZERO); 1433 saddr = (struct sockaddr *)(np + 1); 1434 error = copyin(argp->ex_addr, saddr, argp->ex_addrlen); 1435 if (error) 1436 goto out; 1437 if (saddr->sa_len > argp->ex_addrlen) 1438 saddr->sa_len = argp->ex_addrlen; 1439 if (argp->ex_masklen) { 1440 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1441 error = copyin(argp->ex_mask, smask, argp->ex_masklen); 1442 if (error) 1443 goto out; 1444 if (smask->sa_len > argp->ex_masklen) 1445 smask->sa_len = argp->ex_masklen; 1446 } 1447 i = saddr->sa_family; 1448 if (i < 0 || i > AF_MAX) { 1449 error = EINVAL; 1450 goto out; 1451 } 1452 if ((rnh = nep->ne_rtable[i]) == 0) { 1453 /* 1454 * Seems silly to initialize every AF when most are not 1455 * used, do so on demand here 1456 */ 1457 for (dom = domains; dom; dom = dom->dom_next) 1458 if (dom->dom_family == i && dom->dom_rtattach) { 1459 dom->dom_rtattach((void **)&nep->ne_rtable[i], 1460 dom->dom_rtoffset); 1461 break; 1462 } 1463 if ((rnh = nep->ne_rtable[i]) == 0) { 1464 error = ENOBUFS; 1465 goto out; 1466 } 1467 } 1468 rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, 1469 np->netc_rnodes, 0); 1470 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1471 error = EPERM; 1472 goto out; 1473 } 1474 np->netc_exflags = argp->ex_flags; 1475 np->netc_anon = argp->ex_anon; 1476 np->netc_anon.cr_ref = 1; 1477 return (0); 1478out: 1479 free(np, M_NETADDR); 1480 return (error); 1481} 1482 1483/* ARGSUSED */ 1484int 1485vfs_free_netcred(struct radix_node *rn, void *w) 1486{ 1487 struct radix_node_head *rnh = (struct radix_node_head *)w; 1488 1489 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh, NULL); 1490 free(rn, M_NETADDR); 1491 return (0); 1492} 1493 1494/* 1495 * Free the net address hash lists that are hanging off the mount points. 1496 */ 1497void 1498vfs_free_addrlist(struct netexport *nep) 1499{ 1500 int i; 1501 struct radix_node_head *rnh; 1502 1503 for (i = 0; i <= AF_MAX; i++) 1504 if ((rnh = nep->ne_rtable[i]) != NULL) { 1505 (*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh); 1506 free(rnh, M_RTABLE); 1507 nep->ne_rtable[i] = 0; 1508 } 1509} 1510 1511int 1512vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp) 1513{ 1514 int error; 1515 1516 if (argp->ex_flags & MNT_DELEXPORT) { 1517 vfs_free_addrlist(nep); 1518 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1519 } 1520 if (argp->ex_flags & MNT_EXPORTED) { 1521 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1522 return (error); 1523 mp->mnt_flag |= MNT_EXPORTED; 1524 } 1525 return (0); 1526} 1527 1528struct netcred * 1529vfs_export_lookup(struct mount *mp, struct netexport *nep, struct mbuf *nam) 1530{ 1531 struct netcred *np; 1532 struct radix_node_head *rnh; 1533 struct sockaddr *saddr; 1534 1535 np = NULL; 1536 if (mp->mnt_flag & MNT_EXPORTED) { 1537 /* 1538 * Lookup in the export list first. 1539 */ 1540 if (nam != NULL) { 1541 saddr = mtod(nam, struct sockaddr *); 1542 rnh = nep->ne_rtable[saddr->sa_family]; 1543 if (rnh != NULL) { 1544 np = (struct netcred *) 1545 (*rnh->rnh_matchaddr)((caddr_t)saddr, 1546 rnh); 1547 if (np && np->netc_rnodes->rn_flags & RNF_ROOT) 1548 np = NULL; 1549 } 1550 } 1551 /* 1552 * If no address match, use the default if it exists. 