vfs_subr.c revision 1.263
1/* $OpenBSD: vfs_subr.c,v 1.263 2017/12/11 14:11:22 bluhm Exp $ */ 2/* $NetBSD: vfs_subr.c,v 1.53 1996/04/22 01:39:13 christos Exp $ */ 3 4/* 5 * Copyright (c) 1989, 1993 6 * The Regents of the University of California. All rights reserved. 7 * (c) UNIX System Laboratories, Inc. 8 * All or some portions of this file are derived from material licensed 9 * to the University of California by American Telephone and Telegraph 10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 11 * the permission of UNIX System Laboratories, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 * 37 * @(#)vfs_subr.c 8.13 (Berkeley) 4/18/94 38 */ 39 40/* 41 * External virtual filesystem routines 42 */ 43 44#include <sys/param.h> 45#include <sys/systm.h> 46#include <sys/proc.h> 47#include <sys/sysctl.h> 48#include <sys/mount.h> 49#include <sys/time.h> 50#include <sys/fcntl.h> 51#include <sys/kernel.h> 52#include <sys/conf.h> 53#include <sys/vnode.h> 54#include <sys/lock.h> 55#include <sys/stat.h> 56#include <sys/acct.h> 57#include <sys/namei.h> 58#include <sys/ucred.h> 59#include <sys/buf.h> 60#include <sys/errno.h> 61#include <sys/malloc.h> 62#include <sys/mbuf.h> 63#include <sys/syscallargs.h> 64#include <sys/pool.h> 65#include <sys/tree.h> 66#include <sys/specdev.h> 67 68#include <netinet/in.h> 69 70#include <uvm/uvm_extern.h> 71#include <uvm/uvm_vnode.h> 72 73#include "softraid.h" 74 75void sr_shutdown(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; 117 118#ifdef DEBUG 119void printlockedvnodes(void); 120#endif 121 122struct pool vnode_pool; 123struct pool uvm_vnode_pool; 124 125static inline int rb_buf_compare(const struct buf *b1, const struct buf *b2); 126RBT_GENERATE(buf_rb_bufs, buf, b_rbbufs, rb_buf_compare); 127 128static inline int 129rb_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 */ 141void 142vntblinit(void) 143{ 144 /* buffer cache may need a vnode for each buffer */ 145 maxvnodes = 2 * initialvnodes; 146 pool_init(&vnode_pool, sizeof(struct vnode), 0, IPL_NONE, 147 PR_WAITOK, "vnodes", NULL); 148 pool_init(&uvm_vnode_pool, sizeof(struct uvm_vnode), 0, IPL_NONE, 149 PR_WAITOK, "uvmvnodes", NULL); 150 TAILQ_INIT(&vnode_hold_list); 151 TAILQ_INIT(&vnode_free_list); 152 TAILQ_INIT(&mountlist); 153 /* 154 * Initialize the filesystem syncer. 155 */ 156 vn_initialize_syncerd(); 157 158#ifdef NFSSERVER 159 rn_init(sizeof(struct sockaddr_in)); 160#endif /* NFSSERVER */ 161} 162 163/* 164 * Mark a mount point as busy. Used to synchronize access and to delay 165 * unmounting. 166 * 167 * Default behaviour is to attempt getting a READ lock and in case of an 168 * ongoing unmount, to wait for it to finish and then return failure. 169 */ 170int 171vfs_busy(struct mount *mp, int flags) 172{ 173 int rwflags = 0; 174 175 /* new mountpoints need their lock initialised */ 176 if (mp->mnt_lock.rwl_name == NULL) 177 rw_init_flags(&mp->mnt_lock, "vfslock", RWL_IS_VNODE); 178 179 if (flags & VB_WRITE) 180 rwflags |= RW_WRITE; 181 else 182 rwflags |= RW_READ; 183 184 if (flags & VB_WAIT) 185 rwflags |= RW_SLEEPFAIL; 186 else 187 rwflags |= RW_NOSLEEP; 188 189 if (rw_enter(&mp->mnt_lock, rwflags)) 190 return (EBUSY); 191 192 return (0); 193} 194 195/* 196 * Free a busy file system 197 */ 198void 199vfs_unbusy(struct mount *mp) 200{ 201 rw_exit(&mp->mnt_lock); 202} 203 204int 205vfs_isbusy(struct mount *mp) 206{ 207 if (RWLOCK_OWNER(&mp->mnt_lock) > 0) 208 return (1); 209 else 210 return (0); 211} 212 213/* 214 * Lookup a filesystem type, and if found allocate and initialize 215 * a mount structure for it. 216 * 217 * Devname is usually updated by mount(8) after booting. 