1 /* $OpenBSD: vfs_subr.c,v 1.257 2017/01/15 23:18:05 bluhm 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(void); 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 rn_init(sizeof(struct sockaddr_in)); 159 } 160 161 /* 162 * Mark a mount point as busy. Used to synchronize access and to delay 163 * unmounting. 164 * 165 * Default behaviour is to attempt getting a READ lock and in case of an 166 * ongoing unmount, to wait for it to finish and then return failure. 167 */ 168 int 169 vfs_busy(struct mount *mp, int flags) 170 { 171 int rwflags = 0; 172 173 /* new mountpoints need their lock initialised */ 174 if (mp->mnt_lock.rwl_name == NULL) 175 rw_init(&mp->mnt_lock, "vfslock"); 176 177 if (flags & VB_WRITE) 178 rwflags |= RW_WRITE; 179 else 180 rwflags |= RW_READ; 181 182 if (flags & VB_WAIT) 183 rwflags |= RW_SLEEPFAIL; 184 else 185 rwflags |= RW_NOSLEEP; 186 187 if (rw_enter(&mp->mnt_lock, rwflags)) 188 return (EBUSY); 189 190 return (0); 191 } 192 193 /* 194 * Free a busy file system 195 */ 196 void 197 vfs_unbusy(struct mount *mp) 198 { 199 rw_exit(&mp->mnt_lock); 200 } 201 202 int 203 vfs_isbusy(struct mount *mp) 204 { 205 if (RWLOCK_OWNER(&mp->mnt_lock) > 0) 206 return (1); 207 else 208 return (0); 209 } 210 211 /* 212 * Lookup a filesystem type, and if found allocate and initialize 213 * a mount structure for it. 214 * 215 * Devname is usually updated by mount(8) after booting. 216 */ 217 int 218 vfs_rootmountalloc(char *fstypename, char *devname, struct mount **mpp) 219 { 220 struct vfsconf *vfsp; 221 struct mount *mp; 222 223 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 224 if (!strcmp(vfsp->vfc_name, fstypename)) 225 break; 226 if (vfsp == NULL) 227 return (ENODEV); 228 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 229 (void)vfs_busy(mp, VB_READ|VB_NOWAIT); 230 LIST_INIT(&mp->mnt_vnodelist); 231 mp->mnt_vfc = vfsp; 232 mp->mnt_op = vfsp->vfc_vfsops; 233 mp->mnt_flag = MNT_RDONLY; 234 mp->mnt_vnodecovered = NULLVP; 235 vfsp->vfc_refcount++; 236 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 237 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 238 mp->mnt_stat.f_mntonname[0] = '/'; 239 copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN, 0); 240 copystr(devname, mp->mnt_stat.f_mntfromspec, MNAMELEN, 0); 241 *mpp = mp; 242 return (0); 243 } 244 245 /* 246 * Lookup a mount point by filesystem identifier. 247 */ 248 struct mount * 249 vfs_getvfs(fsid_t *fsid) 250 { 251 struct mount *mp; 252 253 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 254 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 255 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 256 return (mp); 257 } 258 } 259 260 return (NULL); 261 } 262 263 264 /* 265 * Get a new unique fsid 266 */ 267 void 268 vfs_getnewfsid(struct mount *mp) 269 { 270 static u_short xxxfs_mntid; 271 272 fsid_t tfsid; 273 int mtype; 274 275 mtype = mp->mnt_vfc->vfc_typenum; 276 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 277 mp->mnt_stat.f_fsid.val[1] = mtype; 278 if (xxxfs_mntid == 0) 279 ++xxxfs_mntid; 280 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 281 tfsid.val[1] = mtype; 282 if (!TAILQ_EMPTY(&mountlist)) { 283 while (vfs_getvfs(&tfsid)) { 284 tfsid.val[0]++; 285 xxxfs_mntid++; 286 } 287 } 288 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 289 } 290 291 /* 292 * Set vnode attributes to VNOVAL 293 */ 294 void 295 vattr_null(struct vattr *vap) 296 { 297 298 vap->va_type = VNON; 299 /* 300 * Don't get fancy: u_quad_t = u_int = VNOVAL leaves the u_quad_t 301 * with 2^31-1 instead of 2^64-1. Just write'm out and let 302 * the compiler do its job. 303 */ 304 vap->va_mode = VNOVAL; 305 vap->va_nlink = VNOVAL; 306 vap->va_uid = VNOVAL; 307 vap->va_gid = VNOVAL; 308 vap->va_fsid = VNOVAL; 309 vap->va_fileid = VNOVAL; 310 vap->va_size = VNOVAL; 311 vap->va_blocksize = VNOVAL; 312 vap->va_atime.tv_sec = VNOVAL; 313 vap->va_atime.tv_nsec = VNOVAL; 314 vap->va_mtime.tv_sec = VNOVAL; 315 vap->va_mtime.tv_nsec = VNOVAL; 316 vap->va_ctime.tv_sec = VNOVAL; 317 vap->va_ctime.tv_nsec = VNOVAL; 318 vap->va_gen = VNOVAL; 319 vap->va_flags = VNOVAL; 320 vap->va_rdev = VNOVAL; 321 vap->va_bytes = VNOVAL; 322 vap->va_filerev = VNOVAL; 323 vap->va_vaflags = 0; 324 } 325 326 /* 327 * Routines having to do with the management of the vnode table. 328 */ 329 long numvnodes; 330 331 /* 332 * Return the next vnode from the free list. 333 */ 334 int 335 getnewvnode(enum vtagtype tag, struct mount *mp, struct vops *vops, 336 struct vnode **vpp) 337 { 338 struct proc *p = curproc; 339 struct freelst *listhd; 340 static int toggle; 341 struct vnode *vp; 342 int s; 343 344 /* 345 * allow maxvnodes to increase if the buffer cache itself 346 * is big enough to justify it. (we don't shrink it ever) 347 */ 348 maxvnodes = maxvnodes < bcstats.numbufs ? bcstats.numbufs 349 : maxvnodes; 350 351 /* 352 * We must choose whether to allocate a new vnode or recycle an 353 * existing one. The criterion for allocating a new one is that 354 * the total number of vnodes is less than the number desired or 355 * there are no vnodes on either free list. Generally we only 356 * want to recycle vnodes that have no buffers associated with 357 * them, so we look first on the vnode_free_list. If it is empty, 358 * we next consider vnodes with referencing buffers on the 359 * vnode_hold_list. The toggle ensures that half the time we 360 * will use a buffer from the vnode_hold_list, and half the time 361 * we will allocate a new one unless the list has grown to twice 362 * the desired size. We are reticent to recycle vnodes from the 363 * vnode_hold_list because we will lose the identity of all its 364 * referencing buffers. 365 */ 366 toggle ^= 1; 367 if (numvnodes / 2 > maxvnodes) 368 toggle = 0; 369 370 s = splbio(); 371 if ((numvnodes < maxvnodes) || 372 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 373 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 374 splx(s); 375 vp = pool_get(&vnode_pool, PR_WAITOK | PR_ZERO); 376 vp->v_uvm = pool_get(&uvm_vnode_pool, PR_WAITOK | PR_ZERO); 377 vp->v_uvm->u_vnode = vp; 378 RBT_INIT(buf_rb_bufs, &vp->v_bufs_tree); 379 cache_tree_init(&vp->v_nc_tree); 380 TAILQ_INIT(&vp->v_cache_dst); 381 numvnodes++; 382 } else { 383 TAILQ_FOREACH(vp, listhd, v_freelist) { 384 if (VOP_ISLOCKED(vp) == 0) 385 break; 386 } 387 /* 388 * Unless this is a bad time of the month, at most 389 * the first NCPUS items on the free list are 390 * locked, so this is close enough to being empty. 