1 /* $OpenBSD: vfs_subr.c,v 1.265 2017/12/14 20:23:15 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/sysctl.h> 48 #include <sys/mount.h> 49 #include <sys/time.h> 50 #include <sys/fcntl.h> 51 #include <sys/kernel.h> 52 #include <sys/conf.h> 53 #include <sys/vnode.h> 54 #include <sys/lock.h> 55 #include <sys/stat.h> 56 #include <sys/acct.h> 57 #include <sys/namei.h> 58 #include <sys/ucred.h> 59 #include <sys/buf.h> 60 #include <sys/errno.h> 61 #include <sys/malloc.h> 62 #include <sys/mbuf.h> 63 #include <sys/syscallargs.h> 64 #include <sys/pool.h> 65 #include <sys/tree.h> 66 #include <sys/specdev.h> 67 68 #include <netinet/in.h> 69 70 #include <uvm/uvm_extern.h> 71 #include <uvm/uvm_vnode.h> 72 73 #include "softraid.h" 74 75 void sr_shutdown(int); 76 77 enum vtype iftovt_tab[16] = { 78 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, 79 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, 80 }; 81 82 int vttoif_tab[9] = { 83 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, 84 S_IFSOCK, S_IFIFO, S_IFMT, 85 }; 86 87 int prtactive = 0; /* 1 => print out reclaim of active vnodes */ 88 int suid_clear = 1; /* 1 => clear SUID / SGID on owner change */ 89 90 /* 91 * Insq/Remq for the vnode usage lists. 92 */ 93 #define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) 94 #define bufremvn(bp) { \ 95 LIST_REMOVE(bp, b_vnbufs); \ 96 LIST_NEXT(bp, b_vnbufs) = NOLIST; \ 97 } 98 99 struct freelst vnode_hold_list; /* list of vnodes referencing buffers */ 100 struct freelst vnode_free_list; /* vnode free list */ 101 102 struct mntlist mountlist; /* mounted filesystem list */ 103 104 void vclean(struct vnode *, int, struct proc *); 105 106 void insmntque(struct vnode *, struct mount *); 107 int getdevvp(dev_t, struct vnode **, enum vtype); 108 109 int vfs_hang_addrlist(struct mount *, struct netexport *, 110 struct export_args *); 111 int vfs_free_netcred(struct radix_node *, void *, u_int); 112 void vfs_free_addrlist(struct netexport *); 113 void vputonfreelist(struct vnode *); 114 115 int vflush_vnode(struct vnode *, void *); 116 int maxvnodes; 117 118 #ifdef DEBUG 119 void printlockedvnodes(void); 120 #endif 121 122 struct pool vnode_pool; 123 struct pool uvm_vnode_pool; 124 125 static inline int rb_buf_compare(const struct buf *b1, const struct buf *b2); 126 RBT_GENERATE(buf_rb_bufs, buf, b_rbbufs, rb_buf_compare); 127 128 static inline int 129 rb_buf_compare(const struct buf *b1, const struct buf *b2) 130 { 131 if (b1->b_lblkno < b2->b_lblkno) 132 return(-1); 133 if (b1->b_lblkno > b2->b_lblkno) 134 return(1); 135 return(0); 136 } 137 138 /* 139 * Initialize the vnode management data structures. 140 */ 141 void 142 vntblinit(void) 143 { 144 /* buffer cache may need a vnode for each buffer */ 145 maxvnodes = 2 * initialvnodes; 146 pool_init(&vnode_pool, sizeof(struct vnode), 0, IPL_NONE, 147 PR_WAITOK, "vnodes", NULL); 148 pool_init(&uvm_vnode_pool, sizeof(struct uvm_vnode), 0, IPL_NONE, 149 PR_WAITOK, "uvmvnodes", NULL); 150 TAILQ_INIT(&vnode_hold_list); 151 TAILQ_INIT(&vnode_free_list); 152 TAILQ_INIT(&mountlist); 153 /* 154 * Initialize the filesystem syncer. 155 */ 156 vn_initialize_syncerd(); 157 158 #ifdef NFSSERVER 159 rn_init(sizeof(struct sockaddr_in)); 160 #endif /* NFSSERVER */ 161 } 162 163 /* 164 * Mark a mount point as busy. Used to synchronize access and to delay 165 * unmounting. 166 * 167 * Default behaviour is to attempt getting a READ lock and in case of an 168 * ongoing unmount, to wait for it to finish and then return failure. 169 */ 170 int 171 vfs_busy(struct mount *mp, int flags) 172 { 173 int rwflags = 0; 174 175 /* new mountpoints need their lock initialised */ 176 if (mp->mnt_lock.rwl_name == NULL) 177 rw_init_flags(&mp->mnt_lock, "vfslock", RWL_IS_VNODE); 178 179 if (flags & VB_WRITE) 180 rwflags |= RW_WRITE; 181 else 182 rwflags |= RW_READ; 183 184 if (flags & VB_WAIT) 185 rwflags |= RW_SLEEPFAIL; 186 else 187 rwflags |= RW_NOSLEEP; 188 189 if (rw_enter(&mp->mnt_lock, rwflags)) 190 return (EBUSY); 191 192 return (0); 193 } 194 195 /* 196 * Free a busy file system 197 */ 198 void 199 vfs_unbusy(struct mount *mp) 200 { 201 rw_exit(&mp->mnt_lock); 202 } 203 204 int 205 vfs_isbusy(struct mount *mp) 206 { 207 if (RWLOCK_OWNER(&mp->mnt_lock) > 0) 208 return (1); 209 else 210 return (0); 211 } 212 213 /* 214 * Lookup a filesystem type, and if found allocate and initialize 215 * a mount structure for it. 216 * 217 * Devname is usually updated by mount(8) after booting. 218 */ 219 int 220 vfs_rootmountalloc(char *fstypename, char *devname, struct mount **mpp) 221 { 222 struct vfsconf *vfsp; 223 struct mount *mp; 224 225 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 226 if (!strcmp(vfsp->vfc_name, fstypename)) 227 break; 228 if (vfsp == NULL) 229 return (ENODEV); 230 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 231 (void)vfs_busy(mp, VB_READ|VB_NOWAIT); 232 LIST_INIT(&mp->mnt_vnodelist); 233 mp->mnt_vfc = vfsp; 234 mp->mnt_op = vfsp->vfc_vfsops; 235 mp->mnt_flag = MNT_RDONLY; 236 mp->mnt_vnodecovered = NULLVP; 237 vfsp->vfc_refcount++; 238 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 239 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 240 mp->mnt_stat.f_mntonname[0] = '/'; 241 copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN, 0); 242 copystr(devname, mp->mnt_stat.f_mntfromspec, MNAMELEN, 0); 243 *mpp = mp; 244 return (0); 245 } 246 247 /* 248 * Lookup a mount point by filesystem identifier. 249 */ 250 struct mount * 251 vfs_getvfs(fsid_t *fsid) 252 { 253 struct mount *mp; 254 255 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 256 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 257 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 258 return (mp); 259 } 260 } 261 262 return (NULL); 263 } 264 265 266 /* 267 * Get a new unique fsid 268 */ 269 void 270 vfs_getnewfsid(struct mount *mp) 271 { 272 static u_short xxxfs_mntid; 273 274 fsid_t tfsid; 275 int mtype; 276 277 mtype = mp->mnt_vfc->vfc_typenum; 278 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 279 mp->mnt_stat.f_fsid.val[1] = mtype; 280 if (xxxfs_mntid == 0) 281 ++xxxfs_mntid; 282 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 283 tfsid.val[1] = mtype; 284 if (!TAILQ_EMPTY(&mountlist)) { 285 while (vfs_getvfs(&tfsid)) { 286 tfsid.val[0]++; 287 xxxfs_mntid++; 288 } 289 } 290 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 291 } 292 293 /* 294 * Set vnode attributes to VNOVAL 295 */ 296 void 297 vattr_null(struct vattr *vap) 298 { 299 300 vap->va_type = VNON; 301 /* 302 * Don't get fancy: u_quad_t = u_int = VNOVAL leaves the u_quad_t 303 * with 2^31-1 instead of 2^64-1. Just write'm out and let 304 * the compiler do its job. 305 */ 306 vap->va_mode = VNOVAL; 307 vap->va_nlink = VNOVAL; 308 vap->va_uid = VNOVAL; 309 vap->va_gid = VNOVAL; 310 vap->va_fsid = VNOVAL; 311 vap->va_fileid = VNOVAL; 312 vap->va_size = VNOVAL; 313 vap->va_blocksize = VNOVAL; 314 vap->va_atime.tv_sec = VNOVAL; 315 vap->va_atime.tv_nsec = VNOVAL; 316 vap->va_mtime.tv_sec = VNOVAL; 317 vap->va_mtime.tv_nsec = VNOVAL; 318 vap->va_ctime.tv_sec = VNOVAL; 319 vap->va_ctime.tv_nsec = VNOVAL; 320 vap->va_gen = VNOVAL; 321 vap->va_flags = VNOVAL; 322 vap->va_rdev = VNOVAL; 323 vap->va_bytes = VNOVAL; 324 vap->va_filerev = VNOVAL; 325 vap->va_vaflags = 0; 326 } 327 328 /* 329 * Routines having to do with the management of the vnode table. 