1553 */ 1554 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1555 np = &nep->ne_defexported; 1556 } 1557 return (np); 1558} 1559 1560/* 1561 * Do the usual access checking. 1562 * file_mode, uid and gid are from the vnode in question, 1563 * while acc_mode and cred are from the VOP_ACCESS parameter list 1564 */ 1565int 1566vaccess(enum vtype type, mode_t file_mode, uid_t uid, gid_t gid, 1567 mode_t acc_mode, struct ucred *cred) 1568{ 1569 mode_t mask; 1570 1571 /* User id 0 always gets read/write access. */ 1572 if (cred->cr_uid == 0) { 1573 /* For VEXEC, at least one of the execute bits must be set. */ 1574 if ((acc_mode & VEXEC) && type != VDIR && 1575 (file_mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0) 1576 return EACCES; 1577 return 0; 1578 } 1579 1580 mask = 0; 1581 1582 /* Otherwise, check the owner. */ 1583 if (cred->cr_uid == uid) { 1584 if (acc_mode & VEXEC) 1585 mask |= S_IXUSR; 1586 if (acc_mode & VREAD) 1587 mask |= S_IRUSR; 1588 if (acc_mode & VWRITE) 1589 mask |= S_IWUSR; 1590 return (file_mode & mask) == mask ? 0 : EACCES; 1591 } 1592 1593 /* Otherwise, check the groups. */ 1594 if (cred->cr_gid == gid || groupmember(gid, cred)) { 1595 if (acc_mode & VEXEC) 1596 mask |= S_IXGRP; 1597 if (acc_mode & VREAD) 1598 mask |= S_IRGRP; 1599 if (acc_mode & VWRITE) 1600 mask |= S_IWGRP; 1601 return (file_mode & mask) == mask ? 0 : EACCES; 1602 } 1603 1604 /* Otherwise, check everyone else. */ 1605 if (acc_mode & VEXEC) 1606 mask |= S_IXOTH; 1607 if (acc_mode & VREAD) 1608 mask |= S_IROTH; 1609 if (acc_mode & VWRITE) 1610 mask |= S_IWOTH; 1611 return (file_mode & mask) == mask ? 0 : EACCES; 1612} 1613 1614/* 1615 * Unmount all file systems. 1616 * We traverse the list in reverse order under the assumption that doing so 1617 * will avoid needing to worry about dependencies. 1618 */ 1619void 1620vfs_unmountall(void) 1621{ 1622 struct mount *mp, *nmp; 1623 int allerror, error, again = 1; 1624 1625 retry: 1626 allerror = 0; 1627 for (mp = CIRCLEQ_LAST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1628 mp = nmp) { 1629 nmp = CIRCLEQ_PREV(mp, mnt_list); 1630 if ((vfs_busy(mp, VB_WRITE|VB_NOWAIT)) != 0) 1631 continue; 1632 if ((error = dounmount(mp, MNT_FORCE, curproc, NULL)) != 0) { 1633 printf("unmount of %s failed with error %d\n", 1634 mp->mnt_stat.f_mntonname, error); 1635 allerror = 1; 1636 } 1637 } 1638 1639 if (allerror) { 1640 printf("WARNING: some file systems would not unmount\n"); 1641 if (again) { 1642 printf("retrying\n"); 1643 again = 0; 1644 goto retry; 1645 } 1646 } 1647} 1648 1649/* 1650 * Sync and unmount file systems before shutting down. 1651 */ 1652void 1653vfs_shutdown(void) 1654{ 1655#ifdef ACCOUNTING 1656 extern void acct_shutdown(void); 1657 1658 acct_shutdown(); 1659#endif 1660 1661 /* XXX Should suspend scheduling. */ 1662 (void) spl0(); 1663 1664 printf("syncing disks... "); 1665 1666 if (panicstr == 0) { 1667 /* Sync before unmount, in case we hang on something. */ 1668 sys_sync(&proc0, (void *)0, (register_t *)0); 1669 1670 /* Unmount file systems. */ 1671 vfs_unmountall(); 1672 } 1673 1674 if (vfs_syncwait(1)) 1675 printf("giving up\n"); 1676 else 1677 printf("done\n"); 1678} 1679 1680/* 1681 * perform sync() operation and wait for buffers to flush. 