391 */ 392 if (vp == NULL) { 393 splx(s); 394 tablefull("vnode"); 395 *vpp = 0; 396 return (ENFILE); 397 } 398 399 #ifdef DIAGNOSTIC 400 if (vp->v_usecount) { 401 vprint("free vnode", vp); 402 panic("free vnode isn't"); 403 } 404 #endif 405 406 TAILQ_REMOVE(listhd, vp, v_freelist); 407 vp->v_bioflag &= ~VBIOONFREELIST; 408 splx(s); 409 410 if (vp->v_type != VBAD) 411 vgonel(vp, p); 412 #ifdef DIAGNOSTIC 413 if (vp->v_data) { 414 vprint("cleaned vnode", vp); 415 panic("cleaned vnode isn't"); 416 } 417 s = splbio(); 418 if (vp->v_numoutput) 419 panic("Clean vnode has pending I/O's"); 420 splx(s); 421 #endif 422 vp->v_flag = 0; 423 vp->v_socket = 0; 424 } 425 cache_purge(vp); 426 vp->v_type = VNON; 427 vp->v_tag = tag; 428 vp->v_op = vops; 429 insmntque(vp, mp); 430 *vpp = vp; 431 vp->v_usecount = 1; 432 vp->v_data = 0; 433 return (0); 434 } 435 436 /* 437 * Move a vnode from one mount queue to another. 438 */ 439 void 440 insmntque(struct vnode *vp, struct mount *mp) 441 { 442 /* 443 * Delete from old mount point vnode list, if on one. 444 */ 445 if (vp->v_mount != NULL) 446 LIST_REMOVE(vp, v_mntvnodes); 447 /* 448 * Insert into list of vnodes for the new mount point, if available. 449 */ 450 if ((vp->v_mount = mp) != NULL) 451 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 452 } 453 454 /* 455 * Create a vnode for a block device. 456 * Used for root filesystem, argdev, and swap areas. 457 * Also used for memory file system special devices. 458 */ 459 int 460 bdevvp(dev_t dev, struct vnode **vpp) 461 { 462 return (getdevvp(dev, vpp, VBLK)); 463 } 464 465 /* 466 * Create a vnode for a character device. 467 * Used for console handling. 468 */ 469 int 470 cdevvp(dev_t dev, struct vnode **vpp) 471 { 472 return (getdevvp(dev, vpp, VCHR)); 473 } 474 475 /* 476 * Create a vnode for a device. 477 * Used by bdevvp (block device) for root file system etc., 478 * and by cdevvp (character device) for console. 479 */ 480 int 481 getdevvp(dev_t dev, struct vnode **vpp, enum vtype type) 482 { 483 struct vnode *vp; 484 struct vnode *nvp; 485 int error; 486 487 if (dev == NODEV) { 488 *vpp = NULLVP; 489 return (0); 490 } 491 error = getnewvnode(VT_NON, NULL, &spec_vops, &nvp); 492 if (error) { 493 *vpp = NULLVP; 494 return (error); 495 } 496 vp = nvp; 497 vp->v_type = type; 498 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 499 vput(vp); 500 vp = nvp; 501 } 502 if (vp->v_type == VCHR && cdevsw[major(vp->v_rdev)].d_type == D_TTY) 503 vp->v_flag |= VISTTY; 504 *vpp = vp; 505 return (0); 506 } 507 508 /* 509 * Check to see if the new vnode represents a special device 510 * for which we already have a vnode (either because of 511 * bdevvp() or because of a different vnode representing 512 * the same block device). If such an alias exists, deallocate 513 * the existing contents and return the aliased vnode. The 514 * caller is responsible for filling it with its new contents. 515 */ 516 struct vnode * 517 checkalias(struct vnode *nvp, dev_t nvp_rdev, struct mount *mp) 518 { 519 struct proc *p = curproc; 520 struct vnode *vp; 521 struct vnode **vpp; 522 523 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 524 return (NULLVP); 525 526 vpp = &speclisth[SPECHASH(nvp_rdev)]; 527 loop: 528 for (vp = *vpp; vp; vp = vp->v_specnext) { 529 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 530 continue; 531 } 532 /* 533 * Alias, but not in use, so flush it out. 534 */ 535 if (vp->v_usecount == 0) { 536 vgonel(vp, p); 537 goto loop; 538 } 539 if (vget(vp, LK_EXCLUSIVE, p)) { 540 goto loop; 541 } 542 break; 543 } 544 545 /* 546 * Common case is actually in the if statement 547 */ 548 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 549 nvp->v_specinfo = malloc(sizeof(struct specinfo), M_VNODE, 550 M_WAITOK); 551 nvp->v_rdev = nvp_rdev; 552 nvp->v_hashchain = vpp; 553 nvp->v_specnext = *vpp; 554 nvp->v_specmountpoint = NULL; 555 nvp->v_speclockf = NULL; 556 nvp->v_specbitmap = NULL; 557 if (nvp->v_type == VCHR && 558 (cdevsw[major(nvp_rdev)].d_flags & D_CLONE) && 559 (minor(nvp_rdev) >> CLONE_SHIFT == 0)) { 560 if (vp != NULLVP) 561 nvp->v_specbitmap = vp->v_specbitmap; 562 else 563 nvp->v_specbitmap = malloc(CLONE_MAPSZ, 564 M_VNODE, M_WAITOK | M_ZERO); 565 } 566 *vpp = nvp; 567 if (vp != NULLVP) { 568 nvp->v_flag |= VALIASED; 569 vp->v_flag |= VALIASED; 570 vput(vp); 571 } 572 return (NULLVP); 573 } 574 575 /* 576 * This code is the uncommon case. It is called in case 577 * we found an alias that was VT_NON && vtype of VBLK 578 * This means we found a block device that was created 579 * using bdevvp. 580 * An example of such a vnode is the root partition device vnode 581 * created in ffs_mountroot. 582 * 583 * The vnodes created by bdevvp should not be aliased (why?). 584 */ 585 586 VOP_UNLOCK(vp, p); 587 vclean(vp, 0, p); 588 vp->v_op = nvp->v_op; 589 vp->v_tag = nvp->v_tag; 590 nvp->v_type = VNON; 591 insmntque(vp, mp); 592 return (vp); 593 } 594 595 /* 596 * Grab a particular vnode from the free list, increment its 597 * reference count and lock it. If the vnode lock bit is set, 598 * the vnode is being eliminated in vgone. In that case, we 599 * cannot grab it, so the process is awakened when the 600 * transition is completed, and an error code is returned to 601 * indicate that the vnode is no longer usable, possibly 602 * having been changed to a new file system type. 603 */ 604 int 605 vget(struct vnode *vp, int flags, struct proc *p) 606 { 607 int error, s, onfreelist; 608 609 /* 610 * If the vnode is in the process of being cleaned out for 611 * another use, we wait for the cleaning to finish and then 612 * return failure. Cleaning is determined by checking that 613 * the VXLOCK flag is set. 614 */ 615 616 if (vp->v_flag & VXLOCK) { 617 if (flags & LK_NOWAIT) { 618 return (EBUSY); 619 } 620 621 vp->v_flag |= VXWANT; 622 tsleep(vp, PINOD, "vget", 0); 623 return (ENOENT); 624 } 625 626 onfreelist = vp->v_bioflag & VBIOONFREELIST; 627 if (vp->v_usecount == 0 && onfreelist) { 628 s = splbio(); 629 if (vp->v_holdcnt > 0) 630 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 631 else 632 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 633 vp->v_bioflag &= ~VBIOONFREELIST; 634 splx(s); 635 } 636 637 vp->v_usecount++; 638 if (flags & LK_TYPE_MASK) { 639 if ((error = vn_lock(vp, flags, p)) != 0) { 640 vp->v_usecount--; 641 if (vp->v_usecount == 0 && onfreelist) 642 vputonfreelist(vp); 643 } 644 return (error); 645 } 646 647 return (0); 648 } 649 650 651 /* Vnode reference. */ 652 void 653 vref(struct vnode *vp) 654 { 655 #ifdef DIAGNOSTIC 656 if (vp->v_usecount == 0) 657 panic("vref used where vget required"); 658 if (vp->v_type == VNON) 659 panic("vref on a VNON vnode"); 660 #endif 661 vp->v_usecount++; 662 } 663 664 void 665 vputonfreelist(struct vnode *vp) 666 { 667 int s; 668 struct freelst *lst; 669 670 s = splbio(); 671 #ifdef DIAGNOSTIC 672 if (vp->v_usecount != 0) 673 panic("Use count is not zero!"); 674 675 if (vp->v_bioflag & VBIOONFREELIST) { 676 vprint("vnode already on free list: ", vp); 677 panic("vnode already on free list"); 