1 /* $OpenBSD: vfs_subr.c,v 1.177 2009/06/06 18:06:22 art Exp $ */ 2 /* $NetBSD: vfs_subr.c,v 1.53 1996/04/22 01:39:13 christos Exp $ */ 3 4 /* 5 * Copyright (c) 1989, 1993 6 * The Regents of the University of California. All rights reserved. 7 * (c) UNIX System Laboratories, Inc. 8 * All or some portions of this file are derived from material licensed 9 * to the University of California by American Telephone and Telegraph 10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 11 * the permission of UNIX System Laboratories, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. 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 69 enum vtype iftovt_tab[16] = { 70 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, 71 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, 72 }; 73 74 int vttoif_tab[9] = { 75 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, 76 S_IFSOCK, S_IFIFO, S_IFMT, 77 }; 78 79 int doforce = 1; /* 1 => permit forcible unmounting */ 80 int prtactive = 0; /* 1 => print out reclaim of active vnodes */ 81 int 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 92 struct freelst vnode_hold_list; /* list of vnodes referencing buffers */ 93 struct freelst vnode_free_list; /* vnode free list */ 94 95 struct mntlist mountlist; /* mounted filesystem list */ 96 97 void vclean(struct vnode *, int, struct proc *); 98 void vhold(struct vnode *); 99 void vdrop(struct vnode *); 100 101 void insmntque(struct vnode *, struct mount *); 102 int getdevvp(dev_t, struct vnode **, enum vtype); 103 104 int vfs_hang_addrlist(struct mount *, struct netexport *, 105 struct export_args *); 106 int vfs_free_netcred(struct radix_node *, void *); 107 void vfs_free_addrlist(struct netexport *); 108 void vputonfreelist(struct vnode *); 109 110 int vflush_vnode(struct vnode *, void *); 111 int maxvnodes; 112 113 #ifdef DEBUG 114 void printlockedvnodes(void); 115 #endif 116 117 struct pool vnode_pool; 118 119 static int rb_buf_compare(struct buf *b1, struct buf *b2); 120 RB_GENERATE(buf_rb_bufs, buf, b_rbbufs, rb_buf_compare); 121 122 static int 123 rb_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 */ 135 void 136 vntblinit(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 */ 158 int 159 vfs_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 */ 186 void 187 vfs_unbusy(struct mount *mp) 188 { 189 rw_exit(&mp->mnt_lock); 190 } 191 192 int 193 vfs_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 */ 207 int 208 vfs_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 */ 237 struct mount * 238 vfs_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 */ 256 void 257 vfs_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 */ 285 long 286 makefstype(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 */ 300 void 301 vattr_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 */ 321 extern int (**dead_vnodeop_p)(void *); 322 long numvnodes; 323 324 /* 325 * Return the next vnode from the free list. 326 */ 327 int 328 getnewvnode(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 numvnodes++; 364 } else { 365 for (vp = TAILQ_FIRST(listhd); vp != NULLVP; 366 vp = TAILQ_NEXT(vp, v_freelist)) { 367 if (VOP_ISLOCKED(vp) == 0) 368 break; 369 } 370 /* 371 * Unless this is a bad time of the month, at most 372 * the first NCPUS items on the free list are 373 * locked, so this is close enough to being empty. 374 */ 375 if (vp == NULL) { 376 splx(s); 377 tablefull("vnode"); 378 *vpp = 0; 379 return (ENFILE); 380 } 381 382 #ifdef DIAGNOSTIC 383 if (vp->v_usecount) { 384 vprint("free vnode", vp); 385 panic("free vnode isn't"); 386 } 387 #endif 388 389 TAILQ_REMOVE(listhd, vp, v_freelist); 390 vp->v_bioflag &= ~VBIOONFREELIST; 391 splx(s); 392 393 if (vp->v_type != VBAD) 394 vgonel(vp, p); 395 #ifdef DIAGNOSTIC 396 if (vp->v_data) { 397 vprint("cleaned vnode", vp); 398 panic("cleaned vnode isn't"); 399 } 400 s = splbio(); 401 if (vp->v_numoutput) 402 panic("Clean vnode has pending I/O's"); 403 splx(s); 404 #endif 405 vp->v_flag = 0; 406 vp->v_socket = 0; 407 } 408 vp->v_type = VNON; 409 cache_purge(vp); 410 vp->v_tag = tag; 411 vp->v_op = vops; 412 insmntque(vp, mp); 413 *vpp = vp; 414 vp->v_usecount = 1; 415 vp->v_data = 0; 416 simple_lock_init(&vp->v_uvm.u_obj.vmobjlock); 417 return (0); 418 } 419 420 /* 421 * Move a vnode from one mount queue to another. 422 */ 423 void 424 insmntque(struct vnode *vp, struct mount *mp) 425 { 426 /* 427 * Delete from old mount point vnode list, if on one. 428 */ 429 if (vp->v_mount != NULL) 430 LIST_REMOVE(vp, v_mntvnodes); 431 /* 432 * Insert into list of vnodes for the new mount point, if available. 433 */ 434 if ((vp->v_mount = mp) != NULL) 435 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 436 } 437 438 /* 439 * Create a vnode for a block device. 440 * Used for root filesystem, argdev, and swap areas. 441 * Also used for memory file system special devices. 442 */ 443 int 444 bdevvp(dev_t dev, struct vnode **vpp) 445 { 446 return (getdevvp(dev, vpp, VBLK)); 447 } 448 449 /* 450 * Create a vnode for a character device. 451 * Used for console handling. 452 */ 453 int 454 cdevvp(dev_t dev, struct vnode **vpp) 455 { 456 return (getdevvp(dev, vpp, VCHR)); 457 } 458 459 /* 460 * Create a vnode for a device. 461 * Used by bdevvp (block device) for root file system etc., 462 * and by cdevvp (character device) for console. 463 */ 464 int 465 getdevvp(dev_t dev, struct vnode **vpp, enum vtype type) 466 { 467 struct vnode *vp; 468 struct vnode *nvp; 469 int error; 470 471 if (dev == NODEV) { 472 *vpp = NULLVP; 473 return (0); 474 } 475 error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); 476 if (error) { 477 *vpp = NULLVP; 478 return (error); 479 } 480 vp = nvp; 481 vp->v_type = type; 482 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 483 vput(vp); 484 vp = nvp; 485 } 486 *vpp = vp; 487 return (0); 488 } 489 490 /* 491 * Check to see if the new vnode represents a special device 492 * for which we already have a vnode (either because of 493 * bdevvp() or because of a different vnode representing 494 * the same block device). If such an alias exists, deallocate 495 * the existing contents and return the aliased vnode. The 496 * caller is responsible for filling it with its new contents. 