330 */ 331 long numvnodes; 332 333 /* 334 * Return the next vnode from the free list. 335 */ 336 int 337 getnewvnode(enum vtagtype tag, struct mount *mp, struct vops *vops, 338 struct vnode **vpp) 339 { 340 struct proc *p = curproc; 341 struct freelst *listhd; 342 static int toggle; 343 struct vnode *vp; 344 int s; 345 346 /* 347 * allow maxvnodes to increase if the buffer cache itself 348 * is big enough to justify it. (we don't shrink it ever) 349 */ 350 maxvnodes = maxvnodes < bcstats.numbufs ? bcstats.numbufs 351 : maxvnodes; 352 353 /* 354 * We must choose whether to allocate a new vnode or recycle an 355 * existing one. The criterion for allocating a new one is that 356 * the total number of vnodes is less than the number desired or 357 * there are no vnodes on either free list. Generally we only 358 * want to recycle vnodes that have no buffers associated with 359 * them, so we look first on the vnode_free_list. If it is empty, 360 * we next consider vnodes with referencing buffers on the 361 * vnode_hold_list. The toggle ensures that half the time we 362 * will use a buffer from the vnode_hold_list, and half the time 363 * we will allocate a new one unless the list has grown to twice 364 * the desired size. We are reticent to recycle vnodes from the 365 * vnode_hold_list because we will lose the identity of all its 366 * referencing buffers. 367 */ 368 toggle ^= 1; 369 if (numvnodes / 2 > maxvnodes) 370 toggle = 0; 371 372 s = splbio(); 373 if ((numvnodes < maxvnodes) || 374 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 375 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 376 splx(s); 377 vp = pool_get(&vnode_pool, PR_WAITOK | PR_ZERO); 378 vp->v_uvm = pool_get(&uvm_vnode_pool, PR_WAITOK | PR_ZERO); 379 vp->v_uvm->u_vnode = vp; 380 RBT_INIT(buf_rb_bufs, &vp->v_bufs_tree); 381 cache_tree_init(&vp->v_nc_tree); 382 TAILQ_INIT(&vp->v_cache_dst); 383 numvnodes++; 384 } else { 385 TAILQ_FOREACH(vp, listhd, v_freelist) { 386 if (VOP_ISLOCKED(vp) == 0) 387 break; 388 } 389 /* 390 * Unless this is a bad time of the month, at most 391 * the first NCPUS items on the free list are 392 * locked, so this is close enough to being empty. 393 */ 394 if (vp == NULL) { 395 splx(s); 396 tablefull("vnode"); 397 *vpp = 0; 398 return (ENFILE); 399 } 400 401 #ifdef DIAGNOSTIC 402 if (vp->v_usecount) { 403 vprint("free vnode", vp); 404 panic("free vnode isn't"); 405 } 406 #endif 407 408 TAILQ_REMOVE(listhd, vp, v_freelist); 409 vp->v_bioflag &= ~VBIOONFREELIST; 410 splx(s); 411 412 if (vp->v_type != VBAD) 413 vgonel(vp, p); 414 #ifdef DIAGNOSTIC 415 if (vp->v_data) { 416 vprint("cleaned vnode", vp); 417 panic("cleaned vnode isn't"); 418 } 419 s = splbio(); 420 if (vp->v_numoutput) 421 panic("Clean vnode has pending I/O's"); 422 splx(s); 423 #endif 424 vp->v_flag = 0; 425 vp->v_socket = 0; 426 } 427 cache_purge(vp); 428 vp->v_type = VNON; 429 vp->v_tag = tag; 430 vp->v_op = vops; 431 insmntque(vp, mp); 432 *vpp = vp; 433 vp->v_usecount = 1; 434 vp->v_data = 0; 435 return (0); 436 } 437 438 /* 439 * Move a vnode from one mount queue to another. 440 */ 441 void 442 insmntque(struct vnode *vp, struct mount *mp) 443 { 444 /* 445 * Delete from old mount point vnode list, if on one. 446 */ 447 if (vp->v_mount != NULL) 448 LIST_REMOVE(vp, v_mntvnodes); 449 /* 450 * Insert into list of vnodes for the new mount point, if available. 451 */ 452 if ((vp->v_mount = mp) != NULL) 453 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 454 } 455 456 /* 457 * Create a vnode for a block device. 458 * Used for root filesystem, argdev, and swap areas. 459 * Also used for memory file system special devices. 460 */ 461 int 462 bdevvp(dev_t dev, struct vnode **vpp) 463 { 464 return (getdevvp(dev, vpp, VBLK)); 465 } 466 467 /* 468 * Create a vnode for a character device. 469 * Used for console handling. 470 */ 471 int 472 cdevvp(dev_t dev, struct vnode **vpp) 473 { 474 return (getdevvp(dev, vpp, VCHR)); 475 } 476 477 /* 478 * Create a vnode for a device. 479 * Used by bdevvp (block device) for root file system etc., 480 * and by cdevvp (character device) for console. 481 */ 482 int 483 getdevvp(dev_t dev, struct vnode **vpp, enum vtype type) 484 { 485 struct vnode *vp; 486 struct vnode *nvp; 487 int error; 488 489 if (dev == NODEV) { 490 *vpp = NULLVP; 491 return (0); 492 } 493 error = getnewvnode(VT_NON, NULL, &spec_vops, &nvp); 494 if (error) { 495 *vpp = NULLVP; 496 return (error); 497 } 498 vp = nvp; 499 vp->v_type = type; 500 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 501 vput(vp); 502 vp = nvp; 503 } 504 if (vp->v_type == VCHR && cdevsw[major(vp->v_rdev)].d_type == D_TTY) 505 vp->v_flag |= VISTTY; 506 *vpp = vp; 507 return (0); 508 } 509 510 /* 511 * Check to see if the new vnode represents a special device 512 * for which we already have a vnode (either because of 513 * bdevvp() or because of a different vnode representing 514 * the same block device). If such an alias exists, deallocate 515 * the existing contents and return the aliased vnode. The 516 * caller is responsible for filling it with its new contents. 517 */ 518 struct vnode * 519 checkalias(struct vnode *nvp, dev_t nvp_rdev, struct mount *mp) 520 { 521 struct proc *p = curproc; 522 struct vnode *vp; 523 struct vnode **vpp; 524 525 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 526 return (NULLVP); 527 528 vpp = &speclisth[SPECHASH(nvp_rdev)]; 529 loop: 530 for (vp = *vpp; vp; vp = vp->v_specnext) { 531 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 532 continue; 533 } 534 /* 535 * Alias, but not in use, so flush it out. 536 */ 537 if (vp->v_usecount == 0) { 538 vgonel(vp, p); 539 goto loop; 540 } 541 if (vget(vp, LK_EXCLUSIVE, p)) { 542 goto loop; 543 } 544 break; 545 } 546 547 /* 548 * Common case is actually in the if statement 549 */ 550 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 551 nvp->v_specinfo = malloc(sizeof(struct specinfo), M_VNODE, 552 M_WAITOK); 553 nvp->v_rdev = nvp_rdev; 554 nvp->v_hashchain = vpp; 555 nvp->v_specnext = *vpp; 556 nvp->v_specmountpoint = NULL; 557 nvp->v_speclockf = NULL; 558 nvp->v_specbitmap = NULL; 559 if (nvp->v_type == VCHR && 560 (cdevsw[major(nvp_rdev)].d_flags & D_CLONE) && 561 (minor(nvp_rdev) >> CLONE_SHIFT == 0)) { 562 if (vp != NULLVP) 563 nvp->v_specbitmap = vp->v_specbitmap; 564 else 565 nvp->v_specbitmap = malloc(CLONE_MAPSZ, 566 M_VNODE, M_WAITOK | M_ZERO); 567 } 568 *vpp = nvp; 569 if (vp != NULLVP) { 570 nvp->v_flag |= VALIASED; 571 vp->v_flag |= VALIASED; 572 vput(vp); 573 } 574 return (NULLVP); 575 } 576 577 /* 578 * This code is the uncommon case. It is called in case 579 * we found an alias that was VT_NON && vtype of VBLK 580 * This means we found a block device that was created 581 * using bdevvp. 582 * An example of such a vnode is the root partition device vnode 583 * created in ffs_mountroot. 584 * 585 * The vnodes created by bdevvp should not be aliased (why?). 