218 */ 219int 220vfs_rootmountalloc(char *fstypename, char *devname, struct mount **mpp) 221{ 222 struct vfsconf *vfsp; 223 struct mount *mp; 224 225 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 226 if (!strcmp(vfsp->vfc_name, fstypename)) 227 break; 228 if (vfsp == NULL) 229 return (ENODEV); 230 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 231 (void)vfs_busy(mp, VB_READ|VB_NOWAIT); 232 LIST_INIT(&mp->mnt_vnodelist); 233 mp->mnt_vfc = vfsp; 234 mp->mnt_op = vfsp->vfc_vfsops; 235 mp->mnt_flag = MNT_RDONLY; 236 mp->mnt_vnodecovered = NULLVP; 237 vfsp->vfc_refcount++; 238 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 239 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 240 mp->mnt_stat.f_mntonname[0] = '/'; 241 copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN, 0); 242 copystr(devname, mp->mnt_stat.f_mntfromspec, MNAMELEN, 0); 243 *mpp = mp; 244 return (0); 245 } 246 247/* 248 * Lookup a mount point by filesystem identifier. 249 */ 250struct mount * 251vfs_getvfs(fsid_t *fsid) 252{ 253 struct mount *mp; 254 255 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 256 if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && 257 mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { 258 return (mp); 259 } 260 } 261 262 return (NULL); 263} 264 265 266/* 267 * Get a new unique fsid 268 */ 269void 270vfs_getnewfsid(struct mount *mp) 271{ 272 static u_short xxxfs_mntid; 273 274 fsid_t tfsid; 275 int mtype; 276 277 mtype = mp->mnt_vfc->vfc_typenum; 278 mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); 279 mp->mnt_stat.f_fsid.val[1] = mtype; 280 if (xxxfs_mntid == 0) 281 ++xxxfs_mntid; 282 tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); 283 tfsid.val[1] = mtype; 284 if (!TAILQ_EMPTY(&mountlist)) { 285 while (vfs_getvfs(&tfsid)) { 286 tfsid.val[0]++; 287 xxxfs_mntid++; 288 } 289 } 290 mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; 291} 292 293/* 294 * Set vnode attributes to VNOVAL 295 */ 296void 297vattr_null(struct vattr *vap) 298{ 299 300 vap->va_type = VNON; 301 /* 302 * Don't get fancy: u_quad_t = u_int = VNOVAL leaves the u_quad_t 303 * with 2^31-1 instead of 2^64-1. Just write'm out and let 304 * the compiler do its job. 305 */ 306 vap->va_mode = VNOVAL; 307 vap->va_nlink = VNOVAL; 308 vap->va_uid = VNOVAL; 309 vap->va_gid = VNOVAL; 310 vap->va_fsid = VNOVAL; 311 vap->va_fileid = VNOVAL; 312 vap->va_size = VNOVAL; 313 vap->va_blocksize = VNOVAL; 314 vap->va_atime.tv_sec = VNOVAL; 315 vap->va_atime.tv_nsec = VNOVAL; 316 vap->va_mtime.tv_sec = VNOVAL; 317 vap->va_mtime.tv_nsec = VNOVAL; 318 vap->va_ctime.tv_sec = VNOVAL; 319 vap->va_ctime.tv_nsec = VNOVAL; 320 vap->va_gen = VNOVAL; 321 vap->va_flags = VNOVAL; 322 vap->va_rdev = VNOVAL; 323 vap->va_bytes = VNOVAL; 324 vap->va_filerev = VNOVAL; 325 vap->va_vaflags = 0; 326} 327 328/* 329 * Routines having to do with the management of the vnode table. 330 */ 331long numvnodes; 332 333/* 334 * Return the next vnode from the free list. 335 */ 336int 337getnewvnode(enum vtagtype tag, struct mount *mp, struct vops *vops, 338 struct vnode **vpp) 339{ 340 struct proc *p = curproc; 341 struct freelst *listhd; 342 static int toggle; 343 struct vnode *vp; 344 int s; 345 346 /* 347 * allow maxvnodes to increase if the buffer cache itself 348 * is big enough to justify it. (we don't shrink it ever) 349 */ 350 maxvnodes = maxvnodes < bcstats.numbufs ? bcstats.numbufs 351 : maxvnodes; 352 353 /* 354 * We must choose whether to allocate a new vnode or recycle an 355 * existing one. The criterion for allocating a new one is that 356 * the total number of vnodes is less than the number desired or 357 * there are no vnodes on either