1682 * assumptions: called w/ scheduler disabled and physical io enabled 1683 * for now called at spl0() XXX 1684 */ 1685int 1686vfs_syncwait(int verbose) 1687{ 1688 struct buf *bp; 1689 int iter, nbusy, dcount, s; 1690 struct proc *p; 1691 1692 p = curproc? curproc : &proc0; 1693 sys_sync(p, (void *)0, (register_t *)0); 1694 1695 /* Wait for sync to finish. */ 1696 dcount = 10000; 1697 for (iter = 0; iter < 20; iter++) { 1698 nbusy = 0; 1699 LIST_FOREACH(bp, &bufhead, b_list) { 1700 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1701 nbusy++; 1702 /* 1703 * With soft updates, some buffers that are 1704 * written will be remarked as dirty until other 1705 * buffers are written. 1706 */ 1707 if (bp->b_flags & B_DELWRI) { 1708 s = splbio(); 1709 bremfree(bp); 1710 buf_acquire(bp); 1711 splx(s); 1712 nbusy++; 1713 bawrite(bp); 1714 if (dcount-- <= 0) { 1715 if (verbose) 1716 printf("softdep "); 1717 return 1; 1718 } 1719 } 1720 } 1721 if (nbusy == 0) 1722 break; 1723 if (verbose) 1724 printf("%d ", nbusy); 1725 DELAY(40000 * iter); 1726 } 1727 1728 return nbusy; 1729} 1730 1731/* 1732 * posix file system related system variables. 1733 */ 1734int 1735fs_posix_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, 1736 void *newp, size_t newlen, struct proc *p) 1737{ 1738 /* all sysctl names at this level are terminal */ 1739 if (namelen != 1) 1740 return (ENOTDIR); 1741 1742 switch (name[0]) { 1743 case FS_POSIX_SETUID: 1744 if (newp && securelevel > 0) 1745 return (EPERM); 1746 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1747 default: 1748 return (EOPNOTSUPP); 1749 } 1750 /* NOTREACHED */ 1751} 1752 1753/* 1754 * file system related system variables. 1755 */ 1756int 1757fs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1758 size_t newlen, struct proc *p) 1759{ 1760 sysctlfn *fn; 1761 1762 switch (name[0]) { 1763 case FS_POSIX: 1764 fn = fs_posix_sysctl; 1765 break; 1766 default: 1767 return (EOPNOTSUPP); 1768 } 1769 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1770} 1771 1772 1773/* 1774 * Routines dealing with vnodes and buffers 1775 */ 1776 1777/* 1778 * Wait for all outstanding I/Os to complete 1779 * 1780 * Manipulates v_numoutput. Must be called at splbio() 1781 */ 1782int 1783vwaitforio(struct vnode *vp, int slpflag, char *wmesg, int timeo) 1784{ 1785 int error = 0; 1786 1787 splassert(IPL_BIO); 1788 1789 while (vp->v_numoutput) { 1790 vp->v_bioflag |= VBIOWAIT; 1791 error = tsleep(&vp->v_numoutput, 1792 slpflag | (PRIBIO + 1), wmesg, timeo); 1793 if (error) 1794 break; 1795 } 1796 1797 return (error); 1798} 1799 1800/* 1801 * Update outstanding I/O count and do wakeup if requested. 1802 * 1803 * Manipulates v_numoutput. Must be called at splbio() 1804 */ 1805void 1806vwakeup(struct vnode *vp) 1807{ 1808 splassert(IPL_BIO); 1809 1810 if (vp != NULL) { 1811 if (vp->v_numoutput-- == 0) 1812 panic("vwakeup: neg numoutput"); 1813 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1814 vp->v_bioflag &= ~VBIOWAIT; 1815 wakeup(&vp->v_numoutput); 1816 } 1817 } 1818} 1819 1820/* 1821 * Flush out and invalidate all buffers associated with a vnode. 1822 * Called with the underlying object locked. 