586 */ 587 588 VOP_UNLOCK(vp, p); 589 vclean(vp, 0, p); 590 vp->v_op = nvp->v_op; 591 vp->v_tag = nvp->v_tag; 592 nvp->v_type = VNON; 593 insmntque(vp, mp); 594 return (vp); 595 } 596 597 /* 598 * Grab a particular vnode from the free list, increment its 599 * reference count and lock it. If the vnode lock bit is set, 600 * the vnode is being eliminated in vgone. In that case, we 601 * cannot grab it, so the process is awakened when the 602 * transition is completed, and an error code is returned to 603 * indicate that the vnode is no longer usable, possibly 604 * having been changed to a new file system type. 605 */ 606 int 607 vget(struct vnode *vp, int flags, struct proc *p) 608 { 609 int error, s, onfreelist; 610 611 /* 612 * If the vnode is in the process of being cleaned out for 613 * another use, we wait for the cleaning to finish and then 614 * return failure. Cleaning is determined by checking that 615 * the VXLOCK flag is set. 616 */ 617 618 if (vp->v_flag & VXLOCK) { 619 if (flags & LK_NOWAIT) { 620 return (EBUSY); 621 } 622 623 vp->v_flag |= VXWANT; 624 tsleep(vp, PINOD, "vget", 0); 625 return (ENOENT); 626 } 627 628 onfreelist = vp->v_bioflag & VBIOONFREELIST; 629 if (vp->v_usecount == 0 && onfreelist) { 630 s = splbio(); 631 if (vp->v_holdcnt > 0) 632 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 633 else 634 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 635 vp->v_bioflag &= ~VBIOONFREELIST; 636 splx(s); 637 } 638 639 vp->v_usecount++; 640 if (flags & LK_TYPE_MASK) { 641 if ((error = vn_lock(vp, flags, p)) != 0) { 642 vp->v_usecount--; 643 if (vp->v_usecount == 0 && onfreelist) 644 vputonfreelist(vp); 645 } 646 return (error); 647 } 648 649 return (0); 650 } 651 652 653 /* Vnode reference. */ 654 void 655 vref(struct vnode *vp) 656 { 657 #ifdef DIAGNOSTIC 658 if (vp->v_usecount == 0) 659 panic("vref used where vget required"); 660 if (vp->v_type == VNON) 661 panic("vref on a VNON vnode"); 662 #endif 663 vp->v_usecount++; 664 } 665 666 void 667 vputonfreelist(struct vnode *vp) 668 { 669 int s; 670 struct freelst *lst; 671 672 s = splbio(); 673 #ifdef DIAGNOSTIC 674 if (vp->v_usecount != 0) 675 panic("Use count is not zero!"); 676 677 if (vp->v_bioflag & VBIOONFREELIST) { 678 vprint("vnode already on free list: ", vp); 679 panic("vnode already on free list"); 680 } 681 #endif 682 683 vp->v_bioflag |= VBIOONFREELIST; 684 685 if (vp->v_holdcnt > 0) 686 lst = &vnode_hold_list; 687 else 688 lst = &vnode_free_list; 689 690 if (vp->v_type == VBAD) 691 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 692 else 693 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 694 695 splx(s); 696 } 697 698 /* 699 * vput(), just unlock and vrele() 700 */ 701 void 702 vput(struct vnode *vp) 703 { 704 struct proc *p = curproc; 705 706 #ifdef DIAGNOSTIC 707 if (vp == NULL) 708 panic("vput: null vp"); 709 #endif 710 711 #ifdef DIAGNOSTIC 712 if (vp->v_usecount == 0) { 713 vprint("vput: bad ref count", vp); 714 panic("vput: ref cnt"); 715 } 716 #endif 717 vp->v_usecount--; 718 if (vp->v_usecount > 0) { 719 VOP_UNLOCK(vp, p); 720 return; 721 } 722 723 #ifdef DIAGNOSTIC 724 if (vp->v_writecount != 0) { 725 vprint("vput: bad writecount", vp); 726 panic("vput: v_writecount != 0"); 727 } 728 #endif 729 730 VOP_INACTIVE(vp, p); 731 732 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 733 vputonfreelist(vp); 734 } 735 736 /* 737 * Vnode release - use for active VNODES. 738 * If count drops to zero, call inactive routine and return to freelist. 739 * Returns 0 if it did not sleep. 740 */ 741 int 742 vrele(struct vnode *vp) 743 { 744 struct proc *p = curproc; 745 746 #ifdef DIAGNOSTIC 747 if (vp == NULL) 748 panic("vrele: null vp"); 749 #endif 750 #ifdef DIAGNOSTIC 751 if (vp->v_usecount == 0) { 752 vprint("vrele: bad ref count", vp); 753 panic("vrele: ref cnt"); 754 } 755 #endif 756 vp->v_usecount--; 757 if (vp->v_usecount > 0) { 758 return (0); 759 } 760 761 #ifdef DIAGNOSTIC 762 if (vp->v_writecount != 0) { 763 vprint("vrele: bad writecount", vp); 764 panic("vrele: v_writecount != 0"); 765 } 766 #endif 767 768 if (vn_lock(vp, LK_EXCLUSIVE, p)) { 769 #ifdef DIAGNOSTIC 770 vprint("vrele: cannot lock", vp); 771 #endif 772 return (1); 773 } 774 775 VOP_INACTIVE(vp, p); 776 777 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 778 vputonfreelist(vp); 779 return (1); 780 } 781 782 /* Page or buffer structure gets a reference. */ 783 void 784 vhold(struct vnode *vp) 785 { 786 /* 787 * If it is on the freelist and the hold count is currently 788 * zero, move it to the hold list. 789 */ 790 if ((vp->v_bioflag & VBIOONFREELIST) && 791 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 792 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 793 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 794 } 795 vp->v_holdcnt++; 796 } 797 798 /* Lose interest in a vnode. */ 799 void 800 vdrop(struct vnode *vp) 801 { 802 #ifdef DIAGNOSTIC 803 if (vp->v_holdcnt == 0) 804 panic("vdrop: zero holdcnt"); 805 #endif 806 807 vp->v_holdcnt--; 808 809 /* 810 * If it is on the holdlist and the hold count drops to 811 * zero, move it to the free list. 812 */ 813 if ((vp->v_bioflag & VBIOONFREELIST) && 814 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 815 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 816 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 817 } 818 } 819 820 /* 821 * Remove any vnodes in the vnode table belonging to mount point mp. 822 * 823 * If MNT_NOFORCE is specified, there should not be any active ones, 824 * return error if any are found (nb: this is a user error, not a 825 * system error). If MNT_FORCE is specified, detach any active vnodes 826 * that are found. 827 */ 828 #ifdef DEBUG 829 int busyprt = 0; /* print out busy vnodes */ 830 struct ctldebug debug1 = { "busyprt", &busyprt }; 831 #endif 832 833 int 834 vfs_mount_foreach_vnode(struct mount *mp, 835 int (*func)(struct vnode *, void *), void *arg) { 836 struct vnode *vp, *nvp; 837 int error = 0; 838 839 loop: 840 LIST_FOREACH_SAFE(vp , &mp->mnt_vnodelist, v_mntvnodes, nvp) { 841 if (vp->v_mount != mp) 842 goto loop; 843 844 error = func(vp, arg); 845 846 if (error != 0) 847 break; 848 } 849 850 return (error); 851 } 852 853 struct vflush_args { 854 struct vnode *skipvp; 855 int busy; 856 int flags; 857 }; 858 859 int 860 vflush_vnode(struct vnode *vp, void *arg) 861 { 862 struct vflush_args *va = arg; 863 struct proc *p = curproc; 864 865 if (vp == va->skipvp) { 866 return (0); 867 } 868 869 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 870 return (0); 871 } 872 873 /* 874 * If WRITECLOSE is set, only flush out regular file 875 * vnodes open for writing. 876 */ 877 if ((va->flags & WRITECLOSE) && 878 (vp->v_writecount == 0 || vp->v_type != VREG)) { 879 return (0); 880 } 881 882 /* 883 * With v_usecount == 0, all we need to do is clear 884 * out the vnode data structures and we are done. 885 */ 886 if (vp->v_usecount == 0) { 887 vgonel(vp, p); 888 return (0); 889 } 890 891 /* 892 * If FORCECLOSE is set, forcibly close the vnode. 