497 */ 498 struct vnode * 499 checkalias(struct vnode *nvp, dev_t nvp_rdev, struct mount *mp) 500 { 501 struct proc *p = curproc; 502 struct vnode *vp; 503 struct vnode **vpp; 504 505 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 506 return (NULLVP); 507 508 vpp = &speclisth[SPECHASH(nvp_rdev)]; 509 loop: 510 for (vp = *vpp; vp; vp = vp->v_specnext) { 511 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 512 continue; 513 } 514 /* 515 * Alias, but not in use, so flush it out. 516 */ 517 if (vp->v_usecount == 0) { 518 vgonel(vp, p); 519 goto loop; 520 } 521 if (vget(vp, LK_EXCLUSIVE, p)) { 522 goto loop; 523 } 524 break; 525 } 526 527 /* 528 * Common case is actually in the if statement 529 */ 530 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 531 nvp->v_specinfo = malloc(sizeof(struct specinfo), M_VNODE, 532 M_WAITOK); 533 nvp->v_rdev = nvp_rdev; 534 nvp->v_hashchain = vpp; 535 nvp->v_specnext = *vpp; 536 nvp->v_specmountpoint = NULL; 537 nvp->v_speclockf = NULL; 538 bzero(nvp->v_specbitmap, sizeof(nvp->v_specbitmap)); 539 *vpp = nvp; 540 if (vp != NULLVP) { 541 nvp->v_flag |= VALIASED; 542 vp->v_flag |= VALIASED; 543 vput(vp); 544 } 545 return (NULLVP); 546 } 547 548 /* 549 * This code is the uncommon case. It is called in case 550 * we found an alias that was VT_NON && vtype of VBLK 551 * This means we found a block device that was created 552 * using bdevvp. 553 * An example of such a vnode is the root partition device vnode 554 * created in ffs_mountroot. 555 * 556 * The vnodes created by bdevvp should not be aliased (why?). 557 */ 558 559 VOP_UNLOCK(vp, 0, p); 560 vclean(vp, 0, p); 561 vp->v_op = nvp->v_op; 562 vp->v_tag = nvp->v_tag; 563 nvp->v_type = VNON; 564 insmntque(vp, mp); 565 return (vp); 566 } 567 568 /* 569 * Grab a particular vnode from the free list, increment its 570 * reference count and lock it. If the vnode lock bit is set, 571 * the vnode is being eliminated in vgone. In that case, we 572 * cannot grab it, so the process is awakened when the 573 * transition is completed, and an error code is returned to 574 * indicate that the vnode is no longer usable, possibly 575 * having been changed to a new file system type. 576 */ 577 int 578 vget(struct vnode *vp, int flags, struct proc *p) 579 { 580 int error, s, onfreelist; 581 582 /* 583 * If the vnode is in the process of being cleaned out for 584 * another use, we wait for the cleaning to finish and then 585 * return failure. Cleaning is determined by checking that 586 * the VXLOCK flag is set. 587 */ 588 589 if (vp->v_flag & VXLOCK) { 590 if (flags & LK_NOWAIT) { 591 return (EBUSY); 592 } 593 594 vp->v_flag |= VXWANT; 595 tsleep(vp, PINOD, "vget", 0); 596 return (ENOENT); 597 } 598 599 onfreelist = vp->v_bioflag & VBIOONFREELIST; 600 if (vp->v_usecount == 0 && onfreelist) { 601 s = splbio(); 602 if (vp->v_holdcnt > 0) 603 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 604 else 605 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 606 vp->v_bioflag &= ~VBIOONFREELIST; 607 splx(s); 608 } 609 610 vp->v_usecount++; 611 if (flags & LK_TYPE_MASK) { 612 if ((error = vn_lock(vp, flags, p)) != 0) { 613 vp->v_usecount--; 614 if (vp->v_usecount == 0 && onfreelist) 615 vputonfreelist(vp); 616 } 617 return (error); 618 } 619 620 return (0); 621 } 622 623 624 /* Vnode reference. */ 625 void 626 vref(struct vnode *vp) 627 { 628 #ifdef DIAGNOSTIC 629 if (vp->v_usecount == 0) 630 panic("vref used where vget required"); 631 #endif 632 vp->v_usecount++; 633 } 634 635 void 636 vputonfreelist(struct vnode *vp) 637 { 638 int s; 639 struct freelst *lst; 640 641 s = splbio(); 642 #ifdef DIAGNOSTIC 643 if (vp->v_usecount != 0) 644 panic("Use count is not zero!"); 645 646 if (vp->v_bioflag & VBIOONFREELIST) { 647 vprint("vnode already on free list: ", vp); 648 panic("vnode already on free list"); 649 } 650 #endif 651 652 vp->v_bioflag |= VBIOONFREELIST; 653 654 if (vp->v_holdcnt > 0) 655 lst = &vnode_hold_list; 656 else 657 lst = &vnode_free_list; 658 659 if (vp->v_type == VBAD) 660 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 661 else 662 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 663 664 splx(s); 665 } 666 667 /* 668 * vput(), just unlock and vrele() 669 */ 670 void 671 vput(struct vnode *vp) 672 { 673 struct proc *p = curproc; 674 675 #ifdef DIAGNOSTIC 676 if (vp == NULL) 677 panic("vput: null vp"); 678 #endif 679 680 #ifdef DIAGNOSTIC 681 if (vp->v_usecount == 0) { 682 vprint("vput: bad ref count", vp); 683 panic("vput: ref cnt"); 684 } 685 #endif 686 vp->v_usecount--; 687 if (vp->v_usecount > 0) { 688 VOP_UNLOCK(vp, 0, p); 689 return; 690 } 691 692 #ifdef DIAGNOSTIC 693 if (vp->v_writecount != 0) { 694 vprint("vput: bad writecount", vp); 695 panic("vput: v_writecount != 0"); 696 } 697 #endif 698 699 VOP_INACTIVE(vp, p); 700 701 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 702 vputonfreelist(vp); 703 } 704 705 /* 706 * Vnode release - use for active VNODES. 707 * If count drops to zero, call inactive routine and return to freelist. 708 * Returns 0 if it did not sleep. 709 */ 710 int 711 vrele(struct vnode *vp) 712 { 713 struct proc *p = curproc; 714 715 #ifdef DIAGNOSTIC 716 if (vp == NULL) 717 panic("vrele: null vp"); 718 #endif 719 #ifdef DIAGNOSTIC 720 if (vp->v_usecount == 0) { 721 vprint("vrele: bad ref count", vp); 722 panic("vrele: ref cnt"); 723 } 724 #endif 725 vp->v_usecount--; 726 if (vp->v_usecount > 0) { 727 return (0); 728 } 729 730 #ifdef DIAGNOSTIC 731 if (vp->v_writecount != 0) { 732 vprint("vrele: bad writecount", vp); 733 panic("vrele: v_writecount != 0"); 734 } 735 #endif 736 737 if (vn_lock(vp, LK_EXCLUSIVE, p)) { 738 #ifdef DIAGNOSTIC 739 vprint("vrele: cannot lock", vp); 740 #endif 741 return (1); 742 } 743 744 VOP_INACTIVE(vp, p); 745 746 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 747 vputonfreelist(vp); 748 return (1); 749 } 750 751 /* Page or buffer structure gets a reference. */ 752 void 753 vhold(struct vnode *vp) 754 { 755 /* 756 * If it is on the freelist and the hold count is currently 757 * zero, move it to the hold list. 758 */ 759 if ((vp->v_bioflag & VBIOONFREELIST) && 760 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 761 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 762 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 763 } 764 vp->v_holdcnt++; 765 } 766 767 /* Lose interest in a vnode. */ 768 void 769 vdrop(struct vnode *vp) 770 { 771 #ifdef DIAGNOSTIC 772 if (vp->v_holdcnt == 0) 773 panic("vdrop: zero holdcnt"); 774 #endif 775 776 vp->v_holdcnt--; 777 778 /* 779 * If it is on the holdlist and the hold count drops to 780 * zero, move it to the free list. 781 */ 782 if ((vp->v_bioflag & VBIOONFREELIST) && 783 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 784 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 785 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 786 } 787 } 788 789 /* 790 * Remove any vnodes in the vnode table belonging to mount point mp. 791 * 792 * If MNT_NOFORCE is specified, there should not be any active ones, 793 * return error if any are found (nb: this is a user error, not a 794 * system error). If MNT_FORCE is specified, detach any active vnodes 795 * that are found. 