1823 */ 1824int 1825vinvalbuf(struct vnode *vp, int flags, struct ucred *cred, struct proc *p, 1826 int slpflag, int slptimeo) 1827{ 1828 struct buf *bp; 1829 struct buf *nbp, *blist; 1830 int s, error; 1831 1832#ifdef VFSDEBUG 1833 if ((vp->v_flag & VLOCKSWORK) && !VOP_ISLOCKED(vp)) 1834 panic("vinvalbuf(): vp isn't locked"); 1835#endif 1836 1837 if (flags & V_SAVE) { 1838 s = splbio(); 1839 vwaitforio(vp, 0, "vinvalbuf", 0); 1840 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1841 splx(s); 1842 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1843 return (error); 1844 s = splbio(); 1845 if (vp->v_numoutput > 0 || 1846 !LIST_EMPTY(&vp->v_dirtyblkhd)) 1847 panic("vinvalbuf: dirty bufs"); 1848 } 1849 splx(s); 1850 } 1851loop: 1852 s = splbio(); 1853 for (;;) { 1854 if ((blist = LIST_FIRST(&vp->v_cleanblkhd)) && 1855 (flags & V_SAVEMETA)) 1856 while (blist && blist->b_lblkno < 0) 1857 blist = LIST_NEXT(blist, b_vnbufs); 1858 if (blist == NULL && 1859 (blist = LIST_FIRST(&vp->v_dirtyblkhd)) && 1860 (flags & V_SAVEMETA)) 1861 while (blist && blist->b_lblkno < 0) 1862 blist = LIST_NEXT(blist, b_vnbufs); 1863 if (!blist) 1864 break; 1865 1866 for (bp = blist; bp; bp = nbp) { 1867 nbp = LIST_NEXT(bp, b_vnbufs); 1868 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1869 continue; 1870 if (bp->b_flags & B_BUSY) { 1871 bp->b_flags |= B_WANTED; 1872 error = tsleep(bp, slpflag | (PRIBIO + 1), 1873 "vinvalbuf", slptimeo); 1874 if (error) { 1875 splx(s); 1876 return (error); 1877 } 1878 break; 1879 } 1880 bremfree(bp); 1881 buf_acquire(bp); 1882 /* 1883 * XXX Since there are no node locks for NFS, I believe 1884 * there is a slight chance that a delayed write will 1885 * occur while sleeping just above, so check for it. 1886 */ 1887 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1888 splx(s); 1889 (void) VOP_BWRITE(bp); 1890 goto loop; 1891 } 1892 bp->b_flags |= B_INVAL; 1893 brelse(bp); 1894 } 1895 } 1896 if (!(flags & V_SAVEMETA) && 1897 (!LIST_EMPTY(&vp->v_dirtyblkhd) || !LIST_EMPTY(&vp->v_cleanblkhd))) 1898 panic("vinvalbuf: flush failed"); 1899 splx(s); 1900 return (0); 1901} 1902 1903void 1904vflushbuf(struct vnode *vp, int sync) 1905{ 1906 struct buf *bp, *nbp; 1907 int s; 1908 1909loop: 1910 s = splbio(); 1911 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); 1912 bp != LIST_END(&vp->v_dirtyblkhd); bp = nbp) { 1913 nbp = LIST_NEXT(bp, b_vnbufs); 1914 if ((bp->b_flags & B_BUSY)) 1915 continue; 1916 if ((bp->b_flags & B_DELWRI) == 0) 1917 panic("vflushbuf: not dirty"); 1918 bremfree(bp); 1919 buf_acquire(bp); 1920 splx(s); 1921 /* 1922 * Wait for I/O associated with indirect blocks to complete, 1923 * since there is no way to quickly wait for them below. 1924 */ 1925 if (bp->b_vp == vp || sync == 0) 1926 (void) bawrite(bp); 1927 else 1928 (void) bwrite(bp); 1929 goto loop; 1930 } 1931 if (sync == 0) { 1932 splx(s); 1933 return; 1934 } 1935 vwaitforio(vp, 0, "vflushbuf", 0); 1936 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1937 splx(s); 1938#ifdef DIAGNOSTIC 1939 vprint("vflushbuf: dirty", vp); 1940#endif 1941 goto loop; 1942 } 1943 splx(s); 1944} 1945 1946/* 1947 * Associate a buffer with a vnode. 1948 * 1949 * Manipulates buffer vnode queues. Must be called at splbio(). 1950 */ 1951void 1952bgetvp(struct vnode *vp, struct buf *bp) 1953{ 1954 splassert(IPL_BIO); 1955 1956 1957 if (bp->b_vp) 1958 panic("bgetvp: not free"); 1959 vhold(vp); 1960 bp->b_vp = vp; 1961 if (vp->v_type == VBLK || vp->v_type == VCHR) 1962 bp->b_dev = vp->v_rdev; 1963 else 1964 bp->b_dev = NODEV; 1965 /* 1966 * Insert onto list for new vnode. 