678 } 679 #endif 680 681 vp->v_bioflag |= VBIOONFREELIST; 682 683 if (vp->v_holdcnt > 0) 684 lst = &vnode_hold_list; 685 else 686 lst = &vnode_free_list; 687 688 if (vp->v_type == VBAD) 689 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 690 else 691 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 692 693 splx(s); 694 } 695 696 /* 697 * vput(), just unlock and vrele() 698 */ 699 void 700 vput(struct vnode *vp) 701 { 702 struct proc *p = curproc; 703 704 #ifdef DIAGNOSTIC 705 if (vp == NULL) 706 panic("vput: null vp"); 707 #endif 708 709 #ifdef DIAGNOSTIC 710 if (vp->v_usecount == 0) { 711 vprint("vput: bad ref count", vp); 712 panic("vput: ref cnt"); 713 } 714 #endif 715 vp->v_usecount--; 716 if (vp->v_usecount > 0) { 717 VOP_UNLOCK(vp, p); 718 return; 719 } 720 721 #ifdef DIAGNOSTIC 722 if (vp->v_writecount != 0) { 723 vprint("vput: bad writecount", vp); 724 panic("vput: v_writecount != 0"); 725 } 726 #endif 727 728 VOP_INACTIVE(vp, p); 729 730 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 731 vputonfreelist(vp); 732 } 733 734 /* 735 * Vnode release - use for active VNODES. 736 * If count drops to zero, call inactive routine and return to freelist. 737 * Returns 0 if it did not sleep. 738 */ 739 int 740 vrele(struct vnode *vp) 741 { 742 struct proc *p = curproc; 743 744 #ifdef DIAGNOSTIC 745 if (vp == NULL) 746 panic("vrele: null vp"); 747 #endif 748 #ifdef DIAGNOSTIC 749 if (vp->v_usecount == 0) { 750 vprint("vrele: bad ref count", vp); 751 panic("vrele: ref cnt"); 752 } 753 #endif 754 vp->v_usecount--; 755 if (vp->v_usecount > 0) { 756 return (0); 757 } 758 759 #ifdef DIAGNOSTIC 760 if (vp->v_writecount != 0) { 761 vprint("vrele: bad writecount", vp); 762 panic("vrele: v_writecount != 0"); 763 } 764 #endif 765 766 if (vn_lock(vp, LK_EXCLUSIVE, p)) { 767 #ifdef DIAGNOSTIC 768 vprint("vrele: cannot lock", vp); 769 #endif 770 return (1); 771 } 772 773 VOP_INACTIVE(vp, p); 774 775 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 776 vputonfreelist(vp); 777 return (1); 778 } 779 780 /* Page or buffer structure gets a reference. */ 781 void 782 vhold(struct vnode *vp) 783 { 784 /* 785 * If it is on the freelist and the hold count is currently 786 * zero, move it to the hold list. 787 */ 788 if ((vp->v_bioflag & VBIOONFREELIST) && 789 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 790 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 791 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 792 } 793 vp->v_holdcnt++; 794 } 795 796 /* Lose interest in a vnode. */ 797 void 798 vdrop(struct vnode *vp) 799 { 800 #ifdef DIAGNOSTIC 801 if (vp->v_holdcnt == 0) 802 panic("vdrop: zero holdcnt"); 803 #endif 804 805 vp->v_holdcnt--; 806 807 /* 808 * If it is on the holdlist and the hold count drops to 809 * zero, move it to the free list. 810 */ 811 if ((vp->v_bioflag & VBIOONFREELIST) && 812 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 813 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 814 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 815 } 816 } 817 818 /* 819 * Remove any vnodes in the vnode table belonging to mount point mp. 820 * 821 * If MNT_NOFORCE is specified, there should not be any active ones, 822 * return error if any are found (nb: this is a user error, not a 823 * system error). If MNT_FORCE is specified, detach any active vnodes 824 * that are found. 825 */ 826 #ifdef DEBUG 827 int busyprt = 0; /* print out busy vnodes */ 828 struct ctldebug debug1 = { "busyprt", &busyprt }; 829 #endif 830 831 int 832 vfs_mount_foreach_vnode(struct mount *mp, 833 int (*func)(struct vnode *, void *), void *arg) { 834 struct vnode *vp, *nvp; 835 int error = 0; 836 837 loop: 838 LIST_FOREACH_SAFE(vp , &mp->mnt_vnodelist, v_mntvnodes, nvp) { 839 if (vp->v_mount != mp) 840 goto loop; 841 842 error = func(vp, arg); 843 844 if (error != 0) 845 break; 846 } 847 848 return (error); 849 } 850 851 struct vflush_args { 852 struct vnode *skipvp; 853 int busy; 854 int flags; 855 }; 856 857 int 858 vflush_vnode(struct vnode *vp, void *arg) { 859 struct vflush_args *va = arg; 860 struct proc *p = curproc; 861 862 if (vp == va->skipvp) { 863 return (0); 864 } 865 866 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 867 return (0); 868 } 869 870 /* 871 * If WRITECLOSE is set, only flush out regular file 872 * vnodes open for writing. 873 */ 874 if ((va->flags & WRITECLOSE) && 875 (vp->v_writecount == 0 || vp->v_type != VREG)) { 876 return (0); 877 } 878 879 /* 880 * With v_usecount == 0, all we need to do is clear 881 * out the vnode data structures and we are done. 882 */ 883 if (vp->v_usecount == 0) { 884 vgonel(vp, p); 885 return (0); 886 } 887 888 /* 889 * If FORCECLOSE is set, forcibly close the vnode. 890 * For block or character devices, revert to an 891 * anonymous device. For all other files, just kill them. 892 */ 893 if (va->flags & FORCECLOSE) { 894 if (vp->v_type != VBLK && vp->v_type != VCHR) { 895 vgonel(vp, p); 896 } else { 897 vclean(vp, 0, p); 898 vp->v_op = &spec_vops; 899 insmntque(vp, NULL); 900 } 901 return (0); 902 } 903 904 #ifdef DEBUG 905 if (busyprt) 906 vprint("vflush: busy vnode", vp); 907 #endif 908 va->busy++; 909 return (0); 910 } 911 912 int 913 vflush(struct mount *mp, struct vnode *skipvp, int flags) 914 { 915 struct vflush_args va; 916 va.skipvp = skipvp; 917 va.busy = 0; 918 va.flags = flags; 919 920 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 921 922 if (va.busy) 923 return (EBUSY); 924 return (0); 925 } 926 927 /* 928 * Disassociate the underlying file system from a vnode. 929 */ 930 void 931 vclean(struct vnode *vp, int flags, struct proc *p) 932 { 933 int active; 934 935 /* 936 * Check to see if the vnode is in use. 937 * If so we have to reference it before we clean it out 938 * so that its count cannot fall to zero and generate a 939 * race against ourselves to recycle it. 940 */ 941 if ((active = vp->v_usecount) != 0) 942 vp->v_usecount++; 943 944 /* 945 * Prevent the vnode from being recycled or 946 * brought into use while we clean it out. 947 */ 948 if (vp->v_flag & VXLOCK) 949 panic("vclean: deadlock"); 950 vp->v_flag |= VXLOCK; 951 /* 952 * Even if the count is zero, the VOP_INACTIVE routine may still 953 * have the object locked while it cleans it out. The VOP_LOCK 954 * ensures that the VOP_INACTIVE routine is done with its work. 955 * For active vnodes, it ensures that no other activity can 956 * occur while the underlying object is being cleaned out. 957 */ 958 VOP_LOCK(vp, LK_DRAIN | LK_EXCLUSIVE, p); 959 960 /* 961 * Clean out any VM data associated with the vnode. 962 */ 963 uvm_vnp_terminate(vp); 964 /* 965 * Clean out any buffers associated with the vnode. 