796 */ 797 #ifdef DEBUG 798 int busyprt = 0; /* print out busy vnodes */ 799 struct ctldebug debug1 = { "busyprt", &busyprt }; 800 #endif 801 802 int 803 vfs_mount_foreach_vnode(struct mount *mp, 804 int (*func)(struct vnode *, void *), void *arg) { 805 struct vnode *vp, *nvp; 806 int error = 0; 807 808 loop: 809 for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; vp = nvp) { 810 if (vp->v_mount != mp) 811 goto loop; 812 nvp = LIST_NEXT(vp, v_mntvnodes); 813 814 error = func(vp, arg); 815 816 if (error != 0) 817 break; 818 } 819 820 return (error); 821 } 822 823 struct vflush_args { 824 struct vnode *skipvp; 825 int busy; 826 int flags; 827 }; 828 829 int 830 vflush_vnode(struct vnode *vp, void *arg) { 831 struct vflush_args *va = arg; 832 struct proc *p = curproc; 833 834 if (vp == va->skipvp) { 835 return (0); 836 } 837 838 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 839 return (0); 840 } 841 842 /* 843 * If WRITECLOSE is set, only flush out regular file 844 * vnodes open for writing. 845 */ 846 if ((va->flags & WRITECLOSE) && 847 (vp->v_writecount == 0 || vp->v_type != VREG)) { 848 return (0); 849 } 850 851 /* 852 * With v_usecount == 0, all we need to do is clear 853 * out the vnode data structures and we are done. 854 */ 855 if (vp->v_usecount == 0) { 856 vgonel(vp, p); 857 return (0); 858 } 859 860 /* 861 * If FORCECLOSE is set, forcibly close the vnode. 862 * For block or character devices, revert to an 863 * anonymous device. For all other files, just kill them. 864 */ 865 if (va->flags & FORCECLOSE) { 866 if (vp->v_type != VBLK && vp->v_type != VCHR) { 867 vgonel(vp, p); 868 } else { 869 vclean(vp, 0, p); 870 vp->v_op = spec_vnodeop_p; 871 insmntque(vp, (struct mount *)0); 872 } 873 return (0); 874 } 875 876 #ifdef DEBUG 877 if (busyprt) 878 vprint("vflush: busy vnode", vp); 879 #endif 880 va->busy++; 881 return (0); 882 } 883 884 int 885 vflush(struct mount *mp, struct vnode *skipvp, int flags) 886 { 887 struct vflush_args va; 888 va.skipvp = skipvp; 889 va.busy = 0; 890 va.flags = flags; 891 892 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 893 894 if (va.busy) 895 return (EBUSY); 896 return (0); 897 } 898 899 /* 900 * Disassociate the underlying file system from a vnode. 901 */ 902 void 903 vclean(struct vnode *vp, int flags, struct proc *p) 904 { 905 int active; 906 907 /* 908 * Check to see if the vnode is in use. 909 * If so we have to reference it before we clean it out 910 * so that its count cannot fall to zero and generate a 911 * race against ourselves to recycle it. 912 */ 913 if ((active = vp->v_usecount) != 0) 914 vp->v_usecount++; 915 916 /* 917 * Prevent the vnode from being recycled or 918 * brought into use while we clean it out. 919 */ 920 if (vp->v_flag & VXLOCK) 921 panic("vclean: deadlock"); 922 vp->v_flag |= VXLOCK; 923 /* 924 * Even if the count is zero, the VOP_INACTIVE routine may still 925 * have the object locked while it cleans it out. The VOP_LOCK 926 * ensures that the VOP_INACTIVE routine is done with its work. 927 * For active vnodes, it ensures that no other activity can 928 * occur while the underlying object is being cleaned out. 929 */ 930 VOP_LOCK(vp, LK_DRAIN, p); 931 932 /* 933 * Clean out any VM data associated with the vnode. 934 */ 935 uvm_vnp_terminate(vp); 936 /* 937 * Clean out any buffers associated with the vnode. 938 */ 939 if (flags & DOCLOSE) 940 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 941 /* 942 * If purging an active vnode, it must be closed and 943 * deactivated before being reclaimed. Note that the 944 * VOP_INACTIVE will unlock the vnode 945 */ 946 if (active) { 947 if (flags & DOCLOSE) 948 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 949 VOP_INACTIVE(vp, p); 950 } else { 951 /* 952 * Any other processes trying to obtain this lock must first 953 * wait for VXLOCK to clear, then call the new lock operation. 954 */ 955 VOP_UNLOCK(vp, 0, p); 956 } 957 958 /* 959 * Reclaim the vnode. 960 */ 961 if (VOP_RECLAIM(vp, p)) 962 panic("vclean: cannot reclaim"); 963 if (active) { 964 vp->v_usecount--; 965 if (vp->v_usecount == 0) { 966 if (vp->v_holdcnt > 0) 967 panic("vclean: not clean"); 968 vputonfreelist(vp); 969 } 970 } 971 cache_purge(vp); 972 973 /* 974 * Done with purge, notify sleepers of the grim news. 975 */ 976 vp->v_op = dead_vnodeop_p; 977 VN_KNOTE(vp, NOTE_REVOKE); 978 vp->v_tag = VT_NON; 979 vp->v_flag &= ~VXLOCK; 980 #ifdef VFSDEBUG 981 vp->v_flag &= ~VLOCKSWORK; 982 #endif 983 if (vp->v_flag & VXWANT) { 984 vp->v_flag &= ~VXWANT; 985 wakeup(vp); 986 } 987 } 988 989 /* 990 * Recycle an unused vnode to the front of the free list. 991 */ 992 int 993 vrecycle(struct vnode *vp, struct proc *p) 994 { 995 if (vp->v_usecount == 0) { 996 vgonel(vp, p); 997 return (1); 998 } 999 return (0); 1000 } 1001 1002 /* 1003 * Eliminate all activity associated with a vnode 1004 * in preparation for reuse. 1005 */ 1006 void 1007 vgone(struct vnode *vp) 1008 { 1009 struct proc *p = curproc; 1010 vgonel(vp, p); 1011 } 1012 1013 /* 1014 * vgone, with struct proc. 1015 */ 1016 void 1017 vgonel(struct vnode *vp, struct proc *p) 1018 { 1019 struct vnode *vq; 1020 struct vnode *vx; 1021 1022 /* 1023 * If a vgone (or vclean) is already in progress, 1024 * wait until it is done and return. 