1967 */ 1968 bufinsvn(bp, &vp->v_cleanblkhd); 1969} 1970 1971/* 1972 * Disassociate a buffer from a vnode. 1973 * 1974 * Manipulates vnode buffer queues. Must be called at splbio(). 1975 */ 1976void 1977brelvp(struct buf *bp) 1978{ 1979 struct vnode *vp; 1980 1981 splassert(IPL_BIO); 1982 1983 if ((vp = bp->b_vp) == (struct vnode *) 0) 1984 panic("brelvp: NULL"); 1985 /* 1986 * Delete from old vnode list, if on one. 1987 */ 1988 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 1989 bufremvn(bp); 1990 if ((vp->v_bioflag & VBIOONSYNCLIST) && 1991 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 1992 vp->v_bioflag &= ~VBIOONSYNCLIST; 1993 LIST_REMOVE(vp, v_synclist); 1994 } 1995 bp->b_vp = NULL; 1996 1997 vdrop(vp); 1998} 1999 2000/* 2001 * Replaces the current vnode associated with the buffer, if any, 2002 * with a new vnode. 2003 * 2004 * If an output I/O is pending on the buffer, the old vnode 2005 * I/O count is adjusted. 2006 * 2007 * Ignores vnode buffer queues. Must be called at splbio(). 2008 */ 2009void 2010buf_replacevnode(struct buf *bp, struct vnode *newvp) 2011{ 2012 struct vnode *oldvp = bp->b_vp; 2013 2014 splassert(IPL_BIO); 2015 2016 if (oldvp) 2017 brelvp(bp); 2018 2019 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2020 newvp->v_numoutput++; /* put it on swapdev */ 2021 vwakeup(oldvp); 2022 } 2023 2024 bgetvp(newvp, bp); 2025 bufremvn(bp); 2026} 2027 2028/* 2029 * Used to assign buffers to the appropriate clean or dirty list on 2030 * the vnode and to add newly dirty vnodes to the appropriate 2031 * filesystem syncer list. 2032 * 2033 * Manipulates vnode buffer queues. Must be called at splbio(). 2034 */ 2035void 2036reassignbuf(struct buf *bp) 2037{ 2038 struct buflists *listheadp; 2039 int delay; 2040 struct vnode *vp = bp->b_vp; 2041 2042 splassert(IPL_BIO); 2043 2044 /* 2045 * Delete from old vnode list, if on one. 2046 */ 2047 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2048 bufremvn(bp); 2049 2050 /* 2051 * If dirty, put on list of dirty buffers; 2052 * otherwise insert onto list of clean buffers. 2053 */ 2054 if ((bp->b_flags & B_DELWRI) == 0) { 2055 listheadp = &vp->v_cleanblkhd; 2056 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2057 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2058 vp->v_bioflag &= ~VBIOONSYNCLIST; 2059 LIST_REMOVE(vp, v_synclist); 2060 } 2061 } else { 2062 listheadp = &vp->v_dirtyblkhd; 2063 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2064 switch (vp->v_type) { 2065 case VDIR: 2066 delay = syncdelay / 2; 2067 break; 2068 case VBLK: 2069 if (vp->v_specmountpoint != NULL) { 2070 delay = syncdelay / 3; 2071 break; 2072 } 2073 /* FALLTHROUGH */ 2074 default: 2075 delay = syncdelay; 2076 } 2077 vn_syncer_add_to_worklist(vp, delay); 2078 } 2079 } 2080 bufinsvn(bp, listheadp); 2081} 2082 2083int 2084vfs_register(struct vfsconf *vfs) 2085{ 2086 struct vfsconf *vfsp; 2087 struct vfsconf **vfspp; 2088 2089#ifdef DIAGNOSTIC 2090 /* Paranoia? */ 2091 if (vfs->vfc_refcount != 0) 2092 printf("vfs_register called with vfc_refcount > 0\n"); 2093#endif 2094 2095 /* Check if filesystem already known */ 2096 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2097 