1 /* $OpenBSD: vfs_subr.c,v 1.253 2016/09/16 03:21:16 dlg 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 for (vp = TAILQ_FIRST(listhd); vp != NULLVP; 384 vp = TAILQ_NEXT(vp, v_freelist)) { 385 if (VOP_ISLOCKED(vp) == 0) 386 break; 387 } 388 /* 389 * Unless this is a bad time of the month, at most 390 * the first NCPUS items on the free list are 391 * locked, so this is close enough to being empty. 392 */ 393 if (vp == NULL) { 394 splx(s); 395 tablefull("vnode"); 396 *vpp = 0; 397 return (ENFILE); 398 } 399 400 #ifdef DIAGNOSTIC 401 if (vp->v_usecount) { 402 vprint("free vnode", vp); 403 panic("free vnode isn't"); 404 } 405 #endif 406 407 TAILQ_REMOVE(listhd, vp, v_freelist); 408 vp->v_bioflag &= ~VBIOONFREELIST; 409 splx(s); 410 411 if (vp->v_type != VBAD) 412 vgonel(vp, p); 413 #ifdef DIAGNOSTIC 414 if (vp->v_data) { 415 vprint("cleaned vnode", vp); 416 panic("cleaned vnode isn't"); 417 } 418 s = splbio(); 419 if (vp->v_numoutput) 420 panic("Clean vnode has pending I/O's"); 421 splx(s); 422 #endif 423 vp->v_flag = 0; 424 vp->v_socket = 0; 425 } 426 cache_purge(vp); 427 vp->v_type = VNON; 428 vp->v_tag = tag; 429 vp->v_op = vops; 430 insmntque(vp, mp); 431 *vpp = vp; 432 vp->v_usecount = 1; 433 vp->v_data = 0; 434 return (0); 435 } 436 437 /* 438 * Move a vnode from one mount queue to another. 439 */ 440 void 441 insmntque(struct vnode *vp, struct mount *mp) 442 { 443 /* 444 * Delete from old mount point vnode list, if on one. 445 */ 446 if (vp->v_mount != NULL) 447 LIST_REMOVE(vp, v_mntvnodes); 448 /* 449 * Insert into list of vnodes for the new mount point, if available. 450 */ 451 if ((vp->v_mount = mp) != NULL) 452 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 453 } 454 455 /* 456 * Create a vnode for a block device. 457 * Used for root filesystem, argdev, and swap areas. 458 * Also used for memory file system special devices. 459 */ 460 int 461 bdevvp(dev_t dev, struct vnode **vpp) 462 { 463 return (getdevvp(dev, vpp, VBLK)); 464 } 465 466 /* 467 * Create a vnode for a character device. 468 * Used for console handling. 469 */ 470 int 471 cdevvp(dev_t dev, struct vnode **vpp) 472 { 473 return (getdevvp(dev, vpp, VCHR)); 474 } 475 476 /* 477 * Create a vnode for a device. 478 * Used by bdevvp (block device) for root file system etc., 479 * and by cdevvp (character device) for console. 480 */ 481 int 482 getdevvp(dev_t dev, struct vnode **vpp, enum vtype type) 483 { 484 struct vnode *vp; 485 struct vnode *nvp; 486 int error; 487 488 if (dev == NODEV) { 489 *vpp = NULLVP; 490 return (0); 491 } 492 error = getnewvnode(VT_NON, NULL, &spec_vops, &nvp); 493 if (error) { 494 *vpp = NULLVP; 495 return (error); 496 } 497 vp = nvp; 498 vp->v_type = type; 499 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 500 vput(vp); 501 vp = nvp; 502 } 503 if (vp->v_type == VCHR && cdevsw[major(vp->v_rdev)].d_type == D_TTY) 504 vp->v_flag |= VISTTY; 505 *vpp = vp; 506 return (0); 507 } 508 509 /* 510 * Check to see if the new vnode represents a special device 511 * for which we already have a vnode (either because of 512 * bdevvp() or because of a different vnode representing 513 * the same block device). If such an alias exists, deallocate 514 * the existing contents and return the aliased vnode. The 515 * caller is responsible for filling it with its new contents. 516 */ 517 struct vnode * 518 checkalias(struct vnode *nvp, dev_t nvp_rdev, struct mount *mp) 519 { 520 struct proc *p = curproc; 521 struct vnode *vp; 522 struct vnode **vpp; 523 524 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 525 return (NULLVP); 526 527 vpp = &speclisth[SPECHASH(nvp_rdev)]; 528 loop: 529 for (vp = *vpp; vp; vp = vp->v_specnext) { 530 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 531 continue; 532 } 533 /* 534 * Alias, but not in use, so flush it out. 535 */ 536 if (vp->v_usecount == 0) { 537 vgonel(vp, p); 538 goto loop; 539 } 540 if (vget(vp, LK_EXCLUSIVE, p)) { 541 goto loop; 542 } 543 break; 544 } 545 546 /* 547 * Common case is actually in the if statement 548 */ 549 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 550 nvp->v_specinfo = malloc(sizeof(struct specinfo), M_VNODE, 551 M_WAITOK); 552 nvp->v_rdev = nvp_rdev; 553 nvp->v_hashchain = vpp; 554 nvp->v_specnext = *vpp; 555 nvp->v_specmountpoint = NULL; 556 nvp->v_speclockf = NULL; 557 nvp->v_specbitmap = NULL; 558 if (nvp->v_type == VCHR && 559 (cdevsw[major(nvp_rdev)].d_flags & D_CLONE) && 560 (minor(nvp_rdev) >> CLONE_SHIFT == 0)) { 561 if (vp != NULLVP) 562 nvp->v_specbitmap = vp->v_specbitmap; 563 else 564 nvp->v_specbitmap = malloc(CLONE_MAPSZ, 565 M_VNODE, M_WAITOK | M_ZERO); 566 } 567 *vpp = nvp; 568 if (vp != NULLVP) { 569 nvp->v_flag |= VALIASED; 570 vp->v_flag |= VALIASED; 571 vput(vp); 572 } 573 return (NULLVP); 574 } 575 576 /* 577 * This code is the uncommon case. It is called in case 578 * we found an alias that was VT_NON && vtype of VBLK 579 * This means we found a block device that was created 580 * using bdevvp. 581 * An example of such a vnode is the root partition device vnode 582 * created in ffs_mountroot. 583 * 584 * The vnodes created by bdevvp should not be aliased (why?). 