1025 */ 1026 if (vp->v_flag & VXLOCK) { 1027 vp->v_flag |= VXWANT; 1028 tsleep(vp, PINOD, "vgone", 0); 1029 return; 1030 } 1031 1032 /* 1033 * Clean out the filesystem specific data. 1034 */ 1035 vclean(vp, DOCLOSE, p); 1036 /* 1037 * Delete from old mount point vnode list, if on one. 1038 */ 1039 if (vp->v_mount != NULL) 1040 insmntque(vp, (struct mount *)0); 1041 /* 1042 * If special device, remove it from special device alias list 1043 * if it is on one. 1044 */ 1045 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1046 if (*vp->v_hashchain == vp) { 1047 *vp->v_hashchain = vp->v_specnext; 1048 } else { 1049 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1050 if (vq->v_specnext != vp) 1051 continue; 1052 vq->v_specnext = vp->v_specnext; 1053 break; 1054 } 1055 if (vq == NULL) 1056 panic("missing bdev"); 1057 } 1058 if (vp->v_flag & VALIASED) { 1059 vx = NULL; 1060 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1061 if (vq->v_rdev != vp->v_rdev || 1062 vq->v_type != vp->v_type) 1063 continue; 1064 if (vx) 1065 break; 1066 vx = vq; 1067 } 1068 if (vx == NULL) 1069 panic("missing alias"); 1070 if (vq == NULL) 1071 vx->v_flag &= ~VALIASED; 1072 vp->v_flag &= ~VALIASED; 1073 } 1074 free(vp->v_specinfo, M_VNODE); 1075 vp->v_specinfo = NULL; 1076 } 1077 /* 1078 * If it is on the freelist and not already at the head, 1079 * move it to the head of the list. 1080 */ 1081 vp->v_type = VBAD; 1082 1083 /* 1084 * Move onto the free list, unless we were called from 1085 * getnewvnode and we're not on any free list 1086 */ 1087 if (vp->v_usecount == 0 && 1088 (vp->v_bioflag & VBIOONFREELIST)) { 1089 int s; 1090 1091 s = splbio(); 1092 1093 if (vp->v_holdcnt > 0) 1094 panic("vgonel: not clean"); 1095 1096 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1097 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1098 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1099 } 1100 splx(s); 1101 } 1102 } 1103 1104 /* 1105 * Lookup a vnode by device number. 1106 */ 1107 int 1108 vfinddev(dev_t dev, enum vtype type, struct vnode **vpp) 1109 { 1110 struct vnode *vp; 1111 int rc =0; 1112 1113 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1114 if (dev != vp->v_rdev || type != vp->v_type) 1115 continue; 1116 *vpp = vp; 1117 rc = 1; 1118 break; 1119 } 1120 return (rc); 1121 } 1122 1123 /* 1124 * Revoke all the vnodes corresponding to the specified minor number 1125 * range (endpoints inclusive) of the specified major. 1126 */ 1127 void 1128 vdevgone(int maj, int minl, int minh, enum vtype type) 1129 { 1130 struct vnode *vp; 1131 int mn; 1132 1133 for (mn = minl; mn <= minh; mn++) 1134 if (vfinddev(makedev(maj, mn), type, &vp)) 1135 VOP_REVOKE(vp, REVOKEALL); 1136 } 1137 1138 /* 1139 * Calculate the total number of references to a special device. 1140 */ 1141 int 1142 vcount(struct vnode *vp) 1143 { 1144 struct vnode *vq, *vnext; 1145 int count; 1146 1147 loop: 1148 if ((vp->v_flag & VALIASED) == 0) 1149 return (vp->v_usecount); 1150 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1151 vnext = vq->v_specnext; 1152 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1153 continue; 1154 /* 1155 * Alias, but not in use, so flush it out. 1156 */ 1157 if (vq->v_usecount == 0 && vq != vp) { 1158 vgone(vq); 1159 goto loop; 1160 } 1161 count += vq->v_usecount; 1162 } 1163 return (count); 1164 } 1165 1166 #if defined(DEBUG) || defined(DIAGNOSTIC) 1167 /* 1168 * Print out a description of a vnode. 1169 */ 1170 static char *typename[] = 1171 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1172 1173 void 1174 vprint(char *label, struct vnode *vp) 1175 { 1176 char buf[64]; 1177 1178 if (label != NULL) 1179 printf("%s: ", label); 1180 printf("%p, type %s, use %u, write %u, hold %u,", 1181 vp, typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1182 vp->v_holdcnt); 1183 buf[0] = '\0'; 1184 if (vp->v_flag & VROOT) 1185 strlcat(buf, "|VROOT", sizeof buf); 1186 if (vp->v_flag & VTEXT) 1187 strlcat(buf, "|VTEXT", sizeof buf); 1188 if (vp->v_flag & VSYSTEM) 1189 strlcat(buf, "|VSYSTEM", sizeof buf); 1190 if (vp->v_flag & VXLOCK) 1191 strlcat(buf, "|VXLOCK", sizeof buf); 1192 if (vp->v_flag & VXWANT) 1193 strlcat(buf, "|VXWANT", sizeof buf); 1194 if (vp->v_bioflag & VBIOWAIT) 1195 strlcat(buf, "|VBIOWAIT", sizeof buf); 1196 if (vp->v_bioflag & VBIOONFREELIST) 1197 strlcat(buf, "|VBIOONFREELIST", sizeof buf); 1198 if (vp->v_bioflag & VBIOONSYNCLIST) 1199 strlcat(buf, "|VBIOONSYNCLIST", sizeof buf); 1200 if (vp->v_flag & VALIASED) 1201 strlcat(buf, "|VALIASED", sizeof buf); 1202 if (buf[0] != '\0') 1203 printf(" flags (%s)", &buf[1]); 1204 if (vp->v_data == NULL) { 1205 printf("\n"); 1206 } else { 1207 printf("\n\t"); 1208 VOP_PRINT(vp); 1209 } 1210 } 1211 #endif /* DEBUG || DIAGNOSTIC */ 1212 1213 #ifdef DEBUG 1214 /* 1215 * List all of the locked vnodes in the system. 1216 * Called when debugging the kernel. 1217 */ 1218 void 1219 printlockedvnodes(void) 1220 { 1221 struct mount *mp, *nmp; 1222 struct vnode *vp; 1223 1224 printf("Locked vnodes\n"); 1225 1226 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1227 mp = nmp) { 1228 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) { 1229 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1230 continue; 1231 } 1232 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 1233 if (VOP_ISLOCKED(vp)) 1234 vprint((char *)0, vp); 1235 } 1236 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1237 vfs_unbusy(mp); 1238 } 1239 1240 } 1241 #endif 1242 1243 /* 1244 * Top level filesystem related information gathering. 1245 */ 1246 int 1247 vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1248 size_t newlen, struct proc *p) 1249 { 1250 struct vfsconf *vfsp, *tmpvfsp; 1251 int ret; 1252 1253 /* all sysctl names at this level are at least name and field */ 1254 if (namelen < 2) 1255 return (ENOTDIR); /* overloaded */ 1256 1257 if (name[0] != VFS_GENERIC) { 1258 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1259 if (vfsp->vfc_typenum == name[0]) 1260 break; 1261 1262 if (vfsp == NULL) 1263 return (EOPNOTSUPP); 1264 1265 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1266 oldp, oldlenp, newp, newlen, p)); 1267 } 1268 1269 switch (name[1]) { 1270 case VFS_MAXTYPENUM: 1271 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1272 1273 case VFS_CONF: 1274 if (namelen < 3) 1275 return (ENOTDIR); /* overloaded */ 1276 1277 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1278 if (vfsp->vfc_typenum == name[2]) 1279 break; 1280 1281 if (vfsp == NULL) 1282 return (EOPNOTSUPP); 1283 1284 /* Make a copy, clear out kernel pointers */ 1285 tmpvfsp = malloc(sizeof(*tmpvfsp), M_TEMP, M_WAITOK); 1286 bcopy(vfsp, tmpvfsp, sizeof(*tmpvfsp)); 1287 tmpvfsp->vfc_vfsops = NULL; 1288 tmpvfsp->vfc_next = NULL; 1289 1290 ret = sysctl_rdstruct(oldp, oldlenp, newp, tmpvfsp, 1291 sizeof(struct vfsconf)); 1292 1293 free(tmpvfsp, M_TEMP); 1294 return (ret); 1295 case VFS_BCACHESTAT: /* buffer cache statistics */ 1296 ret = sysctl_rdstruct(oldp, oldlenp, newp, &bcstats, 1297 sizeof(struct bcachestats)); 1298 return(ret); 1299 } 1300 return (EOPNOTSUPP); 1301 } 1302 1303 int kinfo_vdebug = 1; 1304 #define KINFO_VNODESLOP 10 1305 /* 1306 * Dump vnode list (via sysctl). 1307 * Copyout address of vnode followed by vnode. 