893 * For block or character devices, revert to an 894 * anonymous device. For all other files, just kill them. 895 */ 896 if (va->flags & FORCECLOSE) { 897 if (vp->v_type != VBLK && vp->v_type != VCHR) { 898 vgonel(vp, p); 899 } else { 900 vclean(vp, 0, p); 901 vp->v_op = &spec_vops; 902 insmntque(vp, NULL); 903 } 904 return (0); 905 } 906 907 /* 908 * If set, this is allowed to ignore vnodes which don't 909 * have changes pending to disk. 910 * XXX Might be nice to check per-fs "inode" flags, but 911 * generally the filesystem is sync'd already, right? 912 */ 913 if ((va->flags & IGNORECLEAN) && 914 LIST_EMPTY(&vp->v_dirtyblkhd)) 915 return (0); 916 917 #ifdef DEBUG 918 if (busyprt) 919 vprint("vflush: busy vnode", vp); 920 #endif 921 va->busy++; 922 return (0); 923 } 924 925 int 926 vflush(struct mount *mp, struct vnode *skipvp, int flags) 927 { 928 struct vflush_args va; 929 va.skipvp = skipvp; 930 va.busy = 0; 931 va.flags = flags; 932 933 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 934 935 if (va.busy) 936 return (EBUSY); 937 return (0); 938 } 939 940 /* 941 * Disassociate the underlying file system from a vnode. 942 */ 943 void 944 vclean(struct vnode *vp, int flags, struct proc *p) 945 { 946 int active; 947 948 /* 949 * Check to see if the vnode is in use. 950 * If so we have to reference it before we clean it out 951 * so that its count cannot fall to zero and generate a 952 * race against ourselves to recycle it. 953 */ 954 if ((active = vp->v_usecount) != 0) 955 vp->v_usecount++; 956 957 /* 958 * Prevent the vnode from being recycled or 959 * brought into use while we clean it out. 960 */ 961 if (vp->v_flag & VXLOCK) 962 panic("vclean: deadlock"); 963 vp->v_flag |= VXLOCK; 964 /* 965 * Even if the count is zero, the VOP_INACTIVE routine may still 966 * have the object locked while it cleans it out. The VOP_LOCK 967 * ensures that the VOP_INACTIVE routine is done with its work. 968 * For active vnodes, it ensures that no other activity can 969 * occur while the underlying object is being cleaned out. 970 */ 971 VOP_LOCK(vp, LK_DRAIN | LK_EXCLUSIVE, p); 972 973 /* 974 * Clean out any VM data associated with the vnode. 975 */ 976 uvm_vnp_terminate(vp); 977 /* 978 * Clean out any buffers associated with the vnode. 979 */ 980 if (flags & DOCLOSE) 981 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 982 /* 983 * If purging an active vnode, it must be closed and 984 * deactivated before being reclaimed. Note that the 985 * VOP_INACTIVE will unlock the vnode 986 */ 987 if (active) { 988 if (flags & DOCLOSE) 989 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 990 VOP_INACTIVE(vp, p); 991 } else { 992 /* 993 * Any other processes trying to obtain this lock must first 994 * wait for VXLOCK to clear, then call the new lock operation. 995 */ 996 VOP_UNLOCK(vp, p); 997 } 998 999 /* 1000 * Reclaim the vnode. 1001 */ 1002 if (VOP_RECLAIM(vp, p)) 1003 panic("vclean: cannot reclaim"); 1004 if (active) { 1005 vp->v_usecount--; 1006 if (vp->v_usecount == 0) { 1007 if (vp->v_holdcnt > 0) 1008 panic("vclean: not clean"); 1009 vputonfreelist(vp); 1010 } 1011 } 1012 cache_purge(vp); 1013 1014 /* 1015 * Done with purge, notify sleepers of the grim news. 1016 */ 1017 vp->v_op = &dead_vops; 1018 VN_KNOTE(vp, NOTE_REVOKE); 1019 vp->v_tag = VT_NON; 1020 vp->v_flag &= ~VXLOCK; 1021 #ifdef VFSLCKDEBUG 1022 vp->v_flag &= ~VLOCKSWORK; 1023 #endif 1024 if (vp->v_flag & VXWANT) { 1025 vp->v_flag &= ~VXWANT; 1026 wakeup(vp); 1027 } 1028 } 1029 1030 /* 1031 * Recycle an unused vnode to the front of the free list. 1032 */ 1033 int 1034 vrecycle(struct vnode *vp, struct proc *p) 1035 { 1036 if (vp->v_usecount == 0) { 1037 vgonel(vp, p); 1038 return (1); 1039 } 1040 return (0); 1041 } 1042 1043 /* 1044 * Eliminate all activity associated with a vnode 1045 * in preparation for reuse. 1046 */ 1047 void 1048 vgone(struct vnode *vp) 1049 { 1050 struct proc *p = curproc; 1051 vgonel(vp, p); 1052 } 1053 1054 /* 1055 * vgone, with struct proc. 1056 */ 1057 void 1058 vgonel(struct vnode *vp, struct proc *p) 1059 { 1060 struct vnode *vq; 1061 struct vnode *vx; 1062 1063 /* 1064 * If a vgone (or vclean) is already in progress, 1065 * wait until it is done and return. 1066 */ 1067 if (vp->v_flag & VXLOCK) { 1068 vp->v_flag |= VXWANT; 1069 tsleep(vp, PINOD, "vgone", 0); 1070 return; 1071 } 1072 1073 /* 1074 * Clean out the filesystem specific data. 1075 */ 1076 vclean(vp, DOCLOSE, p); 1077 /* 1078 * Delete from old mount point vnode list, if on one. 1079 */ 1080 if (vp->v_mount != NULL) 1081 insmntque(vp, NULL); 1082 /* 1083 * If special device, remove it from special device alias list 1084 * if it is on one. 1085 */ 1086 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1087 if ((vp->v_flag & VALIASED) == 0 && vp->v_type == VCHR && 1088 (cdevsw[major(vp->v_rdev)].d_flags & D_CLONE) && 1089 (minor(vp->v_rdev) >> CLONE_SHIFT == 0)) { 1090 free(vp->v_specbitmap, M_VNODE, CLONE_MAPSZ); 1091 } 1092 if (*vp->v_hashchain == vp) { 1093 *vp->v_hashchain = vp->v_specnext; 1094 } else { 1095 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1096 if (vq->v_specnext != vp) 1097 continue; 1098 vq->v_specnext = vp->v_specnext; 1099 break; 1100 } 1101 if (vq == NULL) 1102 panic("missing bdev"); 1103 } 1104 if (vp->v_flag & VALIASED) { 1105 vx = NULL; 1106 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1107 if (vq->v_rdev != vp->v_rdev || 1108 vq->v_type != vp->v_type) 1109 continue; 1110 if (vx) 1111 break; 1112 vx = vq; 1113 } 1114 if (vx == NULL) 1115 panic("missing alias"); 1116 if (vq == NULL) 1117 vx->v_flag &= ~VALIASED; 1118 vp->v_flag &= ~VALIASED; 1119 } 1120 free(vp->v_specinfo, M_VNODE, sizeof(struct specinfo)); 1121 vp->v_specinfo = NULL; 1122 } 1123 /* 1124 * If it is on the freelist and not already at the head, 1125 * move it to the head of the list. 1126 */ 1127 vp->v_type = VBAD; 1128 1129 /* 1130 * Move onto the free list, unless we were called from 1131 * getnewvnode and we're not on any free list 1132 */ 1133 if (vp->v_usecount == 0 && 1134 (vp->v_bioflag & VBIOONFREELIST)) { 1135 int s; 1136 1137 s = splbio(); 1138 1139 if (vp->v_holdcnt > 0) 1140 panic("vgonel: not clean"); 1141 1142 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1143 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1144 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1145 } 1146 splx(s); 1147 } 1148 } 1149 1150 /* 1151 * Lookup a vnode by device number. 1152 */ 1153 int 1154 vfinddev(dev_t dev, enum vtype type, struct vnode **vpp) 1155 { 1156 struct vnode *vp; 1157 int rc =0; 1158 1159 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1160 if (dev != vp->v_rdev || type != vp->v_type) 1161 continue; 1162 *vpp = vp; 1163 rc = 1; 1164 break; 1165 } 1166 return (rc); 1167 } 1168 1169 /* 1170 * Revoke all the vnodes corresponding to the specified minor number 1171 * range (endpoints inclusive) of the specified major. 1172 */ 1173 void 1174 vdevgone(int maj, int minl, int minh, enum vtype type) 1175 { 1176 struct vnode *vp; 1177 int mn; 1178 1179 for (mn = minl; mn <= minh; mn++) 1180 if (vfinddev(makedev(maj, mn), type, &vp)) 1181 VOP_REVOKE(vp, REVOKEALL); 1182 } 1183 1184 /* 1185 * Calculate the total number of references to a special device. 