vfs_subr.c revision 1.281
1/* $OpenBSD: vfs_subr.c,v 1.281 2018/09/26 14:51:44 visa 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 75void sr_quiesce(void); 76 77enum vtype iftovt_tab[16] = { 78 VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, 79 VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, 80}; 81 82int vttoif_tab[9] = { 83 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, 84 S_IFSOCK, S_IFIFO, S_IFMT, 85}; 86 87int prtactive = 0; /* 1 => print out reclaim of active vnodes */ 88int 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 99struct freelst vnode_hold_list; /* list of vnodes referencing buffers */ 100struct freelst vnode_free_list; /* vnode free list */ 101 102struct mntlist mountlist; /* mounted filesystem list */ 103 104void vclean(struct vnode *, int, struct proc *); 105 106void insmntque(struct vnode *, struct mount *); 107int getdevvp(dev_t, struct vnode **, enum vtype); 108 109int vfs_hang_addrlist(struct mount *, struct netexport *, 110 struct export_args *); 111int vfs_free_netcred(struct radix_node *, void *, u_int); 112void vfs_free_addrlist(struct netexport *); 113void vputonfreelist(struct vnode *); 114 115int vflush_vnode(struct vnode *, void *); 116int maxvnodes; 117 118void vfs_unmountall(void); 119 120#ifdef DEBUG 121void printlockedvnodes(void); 122#endif 123 124struct pool vnode_pool; 125struct pool uvm_vnode_pool; 126 127static inline int rb_buf_compare(const struct buf *b1, const struct buf *b2); 128RBT_GENERATE(buf_rb_bufs, buf, b_rbbufs, rb_buf_compare); 129 130static inline int 131rb_buf_compare(const struct buf *b1, const struct buf *b2) 132{ 133 if (b1->b_lblkno < b2->b_lblkno) 134 return(-1); 135 if (b1->b_lblkno > b2->b_lblkno) 136 return(1); 137 return(0); 138} 139 140/* 141 * Initialize the vnode management data structures. 142 */ 143void 144vntblinit(void) 145{ 146 /* buffer cache may need a vnode for each buffer */ 147 maxvnodes = 2 * initialvnodes; 148 pool_init(&vnode_pool, sizeof(struct vnode), 0, IPL_NONE, 149 PR_WAITOK, "vnodes", NULL); 150 pool_init(&uvm_vnode_pool, sizeof(struct uvm_vnode), 0, IPL_NONE, 151 PR_WAITOK, "uvmvnodes", NULL); 152 TAILQ_INIT(&vnode_hold_list); 153 TAILQ_INIT(&vnode_free_list); 154 TAILQ_INIT(&mountlist); 155 /* 156 * Initialize the filesystem syncer. 157 */ 158 vn_initialize_syncerd(); 159 160#ifdef NFSSERVER 161 rn_init(sizeof(struct sockaddr_in)); 162#endif /* NFSSERVER */ 163} 164 165/* 166 * Allocate a mount point. 167 * 168 * The returned mount point is marked as busy. 169 */ 170struct mount * 171vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp) 172{ 173 struct mount *mp; 174 175 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 176 rw_init_flags(&mp->mnt_lock, "vfslock", RWL_IS_VNODE); 177 (void)vfs_busy(mp, VB_READ|VB_NOWAIT); 178 179 LIST_INIT(&mp->mnt_vnodelist); 180 mp->mnt_vnodecovered = vp; 181 182 vfsp->vfc_refcount++; 183 mp->mnt_vfc = vfsp; 184 mp->mnt_op = vfsp->vfc_vfsops; 185 mp->mnt_flag = vfsp->vfc_flags & MNT_VISFLAGMASK; 186 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 187 188 return (mp); 189} 190 191/* 192 * Release a mount point. 193 */ 194void 195vfs_mount_free(struct mount *mp) 196{ 197 mp->mnt_vfc->vfc_refcount--; 198 free(mp, M_MOUNT, sizeof(*mp)); 199} 200 201/* 202 * Mark a mount point as busy. Used to synchronize access and to delay 203 * unmounting. 204 * 205 * Default behaviour is to attempt getting a READ lock and in case of an 206 * ongoing unmount, to wait for it to finish and then return failure. 207 */ 208int 209vfs_busy(struct mount *mp, int flags) 210{ 211 int rwflags = 0; 212 213 if (flags & VB_WRITE) 214 rwflags |= RW_WRITE; 215 else 216 rwflags |= RW_READ; 217 218 if (flags & VB_WAIT) 219 rwflags |= RW_SLEEPFAIL; 220 else 221 rwflags |= RW_NOSLEEP; 222 223#ifdef WITNESS 224 if (flags & VB_DUPOK) 225 rwflags |= RW_DUPOK; 226#endif 227 228 if (rw_enter(&mp->mnt_lock, rwflags)) 229 return (EBUSY); 230 231 return (0); 232} 233 234/* 235 * Free a busy file system 236 */ 237void 238vfs_unbusy(struct mount *mp) 239{ 240 rw_exit(&mp->mnt_lock); 241} 242 243int 244vfs_isbusy(struct mount *mp) 245{ 246 if (RWLOCK_OWNER(&mp->mnt_lock) > 0) 247 return (1); 248 else 249 return (0); 250} 251 252/* 253 * Lookup a filesystem type, and if found allocate and initialize 254 * a mount structure for it. 255 * 256 * Devname is usually updated by mount(8) after booting. 257 */ 258int 259vfs_rootmountalloc(char *fstypename, char *devname, struct mount **mpp) 260{ 261 struct vfsconf *vfsp; 262 struct mount *mp; 263 264 vfsp = vfs_byname(fstypename); 265 if (vfsp == NULL) 266 return (ENODEV); 267 mp = vfs_mount_alloc(NULLVP, vfsp); 268 mp->mnt_flag |= MNT_RDONLY; 269 mp->mnt_stat.f_mntonname[0] = '/'; 270 copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN, 0); 271 copystr(devname, mp->mnt_stat.f_mntfromspec, MNAMELEN, 0); 272 *mpp = mp; 273 return (0); 274 } 275 276/* 277 * Lookup a mount point by filesystem identifier. 278 */ 279struct mount * 280vfs_getvfs(fsid_t *fsid) 281{ 282 struct mount *mp; 283 284 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 285 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 286 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 287 return (mp); 288 } 289 } 290 291 return (NULL); 292} 293 294 295/* 296 * Get a new unique fsid 297 */ 298void 299vfs_getnewfsid(struct mount *mp) 300{ 301 static u_short xxxfs_mntid; 302 303 fsid_t tfsid; 304 int mtype; 305 306 mtype = mp->mnt_vfc->vfc_typenum; 307 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 308 mp->mnt_stat.f_fsid.val[1] = mtype; 309 if (xxxfs_mntid == 0) 310 ++xxxfs_mntid; 311 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 312 tfsid.val[1] = mtype; 313 if (!TAILQ_EMPTY(&mountlist)) { 314 while (vfs_getvfs(&tfsid)) { 315 tfsid.val[0]++; 316 xxxfs_mntid++; 317 } 318 } 319 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 320} 321 322/* 323 * Set vnode attributes to VNOVAL 324 */ 325void 326vattr_null(struct vattr *vap) 327{ 328 329 vap->va_type = VNON; 330 /* 331 * Don't get fancy: u_quad_t = u_int = VNOVAL leaves the u_quad_t 332 * with 2^31-1 instead of 2^64-1. Just write'm out and let 333 * the compiler do its job. 334 */ 335 vap->va_mode = VNOVAL; 336 vap->va_nlink = VNOVAL; 337 vap->va_uid = VNOVAL; 338 vap->va_gid = VNOVAL; 339 vap->va_fsid = VNOVAL; 340 vap->va_fileid = VNOVAL; 341 vap->va_size = VNOVAL; 342 vap->va_blocksize = VNOVAL; 343 vap->va_atime.tv_sec = VNOVAL; 344 vap->va_atime.tv_nsec = VNOVAL; 345 vap->va_mtime.tv_sec = VNOVAL; 346 vap->va_mtime.tv_nsec = VNOVAL; 347 vap->va_ctime.tv_sec = VNOVAL; 348 vap->va_ctime.tv_nsec = VNOVAL; 349 vap->va_gen = VNOVAL; 350 vap->va_flags = VNOVAL; 351 vap->va_rdev = VNOVAL; 352 vap->va_bytes = VNOVAL; 353 vap->va_filerev = VNOVAL; 354 vap->va_vaflags = 0; 355} 356 357/* 358 * Routines having to do with the management of the vnode table. 