1308 */ 1309 /* ARGSUSED */ 1310 int 1311 sysctl_vnode(char *where, size_t *sizep, struct proc *p) 1312 { 1313 struct mount *mp, *nmp; 1314 struct vnode *vp, *nvp; 1315 char *bp = where, *savebp; 1316 char *ewhere; 1317 int error; 1318 1319 if (where == NULL) { 1320 *sizep = (numvnodes + KINFO_VNODESLOP) * sizeof(struct e_vnode); 1321 return (0); 1322 } 1323 ewhere = where + *sizep; 1324 1325 for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1326 mp = nmp) { 1327 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) { 1328 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1329 continue; 1330 } 1331 savebp = bp; 1332 again: 1333 for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; 1334 vp = nvp) { 1335 /* 1336 * Check that the vp is still associated with 1337 * this filesystem. RACE: could have been 1338 * recycled onto the same filesystem. 1339 */ 1340 if (vp->v_mount != mp) { 1341 if (kinfo_vdebug) 1342 printf("kinfo: vp changed\n"); 1343 bp = savebp; 1344 goto again; 1345 } 1346 nvp = LIST_NEXT(vp, v_mntvnodes); 1347 if (bp + sizeof(struct e_vnode) > ewhere) { 1348 *sizep = bp - where; 1349 vfs_unbusy(mp); 1350 return (ENOMEM); 1351 } 1352 if ((error = copyout(&vp, 1353 &((struct e_vnode *)bp)->vptr, 1354 sizeof(struct vnode *))) || 1355 (error = copyout(vp, 1356 &((struct e_vnode *)bp)->vnode, 1357 sizeof(struct vnode)))) { 1358 vfs_unbusy(mp); 1359 return (error); 1360 } 1361 bp += sizeof(struct e_vnode); 1362 } 1363 1364 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1365 vfs_unbusy(mp); 1366 } 1367 1368 *sizep = bp - where; 1369 1370 return (0); 1371 } 1372 1373 /* 1374 * Check to see if a filesystem is mounted on a block device. 1375 */ 1376 int 1377 vfs_mountedon(struct vnode *vp) 1378 { 1379 struct vnode *vq; 1380 int error = 0; 1381 1382 if (vp->v_specmountpoint != NULL) 1383 return (EBUSY); 1384 if (vp->v_flag & VALIASED) { 1385 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1386 if (vq->v_rdev != vp->v_rdev || 1387 vq->v_type != vp->v_type) 1388 continue; 1389 if (vq->v_specmountpoint != NULL) { 1390 error = EBUSY; 1391 break; 1392 } 1393 } 1394 } 1395 return (error); 1396 } 1397 1398 /* 1399 * Build hash lists of net addresses and hang them off the mount point. 1400 * Called by ufs_mount() to set up the lists of export addresses. 1401 */ 1402 int 1403 vfs_hang_addrlist(struct mount *mp, struct netexport *nep, 1404 struct export_args *argp) 1405 { 1406 struct netcred *np; 1407 struct radix_node_head *rnh; 1408 int i; 1409 struct radix_node *rn; 1410 struct sockaddr *saddr, *smask = 0; 1411 struct domain *dom; 1412 int error; 1413 1414 if (argp->ex_addrlen == 0) { 1415 if (mp->mnt_flag & MNT_DEFEXPORTED) 1416 return (EPERM); 1417 np = &nep->ne_defexported; 1418 np->netc_exflags = argp->ex_flags; 1419 np->netc_anon = argp->ex_anon; 1420 np->netc_anon.cr_ref = 1; 1421 mp->mnt_flag |= MNT_DEFEXPORTED; 1422 return (0); 1423 } 1424 if (argp->ex_addrlen > MLEN || argp->ex_masklen > MLEN || 1425 argp->ex_addrlen < 0 || argp->ex_masklen < 0) 1426 return (EINVAL); 1427 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1428 np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK|M_ZERO); 1429 saddr = (struct sockaddr *)(np + 1); 1430 error = copyin(argp->ex_addr, saddr, argp->ex_addrlen); 1431 if (error) 1432 goto out; 1433 if (saddr->sa_len > argp->ex_addrlen) 1434 saddr->sa_len = argp->ex_addrlen; 1435 if (argp->ex_masklen) { 1436 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1437 error = copyin(argp->ex_mask, smask, argp->ex_masklen); 1438 if (error) 1439 goto out; 1440 if (smask->sa_len > argp->ex_masklen) 1441 smask->sa_len = argp->ex_masklen; 1442 } 1443 i = saddr->sa_family; 1444 if (i < 0 || i > AF_MAX) { 1445 error = EINVAL; 1446 goto out; 1447 } 1448 if ((rnh = nep->ne_rtable[i]) == 0) { 1449 /* 1450 * Seems silly to initialize every AF when most are not 1451 * used, do so on demand here 1452 */ 1453 for (dom = domains; dom; dom = dom->dom_next) 1454 if (dom->dom_family == i && dom->dom_rtattach) { 1455 dom->dom_rtattach((void **)&nep->ne_rtable[i], 1456 dom->dom_rtoffset); 1457 break; 1458 } 1459 if ((rnh = nep->ne_rtable[i]) == 0) { 1460 error = ENOBUFS; 1461 goto out; 1462 } 1463 } 1464 rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, 1465 np->netc_rnodes, 0); 1466 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1467 error = EPERM; 1468 goto out; 1469 } 1470 np->netc_exflags = argp->ex_flags; 1471 np->netc_anon = argp->ex_anon; 1472 np->netc_anon.cr_ref = 1; 1473 return (0); 1474 out: 1475 free(np, M_NETADDR); 1476 return (error); 1477 } 1478 1479 /* ARGSUSED */ 1480 int 1481 vfs_free_netcred(struct radix_node *rn, void *w) 1482 { 1483 struct radix_node_head *rnh = (struct radix_node_head *)w; 1484 1485 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh, NULL); 1486 free(rn, M_NETADDR); 1487 return (0); 1488 } 1489 1490 /* 1491 * Free the net address hash lists that are hanging off the mount points. 1492 */ 1493 void 1494 vfs_free_addrlist(struct netexport *nep) 1495 { 1496 int i; 1497 struct radix_node_head *rnh; 1498 1499 for (i = 0; i <= AF_MAX; i++) 1500 if ((rnh = nep->ne_rtable[i]) != NULL) { 1501 (*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh); 1502 free(rnh, M_RTABLE); 1503 nep->ne_rtable[i] = 0; 1504 } 1505 } 1506 1507 int 1508 vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp) 1509 { 1510 int error; 1511 1512 if (argp->ex_flags & MNT_DELEXPORT) { 1513 vfs_free_addrlist(nep); 1514 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1515 } 1516 if (argp->ex_flags & MNT_EXPORTED) { 1517 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1518 return (error); 1519 mp->mnt_flag |= MNT_EXPORTED; 1520 } 1521 return (0); 1522 } 1523 1524 struct netcred * 1525 vfs_export_lookup(struct mount *mp, struct netexport *nep, struct mbuf *nam) 1526 { 1527 struct netcred *np; 1528 struct radix_node_head *rnh; 1529 struct sockaddr *saddr; 1530 1531 np = NULL; 1532 if (mp->mnt_flag & MNT_EXPORTED) { 1533 /* 1534 * Lookup in the export list first. 1535 */ 1536 if (nam != NULL) { 1537 saddr = mtod(nam, struct sockaddr *); 1538 rnh = nep->ne_rtable[saddr->sa_family]; 1539 if (rnh != NULL) { 1540 np = (struct netcred *) 1541 (*rnh->rnh_matchaddr)((caddr_t)saddr, 1542 rnh); 1543 if (np && np->netc_rnodes->rn_flags & RNF_ROOT) 1544 np = NULL; 1545 } 1546 } 1547 /* 1548 * If no address match, use the default if it exists. 1549 */ 1550 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1551 np = &nep->ne_defexported; 1552 } 1553 return (np); 1554 } 1555 1556 /* 1557 * Do the usual access checking. 