1186 */ 1187 int 1188 vcount(struct vnode *vp) 1189 { 1190 struct vnode *vq, *vnext; 1191 int count; 1192 1193 loop: 1194 if ((vp->v_flag & VALIASED) == 0) 1195 return (vp->v_usecount); 1196 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1197 vnext = vq->v_specnext; 1198 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1199 continue; 1200 /* 1201 * Alias, but not in use, so flush it out. 1202 */ 1203 if (vq->v_usecount == 0 && vq != vp) { 1204 vgone(vq); 1205 goto loop; 1206 } 1207 count += vq->v_usecount; 1208 } 1209 return (count); 1210 } 1211 1212 #if defined(DEBUG) || defined(DIAGNOSTIC) 1213 /* 1214 * Print out a description of a vnode. 1215 */ 1216 static char *typename[] = 1217 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1218 1219 void 1220 vprint(char *label, struct vnode *vp) 1221 { 1222 char buf[64]; 1223 1224 if (label != NULL) 1225 printf("%s: ", label); 1226 printf("%p, type %s, use %u, write %u, hold %u,", 1227 vp, typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1228 vp->v_holdcnt); 1229 buf[0] = '\0'; 1230 if (vp->v_flag & VROOT) 1231 strlcat(buf, "|VROOT", sizeof buf); 1232 if (vp->v_flag & VTEXT) 1233 strlcat(buf, "|VTEXT", sizeof buf); 1234 if (vp->v_flag & VSYSTEM) 1235 strlcat(buf, "|VSYSTEM", sizeof buf); 1236 if (vp->v_flag & VXLOCK) 1237 strlcat(buf, "|VXLOCK", sizeof buf); 1238 if (vp->v_flag & VXWANT) 1239 strlcat(buf, "|VXWANT", sizeof buf); 1240 if (vp->v_bioflag & VBIOWAIT) 1241 strlcat(buf, "|VBIOWAIT", sizeof buf); 1242 if (vp->v_bioflag & VBIOONFREELIST) 1243 strlcat(buf, "|VBIOONFREELIST", sizeof buf); 1244 if (vp->v_bioflag & VBIOONSYNCLIST) 1245 strlcat(buf, "|VBIOONSYNCLIST", sizeof buf); 1246 if (vp->v_flag & VALIASED) 1247 strlcat(buf, "|VALIASED", sizeof buf); 1248 if (buf[0] != '\0') 1249 printf(" flags (%s)", &buf[1]); 1250 if (vp->v_data == NULL) { 1251 printf("\n"); 1252 } else { 1253 printf("\n\t"); 1254 VOP_PRINT(vp); 1255 } 1256 } 1257 #endif /* DEBUG || DIAGNOSTIC */ 1258 1259 #ifdef DEBUG 1260 /* 1261 * List all of the locked vnodes in the system. 1262 * Called when debugging the kernel. 1263 */ 1264 void 1265 printlockedvnodes(void) 1266 { 1267 struct mount *mp; 1268 struct vnode *vp; 1269 1270 printf("Locked vnodes\n"); 1271 1272 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 1273 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) 1274 continue; 1275 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 1276 if (VOP_ISLOCKED(vp)) 1277 vprint(NULL, vp); 1278 } 1279 vfs_unbusy(mp); 1280 } 1281 1282 } 1283 #endif 1284 1285 /* 1286 * Top level filesystem related information gathering. 1287 */ 1288 int 1289 vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1290 size_t newlen, struct proc *p) 1291 { 1292 struct vfsconf *vfsp, *tmpvfsp; 1293 int ret; 1294 1295 /* all sysctl names at this level are at least name and field */ 1296 if (namelen < 2) 1297 return (ENOTDIR); /* overloaded */ 1298 1299 if (name[0] != VFS_GENERIC) { 1300 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1301 if (vfsp->vfc_typenum == name[0]) 1302 break; 1303 1304 if (vfsp == NULL || vfsp->vfc_vfsops->vfs_sysctl == NULL) 1305 return (EOPNOTSUPP); 1306 1307 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1308 oldp, oldlenp, newp, newlen, p)); 1309 } 1310 1311 switch (name[1]) { 1312 case VFS_MAXTYPENUM: 1313 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1314 1315 case VFS_CONF: 1316 if (namelen < 3) 1317 return (ENOTDIR); /* overloaded */ 1318 1319 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1320 if (vfsp->vfc_typenum == name[2]) 1321 break; 1322 1323 if (vfsp == NULL) 1324 return (EOPNOTSUPP); 1325 1326 /* Make a copy, clear out kernel pointers */ 1327 tmpvfsp = malloc(sizeof(*tmpvfsp), M_TEMP, M_WAITOK|M_ZERO); 1328 memcpy(tmpvfsp, vfsp, sizeof(*tmpvfsp)); 1329 tmpvfsp->vfc_vfsops = NULL; 1330 tmpvfsp->vfc_next = NULL; 1331 1332 ret = sysctl_rdstruct(oldp, oldlenp, newp, tmpvfsp, 1333 sizeof(struct vfsconf)); 1334 1335 free(tmpvfsp, M_TEMP, sizeof(*tmpvfsp)); 1336 return (ret); 1337 case VFS_BCACHESTAT: /* buffer cache statistics */ 1338 ret = sysctl_rdstruct(oldp, oldlenp, newp, &bcstats, 1339 sizeof(struct bcachestats)); 1340 return(ret); 1341 } 1342 return (EOPNOTSUPP); 1343 } 1344 1345 /* 1346 * Check to see if a filesystem is mounted on a block device. 1347 */ 1348 int 1349 vfs_mountedon(struct vnode *vp) 1350 { 1351 struct vnode *vq; 1352 int error = 0; 1353 1354 if (vp->v_specmountpoint != NULL) 1355 return (EBUSY); 1356 if (vp->v_flag & VALIASED) { 1357 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1358 if (vq->v_rdev != vp->v_rdev || 1359 vq->v_type != vp->v_type) 1360 continue; 1361 if (vq->v_specmountpoint != NULL) { 1362 error = EBUSY; 1363 break; 1364 } 1365 } 1366 } 1367 return (error); 1368 } 1369 1370 #ifdef NFSSERVER 1371 /* 1372 * Build hash lists of net addresses and hang them off the mount point. 1373 * Called by vfs_export() to set up the lists of export addresses. 1374 */ 1375 int 1376 vfs_hang_addrlist(struct mount *mp, struct netexport *nep, 1377 struct export_args *argp) 1378 { 1379 struct netcred *np; 1380 struct radix_node_head *rnh; 1381 int nplen, i; 1382 struct radix_node *rn; 1383 struct sockaddr *saddr, *smask = 0; 1384 int error; 1385 1386 if (argp->ex_addrlen == 0) { 1387 if (mp->mnt_flag & MNT_DEFEXPORTED) 1388 return (EPERM); 1389 np = &nep->ne_defexported; 1390 /* fill in the kernel's ucred from userspace's xucred */ 1391 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1392 return (error); 1393 mp->mnt_flag |= MNT_DEFEXPORTED; 1394 goto finish; 1395 } 1396 if (argp->ex_addrlen > MLEN || argp->ex_masklen > MLEN || 1397 argp->ex_addrlen < 0 || argp->ex_masklen < 0) 1398 return (EINVAL); 1399 nplen = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1400 np = (struct netcred *)malloc(nplen, M_NETADDR, M_WAITOK|M_ZERO); 1401 saddr = (struct sockaddr *)(np + 1); 1402 error = copyin(argp->ex_addr, saddr, argp->ex_addrlen); 1403 if (error) 1404 goto out; 1405 if (saddr->sa_len > argp->ex_addrlen) 1406 saddr->sa_len = argp->ex_addrlen; 1407 if (argp->ex_masklen) { 1408 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1409 error = copyin(argp->ex_mask, smask, argp->ex_masklen); 1410 if (error) 1411 goto out; 1412 if (smask->sa_len > argp->ex_masklen) 1413 smask->sa_len = argp->ex_masklen; 1414 } 1415 /* fill in the kernel's ucred from userspace's xucred */ 1416 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1417 goto out; 1418 i = saddr->sa_family; 1419 switch (i) { 1420 case AF_INET: 1421 if ((rnh = nep->ne_rtable_inet) == NULL) { 1422 if (!rn_inithead((void **)&nep->ne_rtable_inet, 1423 offsetof(struct sockaddr_in, sin_addr))) { 1424 error = ENOBUFS; 1425 goto out; 1426 } 1427 rnh = nep->ne_rtable_inet; 1428 } 1429 break; 1430 default: 1431 error = EINVAL; 1432 goto out; 1433 } 1434 rn = rn_addroute(saddr, smask, rnh, np->netc_rnodes, 0); 1435 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1436 error = EPERM; 1437 goto out; 1438 } 1439 finish: 1440 np->netc_exflags = argp->ex_flags; 1441 return (0); 1442 out: 1443 free(np, M_NETADDR, nplen); 1444 return (error); 1445 } 1446 1447 int 1448 vfs_free_netcred(struct radix_node *rn, void *w, u_int id) 1449 { 1450 struct radix_node_head *rnh = (struct radix_node_head *)w; 1451 1452 rn_delete(rn->rn_key, rn->rn_mask, rnh, NULL); 1453 free(rn, M_NETADDR, 0); 1454 return (0); 1455 } 1456 1457 /* 1458 * Free the net address hash lists that are hanging off the mount points. 