359 */ 360long numvnodes; 361 362/* 363 * Return the next vnode from the free list. 364 */ 365int 366getnewvnode(enum vtagtype tag, struct mount *mp, struct vops *vops, 367 struct vnode **vpp) 368{ 369 struct proc *p = curproc; 370 struct freelst *listhd; 371 static int toggle; 372 struct vnode *vp; 373 int s; 374 375 /* 376 * allow maxvnodes to increase if the buffer cache itself 377 * is big enough to justify it. (we don't shrink it ever) 378 */ 379 maxvnodes = maxvnodes < bcstats.numbufs ? bcstats.numbufs 380 : maxvnodes; 381 382 /* 383 * We must choose whether to allocate a new vnode or recycle an 384 * existing one. The criterion for allocating a new one is that 385 * the total number of vnodes is less than the number desired or 386 * there are no vnodes on either free list. Generally we only 387 * want to recycle vnodes that have no buffers associated with 388 * them, so we look first on the vnode_free_list. If it is empty, 389 * we next consider vnodes with referencing buffers on the 390 * vnode_hold_list. The toggle ensures that half the time we 391 * will use a buffer from the vnode_hold_list, and half the time 392 * we will allocate a new one unless the list has grown to twice 393 * the desired size. We are reticent to recycle vnodes from the 394 * vnode_hold_list because we will lose the identity of all its 395 * referencing buffers. 396 */ 397 toggle ^= 1; 398 if (numvnodes / 2 > maxvnodes) 399 toggle = 0; 400 401 s = splbio(); 402 if ((numvnodes < maxvnodes) || 403 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 404 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 405 splx(s); 406 vp = pool_get(&vnode_pool, PR_WAITOK | PR_ZERO); 407 vp->v_uvm = pool_get(&uvm_vnode_pool, PR_WAITOK | PR_ZERO); 408 vp->v_uvm->u_vnode = vp; 409 RBT_INIT(buf_rb_bufs, &vp->v_bufs_tree); 410 cache_tree_init(&vp->v_nc_tree); 411 TAILQ_INIT(&vp->v_cache_dst); 412 numvnodes++; 413 } else { 414 TAILQ_FOREACH(vp, listhd, v_freelist) { 415 if (VOP_ISLOCKED(vp) == 0) 416 break; 417 } 418 /* 419 * Unless this is a bad time of the month, at most 420 * the first NCPUS items on the free list are 421 * locked, so this is close enough to being empty. 422 */ 423 if (vp == NULL) { 424 splx(s); 425 tablefull("vnode"); 426 *vpp = 0; 427 return (ENFILE); 428 } 429 430#ifdef DIAGNOSTIC 431 if (vp->v_usecount) { 432 vprint("free vnode", vp); 433 panic("free vnode isn't"); 434 } 435#endif 436 437 TAILQ_REMOVE(listhd, vp, v_freelist); 438 vp->v_bioflag &= ~VBIOONFREELIST; 439 splx(s); 440 441 if (vp->v_type != VBAD) 442 vgonel(vp, p); 443#ifdef DIAGNOSTIC 444 if (vp->v_data) { 445 vprint("cleaned vnode", vp); 446 panic("cleaned vnode isn't"); 447 } 448 s = splbio(); 449 if (vp->v_numoutput) 450 panic("Clean vnode has pending I/O's"); 451 splx(s); 452#endif 453 vp->v_flag = 0; 454 vp->v_socket = 0; 455 } 456 cache_purge(vp); 457 vp->v_type = VNON; 458 vp->v_tag = tag; 459 vp->v_op = vops; 460 insmntque(vp, mp); 461 *vpp = vp; 462 vp->v_usecount = 1; 463 vp->v_data = 0; 464 return (0); 465} 466 467/* 468 * Move a vnode from one mount queue to another. 469 */ 470void 471insmntque(struct vnode *vp, struct mount *mp) 472{ 473 /* 474 * Delete from old mount point vnode list, if on one. 475 */ 476 if (vp->v_mount != NULL) 477 LIST_REMOVE(vp, v_mntvnodes); 478 /* 479 * Insert into list of vnodes for the new mount point, if available. 480 */ 481 if ((vp->v_mount = mp) != NULL) 482 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 483} 484 485/* 486 * Create a vnode for a block device. 487 * Used for root filesystem, argdev, and swap areas. 488 * Also used for memory file system special devices. 489 */ 490int 491bdevvp(dev_t dev, struct vnode **vpp) 492{ 493 return (getdevvp(dev, vpp, VBLK)); 494} 495 496/* 497 * Create a vnode for a character device. 498 * Used for console handling. 499 */ 500int 501cdevvp(dev_t dev, struct vnode **vpp) 502{ 503 return (getdevvp(dev, vpp, VCHR)); 504} 505 506/* 507 * Create a vnode for a device. 508 * Used by bdevvp (block device) for root file system etc., 509 * and by cdevvp (character device) for console. 510 */ 511int 512getdevvp(dev_t dev, struct vnode **vpp, enum vtype type) 513{ 514 struct vnode *vp; 515 struct vnode *nvp; 516 int error; 517 518 if (dev == NODEV) { 519 *vpp = NULLVP; 520 return (0); 521 } 522 error = getnewvnode(VT_NON, NULL, &spec_vops, &nvp); 523 if (error) { 524 *vpp = NULLVP; 525 return (error); 526 } 527 vp = nvp; 528 vp->v_type = type; 529 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 530 vput(vp); 531 vp = nvp; 532 } 533 if (vp->v_type == VCHR && cdevsw[major(vp->v_rdev)].d_type == D_TTY) 534 vp->v_flag |= VISTTY; 535 *vpp = vp; 536 return (0); 537} 538 539/* 540 * Check to see if the new vnode represents a special device 541 * for which we already have a vnode (either because of 542 * bdevvp() or because of a different vnode representing 543 * the same block device). If such an alias exists, deallocate 544 * the existing contents and return the aliased vnode. The 545 * caller is responsible for filling it with its new contents. 546 */ 547struct vnode * 548checkalias(struct vnode *nvp, dev_t nvp_rdev, struct mount *mp) 549{ 550 struct proc *p = curproc; 551 struct vnode *vp; 552 struct vnode **vpp; 553 554 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 555 return (NULLVP); 556 557 vpp = &speclisth[SPECHASH(nvp_rdev)]; 558loop: 559 for (vp = *vpp; vp; vp = vp->v_specnext) { 560 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 561 continue; 562 } 563 /* 564 * Alias, but not in use, so flush it out. 565 */ 566 if (vp->v_usecount == 0) { 567 vgonel(vp, p); 568 goto loop; 569 } 570 if (vget(vp, LK_EXCLUSIVE)) { 571 goto loop; 572 } 573 break; 574 } 575 576 /* 577 * Common case is actually in the if statement 578 */ 579 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 580 nvp->v_specinfo = malloc(sizeof(struct specinfo), M_VNODE, 581 M_WAITOK); 582 nvp->v_rdev = nvp_rdev; 583 nvp->v_hashchain = vpp; 584 nvp->v_specnext = *vpp; 585 nvp->v_specmountpoint = NULL; 586 nvp->v_speclockf = NULL; 587 nvp->v_specbitmap = NULL; 588 if (nvp->v_type == VCHR && 589 (cdevsw[major(nvp_rdev)].d_flags & D_CLONE) && 590 (minor(nvp_rdev) >> CLONE_SHIFT == 0)) { 591 if (vp != NULLVP) 592 nvp->v_specbitmap = vp->v_specbitmap; 593 else 594 nvp->v_specbitmap = malloc(CLONE_MAPSZ, 595 M_VNODE, M_WAITOK | M_ZERO); 596 } 597 *vpp = nvp; 598 if (vp != NULLVP) { 599 nvp->v_flag |= VALIASED; 600 vp->v_flag |= VALIASED; 601 vput(vp); 602 } 603 return (NULLVP); 604 } 605 606 /* 607 * This code is the uncommon case. It is called in case 608 * we found an alias that was VT_NON && vtype of VBLK 609 * This means we found a block device that was created 610 * using bdevvp. 611 * An example of such a vnode is the root partition device vnode 612 * created in ffs_mountroot. 613 * 614 * The vnodes created by bdevvp should not be aliased (why?). 