585 */ 586 587 VOP_UNLOCK(vp, p); 588 vclean(vp, 0, p); 589 vp->v_op = nvp->v_op; 590 vp->v_tag = nvp->v_tag; 591 nvp->v_type = VNON; 592 insmntque(vp, mp); 593 return (vp); 594 } 595 596 /* 597 * Grab a particular vnode from the free list, increment its 598 * reference count and lock it. If the vnode lock bit is set, 599 * the vnode is being eliminated in vgone. In that case, we 600 * cannot grab it, so the process is awakened when the 601 * transition is completed, and an error code is returned to 602 * indicate that the vnode is no longer usable, possibly 603 * having been changed to a new file system type. 604 */ 605 int 606 vget(struct vnode *vp, int flags, struct proc *p) 607 { 608 int error, s, onfreelist; 609 610 /* 611 * If the vnode is in the process of being cleaned out for 612 * another use, we wait for the cleaning to finish and then 613 * return failure. Cleaning is determined by checking that 614 * the VXLOCK flag is set. 615 */ 616 617 if (vp->v_flag & VXLOCK) { 618 if (flags & LK_NOWAIT) { 619 return (EBUSY); 620 } 621 622 vp->v_flag |= VXWANT; 623 tsleep(vp, PINOD, "vget", 0); 624 return (ENOENT); 625 } 626 627 onfreelist = vp->v_bioflag & VBIOONFREELIST; 628 if (vp->v_usecount == 0 && onfreelist) { 629 s = splbio(); 630 if (vp->v_holdcnt > 0) 631 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 632 else 633 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 634 vp->v_bioflag &= ~VBIOONFREELIST; 635 splx(s); 636 } 637 638 vp->v_usecount++; 639 if (flags & LK_TYPE_MASK) { 640 if ((error = vn_lock(vp, flags, p)) != 0) { 641 vp->v_usecount--; 642 if (vp->v_usecount == 0 && onfreelist) 643 vputonfreelist(vp); 644 } 645 return (error); 646 } 647 648 return (0); 649 } 650 651 652 /* Vnode reference. */ 653 void 654 vref(struct vnode *vp) 655 { 656 #ifdef DIAGNOSTIC 657 if (vp->v_usecount == 0) 658 panic("vref used where vget required"); 659 if (vp->v_type == VNON) 660 panic("vref on a VNON vnode"); 661 #endif 662 vp->v_usecount++; 663 } 664 665 void 666 vputonfreelist(struct vnode *vp) 667 { 668 int s; 669 struct freelst *lst; 670 671 s = splbio(); 672 #ifdef DIAGNOSTIC 673 if (vp->v_usecount != 0) 674 panic("Use count is not zero!"); 675 676 if (vp->v_bioflag & VBIOONFREELIST) { 677 vprint("vnode already on free list: ", vp); 678 panic("vnode already on free list"); 679 } 680 #endif 681 682 vp->v_bioflag |= VBIOONFREELIST; 683 684 if (vp->v_holdcnt > 0) 685 lst = &vnode_hold_list; 686 else 687 lst = &vnode_free_list; 688 689 if (vp->v_type == VBAD) 690 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 691 else 692 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 693 694 splx(s); 695 } 696 697 /* 698 * vput(), just unlock and vrele() 699 */ 700 void 701 vput(struct vnode *vp) 702 { 703 struct proc *p = curproc; 704 705 #ifdef DIAGNOSTIC 706 if (vp == NULL) 707 panic("vput: null vp"); 708 #endif 709 710 #ifdef DIAGNOSTIC 711 if (vp->v_usecount == 0) { 712 vprint("vput: bad ref count", vp); 713 panic("vput: ref cnt"); 714 } 715 #endif 716 vp->v_usecount--; 717 if (vp->v_usecount > 0) { 718 VOP_UNLOCK(vp, p); 719 return; 720 } 721 722 #ifdef DIAGNOSTIC 723 if (vp->v_writecount != 0) { 724 vprint("vput: bad writecount", vp); 725 panic("vput: v_writecount != 0"); 726 } 727 #endif 728 729 VOP_INACTIVE(vp, p); 730 731 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 732 vputonfreelist(vp); 733 } 734 735 /* 736 * Vnode release - use for active VNODES. 737 * If count drops to zero, call inactive routine and return to freelist. 738 * Returns 0 if it did not sleep. 739 */ 740 int 741 vrele(struct vnode *vp) 742 { 743 struct proc *p = curproc; 744 745 #ifdef DIAGNOSTIC 746 if (vp == NULL) 747 panic("vrele: null vp"); 748 #endif 749 #ifdef DIAGNOSTIC 750 if (vp->v_usecount == 0) { 751 vprint("vrele: bad ref count", vp); 752 panic("vrele: ref cnt"); 753 } 754 #endif 755 vp->v_usecount--; 756 if (vp->v_usecount > 0) { 757 return (0); 758 } 759 760 #ifdef DIAGNOSTIC 761 if (vp->v_writecount != 0) { 762 vprint("vrele: bad writecount", vp); 763 panic("vrele: v_writecount != 0"); 764 } 765 #endif 766 767 if (vn_lock(vp, LK_EXCLUSIVE, p)) { 768 #ifdef DIAGNOSTIC 769 vprint("vrele: cannot lock", vp); 770 #endif 771 return (1); 772 } 773 774 VOP_INACTIVE(vp, p); 775 776 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 777 vputonfreelist(vp); 778 return (1); 779 } 780 781 /* Page or buffer structure gets a reference. */ 782 void 783 vhold(struct vnode *vp) 784 { 785 /* 786 * If it is on the freelist and the hold count is currently 787 * zero, move it to the hold list. 788 */ 789 if ((vp->v_bioflag & VBIOONFREELIST) && 790 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 791 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 792 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 793 } 794 vp->v_holdcnt++; 795 } 796 797 /* Lose interest in a vnode. */ 798 void 799 vdrop(struct vnode *vp) 800 { 801 #ifdef DIAGNOSTIC 802 if (vp->v_holdcnt == 0) 803 panic("vdrop: zero holdcnt"); 804 #endif 805 806 vp->v_holdcnt--; 807 808 /* 809 * If it is on the holdlist and the hold count drops to 810 * zero, move it to the free list. 811 */ 812 if ((vp->v_bioflag & VBIOONFREELIST) && 813 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 814 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 815 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 816 } 817 } 818 819 /* 820 * Remove any vnodes in the vnode table belonging to mount point mp. 821 * 822 * If MNT_NOFORCE is specified, there should not be any active ones, 823 * return error if any are found (nb: this is a user error, not a 824 * system error). If MNT_FORCE is specified, detach any active vnodes 825 * that are found. 826 */ 827 #ifdef DEBUG 828 int busyprt = 0; /* print out busy vnodes */ 829 struct ctldebug debug1 = { "busyprt", &busyprt }; 830 #endif 831 832 int 833 vfs_mount_foreach_vnode(struct mount *mp, 834 int (*func)(struct vnode *, void *), void *arg) { 835 struct vnode *vp, *nvp; 836 int error = 0; 837 838 loop: 839 for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; vp = nvp) { 840 if (vp->v_mount != mp) 841 goto loop; 842 nvp = LIST_NEXT(vp, v_mntvnodes); 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 struct vflush_args *va = arg; 862 struct proc *p = curproc; 863 864 if (vp == va->skipvp) { 865 return (0); 866 } 867 868 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 869 return (0); 870 } 871 872 /* 873 * If WRITECLOSE is set, only flush out regular file 874 * vnodes open for writing. 875 */ 876 if ((va->flags & WRITECLOSE) && 877 (vp->v_writecount == 0 || vp->v_type != VREG)) { 878 return (0); 879 } 880 881 /* 882 * With v_usecount == 0, all we need to do is clear 883 * out the vnode data structures and we are done. 884 */ 885 if (vp->v_usecount == 0) { 886 vgonel(vp, p); 887 return (0); 888 } 889 890 /* 891 * If FORCECLOSE is set, forcibly close the vnode. 