1558 * file_mode, uid and gid are from the vnode in question, 1559 * while acc_mode and cred are from the VOP_ACCESS parameter list 1560 */ 1561 int 1562 vaccess(enum vtype type, mode_t file_mode, uid_t uid, gid_t gid, 1563 mode_t acc_mode, struct ucred *cred) 1564 { 1565 mode_t mask; 1566 1567 /* User id 0 always gets read/write access. */ 1568 if (cred->cr_uid == 0) { 1569 /* For VEXEC, at least one of the execute bits must be set. */ 1570 if ((acc_mode & VEXEC) && type != VDIR && 1571 (file_mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0) 1572 return EACCES; 1573 return 0; 1574 } 1575 1576 mask = 0; 1577 1578 /* Otherwise, check the owner. */ 1579 if (cred->cr_uid == uid) { 1580 if (acc_mode & VEXEC) 1581 mask |= S_IXUSR; 1582 if (acc_mode & VREAD) 1583 mask |= S_IRUSR; 1584 if (acc_mode & VWRITE) 1585 mask |= S_IWUSR; 1586 return (file_mode & mask) == mask ? 0 : EACCES; 1587 } 1588 1589 /* Otherwise, check the groups. */ 1590 if (cred->cr_gid == gid || groupmember(gid, cred)) { 1591 if (acc_mode & VEXEC) 1592 mask |= S_IXGRP; 1593 if (acc_mode & VREAD) 1594 mask |= S_IRGRP; 1595 if (acc_mode & VWRITE) 1596 mask |= S_IWGRP; 1597 return (file_mode & mask) == mask ? 0 : EACCES; 1598 } 1599 1600 /* Otherwise, check everyone else. */ 1601 if (acc_mode & VEXEC) 1602 mask |= S_IXOTH; 1603 if (acc_mode & VREAD) 1604 mask |= S_IROTH; 1605 if (acc_mode & VWRITE) 1606 mask |= S_IWOTH; 1607 return (file_mode & mask) == mask ? 0 : EACCES; 1608 } 1609 1610 /* 1611 * Unmount all file systems. 1612 * We traverse the list in reverse order under the assumption that doing so 1613 * will avoid needing to worry about dependencies. 1614 */ 1615 void 1616 vfs_unmountall(void) 1617 { 1618 struct mount *mp, *nmp; 1619 int allerror, error, again = 1; 1620 1621 retry: 1622 allerror = 0; 1623 for (mp = CIRCLEQ_LAST(&mountlist); mp != CIRCLEQ_END(&mountlist); 1624 mp = nmp) { 1625 nmp = CIRCLEQ_PREV(mp, mnt_list); 1626 if ((vfs_busy(mp, VB_WRITE|VB_NOWAIT)) != 0) 1627 continue; 1628 if ((error = dounmount(mp, MNT_FORCE, curproc, NULL)) != 0) { 1629 printf("unmount of %s failed with error %d\n", 1630 mp->mnt_stat.f_mntonname, error); 1631 allerror = 1; 1632 } 1633 } 1634 1635 if (allerror) { 1636 printf("WARNING: some file systems would not unmount\n"); 1637 if (again) { 1638 printf("retrying\n"); 1639 again = 0; 1640 goto retry; 1641 } 1642 } 1643 } 1644 1645 /* 1646 * Sync and unmount file systems before shutting down. 1647 */ 1648 void 1649 vfs_shutdown(void) 1650 { 1651 #ifdef ACCOUNTING 1652 extern void acct_shutdown(void); 1653 1654 acct_shutdown(); 1655 #endif 1656 1657 /* XXX Should suspend scheduling. */ 1658 (void) spl0(); 1659 1660 printf("syncing disks... "); 1661 1662 if (panicstr == 0) { 1663 /* Sync before unmount, in case we hang on something. */ 1664 sys_sync(&proc0, (void *)0, (register_t *)0); 1665 1666 /* Unmount file systems. */ 1667 vfs_unmountall(); 1668 } 1669 1670 if (vfs_syncwait(1)) 1671 printf("giving up\n"); 1672 else 1673 printf("done\n"); 1674 } 1675 1676 /* 1677 * perform sync() operation and wait for buffers to flush. 1678 * assumptions: called w/ scheduler disabled and physical io enabled 1679 * for now called at spl0() XXX 1680 */ 1681 int 1682 vfs_syncwait(int verbose) 1683 { 1684 struct buf *bp; 1685 int iter, nbusy, dcount, s; 1686 struct proc *p; 1687 1688 p = curproc? curproc : &proc0; 1689 sys_sync(p, (void *)0, (register_t *)0); 1690 1691 /* Wait for sync to finish. */ 1692 dcount = 10000; 1693 for (iter = 0; iter < 20; iter++) { 1694 nbusy = 0; 1695 LIST_FOREACH(bp, &bufhead, b_list) { 1696 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1697 nbusy++; 1698 /* 1699 * With soft updates, some buffers that are 1700 * written will be remarked as dirty until other 1701 * buffers are written. 1702 */ 1703 if (bp->b_flags & B_DELWRI) { 1704 s = splbio(); 1705 buf_acquire(bp); 1706 splx(s); 1707 nbusy++; 1708 bawrite(bp); 1709 if (dcount-- <= 0) { 1710 if (verbose) 1711 printf("softdep "); 1712 return 1; 1713 } 1714 } 1715 } 1716 if (nbusy == 0) 1717 break; 1718 if (verbose) 1719 printf("%d ", nbusy); 1720 DELAY(40000 * iter); 1721 } 1722 1723 return nbusy; 1724 } 1725 1726 /* 1727 * posix file system related system variables. 1728 */ 1729 int 1730 fs_posix_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, 1731 void *newp, size_t newlen, struct proc *p) 1732 { 1733 /* all sysctl names at this level are terminal */ 1734 if (namelen != 1) 1735 return (ENOTDIR); 1736 1737 switch (name[0]) { 1738 case FS_POSIX_SETUID: 1739 if (newp && securelevel > 0) 1740 return (EPERM); 1741 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1742 default: 1743 return (EOPNOTSUPP); 1744 } 1745 /* NOTREACHED */ 1746 } 1747 1748 /* 1749 * file system related system variables. 1750 */ 1751 int 1752 fs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1753 size_t newlen, struct proc *p) 1754 { 1755 sysctlfn *fn; 1756 1757 switch (name[0]) { 1758 case FS_POSIX: 1759 fn = fs_posix_sysctl; 1760 break; 1761 default: 1762 return (EOPNOTSUPP); 1763 } 1764 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1765 } 1766 1767 1768 /* 1769 * Routines dealing with vnodes and buffers 1770 */ 1771 1772 /* 1773 * Wait for all outstanding I/Os to complete 1774 * 1775 * Manipulates v_numoutput. Must be called at splbio() 1776 */ 1777 int 1778 vwaitforio(struct vnode *vp, int slpflag, char *wmesg, int timeo) 1779 { 1780 int error = 0; 1781 1782 splassert(IPL_BIO); 1783 1784 while (vp->v_numoutput) { 1785 vp->v_bioflag |= VBIOWAIT; 1786 error = tsleep(&vp->v_numoutput, 1787 slpflag | (PRIBIO + 1), wmesg, timeo); 1788 if (error) 1789 break; 1790 } 1791 1792 return (error); 1793 } 1794 1795 /* 1796 * Update outstanding I/O count and do wakeup if requested. 1797 * 1798 * Manipulates v_numoutput. Must be called at splbio() 1799 */ 1800 void 1801 vwakeup(struct vnode *vp) 1802 { 1803 splassert(IPL_BIO); 1804 1805 if (vp != NULL) { 1806 if (vp->v_numoutput-- == 0) 1807 panic("vwakeup: neg numoutput"); 1808 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1809 vp->v_bioflag &= ~VBIOWAIT; 1810 wakeup(&vp->v_numoutput); 1811 } 1812 } 1813 } 1814 1815 /* 1816 * Flush out and invalidate all buffers associated with a vnode. 1817 * Called with the underlying object locked. 