615 */ 616 617 VOP_UNLOCK(vp); 618 vclean(vp, 0, p); 619 vp->v_op = nvp->v_op; 620 vp->v_tag = nvp->v_tag; 621 nvp->v_type = VNON; 622 insmntque(vp, mp); 623 return (vp); 624} 625 626/* 627 * Grab a particular vnode from the free list, increment its 628 * reference count and lock it. If the vnode lock bit is set, 629 * the vnode is being eliminated in vgone. In that case, we 630 * cannot grab it, so the process is awakened when the 631 * transition is completed, and an error code is returned to 632 * indicate that the vnode is no longer usable, possibly 633 * having been changed to a new file system type. 634 */ 635int 636vget(struct vnode *vp, int flags) 637{ 638 int error, s, onfreelist; 639 640 /* 641 * If the vnode is in the process of being cleaned out for 642 * another use, we wait for the cleaning to finish and then 643 * return failure. Cleaning is determined by checking that 644 * the VXLOCK flag is set. 645 */ 646 647 if (vp->v_flag & VXLOCK) { 648 if (flags & LK_NOWAIT) { 649 return (EBUSY); 650 } 651 652 vp->v_flag |= VXWANT; 653 tsleep(vp, PINOD, "vget", 0); 654 return (ENOENT); 655 } 656 657 onfreelist = vp->v_bioflag & VBIOONFREELIST; 658 if (vp->v_usecount == 0 && onfreelist) { 659 s = splbio(); 660 if (vp->v_holdcnt > 0) 661 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 662 else 663 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 664 vp->v_bioflag &= ~VBIOONFREELIST; 665 splx(s); 666 } 667 668 vp->v_usecount++; 669 if (flags & LK_TYPE_MASK) { 670 if ((error = vn_lock(vp, flags)) != 0) { 671 vp->v_usecount--; 672 if (vp->v_usecount == 0 && onfreelist) 673 vputonfreelist(vp); 674 } 675 return (error); 676 } 677 678 return (0); 679} 680 681 682/* Vnode reference. */ 683void 684vref(struct vnode *vp) 685{ 686#ifdef DIAGNOSTIC 687 if (vp->v_usecount == 0) 688 panic("vref used where vget required"); 689 if (vp->v_type == VNON) 690 panic("vref on a VNON vnode"); 691#endif 692 vp->v_usecount++; 693} 694 695void 696vputonfreelist(struct vnode *vp) 697{ 698 int s; 699 struct freelst *lst; 700 701 s = splbio(); 702#ifdef DIAGNOSTIC 703 if (vp->v_usecount != 0) 704 panic("Use count is not zero!"); 705 706 if (vp->v_bioflag & VBIOONFREELIST) { 707 vprint("vnode already on free list: ", vp); 708 panic("vnode already on free list"); 709 } 710#endif 711 712 vp->v_bioflag |= VBIOONFREELIST; 713 714 if (vp->v_holdcnt > 0) 715 lst = &vnode_hold_list; 716 else 717 lst = &vnode_free_list; 718 719 if (vp->v_type == VBAD) 720 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 721 else 722 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 723 724 splx(s); 725} 726 727/* 728 * vput(), just unlock and vrele() 729 */ 730void 731vput(struct vnode *vp) 732{ 733 struct proc *p = curproc; 734 735#ifdef DIAGNOSTIC 736 if (vp == NULL) 737 panic("vput: null vp"); 738#endif 739 740#ifdef DIAGNOSTIC 741 if (vp->v_usecount == 0) { 742 vprint("vput: bad ref count", vp); 743 panic("vput: ref cnt"); 744 } 745#endif 746 vp->v_usecount--; 747 KASSERT(vp->v_usecount > 0 || vp->v_uvcount == 0); 748 if (vp->v_usecount > 0) { 749 VOP_UNLOCK(vp); 750 return; 751 } 752 753#ifdef DIAGNOSTIC 754 if (vp->v_writecount != 0) { 755 vprint("vput: bad writecount", vp); 756 panic("vput: v_writecount != 0"); 757 } 758#endif 759 760 VOP_INACTIVE(vp, p); 761 762 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 763 vputonfreelist(vp); 764} 765 766/* 767 * Vnode release - use for active VNODES. 768 * If count drops to zero, call inactive routine and return to freelist. 769 * Returns 0 if it did not sleep. 770 */ 771int 772vrele(struct vnode *vp) 773{ 774 struct proc *p = curproc; 775 776#ifdef DIAGNOSTIC 777 if (vp == NULL) 778 panic("vrele: null vp"); 779#endif 780#ifdef DIAGNOSTIC 781 if (vp->v_usecount == 0) { 782 vprint("vrele: bad ref count", vp); 783 panic("vrele: ref cnt"); 784 } 785#endif 786 vp->v_usecount--; 787 if (vp->v_usecount > 0) { 788 return (0); 789 } 790 791#ifdef DIAGNOSTIC 792 if (vp->v_writecount != 0) { 793 vprint("vrele: bad writecount", vp); 794 panic("vrele: v_writecount != 0"); 795 } 796#endif 797 798 if (vn_lock(vp, LK_EXCLUSIVE)) { 799#ifdef DIAGNOSTIC 800 vprint("vrele: cannot lock", vp); 801#endif 802 return (1); 803 } 804 805 VOP_INACTIVE(vp, p); 806 807 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 808 vputonfreelist(vp); 809 return (1); 810} 811 812/* Page or buffer structure gets a reference. */ 813void 814vhold(struct vnode *vp) 815{ 816 /* 817 * If it is on the freelist and the hold count is currently 818 * zero, move it to the hold list. 819 */ 820 if ((vp->v_bioflag & VBIOONFREELIST) && 821 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 822 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 823 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 824 } 825 vp->v_holdcnt++; 826} 827 828/* Lose interest in a vnode. */ 829void 830vdrop(struct vnode *vp) 831{ 832#ifdef DIAGNOSTIC 833 if (vp->v_holdcnt == 0) 834 panic("vdrop: zero holdcnt"); 835#endif 836 837 vp->v_holdcnt--; 838 839 /* 840 * If it is on the holdlist and the hold count drops to 841 * zero, move it to the free list. 842 */ 843 if ((vp->v_bioflag & VBIOONFREELIST) && 844 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 845 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 846 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 847 } 848} 849 850/* 851 * Remove any vnodes in the vnode table belonging to mount point mp. 852 * 853 * If MNT_NOFORCE is specified, there should not be any active ones, 854 * return error if any are found (nb: this is a user error, not a 855 * system error). If MNT_FORCE is specified, detach any active vnodes 856 * that are found. 857 */ 858#ifdef DEBUG 859int busyprt = 0; /* print out busy vnodes */ 860struct ctldebug debug1 = { "busyprt", &busyprt }; 861#endif 862 863int 864vfs_mount_foreach_vnode(struct mount *mp, 865 int (*func)(struct vnode *, void *), void *arg) { 866 struct vnode *vp, *nvp; 867 int error = 0; 868 869loop: 870 LIST_FOREACH_SAFE(vp , &mp->mnt_vnodelist, v_mntvnodes, nvp) { 871 if (vp->v_mount != mp) 872 goto loop; 873 874 error = func(vp, arg); 875 876 if (error != 0) 877 break; 878 } 879 880 return (error); 881} 882 883struct vflush_args { 884 struct vnode *skipvp; 885 int busy; 886 int flags; 887}; 888 889int 890vflush_vnode(struct vnode *vp, void *arg) 891{ 892 struct vflush_args *va = arg; 893 struct proc *p = curproc; 894 895 if (vp == va->skipvp) { 896 return (0); 897 } 898 899 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 900 return (0); 901 } 902 903 /* 904 * If WRITECLOSE is set, only flush out regular file 905 * vnodes open for writing. 906 */ 907 if ((va->flags & WRITECLOSE) && 908 (vp->v_writecount == 0 || vp->v_type != VREG)) { 909 return (0); 910 } 911 912 /* 913 * With v_usecount == 0, all we need to do is clear 914 * out the vnode data structures and we are done. 915 */ 916 if (vp->v_usecount == 0) { 917 vgonel(vp, p); 918 return (0); 919 } 920 921 /* 922 * If FORCECLOSE is set, forcibly close the vnode. 923 * For block or character devices, revert to an 924 * anonymous device. For all other files, just kill them. 925 */ 926 if (va->flags & FORCECLOSE) { 927 if (vp->v_type != VBLK && vp->v_type != VCHR) { 928 vgonel(vp, p); 929 } else { 930 vclean(vp, 0, p); 931 vp->v_op = &spec_vops; 932 insmntque(vp, NULL); 933 } 934 return (0); 935 } 936 937 /* 938 * If set, this is allowed to ignore vnodes which don't 939 * have changes pending to disk. 940 * XXX Might be nice to check per-fs "inode" flags, but 941 * generally the filesystem is sync'd already, right? 