1459 */ 1460 void 1461 vfs_free_addrlist(struct netexport *nep) 1462 { 1463 struct radix_node_head *rnh; 1464 1465 if ((rnh = nep->ne_rtable_inet) != NULL) { 1466 rn_walktree(rnh, vfs_free_netcred, rnh); 1467 free(rnh, M_RTABLE, 0); 1468 nep->ne_rtable_inet = NULL; 1469 } 1470 } 1471 #endif /* NFSSERVER */ 1472 1473 int 1474 vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp) 1475 { 1476 #ifdef NFSSERVER 1477 int error; 1478 1479 if (argp->ex_flags & MNT_DELEXPORT) { 1480 vfs_free_addrlist(nep); 1481 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1482 } 1483 if (argp->ex_flags & MNT_EXPORTED) { 1484 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1485 return (error); 1486 mp->mnt_flag |= MNT_EXPORTED; 1487 } 1488 return (0); 1489 #else 1490 return (ENOTSUP); 1491 #endif /* NFSSERVER */ 1492 } 1493 1494 struct netcred * 1495 vfs_export_lookup(struct mount *mp, struct netexport *nep, struct mbuf *nam) 1496 { 1497 #ifdef NFSSERVER 1498 struct netcred *np; 1499 struct radix_node_head *rnh; 1500 struct sockaddr *saddr; 1501 1502 np = NULL; 1503 if (mp->mnt_flag & MNT_EXPORTED) { 1504 /* 1505 * Lookup in the export list first. 1506 */ 1507 if (nam != NULL) { 1508 saddr = mtod(nam, struct sockaddr *); 1509 switch(saddr->sa_family) { 1510 case AF_INET: 1511 rnh = nep->ne_rtable_inet; 1512 break; 1513 default: 1514 rnh = NULL; 1515 break; 1516 } 1517 if (rnh != NULL) 1518 np = (struct netcred *)rn_match(saddr, rnh); 1519 } 1520 /* 1521 * If no address match, use the default if it exists. 1522 */ 1523 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1524 np = &nep->ne_defexported; 1525 } 1526 return (np); 1527 #else 1528 return (NULL); 1529 #endif /* NFSSERVER */ 1530 } 1531 1532 /* 1533 * Do the usual access checking. 1534 * file_mode, uid and gid are from the vnode in question, 1535 * while acc_mode and cred are from the VOP_ACCESS parameter list 1536 */ 1537 int 1538 vaccess(enum vtype type, mode_t file_mode, uid_t uid, gid_t gid, 1539 mode_t acc_mode, struct ucred *cred) 1540 { 1541 mode_t mask; 1542 1543 /* User id 0 always gets read/write access. */ 1544 if (cred->cr_uid == 0) { 1545 /* For VEXEC, at least one of the execute bits must be set. */ 1546 if ((acc_mode & VEXEC) && type != VDIR && 1547 (file_mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0) 1548 return EACCES; 1549 return 0; 1550 } 1551 1552 mask = 0; 1553 1554 /* Otherwise, check the owner. */ 1555 if (cred->cr_uid == uid) { 1556 if (acc_mode & VEXEC) 1557 mask |= S_IXUSR; 1558 if (acc_mode & VREAD) 1559 mask |= S_IRUSR; 1560 if (acc_mode & VWRITE) 1561 mask |= S_IWUSR; 1562 return (file_mode & mask) == mask ? 0 : EACCES; 1563 } 1564 1565 /* Otherwise, check the groups. */ 1566 if (groupmember(gid, cred)) { 1567 if (acc_mode & VEXEC) 1568 mask |= S_IXGRP; 1569 if (acc_mode & VREAD) 1570 mask |= S_IRGRP; 1571 if (acc_mode & VWRITE) 1572 mask |= S_IWGRP; 1573 return (file_mode & mask) == mask ? 0 : EACCES; 1574 } 1575 1576 /* Otherwise, check everyone else. */ 1577 if (acc_mode & VEXEC) 1578 mask |= S_IXOTH; 1579 if (acc_mode & VREAD) 1580 mask |= S_IROTH; 1581 if (acc_mode & VWRITE) 1582 mask |= S_IWOTH; 1583 return (file_mode & mask) == mask ? 0 : EACCES; 1584 } 1585 1586 int 1587 vfs_readonly(struct mount *mp, struct proc *p) 1588 { 1589 int error; 1590 1591 error = vfs_busy(mp, VB_WRITE|VB_WAIT); 1592 if (error) { 1593 printf("%s: busy\n", mp->mnt_stat.f_mntonname); 1594 return (error); 1595 } 1596 uvm_vnp_sync(mp); 1597 error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p); 1598 if (error) { 1599 printf("%s: failed to sync\n", mp->mnt_stat.f_mntonname); 1600 vfs_unbusy(mp); 1601 return (error); 1602 } 1603 1604 mp->mnt_flag |= MNT_UPDATE | MNT_RDONLY; 1605 mp->mnt_flag &= ~MNT_SOFTDEP; 1606 error = VFS_MOUNT(mp, mp->mnt_stat.f_mntonname, NULL, NULL, curproc); 1607 if (error) { 1608 printf("%s: failed to remount rdonly, error %d\n", 1609 mp->mnt_stat.f_mntonname, error); 1610 vfs_unbusy(mp); 1611 return (error); 1612 } 1613 if (mp->mnt_syncer != NULL) 1614 vgone(mp->mnt_syncer); 1615 mp->mnt_syncer = NULL; 1616 vfs_unbusy(mp); 1617 return (error); 1618 } 1619 1620 /* 1621 * Read-only all file systems. 1622 * We traverse the list in reverse order under the assumption that doing so 1623 * will avoid needing to worry about dependencies. 1624 */ 1625 void 1626 vfs_rofs(struct proc *p) 1627 { 1628 struct mount *mp, *nmp; 1629 1630 TAILQ_FOREACH_REVERSE_SAFE(mp, &mountlist, mntlist, mnt_list, nmp) { 1631 /* XXX Here is a race, the next pointer is not locked. */ 1632 (void) vfs_readonly(mp, p); 1633 } 1634 } 1635 1636 /* 1637 * Sync and unmount file systems before shutting down. 1638 */ 1639 void 1640 vfs_shutdown(struct proc *p) 1641 { 1642 #ifdef ACCOUNTING 1643 acct_shutdown(); 1644 #endif 1645 1646 printf("syncing disks... "); 1647 1648 if (panicstr == 0) { 1649 /* Take all filesystems to read-only */ 1650 sys_sync(p, NULL, NULL); 1651 vfs_rofs(p); 1652 } 1653 1654 if (vfs_syncwait(p, 1)) 1655 printf("giving up\n"); 1656 else 1657 printf("done\n"); 1658 1659 #if NSOFTRAID > 0 1660 sr_shutdown(1); 1661 #endif 1662 } 1663 1664 /* 1665 * perform sync() operation and wait for buffers to flush. 1666 */ 1667 int 1668 vfs_syncwait(struct proc *p, int verbose) 1669 { 1670 struct buf *bp; 1671 int iter, nbusy, dcount, s; 1672 #ifdef MULTIPROCESSOR 1673 int hold_count; 1674 #endif 1675 1676 sys_sync(p, NULL, NULL); 1677 1678 /* Wait for sync to finish. */ 1679 dcount = 10000; 1680 for (iter = 0; iter < 20; iter++) { 1681 nbusy = 0; 1682 LIST_FOREACH(bp, &bufhead, b_list) { 1683 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1684 nbusy++; 1685 /* 1686 * With soft updates, some buffers that are 1687 * written will be remarked as dirty until other 1688 * buffers are written. 1689 */ 1690 if (bp->b_flags & B_DELWRI) { 1691 s = splbio(); 1692 bremfree(bp); 1693 buf_acquire(bp); 1694 splx(s); 1695 nbusy++; 1696 bawrite(bp); 1697 if (dcount-- <= 0) { 1698 if (verbose) 1699 printf("softdep "); 1700 return 1; 1701 } 1702 } 1703 } 1704 if (nbusy == 0) 1705 break; 1706 if (verbose) 1707 printf("%d ", nbusy); 1708 #ifdef MULTIPROCESSOR 1709 if (_kernel_lock_held()) 1710 hold_count = __mp_release_all(&kernel_lock); 1711 else 1712 hold_count = 0; 1713 #endif 1714 DELAY(40000 * iter); 1715 #ifdef MULTIPROCESSOR 1716 if (hold_count) 1717 __mp_acquire_count(&kernel_lock, hold_count); 1718 #endif 1719 } 1720 1721 return nbusy; 1722 } 1723 1724 /* 1725 * posix file system related system variables. 1726 */ 1727 int 1728 fs_posix_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, 1729 void *newp, size_t newlen, struct proc *p) 1730 { 1731 /* all sysctl names at this level are terminal */ 1732 if (namelen != 1) 1733 return (ENOTDIR); 1734 1735 switch (name[0]) { 1736 case FS_POSIX_SETUID: 1737 if (newp && securelevel > 0) 1738 return (EPERM); 1739 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1740 default: 1741 return (EOPNOTSUPP); 1742 } 1743 /* NOTREACHED */ 1744 } 1745 1746 /* 1747 * file system related system variables. 