892 * For block or character devices, revert to an 893 * anonymous device. For all other files, just kill them. 894 */ 895 if (va->flags & FORCECLOSE) { 896 if (vp->v_type != VBLK && vp->v_type != VCHR) { 897 vgonel(vp, p); 898 } else { 899 vclean(vp, 0, p); 900 vp->v_op = &spec_vops; 901 insmntque(vp, NULL); 902 } 903 return (0); 904 } 905 906 #ifdef DEBUG 907 if (busyprt) 908 vprint("vflush: busy vnode", vp); 909 #endif 910 va->busy++; 911 return (0); 912 } 913 914 int 915 vflush(struct mount *mp, struct vnode *skipvp, int flags) 916 { 917 struct vflush_args va; 918 va.skipvp = skipvp; 919 va.busy = 0; 920 va.flags = flags; 921 922 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 923 924 if (va.busy) 925 return (EBUSY); 926 return (0); 927 } 928 929 /* 930 * Disassociate the underlying file system from a vnode. 931 */ 932 void 933 vclean(struct vnode *vp, int flags, struct proc *p) 934 { 935 int active; 936 937 /* 938 * Check to see if the vnode is in use. 939 * If so we have to reference it before we clean it out 940 * so that its count cannot fall to zero and generate a 941 * race against ourselves to recycle it. 942 */ 943 if ((active = vp->v_usecount) != 0) 944 vp->v_usecount++; 945 946 /* 947 * Prevent the vnode from being recycled or 948 * brought into use while we clean it out. 949 */ 950 if (vp->v_flag & VXLOCK) 951 panic("vclean: deadlock"); 952 vp->v_flag |= VXLOCK; 953 /* 954 * Even if the count is zero, the VOP_INACTIVE routine may still 955 * have the object locked while it cleans it out. The VOP_LOCK 956 * ensures that the VOP_INACTIVE routine is done with its work. 957 * For active vnodes, it ensures that no other activity can 958 * occur while the underlying object is being cleaned out. 959 */ 960 VOP_LOCK(vp, LK_DRAIN | LK_EXCLUSIVE, p); 961 962 /* 963 * Clean out any VM data associated with the vnode. 964 */ 965 uvm_vnp_terminate(vp); 966 /* 967 * Clean out any buffers associated with the vnode. 968 */ 969 if (flags & DOCLOSE) 970 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 971 /* 972 * If purging an active vnode, it must be closed and 973 * deactivated before being reclaimed. Note that the 974 * VOP_INACTIVE will unlock the vnode 975 */ 976 if (active) { 977 if (flags & DOCLOSE) 978 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 979 VOP_INACTIVE(vp, p); 980 } else { 981 /* 982 * Any other processes trying to obtain this lock must first 983 * wait for VXLOCK to clear, then call the new lock operation. 984 */ 985 VOP_UNLOCK(vp, p); 986 } 987 988 /* 989 * Reclaim the vnode. 990 */ 991 if (VOP_RECLAIM(vp, p)) 992 panic("vclean: cannot reclaim"); 993 if (active) { 994 vp->v_usecount--; 995 if (vp->v_usecount == 0) { 996 if (vp->v_holdcnt > 0) 997 panic("vclean: not clean"); 998 vputonfreelist(vp); 999 } 1000 } 1001 cache_purge(vp); 1002 1003 /* 1004 * Done with purge, notify sleepers of the grim news. 1005 */ 1006 vp->v_op = &dead_vops; 1007 VN_KNOTE(vp, NOTE_REVOKE); 1008 vp->v_tag = VT_NON; 1009 vp->v_flag &= ~VXLOCK; 1010 #ifdef VFSLCKDEBUG 1011 vp->v_flag &= ~VLOCKSWORK; 1012 #endif 1013 if (vp->v_flag & VXWANT) { 1014 vp->v_flag &= ~VXWANT; 1015 wakeup(vp); 1016 } 1017 } 1018 1019 /* 1020 * Recycle an unused vnode to the front of the free list. 1021 */ 1022 int 1023 vrecycle(struct vnode *vp, struct proc *p) 1024 { 1025 if (vp->v_usecount == 0) { 1026 vgonel(vp, p); 1027 return (1); 1028 } 1029 return (0); 1030 } 1031 1032 /* 1033 * Eliminate all activity associated with a vnode 1034 * in preparation for reuse. 1035 */ 1036 void 1037 vgone(struct vnode *vp) 1038 { 1039 struct proc *p = curproc; 1040 vgonel(vp, p); 1041 } 1042 1043 /* 1044 * vgone, with struct proc. 1045 */ 1046 void 1047 vgonel(struct vnode *vp, struct proc *p) 1048 { 1049 struct vnode *vq; 1050 struct vnode *vx; 1051 1052 /* 1053 * If a vgone (or vclean) is already in progress, 1054 * wait until it is done and return. 1055 */ 1056 if (vp->v_flag & VXLOCK) { 1057 vp->v_flag |= VXWANT; 1058 tsleep(vp, PINOD, "vgone", 0); 1059 return; 1060 } 1061 1062 /* 1063 * Clean out the filesystem specific data. 1064 */ 1065 vclean(vp, DOCLOSE, p); 1066 /* 1067 * Delete from old mount point vnode list, if on one. 1068 */ 1069 if (vp->v_mount != NULL) 1070 insmntque(vp, NULL); 1071 /* 1072 * If special device, remove it from special device alias list 1073 * if it is on one. 1074 */ 1075 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1076 if ((vp->v_flag & VALIASED) == 0 && vp->v_type == VCHR && 1077 (cdevsw[major(vp->v_rdev)].d_flags & D_CLONE) && 1078 (minor(vp->v_rdev) >> CLONE_SHIFT == 0)) { 1079 free(vp->v_specbitmap, M_VNODE, CLONE_MAPSZ); 1080 } 1081 if (*vp->v_hashchain == vp) { 1082 *vp->v_hashchain = vp->v_specnext; 1083 } else { 1084 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1085 if (vq->v_specnext != vp) 1086 continue; 1087 vq->v_specnext = vp->v_specnext; 1088 break; 1089 } 1090 if (vq == NULL) 1091 panic("missing bdev"); 1092 } 1093 if (vp->v_flag & VALIASED) { 1094 vx = NULL; 1095 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1096 if (vq->v_rdev != vp->v_rdev || 1097 vq->v_type != vp->v_type) 1098 continue; 1099 if (vx) 1100 break; 1101 vx = vq; 1102 } 1103 if (vx == NULL) 1104 panic("missing alias"); 1105 if (vq == NULL) 1106 vx->v_flag &= ~VALIASED; 1107 vp->v_flag &= ~VALIASED; 1108 } 1109 free(vp->v_specinfo, M_VNODE, sizeof(struct specinfo)); 1110 vp->v_specinfo = NULL; 1111 } 1112 /* 1113 * If it is on the freelist and not already at the head, 1114 * move it to the head of the list. 1115 */ 1116 vp->v_type = VBAD; 1117 1118 /* 1119 * Move onto the free list, unless we were called from 1120 * getnewvnode and we're not on any free list 1121 */ 1122 if (vp->v_usecount == 0 && 1123 (vp->v_bioflag & VBIOONFREELIST)) { 1124 int s; 1125 1126 s = splbio(); 1127 1128 if (vp->v_holdcnt > 0) 1129 panic("vgonel: not clean"); 1130 1131 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1132 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1133 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1134 } 1135 splx(s); 1136 } 1137 } 1138 1139 /* 1140 * Lookup a vnode by device number. 