942 */ 943 if ((va->flags & IGNORECLEAN) && 944 LIST_EMPTY(&vp->v_dirtyblkhd)) 945 return (0); 946 947#ifdef DEBUG 948 if (busyprt) 949 vprint("vflush: busy vnode", vp); 950#endif 951 va->busy++; 952 return (0); 953} 954 955int 956vflush(struct mount *mp, struct vnode *skipvp, int flags) 957{ 958 struct vflush_args va; 959 va.skipvp = skipvp; 960 va.busy = 0; 961 va.flags = flags; 962 963 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 964 965 if (va.busy) 966 return (EBUSY); 967 return (0); 968} 969 970/* 971 * Disassociate the underlying file system from a vnode. 972 */ 973void 974vclean(struct vnode *vp, int flags, struct proc *p) 975{ 976 int active; 977 978 /* 979 * Check to see if the vnode is in use. 980 * If so we have to reference it before we clean it out 981 * so that its count cannot fall to zero and generate a 982 * race against ourselves to recycle it. 983 */ 984 if ((active = vp->v_usecount) != 0) 985 vp->v_usecount++; 986 987 /* 988 * Prevent the vnode from being recycled or 989 * brought into use while we clean it out. 990 */ 991 if (vp->v_flag & VXLOCK) 992 panic("vclean: deadlock"); 993 vp->v_flag |= VXLOCK; 994 /* 995 * Even if the count is zero, the VOP_INACTIVE routine may still 996 * have the object locked while it cleans it out. The VOP_LOCK 997 * ensures that the VOP_INACTIVE routine is done with its work. 998 * For active vnodes, it ensures that no other activity can 999 * occur while the underlying object is being cleaned out. 1000 */ 1001 VOP_LOCK(vp, LK_DRAIN | LK_EXCLUSIVE); 1002 1003 /* 1004 * Clean out any VM data associated with the vnode. 1005 */ 1006 uvm_vnp_terminate(vp); 1007 /* 1008 * Clean out any buffers associated with the vnode. 1009 */ 1010 if (flags & DOCLOSE) 1011 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 1012 /* 1013 * If purging an active vnode, it must be closed and 1014 * deactivated before being reclaimed. Note that the 1015 * VOP_INACTIVE will unlock the vnode 1016 */ 1017 if (active) { 1018 if (flags & DOCLOSE) 1019 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 1020 VOP_INACTIVE(vp, p); 1021 } else { 1022 /* 1023 * Any other processes trying to obtain this lock must first 1024 * wait for VXLOCK to clear, then call the new lock operation. 1025 */ 1026 VOP_UNLOCK(vp); 1027 } 1028 1029 /* 1030 * Reclaim the vnode. 1031 */ 1032 if (VOP_RECLAIM(vp, p)) 1033 panic("vclean: cannot reclaim"); 1034 if (active) { 1035 vp->v_usecount--; 1036 if (vp->v_usecount == 0) { 1037 if (vp->v_holdcnt > 0) 1038 panic("vclean: not clean"); 1039 vputonfreelist(vp); 1040 } 1041 } 1042 cache_purge(vp); 1043 1044 /* 1045 * Done with purge, notify sleepers of the grim news. 1046 */ 1047 vp->v_op = &dead_vops; 1048 VN_KNOTE(vp, NOTE_REVOKE); 1049 vp->v_tag = VT_NON; 1050 vp->v_flag &= ~VXLOCK; 1051#ifdef VFSLCKDEBUG 1052 vp->v_flag &= ~VLOCKSWORK; 1053#endif 1054 if (vp->v_flag & VXWANT) { 1055 vp->v_flag &= ~VXWANT; 1056 wakeup(vp); 1057 } 1058} 1059 1060/* 1061 * Recycle an unused vnode to the front of the free list. 1062 */ 1063int 1064vrecycle(struct vnode *vp, struct proc *p) 1065{ 1066 if (vp->v_usecount == 0) { 1067 vgonel(vp, p); 1068 return (1); 1069 } 1070 return (0); 1071} 1072 1073/* 1074 * Eliminate all activity associated with a vnode 1075 * in preparation for reuse. 1076 */ 1077void 1078vgone(struct vnode *vp) 1079{ 1080 struct proc *p = curproc; 1081 vgonel(vp, p); 1082} 1083 1084/* 1085 * vgone, with struct proc. 1086 */ 1087void 1088vgonel(struct vnode *vp, struct proc *p) 1089{ 1090 struct vnode *vq; 1091 struct vnode *vx; 1092 1093 KASSERT(vp->v_uvcount == 0); 1094 1095 /* 1096 * If a vgone (or vclean) is already in progress, 1097 * wait until it is done and return. 1098 */ 1099 if (vp->v_flag & VXLOCK) { 1100 vp->v_flag |= VXWANT; 1101 tsleep(vp, PINOD, "vgone", 0); 1102 return; 1103 } 1104 1105 /* 1106 * Clean out the filesystem specific data. 1107 */ 1108 vclean(vp, DOCLOSE, p); 1109 /* 1110 * Delete from old mount point vnode list, if on one. 1111 */ 1112 if (vp->v_mount != NULL) 1113 insmntque(vp, NULL); 1114 /* 1115 * If special device, remove it from special device alias list 1116 * if it is on one. 1117 */ 1118 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1119 if ((vp->v_flag & VALIASED) == 0 && vp->v_type == VCHR && 1120 (cdevsw[major(vp->v_rdev)].d_flags & D_CLONE) && 1121 (minor(vp->v_rdev) >> CLONE_SHIFT == 0)) { 1122 free(vp->v_specbitmap, M_VNODE, CLONE_MAPSZ); 1123 } 1124 if (*vp->v_hashchain == vp) { 1125 *vp->v_hashchain = vp->v_specnext; 1126 } else { 1127 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1128 if (vq->v_specnext != vp) 1129 continue; 1130 vq->v_specnext = vp->v_specnext; 1131 break; 1132 } 1133 if (vq == NULL) 1134 panic("missing bdev"); 1135 } 1136 if (vp->v_flag & VALIASED) { 1137 vx = NULL; 1138 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1139 if (vq->v_rdev != vp->v_rdev || 1140 vq->v_type != vp->v_type) 1141 continue; 1142 if (vx) 1143 break; 1144 vx = vq; 1145 } 1146 if (vx == NULL) 1147 panic("missing alias"); 1148 if (vq == NULL) 1149 vx->v_flag &= ~VALIASED; 1150 vp->v_flag &= ~VALIASED; 1151 } 1152 free(vp->v_specinfo, M_VNODE, sizeof(struct specinfo)); 1153 vp->v_specinfo = NULL; 1154 } 1155 /* 1156 * If it is on the freelist and not already at the head, 1157 * move it to the head of the list. 1158 */ 1159 vp->v_type = VBAD; 1160 1161 /* 1162 * Move onto the free list, unless we were called from 1163 * getnewvnode and we're not on any free list 1164 */ 1165 if (vp->v_usecount == 0 && 1166 (vp->v_bioflag & VBIOONFREELIST)) { 1167 int s; 1168 1169 s = splbio(); 1170 1171 if (vp->v_holdcnt > 0) 1172 panic("vgonel: not clean"); 1173 1174 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1175 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1176 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1177 } 1178 splx(s); 1179 } 1180} 1181 1182/* 1183 * Lookup a vnode by device number. 1184 */ 1185int 1186vfinddev(dev_t dev, enum vtype type, struct vnode **vpp) 1187{ 1188 struct vnode *vp; 1189 int rc =0; 1190 1191 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1192 if (dev != vp->v_rdev || type != vp->v_type) 1193 continue; 1194 *vpp = vp; 1195 rc = 1; 1196 break; 1197 } 1198 return (rc); 1199} 1200 1201/* 1202 * Revoke all the vnodes corresponding to the specified minor number 1203 * range (endpoints inclusive) of the specified major. 1204 */ 1205void 1206vdevgone(int maj, int minl, int minh, enum vtype type) 1207{ 1208 struct vnode *vp; 1209 int mn; 1210 1211 for (mn = minl; mn <= minh; mn++) 1212 if (vfinddev(makedev(maj, mn), type, &vp)) 1213 VOP_REVOKE(vp, REVOKEALL); 1214} 1215 1216/* 1217 * Calculate the total number of references to a special device. 