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2098 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2099 return (EEXIST); 2100 2101 if (vfs->vfc_typenum > maxvfsconf) 2102 maxvfsconf = vfs->vfc_typenum; 2103 2104 vfs->vfc_next = NULL; 2105 2106 /* Add to the end of the list */ 2107 *vfspp = vfs; 2108 2109 /* Call vfs_init() */ 2110 if (vfs->vfc_vfsops->vfs_init) 2111 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2112 2113 return 0; 2114} 2115 2116int 2117vfs_unregister(struct vfsconf *vfs) 2118{ 2119 struct vfsconf *vfsp; 2120 struct vfsconf **vfspp; 2121 int maxtypenum; 2122 2123 /* Find our vfsconf struct */ 2124 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2125 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2126 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2127 break; 2128 } 2129 2130 if (!vfsp) /* Not found */ 2131 return (ENOENT); 2132 2133 if (vfsp->vfc_refcount) /* In use */ 2134 return (EBUSY); 2135 2136 /* Remove from list and free */ 2137 *vfspp = vfsp->vfc_next; 2138 2139 maxtypenum = 0; 2140 2141 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2142 if (vfsp->vfc_typenum > maxtypenum) 2143 maxtypenum = vfsp->vfc_typenum; 2144 2145 maxvfsconf = maxtypenum; 2146 return 0; 2147} 2148 2149/* 2150 * Check if vnode represents a disk device 2151 */ 2152int 2153vn_isdisk(struct vnode *vp, int *errp) 2154{ 2155 if (vp->v_type != VBLK && vp->v_type != VCHR) 2156 return (0); 2157 2158 return (1); 2159} 2160 2161#ifdef DDB 2162#include <machine/db_machdep.h> 2163#include <ddb/db_interface.h> 2164#include <ddb/db_output.h> 2165 2166void 2167vfs_buf_print(void *b, int full, int (*pr)(const char *, ...)) 2168{ 2169 struct buf *bp = b; 2170 2171 (*pr)(" vp %p lblkno 0x%llx blkno 0x%llx dev 0x%x\n" 2172 " proc %p error %d flags %b\n", 2173 bp->b_vp, (int64_t)bp->b_lblkno, (int64_t)bp->b_blkno, bp->b_dev, 2174 bp->b_proc, bp->b_error, bp->b_flags, B_BITS); 2175 2176 (*pr)(" bufsize 0x%lx bcount 0x%lx resid 0x%lx sync 0x%x\n" 2177 " data %p saveaddr %p dep %p iodone %p\n", 2178 bp->b_bufsize, bp->b_bcount, (long)bp->b_resid, bp->b_synctime, 2179 bp->b_data, bp->b_saveaddr, LIST_FIRST(&bp->b_dep), bp->b_iodone); 2180 2181 (*pr)(" dirty {off 0x%x end 0x%x} valid {off 0x%x end 0x%x}\n", 2182 bp->b_dirtyoff, bp->b_dirtyend, bp->b_validoff, bp->b_validend); 2183 2184#ifdef FFS_SOFTUPDATES 2185 if (full) 2186 softdep_print(bp, full, pr); 2187#endif 2188} 2189 2190const char *vtypes[] = { VTYPE_NAMES }; 2191const char *vtags[] = { VTAG_NAMES }; 2192 2193void 2194vfs_vnode_print(void *v, int full, int (*pr)(const char *, ...)) 2195{ 2196 struct vnode *vp = v; 2197 2198#define NENTS(n) (sizeof n / sizeof(n[0])) 2199 (*pr)("tag %s(%d) type %s(%d) mount %p typedata %p\n", 2200 vp->v_tag > NENTS(vtags)? "<unk>":vtags[vp->v_tag], vp->v_tag, 2201 vp->v_type > NENTS(vtypes)? "<unk>":vtypes[vp->v_type], 2202 vp->v_type, vp->v_mount, vp->v_mountedhere); 2203 2204 (*pr)("data %p usecount %d writecount %ld holdcnt %ld numoutput %d\n", 2205 vp->v_data, vp->v_usecount, vp->v_writecount, 2206 vp->v_holdcnt, vp->v_numoutput); 2207 2208 /* uvm_object_printit(&vp->v_uobj, full, pr); */ 2209 2210 if (full) { 2211 struct buf *bp; 2212 2213 (*pr)("clean bufs:\n"); 2214 LIST_FOREACH(bp, &vp->v_cleanblkhd, b_vnbufs) { 2215 (*pr)(" bp %p\n", bp); 2216 vfs_buf_print(bp, full, pr); 2217 } 2218 2219 (*pr)("dirty bufs:\n"); 2220 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) { 2221 (*pr)(" bp %p\n", bp); 2222 vfs_buf_print(bp, full, pr); 2223 } 2224 } 2225} 2226 2227void 2228vfs_mount_print(struct mount *mp, int full, int (*pr)(const char *, ...)) 