1 /* $OpenBSD: vfs_subr.c,v 1.159 2007/11/15 16:50:28 deraadt Exp $ */ 2 /* $NetBSD: vfs_subr.c,v 1.53 1996/04/22 01:39:13 christos Exp $ */ 3 4 /* 5 * Copyright (c) 1989, 1993 6 * The Regents of the University of California. All rights reserved. 7 * (c) UNIX System Laboratories, Inc. 8 * All or some portions of this file are derived from material licensed 9 * to the University of California by American Telephone and Telegraph 10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 11 * the permission of UNIX System Laboratories, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 * 37 * @(#)vfs_subr.c 8.13 (Berkeley) 4/18/94 38 */ 39 40 /* 41 * External virtual filesystem routines 42 */ 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/proc.h> 47 #include <sys/mount.h> 48 #include <sys/time.h> 49 #include <sys/fcntl.h> 50 #include <sys/kernel.h> 51 #include <sys/vnode.h> 52 #include <sys/stat.h> 53 #include <sys/namei.h> 54 #include <sys/ucred.h> 55 #include <sys/buf.h> 56 #include <sys/errno.h> 57 #include <sys/malloc.h> 58 #include <sys/domain.h> 59 #include <sys/mbuf.h> 60 #include <sys/syscallargs.h> 61 #include <sys/pool.h> 62 63 #include <uvm/uvm_extern.h> 64 #include <sys/sysctl.h> 65 66 #include <miscfs/specfs/specdev.h> 67 68 enum vtype iftovt_tab[16] = { 69 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, 70 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, 71 }; 72 73 int vttoif_tab[9] = { 74 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, 75 S_IFSOCK, S_IFIFO, S_IFMT, 76 }; 77 78 int doforce = 1; /* 1 => permit forcible unmounting */ 79 int prtactive = 0; /* 1 => print out reclaim of active vnodes */ 80 int suid_clear = 1; /* 1 => clear SUID / SGID on owner change */ 81 82 /* 83 * Insq/Remq for the vnode usage lists. 84 */ 85 #define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) 86 #define bufremvn(bp) { \ 87 LIST_REMOVE(bp, b_vnbufs); \ 88 LIST_NEXT(bp, b_vnbufs) = NOLIST; \ 89 } 90 91 struct freelst vnode_hold_list; /* list of vnodes referencing buffers */ 92 struct freelst vnode_free_list; /* vnode free list */ 93 94 struct mntlist mountlist; /* mounted filesystem list */ 95 96 void vclean(struct vnode *, int, struct proc *); 97 98 void insmntque(struct vnode *, struct mount *); 99 int getdevvp(dev_t, struct vnode **, enum vtype); 100 101 int vfs_hang_addrlist(struct mount *, struct netexport *, 102 struct export_args *); 103 int vfs_free_netcred(struct radix_node *, void *); 104 void vfs_free_addrlist(struct netexport *); 105 void vputonfreelist(struct vnode *); 106 107 int vflush_vnode(struct vnode *, void *); 108 int maxvnodes; 109 110 #ifdef DEBUG 111 void printlockedvnodes(void); 112 #endif 113 114 struct pool vnode_pool; 115 116 /* 117 * Initialize the vnode management data structures. 118 */ 119 void 120 vntblinit(void) 121 { 122 /* buffer cache may need a vnode for each buffer */ 123 maxvnodes = desiredvnodes; 124 pool_init(&vnode_pool, sizeof(struct vnode), 0, 0, 0, "vnodes", 125 &pool_allocator_nointr); 126 TAILQ_INIT(&vnode_hold_list); 127 TAILQ_INIT(&vnode_free_list); 128 CIRCLEQ_INIT(&mountlist); 129 /* 130 * Initialize the filesystem syncer. 131 */ 132 vn_initialize_syncerd(); 133 } 134 135 /* 136 * Mark a mount point as busy. Used to synchronize access and to delay 137 * unmounting. 138 * 139 * Default behaviour is to attempt getting a READ lock and in case of an 140 * ongoing unmount, to wait for it to finish and then return failure. 141 */ 142 int 143 vfs_busy(struct mount *mp, int flags) 144 { 145 int rwflags = 0; 146 147 /* new mountpoints need their lock initialised */ 148 if (mp->mnt_lock.rwl_name == NULL) 149 rw_init(&mp->mnt_lock, "vfslock"); 150 151 if (flags & VB_WRITE) 152 rwflags |= RW_WRITE; 153 else 154 rwflags |= RW_READ; 155 156 if (flags & VB_WAIT) 157 rwflags |= RW_SLEEPFAIL; 158 else 159 rwflags |= RW_NOSLEEP; 160 161 if (rw_enter(&mp->mnt_lock, rwflags)) 162 return (EBUSY); 163 164 return (0); 165 } 166 167 /* 168 * Free a busy file system 169 */ 170 void 171 vfs_unbusy(struct mount *mp) 172 { 173 rw_exit(&mp->mnt_lock); 174 } 175 176 int 177 vfs_isbusy(struct mount *mp) 178 { 179 if (RWLOCK_OWNER(&mp->mnt_lock) > 0) 180 return (1); 181 else 182 return (0); 183 } 184 185 /* 186 * Lookup a filesystem type, and if found allocate and initialize 187 * a mount structure for it. 188 * 189 * Devname is usually updated by mount(8) after booting. 190 */ 191 int 192 vfs_rootmountalloc(char *fstypename, char *devname, struct mount **mpp) 193 { 194 struct vfsconf *vfsp; 195 struct mount *mp; 196 197 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 198 if (!strcmp(vfsp->vfc_name, fstypename)) 199 break; 200 if (vfsp == NULL) 201 return (ENODEV); 202 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK|M_ZERO); 203 (void)vfs_busy(mp, VB_READ|VB_NOWAIT); 204 LIST_INIT(&mp->mnt_vnodelist); 205 mp->mnt_vfc = vfsp; 206 mp->mnt_op = vfsp->vfc_vfsops; 207 mp->mnt_flag = MNT_RDONLY; 208 mp->mnt_vnodecovered = NULLVP; 209 vfsp->vfc_refcount++; 210 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 211 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 212 mp->mnt_stat.f_mntonname[0] = '/'; 213 (void)copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); 214 *mpp = mp; 215 return (0); 216 } 217 218 /* 219 * Find an appropriate filesystem to use for the root. If a filesystem 220 * has not been preselected, walk through the list of known filesystems 221 * trying those that have mountroot routines, and try them until one 222 * works or we have tried them all. 223 */ 224 int 225 vfs_mountroot(void) 226 { 227 struct vfsconf *vfsp; 228 int error; 229 230 if (mountroot != NULL) 231 return ((*mountroot)()); 232 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) { 233 if (vfsp->vfc_mountroot == NULL) 234 continue; 235 if ((error = (*vfsp->vfc_mountroot)()) == 0) 236 return (0); 237 printf("%s_mountroot failed: %d\n", vfsp->vfc_name, error); 238 } 239 return (ENODEV); 240 } 241 242 /* 243 * Lookup a mount point by filesystem identifier. 244 */ 245 struct mount * 246 vfs_getvfs(fsid_t *fsid) 247 { 248 struct mount *mp; 249 250 CIRCLEQ_FOREACH(mp, &mountlist, mnt_list) { 251 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 252 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 253 return (mp); 254 } 255 } 256 257 return (NULL); 258 } 259 260 261 /* 262 * Get a new unique fsid 263 */ 264 void 265 vfs_getnewfsid(struct mount *mp) 266 { 267 static u_short xxxfs_mntid; 268 269 fsid_t tfsid; 270 int mtype; 271 272 mtype = mp->mnt_vfc->vfc_typenum; 273 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 274 mp->mnt_stat.f_fsid.val[1] = mtype; 275 if (xxxfs_mntid == 0) 276 ++xxxfs_mntid; 277 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 278 tfsid.val[1] = mtype; 279 if (!CIRCLEQ_EMPTY(&mountlist)) { 280 while (vfs_getvfs(&tfsid)) { 281 tfsid.val[0]++; 282 xxxfs_mntid++; 283 } 284 } 285 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 286 } 287 288 /* 289 * Make a 'unique' number from a mount type name. 290 * Note that this is no longer used for ffs which 291 * now has an on-disk filesystem id. 292 */ 293 long 294 makefstype(char *type) 295 { 296 long rv; 297 298 for (rv = 0; *type; type++) { 299 rv <<= 2; 300 rv ^= *type; 301 } 302 return rv; 303 } 304 305 /* 306 * Set vnode attributes to VNOVAL 307 */ 308 void 309 vattr_null(struct vattr *vap) 310 { 311 312 vap->va_type = VNON; 313 /* XXX These next two used to be one line, but for a GCC bug. */ 314 vap->va_size = VNOVAL; 315 vap->va_bytes = VNOVAL; 316 vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = 317 vap->va_fsid = vap->va_fileid = 318 vap->va_blocksize = vap->va_rdev = 319 vap->va_atime.tv_sec = vap->va_atime.tv_nsec = 320 vap->va_mtime.tv_sec = vap->va_mtime.tv_nsec = 321 vap->va_ctime.tv_sec = vap->va_ctime.tv_nsec = 322 vap->va_flags = vap->va_gen = VNOVAL; 323 vap->va_vaflags = 0; 324 } 325 326 /* 327 * Routines having to do with the management of the vnode table. 