1748 */ 1749 int 1750 fs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1751 size_t newlen, struct proc *p) 1752 { 1753 sysctlfn *fn; 1754 1755 switch (name[0]) { 1756 case FS_POSIX: 1757 fn = fs_posix_sysctl; 1758 break; 1759 default: 1760 return (EOPNOTSUPP); 1761 } 1762 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1763 } 1764 1765 1766 /* 1767 * Routines dealing with vnodes and buffers 1768 */ 1769 1770 /* 1771 * Wait for all outstanding I/Os to complete 1772 * 1773 * Manipulates v_numoutput. Must be called at splbio() 1774 */ 1775 int 1776 vwaitforio(struct vnode *vp, int slpflag, char *wmesg, int timeo) 1777 { 1778 int error = 0; 1779 1780 splassert(IPL_BIO); 1781 1782 while (vp->v_numoutput) { 1783 vp->v_bioflag |= VBIOWAIT; 1784 error = tsleep(&vp->v_numoutput, 1785 slpflag | (PRIBIO + 1), wmesg, timeo); 1786 if (error) 1787 break; 1788 } 1789 1790 return (error); 1791 } 1792 1793 /* 1794 * Update outstanding I/O count and do wakeup if requested. 1795 * 1796 * Manipulates v_numoutput. Must be called at splbio() 1797 */ 1798 void 1799 vwakeup(struct vnode *vp) 1800 { 1801 splassert(IPL_BIO); 1802 1803 if (vp != NULL) { 1804 if (vp->v_numoutput-- == 0) 1805 panic("vwakeup: neg numoutput"); 1806 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1807 vp->v_bioflag &= ~VBIOWAIT; 1808 wakeup(&vp->v_numoutput); 1809 } 1810 } 1811 } 1812 1813 /* 1814 * Flush out and invalidate all buffers associated with a vnode. 1815 * Called with the underlying object locked. 1816 */ 1817 int 1818 vinvalbuf(struct vnode *vp, int flags, struct ucred *cred, struct proc *p, 1819 int slpflag, int slptimeo) 1820 { 1821 struct buf *bp; 1822 struct buf *nbp, *blist; 1823 int s, error; 1824 1825 #ifdef VFSLCKDEBUG 1826 if ((vp->v_flag & VLOCKSWORK) && !VOP_ISLOCKED(vp)) 1827 panic("vinvalbuf(): vp isn't locked"); 1828 #endif 1829 1830 if (flags & V_SAVE) { 1831 s = splbio(); 1832 vwaitforio(vp, 0, "vinvalbuf", 0); 1833 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1834 splx(s); 1835 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1836 return (error); 1837 s = splbio(); 1838 if (vp->v_numoutput > 0 || 1839 !LIST_EMPTY(&vp->v_dirtyblkhd)) 1840 panic("vinvalbuf: dirty bufs"); 1841 } 1842 splx(s); 1843 } 1844 loop: 1845 s = splbio(); 1846 for (;;) { 1847 if ((blist = LIST_FIRST(&vp->v_cleanblkhd)) && 1848 (flags & V_SAVEMETA)) 1849 while (blist && blist->b_lblkno < 0) 1850 blist = LIST_NEXT(blist, b_vnbufs); 1851 if (blist == NULL && 1852 (blist = LIST_FIRST(&vp->v_dirtyblkhd)) && 1853 (flags & V_SAVEMETA)) 1854 while (blist && blist->b_lblkno < 0) 1855 blist = LIST_NEXT(blist, b_vnbufs); 1856 if (!blist) 1857 break; 1858 1859 for (bp = blist; bp; bp = nbp) { 1860 nbp = LIST_NEXT(bp, b_vnbufs); 1861 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1862 continue; 1863 if (bp->b_flags & B_BUSY) { 1864 bp->b_flags |= B_WANTED; 1865 error = tsleep(bp, slpflag | (PRIBIO + 1), 1866 "vinvalbuf", slptimeo); 1867 if (error) { 1868 splx(s); 1869 return (error); 1870 } 1871 break; 1872 } 1873 bremfree(bp); 1874 /* 1875 * XXX Since there are no node locks for NFS, I believe 1876 * there is a slight chance that a delayed write will 1877 * occur while sleeping just above, so check for it. 1878 */ 1879 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1880 buf_acquire(bp); 1881 splx(s); 1882 (void) VOP_BWRITE(bp); 1883 goto loop; 1884 } 1885 buf_acquire_nomap(bp); 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 LIST_FOREACH_SAFE(bp, &vp->v_dirtyblkhd, b_vnbufs, nbp) { 1906 if ((bp->b_flags & B_BUSY)) 1907 continue; 1908 if ((bp->b_flags & B_DELWRI) == 0) 1909 panic("vflushbuf: not dirty"); 1910 bremfree(bp); 1911 buf_acquire(bp); 1912 splx(s); 1913 /* 1914 * Wait for I/O associated with indirect blocks to complete, 1915 * since there is no way to quickly wait for them below. 1916 */ 1917 if (bp->b_vp == vp || sync == 0) 1918 (void) bawrite(bp); 1919 else 1920 (void) bwrite(bp); 1921 goto loop; 1922 } 1923 if (sync == 0) { 1924 splx(s); 1925 return; 1926 } 1927 vwaitforio(vp, 0, "vflushbuf", 0); 1928 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1929 splx(s); 1930 #ifdef DIAGNOSTIC 1931 vprint("vflushbuf: dirty", vp); 1932 #endif 1933 goto loop; 1934 } 1935 splx(s); 1936 } 1937 1938 /* 1939 * Associate a buffer with a vnode. 1940 * 1941 * Manipulates buffer vnode queues. Must be called at splbio(). 1942 */ 1943 void 1944 bgetvp(struct vnode *vp, struct buf *bp) 1945 { 1946 splassert(IPL_BIO); 1947 1948 1949 if (bp->b_vp) 1950 panic("bgetvp: not free"); 1951 vhold(vp); 1952 bp->b_vp = vp; 1953 if (vp->v_type == VBLK || vp->v_type == VCHR) 1954 bp->b_dev = vp->v_rdev; 1955 else 1956 bp->b_dev = NODEV; 1957 /* 1958 * Insert onto list for new vnode. 1959 */ 1960 bufinsvn(bp, &vp->v_cleanblkhd); 1961 } 1962 1963 /* 1964 * Disassociate a buffer from a vnode. 1965 * 1966 * Manipulates vnode buffer queues. Must be called at splbio(). 1967 */ 1968 void 1969 brelvp(struct buf *bp) 1970 { 1971 struct vnode *vp; 1972 1973 splassert(IPL_BIO); 1974 1975 if ((vp = bp->b_vp) == (struct vnode *) 0) 1976 panic("brelvp: NULL"); 1977 /* 1978 * Delete from old vnode list, if on one. 1979 */ 1980 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 1981 bufremvn(bp); 1982 if ((vp->v_bioflag & VBIOONSYNCLIST) && 1983 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 1984 vp->v_bioflag &= ~VBIOONSYNCLIST; 1985 LIST_REMOVE(vp, v_synclist); 1986 } 1987 bp->b_vp = NULL; 1988 1989 vdrop(vp); 1990 } 1991 1992 /* 1993 * Replaces the current vnode associated with the buffer, if any, 1994 * with a new vnode. 1995 * 1996 * If an output I/O is pending on the buffer, the old vnode 1997 * I/O count is adjusted. 1998 * 1999 * Ignores vnode buffer queues. Must be called at splbio(). 2000 */ 2001 void 2002 buf_replacevnode(struct buf *bp, struct vnode *newvp) 2003 { 2004 struct vnode *oldvp = bp->b_vp; 2005 2006 splassert(IPL_BIO); 2007 2008 if (oldvp) 2009 brelvp(bp); 2010 2011 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2012 newvp->v_numoutput++; /* put it on swapdev */ 2013 vwakeup(oldvp); 2014 } 2015 2016 bgetvp(newvp, bp); 2017 bufremvn(bp); 2018 } 2019 2020 /* 2021 * Used to assign buffers to the appropriate clean or dirty list on 2022 * the vnode and to add newly dirty vnodes to the appropriate 2023 * filesystem syncer list. 2024 * 2025 * Manipulates vnode buffer queues. Must be called at splbio(). 2026 */ 2027 void 2028 reassignbuf(struct buf *bp) 2029 { 2030 struct buflists *listheadp; 2031 int delay; 2032 struct vnode *vp = bp->b_vp; 2033 2034 splassert(IPL_BIO); 2035 2036 /* 2037 * Delete from old vnode list, if on one. 2038 */ 2039 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2040 bufremvn(bp); 2041 2042 /* 2043 * If dirty, put on list of dirty buffers; 2044 * otherwise insert onto list of clean buffers. 