1141 */ 1142 int 1143 vfinddev(dev_t dev, enum vtype type, struct vnode **vpp) 1144 { 1145 struct vnode *vp; 1146 int rc =0; 1147 1148 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1149 if (dev != vp->v_rdev || type != vp->v_type) 1150 continue; 1151 *vpp = vp; 1152 rc = 1; 1153 break; 1154 } 1155 return (rc); 1156 } 1157 1158 /* 1159 * Revoke all the vnodes corresponding to the specified minor number 1160 * range (endpoints inclusive) of the specified major. 1161 */ 1162 void 1163 vdevgone(int maj, int minl, int minh, enum vtype type) 1164 { 1165 struct vnode *vp; 1166 int mn; 1167 1168 for (mn = minl; mn <= minh; mn++) 1169 if (vfinddev(makedev(maj, mn), type, &vp)) 1170 VOP_REVOKE(vp, REVOKEALL); 1171 } 1172 1173 /* 1174 * Calculate the total number of references to a special device. 1175 */ 1176 int 1177 vcount(struct vnode *vp) 1178 { 1179 struct vnode *vq, *vnext; 1180 int count; 1181 1182 loop: 1183 if ((vp->v_flag & VALIASED) == 0) 1184 return (vp->v_usecount); 1185 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1186 vnext = vq->v_specnext; 1187 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1188 continue; 1189 /* 1190 * Alias, but not in use, so flush it out. 1191 */ 1192 if (vq->v_usecount == 0 && vq != vp) { 1193 vgone(vq); 1194 goto loop; 1195 } 1196 count += vq->v_usecount; 1197 } 1198 return (count); 1199 } 1200 1201 #if defined(DEBUG) || defined(DIAGNOSTIC) 1202 /* 1203 * Print out a description of a vnode. 1204 */ 1205 static char *typename[] = 1206 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1207 1208 void 1209 vprint(char *label, struct vnode *vp) 1210 { 1211 char buf[64]; 1212 1213 if (label != NULL) 1214 printf("%s: ", label); 1215 printf("%p, type %s, use %u, write %u, hold %u,", 1216 vp, typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1217 vp->v_holdcnt); 1218 buf[0] = '\0'; 1219 if (vp->v_flag & VROOT) 1220 strlcat(buf, "|VROOT", sizeof buf); 1221 if (vp->v_flag & VTEXT) 1222 strlcat(buf, "|VTEXT", sizeof buf); 1223 if (vp->v_flag & VSYSTEM) 1224 strlcat(buf, "|VSYSTEM", sizeof buf); 1225 if (vp->v_flag & VXLOCK) 1226 strlcat(buf, "|VXLOCK", sizeof buf); 1227 if (vp->v_flag & VXWANT) 1228 strlcat(buf, "|VXWANT", sizeof buf); 1229 if (vp->v_bioflag & VBIOWAIT) 1230 strlcat(buf, "|VBIOWAIT", sizeof buf); 1231 if (vp->v_bioflag & VBIOONFREELIST) 1232 strlcat(buf, "|VBIOONFREELIST", sizeof buf); 1233 if (vp->v_bioflag & VBIOONSYNCLIST) 1234 strlcat(buf, "|VBIOONSYNCLIST", sizeof buf); 1235 if (vp->v_flag & VALIASED) 1236 strlcat(buf, "|VALIASED", sizeof buf); 1237 if (buf[0] != '\0') 1238 printf(" flags (%s)", &buf[1]); 1239 if (vp->v_data == NULL) { 1240 printf("\n"); 1241 } else { 1242 printf("\n\t"); 1243 VOP_PRINT(vp); 1244 } 1245 } 1246 #endif /* DEBUG || DIAGNOSTIC */ 1247 1248 #ifdef DEBUG 1249 /* 1250 * List all of the locked vnodes in the system. 1251 * Called when debugging the kernel. 1252 */ 1253 void 1254 printlockedvnodes(void) 1255 { 1256 struct mount *mp, *nmp; 1257 struct vnode *vp; 1258 1259 printf("Locked vnodes\n"); 1260 1261 TAILQ_FOREACH_SAFE(mp, &mountlist, mnt_list, nmp) { 1262 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) 1263 continue; 1264 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 1265 if (VOP_ISLOCKED(vp)) 1266 vprint(NULL, vp); 1267 } 1268 vfs_unbusy(mp); 1269 } 1270 1271 } 1272 #endif 1273 1274 /* 1275 * Top level filesystem related information gathering. 1276 */ 1277 int 1278 vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1279 size_t newlen, struct proc *p) 1280 { 1281 struct vfsconf *vfsp, *tmpvfsp; 1282 int ret; 1283 1284 /* all sysctl names at this level are at least name and field */ 1285 if (namelen < 2) 1286 return (ENOTDIR); /* overloaded */ 1287 1288 if (name[0] != VFS_GENERIC) { 1289 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1290 if (vfsp->vfc_typenum == name[0]) 1291 break; 1292 1293 if (vfsp == NULL || vfsp->vfc_vfsops->vfs_sysctl == NULL) 1294 return (EOPNOTSUPP); 1295 1296 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1297 oldp, oldlenp, newp, newlen, p)); 1298 } 1299 1300 switch (name[1]) { 1301 case VFS_MAXTYPENUM: 1302 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1303 1304 case VFS_CONF: 1305 if (namelen < 3) 1306 return (ENOTDIR); /* overloaded */ 1307 1308 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1309 if (vfsp->vfc_typenum == name[2]) 1310 break; 1311 1312 if (vfsp == NULL) 1313 return (EOPNOTSUPP); 1314 1315 /* Make a copy, clear out kernel pointers */ 1316 tmpvfsp = malloc(sizeof(*tmpvfsp), M_TEMP, M_WAITOK); 1317 memcpy(tmpvfsp, vfsp, sizeof(*tmpvfsp)); 1318 tmpvfsp->vfc_vfsops = NULL; 1319 tmpvfsp->vfc_next = NULL; 1320 1321 ret = sysctl_rdstruct(oldp, oldlenp, newp, tmpvfsp, 1322 sizeof(struct vfsconf)); 1323 1324 free(tmpvfsp, M_TEMP, sizeof(*tmpvfsp)); 1325 return (ret); 1326 case VFS_BCACHESTAT: /* buffer cache statistics */ 1327 ret = sysctl_rdstruct(oldp, oldlenp, newp, &bcstats, 1328 sizeof(struct bcachestats)); 1329 return(ret); 1330 } 1331 return (EOPNOTSUPP); 1332 } 1333 1334 /* 1335 * Check to see if a filesystem is mounted on a block device. 1336 */ 1337 int 1338 vfs_mountedon(struct vnode *vp) 1339 { 1340 struct vnode *vq; 1341 int error = 0; 1342 1343 if (vp->v_specmountpoint != NULL) 1344 return (EBUSY); 1345 if (vp->v_flag & VALIASED) { 1346 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1347 if (vq->v_rdev != vp->v_rdev || 1348 vq->v_type != vp->v_type) 1349 continue; 1350 if (vq->v_specmountpoint != NULL) { 1351 error = EBUSY; 1352 break; 1353 } 1354 } 1355 } 1356 return (error); 1357 } 1358 1359 /* 1360 * Build hash lists of net addresses and hang them off the mount point. 1361 * Called by ufs_mount() to set up the lists of export addresses. 