1218 */ 1219int 1220vcount(struct vnode *vp) 1221{ 1222 struct vnode *vq, *vnext; 1223 int count; 1224 1225loop: 1226 if ((vp->v_flag & VALIASED) == 0) 1227 return (vp->v_usecount); 1228 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1229 vnext = vq->v_specnext; 1230 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1231 continue; 1232 /* 1233 * Alias, but not in use, so flush it out. 1234 */ 1235 if (vq->v_usecount == 0 && vq != vp) { 1236 vgone(vq); 1237 goto loop; 1238 } 1239 count += vq->v_usecount; 1240 } 1241 return (count); 1242} 1243 1244#if defined(DEBUG) || defined(DIAGNOSTIC) 1245/* 1246 * Print out a description of a vnode. 1247 */ 1248static char *typename[] = 1249 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1250 1251void 1252vprint(char *label, struct vnode *vp) 1253{ 1254 char buf[64]; 1255 1256 if (label != NULL) 1257 printf("%s: ", label); 1258 printf("%p, type %s, use %u, write %u, hold %u,", 1259 vp, typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1260 vp->v_holdcnt); 1261 buf[0] = '\0'; 1262 if (vp->v_flag & VROOT) 1263 strlcat(buf, "|VROOT", sizeof buf); 1264 if (vp->v_flag & VTEXT) 1265 strlcat(buf, "|VTEXT", sizeof buf); 1266 if (vp->v_flag & VSYSTEM) 1267 strlcat(buf, "|VSYSTEM", sizeof buf); 1268 if (vp->v_flag & VXLOCK) 1269 strlcat(buf, "|VXLOCK", sizeof buf); 1270 if (vp->v_flag & VXWANT) 1271 strlcat(buf, "|VXWANT", sizeof buf); 1272 if (vp->v_bioflag & VBIOWAIT) 1273 strlcat(buf, "|VBIOWAIT", sizeof buf); 1274 if (vp->v_bioflag & VBIOONFREELIST) 1275 strlcat(buf, "|VBIOONFREELIST", sizeof buf); 1276 if (vp->v_bioflag & VBIOONSYNCLIST) 1277 strlcat(buf, "|VBIOONSYNCLIST", sizeof buf); 1278 if (vp->v_flag & VALIASED) 1279 strlcat(buf, "|VALIASED", sizeof buf); 1280 if (buf[0] != '\0') 1281 printf(" flags (%s)", &buf[1]); 1282 if (vp->v_data == NULL) { 1283 printf("\n"); 1284 } else { 1285 printf("\n\t"); 1286 VOP_PRINT(vp); 1287 } 1288} 1289#endif /* DEBUG || DIAGNOSTIC */ 1290 1291#ifdef DEBUG 1292/* 1293 * List all of the locked vnodes in the system. 1294 * Called when debugging the kernel. 1295 */ 1296void 1297printlockedvnodes(void) 1298{ 1299 struct mount *mp; 1300 struct vnode *vp; 1301 1302 printf("Locked vnodes\n"); 1303 1304 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 1305 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) 1306 continue; 1307 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 1308 if (VOP_ISLOCKED(vp)) 1309 vprint(NULL, vp); 1310 } 1311 vfs_unbusy(mp); 1312 } 1313 1314} 1315#endif 1316 1317/* 1318 * Top level filesystem related information gathering. 1319 */ 1320int 1321vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1322 size_t newlen, struct proc *p) 1323{ 1324 struct vfsconf *vfsp, *tmpvfsp; 1325 int ret; 1326 1327 /* all sysctl names at this level are at least name and field */ 1328 if (namelen < 2) 1329 return (ENOTDIR); /* overloaded */ 1330 1331 if (name[0] != VFS_GENERIC) { 1332 vfsp = vfs_bytypenum(name[0]); 1333 if (vfsp == NULL || vfsp->vfc_vfsops->vfs_sysctl == NULL) 1334 return (EOPNOTSUPP); 1335 1336 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1337 oldp, oldlenp, newp, newlen, p)); 1338 } 1339 1340 switch (name[1]) { 1341 case VFS_MAXTYPENUM: 1342 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1343 1344 case VFS_CONF: 1345 if (namelen < 3) 1346 return (ENOTDIR); /* overloaded */ 1347 1348 vfsp = vfs_bytypenum(name[2]); 1349 if (vfsp == NULL) 1350 return (EOPNOTSUPP); 1351 1352 /* Make a copy, clear out kernel pointers */ 1353 tmpvfsp = malloc(sizeof(*tmpvfsp), M_TEMP, M_WAITOK|M_ZERO); 1354 memcpy(tmpvfsp, vfsp, sizeof(*tmpvfsp)); 1355 tmpvfsp->vfc_vfsops = NULL; 1356 1357 ret = sysctl_rdstruct(oldp, oldlenp, newp, tmpvfsp, 1358 sizeof(struct vfsconf)); 1359 1360 free(tmpvfsp, M_TEMP, sizeof(*tmpvfsp)); 1361 return (ret); 1362 case VFS_BCACHESTAT: /* buffer cache statistics */ 1363 ret = sysctl_rdstruct(oldp, oldlenp, newp, &bcstats, 1364 sizeof(struct bcachestats)); 1365 return(ret); 1366 } 1367 return (EOPNOTSUPP); 1368} 1369 1370/* 1371 * Check to see if a filesystem is mounted on a block device. 1372 */ 1373int 1374vfs_mountedon(struct vnode *vp) 1375{ 1376 struct vnode *vq; 1377 int error = 0; 1378 1379 if (vp->v_specmountpoint != NULL) 1380 return (EBUSY); 1381 if (vp->v_flag & VALIASED) { 1382 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1383 if (vq->v_rdev != vp->v_rdev || 1384 vq->v_type != vp->v_type) 1385 continue; 1386 if (vq->v_specmountpoint != NULL) { 1387 error = EBUSY; 1388 break; 1389 } 1390 } 1391 } 1392 return (error); 1393} 1394 1395#ifdef NFSSERVER 1396/* 1397 * Build hash lists of net addresses and hang them off the mount point. 1398 * Called by vfs_export() to set up the lists of export addresses. 1399 */ 1400int 1401vfs_hang_addrlist(struct mount *mp, struct netexport *nep, 1402 struct export_args *argp) 1403{ 1404 struct netcred *np; 1405 struct radix_node_head *rnh; 1406 int nplen, i; 1407 struct radix_node *rn; 1408 struct sockaddr *saddr, *smask = 0; 1409 int error; 1410 1411 if (argp->ex_addrlen == 0) { 1412 if (mp->mnt_flag & MNT_DEFEXPORTED) 1413 return (EPERM); 1414 np = &nep->ne_defexported; 1415 /* fill in the kernel's ucred from userspace's xucred */ 1416 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1417 return (error); 1418 mp->mnt_flag |= MNT_DEFEXPORTED; 1419 goto finish; 1420 } 1421 if (argp->ex_addrlen > MLEN || argp->ex_masklen > MLEN || 1422 argp->ex_addrlen < 0 || argp->ex_masklen < 0) 1423 return (EINVAL); 1424 nplen = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1425 np = (struct netcred *)malloc(nplen, M_NETADDR, M_WAITOK|M_ZERO); 1426 saddr = (struct sockaddr *)(np + 1); 1427 error = copyin(argp->ex_addr, saddr, argp->ex_addrlen); 1428 if (error) 1429 goto out; 1430 if (saddr->sa_len > argp->ex_addrlen) 1431 saddr->sa_len = argp->ex_addrlen; 1432 if (argp->ex_masklen) { 1433 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1434 error = copyin(argp->ex_mask, smask, argp->ex_masklen); 1435 if (error) 1436 goto out; 1437 if (smask->sa_len > argp->ex_masklen) 1438 smask->sa_len = argp->ex_masklen; 1439 } 1440 /* fill in the kernel's ucred from userspace's xucred */ 1441 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1442 goto out; 1443 i = saddr->sa_family; 1444 switch (i) { 1445 case AF_INET: 1446 if ((rnh = nep->ne_rtable_inet) == NULL) { 1447 if (!rn_inithead((void **)&nep->ne_rtable_inet, 1448 offsetof(struct sockaddr_in, sin_addr))) { 1449 error = ENOBUFS; 1450 goto out; 1451 } 1452 rnh = nep->ne_rtable_inet; 1453 } 1454 break; 1455 default: 1456 error = EINVAL; 1457 goto out; 1458 } 1459 rn = rn_addroute(saddr, smask, rnh, np->netc_rnodes, 0); 1460 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1461 error = EPERM; 1462 goto out; 1463 } 1464finish: 1465 np->netc_exflags = argp->ex_flags; 1466 return (0); 1467out: 1468 free(np, M_NETADDR, nplen); 1469 return (error); 1470} 1471 1472int 1473vfs_free_netcred(struct radix_node *rn, void *w, u_int id) 1474{ 1475 struct radix_node_head *rnh = (struct radix_node_head *)w; 1476 1477 rn_delete(rn->rn_key, rn->rn_mask, rnh, NULL); 1478 free(rn, M_NETADDR, 0); 1479 return (0); 1480} 1481 1482/* 1483 * Free the net address hash lists that are hanging off the mount points. 