free list. Generally we only 358 * want to recycle vnodes that have no buffers associated with 359 * them, so we look first on the vnode_free_list. If it is empty, 360 * we next consider vnodes with referencing buffers on the 361 * vnode_hold_list. The toggle ensures that half the time we 362 * will use a buffer from the vnode_hold_list, and half the time 363 * we will allocate a new one unless the list has grown to twice 364 * the desired size. We are reticent to recycle vnodes from the 365 * vnode_hold_list because we will lose the identity of all its 366 * referencing buffers. 367 */ 368 toggle ^= 1; 369 if (numvnodes / 2 > maxvnodes) 370 toggle = 0; 371 372 s = splbio(); 373 if ((numvnodes < maxvnodes) || 374 ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && 375 ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { 376 splx(s); 377 vp = pool_get(&vnode_pool, PR_WAITOK | PR_ZERO); 378 vp->v_uvm = pool_get(&uvm_vnode_pool, PR_WAITOK | PR_ZERO); 379 vp->v_uvm->u_vnode = vp; 380 RBT_INIT(buf_rb_bufs, &vp->v_bufs_tree); 381 cache_tree_init(&vp->v_nc_tree); 382 TAILQ_INIT(&vp->v_cache_dst); 383 numvnodes++; 384 } else { 385 TAILQ_FOREACH(vp, listhd, v_freelist) { 386 if (VOP_ISLOCKED(vp) == 0) 387 break; 388 } 389 /* 390 * Unless this is a bad time of the month, at most 391 * the first NCPUS items on the free list are 392 * locked, so this is close enough to being empty. 393 */ 394 if (vp == NULL) { 395 splx(s); 396 tablefull("vnode"); 397 *vpp = 0; 398 return (ENFILE); 399 } 400 401#ifdef DIAGNOSTIC 402 if (vp->v_usecount) { 403 vprint("free vnode", vp); 404 panic("free vnode isn't"); 405 } 406#endif 407 408 TAILQ_REMOVE(listhd, vp, v_freelist); 409 vp->v_bioflag &= ~VBIOONFREELIST; 410 splx(s); 411 412 if (vp->v_type != VBAD) 413 vgonel(vp, p); 414#ifdef DIAGNOSTIC 415 if (vp->v_data) { 416 vprint("cleaned vnode", vp); 417 panic("cleaned vnode isn't"); 418 } 419 s = splbio(); 420 if (vp->v_numoutput) 421 panic("Clean vnode has pending I/O's"); 422 splx(s); 423#endif 424 vp->v_flag = 0; 425 vp->v_socket = 0; 426 } 427 cache_purge(vp); 428 vp->v_type = VNON; 429 vp->v_tag = tag; 430 vp->v_op = vops; 431 insmntque(vp, mp); 432 *vpp = vp; 433 vp->v_usecount = 1; 434 vp->v_data = 0; 435 return (0); 436} 437 438/* 439 * Move a vnode from one mount queue to another. 440 */ 441void 442insmntque(struct vnode *vp, struct mount *mp) 443{ 444 /* 445 * Delete from old mount point vnode list, if on one. 446 */ 447 if (vp->v_mount != NULL) 448 LIST_REMOVE(vp, v_mntvnodes); 449 /* 450 * Insert into list of vnodes for the new mount point, if available. 451 */ 452 if ((vp->v_mount = mp) != NULL) 453 LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); 454} 455 456/* 457 * Create a vnode for a block device. 458 * Used for root filesystem, argdev, and swap areas. 459 * Also used for memory file system special devices. 460 */ 461int 462bdevvp(dev_t dev, struct vnode **vpp) 463{ 464 return (getdevvp(dev, vpp, VBLK)); 465} 466 467/* 468 * Create a vnode for a character device. 469 * Used for console handling. 470 */ 471int 472cdevvp(dev_t dev, struct vnode **vpp) 473{ 474 return (getdevvp(dev, vpp, VCHR)); 475} 476 477/* 478 * Create a vnode for a device. 479 * Used by bdevvp (block device) for root file system etc., 480 * and by cdevvp (character device) for console. 