2229{ 2230 struct vfsconf *vfc = mp->mnt_vfc; 2231 struct vnode *vp; 2232 int cnt = 0; 2233 2234 (*pr)("flags %b\nvnodecovered %p syncer %p data %p\n", 2235 mp->mnt_flag, MNT_BITS, 2236 mp->mnt_vnodecovered, mp->mnt_syncer, mp->mnt_data); 2237 2238 (*pr)("vfsconf: ops %p name \"%s\" num %d ref %d flags 0x%x\n", 2239 vfc->vfc_vfsops, vfc->vfc_name, vfc->vfc_typenum, 2240 vfc->vfc_refcount, vfc->vfc_flags); 2241 2242 (*pr)("statvfs cache: bsize %x iosize %x\nblocks %llu free %llu avail %lld\n", 2243 mp->mnt_stat.f_bsize, mp->mnt_stat.f_iosize, mp->mnt_stat.f_blocks, 2244 mp->mnt_stat.f_bfree, mp->mnt_stat.f_bavail); 2245 2246 (*pr)(" files %llu ffiles %llu favail $lld\n", mp->mnt_stat.f_files, 2247 mp->mnt_stat.f_ffree, mp->mnt_stat.f_favail); 2248 2249 (*pr)(" f_fsidx {0x%x, 0x%x} owner %u ctime 0x%x\n", 2250 mp->mnt_stat.f_fsid.val[0], mp->mnt_stat.f_fsid.val[1], 2251 mp->mnt_stat.f_owner, mp->mnt_stat.f_ctime); 2252 2253 (*pr)(" syncwrites %llu asyncwrites = %llu\n", 2254 mp->mnt_stat.f_syncwrites, mp->mnt_stat.f_asyncwrites); 2255 2256 (*pr)(" syncreads %llu asyncreads = %llu\n", 2257 mp->mnt_stat.f_syncreads, mp->mnt_stat.f_asyncreads); 2258 2259 (*pr)(" fstype \"%s\" mnton \"%s\" mntfrom \"%s\"\n", 2260 mp->mnt_stat.f_fstypename, mp->mnt_stat.f_mntonname, 2261 mp->mnt_stat.f_mntfromname); 2262 2263 (*pr)("locked vnodes:"); 2264 /* XXX would take mountlist lock, except ddb has no context */ 2265 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2266 if (VOP_ISLOCKED(vp)) { 2267 if (!LIST_NEXT(vp, v_mntvnodes)) 2268 (*pr)(" %p", vp); 2269 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2270 (*pr)("\n\t%p", vp); 2271 else 2272 (*pr)(", %p", vp); 2273 } 2274 (*pr)("\n"); 2275 2276 if (full) { 2277 (*pr)("all vnodes:\n\t"); 2278 /* XXX would take mountlist lock, except ddb has no context */ 2279 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2280 if (!LIST_NEXT(vp, v_mntvnodes)) 2281 (*pr)(" %p", vp); 2282 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2283 (*pr)(" %p,\n\t", vp); 2284 else 2285 (*pr)(" %p,", vp); 2286 (*pr)("\n"); 2287 } 2288} 2289#endif /* DDB */ 2290 2291void 2292copy_statfs_info(struct statfs *sbp, const struct mount *mp) 2293{ 2294 const struct statfs *mbp; 2295 2296 strncpy(sbp->f_fstypename, mp->mnt_vfc->vfc_name, MFSNAMELEN); 2297 2298 if (sbp == (mbp = &mp->mnt_stat)) 2299 return; 2300 2301 sbp->f_fsid = mbp->f_fsid; 2302 sbp->f_owner = mbp->f_owner; 2303 sbp->f_flags = mbp->f_flags; 2304 sbp->f_syncwrites = mbp->f_syncwrites; 2305 sbp->f_asyncwrites = mbp->f_asyncwrites; 2306 sbp->f_syncreads = mbp->f_syncreads; 2307 sbp->f_asyncreads = mbp->f_asyncreads; 2308 sbp->f_namemax = mbp->f_namemax; 2309 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN); 2310 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); 2311 bcopy(&mp->mnt_stat.mount_info.ufs_args, &sbp->mount_info.ufs_args, 2312 sizeof(struct ufs_args)); 2313} 2314