2045 */ 2046 if ((bp->b_flags & B_DELWRI) == 0) { 2047 listheadp = &vp->v_cleanblkhd; 2048 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2049 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2050 vp->v_bioflag &= ~VBIOONSYNCLIST; 2051 LIST_REMOVE(vp, v_synclist); 2052 } 2053 } else { 2054 listheadp = &vp->v_dirtyblkhd; 2055 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2056 switch (vp->v_type) { 2057 case VDIR: 2058 delay = syncdelay / 2; 2059 break; 2060 case VBLK: 2061 if (vp->v_specmountpoint != NULL) { 2062 delay = syncdelay / 3; 2063 break; 2064 } 2065 /* FALLTHROUGH */ 2066 default: 2067 delay = syncdelay; 2068 } 2069 vn_syncer_add_to_worklist(vp, delay); 2070 } 2071 } 2072 bufinsvn(bp, listheadp); 2073 } 2074 2075 int 2076 vfs_register(struct vfsconf *vfs) 2077 { 2078 struct vfsconf *vfsp; 2079 struct vfsconf **vfspp; 2080 2081 #ifdef DIAGNOSTIC 2082 /* Paranoia? */ 2083 if (vfs->vfc_refcount != 0) 2084 printf("vfs_register called with vfc_refcount > 0\n"); 2085 #endif 2086 2087 /* Check if filesystem already known */ 2088 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2089 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2090 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2091 return (EEXIST); 2092 2093 if (vfs->vfc_typenum > maxvfsconf) 2094 maxvfsconf = vfs->vfc_typenum; 2095 2096 vfs->vfc_next = NULL; 2097 2098 /* Add to the end of the list */ 2099 *vfspp = vfs; 2100 2101 /* Call vfs_init() */ 2102 if (vfs->vfc_vfsops->vfs_init) 2103 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2104 2105 return 0; 2106 } 2107 2108 int 2109 vfs_unregister(struct vfsconf *vfs) 2110 { 2111 struct vfsconf *vfsp; 2112 struct vfsconf **vfspp; 2113 int maxtypenum; 2114 2115 /* Find our vfsconf struct */ 2116 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2117 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2118 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2119 break; 2120 } 2121 2122 if (!vfsp) /* Not found */ 2123 return (ENOENT); 2124 2125 if (vfsp->vfc_refcount) /* In use */ 2126 return (EBUSY); 2127 2128 /* Remove from list and free */ 2129 *vfspp = vfsp->vfc_next; 2130 2131 maxtypenum = 0; 2132 2133 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2134 if (vfsp->vfc_typenum > maxtypenum) 2135 maxtypenum = vfsp->vfc_typenum; 2136 2137 maxvfsconf = maxtypenum; 2138 return 0; 2139 } 2140 2141 /* 2142 * Check if vnode represents a disk device 2143 */ 2144 int 2145 vn_isdisk(struct vnode *vp, int *errp) 2146 { 2147 if (vp->v_type != VBLK && vp->v_type != VCHR) 2148 return (0); 2149 2150 return (1); 2151 } 2152 2153 #ifdef DDB 2154 #include <machine/db_machdep.h> 2155 #include <ddb/db_interface.h> 2156 2157 void 2158 vfs_buf_print(void *b, int full, 2159 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2160 { 2161 struct buf *bp = b; 2162 2163 (*pr)(" vp %p lblkno 0x%llx blkno 0x%llx dev 0x%x\n" 2164 " proc %p error %d flags %lb\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\n" 2169 " data %p saveaddr %p dep %p iodone %p\n", 2170 bp->b_bufsize, bp->b_bcount, (long)bp->b_resid, 2171 bp->b_data, bp->b_saveaddr, 2172 LIST_FIRST(&bp->b_dep), bp->b_iodone); 2173 2174 (*pr)(" dirty {off 0x%x end 0x%x} valid {off 0x%x end 0x%x}\n", 2175 bp->b_dirtyoff, bp->b_dirtyend, bp->b_validoff, bp->b_validend); 2176 2177 #ifdef FFS_SOFTUPDATES 2178 if (full) 2179 softdep_print(bp, full, pr); 2180 #endif 2181 } 2182 2183 const char *vtypes[] = { VTYPE_NAMES }; 2184 const char *vtags[] = { VTAG_NAMES }; 2185 2186 void 2187 vfs_vnode_print(void *v, int full, 2188 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2189 { 2190 struct vnode *vp = v; 2191 2192 (*pr)("tag %s(%d) type %s(%d) mount %p typedata %p\n", 2193 (u_int)vp->v_tag >= nitems(vtags)? "<unk>":vtags[vp->v_tag], 2194 vp->v_tag, 2195 (u_int)vp->v_type >= nitems(vtypes)? "<unk>":vtypes[vp->v_type], 2196 vp->v_type, vp->v_mount, vp->v_mountedhere); 2197 2198 (*pr)("data %p usecount %d writecount %d holdcnt %d numoutput %d\n", 2199 vp->v_data, vp->v_usecount, vp->v_writecount, 2200 vp->v_holdcnt, vp->v_numoutput); 2201 2202 /* uvm_object_printit(&vp->v_uobj, full, pr); */ 2203 2204 if (full) { 2205 struct buf *bp; 2206 2207 (*pr)("clean bufs:\n"); 2208 LIST_FOREACH(bp, &vp->v_cleanblkhd, b_vnbufs) { 2209 (*pr)(" bp %p\n", bp); 2210 vfs_buf_print(bp, full, pr); 2211 } 2212 2213 (*pr)("dirty bufs:\n"); 2214 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) { 2215 (*pr)(" bp %p\n", bp); 2216 vfs_buf_print(bp, full, pr); 2217 } 2218 } 2219 } 2220 2221 void 2222 vfs_mount_print(struct mount *mp, int full, 2223 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2224 { 2225 struct vfsconf *vfc = mp->mnt_vfc; 2226 struct vnode *vp; 2227 int cnt; 2228 2229 (*pr)("flags %b\nvnodecovered %p syncer %p data %p\n", 2230 mp->mnt_flag, MNT_BITS, 2231 mp->mnt_vnodecovered, mp->mnt_syncer, mp->mnt_data); 2232 2233 (*pr)("vfsconf: ops %p name \"%s\" num %d ref %d flags 0x%x\n", 2234 vfc->vfc_vfsops, vfc->vfc_name, vfc->vfc_typenum, 2235 vfc->vfc_refcount, vfc->vfc_flags); 2236 2237 (*pr)("statvfs cache: bsize %x iosize %x\nblocks %llu free %llu avail %lld\n", 2238 mp->mnt_stat.f_bsize, mp->mnt_stat.f_iosize, mp->mnt_stat.f_blocks, 2239 mp->mnt_stat.f_bfree, mp->mnt_stat.f_bavail); 2240 2241 (*pr)(" files %llu ffiles %llu favail %lld\n", mp->mnt_stat.f_files, 2242 mp->mnt_stat.f_ffree, mp->mnt_stat.f_favail); 2243 2244 (*pr)(" f_fsidx {0x%x, 0x%x} owner %u ctime 0x%llx\n", 2245 mp->mnt_stat.f_fsid.val[0], mp->mnt_stat.f_fsid.val[1], 2246 mp->mnt_stat.f_owner, mp->mnt_stat.f_ctime); 2247 2248 (*pr)(" syncwrites %llu asyncwrites = %llu\n", 2249 mp->mnt_stat.f_syncwrites, mp->mnt_stat.f_asyncwrites); 2250 2251 (*pr)(" syncreads %llu asyncreads = %llu\n", 2252 mp->mnt_stat.f_syncreads, mp->mnt_stat.f_asyncreads); 2253 2254 (*pr)(" fstype \"%s\" mnton \"%s\" mntfrom \"%s\" mntspec \"%s\"\n", 2255 mp->mnt_stat.f_fstypename, mp->mnt_stat.f_mntonname, 2256 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntfromspec); 2257 2258 (*pr)("locked vnodes:"); 2259 /* XXX would take mountlist lock, except ddb has no context */ 2260 cnt = 0; 2261 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2262 if (VOP_ISLOCKED(vp)) { 2263 if (cnt == 0) 2264 (*pr)("\n %p", vp); 2265 else if ((cnt % (72 / (sizeof(void *) * 2 + 4))) == 0) 2266 (*pr)(",\n %p", vp); 2267 else 2268 (*pr)(", %p", vp); 2269 cnt++; 2270 } 2271 } 2272 (*pr)("\n"); 2273 2274 if (full) { 2275 (*pr)("all vnodes:"); 2276 /* XXX would take mountlist lock, except ddb has no context */ 2277 cnt = 0; 2278 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2279 if (cnt == 0) 2280 (*pr)("\n %p", vp); 2281 else if ((cnt % (72 / (sizeof(void *) * 2 + 4))) == 0) 2282 (*pr)(",\n %p", vp); 2283 else 2284 (*pr)(", %p", vp); 2285 cnt++; 2286 } 2287 (*pr)("\n"); 2288 } 2289 } 2290 #endif /* DDB */ 2291 2292 void 2293 copy_statfs_info(struct statfs *sbp, const struct mount *mp) 2294 { 2295 const struct statfs *mbp; 2296 2297 strncpy(sbp->f_fstypename, mp->mnt_vfc->vfc_name, MFSNAMELEN); 2298 2299 if (sbp == (mbp = &mp->mnt_stat)) 2300 return; 2301 2302 sbp->f_fsid = mbp->f_fsid; 2303 sbp->f_owner = mbp->f_owner; 2304 sbp->f_flags = mbp->f_flags; 2305 sbp->f_syncwrites = mbp->f_syncwrites; 2306 sbp->f_asyncwrites = mbp->f_asyncwrites; 2307 sbp->f_syncreads = mbp->f_syncreads; 2308 sbp->f_asyncreads = mbp->f_asyncreads; 2309 sbp->f_namemax = mbp->f_namemax; 2310 memcpy(sbp->f_mntonname, mp->mnt_stat.f_mntonname, MNAMELEN); 2311 memcpy(sbp->f_mntfromname, mp->mnt_stat.f_mntfromname, MNAMELEN); 2312 memcpy(sbp->f_mntfromspec, mp->mnt_stat.f_mntfromspec, MNAMELEN); 2313 memcpy(&sbp->mount_info, &mp->mnt_stat.mount_info, 2314 sizeof(union mount_info)); 2315 } 2316