328 */ 329 extern int (**dead_vnodeop_p)(void *); 330 long numvnodes; 331 332 /* 333 * Return the next vnode from the free list. 334 */ 335 int 336 getnewvnode(enum vtagtype tag, struct mount *mp, int (**vops)(void *), 337 struct vnode **vpp) 338 { 339 struct proc *p = curproc; 340 struct freelst *listhd; 341 static int toggle; 342 struct vnode *vp; 343 int s; 344 345 /* 346 * We must choose whether to allocate a new vnode or recycle an 347 * existing one. The criterion for allocating a new one is that 348 * the total number of vnodes is less than the number desired or 349 * there are no vnodes on either free list. Generally we only 350 * want to recycle vnodes that have no buffers associated with 351 * them, so we look first on the vnode_free_list. If it is empty, 352 * we next consider vnodes with referencing buffers on the 353 * vnode_hold_list. The toggle ensures that half the time we 354 * will use a buffer from the vnode_hold_list, and half the time 355 * we will allocate a new one unless the list has grown to twice 356 * the desired size. We are reticent to recycle vnodes from the 357 * vnode_hold_list because we will lose the identity of all its 358 * referencing buffers. 359 */ 360 toggle ^= 1; 361 if (numvnodes > 2 * maxvnodes) 362 toggle = 0; 363 364 s = splbio(); 365 if ((numvnodes < maxvnodes) || 366 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 367 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 368 splx(s); 369 vp = pool_get(&vnode_pool, PR_WAITOK); 370 bzero((char *)vp, sizeof *vp); 371 numvnodes++; 372 } else { 373 for (vp = TAILQ_FIRST(listhd); vp != NULLVP; 374 vp = TAILQ_NEXT(vp, v_freelist)) { 375 if (VOP_ISLOCKED(vp) == 0) 376 break; 377 } 378 /* 379 * Unless this is a bad time of the month, at most 380 * the first NCPUS items on the free list are 381 * locked, so this is close enough to being empty. 382 */ 383 if (vp == NULL) { 384 splx(s); 385 tablefull("vnode"); 386 *vpp = 0; 387 return (ENFILE); 388 } 389 390 #ifdef DIAGNOSTIC 391 if (vp->v_usecount) { 392 vprint("free vnode", vp); 393 panic("free vnode isn't"); 394 } 395 #endif 396 397 TAILQ_REMOVE(listhd, vp, v_freelist); 398 vp->v_bioflag &= ~VBIOONFREELIST; 399 splx(s); 400 401 if (vp->v_type != VBAD) 402 vgonel(vp, p); 403 #ifdef DIAGNOSTIC 404 if (vp->v_data) { 405 vprint("cleaned vnode", vp); 406 panic("cleaned vnode isn't"); 407 } 408 s = splbio(); 409 if (vp->v_numoutput) 410 panic("Clean vnode has pending I/O's"); 411 splx(s); 412 #endif 413 vp->v_flag = 0; 414 vp->v_socket = 0; 415 } 416 vp->v_type = VNON; 417 cache_purge(vp); 418 vp->v_tag = tag; 419 vp->v_op = vops; 420 insmntque(vp, mp); 421 *vpp = vp; 422 vp->v_usecount = 1; 423 vp->v_data = 0; 424 simple_lock_init(&vp->v_uvm.u_obj.vmobjlock); 425 return (0); 426 } 427 428 /* 429 * Move a vnode from one mount queue to another. 430 */ 431 void 432 insmntque(struct vnode *vp, struct mount *mp) 433 { 434 /* 435 * Delete from old mount point vnode list, if on one. 436 */ 437 if (vp->v_mount != NULL) 438 LIST_REMOVE(vp, v_mntvnodes); 439 /* 440 * Insert into list of vnodes for the new mount point, if available. 441 */ 442 if ((vp->v_mount = mp) != NULL) 443 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 444 } 445 446 /* 447 * Create a vnode for a block device. 448 * Used for root filesystem, argdev, and swap areas. 449 * Also used for memory file system special devices. 450 */ 451 int 452 bdevvp(dev_t dev, struct vnode **vpp) 453 { 454 return (getdevvp(dev, vpp, VBLK)); 455 } 456 457 /* 458 * Create a vnode for a character device. 459 * Used for console handling. 460 */ 461 int 462 cdevvp(dev_t dev, struct vnode **vpp) 463 { 464 return (getdevvp(dev, vpp, VCHR)); 465 } 466 467 /* 468 * Create a vnode for a device. 469 * Used by bdevvp (block device) for root file system etc., 470 * and by cdevvp (character device) for console. 471 */ 472 int 473 getdevvp(dev_t dev, struct vnode **vpp, enum vtype type) 474 { 475 struct vnode *vp; 476 struct vnode *nvp; 477 int error; 478 479 if (dev == NODEV) { 480 *vpp = NULLVP; 481 return (0); 482 } 483 error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); 484 if (error) { 485 *vpp = NULLVP; 486 return (error); 487 } 488 vp = nvp; 489 vp->v_type = type; 490 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 491 vput(vp); 492 vp = nvp; 493 } 494 *vpp = vp; 495 return (0); 496 } 497 498 /* 499 * Check to see if the new vnode represents a special device 500 * for which we already have a vnode (either because of 501 * bdevvp() or because of a different vnode representing 502 * the same block device). If such an alias exists, deallocate 503 * the existing contents and return the aliased vnode. The 504 * caller is responsible for filling it with its new contents. 505 */ 506 struct vnode * 507 checkalias(struct vnode *nvp, dev_t nvp_rdev, struct mount *mp) 508 { 509 struct proc *p = curproc; 510 struct vnode *vp; 511 struct vnode **vpp; 512 513 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 514 return (NULLVP); 515 516 vpp = &speclisth[SPECHASH(nvp_rdev)]; 517 loop: 518 for (vp = *vpp; vp; vp = vp->v_specnext) { 519 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 520 continue; 521 } 522 /* 523 * Alias, but not in use, so flush it out. 524 */ 525 if (vp->v_usecount == 0) { 526 vgonel(vp, p); 527 goto loop; 528 } 529 if (vget(vp, LK_EXCLUSIVE, p)) { 530 goto loop; 531 } 532 break; 533 } 534 535 /* 536 * Common case is actually in the if statement 537 */ 538 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 539 nvp->v_specinfo = malloc(sizeof(struct specinfo), M_VNODE, 540 M_WAITOK); 541 nvp->v_rdev = nvp_rdev; 542 nvp->v_hashchain = vpp; 543 nvp->v_specnext = *vpp; 544 nvp->v_specmountpoint = NULL; 545 nvp->v_speclockf = NULL; 546 bzero(nvp->v_specbitmap, sizeof(nvp->v_specbitmap)); 547 *vpp = nvp; 548 if (vp != NULLVP) { 549 nvp->v_flag |= VALIASED; 550 vp->v_flag |= VALIASED; 551 vput(vp); 552 } 553 return (NULLVP); 554 } 555 556 /* 557 * This code is the uncommon case. It is called in case 558 * we found an alias that was VT_NON && vtype of VBLK 559 * This means we found a block device that was created 560 * using bdevvp. 561 * An example of such a vnode is the root partition device vnode 562 * created in ffs_mountroot. 563 * 564 * The vnodes created by bdevvp should not be aliased (why?). 