481 */ 482int 483getdevvp(dev_t dev, struct vnode **vpp, enum vtype type) 484{ 485 struct vnode *vp; 486 struct vnode *nvp; 487 int error; 488 489 if (dev == NODEV) { 490 *vpp = NULLVP; 491 return (0); 492 } 493 error = getnewvnode(VT_NON, NULL, &spec_vops, &nvp); 494 if (error) { 495 *vpp = NULLVP; 496 return (error); 497 } 498 vp = nvp; 499 vp->v_type = type; 500 if ((nvp = checkalias(vp, dev, NULL)) != 0) { 501 vput(vp); 502 vp = nvp; 503 } 504 if (vp->v_type == VCHR && cdevsw[major(vp->v_rdev)].d_type == D_TTY) 505 vp->v_flag |= VISTTY; 506 *vpp = vp; 507 return (0); 508} 509 510/* 511 * Check to see if the new vnode represents a special device 512 * for which we already have a vnode (either because of 513 * bdevvp() or because of a different vnode representing 514 * the same block device). If such an alias exists, deallocate 515 * the existing contents and return the aliased vnode. The 516 * caller is responsible for filling it with its new contents. 517 */ 518struct vnode * 519checkalias(struct vnode *nvp, dev_t nvp_rdev, struct mount *mp) 520{ 521 struct proc *p = curproc; 522 struct vnode *vp; 523 struct vnode **vpp; 524 525 if (nvp->v_type != VBLK && nvp->v_type != VCHR) 526 return (NULLVP); 527 528 vpp = &speclisth[SPECHASH(nvp_rdev)]; 529loop: 530 for (vp = *vpp; vp; vp = vp->v_specnext) { 531 if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { 532 continue; 533 } 534 /* 535 * Alias, but not in use, so flush it out. 536 */ 537 if (vp->v_usecount == 0) { 538 vgonel(vp, p); 539 goto loop; 540 } 541 if (vget(vp, LK_EXCLUSIVE, p)) { 542 goto loop; 543 } 544 break; 545 } 546 547 /* 548 * Common case is actually in the if statement 549 */ 550 if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { 551 nvp->v_specinfo = malloc(sizeof(struct specinfo), M_VNODE, 552 M_WAITOK); 553 nvp->v_rdev = nvp_rdev; 554 nvp->v_hashchain = vpp; 555 nvp->v_specnext = *vpp; 556 nvp->v_specmountpoint = NULL; 557 nvp->v_speclockf = NULL; 558 nvp->v_specbitmap = NULL; 559 if (nvp->v_type == VCHR && 560 (cdevsw[major(nvp_rdev)].d_flags & D_CLONE) && 561 (minor(nvp_rdev) >> CLONE_SHIFT == 0)) { 562 if (vp != NULLVP) 563 nvp->v_specbitmap = vp->v_specbitmap; 564 else 565 nvp->v_specbitmap = malloc(CLONE_MAPSZ, 566 M_VNODE, M_WAITOK | M_ZERO); 567 } 568 *vpp = nvp; 569 if (vp != NULLVP) { 570 nvp->v_flag |= VALIASED; 571 vp->v_flag |= VALIASED; 572 vput(vp); 573 } 574 return (NULLVP); 575 } 576 577 /* 578 * This code is the uncommon case. It is called in case 579 * we found an alias that was VT_NON && vtype of VBLK 580 * This means we found a block device that was created 581 * using bdevvp. 582 * An example of such a vnode is the root partition device vnode 583 * created in ffs_mountroot. 584 * 585 * The vnodes created by bdevvp should not be aliased (why?). 586 */ 587 588 VOP_UNLOCK(vp, p); 589 vclean(vp, 0, p); 590 vp->v_op = nvp->v_op; 591 vp->v_tag = nvp->v_tag; 592 nvp->v_type = VNON; 593 insmntque(vp, mp); 594 return (vp); 595} 596 597/* 598 * Grab a particular vnode from the free list, increment its 599 * reference count and lock it. If the vnode lock bit is set, 600 * the vnode is being eliminated in vgone. In that case, we 601 * cannot grab it, so the process is awakened when the 602 * transition is completed, and an error code is returned to 603 * indicate that the vnode is no longer usable, possibly 604 * having been changed to a new file system type. 605 */ 606int 607vget(struct vnode *vp, int flags, struct proc *p) 608{ 609 int error, s, onfreelist; 610 611 /* 612 * If the vnode is in the process of being cleaned out for 613 * another use, we wait for the cleaning to finish and then 614 * return failure. Cleaning is determined by checking that 615 * the VXLOCK flag is set. 616 */ 617 618 if (vp->v_flag & VXLOCK) { 619 if (flags & LK_NOWAIT) { 620 return (EBUSY); 621 } 622 623 vp->v_flag |= VXWANT; 624 tsleep(vp, PINOD, "vget", 0); 625 return (ENOENT); 626 } 627 628 onfreelist = vp->v_bioflag & VBIOONFREELIST; 629 if (vp->v_usecount == 0 && onfreelist) { 