966 */ 967 if (flags & DOCLOSE) 968 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 969 /* 970 * If purging an active vnode, it must be closed and 971 * deactivated before being reclaimed. Note that the 972 * VOP_INACTIVE will unlock the vnode 973 */ 974 if (active) { 975 if (flags & DOCLOSE) 976 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 977 VOP_INACTIVE(vp, p); 978 } else { 979 /* 980 * Any other processes trying to obtain this lock must first 981 * wait for VXLOCK to clear, then call the new lock operation. 982 */ 983 VOP_UNLOCK(vp, p); 984 } 985 986 /* 987 * Reclaim the vnode. 988 */ 989 if (VOP_RECLAIM(vp, p)) 990 panic("vclean: cannot reclaim"); 991 if (active) { 992 vp->v_usecount--; 993 if (vp->v_usecount == 0) { 994 if (vp->v_holdcnt > 0) 995 panic("vclean: not clean"); 996 vputonfreelist(vp); 997 } 998 } 999 cache_purge(vp); 1000 1001 /* 1002 * Done with purge, notify sleepers of the grim news. 1003 */ 1004 vp->v_op = &dead_vops; 1005 VN_KNOTE(vp, NOTE_REVOKE); 1006 vp->v_tag = VT_NON; 1007 vp->v_flag &= ~VXLOCK; 1008 #ifdef VFSLCKDEBUG 1009 vp->v_flag &= ~VLOCKSWORK; 1010 #endif 1011 if (vp->v_flag & VXWANT) { 1012 vp->v_flag &= ~VXWANT; 1013 wakeup(vp); 1014 } 1015 } 1016 1017 /* 1018 * Recycle an unused vnode to the front of the free list. 1019 */ 1020 int 1021 vrecycle(struct vnode *vp, struct proc *p) 1022 { 1023 if (vp->v_usecount == 0) { 1024 vgonel(vp, p); 1025 return (1); 1026 } 1027 return (0); 1028 } 1029 1030 /* 1031 * Eliminate all activity associated with a vnode 1032 * in preparation for reuse. 1033 */ 1034 void 1035 vgone(struct vnode *vp) 1036 { 1037 struct proc *p = curproc; 1038 vgonel(vp, p); 1039 } 1040 1041 /* 1042 * vgone, with struct proc. 1043 */ 1044 void 1045 vgonel(struct vnode *vp, struct proc *p) 1046 { 1047 struct vnode *vq; 1048 struct vnode *vx; 1049 1050 /* 1051 * If a vgone (or vclean) is already in progress, 1052 * wait until it is done and return. 1053 */ 1054 if (vp->v_flag & VXLOCK) { 1055 vp->v_flag |= VXWANT; 1056 tsleep(vp, PINOD, "vgone", 0); 1057 return; 1058 } 1059 1060 /* 1061 * Clean out the filesystem specific data. 1062 */ 1063 vclean(vp, DOCLOSE, p); 1064 /* 1065 * Delete from old mount point vnode list, if on one. 1066 */ 1067 if (vp->v_mount != NULL) 1068 insmntque(vp, NULL); 1069 /* 1070 * If special device, remove it from special device alias list 1071 * if it is on one. 1072 */ 1073 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1074 if ((vp->v_flag & VALIASED) == 0 && vp->v_type == VCHR && 1075 (cdevsw[major(vp->v_rdev)].d_flags & D_CLONE) && 1076 (minor(vp->v_rdev) >> CLONE_SHIFT == 0)) { 1077 free(vp->v_specbitmap, M_VNODE, CLONE_MAPSZ); 1078 } 1079 if (*vp->v_hashchain == vp) { 1080 *vp->v_hashchain = vp->v_specnext; 1081 } else { 1082 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1083 if (vq->v_specnext != vp) 1084 continue; 1085 vq->v_specnext = vp->v_specnext; 1086 break; 1087 } 1088 if (vq == NULL) 1089 panic("missing bdev"); 1090 } 1091 if (vp->v_flag & VALIASED) { 1092 vx = NULL; 1093 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1094 if (vq->v_rdev != vp->v_rdev || 1095 vq->v_type != vp->v_type) 1096 continue; 1097 if (vx) 1098 break; 1099 vx = vq; 1100 } 1101 if (vx == NULL) 1102 panic("missing alias"); 1103 if (vq == NULL) 1104 vx->v_flag &= ~VALIASED; 1105 vp->v_flag &= ~VALIASED; 1106 } 1107 free(vp->v_specinfo, M_VNODE, sizeof(struct specinfo)); 1108 vp->v_specinfo = NULL; 1109 } 1110 /* 1111 * If it is on the freelist and not already at the head, 1112 * move it to the head of the list. 1113 */ 1114 vp->v_type = VBAD; 1115 1116 /* 1117 * Move onto the free list, unless we were called from 1118 * getnewvnode and we're not on any free list 1119 */ 1120 if (vp->v_usecount == 0 && 1121 (vp->v_bioflag & VBIOONFREELIST)) { 1122 int s; 1123 1124 s = splbio(); 1125 1126 if (vp->v_holdcnt > 0) 1127 panic("vgonel: not clean"); 1128 1129 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1130 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1131 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1132 } 1133 splx(s); 1134 } 1135 } 1136 1137 /* 1138 * Lookup a vnode by device number. 1139 */ 1140 int 1141 vfinddev(dev_t dev, enum vtype type, struct vnode **vpp) 1142 { 1143 struct vnode *vp; 1144 int rc =0; 1145 1146 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1147 if (dev != vp->v_rdev || type != vp->v_type) 1148 continue; 1149 *vpp = vp; 1150 rc = 1; 1151 break; 1152 } 1153 return (rc); 1154 } 1155 1156 /* 1157 * Revoke all the vnodes corresponding to the specified minor number 1158 * range (endpoints inclusive) of the specified major. 1159 */ 1160 void 1161 vdevgone(int maj, int minl, int minh, enum vtype type) 1162 { 1163 struct vnode *vp; 1164 int mn; 1165 1166 for (mn = minl; mn <= minh; mn++) 1167 if (vfinddev(makedev(maj, mn), type, &vp)) 1168 VOP_REVOKE(vp, REVOKEALL); 1169 } 1170 1171 /* 1172 * Calculate the total number of references to a special device. 1173 */ 1174 int 1175 vcount(struct vnode *vp) 1176 { 1177 struct vnode *vq, *vnext; 1178 int count; 1179 1180 loop: 1181 if ((vp->v_flag & VALIASED) == 0) 1182 return (vp->v_usecount); 1183 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1184 vnext = vq->v_specnext; 1185 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1186 continue; 1187 /* 1188 * Alias, but not in use, so flush it out. 1189 */ 1190 if (vq->v_usecount == 0 && vq != vp) { 1191 vgone(vq); 1192 goto loop; 1193 } 1194 count += vq->v_usecount; 1195 } 1196 return (count); 1197 } 1198 1199 #if defined(DEBUG) || defined(DIAGNOSTIC) 1200 /* 1201 * Print out a description of a vnode. 1202 */ 1203 static char *typename[] = 1204 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1205 1206 void 1207 vprint(char *label, struct vnode *vp) 1208 { 1209 char buf[64]; 1210 1211 if (label != NULL) 1212 printf("%s: ", label); 1213 printf("%p, type %s, use %u, write %u, hold %u,", 1214 vp, typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1215 vp->v_holdcnt); 1216 buf[0] = '\0'; 1217 if (vp->v_flag & VROOT) 1218 strlcat(buf, "|VROOT", sizeof buf); 1219 if (vp->v_flag & VTEXT) 1220 strlcat(buf, "|VTEXT", sizeof buf); 1221 if (vp->v_flag & VSYSTEM) 1222 strlcat(buf, "|VSYSTEM", sizeof buf); 1223 if (vp->v_flag & VXLOCK) 1224 strlcat(buf, "|VXLOCK", sizeof buf); 1225 if (vp->v_flag & VXWANT) 1226 strlcat(buf, "|VXWANT", sizeof buf); 1227 if (vp->v_bioflag & VBIOWAIT) 1228 strlcat(buf, "|VBIOWAIT", sizeof buf); 1229 if (vp->v_bioflag & VBIOONFREELIST) 1230 strlcat(buf, "|VBIOONFREELIST", sizeof buf); 1231 if (vp->v_bioflag & VBIOONSYNCLIST) 1232 strlcat(buf, "|VBIOONSYNCLIST", sizeof buf); 1233 if (vp->v_flag & VALIASED) 1234 strlcat(buf, "|VALIASED", sizeof buf); 1235 if (buf[0] != '\0') 1236 printf(" flags (%s)", &buf[1]); 1237 if (vp->v_data == NULL) { 1238 printf("\n"); 1239 } else { 1240 printf("\n\t"); 1241 VOP_PRINT(vp); 1242 } 1243 } 1244 #endif /* DEBUG || DIAGNOSTIC */ 1245 1246 #ifdef DEBUG 1247 /* 1248 * List all of the locked vnodes in the system. 