565 */ 566 567 VOP_UNLOCK(vp, 0, p); 568 vclean(vp, 0, p); 569 vp->v_op = nvp->v_op; 570 vp->v_tag = nvp->v_tag; 571 nvp->v_type = VNON; 572 insmntque(vp, mp); 573 return (vp); 574 } 575 576 /* 577 * Grab a particular vnode from the free list, increment its 578 * reference count and lock it. If the vnode lock bit is set, 579 * the vnode is being eliminated in vgone. In that case, we 580 * cannot grab it, so the process is awakened when the 581 * transition is completed, and an error code is returned to 582 * indicate that the vnode is no longer usable, possibly 583 * having been changed to a new file system type. 584 */ 585 int 586 vget(struct vnode *vp, int flags, struct proc *p) 587 { 588 int error, s, onfreelist; 589 590 /* 591 * If the vnode is in the process of being cleaned out for 592 * another use, we wait for the cleaning to finish and then 593 * return failure. Cleaning is determined by checking that 594 * the VXLOCK flag is set. 595 */ 596 597 if (vp->v_flag & VXLOCK) { 598 if (flags & LK_NOWAIT) { 599 return (EBUSY); 600 } 601 602 vp->v_flag |= VXWANT; 603 ltsleep(vp, PINOD | PNORELOCK, "vget", 0, NULL); 604 return (ENOENT); 605 } 606 607 onfreelist = vp->v_bioflag & VBIOONFREELIST; 608 if (vp->v_usecount == 0 && onfreelist) { 609 s = splbio(); 610 if (vp->v_holdcnt > 0) 611 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 612 else 613 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 614 vp->v_bioflag &= ~VBIOONFREELIST; 615 splx(s); 616 } 617 618 vp->v_usecount++; 619 if (flags & LK_TYPE_MASK) { 620 if ((error = vn_lock(vp, flags, p)) != 0) { 621 vp->v_usecount--; 622 if (vp->v_usecount == 0 && onfreelist) 623 vputonfreelist(vp); 624 } 625 return (error); 626 } 627 628 return (0); 629 } 630 631 632 /* Vnode reference. */ 633 void 634 vref(struct vnode *vp) 635 { 636 #ifdef DIAGNOSTIC 637 if (vp->v_usecount == 0) 638 panic("vref used where vget required"); 639 #endif 640 vp->v_usecount++; 641 } 642 643 void 644 vputonfreelist(struct vnode *vp) 645 { 646 int s; 647 struct freelst *lst; 648 649 s = splbio(); 650 #ifdef DIAGNOSTIC 651 if (vp->v_usecount != 0) 652 panic("Use count is not zero!"); 653 654 if (vp->v_bioflag & VBIOONFREELIST) { 655 vprint("vnode already on free list: ", vp); 656 panic("vnode already on free list"); 657 } 658 #endif 659 660 vp->v_bioflag |= VBIOONFREELIST; 661 662 if (vp->v_holdcnt > 0) 663 lst = &vnode_hold_list; 664 else 665 lst = &vnode_free_list; 666 667 if (vp->v_type == VBAD) 668 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 669 else 670 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 671 672 splx(s); 673 } 674 675 /* 676 * vput(), just unlock and vrele() 677 */ 678 void 679 vput(struct vnode *vp) 680 { 681 struct proc *p = curproc; 682 683 #ifdef DIAGNOSTIC 684 if (vp == NULL) 685 panic("vput: null vp"); 686 #endif 687 688 #ifdef DIAGNOSTIC 689 if (vp->v_usecount == 0) { 690 vprint("vput: bad ref count", vp); 691 panic("vput: ref cnt"); 692 } 693 #endif 694 vp->v_usecount--; 695 if (vp->v_usecount > 0) { 696 VOP_UNLOCK(vp, 0, p); 697 return; 698 } 699 700 #ifdef DIAGNOSTIC 701 if (vp->v_writecount != 0) { 702 vprint("vput: bad writecount", vp); 703 panic("vput: v_writecount != 0"); 704 } 705 #endif 706 707 VOP_INACTIVE(vp, p); 708 709 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 710 vputonfreelist(vp); 711 } 712 713 /* 714 * Vnode release - use for active VNODES. 715 * If count drops to zero, call inactive routine and return to freelist. 716 */ 717 void 718 vrele(struct vnode *vp) 719 { 720 struct proc *p = curproc; 721 722 #ifdef DIAGNOSTIC 723 if (vp == NULL) 724 panic("vrele: null vp"); 725 #endif 726 #ifdef DIAGNOSTIC 727 if (vp->v_usecount == 0) { 728 vprint("vrele: bad ref count", vp); 729 panic("vrele: ref cnt"); 730 } 731 #endif 732 vp->v_usecount--; 733 if (vp->v_usecount > 0) { 734 return; 735 } 736 737 #ifdef DIAGNOSTIC 738 if (vp->v_writecount != 0) { 739 vprint("vrele: bad writecount", vp); 740 panic("vrele: v_writecount != 0"); 741 } 742 #endif 743 744 if (vn_lock(vp, LK_EXCLUSIVE, p)) { 745 #ifdef DIAGNOSTIC 746 vprint("vrele: cannot lock", vp); 747 #endif 748 return; 749 } 750 751 VOP_INACTIVE(vp, p); 752 753 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 754 vputonfreelist(vp); 755 } 756 757 void vhold(struct vnode *vp); 758 759 /* 760 * Page or buffer structure gets a reference. 761 */ 762 void 763 vhold(struct vnode *vp) 764 { 765 /* 766 * If it is on the freelist and the hold count is currently 767 * zero, move it to the hold list. 768 */ 769 if ((vp->v_bioflag & VBIOONFREELIST) && 770 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 771 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 772 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 773 } 774 vp->v_holdcnt++; 775 } 776 777 /* 778 * Remove any vnodes in the vnode table belonging to mount point mp. 779 * 780 * If MNT_NOFORCE is specified, there should not be any active ones, 781 * return error if any are found (nb: this is a user error, not a 782 * system error). If MNT_FORCE is specified, detach any active vnodes 783 * that are found. 784 */ 785 #ifdef DEBUG 786 int busyprt = 0; /* print out busy vnodes */ 787 struct ctldebug debug1 = { "busyprt", &busyprt }; 788 #endif 789 790 int 791 vfs_mount_foreach_vnode(struct mount *mp, 792 int (*func)(struct vnode *, void *), void *arg) { 793 struct vnode *vp, *nvp; 794 int error = 0; 795 796 loop: 797 for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; vp = nvp) { 798 if (vp->v_mount != mp) 799 goto loop; 800 nvp = LIST_NEXT(vp, v_mntvnodes); 801 802 error = func(vp, arg); 803 804 if (error != 0) 805 break; 806 } 807 808 return (error); 809 } 810 811 struct vflush_args { 812 struct vnode *skipvp; 813 int busy; 814 int flags; 815 }; 816 817 int 818 vflush_vnode(struct vnode *vp, void *arg) { 819 struct vflush_args *va = arg; 820 struct proc *p = curproc; 821 822 if (vp == va->skipvp) { 823 return (0); 824 } 825 826 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 827 return (0); 828 } 829 830 /* 831 * If WRITECLOSE is set, only flush out regular file 832 * vnodes open for writing. 833 */ 834 if ((va->flags & WRITECLOSE) && 835 (vp->v_writecount == 0 || vp->v_type != VREG)) { 836 return (0); 837 } 838 839 /* 840 * With v_usecount == 0, all we need to do is clear 841 * out the vnode data structures and we are done. 842 */ 843 if (vp->v_usecount == 0) { 844 vgonel(vp, p); 845 return (0); 846 } 847 848 /* 849 * If FORCECLOSE is set, forcibly close the vnode. 850 * For block or character devices, revert to an 851 * anonymous device. For all other files, just kill them. 852 */ 853 if (va->flags & FORCECLOSE) { 854 if (vp->v_type != VBLK && vp->v_type != VCHR) { 855 vgonel(vp, p); 856 } else { 857 vclean(vp, 0, p); 858 vp->v_op = spec_vnodeop_p; 859 insmntque(vp, (struct mount *)0); 860 } 861 return (0); 862 } 863 864 #ifdef DEBUG 865 if (busyprt) 866 vprint("vflush: busy vnode", vp); 867 #endif 868 va->busy++; 869 return (0); 870 } 871 872 int 873 vflush(struct mount *mp, struct vnode *skipvp, int flags) 874 { 875 struct vflush_args va; 876 va.skipvp = skipvp; 877 va.busy = 0; 878 va.flags = flags; 879 880 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 881 882 if (va.busy) 883 return (EBUSY); 884 return (0); 885 } 886 887 /* 888 * Disassociate the underlying file system from a vnode. 