630 s = splbio(); 631 if (vp->v_holdcnt > 0) 632 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 633 else 634 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 635 vp->v_bioflag &= ~VBIOONFREELIST; 636 splx(s); 637 } 638 639 vp->v_usecount++; 640 if (flags & LK_TYPE_MASK) { 641 if ((error = vn_lock(vp, flags, p)) != 0) { 642 vp->v_usecount--; 643 if (vp->v_usecount == 0 && onfreelist) 644 vputonfreelist(vp); 645 } 646 return (error); 647 } 648 649 return (0); 650} 651 652 653/* Vnode reference. */ 654void 655vref(struct vnode *vp) 656{ 657#ifdef DIAGNOSTIC 658 if (vp->v_usecount == 0) 659 panic("vref used where vget required"); 660 if (vp->v_type == VNON) 661 panic("vref on a VNON vnode"); 662#endif 663 vp->v_usecount++; 664} 665 666void 667vputonfreelist(struct vnode *vp) 668{ 669 int s; 670 struct freelst *lst; 671 672 s = splbio(); 673#ifdef DIAGNOSTIC 674 if (vp->v_usecount != 0) 675 panic("Use count is not zero!"); 676 677 if (vp->v_bioflag & VBIOONFREELIST) { 678 vprint("vnode already on free list: ", vp); 679 panic("vnode already on free list"); 680 } 681#endif 682 683 vp->v_bioflag |= VBIOONFREELIST; 684 685 if (vp->v_holdcnt > 0) 686 lst = &vnode_hold_list; 687 else 688 lst = &vnode_free_list; 689 690 if (vp->v_type == VBAD) 691 TAILQ_INSERT_HEAD(lst, vp, v_freelist); 692 else 693 TAILQ_INSERT_TAIL(lst, vp, v_freelist); 694 695 splx(s); 696} 697 698/* 699 * vput(), just unlock and vrele() 700 */ 701void 702vput(struct vnode *vp) 703{ 704 struct proc *p = curproc; 705 706#ifdef DIAGNOSTIC 707 if (vp == NULL) 708 panic("vput: null vp"); 709#endif 710 711#ifdef DIAGNOSTIC 712 if (vp->v_usecount == 0) { 713 vprint("vput: bad ref count", vp); 714 panic("vput: ref cnt"); 715 } 716#endif 717 vp->v_usecount--; 718 if (vp->v_usecount > 0) { 719 VOP_UNLOCK(vp, p); 720 return; 721 } 722 723#ifdef DIAGNOSTIC 724 if (vp->v_writecount != 0) { 725 vprint("vput: bad writecount", vp); 726 panic("vput: v_writecount != 0"); 727 } 728#endif 729 730 VOP_INACTIVE(vp, p); 731 732 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 733 vputonfreelist(vp); 734} 735 736/* 737 * Vnode release - use for active VNODES. 738 * If count drops to zero, call inactive routine and return to freelist. 739 * Returns 0 if it did not sleep. 740 */ 741int 742vrele(struct vnode *vp) 743{ 744 struct proc *p = curproc; 745 746#ifdef DIAGNOSTIC 747 if (vp == NULL) 748 panic("vrele: null vp"); 749#endif 750#ifdef DIAGNOSTIC 751 if (vp->v_usecount == 0) { 752 vprint("vrele: bad ref count", vp); 753 panic("vrele: ref cnt"); 754 } 755#endif 756 vp->v_usecount--; 757 if (vp->v_usecount > 0) { 758 return (0); 759 } 760 761#ifdef DIAGNOSTIC 762 if (vp->v_writecount != 0) { 763 vprint("vrele: bad writecount", vp); 764 panic("vrele: v_writecount != 0"); 765 } 766#endif 767 768 if (vn_lock(vp, LK_EXCLUSIVE, p)) { 769#ifdef DIAGNOSTIC 770 vprint("vrele: cannot lock", vp); 771#endif 772 return (1); 773 } 774 775 VOP_INACTIVE(vp, p); 776 777 if (vp->v_usecount == 0 && !(vp->v_bioflag & VBIOONFREELIST)) 778 vputonfreelist(vp); 779 return (1); 780} 781 782/* Page or buffer structure gets a reference. */ 783void 784vhold(struct vnode *vp) 785{ 786 /* 787 * If it is on the freelist and the hold count is currently 788 * zero, move it to the hold list. 789 */ 790 if ((vp->v_bioflag & VBIOONFREELIST) && 791 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 792 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 793 TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); 794 } 795 vp->v_holdcnt++; 796} 797 798/* Lose interest in a vnode. */ 799void 800vdrop(struct vnode *vp) 801{ 802#ifdef DIAGNOSTIC 803 if (vp->v_holdcnt == 0) 804 panic("vdrop: zero holdcnt"); 805#endif 806 807 vp->v_holdcnt--; 808 809 /* 810 * If it is on the holdlist and the hold count drops to 811 * zero, move it to the free list. 812 */ 813 if ((vp->v_bioflag & VBIOONFREELIST) && 814 vp->v_holdcnt == 0 && vp->v_usecount == 0) { 815 TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); 816 TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); 817 } 818} 819 820/* 821 * Remove any vnodes in the vnode table belonging to mount point mp. 822 * 823 * If MNT_NOFORCE is specified, there should not be any active ones, 824 * return error if any are found (nb: this is a user error, not a 825 * system error). If MNT_FORCE is specified, detach any active vnodes 826 * that are found. 827 */ 828#ifdef DEBUG 829int busyprt = 0; /* print out busy vnodes */ 830struct ctldebug debug1 = { "busyprt", &busyprt }; 831#endif 832 833int 834vfs_mount_foreach_vnode(struct mount *mp, 835 int (*func)(struct vnode *, void *), void *arg) { 836 struct vnode *vp, *nvp; 837 int error = 0; 838 839loop: 840 LIST_FOREACH_SAFE(vp , &mp->mnt_vnodelist, v_mntvnodes, nvp) { 841 if (vp->v_mount != mp) 842 goto loop; 843 844 error = func(vp, arg); 845 846 if (error != 0) 847 break; 848 } 849 850 return (error); 851} 852 853struct vflush_args { 854 struct vnode *skipvp; 855 int busy; 856 int flags; 857}; 858 859int 860vflush_vnode(struct vnode *vp, void *arg) 861{ 862 struct vflush_args *va = arg; 863 struct proc *p = curproc; 864 865 if (vp == va->skipvp) { 866 return (0); 867 } 868 869 if ((va->flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { 870 return (0); 871 } 872 873 /* 874 * If WRITECLOSE is set, only flush out regular file 875 * vnodes open for writing. 876 */ 877 if ((va->flags & WRITECLOSE) && 878 (vp->v_writecount == 0 || vp->v_type != VREG)) { 879 return (0); 880 } 881 882 /* 883 * With v_usecount == 0, all we need to do is clear 884 * out the vnode data structures and we are done. 885 */ 886 if (vp->v_usecount == 0) { 887 vgonel(vp, p); 888 return (0); 889 } 890 891 /* 892 * If FORCECLOSE is set, forcibly close the vnode. 893 * For block or character devices, revert to an 894 * anonymous device. For all other files, just kill them. 895 */ 896 if (va->flags & FORCECLOSE) { 897 if (vp->v_type != VBLK && vp->v_type != VCHR) { 898 vgonel(vp, p); 899 } else { 900 vclean(vp, 0, p); 901 vp->v_op = &spec_vops; 902 insmntque(vp, NULL); 903 } 904 return (0); 905 } 906 907 if (va->flags & WRITEDEMOTE) { 908 vp->v_op = &dead_vops; 909 vp->v_tag = VT_NON; 910 return (0); 911 } 912 913#ifdef DEBUG 914 if (busyprt) 915 vprint("vflush: busy vnode", vp); 916#endif 917 va->busy++; 918 return (0); 919} 920 921int 922vflush(struct mount *mp, struct vnode *skipvp, int flags) 923{ 924 struct vflush_args va; 925 va.skipvp = skipvp; 926 va.busy = 0; 927 va.flags = flags; 928 929 vfs_mount_foreach_vnode(mp, vflush_vnode, &va); 930 931 if (va.busy) 932 return (EBUSY); 933 return (0); 934} 935 936/* 937 * Disassociate the underlying file system from a vnode. 938 */ 939void 940vclean(struct vnode *vp, int flags, struct proc *p) 941{ 942 int active; 943 944 /* 945 * Check to see if the vnode is in use. 946 * If so we have to reference it before we clean it out 947 * so that its count cannot fall to zero and generate a 948 * race against ourselves to recycle it. 949 */ 950 if ((active = vp->v_usecount) != 0) 951 vp->v_usecount++; 952 953 /* 954 * Prevent the vnode from being recycled or 955 * brought into use while we clean it out. 956 */ 957 if (vp->v_flag & VXLOCK) 958 panic("vclean: deadlock"); 959 vp->v_flag |= VXLOCK; 960 /* 961 * Even if the count is zero, the VOP_INACTIVE routine may still 962 * have the object locked while it cleans it out. The VOP_LOCK 963 * ensures that the VOP_INACTIVE routine is done with its work. 964 * For active vnodes, it ensures that no other activity can 965 * occur while the underlying object is being cleaned out. 966 */ 967 VOP_LOCK(vp, LK_DRAIN | LK_EXCLUSIVE, p); 968 969 /* 970 * Clean out any VM data associated with the vnode. 