1484 */ 1485void 1486vfs_free_addrlist(struct netexport *nep) 1487{ 1488 struct radix_node_head *rnh; 1489 1490 if ((rnh = nep->ne_rtable_inet) != NULL) { 1491 rn_walktree(rnh, vfs_free_netcred, rnh); 1492 free(rnh, M_RTABLE, 0); 1493 nep->ne_rtable_inet = NULL; 1494 } 1495} 1496#endif /* NFSSERVER */ 1497 1498int 1499vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp) 1500{ 1501#ifdef NFSSERVER 1502 int error; 1503 1504 if (argp->ex_flags & MNT_DELEXPORT) { 1505 vfs_free_addrlist(nep); 1506 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1507 } 1508 if (argp->ex_flags & MNT_EXPORTED) { 1509 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1510 return (error); 1511 mp->mnt_flag |= MNT_EXPORTED; 1512 } 1513 return (0); 1514#else 1515 return (ENOTSUP); 1516#endif /* NFSSERVER */ 1517} 1518 1519struct netcred * 1520vfs_export_lookup(struct mount *mp, struct netexport *nep, struct mbuf *nam) 1521{ 1522#ifdef NFSSERVER 1523 struct netcred *np; 1524 struct radix_node_head *rnh; 1525 struct sockaddr *saddr; 1526 1527 np = NULL; 1528 if (mp->mnt_flag & MNT_EXPORTED) { 1529 /* 1530 * Lookup in the export list first. 1531 */ 1532 if (nam != NULL) { 1533 saddr = mtod(nam, struct sockaddr *); 1534 switch(saddr->sa_family) { 1535 case AF_INET: 1536 rnh = nep->ne_rtable_inet; 1537 break; 1538 default: 1539 rnh = NULL; 1540 break; 1541 } 1542 if (rnh != NULL) 1543 np = (struct netcred *)rn_match(saddr, rnh); 1544 } 1545 /* 1546 * If no address match, use the default if it exists. 1547 */ 1548 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1549 np = &nep->ne_defexported; 1550 } 1551 return (np); 1552#else 1553 return (NULL); 1554#endif /* NFSSERVER */ 1555} 1556 1557/* 1558 * Do the usual access checking. 1559 * file_mode, uid and gid are from the vnode in question, 1560 * while acc_mode and cred are from the VOP_ACCESS parameter list 1561 */ 1562int 1563vaccess(enum vtype type, mode_t file_mode, uid_t uid, gid_t gid, 1564 mode_t acc_mode, struct ucred *cred) 1565{ 1566 mode_t mask; 1567 1568 /* User id 0 always gets read/write access. */ 1569 if (cred->cr_uid == 0) { 1570 /* For VEXEC, at least one of the execute bits must be set. */ 1571 if ((acc_mode & VEXEC) && type != VDIR && 1572 (file_mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0) 1573 return EACCES; 1574 return 0; 1575 } 1576 1577 mask = 0; 1578 1579 /* Otherwise, check the owner. */ 1580 if (cred->cr_uid == uid) { 1581 if (acc_mode & VEXEC) 1582 mask |= S_IXUSR; 1583 if (acc_mode & VREAD) 1584 mask |= S_IRUSR; 1585 if (acc_mode & VWRITE) 1586 mask |= S_IWUSR; 1587 return (file_mode & mask) == mask ? 0 : EACCES; 1588 } 1589 1590 /* Otherwise, check the groups. */ 1591 if (groupmember(gid, cred)) { 1592 if (acc_mode & VEXEC) 1593 mask |= S_IXGRP; 1594 if (acc_mode & VREAD) 1595 mask |= S_IRGRP; 1596 if (acc_mode & VWRITE) 1597 mask |= S_IWGRP; 1598 return (file_mode & mask) == mask ? 0 : EACCES; 1599 } 1600 1601 /* Otherwise, check everyone else. */ 1602 if (acc_mode & VEXEC) 1603 mask |= S_IXOTH; 1604 if (acc_mode & VREAD) 1605 mask |= S_IROTH; 1606 if (acc_mode & VWRITE) 1607 mask |= S_IWOTH; 1608 return (file_mode & mask) == mask ? 0 : EACCES; 1609} 1610 1611struct rwlock vfs_stall_lock = RWLOCK_INITIALIZER("vfs_stall"); 1612 1613int 1614vfs_stall(struct proc *p, int stall) 1615{ 1616 struct mount *mp; 1617 int allerror = 0, error; 1618 1619 if (stall) 1620 rw_enter_write(&vfs_stall_lock); 1621 1622 /* 1623 * The loop variable mp is protected by vfs_busy() so that it cannot 1624 * be unmounted while VFS_SYNC() sleeps. Traverse forward to keep the 1625 * lock order consistent with dounmount(). 1626 */ 1627 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 1628 if (stall) { 1629 error = vfs_busy(mp, VB_WRITE|VB_WAIT|VB_DUPOK); 1630 if (error) { 1631 printf("%s: busy\n", mp->mnt_stat.f_mntonname); 1632 allerror = error; 1633 continue; 1634 } 1635 uvm_vnp_sync(mp); 1636 error = VFS_SYNC(mp, MNT_WAIT, stall, p->p_ucred, p); 1637 if (error) { 1638 printf("%s: failed to sync\n", mp->mnt_stat.f_mntonname); 1639 vfs_unbusy(mp); 1640 allerror = error; 1641 continue; 1642 } 1643 mp->mnt_flag |= MNT_STALLED; 1644 } else { 1645 if (mp->mnt_flag & MNT_STALLED) { 1646 vfs_unbusy(mp); 1647 mp->mnt_flag &= ~MNT_STALLED; 1648 } 1649 } 1650 } 1651 1652 if (!stall) 1653 rw_exit_write(&vfs_stall_lock); 1654 1655 return (allerror); 1656} 1657 1658void 1659vfs_stall_barrier(void) 1660{ 1661 rw_enter_read(&vfs_stall_lock); 1662 rw_exit_read(&vfs_stall_lock); 1663} 1664 1665/* 1666 * Unmount all file systems. 1667 * We traverse the list in reverse order under the assumption that doing so 1668 * will avoid needing to worry about dependencies. 1669 */ 1670void 1671vfs_unmountall(void) 1672{ 1673 struct mount *mp, *nmp; 1674 int allerror, error, again = 1; 1675 1676 retry: 1677 allerror = 0; 1678 TAILQ_FOREACH_REVERSE_SAFE(mp, &mountlist, mntlist, mnt_list, nmp) { 1679 if (vfs_busy(mp, VB_WRITE|VB_NOWAIT)) 1680 continue; 1681 /* XXX Here is a race, the next pointer is not locked. */ 1682 if ((error = dounmount(mp, MNT_FORCE, curproc)) != 0) { 1683 printf("unmount of %s failed with error %d\n", 1684 mp->mnt_stat.f_mntonname, error); 1685 allerror = 1; 1686 } 1687 } 1688 1689 if (allerror) { 1690 printf("WARNING: some file systems would not unmount\n"); 1691 if (again) { 1692 printf("retrying\n"); 1693 again = 0; 1694 goto retry; 1695 } 1696 } 1697} 1698 1699/* 1700 * Sync and unmount file systems before shutting down. 1701 */ 1702void 1703vfs_shutdown(struct proc *p) 1704{ 1705#ifdef ACCOUNTING 1706 acct_shutdown(); 1707#endif 1708 1709 printf("syncing disks..."); 1710 1711 if (panicstr == 0) { 1712 /* Sync before unmount, in case we hang on something. */ 1713 sys_sync(p, NULL, NULL); 1714 vfs_unmountall(); 1715 } 1716 1717#if NSOFTRAID > 0 1718 sr_quiesce(); 1719#endif 1720 1721 if (vfs_syncwait(p, 1)) 1722 printf(" giving up\n"); 1723 else 1724 printf(" done\n"); 1725} 1726 1727/* 1728 * perform sync() operation and wait for buffers to flush. 1729 */ 1730int 1731vfs_syncwait(struct proc *p, int verbose) 1732{ 1733 struct buf *bp; 1734 int iter, nbusy, dcount, s; 1735#ifdef MULTIPROCESSOR 1736 int hold_count; 1737#endif 1738 1739 sys_sync(p, NULL, NULL); 1740 1741 /* Wait for sync to finish. */ 1742 dcount = 10000; 1743 for (iter = 0; iter < 20; iter++) { 1744 nbusy = 0; 1745 LIST_FOREACH(bp, &bufhead, b_list) { 1746 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1747 nbusy++; 1748 /* 1749 * With soft updates, some buffers that are 1750 * written will be remarked as dirty until other 1751 * buffers are written. 1752 */ 1753 if (bp->b_flags & B_DELWRI) { 1754 s = splbio(); 1755 bremfree(bp); 1756 buf_acquire(bp); 1757 splx(s); 1758 nbusy++; 1759 bawrite(bp); 1760 if (dcount-- <= 0) { 1761 if (verbose) 1762 printf("softdep "); 1763 return 1; 1764 } 1765 } 1766 } 1767 if (nbusy == 0) 1768 break; 1769 if (verbose) 1770 printf("%d ", nbusy); 1771#ifdef MULTIPROCESSOR 1772 if (_kernel_lock_held()) 1773 hold_count = __mp_release_all(&kernel_lock); 1774 else 1775 hold_count = 0; 1776#endif 1777 DELAY(40000 * iter); 1778#ifdef MULTIPROCESSOR 1779 if (hold_count) 1780 __mp_acquire_count(&kernel_lock, hold_count); 1781#endif 1782 } 1783 1784 return nbusy; 1785} 1786 1787/* 1788 * posix file system related system variables. 