1249 * Called when debugging the kernel. 1250 */ 1251 void 1252 printlockedvnodes(void) 1253 { 1254 struct mount *mp; 1255 struct vnode *vp; 1256 1257 printf("Locked vnodes\n"); 1258 1259 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 1260 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) 1261 continue; 1262 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 1263 if (VOP_ISLOCKED(vp)) 1264 vprint(NULL, vp); 1265 } 1266 vfs_unbusy(mp); 1267 } 1268 1269 } 1270 #endif 1271 1272 /* 1273 * Top level filesystem related information gathering. 1274 */ 1275 int 1276 vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1277 size_t newlen, struct proc *p) 1278 { 1279 struct vfsconf *vfsp, *tmpvfsp; 1280 int ret; 1281 1282 /* all sysctl names at this level are at least name and field */ 1283 if (namelen < 2) 1284 return (ENOTDIR); /* overloaded */ 1285 1286 if (name[0] != VFS_GENERIC) { 1287 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1288 if (vfsp->vfc_typenum == name[0]) 1289 break; 1290 1291 if (vfsp == NULL || vfsp->vfc_vfsops->vfs_sysctl == NULL) 1292 return (EOPNOTSUPP); 1293 1294 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1295 oldp, oldlenp, newp, newlen, p)); 1296 } 1297 1298 switch (name[1]) { 1299 case VFS_MAXTYPENUM: 1300 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1301 1302 case VFS_CONF: 1303 if (namelen < 3) 1304 return (ENOTDIR); /* overloaded */ 1305 1306 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1307 if (vfsp->vfc_typenum == name[2]) 1308 break; 1309 1310 if (vfsp == NULL) 1311 return (EOPNOTSUPP); 1312 1313 /* Make a copy, clear out kernel pointers */ 1314 tmpvfsp = malloc(sizeof(*tmpvfsp), M_TEMP, M_WAITOK); 1315 memcpy(tmpvfsp, vfsp, sizeof(*tmpvfsp)); 1316 tmpvfsp->vfc_vfsops = NULL; 1317 tmpvfsp->vfc_next = NULL; 1318 1319 ret = sysctl_rdstruct(oldp, oldlenp, newp, tmpvfsp, 1320 sizeof(struct vfsconf)); 1321 1322 free(tmpvfsp, M_TEMP, sizeof(*tmpvfsp)); 1323 return (ret); 1324 case VFS_BCACHESTAT: /* buffer cache statistics */ 1325 ret = sysctl_rdstruct(oldp, oldlenp, newp, &bcstats, 1326 sizeof(struct bcachestats)); 1327 return(ret); 1328 } 1329 return (EOPNOTSUPP); 1330 } 1331 1332 /* 1333 * Check to see if a filesystem is mounted on a block device. 1334 */ 1335 int 1336 vfs_mountedon(struct vnode *vp) 1337 { 1338 struct vnode *vq; 1339 int error = 0; 1340 1341 if (vp->v_specmountpoint != NULL) 1342 return (EBUSY); 1343 if (vp->v_flag & VALIASED) { 1344 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1345 if (vq->v_rdev != vp->v_rdev || 1346 vq->v_type != vp->v_type) 1347 continue; 1348 if (vq->v_specmountpoint != NULL) { 1349 error = EBUSY; 1350 break; 1351 } 1352 } 1353 } 1354 return (error); 1355 } 1356 1357 /* 1358 * Build hash lists of net addresses and hang them off the mount point. 1359 * Called by ufs_mount() to set up the lists of export addresses. 1360 */ 1361 int 1362 vfs_hang_addrlist(struct mount *mp, struct netexport *nep, 1363 struct export_args *argp) 1364 { 1365 struct netcred *np; 1366 struct radix_node_head *rnh; 1367 int nplen, i; 1368 struct radix_node *rn; 1369 struct sockaddr *saddr, *smask = 0; 1370 int error; 1371 1372 if (argp->ex_addrlen == 0) { 1373 if (mp->mnt_flag & MNT_DEFEXPORTED) 1374 return (EPERM); 1375 np = &nep->ne_defexported; 1376 /* fill in the kernel's ucred from userspace's xucred */ 1377 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1378 return (error); 1379 mp->mnt_flag |= MNT_DEFEXPORTED; 1380 goto finish; 1381 } 1382 if (argp->ex_addrlen > MLEN || argp->ex_masklen > MLEN || 1383 argp->ex_addrlen < 0 || argp->ex_masklen < 0) 1384 return (EINVAL); 1385 nplen = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1386 np = (struct netcred *)malloc(nplen, M_NETADDR, M_WAITOK|M_ZERO); 1387 saddr = (struct sockaddr *)(np + 1); 1388 error = copyin(argp->ex_addr, saddr, argp->ex_addrlen); 1389 if (error) 1390 goto out; 1391 if (saddr->sa_len > argp->ex_addrlen) 1392 saddr->sa_len = argp->ex_addrlen; 1393 if (argp->ex_masklen) { 1394 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1395 error = copyin(argp->ex_mask, smask, argp->ex_masklen); 1396 if (error) 1397 goto out; 1398 if (smask->sa_len > argp->ex_masklen) 1399 smask->sa_len = argp->ex_masklen; 1400 } 1401 /* fill in the kernel's ucred from userspace's xucred */ 1402 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1403 goto out; 1404 i = saddr->sa_family; 1405 switch (i) { 1406 case AF_INET: 1407 if ((rnh = nep->ne_rtable_inet) == NULL) { 1408 if (!rn_inithead((void **)&nep->ne_rtable_inet, 1409 offsetof(struct sockaddr_in, sin_addr))) { 1410 error = ENOBUFS; 1411 goto out; 1412 } 1413 rnh = nep->ne_rtable_inet; 1414 } 1415 break; 1416 default: 1417 error = EINVAL; 1418 goto out; 1419 } 1420 rn = rn_addroute(saddr, smask, rnh, np->netc_rnodes, 0); 1421 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1422 error = EPERM; 1423 goto out; 1424 } 1425 finish: 1426 np->netc_exflags = argp->ex_flags; 1427 return (0); 1428 out: 1429 free(np, M_NETADDR, nplen); 1430 return (error); 1431 } 1432 1433 int 1434 vfs_free_netcred(struct radix_node *rn, void *w, u_int id) 1435 { 1436 struct radix_node_head *rnh = (struct radix_node_head *)w; 1437 1438 rn_delete(rn->rn_key, rn->rn_mask, rnh, NULL); 1439 free(rn, M_NETADDR, 0); 1440 return (0); 1441 } 1442 1443 /* 1444 * Free the net address hash lists that are hanging off the mount points. 1445 */ 1446 void 1447 vfs_free_addrlist(struct netexport *nep) 1448 { 1449 struct radix_node_head *rnh; 1450 1451 if ((rnh = nep->ne_rtable_inet) != NULL) { 1452 rn_walktree(rnh, vfs_free_netcred, rnh); 1453 free(rnh, M_RTABLE, 0); 1454 nep->ne_rtable_inet = NULL; 1455 } 1456 } 1457 1458 int 1459 vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp) 1460 { 1461 int error; 1462 1463 if (argp->ex_flags & MNT_DELEXPORT) { 1464 vfs_free_addrlist(nep); 1465 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1466 } 1467 if (argp->ex_flags & MNT_EXPORTED) { 1468 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1469 return (error); 1470 mp->mnt_flag |= MNT_EXPORTED; 1471 } 1472 return (0); 1473 } 1474 1475 struct netcred * 1476 vfs_export_lookup(struct mount *mp, struct netexport *nep, struct mbuf *nam) 1477 { 1478 struct netcred *np; 1479 struct radix_node_head *rnh; 1480 struct sockaddr *saddr; 1481 1482 np = NULL; 1483 if (mp->mnt_flag & MNT_EXPORTED) { 1484 /* 1485 * Lookup in the export list first. 1486 */ 1487 if (nam != NULL) { 1488 saddr = mtod(nam, struct sockaddr *); 1489 switch(saddr->sa_family) { 1490 case AF_INET: 1491 rnh = nep->ne_rtable_inet; 1492 break; 1493 default: 1494 rnh = NULL; 1495 break; 1496 } 1497 if (rnh != NULL) 1498 np = (struct netcred *)rn_match(saddr, rnh); 1499 } 1500 /* 1501 * If no address match, use the default if it exists. 