1818 */ 1819 int 1820 vinvalbuf(struct vnode *vp, int flags, struct ucred *cred, struct proc *p, 1821 int slpflag, int slptimeo) 1822 { 1823 struct buf *bp; 1824 struct buf *nbp, *blist; 1825 int s, error; 1826 1827 #ifdef VFSDEBUG 1828 if ((vp->v_flag & VLOCKSWORK) && !VOP_ISLOCKED(vp)) 1829 panic("vinvalbuf(): vp isn't locked"); 1830 #endif 1831 1832 if (flags & V_SAVE) { 1833 s = splbio(); 1834 vwaitforio(vp, 0, "vinvalbuf", 0); 1835 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1836 splx(s); 1837 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1838 return (error); 1839 s = splbio(); 1840 if (vp->v_numoutput > 0 || 1841 !LIST_EMPTY(&vp->v_dirtyblkhd)) 1842 panic("vinvalbuf: dirty bufs"); 1843 } 1844 splx(s); 1845 } 1846 loop: 1847 s = splbio(); 1848 for (;;) { 1849 if ((blist = LIST_FIRST(&vp->v_cleanblkhd)) && 1850 (flags & V_SAVEMETA)) 1851 while (blist && blist->b_lblkno < 0) 1852 blist = LIST_NEXT(blist, b_vnbufs); 1853 if (blist == NULL && 1854 (blist = LIST_FIRST(&vp->v_dirtyblkhd)) && 1855 (flags & V_SAVEMETA)) 1856 while (blist && blist->b_lblkno < 0) 1857 blist = LIST_NEXT(blist, b_vnbufs); 1858 if (!blist) 1859 break; 1860 1861 for (bp = blist; bp; bp = nbp) { 1862 nbp = LIST_NEXT(bp, b_vnbufs); 1863 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1864 continue; 1865 if (bp->b_flags & B_BUSY) { 1866 bp->b_flags |= B_WANTED; 1867 error = tsleep(bp, slpflag | (PRIBIO + 1), 1868 "vinvalbuf", slptimeo); 1869 if (error) { 1870 splx(s); 1871 return (error); 1872 } 1873 break; 1874 } 1875 buf_acquire(bp); 1876 /* 1877 * XXX Since there are no node locks for NFS, I believe 1878 * there is a slight chance that a delayed write will 1879 * occur while sleeping just above, so check for it. 1880 */ 1881 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1882 splx(s); 1883 (void) VOP_BWRITE(bp); 1884 goto loop; 1885 } 1886 bp->b_flags |= B_INVAL; 1887 brelse(bp); 1888 } 1889 } 1890 if (!(flags & V_SAVEMETA) && 1891 (!LIST_EMPTY(&vp->v_dirtyblkhd) || !LIST_EMPTY(&vp->v_cleanblkhd))) 1892 panic("vinvalbuf: flush failed"); 1893 splx(s); 1894 return (0); 1895 } 1896 1897 void 1898 vflushbuf(struct vnode *vp, int sync) 1899 { 1900 struct buf *bp, *nbp; 1901 int s; 1902 1903 loop: 1904 s = splbio(); 1905 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); 1906 bp != LIST_END(&vp->v_dirtyblkhd); bp = nbp) { 1907 nbp = LIST_NEXT(bp, b_vnbufs); 1908 if ((bp->b_flags & B_BUSY)) 1909 continue; 1910 if ((bp->b_flags & B_DELWRI) == 0) 1911 panic("vflushbuf: not dirty"); 1912 buf_acquire(bp); 1913 splx(s); 1914 /* 1915 * Wait for I/O associated with indirect blocks to complete, 1916 * since there is no way to quickly wait for them below. 1917 */ 1918 if (bp->b_vp == vp || sync == 0) 1919 (void) bawrite(bp); 1920 else 1921 (void) bwrite(bp); 1922 goto loop; 1923 } 1924 if (sync == 0) { 1925 splx(s); 1926 return; 1927 } 1928 vwaitforio(vp, 0, "vflushbuf", 0); 1929 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1930 splx(s); 1931 #ifdef DIAGNOSTIC 1932 vprint("vflushbuf: dirty", vp); 1933 #endif 1934 goto loop; 1935 } 1936 splx(s); 1937 } 1938 1939 /* 1940 * Associate a buffer with a vnode. 1941 * 1942 * Manipulates buffer vnode queues. Must be called at splbio(). 1943 */ 1944 void 1945 bgetvp(struct vnode *vp, struct buf *bp) 1946 { 1947 splassert(IPL_BIO); 1948 1949 1950 if (bp->b_vp) 1951 panic("bgetvp: not free"); 1952 vhold(vp); 1953 bp->b_vp = vp; 1954 if (vp->v_type == VBLK || vp->v_type == VCHR) 1955 bp->b_dev = vp->v_rdev; 1956 else 1957 bp->b_dev = NODEV; 1958 /* 1959 * Insert onto list for new vnode. 1960 */ 1961 bufinsvn(bp, &vp->v_cleanblkhd); 1962 } 1963 1964 /* 1965 * Disassociate a buffer from a vnode. 1966 * 1967 * Manipulates vnode buffer queues. Must be called at splbio(). 1968 */ 1969 void 1970 brelvp(struct buf *bp) 1971 { 1972 struct vnode *vp; 1973 1974 splassert(IPL_BIO); 1975 1976 if ((vp = bp->b_vp) == (struct vnode *) 0) 1977 panic("brelvp: NULL"); 1978 /* 1979 * Delete from old vnode list, if on one. 1980 */ 1981 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 1982 bufremvn(bp); 1983 if ((vp->v_bioflag & VBIOONSYNCLIST) && 1984 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 1985 vp->v_bioflag &= ~VBIOONSYNCLIST; 1986 LIST_REMOVE(vp, v_synclist); 1987 } 1988 bp->b_vp = NULL; 1989 1990 vdrop(vp); 1991 } 1992 1993 /* 1994 * Replaces the current vnode associated with the buffer, if any, 1995 * with a new vnode. 1996 * 1997 * If an output I/O is pending on the buffer, the old vnode 1998 * I/O count is adjusted. 1999 * 2000 * Ignores vnode buffer queues. Must be called at splbio(). 2001 */ 2002 void 2003 buf_replacevnode(struct buf *bp, struct vnode *newvp) 2004 { 2005 struct vnode *oldvp = bp->b_vp; 2006 2007 splassert(IPL_BIO); 2008 2009 if (oldvp) 2010 brelvp(bp); 2011 2012 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2013 newvp->v_numoutput++; /* put it on swapdev */ 2014 vwakeup(oldvp); 2015 } 2016 2017 bgetvp(newvp, bp); 2018 bufremvn(bp); 2019 } 2020 2021 /* 2022 * Used to assign buffers to the appropriate clean or dirty list on 2023 * the vnode and to add newly dirty vnodes to the appropriate 2024 * filesystem syncer list. 2025 * 2026 * Manipulates vnode buffer queues. Must be called at splbio(). 2027 */ 2028 void 2029 reassignbuf(struct buf *bp) 2030 { 2031 struct buflists *listheadp; 2032 int delay; 2033 struct vnode *vp = bp->b_vp; 2034 2035 splassert(IPL_BIO); 2036 2037 /* 2038 * Delete from old vnode list, if on one. 2039 */ 2040 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2041 bufremvn(bp); 2042 2043 /* 2044 * If dirty, put on list of dirty buffers; 2045 * otherwise insert onto list of clean buffers. 2046 */ 2047 if ((bp->b_flags & B_DELWRI) == 0) { 2048 listheadp = &vp->v_cleanblkhd; 2049 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2050 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2051 vp->v_bioflag &= ~VBIOONSYNCLIST; 2052 LIST_REMOVE(vp, v_synclist); 2053 } 2054 } else { 2055 listheadp = &vp->v_dirtyblkhd; 2056 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2057 switch (vp->v_type) { 2058 case VDIR: 2059 delay = syncdelay / 2; 2060 break; 2061 case VBLK: 2062 if (vp->v_specmountpoint != NULL) { 2063 delay = syncdelay / 3; 2064 break; 2065 } 2066 /* FALLTHROUGH */ 2067 default: 2068 delay = syncdelay; 2069 } 2070 vn_syncer_add_to_worklist(vp, delay); 2071 } 2072 } 2073 bufinsvn(bp, listheadp); 2074 } 2075 2076 int 2077 vfs_register(struct vfsconf *vfs) 2078 { 2079 struct vfsconf *vfsp; 2080 struct vfsconf **vfspp; 2081 2082 #ifdef DIAGNOSTIC 2083 /* Paranoia? */ 2084 if (vfs->vfc_refcount != 0) 2085 printf("vfs_register called with vfc_refcount > 0\n"); 2086 #endif 2087 2088 /* Check if filesystem already known */ 2089 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2090 