1789 */ 1790int 1791fs_posix_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, 1792 void *newp, size_t newlen, struct proc *p) 1793{ 1794 /* all sysctl names at this level are terminal */ 1795 if (namelen != 1) 1796 return (ENOTDIR); 1797 1798 switch (name[0]) { 1799 case FS_POSIX_SETUID: 1800 if (newp && securelevel > 0) 1801 return (EPERM); 1802 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1803 default: 1804 return (EOPNOTSUPP); 1805 } 1806 /* NOTREACHED */ 1807} 1808 1809/* 1810 * file system related system variables. 1811 */ 1812int 1813fs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1814 size_t newlen, struct proc *p) 1815{ 1816 sysctlfn *fn; 1817 1818 switch (name[0]) { 1819 case FS_POSIX: 1820 fn = fs_posix_sysctl; 1821 break; 1822 default: 1823 return (EOPNOTSUPP); 1824 } 1825 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1826} 1827 1828 1829/* 1830 * Routines dealing with vnodes and buffers 1831 */ 1832 1833/* 1834 * Wait for all outstanding I/Os to complete 1835 * 1836 * Manipulates v_numoutput. Must be called at splbio() 1837 */ 1838int 1839vwaitforio(struct vnode *vp, int slpflag, char *wmesg, int timeo) 1840{ 1841 int error = 0; 1842 1843 splassert(IPL_BIO); 1844 1845 while (vp->v_numoutput) { 1846 vp->v_bioflag |= VBIOWAIT; 1847 error = tsleep(&vp->v_numoutput, 1848 slpflag | (PRIBIO + 1), wmesg, timeo); 1849 if (error) 1850 break; 1851 } 1852 1853 return (error); 1854} 1855 1856/* 1857 * Update outstanding I/O count and do wakeup if requested. 1858 * 1859 * Manipulates v_numoutput. Must be called at splbio() 1860 */ 1861void 1862vwakeup(struct vnode *vp) 1863{ 1864 splassert(IPL_BIO); 1865 1866 if (vp != NULL) { 1867 if (vp->v_numoutput-- == 0) 1868 panic("vwakeup: neg numoutput"); 1869 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1870 vp->v_bioflag &= ~VBIOWAIT; 1871 wakeup(&vp->v_numoutput); 1872 } 1873 } 1874} 1875 1876/* 1877 * Flush out and invalidate all buffers associated with a vnode. 1878 * Called with the underlying object locked. 1879 */ 1880int 1881vinvalbuf(struct vnode *vp, int flags, struct ucred *cred, struct proc *p, 1882 int slpflag, int slptimeo) 1883{ 1884 struct buf *bp; 1885 struct buf *nbp, *blist; 1886 int s, error; 1887 1888#ifdef VFSLCKDEBUG 1889 if ((vp->v_flag & VLOCKSWORK) && !VOP_ISLOCKED(vp)) 1890 panic("%s: vp isn't locked, vp %p", __func__, vp); 1891#endif 1892 1893 if (flags & V_SAVE) { 1894 s = splbio(); 1895 vwaitforio(vp, 0, "vinvalbuf", 0); 1896 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1897 splx(s); 1898 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1899 return (error); 1900 s = splbio(); 1901 if (vp->v_numoutput > 0 || 1902 !LIST_EMPTY(&vp->v_dirtyblkhd)) 1903 panic("%s: dirty bufs, vp %p", __func__, vp); 1904 } 1905 splx(s); 1906 } 1907loop: 1908 s = splbio(); 1909 for (;;) { 1910 if ((blist = LIST_FIRST(&vp->v_cleanblkhd)) && 1911 (flags & V_SAVEMETA)) 1912 while (blist && blist->b_lblkno < 0) 1913 blist = LIST_NEXT(blist, b_vnbufs); 1914 if (blist == NULL && 1915 (blist = LIST_FIRST(&vp->v_dirtyblkhd)) && 1916 (flags & V_SAVEMETA)) 1917 while (blist && blist->b_lblkno < 0) 1918 blist = LIST_NEXT(blist, b_vnbufs); 1919 if (!blist) 1920 break; 1921 1922 for (bp = blist; bp; bp = nbp) { 1923 nbp = LIST_NEXT(bp, b_vnbufs); 1924 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1925 continue; 1926 if (bp->b_flags & B_BUSY) { 1927 bp->b_flags |= B_WANTED; 1928 error = tsleep(bp, slpflag | (PRIBIO + 1), 1929 "vinvalbuf", slptimeo); 1930 if (error) { 1931 splx(s); 1932 return (error); 1933 } 1934 break; 1935 } 1936 bremfree(bp); 1937 /* 1938 * XXX Since there are no node locks for NFS, I believe 1939 * there is a slight chance that a delayed write will 1940 * occur while sleeping just above, so check for it. 1941 */ 1942 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1943 buf_acquire(bp); 1944 splx(s); 1945 (void) VOP_BWRITE(bp); 1946 goto loop; 1947 } 1948 buf_acquire_nomap(bp); 1949 bp->b_flags |= B_INVAL; 1950 brelse(bp); 1951 } 1952 } 1953 if (!(flags & V_SAVEMETA) && 1954 (!LIST_EMPTY(&vp->v_dirtyblkhd) || !LIST_EMPTY(&vp->v_cleanblkhd))) 1955 panic("%s: flush failed, vp %p", __func__, vp); 1956 splx(s); 1957 return (0); 1958} 1959 1960void 1961vflushbuf(struct vnode *vp, int sync) 1962{ 1963 struct buf *bp, *nbp; 1964 int s; 1965 1966loop: 1967 s = splbio(); 1968 LIST_FOREACH_SAFE(bp, &vp->v_dirtyblkhd, b_vnbufs, nbp) { 1969 if ((bp->b_flags & B_BUSY)) 1970 continue; 1971 if ((bp->b_flags & B_DELWRI) == 0) 1972 panic("vflushbuf: not dirty"); 1973 bremfree(bp); 1974 buf_acquire(bp); 1975 splx(s); 1976 /* 1977 * Wait for I/O associated with indirect blocks to complete, 1978 * since there is no way to quickly wait for them below. 1979 */ 1980 if (bp->b_vp == vp || sync == 0) 1981 (void) bawrite(bp); 1982 else 1983 (void) bwrite(bp); 1984 goto loop; 1985 } 1986 if (sync == 0) { 1987 splx(s); 1988 return; 1989 } 1990 vwaitforio(vp, 0, "vflushbuf", 0); 1991 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1992 splx(s); 1993#ifdef DIAGNOSTIC 1994 vprint("vflushbuf: dirty", vp); 1995#endif 1996 goto loop; 1997 } 1998 splx(s); 1999} 2000 2001/* 2002 * Associate a buffer with a vnode. 2003 * 2004 * Manipulates buffer vnode queues. Must be called at splbio(). 2005 */ 2006void 2007bgetvp(struct vnode *vp, struct buf *bp) 2008{ 2009 splassert(IPL_BIO); 2010 2011 2012 if (bp->b_vp) 2013 panic("bgetvp: not free"); 2014 vhold(vp); 2015 bp->b_vp = vp; 2016 if (vp->v_type == VBLK || vp->v_type == VCHR) 2017 bp->b_dev = vp->v_rdev; 2018 else 2019 bp->b_dev = NODEV; 2020 /* 2021 * Insert onto list for new vnode. 2022 */ 2023 bufinsvn(bp, &vp->v_cleanblkhd); 2024} 2025 2026/* 2027 * Disassociate a buffer from a vnode. 2028 * 2029 * Manipulates vnode buffer queues. Must be called at splbio(). 2030 */ 2031void 2032brelvp(struct buf *bp) 2033{ 2034 struct vnode *vp; 2035 2036 splassert(IPL_BIO); 2037 2038 if ((vp = bp->b_vp) == (struct vnode *) 0) 2039 panic("brelvp: NULL"); 2040 /* 2041 * Delete from old vnode list, if on one. 2042 */ 2043 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2044 bufremvn(bp); 2045 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2046 LIST_EMPTY(&vp->v_dirtyblkhd)) { 2047 vp->v_bioflag &= ~VBIOONSYNCLIST; 2048 LIST_REMOVE(vp, v_synclist); 2049 } 2050 bp->b_vp = NULL; 2051 2052 vdrop(vp); 2053} 2054 2055/* 2056 * Replaces the current vnode associated with the buffer, if any, 2057 * with a new vnode. 2058 * 2059 * If an output I/O is pending on the buffer, the old vnode 2060 * I/O count is adjusted. 2061 * 2062 * Ignores vnode buffer queues. Must be called at splbio(). 