1502 */ 1503 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1504 np = &nep->ne_defexported; 1505 } 1506 return (np); 1507 } 1508 1509 /* 1510 * Do the usual access checking. 1511 * file_mode, uid and gid are from the vnode in question, 1512 * while acc_mode and cred are from the VOP_ACCESS parameter list 1513 */ 1514 int 1515 vaccess(enum vtype type, mode_t file_mode, uid_t uid, gid_t gid, 1516 mode_t acc_mode, struct ucred *cred) 1517 { 1518 mode_t mask; 1519 1520 /* User id 0 always gets read/write access. */ 1521 if (cred->cr_uid == 0) { 1522 /* For VEXEC, at least one of the execute bits must be set. */ 1523 if ((acc_mode & VEXEC) && type != VDIR && 1524 (file_mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0) 1525 return EACCES; 1526 return 0; 1527 } 1528 1529 mask = 0; 1530 1531 /* Otherwise, check the owner. */ 1532 if (cred->cr_uid == uid) { 1533 if (acc_mode & VEXEC) 1534 mask |= S_IXUSR; 1535 if (acc_mode & VREAD) 1536 mask |= S_IRUSR; 1537 if (acc_mode & VWRITE) 1538 mask |= S_IWUSR; 1539 return (file_mode & mask) == mask ? 0 : EACCES; 1540 } 1541 1542 /* Otherwise, check the groups. */ 1543 if (groupmember(gid, cred)) { 1544 if (acc_mode & VEXEC) 1545 mask |= S_IXGRP; 1546 if (acc_mode & VREAD) 1547 mask |= S_IRGRP; 1548 if (acc_mode & VWRITE) 1549 mask |= S_IWGRP; 1550 return (file_mode & mask) == mask ? 0 : EACCES; 1551 } 1552 1553 /* Otherwise, check everyone else. */ 1554 if (acc_mode & VEXEC) 1555 mask |= S_IXOTH; 1556 if (acc_mode & VREAD) 1557 mask |= S_IROTH; 1558 if (acc_mode & VWRITE) 1559 mask |= S_IWOTH; 1560 return (file_mode & mask) == mask ? 0 : EACCES; 1561 } 1562 1563 /* 1564 * Unmount all file systems. 1565 * We traverse the list in reverse order under the assumption that doing so 1566 * will avoid needing to worry about dependencies. 1567 */ 1568 void 1569 vfs_unmountall(void) 1570 { 1571 struct mount *mp, *nmp; 1572 int allerror, error, again = 1; 1573 1574 retry: 1575 allerror = 0; 1576 TAILQ_FOREACH_REVERSE_SAFE(mp, &mountlist, mntlist, mnt_list, nmp) { 1577 if (vfs_busy(mp, VB_WRITE|VB_NOWAIT)) 1578 continue; 1579 /* XXX Here is a race, the next pointer is not locked. */ 1580 if ((error = dounmount(mp, MNT_FORCE, curproc)) != 0) { 1581 printf("unmount of %s failed with error %d\n", 1582 mp->mnt_stat.f_mntonname, error); 1583 allerror = 1; 1584 } 1585 } 1586 1587 if (allerror) { 1588 printf("WARNING: some file systems would not unmount\n"); 1589 if (again) { 1590 printf("retrying\n"); 1591 again = 0; 1592 goto retry; 1593 } 1594 } 1595 } 1596 1597 /* 1598 * Sync and unmount file systems before shutting down. 1599 */ 1600 void 1601 vfs_shutdown(void) 1602 { 1603 #ifdef ACCOUNTING 1604 acct_shutdown(); 1605 #endif 1606 1607 /* XXX Should suspend scheduling. */ 1608 (void) spl0(); 1609 1610 printf("syncing disks... "); 1611 1612 if (panicstr == 0) { 1613 /* Sync before unmount, in case we hang on something. */ 1614 sys_sync(&proc0, NULL, NULL); 1615 1616 /* Unmount file systems. */ 1617 vfs_unmountall(); 1618 } 1619 1620 if (vfs_syncwait(1)) 1621 printf("giving up\n"); 1622 else 1623 printf("done\n"); 1624 1625 #if NSOFTRAID > 0 1626 sr_shutdown(); 1627 #endif 1628 } 1629 1630 /* 1631 * perform sync() operation and wait for buffers to flush. 1632 * assumptions: called w/ scheduler disabled and physical io enabled 1633 * for now called at spl0() XXX 1634 */ 1635 int 1636 vfs_syncwait(int verbose) 1637 { 1638 struct buf *bp; 1639 int iter, nbusy, dcount, s; 1640 struct proc *p; 1641 #ifdef MULTIPROCESSOR 1642 int hold_count; 1643 #endif 1644 1645 p = curproc? curproc : &proc0; 1646 sys_sync(p, NULL, NULL); 1647 1648 /* Wait for sync to finish. */ 1649 dcount = 10000; 1650 for (iter = 0; iter < 20; iter++) { 1651 nbusy = 0; 1652 LIST_FOREACH(bp, &bufhead, b_list) { 1653 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1654 nbusy++; 1655 /* 1656 * With soft updates, some buffers that are 1657 * written will be remarked as dirty until other 1658 * buffers are written. 1659 */ 1660 if (bp->b_flags & B_DELWRI) { 1661 s = splbio(); 1662 bremfree(bp); 1663 buf_acquire(bp); 1664 splx(s); 1665 nbusy++; 1666 bawrite(bp); 1667 if (dcount-- <= 0) { 1668 if (verbose) 1669 printf("softdep "); 1670 return 1; 1671 } 1672 } 1673 } 1674 if (nbusy == 0) 1675 break; 1676 if (verbose) 1677 printf("%d ", nbusy); 1678 #ifdef MULTIPROCESSOR 1679 if (__mp_lock_held(&kernel_lock)) 1680 hold_count = __mp_release_all(&kernel_lock); 1681 else 1682 hold_count = 0; 1683 #endif 1684 DELAY(40000 * iter); 1685 #ifdef MULTIPROCESSOR 1686 if (hold_count) 1687 __mp_acquire_count(&kernel_lock, hold_count); 1688 #endif 1689 } 1690 1691 return nbusy; 1692 } 1693 1694 /* 1695 * posix file system related system variables. 1696 */ 1697 int 1698 fs_posix_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, 1699 void *newp, size_t newlen, struct proc *p) 1700 { 1701 /* all sysctl names at this level are terminal */ 1702 if (namelen != 1) 1703 return (ENOTDIR); 1704 1705 switch (name[0]) { 1706 case FS_POSIX_SETUID: 1707 if (newp && securelevel > 0) 1708 return (EPERM); 1709 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1710 default: 1711 return (EOPNOTSUPP); 1712 } 1713 /* NOTREACHED */ 1714 } 1715 1716 /* 1717 * file system related system variables. 1718 */ 1719 int 1720 fs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1721 size_t newlen, struct proc *p) 1722 { 1723 sysctlfn *fn; 1724 1725 switch (name[0]) { 1726 case FS_POSIX: 1727 fn = fs_posix_sysctl; 1728 break; 1729 default: 1730 return (EOPNOTSUPP); 1731 } 1732 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1733 } 1734 1735 1736 /* 1737 * Routines dealing with vnodes and buffers 1738 */ 1739 1740 /* 1741 * Wait for all outstanding I/Os to complete 1742 * 1743 * Manipulates v_numoutput. Must be called at splbio() 1744 */ 1745 int 1746 vwaitforio(struct vnode *vp, int slpflag, char *wmesg, int timeo) 1747 { 1748 int error = 0; 1749 1750 splassert(IPL_BIO); 1751 1752 while (vp->v_numoutput) { 1753 vp->v_bioflag |= VBIOWAIT; 1754 error = tsleep(&vp->v_numoutput, 1755 slpflag | (PRIBIO + 1), wmesg, timeo); 1756 if (error) 1757 break; 1758 } 1759 1760 return (error); 1761 } 1762 1763 /* 1764 * Update outstanding I/O count and do wakeup if requested. 1765 * 1766 * Manipulates v_numoutput. Must be called at splbio() 1767 */ 1768 void 1769 vwakeup(struct vnode *vp) 1770 { 1771 splassert(IPL_BIO); 1772 1773 if (vp != NULL) { 1774 if (vp->v_numoutput-- == 0) 1775 panic("vwakeup: neg numoutput"); 1776 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1777 vp->v_bioflag &= ~VBIOWAIT; 1778 wakeup(&vp->v_numoutput); 1779 } 1780 } 1781 } 1782 1783 /* 1784 * Flush out and invalidate all buffers associated with a vnode. 