1362 */ 1363 int 1364 vfs_hang_addrlist(struct mount *mp, struct netexport *nep, 1365 struct export_args *argp) 1366 { 1367 struct netcred *np; 1368 struct radix_node_head *rnh; 1369 int nplen, i; 1370 struct radix_node *rn; 1371 struct sockaddr *saddr, *smask = 0; 1372 int error; 1373 1374 if (argp->ex_addrlen == 0) { 1375 if (mp->mnt_flag & MNT_DEFEXPORTED) 1376 return (EPERM); 1377 np = &nep->ne_defexported; 1378 /* fill in the kernel's ucred from userspace's xucred */ 1379 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1380 return (error); 1381 mp->mnt_flag |= MNT_DEFEXPORTED; 1382 goto finish; 1383 } 1384 if (argp->ex_addrlen > MLEN || argp->ex_masklen > MLEN || 1385 argp->ex_addrlen < 0 || argp->ex_masklen < 0) 1386 return (EINVAL); 1387 nplen = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1388 np = (struct netcred *)malloc(nplen, M_NETADDR, M_WAITOK|M_ZERO); 1389 saddr = (struct sockaddr *)(np + 1); 1390 error = copyin(argp->ex_addr, saddr, argp->ex_addrlen); 1391 if (error) 1392 goto out; 1393 if (saddr->sa_len > argp->ex_addrlen) 1394 saddr->sa_len = argp->ex_addrlen; 1395 if (argp->ex_masklen) { 1396 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1397 error = copyin(argp->ex_mask, smask, argp->ex_masklen); 1398 if (error) 1399 goto out; 1400 if (smask->sa_len > argp->ex_masklen) 1401 smask->sa_len = argp->ex_masklen; 1402 } 1403 /* fill in the kernel's ucred from userspace's xucred */ 1404 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1405 goto out; 1406 i = saddr->sa_family; 1407 switch (i) { 1408 case AF_INET: 1409 if ((rnh = nep->ne_rtable_inet) == NULL) { 1410 if (!rn_inithead((void **)&nep->ne_rtable_inet, 1411 offsetof(struct sockaddr_in, sin_addr))) { 1412 error = ENOBUFS; 1413 goto out; 1414 } 1415 rnh = nep->ne_rtable_inet; 1416 } 1417 break; 1418 default: 1419 error = EINVAL; 1420 goto out; 1421 } 1422 rn = rn_addroute(saddr, smask, rnh, np->netc_rnodes, 0); 1423 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1424 error = EPERM; 1425 goto out; 1426 } 1427 finish: 1428 np->netc_exflags = argp->ex_flags; 1429 return (0); 1430 out: 1431 free(np, M_NETADDR, nplen); 1432 return (error); 1433 } 1434 1435 int 1436 vfs_free_netcred(struct radix_node *rn, void *w, u_int id) 1437 { 1438 struct radix_node_head *rnh = (struct radix_node_head *)w; 1439 1440 rn_delete(rn->rn_key, rn->rn_mask, rnh, NULL); 1441 free(rn, M_NETADDR, 0); 1442 return (0); 1443 } 1444 1445 /* 1446 * Free the net address hash lists that are hanging off the mount points. 1447 */ 1448 void 1449 vfs_free_addrlist(struct netexport *nep) 1450 { 1451 struct radix_node_head *rnh; 1452 1453 if ((rnh = nep->ne_rtable_inet) != NULL) { 1454 rn_walktree(rnh, vfs_free_netcred, rnh); 1455 free(rnh, M_RTABLE, 0); 1456 nep->ne_rtable_inet = NULL; 1457 } 1458 } 1459 1460 int 1461 vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp) 1462 { 1463 int error; 1464 1465 if (argp->ex_flags & MNT_DELEXPORT) { 1466 vfs_free_addrlist(nep); 1467 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1468 } 1469 if (argp->ex_flags & MNT_EXPORTED) { 1470 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1471 return (error); 1472 mp->mnt_flag |= MNT_EXPORTED; 1473 } 1474 return (0); 1475 } 1476 1477 struct netcred * 1478 vfs_export_lookup(struct mount *mp, struct netexport *nep, struct mbuf *nam) 1479 { 1480 struct netcred *np; 1481 struct radix_node_head *rnh; 1482 struct sockaddr *saddr; 1483 1484 np = NULL; 1485 if (mp->mnt_flag & MNT_EXPORTED) { 1486 /* 1487 * Lookup in the export list first. 1488 */ 1489 if (nam != NULL) { 1490 saddr = mtod(nam, struct sockaddr *); 1491 switch(saddr->sa_family) { 1492 case AF_INET: 1493 rnh = nep->ne_rtable_inet; 1494 break; 1495 default: 1496 rnh = NULL; 1497 break; 1498 } 1499 if (rnh != NULL) 1500 np = (struct netcred *)rn_match(saddr, rnh); 1501 } 1502 /* 1503 * If no address match, use the default if it exists. 1504 */ 1505 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1506 np = &nep->ne_defexported; 1507 } 1508 return (np); 1509 } 1510 1511 /* 1512 * Do the usual access checking. 1513 * file_mode, uid and gid are from the vnode in question, 1514 * while acc_mode and cred are from the VOP_ACCESS parameter list 1515 */ 1516 int 1517 vaccess(enum vtype type, mode_t file_mode, uid_t uid, gid_t gid, 1518 mode_t acc_mode, struct ucred *cred) 1519 { 1520 mode_t mask; 1521 1522 /* User id 0 always gets read/write access. */ 1523 if (cred->cr_uid == 0) { 1524 /* For VEXEC, at least one of the execute bits must be set. */ 1525 if ((acc_mode & VEXEC) && type != VDIR && 1526 (file_mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0) 1527 return EACCES; 1528 return 0; 1529 } 1530 1531 mask = 0; 1532 1533 /* Otherwise, check the owner. */ 1534 if (cred->cr_uid == uid) { 1535 if (acc_mode & VEXEC) 1536 mask |= S_IXUSR; 1537 if (acc_mode & VREAD) 1538 mask |= S_IRUSR; 1539 if (acc_mode & VWRITE) 1540 mask |= S_IWUSR; 1541 return (file_mode & mask) == mask ? 0 : EACCES; 1542 } 1543 1544 /* Otherwise, check the groups. */ 1545 if (groupmember(gid, cred)) { 1546 if (acc_mode & VEXEC) 1547 mask |= S_IXGRP; 1548 if (acc_mode & VREAD) 1549 mask |= S_IRGRP; 1550 if (acc_mode & VWRITE) 1551 mask |= S_IWGRP; 1552 return (file_mode & mask) == mask ? 0 : EACCES; 1553 } 1554 1555 /* Otherwise, check everyone else. */ 1556 if (acc_mode & VEXEC) 1557 mask |= S_IXOTH; 1558 if (acc_mode & VREAD) 1559 mask |= S_IROTH; 1560 if (acc_mode & VWRITE) 1561 mask |= S_IWOTH; 1562 return (file_mode & mask) == mask ? 0 : EACCES; 1563 } 1564 1565 /* 1566 * Unmount all file systems. 1567 * We traverse the list in reverse order under the assumption that doing so 1568 * will avoid needing to worry about dependencies. 