2063 */ 2064void 2065buf_replacevnode(struct buf *bp, struct vnode *newvp) 2066{ 2067 struct vnode *oldvp = bp->b_vp; 2068 2069 splassert(IPL_BIO); 2070 2071 if (oldvp) 2072 brelvp(bp); 2073 2074 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2075 newvp->v_numoutput++; /* put it on swapdev */ 2076 vwakeup(oldvp); 2077 } 2078 2079 bgetvp(newvp, bp); 2080 bufremvn(bp); 2081} 2082 2083/* 2084 * Used to assign buffers to the appropriate clean or dirty list on 2085 * the vnode and to add newly dirty vnodes to the appropriate 2086 * filesystem syncer list. 2087 * 2088 * Manipulates vnode buffer queues. Must be called at splbio(). 2089 */ 2090void 2091reassignbuf(struct buf *bp) 2092{ 2093 struct buflists *listheadp; 2094 int delay; 2095 struct vnode *vp = bp->b_vp; 2096 2097 splassert(IPL_BIO); 2098 2099 /* 2100 * Delete from old vnode list, if on one. 2101 */ 2102 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2103 bufremvn(bp); 2104 2105 /* 2106 * If dirty, put on list of dirty buffers; 2107 * otherwise insert onto list of clean buffers. 2108 */ 2109 if ((bp->b_flags & B_DELWRI) == 0) { 2110 listheadp = &vp->v_cleanblkhd; 2111 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2112 LIST_EMPTY(&vp->v_dirtyblkhd)) { 2113 vp->v_bioflag &= ~VBIOONSYNCLIST; 2114 LIST_REMOVE(vp, v_synclist); 2115 } 2116 } else { 2117 listheadp = &vp->v_dirtyblkhd; 2118 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2119 switch (vp->v_type) { 2120 case VDIR: 2121 delay = syncdelay / 2; 2122 break; 2123 case VBLK: 2124 if (vp->v_specmountpoint != NULL) { 2125 delay = syncdelay / 3; 2126 break; 2127 } 2128 /* FALLTHROUGH */ 2129 default: 2130 delay = syncdelay; 2131 } 2132 vn_syncer_add_to_worklist(vp, delay); 2133 } 2134 } 2135 bufinsvn(bp, listheadp); 2136} 2137 2138/* 2139 * Check if vnode represents a disk device 2140 */ 2141int 2142vn_isdisk(struct vnode *vp, int *errp) 2143{ 2144 if (vp->v_type != VBLK && vp->v_type != VCHR) 2145 return (0); 2146 2147 return (1); 2148} 2149 2150#ifdef DDB 2151#include <machine/db_machdep.h> 2152#include <ddb/db_interface.h> 2153 2154void 2155vfs_buf_print(void *b, int full, 2156 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2157{ 2158 struct buf *bp = b; 2159 2160 (*pr)(" vp %p lblkno 0x%llx blkno 0x%llx dev 0x%x\n" 2161 " proc %p error %d flags %lb\n", 2162 bp->b_vp, (int64_t)bp->b_lblkno, (int64_t)bp->b_blkno, bp->b_dev, 2163 bp->b_proc, bp->b_error, bp->b_flags, B_BITS); 2164 2165 (*pr)(" bufsize 0x%lx bcount 0x%lx resid 0x%lx\n" 2166 " data %p saveaddr %p dep %p iodone %p\n", 2167 bp->b_bufsize, bp->b_bcount, (long)bp->b_resid, 2168 bp->b_data, bp->b_saveaddr, 2169 LIST_FIRST(&bp->b_dep), bp->b_iodone); 2170 2171 (*pr)(" dirty {off 0x%x end 0x%x} valid {off 0x%x end 0x%x}\n", 2172 bp->b_dirtyoff, bp->b_dirtyend, bp->b_validoff, bp->b_validend); 2173 2174#ifdef FFS_SOFTUPDATES 2175 if (full) 2176 softdep_print(bp, full, pr); 2177#endif 2178} 2179 2180const char *vtypes[] = { VTYPE_NAMES }; 2181const char *vtags[] = { VTAG_NAMES }; 2182 2183void 2184vfs_vnode_print(void *v, int full, 2185 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2186{ 2187 struct vnode *vp = v; 2188 2189 (*pr)("tag %s(%d) type %s(%d) mount %p typedata %p\n", 2190 (u_int)vp->v_tag >= nitems(vtags)? "<unk>":vtags[vp->v_tag], 2191 vp->v_tag, 2192 (u_int)vp->v_type >= nitems(vtypes)? "<unk>":vtypes[vp->v_type], 2193 vp->v_type, vp->v_mount, vp->v_mountedhere); 2194 2195 (*pr)("data %p usecount %d writecount %d holdcnt %d numoutput %d\n", 2196 vp->v_data, vp->v_usecount, vp->v_writecount, 2197 vp->v_holdcnt, vp->v_numoutput); 2198 2199 /* uvm_object_printit(&vp->v_uobj, full, pr); */ 2200 2201 if (full) { 2202 struct buf *bp; 2203 2204 (*pr)("clean bufs:\n"); 2205 LIST_FOREACH(bp, &vp->v_cleanblkhd, b_vnbufs) { 2206 (*pr)(" bp %p\n", bp); 2207 vfs_buf_print(bp, full, pr); 2208 } 2209 2210 (*pr)("dirty bufs:\n"); 2211 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) { 2212 (*pr)(" bp %p\n", bp); 2213 vfs_buf_print(bp, full, pr); 2214 } 2215 } 2216} 2217 2218void 2219vfs_mount_print(struct mount *mp, int full, 2220 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2221{ 2222 struct vfsconf *vfc = mp->mnt_vfc; 2223 struct vnode *vp; 2224 int cnt; 2225 2226 (*pr)("flags %b\nvnodecovered %p syncer %p data %p\n", 2227 mp->mnt_flag, MNT_BITS, 2228 mp->mnt_vnodecovered, mp->mnt_syncer, mp->mnt_data); 2229 2230 (*pr)("vfsconf: ops %p name \"%s\" num %d ref %d flags 0x%x\n", 2231 vfc->vfc_vfsops, vfc->vfc_name, vfc->vfc_typenum, 2232 vfc->vfc_refcount, vfc->vfc_flags); 2233 2234 (*pr)("statvfs cache: bsize %x iosize %x\nblocks %llu free %llu avail %lld\n", 2235 mp->mnt_stat.f_bsize, mp->mnt_stat.f_iosize, mp->mnt_stat.f_blocks, 2236 mp->mnt_stat.f_bfree, mp->mnt_stat.f_bavail); 2237 2238 (*pr)(" files %llu ffiles %llu favail %lld\n", mp->mnt_stat.f_files, 2239 mp->mnt_stat.f_ffree, mp->mnt_stat.f_favail); 2240 2241 (*pr)(" f_fsidx {0x%x, 0x%x} owner %u ctime 0x%llx\n", 2242 mp->mnt_stat.f_fsid.val[0], mp->mnt_stat.f_fsid.val[1], 2243 mp->mnt_stat.f_owner, mp->mnt_stat.f_ctime); 2244 2245 (*pr)(" syncwrites %llu asyncwrites = %llu\n", 2246 mp->mnt_stat.f_syncwrites, mp->mnt_stat.f_asyncwrites); 2247 2248 (*pr)(" syncreads %llu asyncreads = %llu\n", 2249 mp->mnt_stat.f_syncreads, mp->mnt_stat.f_asyncreads); 2250 2251 (*pr)(" fstype \"%s\" mnton \"%s\" mntfrom \"%s\" mntspec \"%s\"\n", 2252 mp->mnt_stat.f_fstypename, mp->mnt_stat.f_mntonname, 2253 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntfromspec); 2254 2255 (*pr)("locked vnodes:"); 2256 /* XXX would take mountlist lock, except ddb has no context */ 2257 cnt = 0; 2258 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2259 if (VOP_ISLOCKED(vp)) { 2260 if (cnt == 0) 2261 (*pr)("\n %p", vp); 2262 else if ((cnt % (72 / (sizeof(void *) * 2 + 4))) == 0) 2263 (*pr)(",\n %p", vp); 2264 else 2265 (*pr)(", %p", vp); 2266 cnt++; 2267 } 2268 } 2269 (*pr)("\n"); 2270 2271 if (full) { 2272 (*pr)("all vnodes:"); 2273 /* XXX would take mountlist lock, except ddb has no context */ 2274 cnt = 0; 2275 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2276 if (cnt == 0) 2277 (*pr)("\n %p", vp); 2278 else if ((cnt % (72 / (sizeof(void *) * 2 + 4))) == 0) 2279 (*pr)(",\n %p", vp); 2280 else 2281 (*pr)(", %p", vp); 2282 cnt++; 2283 } 2284 (*pr)("\n"); 2285 } 2286} 2287#endif /* DDB */ 2288 2289void 2290copy_statfs_info(struct statfs *sbp, const struct mount *mp) 2291{ 2292 const struct statfs *mbp; 2293 2294 strncpy(sbp->f_fstypename, mp->mnt_vfc->vfc_name, MFSNAMELEN); 2295 2296 if (sbp == (mbp = &mp->mnt_stat)) 2297 return; 2298 2299 sbp->f_fsid = mbp->f_fsid; 2300 sbp->f_owner = mbp->f_owner; 2301 sbp->f_flags = mbp->f_flags; 2302 sbp->f_syncwrites = mbp->f_syncwrites; 2303 sbp->f_asyncwrites = mbp->f_asyncwrites; 2304 sbp->f_syncreads = mbp->f_syncreads; 2305 sbp->f_asyncreads = mbp->f_asyncreads; 2306 sbp->f_namemax = mbp->f_namemax; 2307 memcpy(sbp->f_mntonname, mp->mnt_stat.f_mntonname, MNAMELEN); 2308 memcpy(sbp->f_mntfromname, mp->mnt_stat.f_mntfromname, MNAMELEN); 2309 memcpy(sbp->f_mntfromspec, mp->mnt_stat.f_mntfromspec, MNAMELEN); 2310 memcpy(&sbp->mount_info, &mp->mnt_stat.mount_info, 2311 sizeof(union mount_info)); 2312} 2313