1785 * Called with the underlying object locked. 1786 */ 1787 int 1788 vinvalbuf(struct vnode *vp, int flags, struct ucred *cred, struct proc *p, 1789 int slpflag, int slptimeo) 1790 { 1791 struct buf *bp; 1792 struct buf *nbp, *blist; 1793 int s, error; 1794 1795 #ifdef VFSLCKDEBUG 1796 if ((vp->v_flag & VLOCKSWORK) && !VOP_ISLOCKED(vp)) 1797 panic("vinvalbuf(): vp isn't locked"); 1798 #endif 1799 1800 if (flags & V_SAVE) { 1801 s = splbio(); 1802 vwaitforio(vp, 0, "vinvalbuf", 0); 1803 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1804 splx(s); 1805 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1806 return (error); 1807 s = splbio(); 1808 if (vp->v_numoutput > 0 || 1809 !LIST_EMPTY(&vp->v_dirtyblkhd)) 1810 panic("vinvalbuf: dirty bufs"); 1811 } 1812 splx(s); 1813 } 1814 loop: 1815 s = splbio(); 1816 for (;;) { 1817 if ((blist = LIST_FIRST(&vp->v_cleanblkhd)) && 1818 (flags & V_SAVEMETA)) 1819 while (blist && blist->b_lblkno < 0) 1820 blist = LIST_NEXT(blist, b_vnbufs); 1821 if (blist == NULL && 1822 (blist = LIST_FIRST(&vp->v_dirtyblkhd)) && 1823 (flags & V_SAVEMETA)) 1824 while (blist && blist->b_lblkno < 0) 1825 blist = LIST_NEXT(blist, b_vnbufs); 1826 if (!blist) 1827 break; 1828 1829 for (bp = blist; bp; bp = nbp) { 1830 nbp = LIST_NEXT(bp, b_vnbufs); 1831 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1832 continue; 1833 if (bp->b_flags & B_BUSY) { 1834 bp->b_flags |= B_WANTED; 1835 error = tsleep(bp, slpflag | (PRIBIO + 1), 1836 "vinvalbuf", slptimeo); 1837 if (error) { 1838 splx(s); 1839 return (error); 1840 } 1841 break; 1842 } 1843 bremfree(bp); 1844 /* 1845 * XXX Since there are no node locks for NFS, I believe 1846 * there is a slight chance that a delayed write will 1847 * occur while sleeping just above, so check for it. 1848 */ 1849 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1850 buf_acquire(bp); 1851 splx(s); 1852 (void) VOP_BWRITE(bp); 1853 goto loop; 1854 } 1855 buf_acquire_nomap(bp); 1856 bp->b_flags |= B_INVAL; 1857 brelse(bp); 1858 } 1859 } 1860 if (!(flags & V_SAVEMETA) && 1861 (!LIST_EMPTY(&vp->v_dirtyblkhd) || !LIST_EMPTY(&vp->v_cleanblkhd))) 1862 panic("vinvalbuf: flush failed"); 1863 splx(s); 1864 return (0); 1865 } 1866 1867 void 1868 vflushbuf(struct vnode *vp, int sync) 1869 { 1870 struct buf *bp, *nbp; 1871 int s; 1872 1873 loop: 1874 s = splbio(); 1875 LIST_FOREACH_SAFE(bp, &vp->v_dirtyblkhd, b_vnbufs, nbp) { 1876 if ((bp->b_flags & B_BUSY)) 1877 continue; 1878 if ((bp->b_flags & B_DELWRI) == 0) 1879 panic("vflushbuf: not dirty"); 1880 bremfree(bp); 1881 buf_acquire(bp); 1882 splx(s); 1883 /* 1884 * Wait for I/O associated with indirect blocks to complete, 1885 * since there is no way to quickly wait for them below. 1886 */ 1887 if (bp->b_vp == vp || sync == 0) 1888 (void) bawrite(bp); 1889 else 1890 (void) bwrite(bp); 1891 goto loop; 1892 } 1893 if (sync == 0) { 1894 splx(s); 1895 return; 1896 } 1897 vwaitforio(vp, 0, "vflushbuf", 0); 1898 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1899 splx(s); 1900 #ifdef DIAGNOSTIC 1901 vprint("vflushbuf: dirty", vp); 1902 #endif 1903 goto loop; 1904 } 1905 splx(s); 1906 } 1907 1908 /* 1909 * Associate a buffer with a vnode. 1910 * 1911 * Manipulates buffer vnode queues. Must be called at splbio(). 1912 */ 1913 void 1914 bgetvp(struct vnode *vp, struct buf *bp) 1915 { 1916 splassert(IPL_BIO); 1917 1918 1919 if (bp->b_vp) 1920 panic("bgetvp: not free"); 1921 vhold(vp); 1922 bp->b_vp = vp; 1923 if (vp->v_type == VBLK || vp->v_type == VCHR) 1924 bp->b_dev = vp->v_rdev; 1925 else 1926 bp->b_dev = NODEV; 1927 /* 1928 * Insert onto list for new vnode. 1929 */ 1930 bufinsvn(bp, &vp->v_cleanblkhd); 1931 } 1932 1933 /* 1934 * Disassociate a buffer from a vnode. 1935 * 1936 * Manipulates vnode buffer queues. Must be called at splbio(). 1937 */ 1938 void 1939 brelvp(struct buf *bp) 1940 { 1941 struct vnode *vp; 1942 1943 splassert(IPL_BIO); 1944 1945 if ((vp = bp->b_vp) == (struct vnode *) 0) 1946 panic("brelvp: NULL"); 1947 /* 1948 * Delete from old vnode list, if on one. 1949 */ 1950 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 1951 bufremvn(bp); 1952 if ((vp->v_bioflag & VBIOONSYNCLIST) && 1953 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 1954 vp->v_bioflag &= ~VBIOONSYNCLIST; 1955 LIST_REMOVE(vp, v_synclist); 1956 } 1957 bp->b_vp = NULL; 1958 1959 vdrop(vp); 1960 } 1961 1962 /* 1963 * Replaces the current vnode associated with the buffer, if any, 1964 * with a new vnode. 1965 * 1966 * If an output I/O is pending on the buffer, the old vnode 1967 * I/O count is adjusted. 1968 * 1969 * Ignores vnode buffer queues. Must be called at splbio(). 1970 */ 1971 void 1972 buf_replacevnode(struct buf *bp, struct vnode *newvp) 1973 { 1974 struct vnode *oldvp = bp->b_vp; 1975 1976 splassert(IPL_BIO); 1977 1978 if (oldvp) 1979 brelvp(bp); 1980 1981 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 1982 newvp->v_numoutput++; /* put it on swapdev */ 1983 vwakeup(oldvp); 1984 } 1985 1986 bgetvp(newvp, bp); 1987 bufremvn(bp); 1988 } 1989 1990 /* 1991 * Used to assign buffers to the appropriate clean or dirty list on 1992 * the vnode and to add newly dirty vnodes to the appropriate 1993 * filesystem syncer list. 1994 * 1995 * Manipulates vnode buffer queues. Must be called at splbio(). 1996 */ 1997 void 1998 reassignbuf(struct buf *bp) 1999 { 2000 struct buflists *listheadp; 2001 int delay; 2002 struct vnode *vp = bp->b_vp; 2003 2004 splassert(IPL_BIO); 2005 2006 /* 2007 * Delete from old vnode list, if on one. 2008 */ 2009 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2010 bufremvn(bp); 2011 2012 /* 2013 * If dirty, put on list of dirty buffers; 2014 * otherwise insert onto list of clean buffers. 2015 */ 2016 if ((bp->b_flags & B_DELWRI) == 0) { 2017 listheadp = &vp->v_cleanblkhd; 2018 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2019 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2020 vp->v_bioflag &= ~VBIOONSYNCLIST; 2021 LIST_REMOVE(vp, v_synclist); 2022 } 2023 } else { 2024 listheadp = &vp->v_dirtyblkhd; 2025 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2026 switch (vp->v_type) { 2027 case VDIR: 2028 delay = syncdelay / 2; 2029 break; 2030 case VBLK: 2031 if (vp->v_specmountpoint != NULL) { 2032 delay = syncdelay / 3; 2033 break; 2034 } 2035 /* FALLTHROUGH */ 2036 default: 2037 delay = syncdelay; 2038 } 2039 vn_syncer_add_to_worklist(vp, delay); 2040 } 2041 } 2042 bufinsvn(bp, listheadp); 2043 } 2044 2045 int 2046 vfs_register(struct vfsconf *vfs) 2047 { 2048 struct vfsconf *vfsp; 2049 struct vfsconf **vfspp; 2050 2051 #ifdef DIAGNOSTIC 2052 /* Paranoia? */ 2053 if (vfs->vfc_refcount != 0) 2054 printf("vfs_register called with vfc_refcount > 0\n"); 2055 #endif 2056 2057 /* Check if filesystem