1569 */ 1570 void 1571 vfs_unmountall(void) 1572 { 1573 struct mount *mp, *nmp; 1574 int allerror, error, again = 1; 1575 1576 retry: 1577 allerror = 0; 1578 TAILQ_FOREACH_REVERSE_SAFE(mp, &mountlist, mntlist, mnt_list, nmp) { 1579 if ((vfs_busy(mp, VB_WRITE|VB_NOWAIT)) != 0) 1580 continue; 1581 if ((error = dounmount(mp, MNT_FORCE, curproc, NULL)) != 0) { 1582 printf("unmount of %s failed with error %d\n", 1583 mp->mnt_stat.f_mntonname, error); 1584 allerror = 1; 1585 } 1586 } 1587 1588 if (allerror) { 1589 printf("WARNING: some file systems would not unmount\n"); 1590 if (again) { 1591 printf("retrying\n"); 1592 again = 0; 1593 goto retry; 1594 } 1595 } 1596 } 1597 1598 /* 1599 * Sync and unmount file systems before shutting down. 1600 */ 1601 void 1602 vfs_shutdown(void) 1603 { 1604 #ifdef ACCOUNTING 1605 acct_shutdown(); 1606 #endif 1607 1608 /* XXX Should suspend scheduling. */ 1609 (void) spl0(); 1610 1611 printf("syncing disks... "); 1612 1613 if (panicstr == 0) { 1614 /* Sync before unmount, in case we hang on something. */ 1615 sys_sync(&proc0, NULL, NULL); 1616 1617 /* Unmount file systems. */ 1618 vfs_unmountall(); 1619 } 1620 1621 if (vfs_syncwait(1)) 1622 printf("giving up\n"); 1623 else 1624 printf("done\n"); 1625 1626 #if NSOFTRAID > 0 1627 sr_shutdown(); 1628 #endif 1629 } 1630 1631 /* 1632 * perform sync() operation and wait for buffers to flush. 1633 * assumptions: called w/ scheduler disabled and physical io enabled 1634 * for now called at spl0() XXX 1635 */ 1636 int 1637 vfs_syncwait(int verbose) 1638 { 1639 struct buf *bp; 1640 int iter, nbusy, dcount, s; 1641 struct proc *p; 1642 #ifdef MULTIPROCESSOR 1643 int hold_count; 1644 #endif 1645 1646 p = curproc? curproc : &proc0; 1647 sys_sync(p, NULL, NULL); 1648 1649 /* Wait for sync to finish. */ 1650 dcount = 10000; 1651 for (iter = 0; iter < 20; iter++) { 1652 nbusy = 0; 1653 LIST_FOREACH(bp, &bufhead, b_list) { 1654 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1655 nbusy++; 1656 /* 1657 * With soft updates, some buffers that are 1658 * written will be remarked as dirty until other 1659 * buffers are written. 1660 */ 1661 if (bp->b_flags & B_DELWRI) { 1662 s = splbio(); 1663 bremfree(bp); 1664 buf_acquire(bp); 1665 splx(s); 1666 nbusy++; 1667 bawrite(bp); 1668 if (dcount-- <= 0) { 1669 if (verbose) 1670 printf("softdep "); 1671 return 1; 1672 } 1673 } 1674 } 1675 if (nbusy == 0) 1676 break; 1677 if (verbose) 1678 printf("%d ", nbusy); 1679 #ifdef MULTIPROCESSOR 1680 if (__mp_lock_held(&kernel_lock)) 1681 hold_count = __mp_release_all(&kernel_lock); 1682 else 1683 hold_count = 0; 1684 #endif 1685 DELAY(40000 * iter); 1686 #ifdef MULTIPROCESSOR 1687 if (hold_count) 1688 __mp_acquire_count(&kernel_lock, hold_count); 1689 #endif 1690 } 1691 1692 return nbusy; 1693 } 1694 1695 /* 1696 * posix file system related system variables. 1697 */ 1698 int 1699 fs_posix_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, 1700 void *newp, size_t newlen, struct proc *p) 1701 { 1702 /* all sysctl names at this level are terminal */ 1703 if (namelen != 1) 1704 return (ENOTDIR); 1705 1706 switch (name[0]) { 1707 case FS_POSIX_SETUID: 1708 if (newp && securelevel > 0) 1709 return (EPERM); 1710 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1711 default: 1712 return (EOPNOTSUPP); 1713 } 1714 /* NOTREACHED */ 1715 } 1716 1717 /* 1718 * file system related system variables. 1719 */ 1720 int 1721 fs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1722 size_t newlen, struct proc *p) 1723 { 1724 sysctlfn *fn; 1725 1726 switch (name[0]) { 1727 case FS_POSIX: 1728 fn = fs_posix_sysctl; 1729 break; 1730 default: 1731 return (EOPNOTSUPP); 1732 } 1733 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1734 } 1735 1736 1737 /* 1738 * Routines dealing with vnodes and buffers 1739 */ 1740 1741 /* 1742 * Wait for all outstanding I/Os to complete 1743 * 1744 * Manipulates v_numoutput. Must be called at splbio() 1745 */ 1746 int 1747 vwaitforio(struct vnode *vp, int slpflag, char *wmesg, int timeo) 1748 { 1749 int error = 0; 1750 1751 splassert(IPL_BIO); 1752 1753 while (vp->v_numoutput) { 1754 vp->v_bioflag |= VBIOWAIT; 1755 error = tsleep(&vp->v_numoutput, 1756 slpflag | (PRIBIO + 1), wmesg, timeo); 1757 if (error) 1758 break; 1759 } 1760 1761 return (error); 1762 } 1763 1764 /* 1765 * Update outstanding I/O count and do wakeup if requested. 1766 * 1767 * Manipulates v_numoutput. Must be called at splbio() 1768 */ 1769 void 1770 vwakeup(struct vnode *vp) 1771 { 1772 splassert(IPL_BIO); 1773 1774 if (vp != NULL) { 1775 if (vp->v_numoutput-- == 0) 1776 panic("vwakeup: neg numoutput"); 1777 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1778 vp->v_bioflag &= ~VBIOWAIT; 1779 wakeup(&vp->v_numoutput); 1780 } 1781 } 1782 } 1783 1784 /* 1785 * Flush out and invalidate all buffers associated with a vnode. 1786 * Called with the underlying object locked. 1787 */ 1788 int 1789 vinvalbuf(struct vnode *vp, int flags, struct ucred *cred, struct proc *p, 1790 int slpflag, int slptimeo) 1791 { 1792 struct buf *bp; 1793 struct buf *nbp, *blist; 1794 int s, error; 1795 1796 #ifdef VFSLCKDEBUG 1797 if ((vp->v_flag & VLOCKSWORK) && !VOP_ISLOCKED(vp)) 1798 panic("vinvalbuf(): vp isn't locked"); 1799 #endif 1800 1801 if (flags & V_SAVE) { 1802 s = splbio(); 1803 vwaitforio(vp, 0, "vinvalbuf", 0); 1804 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1805 splx(s); 1806 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1807 return (error); 1808 s = splbio(); 1809 if (vp->v_numoutput > 0 || 1810 !LIST_EMPTY(&vp->v_dirtyblkhd)) 1811 panic("vinvalbuf: dirty bufs"); 1812 } 1813 splx(s); 1814 } 1815 loop: 1816 s = splbio(); 1817 for (;;) { 1818 if ((blist = LIST_FIRST(&vp->v_cleanblkhd)) && 1819 (flags & V_SAVEMETA)) 1820 while (blist && blist->b_lblkno < 0) 1821 blist = LIST_NEXT(blist, b_vnbufs); 1822 if (blist == NULL && 1823 (blist = LIST_FIRST(&vp->v_dirtyblkhd)) && 1824 (flags & V_SAVEMETA)) 1825 while (blist && blist->b_lblkno < 0) 1826 blist = LIST_NEXT(blist, b_vnbufs); 1827 if (!blist) 1828 break; 1829 1830 for (bp = blist; bp; bp = nbp) { 1831 nbp = LIST_NEXT(bp, b_vnbufs); 1832 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1833 continue; 1834 if (bp->b_flags & B_BUSY) { 1835 bp->b_flags |= B_WANTED; 1836 error = tsleep(bp, slpflag | (PRIBIO + 1), 1837 "vinvalbuf", slptimeo); 1838 if (error) { 1839 splx(s); 1840 return (error); 1841 } 1842 break; 1843 } 1844 bremfree(bp); 1845 /* 1846 * XXX Since there are no node locks for NFS, I believe 1847 * there is a slight chance that a delayed write will 1848 * occur while sleeping just above, so check for it. 1849 */ 1850 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1851 buf_acquire(bp); 1852 splx(s); 1853 (void) VOP_BWRITE(bp); 1854 goto loop; 1855 } 1856 buf_acquire_nomap(bp); 1857 bp->b_flags |= B_INVAL; 1858 brelse(bp); 1859 } 1860 } 1861 if (!