889 */ 890 void 891 vclean(struct vnode *vp, int flags, struct proc *p) 892 { 893 int active; 894 895 /* 896 * Check to see if the vnode is in use. 897 * If so we have to reference it before we clean it out 898 * so that its count cannot fall to zero and generate a 899 * race against ourselves to recycle it. 900 */ 901 if ((active = vp->v_usecount) != 0) 902 vp->v_usecount++; 903 904 /* 905 * Prevent the vnode from being recycled or 906 * brought into use while we clean it out. 907 */ 908 if (vp->v_flag & VXLOCK) 909 panic("vclean: deadlock"); 910 vp->v_flag |= VXLOCK; 911 /* 912 * Even if the count is zero, the VOP_INACTIVE routine may still 913 * have the object locked while it cleans it out. The VOP_LOCK 914 * ensures that the VOP_INACTIVE routine is done with its work. 915 * For active vnodes, it ensures that no other activity can 916 * occur while the underlying object is being cleaned out. 917 */ 918 VOP_LOCK(vp, LK_DRAIN, p); 919 920 /* 921 * Clean out any VM data associated with the vnode. 922 */ 923 uvm_vnp_terminate(vp); 924 /* 925 * Clean out any buffers associated with the vnode. 926 */ 927 if (flags & DOCLOSE) 928 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 929 /* 930 * If purging an active vnode, it must be closed and 931 * deactivated before being reclaimed. Note that the 932 * VOP_INACTIVE will unlock the vnode 933 */ 934 if (active) { 935 if (flags & DOCLOSE) 936 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 937 VOP_INACTIVE(vp, p); 938 } else { 939 /* 940 * Any other processes trying to obtain this lock must first 941 * wait for VXLOCK to clear, then call the new lock operation. 942 */ 943 VOP_UNLOCK(vp, 0, p); 944 } 945 946 /* 947 * Reclaim the vnode. 948 */ 949 if (VOP_RECLAIM(vp, p)) 950 panic("vclean: cannot reclaim"); 951 if (active) { 952 vp->v_usecount--; 953 if (vp->v_usecount == 0) { 954 if (vp->v_holdcnt > 0) 955 panic("vclean: not clean"); 956 vputonfreelist(vp); 957 } 958 } 959 cache_purge(vp); 960 961 /* 962 * Done with purge, notify sleepers of the grim news. 963 */ 964 vp->v_op = dead_vnodeop_p; 965 VN_KNOTE(vp, NOTE_REVOKE); 966 vp->v_tag = VT_NON; 967 vp->v_flag &= ~VXLOCK; 968 #ifdef VFSDEBUG 969 vp->v_flag &= ~VLOCKSWORK; 970 #endif 971 if (vp->v_flag & VXWANT) { 972 vp->v_flag &= ~VXWANT; 973 wakeup(vp); 974 } 975 } 976 977 /* 978 * Recycle an unused vnode to the front of the free list. 979 */ 980 int 981 vrecycle(struct vnode *vp, struct proc *p) 982 { 983 if (vp->v_usecount == 0) { 984 vgonel(vp, p); 985 return (1); 986 } 987 return (0); 988 } 989 990 /* 991 * Eliminate all activity associated with a vnode 992 * in preparation for reuse. 993 */ 994 void 995 vgone(struct vnode *vp) 996 { 997 struct proc *p = curproc; 998 vgonel(vp, p); 999 } 1000 1001 /* 1002 * vgone, with struct proc. 1003 */ 1004 void 1005 vgonel(struct vnode *vp, struct proc *p) 1006 { 1007 struct vnode *vq; 1008 struct vnode *vx; 1009 1010 /* 1011 * If a vgone (or vclean) is already in progress, 1012 * wait until it is done and return. 1013 */ 1014 if (vp->v_flag & VXLOCK) { 1015 vp->v_flag |= VXWANT; 1016 ltsleep(vp, PINOD | PNORELOCK, "vgone", 0, NULL); 1017 return; 1018 } 1019 1020 /* 1021 * Clean out the filesystem specific data. 1022 */ 1023 vclean(vp, DOCLOSE, p); 1024 /* 1025 * Delete from old mount point vnode list, if on one. 1026 */ 1027 if (vp->v_mount != NULL) 1028 insmntque(vp, (struct mount *)0); 1029 /* 1030 * If special device, remove it from special device alias list 1031 * if it is on one. 1032 */ 1033 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1034 if (*vp->v_hashchain == vp) { 1035 *vp->v_hashchain = vp->v_specnext; 1036 } else { 1037 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1038 if (vq->v_specnext != vp) 1039 continue; 1040 vq->v_specnext = vp->v_specnext; 1041 break; 1042 } 1043 if (vq == NULL) 1044 panic("missing bdev"); 1045 } 1046 if (vp->v_flag & VALIASED) { 1047 vx = NULL; 1048 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1049 if (vq->v_rdev != vp->v_rdev || 1050 vq->v_type != vp->v_type) 1051 continue; 1052 if (vx) 1053 break; 1054 vx = vq; 1055 } 1056 if (vx == NULL) 1057 panic("missing alias"); 1058 if (vq == NULL) 1059 vx->v_flag &= ~VALIASED; 1060 vp->v_flag &= ~VALIASED; 1061 } 1062 free(vp->v_specinfo, M_VNODE); 1063 vp->v_specinfo = NULL; 1064 } 1065 /* 1066 * If it is on the freelist and not already at the head, 1067 * move it to the head of the list. 1068 */ 1069 vp->v_type = VBAD; 1070 1071 /* 1072 * Move onto the free list, unless we were called from 1073 * getnewvnode and we're not on any free list 1074 */ 1075 if (vp->v_usecount == 0 && 1076 (vp->v_bioflag & VBIOONFREELIST)) { 1077 int s; 1078 1079 s = splbio(); 1080 1081 if (vp->v_holdcnt > 0) 1082 panic("vgonel: not clean"); 1083 1084 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1085 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1086 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1087 } 1088 splx(s); 1089 } 1090 } 1091 1092 /* 1093 * Lookup a vnode by device number. 1094 */ 1095 int 1096 vfinddev(dev_t dev, enum vtype type, struct vnode **vpp) 1097 { 1098 struct vnode *vp; 1099 int rc =0; 1100 1101 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1102 if (dev != vp->v_rdev || type != vp->v_type) 1103 continue; 1104 *vpp = vp; 1105 rc = 1; 1106 break; 1107 } 1108 return (rc); 1109 } 1110 1111 /* 1112 * Revoke all the vnodes corresponding to the specified minor number 1113 * range (endpoints inclusive) of the specified major. 1114 */ 1115 void 1116 vdevgone(int maj, int minl, int minh, enum vtype type) 1117 { 1118 struct vnode *vp; 1119 int mn; 1120 1121 for (mn = minl; mn <= minh; mn++) 1122 if (vfinddev(makedev(maj, mn), type, &vp)) 1123 VOP_REVOKE(vp, REVOKEALL); 1124 } 1125 1126 /* 1127 * Calculate the total number of references to a special device. 1128 */ 1129 int 1130 vcount(struct vnode *vp) 1131 { 1132 struct vnode *vq, *vnext; 1133 int count; 1134 1135 loop: 1136 if ((vp->v_flag & VALIASED) == 0) 1137 return (vp->v_usecount); 1138 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1139 vnext = vq->v_specnext; 1140 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1141 continue; 1142 /* 1143 * Alias, but not in use, so flush it out. 1144 */ 1145 if (vq->v_usecount == 0 && vq != vp) { 1146 vgone(vq); 1147 goto loop; 1148 } 1149 count += vq->v_usecount; 1150 } 1151 return (count); 1152 } 1153 1154 #if defined(DEBUG) || defined(DIAGNOSTIC) 1155 /* 1156 * Print out a description of a vnode. 