already known */ 2058 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2059 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2060 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2061 return (EEXIST); 2062 2063 if (vfs->vfc_typenum > maxvfsconf) 2064 maxvfsconf = vfs->vfc_typenum; 2065 2066 vfs->vfc_next = NULL; 2067 2068 /* Add to the end of the list */ 2069 *vfspp = vfs; 2070 2071 /* Call vfs_init() */ 2072 if (vfs->vfc_vfsops->vfs_init) 2073 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2074 2075 return 0; 2076 } 2077 2078 int 2079 vfs_unregister(struct vfsconf *vfs) 2080 { 2081 struct vfsconf *vfsp; 2082 struct vfsconf **vfspp; 2083 int maxtypenum; 2084 2085 /* Find our vfsconf struct */ 2086 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2087 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2088 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2089 break; 2090 } 2091 2092 if (!vfsp) /* Not found */ 2093 return (ENOENT); 2094 2095 if (vfsp->vfc_refcount) /* In use */ 2096 return (EBUSY); 2097 2098 /* Remove from list and free */ 2099 *vfspp = vfsp->vfc_next; 2100 2101 maxtypenum = 0; 2102 2103 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2104 if (vfsp->vfc_typenum > maxtypenum) 2105 maxtypenum = vfsp->vfc_typenum; 2106 2107 maxvfsconf = maxtypenum; 2108 return 0; 2109 } 2110 2111 /* 2112 * Check if vnode represents a disk device 2113 */ 2114 int 2115 vn_isdisk(struct vnode *vp, int *errp) 2116 { 2117 if (vp->v_type != VBLK && vp->v_type != VCHR) 2118 return (0); 2119 2120 return (1); 2121 } 2122 2123 #ifdef DDB 2124 #include <machine/db_machdep.h> 2125 #include <ddb/db_interface.h> 2126 2127 void 2128 vfs_buf_print(void *b, int full, 2129 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2130 { 2131 struct buf *bp = b; 2132 2133 (*pr)(" vp %p lblkno 0x%llx blkno 0x%llx dev 0x%x\n" 2134 " proc %p error %d flags %lb\n", 2135 bp->b_vp, (int64_t)bp->b_lblkno, (int64_t)bp->b_blkno, bp->b_dev, 2136 bp->b_proc, bp->b_error, bp->b_flags, B_BITS); 2137 2138 (*pr)(" bufsize 0x%lx bcount 0x%lx resid 0x%lx\n" 2139 " data %p saveaddr %p dep %p iodone %p\n", 2140 bp->b_bufsize, bp->b_bcount, (long)bp->b_resid, 2141 bp->b_data, bp->b_saveaddr, 2142 LIST_FIRST(&bp->b_dep), bp->b_iodone); 2143 2144 (*pr)(" dirty {off 0x%x end 0x%x} valid {off 0x%x end 0x%x}\n", 2145 bp->b_dirtyoff, bp->b_dirtyend, bp->b_validoff, bp->b_validend); 2146 2147 #ifdef FFS_SOFTUPDATES 2148 if (full) 2149 softdep_print(bp, full, pr); 2150 #endif 2151 } 2152 2153 const char *vtypes[] = { VTYPE_NAMES }; 2154 const char *vtags[] = { VTAG_NAMES }; 2155 2156 void 2157 vfs_vnode_print(void *v, int full, 2158 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2159 { 2160 struct vnode *vp = v; 2161 2162 (*pr)("tag %s(%d) type %s(%d) mount %p typedata %p\n", 2163 (u_int)vp->v_tag >= nitems(vtags)? "<unk>":vtags[vp->v_tag], 2164 vp->v_tag, 2165 (u_int)vp->v_type >= nitems(vtypes)? "<unk>":vtypes[vp->v_type], 2166 vp->v_type, vp->v_mount, vp->v_mountedhere); 2167 2168 (*pr)("data %p usecount %d writecount %d holdcnt %d numoutput %d\n", 2169 vp->v_data, vp->v_usecount, vp->v_writecount, 2170 vp->v_holdcnt, vp->v_numoutput); 2171 2172 /* uvm_object_printit(&vp->v_uobj, full, pr); */ 2173 2174 if (full) { 2175 struct buf *bp; 2176 2177 (*pr)("clean bufs:\n"); 2178 LIST_FOREACH(bp, &vp->v_cleanblkhd, b_vnbufs) { 2179 (*pr)(" bp %p\n", bp); 2180 vfs_buf_print(bp, full, pr); 2181 } 2182 2183 (*pr)("dirty bufs:\n"); 2184 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) { 2185 (*pr)(" bp %p\n", bp); 2186 vfs_buf_print(bp, full, pr); 2187 } 2188 } 2189 } 2190 2191 void 2192 vfs_mount_print(struct mount *mp, int full, 2193 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2194 { 2195 struct vfsconf *vfc = mp->mnt_vfc; 2196 struct vnode *vp; 2197 int cnt = 0; 2198 2199 (*pr)("flags %b\nvnodecovered %p syncer %p data %p\n", 2200 mp->mnt_flag, MNT_BITS, 2201 mp->mnt_vnodecovered, mp->mnt_syncer, mp->mnt_data); 2202 2203 (*pr)("vfsconf: ops %p name \"%s\" num %d ref %d flags 0x%x\n", 2204 vfc->vfc_vfsops, vfc->vfc_name, vfc->vfc_typenum, 2205 vfc->vfc_refcount, vfc->vfc_flags); 2206 2207 (*pr)("statvfs cache: bsize %x iosize %x\nblocks %llu free %llu avail %lld\n", 2208 mp->mnt_stat.f_bsize, mp->mnt_stat.f_iosize, mp->mnt_stat.f_blocks, 2209 mp->mnt_stat.f_bfree, mp->mnt_stat.f_bavail); 2210 2211 (*pr)(" files %llu ffiles %llu favail %lld\n", mp->mnt_stat.f_files, 2212 mp->mnt_stat.f_ffree, mp->mnt_stat.f_favail); 2213 2214 (*pr)(" f_fsidx {0x%x, 0x%x} owner %u ctime 0x%llx\n", 2215 mp->mnt_stat.f_fsid.val[0], mp->mnt_stat.f_fsid.val[1], 2216 mp->mnt_stat.f_owner, mp->mnt_stat.f_ctime); 2217 2218 (*pr)(" syncwrites %llu asyncwrites = %llu\n", 2219 mp->mnt_stat.f_syncwrites, mp->mnt_stat.f_asyncwrites); 2220 2221 (*pr)(" syncreads %llu asyncreads = %llu\n", 2222 mp->mnt_stat.f_syncreads, mp->mnt_stat.f_asyncreads); 2223 2224 (*pr)(" fstype \"%s\" mnton \"%s\" mntfrom \"%s\" mntspec \"%s\"\n", 2225 mp->mnt_stat.f_fstypename, mp->mnt_stat.f_mntonname, 2226 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntfromspec); 2227 2228 (*pr)("locked vnodes:"); 2229 /* XXX would take mountlist lock, except ddb has no context */ 2230 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2231 if (VOP_ISLOCKED(vp)) { 2232 if (!LIST_NEXT(vp, v_mntvnodes)) 2233 (*pr)(" %p", vp); 2234 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2235 (*pr)("\n\t%p", vp); 2236 else 2237 (*pr)(", %p", vp); 2238 } 2239 (*pr)("\n"); 2240 2241 if (full) { 2242 (*pr)("all vnodes:\n\t"); 2243 /* XXX would take mountlist lock, except ddb has no context */ 2244 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2245 if (!LIST_NEXT(vp, v_mntvnodes)) 2246 (*pr)(" %p", vp); 2247 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2248 (*pr)(" %p,\n\t", vp); 2249 else 2250 (*pr)(" %p,", vp); 2251 (*pr)("\n"); 2252 } 2253 } 2254 #endif /* DDB */ 2255 2256 void 2257 copy_statfs_info(struct statfs *sbp, const struct mount *mp) 2258 { 2259 const struct statfs *mbp; 2260 2261 strncpy(sbp->f_fstypename, mp->mnt_vfc->vfc_name, MFSNAMELEN); 2262 2263 if (sbp == (mbp = &mp->mnt_stat)) 2264 return; 2265 2266 sbp->f_fsid = mbp->f_fsid; 2267 sbp->f_owner = mbp->f_owner; 2268 sbp->f_flags = mbp->f_flags; 2269 sbp->f_syncwrites = mbp->f_syncwrites; 2270 sbp->f_asyncwrites = mbp->f_asyncwrites; 2271 sbp->f_syncreads = mbp->f_syncreads; 2272 sbp->f_asyncreads = mbp->f_asyncreads; 2273 sbp->f_namemax = mbp->f_namemax; 2274 memcpy(sbp->f_mntonname, mp->mnt_stat.f_mntonname, MNAMELEN); 2275 memcpy(sbp->f_mntfromname, mp->mnt_stat.f_mntfromname, MNAMELEN); 2276 memcpy(sbp->f_mntfromspec, mp->mnt_stat.f_mntfromspec, MNAMELEN); 2277 memcpy(&sbp->mount_info, &mp->mnt_stat.mount_info, 2278 sizeof(union mount_info)); 2279 } 2280