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2091 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2092 return (EEXIST); 2093 2094 if (vfs->vfc_typenum > maxvfsconf) 2095 maxvfsconf = vfs->vfc_typenum; 2096 2097 vfs->vfc_next = NULL; 2098 2099 /* Add to the end of the list */ 2100 *vfspp = vfs; 2101 2102 /* Call vfs_init() */ 2103 if (vfs->vfc_vfsops->vfs_init) 2104 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2105 2106 return 0; 2107 } 2108 2109 int 2110 vfs_unregister(struct vfsconf *vfs) 2111 { 2112 struct vfsconf *vfsp; 2113 struct vfsconf **vfspp; 2114 int maxtypenum; 2115 2116 /* Find our vfsconf struct */ 2117 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2118 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2119 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2120 break; 2121 } 2122 2123 if (!vfsp) /* Not found */ 2124 return (ENOENT); 2125 2126 if (vfsp->vfc_refcount) /* In use */ 2127 return (EBUSY); 2128 2129 /* Remove from list and free */ 2130 *vfspp = vfsp->vfc_next; 2131 2132 maxtypenum = 0; 2133 2134 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2135 if (vfsp->vfc_typenum > maxtypenum) 2136 maxtypenum = vfsp->vfc_typenum; 2137 2138 maxvfsconf = maxtypenum; 2139 return 0; 2140 } 2141 2142 /* 2143 * Check if vnode represents a disk device 2144 */ 2145 int 2146 vn_isdisk(struct vnode *vp, int *errp) 2147 { 2148 if (vp->v_type != VBLK && vp->v_type != VCHR) 2149 return (0); 2150 2151 return (1); 2152 } 2153 2154 #ifdef DDB 2155 #include <machine/db_machdep.h> 2156 #include <ddb/db_interface.h> 2157 #include <ddb/db_output.h> 2158 2159 void 2160 vfs_buf_print(struct buf *bp, int full, int (*pr)(const char *, ...)) 2161 { 2162 2163 (*pr)(" vp %p lblkno 0x%llx blkno 0x%llx dev 0x%x\n" 2164 " proc %p error %d flags %b\n", 2165 bp->b_vp, (int64_t)bp->b_lblkno, (int64_t)bp->b_blkno, bp->b_dev, 2166 bp->b_proc, bp->b_error, bp->b_flags, B_BITS); 2167 2168 (*pr)(" bufsize 0x%lx bcount 0x%lx resid 0x%lx sync 0x%x\n" 2169 " data %p saveaddr %p dep %p iodone %p\n", 2170 bp->b_bufsize, bp->b_bcount, (long)bp->b_resid, bp->b_synctime, 2171 bp->b_data, bp->b_saveaddr, LIST_FIRST(&bp->b_dep), bp->b_iodone); 2172 2173 (*pr)(" dirty {off 0x%x end 0x%x} valid {off 0x%x end 0x%x}\n", 2174 bp->b_dirtyoff, bp->b_dirtyend, bp->b_validoff, bp->b_validend); 2175 2176 #ifdef FFS_SOFTUPDATES 2177 if (full) 2178 softdep_print(bp, full, pr); 2179 #endif 2180 } 2181 2182 const char *vtypes[] = { VTYPE_NAMES }; 2183 const char *vtags[] = { VTAG_NAMES }; 2184 2185 void 2186 vfs_vnode_print(struct vnode *vp, int full, int (*pr)(const char *, ...)) 2187 { 2188 2189 #define NENTS(n) (sizeof n / sizeof(n[0])) 2190 (*pr)("tag %s(%d) type %s(%d) mount %p typedata %p\n", 2191 vp->v_tag > NENTS(vtags)? "<unk>":vtags[vp->v_tag], vp->v_tag, 2192 vp->v_type > NENTS(vtypes)? "<unk>":vtypes[vp->v_type], 2193 vp->v_type, vp->v_mount, vp->v_mountedhere); 2194 2195 (*pr)("data %p usecount %d writecount %ld holdcnt %ld numoutput %d\n", 2196 vp->v_data, vp->v_usecount, vp->v_writecount, 2197 vp->v_holdcnt, vp->v_numoutput); 2198 2199 /* uvm_object_printit(&vp->v_uobj, full, pr); */ 2200 2201 if (full) { 2202 struct buf *bp; 2203 2204 (*pr)("clean bufs:\n"); 2205 LIST_FOREACH(bp, &vp->v_cleanblkhd, b_vnbufs) { 2206 (*pr)(" bp %p\n", bp); 2207 vfs_buf_print(bp, full, pr); 2208 } 2209 2210 (*pr)("dirty bufs:\n"); 2211 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) { 2212 (*pr)(" bp %p\n", bp); 2213 vfs_buf_print(bp, full, pr); 2214 } 2215 } 2216 } 2217 2218 void 2219 vfs_mount_print(struct mount *mp, int full, int (*pr)(const char *, ...)) 2220 { 2221 struct vfsconf *vfc = mp->mnt_vfc; 2222 struct vnode *vp; 2223 int cnt = 0; 2224 2225 (*pr)("flags %b\nvnodecovered %p syncer %p data %p\n", 2226 mp->mnt_flag, MNT_BITS, 2227 mp->mnt_vnodecovered, mp->mnt_syncer, mp->mnt_data); 2228 2229 (*pr)("vfsconf: ops %p name \"%s\" num %d ref %d flags 0x%x\n", 2230 vfc->vfc_vfsops, vfc->vfc_name, vfc->vfc_typenum, 2231 vfc->vfc_refcount, vfc->vfc_flags); 2232 2233 (*pr)("statvfs cache: bsize %x iosize %x\nblocks %llu free %llu avail %lld\n", 2234 mp->mnt_stat.f_bsize, mp->mnt_stat.f_iosize, mp->mnt_stat.f_blocks, 2235 mp->mnt_stat.f_bfree, mp->mnt_stat.f_bavail); 2236 2237 (*pr)(" files %llu ffiles %llu favail $lld\n", mp->mnt_stat.f_files, 2238 mp->mnt_stat.f_ffree, mp->mnt_stat.f_favail); 2239 2240 (*pr)(" f_fsidx {0x%x, 0x%x} owner %u ctime 0x%x\n", 2241 mp->mnt_stat.f_fsid.val[0], mp->mnt_stat.f_fsid.val[1], 2242 mp->mnt_stat.f_owner, mp->mnt_stat.f_ctime); 2243 2244 (*pr)(" syncwrites %llu asyncwrites = %llu\n", 2245 mp->mnt_stat.f_syncwrites, mp->mnt_stat.f_asyncwrites); 2246 2247 (*pr)(" syncreads %llu asyncreads = %llu\n", 2248 mp->mnt_stat.f_syncreads, mp->mnt_stat.f_asyncreads); 2249 2250 (*pr)(" fstype \"%s\" mnton \"%s\" mntfrom \"%s\"\n", 2251 mp->mnt_stat.f_fstypename, mp->mnt_stat.f_mntonname, 2252 mp->mnt_stat.f_mntfromname); 2253 2254 (*pr)("locked vnodes:"); 2255 /* XXX would take mountlist lock, except ddb has no context */ 2256 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2257 if (VOP_ISLOCKED(vp)) { 2258 if (!LIST_NEXT(vp, v_mntvnodes)) 2259 (*pr)(" %p", vp); 2260 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2261 (*pr)("\n\t%p", vp); 2262 else 2263 (*pr)(", %p", vp); 2264 } 2265 (*pr)("\n"); 2266 2267 if (full) { 2268 (*pr)("all vnodes:\n\t"); 2269 /* XXX would take mountlist lock, except ddb has no context */ 2270 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2271 if (!LIST_NEXT(vp, v_mntvnodes)) 2272 (*pr)(" %p", vp); 2273 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2274 (*pr)(" %p,\n\t", vp); 2275 else 2276 (*pr)(" %p,", vp); 2277 (*pr)("\n"); 2278 } 2279 } 2280 #endif /* DDB */ 2281 2282 void 2283 copy_statfs_info(struct statfs *sbp, const struct mount *mp) 2284 { 2285 const struct statfs *mbp; 2286 2287 strncpy(sbp->f_fstypename, mp->mnt_vfc->vfc_name, MFSNAMELEN); 2288 2289 if (sbp == (mbp = &mp->mnt_stat)) 2290 return; 2291 2292 sbp->f_fsid = mbp->f_fsid; 2293 sbp->f_owner = mbp->f_owner; 2294 sbp->f_flags = mbp->f_flags; 2295 sbp->f_syncwrites = mbp->f_syncwrites; 2296 sbp->f_asyncwrites = mbp->f_asyncwrites; 2297 sbp->f_syncreads = mbp->f_syncreads; 2298 sbp->f_asyncreads = mbp->f_asyncreads; 2299 sbp->f_namemax = mbp->f_namemax; 2300 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN); 2301 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); 2302 bcopy(&mp->mnt_stat.mount_info.ufs_args, &sbp->mount_info.ufs_args, 2303 sizeof(struct ufs_args)); 2304 } 2305 2306