1157 */ 1158 static char *typename[] = 1159 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1160 1161 void 1162 vprint(char *label, struct vnode *vp) 1163 { 1164 char buf[64]; 1165 1166 if (label != NULL) 1167 printf("%s: ", label); 1168 printf("%p, type %s, use %u, write %u, hold %u,", 1169 vp, typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1170 vp->v_holdcnt); 1171 buf[0] = '\0'; 1172 if (vp->v_flag & VROOT) 1173 strlcat(buf, "|VROOT", sizeof buf); 1174 if (vp->v_flag & VTEXT) 1175 strlcat(buf, "|VTEXT", sizeof buf); 1176 if (vp->v_flag & VSYSTEM) 1177 strlcat(buf, "|VSYSTEM", sizeof buf); 1178 if (vp->v_flag & VXLOCK) 1179 strlcat(buf, "|VXLOCK", sizeof buf); 1180 if (vp->v_flag & VXWANT) 1181 strlcat(buf, "|VXWANT", sizeof buf); 1182 if (vp->v_bioflag & VBIOWAIT) 1183 strlcat(buf, "|VBIOWAIT", sizeof buf); 1184 if (vp->v_bioflag & VBIOONFREELIST) 1185 strlcat(buf, "|VBIOONFREELIST", sizeof buf); 1186 if (vp->v_bioflag & VBIOONSYNCLIST) 1187 strlcat(buf, "|VBIOONSYNCLIST", sizeof buf); 1188 if (vp->v_flag & VALIASED) 1189 strlcat(buf, "|VALIASED", sizeof buf); 1190 if (buf[0] != '\0') 1191 printf(" flags (%s)", &buf[1]); 1192 if (vp->v_data == NULL) { 1193 printf("\n"); 1194 } else { 1195 printf("\n\t"); 1196 VOP_PRINT(vp); 1197 } 1198 } 1199 #endif /* DEBUG || DIAGNOSTIC */ 1200 1201 #ifdef DEBUG 1202 /* 1203 * List all of the locked vnodes in the system. 1204 * Called when debugging the kernel. 1205 */ 1206 void 1207 printlockedvnodes(void) 1208 { 1209 struct mount *mp, *nmp; 1210 struct vnode *vp; 1211 1212 printf("Locked vnodes\n"); 1213 1214 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1215 mp = nmp) { 1216 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) { 1217 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1218 continue; 1219 } 1220 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 1221 if (VOP_ISLOCKED(vp)) 1222 vprint((char *)0, vp); 1223 } 1224 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1225 vfs_unbusy(mp); 1226 } 1227 1228 } 1229 #endif 1230 1231 /* 1232 * Top level filesystem related information gathering. 1233 */ 1234 int 1235 vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1236 size_t newlen, struct proc *p) 1237 { 1238 struct vfsconf *vfsp, *tmpvfsp; 1239 int ret; 1240 1241 /* all sysctl names at this level are at least name and field */ 1242 if (namelen < 2) 1243 return (ENOTDIR); /* overloaded */ 1244 1245 if (name[0] != VFS_GENERIC) { 1246 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1247 if (vfsp->vfc_typenum == name[0]) 1248 break; 1249 1250 if (vfsp == NULL) 1251 return (EOPNOTSUPP); 1252 1253 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1254 oldp, oldlenp, newp, newlen, p)); 1255 } 1256 1257 switch (name[1]) { 1258 case VFS_MAXTYPENUM: 1259 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1260 1261 case VFS_CONF: 1262 if (namelen < 3) 1263 return (ENOTDIR); /* overloaded */ 1264 1265 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1266 if (vfsp->vfc_typenum == name[2]) 1267 break; 1268 1269 if (vfsp == NULL) 1270 return (EOPNOTSUPP); 1271 1272 /* Make a copy, clear out kernel pointers */ 1273 tmpvfsp = malloc(sizeof(*tmpvfsp), M_TEMP, M_WAITOK); 1274 bcopy(vfsp, tmpvfsp, sizeof(*tmpvfsp)); 1275 tmpvfsp->vfc_vfsops = NULL; 1276 tmpvfsp->vfc_mountroot = NULL; 1277 tmpvfsp->vfc_next = NULL; 1278 1279 ret = sysctl_rdstruct(oldp, oldlenp, newp, tmpvfsp, 1280 sizeof(struct vfsconf)); 1281 1282 free(tmpvfsp, M_TEMP); 1283 return (ret); 1284 } 1285 1286 return (EOPNOTSUPP); 1287 } 1288 1289 int kinfo_vdebug = 1; 1290 #define KINFO_VNODESLOP 10 1291 /* 1292 * Dump vnode list (via sysctl). 1293 * Copyout address of vnode followed by vnode. 1294 */ 1295 /* ARGSUSED */ 1296 int 1297 sysctl_vnode(char *where, size_t *sizep, struct proc *p) 1298 { 1299 struct mount *mp, *nmp; 1300 struct vnode *vp, *nvp; 1301 char *bp = where, *savebp; 1302 char *ewhere; 1303 int error; 1304 1305 if (where == NULL) { 1306 *sizep = (numvnodes + KINFO_VNODESLOP) * sizeof(struct e_vnode); 1307 return (0); 1308 } 1309 ewhere = where + *sizep; 1310 1311 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1312 mp = nmp) { 1313 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) { 1314 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1315 continue; 1316 } 1317 savebp = bp; 1318 again: 1319 for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; 1320 vp = nvp) { 1321 /* 1322 * Check that the vp is still associated with 1323 * this filesystem. RACE: could have been 1324 * recycled onto the same filesystem. 1325 */ 1326 if (vp->v_mount != mp) { 1327 if (kinfo_vdebug) 1328 printf("kinfo: vp changed\n"); 1329 bp = savebp; 1330 goto again; 1331 } 1332 nvp = LIST_NEXT(vp, v_mntvnodes); 1333 if (bp + sizeof(struct e_vnode) > ewhere) { 1334 *sizep = bp - where; 1335 vfs_unbusy(mp); 1336 return (ENOMEM); 1337 } 1338 if ((error = copyout(&vp, 1339 &((struct e_vnode *)bp)->vptr, 1340 sizeof(struct vnode *))) || 1341 (error = copyout(vp, 1342 &((struct e_vnode *)bp)->vnode, 1343 sizeof(struct vnode)))) { 1344 vfs_unbusy(mp); 1345 return (error); 1346 } 1347 bp += sizeof(struct e_vnode); 1348 } 1349 1350 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1351 vfs_unbusy(mp); 1352 } 1353 1354 *sizep = bp - where; 1355 1356 return (0); 1357 } 1358 1359 /* 1360 * Check to see if a filesystem is mounted on a block device. 1361 */ 1362 int 1363 vfs_mountedon(struct vnode *vp) 1364 { 1365 struct vnode *vq; 1366 int error = 0; 1367 1368 if (vp->v_specmountpoint != NULL) 1369 return (EBUSY); 1370 if (vp->v_flag & VALIASED) { 1371 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1372 if (vq->v_rdev != vp->v_rdev || 1373 vq->v_type != vp->v_type) 1374 continue; 1375 if (vq->v_specmountpoint != NULL) { 1376 error = EBUSY; 1377 break; 1378 } 1379 } 1380 } 1381 return (error); 1382 } 1383 1384 /* 1385 * Build hash lists of net addresses and hang them off the mount point. 1386 * Called by ufs_mount() to set up the lists of export addresses. 1387 */ 1388 int 1389 vfs_hang_addrlist(struct mount *mp, struct netexport *nep, 1390 struct export_args *argp) 1391 { 1392 struct netcred *np; 1393 struct radix_node_head *rnh; 1394 int i; 1395 struct radix_node *rn; 1396 struct sockaddr *saddr, *smask = 0; 1397 struct domain *dom; 1398 int error; 1399 1400 if (argp->ex_addrlen == 0) { 1401 if (mp->mnt_flag & MNT_DEFEXPORTED) 1402 return (EPERM); 1403 np = &nep->ne_defexported; 1404 np->netc_exflags = argp->ex_flags; 1405 np->netc_anon = argp->ex_anon; 1406 np->netc_anon.cr_ref = 1; 1407 mp->mnt_flag |= MNT_DEFEXPORTED; 1408 return (0); 1409 } 1410 if (argp->ex_addrlen > MLEN || argp->ex_masklen > MLEN || 1411 argp->ex_addrlen < 0 || argp->ex_masklen < 0) 1412 return (EINVAL); 1413 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1414 np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK|M_ZERO); 1415 saddr = (struct sockaddr *)(np + 1); 1416 error = copyin(argp->ex_addr, saddr, argp->ex_addrlen); 1417 if (error) 1418 goto out; 1419 if (saddr->sa_len > argp->ex_addrlen) 1420 saddr->sa_len = argp->ex_addrlen; 1421 if (argp->ex_masklen) { 1422 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1423 error = copyin(argp->ex_mask, smask, argp->ex_masklen); 1424 if (error) 1425 goto out; 1426 if (smask->sa_len > argp->ex_masklen) 1427 smask->sa_len = argp->ex_masklen; 1428 } 1429 i = saddr->sa_family; 1430 if (i < 0 || i > AF_MAX) { 1431 error = EINVAL; 1432 goto out; 1433 } 1434 if ((rnh = nep->ne_rtable[i]) == 0) { 1435 /* 1436 * Seems silly to initialize every AF when most are not 1437 * used, do so on demand here 1438 */ 1439 for (dom = domains; dom; dom = dom->dom_next) 1440 if (dom->dom_family == i && dom->dom_rtattach) { 1441 dom->dom_rtattach((void **)&nep->ne_rtable[i], 1442 dom->dom_rtoffset); 1443 break; 1444 } 1445 if ((rnh = nep->ne_rtable[i]) == 0) { 1446 error = ENOBUFS; 1447 goto out; 1448 } 1449 } 1450 rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, 1451 np->netc_rnodes); 1452 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1453 error = EPERM; 1454 goto out; 1455 } 1456 np->netc_exflags = argp->ex_flags; 1457 np->netc_anon = argp->ex_anon; 1458 np->netc_anon.cr_ref = 1; 1459 return (0); 1460 out: 1461 free(np, M_NETADDR); 1462 return (error); 1463 } 1464 1465 /* ARGSUSED */ 1466 int 1467 vfs_free_netcred(struct radix_node *rn, void *w) 1468 { 1469 struct radix_node_head *rnh = (struct radix_node_head *)w; 1470 1471 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh, NULL); 1472 free(rn, M_NETADDR); 1473 return (0); 1474 } 1475 1476 /* 1477 * Free the net address hash lists that are hanging off the mount points. 