(flags & V_SAVEMETA) && 1862 (!LIST_EMPTY(&vp->v_dirtyblkhd) || !LIST_EMPTY(&vp->v_cleanblkhd))) 1863 panic("vinvalbuf: flush failed"); 1864 splx(s); 1865 return (0); 1866 } 1867 1868 void 1869 vflushbuf(struct vnode *vp, int sync) 1870 { 1871 struct buf *bp, *nbp; 1872 int s; 1873 1874 loop: 1875 s = splbio(); 1876 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); bp != NULL; bp = nbp) { 1877 nbp = LIST_NEXT(bp, b_vnbufs); 1878 if ((bp->b_flags & B_BUSY)) 1879 continue; 1880 if ((bp->b_flags & B_DELWRI) == 0) 1881 panic("vflushbuf: not dirty"); 1882 bremfree(bp); 1883 buf_acquire(bp); 1884 splx(s); 1885 /* 1886 * Wait for I/O associated with indirect blocks to complete, 1887 * since there is no way to quickly wait for them below. 1888 */ 1889 if (bp->b_vp == vp || sync == 0) 1890 (void) bawrite(bp); 1891 else 1892 (void) bwrite(bp); 1893 goto loop; 1894 } 1895 if (sync == 0) { 1896 splx(s); 1897 return; 1898 } 1899 vwaitforio(vp, 0, "vflushbuf", 0); 1900 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1901 splx(s); 1902 #ifdef DIAGNOSTIC 1903 vprint("vflushbuf: dirty", vp); 1904 #endif 1905 goto loop; 1906 } 1907 splx(s); 1908 } 1909 1910 /* 1911 * Associate a buffer with a vnode. 1912 * 1913 * Manipulates buffer vnode queues. Must be called at splbio(). 1914 */ 1915 void 1916 bgetvp(struct vnode *vp, struct buf *bp) 1917 { 1918 splassert(IPL_BIO); 1919 1920 1921 if (bp->b_vp) 1922 panic("bgetvp: not free"); 1923 vhold(vp); 1924 bp->b_vp = vp; 1925 if (vp->v_type == VBLK || vp->v_type == VCHR) 1926 bp->b_dev = vp->v_rdev; 1927 else 1928 bp->b_dev = NODEV; 1929 /* 1930 * Insert onto list for new vnode. 1931 */ 1932 bufinsvn(bp, &vp->v_cleanblkhd); 1933 } 1934 1935 /* 1936 * Disassociate a buffer from a vnode. 1937 * 1938 * Manipulates vnode buffer queues. Must be called at splbio(). 1939 */ 1940 void 1941 brelvp(struct buf *bp) 1942 { 1943 struct vnode *vp; 1944 1945 splassert(IPL_BIO); 1946 1947 if ((vp = bp->b_vp) == (struct vnode *) 0) 1948 panic("brelvp: NULL"); 1949 /* 1950 * Delete from old vnode list, if on one. 1951 */ 1952 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 1953 bufremvn(bp); 1954 if ((vp->v_bioflag & VBIOONSYNCLIST) && 1955 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 1956 vp->v_bioflag &= ~VBIOONSYNCLIST; 1957 LIST_REMOVE(vp, v_synclist); 1958 } 1959 bp->b_vp = NULL; 1960 1961 vdrop(vp); 1962 } 1963 1964 /* 1965 * Replaces the current vnode associated with the buffer, if any, 1966 * with a new vnode. 1967 * 1968 * If an output I/O is pending on the buffer, the old vnode 1969 * I/O count is adjusted. 1970 * 1971 * Ignores vnode buffer queues. Must be called at splbio(). 1972 */ 1973 void 1974 buf_replacevnode(struct buf *bp, struct vnode *newvp) 1975 { 1976 struct vnode *oldvp = bp->b_vp; 1977 1978 splassert(IPL_BIO); 1979 1980 if (oldvp) 1981 brelvp(bp); 1982 1983 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 1984 newvp->v_numoutput++; /* put it on swapdev */ 1985 vwakeup(oldvp); 1986 } 1987 1988 bgetvp(newvp, bp); 1989 bufremvn(bp); 1990 } 1991 1992 /* 1993 * Used to assign buffers to the appropriate clean or dirty list on 1994 * the vnode and to add newly dirty vnodes to the appropriate 1995 * filesystem syncer list. 1996 * 1997 * Manipulates vnode buffer queues. Must be called at splbio(). 1998 */ 1999 void 2000 reassignbuf(struct buf *bp) 2001 { 2002 struct buflists *listheadp; 2003 int delay; 2004 struct vnode *vp = bp->b_vp; 2005 2006 splassert(IPL_BIO); 2007 2008 /* 2009 * Delete from old vnode list, if on one. 2010 */ 2011 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2012 bufremvn(bp); 2013 2014 /* 2015 * If dirty, put on list of dirty buffers; 2016 * otherwise insert onto list of clean buffers. 2017 */ 2018 if ((bp->b_flags & B_DELWRI) == 0) { 2019 listheadp = &vp->v_cleanblkhd; 2020 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2021 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2022 vp->v_bioflag &= ~VBIOONSYNCLIST; 2023 LIST_REMOVE(vp, v_synclist); 2024 } 2025 } else { 2026 listheadp = &vp->v_dirtyblkhd; 2027 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2028 switch (vp->v_type) { 2029 case VDIR: 2030 delay = syncdelay / 2; 2031 break; 2032 case VBLK: 2033 if (vp->v_specmountpoint != NULL) { 2034 delay = syncdelay / 3; 2035 break; 2036 } 2037 /* FALLTHROUGH */ 2038 default: 2039 delay = syncdelay; 2040 } 2041 vn_syncer_add_to_worklist(vp, delay); 2042 } 2043 } 2044 bufinsvn(bp, listheadp); 2045 } 2046 2047 int 2048 vfs_register(struct vfsconf *vfs) 2049 { 2050 struct vfsconf *vfsp; 2051 struct vfsconf **vfspp; 2052 2053 #ifdef DIAGNOSTIC 2054 /* Paranoia? */ 2055 if (vfs->vfc_refcount != 0) 2056 printf("vfs_register called with vfc_refcount > 0\n"); 2057 #endif 2058 2059 /* Check if filesystem already known */ 2060 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2061 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2062 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2063 return (EEXIST); 2064 2065 if (vfs->vfc_typenum > maxvfsconf) 2066 maxvfsconf = vfs->vfc_typenum; 2067 2068 vfs->vfc_next = NULL; 2069 2070 /* Add to the end of the list */ 2071 *vfspp = vfs; 2072 2073 /* Call vfs_init() */ 2074 if (vfs->vfc_vfsops->vfs_init) 2075 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2076 2077 return 0; 2078 } 2079 2080 int 2081 vfs_unregister(struct vfsconf *vfs) 2082 { 2083 struct vfsconf *vfsp; 2084 struct vfsconf **vfspp; 2085 int maxtypenum; 2086 2087 /* Find our vfsconf struct */ 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 break; 2092 } 2093 2094 if (!vfsp) /* Not found */ 2095 return (ENOENT); 2096 2097 if (vfsp->vfc_refcount) /* In use */ 2098 return (EBUSY); 2099 2100 /* Remove from list and free */ 2101 *vfspp = vfsp->vfc_next; 2102 2103 maxtypenum = 0; 2104 2105 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2106 if (vfsp->vfc_typenum > maxtypenum) 2107 maxtypenum = vfsp->vfc_typenum; 2108 2109 maxvfsconf = maxtypenum; 2110 return 0; 2111 } 2112 2113 /* 2114 * Check if vnode represents a disk device 2115 */ 2116 int 2117 vn_isdisk(struct vnode *vp, int *errp) 2118 { 2119 if (vp->v_type != VBLK && vp->v_type != VCHR) 2120 return (0); 2121 2122 return (1); 2123 } 2124 2125 #ifdef DDB 2126 #include <machine/db_machdep.h> 2127 #include <ddb/db_interface.h> 2128 2129 void 2130 vfs_buf_print(void *b, int full, 2131 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2132 { 2133 struct buf *bp = b; 2134 2135 (*pr)(" vp %p lblkno 0x%llx blkno 0x%llx dev 0x%x\n" 2136 " proc %p error %d flags %lb\n", 2137 bp->b_vp, (int64_t)bp->b_lblkno, (int64_t)bp->b_blkno, bp->b_dev, 2138 bp->b_proc, bp->b_error, bp->b_flags, B_BITS); 2139 2140 (*pr)(" bufsize 0x%lx bcount 0x%lx resid 0x%lx\n" 2141 " data %p saveaddr %p dep %p iodone %p\n", 2142 bp->b_bufsize, bp->b_bcount, (long)bp->b_resid, 2143 bp->b_data, bp->b_saveaddr, 2144 LIST_FIRST(&bp->b_dep), bp->b_iodone); 2145 2146 (*pr)(" dirty {off 0x%x end 0x%x} valid {off 0x%x end 0x%x}\n", 2147 bp->b_dirtyoff, bp->b_dirtyend, bp->b_validoff, bp->b_validend); 2148 2149 #ifdef FFS_SOFTUPDATES 2150 if (full) 2151 softdep_print(bp, full, pr); 2152 #endif 2153 } 2154 2155 const char *vtypes[] = { VTYPE_NAMES }; 2156 const char *vtags[] = { VTAG_NAMES }; 2157 2158 void 2159 vfs_vnode_print(void *v, int full, 2160 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2161 { 2162 struct vnode *vp = v; 2163 2164 (*pr)("tag %s(%d) type %s(%d) mount %p typedata %p\n", 2165 vp->v_tag >= nitems(vtags)? "<unk>":vtags[vp->v_tag], vp->v_tag, 2166 vp->v_type >= nitems(vtypes)? "<unk>":vtypes[vp->v_type], 2167 vp->v_type, vp->v_mount, vp->v_mountedhere); 2168 2169 (*pr)("data %p usecount %d writecount %d holdcnt %d numoutput %d\n", 2170 vp->v_data, vp->v_usecount, vp->v_writecount, 2171 vp->v_holdcnt, vp->v_numoutput); 2172 2173 /* uvm_object_printit(&vp->v_uobj, full, pr); */ 2174 2175 if (full) { 2176 struct buf *bp; 2177 2178 (*pr)("clean bufs:\n"); 2179 LIST_FOREACH(bp, &vp->v_cleanblkhd, b_vnbufs) { 2180 (*pr)(" bp %p\n", bp); 2181 vfs_buf_print(bp, full, pr); 2182 } 2183 2184 (*pr)("dirty bufs:\n"); 2185 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) { 2186 (*pr)(" bp %p\n", bp); 2187 vfs_buf_print(bp, full, pr); 2188 } 2189 } 2190 } 2191 2192 void 2193 vfs_mount_print(struct mount *mp, int full, 2194 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2195 { 2196 struct vfsconf *vfc = mp->mnt_vfc; 2197 struct vnode *vp; 2198 int cnt = 0; 2199 2200 (*pr)("flags %b\nvnodecovered %p syncer %p data %p\n", 2201 mp->mnt_flag, MNT_BITS, 2202 mp->mnt_vnodecovered, mp->mnt_syncer, mp->mnt_data); 2203 2204 (*pr)("vfsconf: ops %p name \"%s\" num %d ref %d flags 0x%x\n", 2205 vfc->vfc_vfsops, vfc->vfc_name, vfc->vfc_typenum, 2206 vfc->vfc_refcount, vfc->vfc_flags); 2207 2208 (*pr)("statvfs cache: bsize %x iosize %x\nblocks %llu free %llu avail %lld\n", 2209 mp->mnt_stat.f_bsize, mp->mnt_stat.f_iosize, mp->mnt_stat.f_blocks, 2210 mp->mnt_stat.f_bfree, mp->mnt_stat.f_bavail); 2211 2212 (*pr)(" files %llu ffiles %llu favail %lld\n", mp->mnt_stat.f_files, 2213 mp->mnt_stat.f_ffree, mp->mnt_stat.f_favail); 2214 2215 (*pr)(" f_fsidx {0x%x, 0x%x} owner %u ctime 0x%llx\n", 2216 mp->mnt_stat.f_fsid.val[0], mp->mnt_stat.f_fsid.val[1], 2217 mp->mnt_stat.f_owner, mp->mnt_stat.f_ctime); 2218 2219 (*pr)(" syncwrites %llu asyncwrites = %llu\n", 2220 mp->mnt_stat.f_syncwrites, mp->mnt_stat.f_asyncwrites); 2221 2222 (*pr)(" syncreads %llu asyncreads = %llu\n", 2223 mp->mnt_stat.f_syncreads, mp->mnt_stat.f_asyncreads); 2224 2225 (*pr)(" fstype \"%s\" mnton \"%s\" mntfrom \"%s\" mntspec \"%s\"\n", 2226 mp->mnt_stat.f_fstypename, mp->mnt_stat.f_mntonname, 2227 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntfromspec); 2228 2229 (*pr)("locked vnodes:"); 2230 /* XXX would take mountlist lock, except ddb has no context */ 2231 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2232 if (VOP_ISLOCKED(vp)) { 2233 if (!LIST_NEXT(vp, v_mntvnodes)) 2234 (*pr)(" %p", vp); 2235 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2236 (*pr)("\n\t%p", vp); 2237 else 2238 (*pr)(", %p", vp); 2239 } 2240 (*pr)("\n"); 2241 2242 if (full) { 2243 (*pr)("all vnodes:\n\t"); 2244 /* XXX would take mountlist lock, except ddb has no context */ 2245 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 2246 if (!LIST_NEXT(vp, v_mntvnodes)) 2247 (*pr)(" %p", vp); 2248 else if (!(cnt++ % (72 / (sizeof(void *) * 2 + 4)))) 2249 (*pr)(" %p,\n\t", vp); 2250 else 2251 (*pr)(" %p,", vp); 2252 (*pr)("\n"); 2253 } 2254 } 2255 #endif /* DDB */ 2256 2257 void 2258 copy_statfs_info(struct statfs *sbp, const struct mount *mp) 2259 { 2260 const struct statfs *mbp; 2261 2262 strncpy(sbp->f_fstypename, mp->mnt_vfc->vfc_name, MFSNAMELEN); 2263 2264 if (sbp == (mbp = &mp->mnt_stat)) 2265 return; 2266 2267 sbp->f_fsid = mbp->f_fsid; 2268 sbp->f_owner = mbp->f_owner; 2269 sbp->f_flags = mbp->f_flags; 2270 sbp->f_syncwrites = mbp->f_syncwrites; 2271 sbp->f_asyncwrites = mbp->f_asyncwrites; 2272 sbp->f_syncreads = mbp->f_syncreads; 2273 sbp->f_asyncreads = mbp->f_asyncreads; 2274 sbp->f_namemax = mbp->f_namemax; 2275 memcpy(sbp->f_mntonname, mp->mnt_stat.f_mntonname, MNAMELEN); 2276 memcpy(sbp->f_mntfromname, mp->mnt_stat.f_mntfromname, MNAMELEN); 2277 memcpy(sbp->f_mntfromspec, mp->mnt_stat.f_mntfromspec, MNAMELEN); 2278 memcpy(&sbp->mount_info, &mp->mnt_stat.mount_info, 2279 sizeof(union mount_info)); 2280 } 2281