1478 */ 1479 void 1480 vfs_free_addrlist(struct netexport *nep) 1481 { 1482 int i; 1483 struct radix_node_head *rnh; 1484 1485 for (i = 0; i <= AF_MAX; i++) 1486 if ((rnh = nep->ne_rtable[i]) != NULL) { 1487 (*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh); 1488 free(rnh, M_RTABLE); 1489 nep->ne_rtable[i] = 0; 1490 } 1491 } 1492 1493 int 1494 vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp) 1495 { 1496 int error; 1497 1498 if (argp->ex_flags & MNT_DELEXPORT) { 1499 vfs_free_addrlist(nep); 1500 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1501 } 1502 if (argp->ex_flags & MNT_EXPORTED) { 1503 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1504 return (error); 1505 mp->mnt_flag |= MNT_EXPORTED; 1506 } 1507 return (0); 1508 } 1509 1510 struct netcred * 1511 vfs_export_lookup(struct mount *mp, struct netexport *nep, struct mbuf *nam) 1512 { 1513 struct netcred *np; 1514 struct radix_node_head *rnh; 1515 struct sockaddr *saddr; 1516 1517 np = NULL; 1518 if (mp->mnt_flag & MNT_EXPORTED) { 1519 /* 1520 * Lookup in the export list first. 1521 */ 1522 if (nam != NULL) { 1523 saddr = mtod(nam, struct sockaddr *); 1524 rnh = nep->ne_rtable[saddr->sa_family]; 1525 if (rnh != NULL) { 1526 np = (struct netcred *) 1527 (*rnh->rnh_matchaddr)((caddr_t)saddr, 1528 rnh); 1529 if (np && np->netc_rnodes->rn_flags & RNF_ROOT) 1530 np = NULL; 1531 } 1532 } 1533 /* 1534 * If no address match, use the default if it exists. 1535 */ 1536 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1537 np = &nep->ne_defexported; 1538 } 1539 return (np); 1540 } 1541 1542 /* 1543 * Do the usual access checking. 1544 * file_mode, uid and gid are from the vnode in question, 1545 * while acc_mode and cred are from the VOP_ACCESS parameter list 1546 */ 1547 int 1548 vaccess(mode_t file_mode, uid_t uid, gid_t gid, mode_t acc_mode, 1549 struct ucred *cred) 1550 { 1551 mode_t mask; 1552 1553 /* User id 0 always gets access. */ 1554 if (cred->cr_uid == 0) 1555 return 0; 1556 1557 mask = 0; 1558 1559 /* Otherwise, check the owner. */ 1560 if (cred->cr_uid == uid) { 1561 if (acc_mode & VEXEC) 1562 mask |= S_IXUSR; 1563 if (acc_mode & VREAD) 1564 mask |= S_IRUSR; 1565 if (acc_mode & VWRITE) 1566 mask |= S_IWUSR; 1567 return (file_mode & mask) == mask ? 0 : EACCES; 1568 } 1569 1570 /* Otherwise, check the groups. */ 1571 if (cred->cr_gid == gid || groupmember(gid, cred)) { 1572 if (acc_mode & VEXEC) 1573 mask |= S_IXGRP; 1574 if (acc_mode & VREAD) 1575 mask |= S_IRGRP; 1576 if (acc_mode & VWRITE) 1577 mask |= S_IWGRP; 1578 return (file_mode & mask) == mask ? 0 : EACCES; 1579 } 1580 1581 /* Otherwise, check everyone else. */ 1582 if (acc_mode & VEXEC) 1583 mask |= S_IXOTH; 1584 if (acc_mode & VREAD) 1585 mask |= S_IROTH; 1586 if (acc_mode & VWRITE) 1587 mask |= S_IWOTH; 1588 return (file_mode & mask) == mask ? 0 : EACCES; 1589 } 1590 1591 /* 1592 * Unmount all file systems. 1593 * We traverse the list in reverse order under the assumption that doing so 1594 * will avoid needing to worry about dependencies. 1595 */ 1596 void 1597 vfs_unmountall(void) 1598 { 1599 struct mount *mp, *nmp; 1600 int allerror, error, again = 1; 1601 1602 retry: 1603 allerror = 0; 1604 for (mp = CIRCLEQ_LAST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1605 mp = nmp) { 1606 nmp = CIRCLEQ_PREV(mp, mnt_list); 1607 if ((vfs_busy(mp, VB_WRITE|VB_NOWAIT)) != 0) 1608 continue; 1609 if ((error = dounmount(mp, MNT_FORCE, curproc, NULL)) != 0) { 1610 printf("unmount of %s failed with error %d\n", 1611 mp->mnt_stat.f_mntonname, error); 1612 allerror = 1; 1613 } 1614 } 1615 1616 if (allerror) { 1617 printf("WARNING: some file systems would not unmount\n"); 1618 if (again) { 1619 printf("retrying\n"); 1620 again = 0; 1621 goto retry; 1622 } 1623 } 1624 } 1625 1626 /* 1627 * Sync and unmount file systems before shutting down. 1628 */ 1629 void 1630 vfs_shutdown(void) 1631 { 1632 #ifdef ACCOUNTING 1633 extern void acct_shutdown(void); 1634 1635 acct_shutdown(); 1636 #endif 1637 1638 /* XXX Should suspend scheduling. */ 1639 (void) spl0(); 1640 1641 printf("\nsyncing disks... "); 1642 1643 if (panicstr == 0) { 1644 /* Sync before unmount, in case we hang on something. */ 1645 sys_sync(&proc0, (void *)0, (register_t *)0); 1646 1647 /* Unmount file systems. */ 1648 vfs_unmountall(); 1649 } 1650 1651 if (vfs_syncwait(1)) 1652 printf("giving up\n"); 1653 else 1654 printf("done\n"); 1655 } 1656 1657 /* 1658 * perform sync() operation and wait for buffers to flush. 1659 * assumtions: called w/ scheduler disabled and physical io enabled 1660 * for now called at spl0() XXX 1661 */ 1662 int 1663 vfs_syncwait(int verbose) 1664 { 1665 struct buf *bp; 1666 int iter, nbusy, dcount, s; 1667 struct proc *p; 1668 1669 p = curproc? curproc : &proc0; 1670 sys_sync(p, (void *)0, (register_t *)0); 1671 1672 /* Wait for sync to finish. */ 1673 dcount = 10000; 1674 for (iter = 0; iter < 20; iter++) { 1675 nbusy = 0; 1676 LIST_FOREACH(bp, &bufhead, b_list) { 1677 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1678 nbusy++; 1679 /* 1680 * With soft updates, some buffers that are 1681 * written will be remarked as dirty until other 1682 * buffers are written. 1683 */ 1684 if (bp->b_flags & B_DELWRI) { 1685 s = splbio(); 1686 bremfree(bp); 1687 bp->b_flags |= B_BUSY; 1688 splx(s); 1689 nbusy++; 1690 bawrite(bp); 1691 if (dcount-- <= 0) { 1692 if (verbose) 1693 printf("softdep "); 1694 return 1; 1695 } 1696 } 1697 } 1698 if (nbusy == 0) 1699 break; 1700 if (verbose) 1701 printf("%d ", nbusy); 1702 DELAY(40000 * iter); 1703 } 1704 1705 return nbusy; 1706 } 1707 1708 /* 1709 * posix file system related system variables. 1710 */ 1711 int 1712 fs_posix_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, 1713 void *newp, size_t newlen, struct proc *p) 1714 { 1715 /* all sysctl names at this level are terminal */ 1716 if (namelen != 1) 1717 return (ENOTDIR); 1718 1719 switch (name[0]) { 1720 case FS_POSIX_SETUID: 1721 if (newp && securelevel > 0) 1722 return (EPERM); 1723 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1724 default: 1725 return (EOPNOTSUPP); 1726 } 1727 /* NOTREACHED */ 1728 } 1729 1730 /* 1731 * file system related system variables. 1732 */ 1733 int 1734 fs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1735 size_t newlen, struct proc *p) 1736 { 1737 sysctlfn *fn; 1738 1739 switch (name[0]) { 1740 case FS_POSIX: 1741 fn = fs_posix_sysctl; 1742 break; 1743 default: 1744 return (EOPNOTSUPP); 1745 } 1746 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1747 } 1748 1749 1750 /* 1751 * Routines dealing with vnodes and buffers 1752 */ 1753 1754 /* 1755 * Wait for all outstanding I/Os to complete 1756 * 1757 * Manipulates v_numoutput. Must be called at splbio() 1758 */ 1759 int 1760 vwaitforio(struct vnode *vp, int slpflag, char *wmesg, int timeo) 1761 { 1762 int error = 0; 1763 1764 splassert(IPL_BIO); 1765 1766 while (vp->v_numoutput) { 1767 vp->v_bioflag |= VBIOWAIT; 1768 error = tsleep(&vp->v_numoutput, 1769 slpflag | (PRIBIO + 1), wmesg, timeo); 1770 if (error) 1771 break; 1772 } 1773 1774 return (error); 1775 } 1776 1777 /* 1778 * Update outstanding I/O count and do wakeup if requested. 1779 * 1780 * Manipulates v_numoutput. Must be called at splbio() 1781 */ 1782 void 1783 vwakeup(struct vnode *vp) 1784 { 1785 splassert(IPL_BIO); 1786 1787 if (vp != NULL) { 1788 if (vp->v_numoutput-- == 0) 1789 panic("vwakeup: neg numoutput"); 1790 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1791 vp->v_bioflag &= ~VBIOWAIT; 1792 wakeup(&vp->v_numoutput); 1793 } 1794 } 1795 } 1796 1797 /* 1798 * Flush out and invalidate all buffers associated with a vnode. 1799 * Called with the underlying object locked. 1800 */ 1801 int 1802 vinvalbuf(struct vnode *vp, int flags, struct ucred *cred, struct proc *p, 1803 int slpflag, int slptimeo) 1804 { 1805 struct buf *bp; 1806 struct buf *nbp, *blist; 1807 int s, error; 1808 1809 #ifdef VFSDEBUG 1810 if ((vp->v_flag & VLOCKSWORK) && !VOP_ISLOCKED(vp)) 1811 panic("vinvalbuf(): vp isn't locked"); 1812 #endif 1813 1814 if (flags & V_SAVE) { 1815 s = splbio(); 1816 vwaitforio(vp, 0, "vinvalbuf", 0); 1817 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1818 splx(s); 1819 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1820 return (error); 1821 s = splbio(); 1822 if (vp->v_numoutput > 0 || 1823 !LIST_EMPTY(&vp->v_dirtyblkhd)) 1824 panic("vinvalbuf: dirty bufs"); 1825 } 1826 splx(s); 1827 } 1828 loop: 1829 s = splbio(); 1830 for (;;) { 1831 if ((blist = LIST_FIRST(&vp->v_cleanblkhd)) && 1832 (flags & V_SAVEMETA)) 1833 while (blist && blist->b_lblkno < 0) 1834 blist = LIST_NEXT(blist, b_vnbufs); 1835 if (blist == NULL && 1836 (blist = LIST_FIRST(&vp->v_dirtyblkhd)) && 1837 (flags & V_SAVEMETA)) 1838 while (blist && blist->b_lblkno < 0) 1839 blist = LIST_NEXT(blist, b_vnbufs); 1840 if (!blist) 1841 break; 1842 1843 for (bp = blist; bp; bp = nbp) { 1844 nbp = LIST_NEXT(bp, b_vnbufs); 1845 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1846 continue; 1847 if (bp->b_flags & B_BUSY) { 1848 bp->b_flags |= B_WANTED; 1849 error = tsleep(bp, slpflag | (PRIBIO + 1), 1850 "vinvalbuf", slptimeo); 1851 if (error) { 1852 splx(s); 1853 return (error); 1854 } 1855 break; 1856 } 1857 bremfree(bp); 1858 bp->b_flags |= B_BUSY; 1859 /* 1860 * XXX Since there are no node locks for NFS, I believe 1861 * there is a slight chance that a delayed write will 1862 * occur while sleeping just above, so check for it. 1863 */ 1864 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1865 splx(s); 1866 (void) VOP_BWRITE(bp); 1867 goto loop; 1868 } 1869 bp->b_flags |= B_INVAL; 1870 brelse(bp); 1871 } 1872 } 1873 if (!(flags & V_SAVEMETA) && 1874 (!LIST_EMPTY(&vp->v_dirtyblkhd) || !LIST_EMPTY(&vp->v_cleanblkhd))) 1875 panic("vinvalbuf: flush failed"); 1876 splx(s); 1877 return (0); 1878 } 1879 1880 void 1881 vflushbuf(struct vnode *vp, int sync) 1882 { 1883 struct buf *bp, *nbp; 1884 int s; 1885 1886 loop: 1887 s = splbio(); 1888 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); 1889 bp != LIST_END(&vp->v_dirtyblkhd); bp = nbp) { 1890 nbp = LIST_NEXT(bp, b_vnbufs); 1891 if ((bp->b_flags & B_BUSY)) 1892 continue; 1893 if ((bp->b_flags & B_DELWRI) == 0) 1894 panic("vflushbuf: not dirty"); 1895 bremfree(bp); 1896 bp->b_flags |= B_BUSY; 1897 splx(s); 1898 /* 1899 * Wait for I/O associated with indirect blocks to complete, 1900 * since there is no way to quickly wait for them below. 1901 */ 1902 if (bp->b_vp == vp || sync == 0) 1903 (void) bawrite(bp); 1904 else 1905 (void) bwrite(bp); 1906 goto loop; 1907 } 1908 if (sync == 0) { 1909 splx(s); 1910 return; 1911 } 1912 vwaitforio(vp, 0, "vflushbuf", 0); 1913 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1914 splx(s); 1915 #ifdef DIAGNOSTIC 1916 vprint("vflushbuf: dirty", vp); 1917 #endif 1918 goto loop; 1919 } 1920 splx(s); 1921 } 1922 1923 /* 1924 * Associate a buffer with a vnode. 1925 * 1926 * Manipulates buffer vnode queues. Must be called at splbio(). 1927 */ 1928 void 1929 bgetvp(struct vnode *vp, struct buf *bp) 1930 { 1931 splassert(IPL_BIO); 1932 1933 1934 if (bp->b_vp) 1935 panic("bgetvp: not free"); 1936 vhold(vp); 1937 bp->b_vp = vp; 1938 if (vp->v_type == VBLK || vp->v_type == VCHR) 1939 bp->b_dev = vp->v_rdev; 1940 else 1941 bp->b_dev = NODEV; 1942 /* 1943 * Insert onto list for new vnode. 1944 */ 1945 bufinsvn(bp, &vp->v_cleanblkhd); 1946 } 1947 1948 /* 1949 * Disassociate a buffer from a vnode. 1950 * 1951 * Manipulates vnode buffer queues. Must be called at splbio(). 1952 */ 1953 void 1954 brelvp(struct buf *bp) 1955 { 1956 struct vnode *vp; 1957 1958 splassert(IPL_BIO); 1959 1960 if ((vp = bp->b_vp) == (struct vnode *) 0) 1961 panic("brelvp: NULL"); 1962 /* 1963 * Delete from old vnode list, if on one. 1964 */ 1965 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 1966 bufremvn(bp); 1967 if ((vp->v_bioflag & VBIOONSYNCLIST) && 1968 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 1969 vp->v_bioflag &= ~VBIOONSYNCLIST; 1970 LIST_REMOVE(vp, v_synclist); 1971 } 1972 bp->b_vp = (struct vnode *) 0; 1973 1974 #ifdef DIAGNOSTIC 1975 if (vp->v_holdcnt == 0) 1976 panic("brelvp: holdcnt"); 1977 #endif 1978 vp->v_holdcnt--; 1979 1980 /* 1981 * If it is on the holdlist and the hold count drops to 1982 * zero, move it to the free list. 1983 */ 1984 if ((vp->v_bioflag & VBIOONFREELIST) && 1985 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 1986 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 1987 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 1988 } 1989 } 1990 1991 /* 1992 * Replaces the current vnode associated with the buffer, if any, 1993 * with a new vnode. 1994 * 1995 * If an output I/O is pending on the buffer, the old vnode 1996 * I/O count is adjusted. 1997 * 1998 * Ignores vnode buffer queues. Must be called at splbio(). 1999 */ 2000 void 2001 buf_replacevnode(struct buf *bp, struct vnode *newvp) 2002 { 2003 struct vnode *oldvp = bp->b_vp; 2004 2005 splassert(IPL_BIO); 2006 2007 if (oldvp) 2008 brelvp(bp); 2009 2010 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2011 newvp->v_numoutput++; /* put it on swapdev */ 2012 vwakeup(oldvp); 2013 } 2014 2015 bgetvp(newvp, bp); 2016 bufremvn(bp); 2017 } 2018 2019 /* 2020 * Used to assign buffers to the appropriate clean or dirty list on 2021 * the vnode and to add newly dirty vnodes to the appropriate 2022 * filesystem syncer list. 2023 * 2024 * Manipulates vnode buffer queues. Must be called at splbio(). 2025 */ 2026 void 2027 reassignbuf(struct buf *bp) 2028 { 2029 struct buflists *listheadp; 2030 int delay; 2031 struct vnode *vp = bp->b_vp; 2032 2033 splassert(IPL_BIO); 2034 2035 /* 2036 * Delete from old vnode list, if on one. 2037 */ 2038 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2039 bufremvn(bp); 2040 2041 /* 2042 * If dirty, put on list of dirty buffers; 2043 * otherwise insert onto list of clean buffers. 2044 */ 2045 if ((bp->b_flags & B_DELWRI) == 0) { 2046 listheadp = &vp->v_cleanblkhd; 2047 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2048 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2049 vp->v_bioflag &= ~VBIOONSYNCLIST; 2050 LIST_REMOVE(vp, v_synclist); 2051 } 2052 } else { 2053 listheadp = &vp->v_dirtyblkhd; 2054 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2055 switch (vp->v_type) { 2056 case VDIR: 2057 delay = syncdelay / 2; 2058 break; 2059 case VBLK: 2060 if (vp->v_specmountpoint != NULL) { 2061 delay = syncdelay / 3; 2062 break; 2063 } 2064 /* FALLTHROUGH */ 2065 default: 2066 delay = syncdelay; 2067 } 2068 vn_syncer_add_to_worklist(vp, delay); 2069 } 2070 } 2071 bufinsvn(bp, listheadp); 2072 } 2073 2074 int 2075 vfs_register(struct vfsconf *vfs) 2076 { 2077 struct vfsconf *vfsp; 2078 struct vfsconf **vfspp; 2079 2080 #ifdef DIAGNOSTIC 2081 /* Paranoia? */ 2082 if (vfs->vfc_refcount != 0) 2083 printf("vfs_register called with vfc_refcount > 0\n"); 2084 #endif 2085 2086 /* Check if filesystem already known */ 2087 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2088 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2089 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2090 return (EEXIST); 2091 2092 if (vfs->vfc_typenum > maxvfsconf) 2093 maxvfsconf = vfs->vfc_typenum; 2094 2095 vfs->vfc_next = NULL; 2096 2097 /* Add to the end of the list */ 2098 *vfspp = vfs; 2099 2100 /* Call vfs_init() */ 2101 if (vfs->vfc_vfsops->vfs_init) 2102 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2103 2104 return 0; 2105 } 2106 2107 int 2108 vfs_unregister(struct vfsconf *vfs) 2109 { 2110 struct vfsconf *vfsp; 2111 struct vfsconf **vfspp; 2112 int maxtypenum; 2113 2114 /* Find our vfsconf struct */ 2115 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2116 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2117 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2118 break; 2119 } 2120 2121 if (!vfsp) /* Not found */ 2122 return (ENOENT); 2123 2124 if (vfsp->vfc_refcount) /* In use */ 2125 return (EBUSY); 2126 2127 /* Remove from list and free */ 2128 *vfspp = vfsp->vfc_next; 2129 2130 maxtypenum = 0; 2131 2132 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2133 if (vfsp->vfc_typenum > maxtypenum) 2134 maxtypenum = vfsp->vfc_typenum; 2135 2136 maxvfsconf = maxtypenum; 2137 return 0; 2138 } 2139 2140 /* 2141 * Check if vnode represents a disk device 2142 */ 2143 int 2144 vn_isdisk(struct vnode *vp, int *errp) 2145 { 2146 if (vp->v_type != VBLK && vp->v_type != VCHR) 2147 return (0); 2148 2149 return (1); 2150 } 2151 2152 #ifdef DDB 2153 #include <machine/db_machdep.h> 2154 #include <ddb/db_interface.h> 2155 #include <ddb/db_output.h> 2156 2157 void 2158 vfs_buf_print(struct buf *bp, int full, int (*pr)(const char *, ...)) 2159 { 2160 2161 (*pr)(" vp %p lblkno 0x%llx blkno 0x%llx dev 0x%x\n" 2162 " proc %p error %d flags %b\n", 2163 bp->b_vp, (int64_t)bp->b_lblkno, (int64_t)bp->b_blkno, bp->b_dev, 2164 bp->b_proc, bp->b_error, bp->b_flags, B_BITS); 2165 2166 (*pr)(" bufsize 0x%lx bcount 0x%lx resid 0x%lx sync 0x%x\n" 2167 " data %p saveaddr %p dep %p iodone %p\n", 2168 bp->b_bufsize, bp->b_bcount, (long)bp->b_resid, bp->b_synctime, 2169 bp->b_data, bp->b_saveaddr, LIST_FIRST(&bp->b_dep), bp->b_iodone); 2170 2171 (*pr)(" dirty {off 0x%x end 0x%x} valid {off 0x%x end 0x%x}\n", 2172 bp->b_dirtyoff, bp->b_dirtyend, bp->b_validoff, bp->b_validend); 2173 2174 #ifdef FFS_SOFTUPDATES 2175 if (full) 2176 softdep_print(bp, full, pr); 2177 #endif 2178 } 2179 2180 const char *vtypes[] = { VTYPE_NAMES }; 2181 const char *vtags[] = { VTAG_NAMES }; 2182 2183 void 2184 vfs_vnode_print(struct vnode *vp, int full, int (*pr)(const char *, ...)) 2185 { 2186 2187 #define NENTS(n) (sizeof n / sizeof(n[0])) 2188 (*pr)("tag %s(%d) type %s(%d) mount %p typedata %p\n", 2189 vp->v_tag > NENTS(vtags)? "<unk>":vtags[vp->v_tag], vp->v_tag, 2190 vp->v_type > NENTS(vtypes)? "<unk>":vtypes[vp->v_type], 2191 vp->v_type, vp->v_mount, vp->v_mountedhere); 2192 2193 (*pr)("data %p usecount %d writecount %ld holdcnt %ld numoutput %d\n", 2194 vp->v_data, vp->v_usecount, vp->v_writecount, 2195 vp->v_holdcnt, vp->v_numoutput); 2196 2197 /* uvm_object_printit(&vp->v_uobj, full, pr); */ 2198 2199 if (full) { 2200 struct buf *bp; 2201 2202 (*pr)("clean bufs:\n"); 2203 LIST_FOREACH(bp, &vp->v_cleanblkhd, b_vnbufs) { 2204 (*pr)(" bp %p\n", bp); 2205 vfs_buf_print(bp, full, pr); 2206 } 2207 2208 (*pr)("dirty bufs:\n"); 2209 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) { 2210 (*pr)(" bp %p\n", bp); 2211 vfs_buf_print(bp, full, pr); 2212 } 2213 } 2214 } 2215 2216 void 2217 vfs_mount_print(struct mount *mp, int full, int (*pr)(const char *, ...)) 2218 { 2219 struct vfsconf *vfc = mp->mnt_vfc; 2220 struct vnode *vp; 2221 int cnt = 0; 2222 2223 (*pr)("flags %b\nvnodecovered %p syncer %p data %p\n", 2224 mp->mnt_flag, MNT_BITS, 2225 mp->mnt_vnodecovered, mp->mnt_syncer, mp->mnt_data); 2226 2227 (*pr)("vfsconf: ops %p name \"%s\" num %d ref %d flags 0x%x\n", 2228 vfc->vfc_vfsops, vfc->vfc_name, vfc->vfc_typenum, 2229 vfc->vfc_refcount, vfc->vfc_flags); 2230 2231 (*pr)("statvfs cache: bsize %x iosize %x\nblocks %u free %u avail %u\n", 2232 mp->mnt_stat.f_bsize, mp->mnt_stat.f_iosize, mp->mnt_stat.f_blocks, 2233 mp->mnt_stat.f_bfree, mp->mnt_stat.f_bavail); 2234 2235 (*pr)(" files %u ffiles %u\n", mp->mnt_stat.f_files, 2236 mp->mnt_stat.f_ffree); 2237 2238 (*pr)(" f_fsidx {0x%x, 0x%x} owner %u ctime 0x%x\n", 2239 mp->mnt_stat.f_fsid.val[0], mp->mnt_stat.f_fsid.val[1], 2240 mp->mnt_stat.f_owner, mp->mnt_stat.f_ctime); 2241 2242 (*pr)(" syncwrites %lu asyncwrites = %lu\n", 2243 mp->mnt_stat.f_syncwrites, mp->mnt_stat.f_asyncwrites); 2244 2245 (*pr)(" fstype \"%s\" mnton \"%s\" mntfrom \"%s\"\n", 2246 mp->mnt_stat.f_fstypename, mp->mnt_stat.f_mntonname, 2247 mp->mnt_stat.f_mntfromname); 2248 2249 (*pr)("locked vnodes:"); 2250 /* XXX would take mountlist lock, except ddb has no context */ 2251 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2252 if (VOP_ISLOCKED(vp)) { 2253 if (!LIST_NEXT(vp, v_mntvnodes)) 2254 (*pr)(" %p", vp); 2255 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2256 (*pr)("\n\t%p", vp); 2257 else 2258 (*pr)(", %p", vp); 2259 } 2260 (*pr)("\n"); 2261 2262 if (full) { 2263 (*pr)("all vnodes:\n\t"); 2264 /* XXX would take mountlist lock, except ddb has no context */ 2265 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2266 if (!LIST_NEXT(vp, v_mntvnodes)) 2267 (*pr)(" %p", vp); 2268 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2269 (*pr)(" %p,\n\t", vp); 2270 else 2271 (*pr)(" %p,", vp); 2272 (*pr)("\n", vp); 2273 } 2274 } 2275 #endif /* DDB */ 2276