971 */ 972 uvm_vnp_terminate(vp); 973 /* 974 * Clean out any buffers associated with the vnode. 975 */ 976 if (flags & DOCLOSE) 977 vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); 978 /* 979 * If purging an active vnode, it must be closed and 980 * deactivated before being reclaimed. Note that the 981 * VOP_INACTIVE will unlock the vnode 982 */ 983 if (active) { 984 if (flags & DOCLOSE) 985 VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); 986 VOP_INACTIVE(vp, p); 987 } else { 988 /* 989 * Any other processes trying to obtain this lock must first 990 * wait for VXLOCK to clear, then call the new lock operation. 991 */ 992 VOP_UNLOCK(vp, p); 993 } 994 995 /* 996 * Reclaim the vnode. 997 */ 998 if (VOP_RECLAIM(vp, p)) 999 panic("vclean: cannot reclaim"); 1000 if (active) { 1001 vp->v_usecount--; 1002 if (vp->v_usecount == 0) { 1003 if (vp->v_holdcnt > 0) 1004 panic("vclean: not clean"); 1005 vputonfreelist(vp); 1006 } 1007 } 1008 cache_purge(vp); 1009 1010 /* 1011 * Done with purge, notify sleepers of the grim news. 1012 */ 1013 vp->v_op = &dead_vops; 1014 VN_KNOTE(vp, NOTE_REVOKE); 1015 vp->v_tag = VT_NON; 1016 vp->v_flag &= ~VXLOCK; 1017#ifdef VFSLCKDEBUG 1018 vp->v_flag &= ~VLOCKSWORK; 1019#endif 1020 if (vp->v_flag & VXWANT) { 1021 vp->v_flag &= ~VXWANT; 1022 wakeup(vp); 1023 } 1024} 1025 1026/* 1027 * Recycle an unused vnode to the front of the free list. 1028 */ 1029int 1030vrecycle(struct vnode *vp, struct proc *p) 1031{ 1032 if (vp->v_usecount == 0) { 1033 vgonel(vp, p); 1034 return (1); 1035 } 1036 return (0); 1037} 1038 1039/* 1040 * Eliminate all activity associated with a vnode 1041 * in preparation for reuse. 1042 */ 1043void 1044vgone(struct vnode *vp) 1045{ 1046 struct proc *p = curproc; 1047 vgonel(vp, p); 1048} 1049 1050/* 1051 * vgone, with struct proc. 1052 */ 1053void 1054vgonel(struct vnode *vp, struct proc *p) 1055{ 1056 struct vnode *vq; 1057 struct vnode *vx; 1058 1059 /* 1060 * If a vgone (or vclean) is already in progress, 1061 * wait until it is done and return. 1062 */ 1063 if (vp->v_flag & VXLOCK) { 1064 vp->v_flag |= VXWANT; 1065 tsleep(vp, PINOD, "vgone", 0); 1066 return; 1067 } 1068 1069 /* 1070 * Clean out the filesystem specific data. 1071 */ 1072 vclean(vp, DOCLOSE, p); 1073 /* 1074 * Delete from old mount point vnode list, if on one. 1075 */ 1076 if (vp->v_mount != NULL) 1077 insmntque(vp, NULL); 1078 /* 1079 * If special device, remove it from special device alias list 1080 * if it is on one. 1081 */ 1082 if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { 1083 if ((vp->v_flag & VALIASED) == 0 && vp->v_type == VCHR && 1084 (cdevsw[major(vp->v_rdev)].d_flags & D_CLONE) && 1085 (minor(vp->v_rdev) >> CLONE_SHIFT == 0)) { 1086 free(vp->v_specbitmap, M_VNODE, CLONE_MAPSZ); 1087 } 1088 if (*vp->v_hashchain == vp) { 1089 *vp->v_hashchain = vp->v_specnext; 1090 } else { 1091 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1092 if (vq->v_specnext != vp) 1093 continue; 1094 vq->v_specnext = vp->v_specnext; 1095 break; 1096 } 1097 if (vq == NULL) 1098 panic("missing bdev"); 1099 } 1100 if (vp->v_flag & VALIASED) { 1101 vx = NULL; 1102 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1103 if (vq->v_rdev != vp->v_rdev || 1104 vq->v_type != vp->v_type) 1105 continue; 1106 if (vx) 1107 break; 1108 vx = vq; 1109 } 1110 if (vx == NULL) 1111 panic("missing alias"); 1112 if (vq == NULL) 1113 vx->v_flag &= ~VALIASED; 1114 vp->v_flag &= ~VALIASED; 1115 } 1116 free(vp->v_specinfo, M_VNODE, sizeof(struct specinfo)); 1117 vp->v_specinfo = NULL; 1118 } 1119 /* 1120 * If it is on the freelist and not already at the head, 1121 * move it to the head of the list. 1122 */ 1123 vp->v_type = VBAD; 1124 1125 /* 1126 * Move onto the free list, unless we were called from 1127 * getnewvnode and we're not on any free list 1128 */ 1129 if (vp->v_usecount == 0 && 1130 (vp->v_bioflag & VBIOONFREELIST)) { 1131 int s; 1132 1133 s = splbio(); 1134 1135 if (vp->v_holdcnt > 0) 1136 panic("vgonel: not clean"); 1137 1138 if (TAILQ_FIRST(&vnode_free_list) != vp) { 1139 TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); 1140 TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); 1141 } 1142 splx(s); 1143 } 1144} 1145 1146/* 1147 * Lookup a vnode by device number. 1148 */ 1149int 1150vfinddev(dev_t dev, enum vtype type, struct vnode **vpp) 1151{ 1152 struct vnode *vp; 1153 int rc =0; 1154 1155 for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { 1156 if (dev != vp->v_rdev || type != vp->v_type) 1157 continue; 1158 *vpp = vp; 1159 rc = 1; 1160 break; 1161 } 1162 return (rc); 1163} 1164 1165/* 1166 * Revoke all the vnodes corresponding to the specified minor number 1167 * range (endpoints inclusive) of the specified major. 1168 */ 1169void 1170vdevgone(int maj, int minl, int minh, enum vtype type) 1171{ 1172 struct vnode *vp; 1173 int mn; 1174 1175 for (mn = minl; mn <= minh; mn++) 1176 if (vfinddev(makedev(maj, mn), type, &vp)) 1177 VOP_REVOKE(vp, REVOKEALL); 1178} 1179 1180/* 1181 * Calculate the total number of references to a special device. 1182 */ 1183int 1184vcount(struct vnode *vp) 1185{ 1186 struct vnode *vq, *vnext; 1187 int count; 1188 1189loop: 1190 if ((vp->v_flag & VALIASED) == 0) 1191 return (vp->v_usecount); 1192 for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { 1193 vnext = vq->v_specnext; 1194 if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) 1195 continue; 1196 /* 1197 * Alias, but not in use, so flush it out. 1198 */ 1199 if (vq->v_usecount == 0 && vq != vp) { 1200 vgone(vq); 1201 goto loop; 1202 } 1203 count += vq->v_usecount; 1204 } 1205 return (count); 1206} 1207 1208#if defined(DEBUG) || defined(DIAGNOSTIC) 1209/* 1210 * Print out a description of a vnode. 1211 */ 1212static char *typename[] = 1213 { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; 1214 1215void 1216vprint(char *label, struct vnode *vp) 1217{ 1218 char buf[64]; 1219 1220 if (label != NULL) 1221 printf("%s: ", label); 1222 printf("%p, type %s, use %u, write %u, hold %u,", 1223 vp, typename[vp->v_type], vp->v_usecount, vp->v_writecount, 1224 vp->v_holdcnt); 1225 buf[0] = '\0'; 1226 if (vp->v_flag & VROOT) 1227 strlcat(buf, "|VROOT", sizeof buf); 1228 if (vp->v_flag & VTEXT) 1229 strlcat(buf, "|VTEXT", sizeof buf); 1230 if (vp->v_flag & VSYSTEM) 1231 strlcat(buf, "|VSYSTEM", sizeof buf); 1232 if (vp->v_flag & VXLOCK) 1233 strlcat(buf, "|VXLOCK", sizeof buf); 1234 if (vp->v_flag & VXWANT) 1235 strlcat(buf, "|VXWANT", sizeof buf); 1236 if (vp->v_bioflag & VBIOWAIT) 1237 strlcat(buf, "|VBIOWAIT", sizeof buf); 1238 if (vp->v_bioflag & VBIOONFREELIST) 1239 strlcat(buf, "|VBIOONFREELIST", sizeof buf); 1240 if (vp->v_bioflag & VBIOONSYNCLIST) 1241 strlcat(buf, "|VBIOONSYNCLIST", sizeof buf); 1242 if (vp->v_flag & VALIASED) 1243 strlcat(buf, "|VALIASED", sizeof buf); 1244 if (buf[0] != '\0') 1245 printf(" flags (%s)", &buf[1]); 1246 if (vp->v_data == NULL) { 1247 printf("\n"); 1248 } else { 1249 printf("\n\t"); 1250 VOP_PRINT(vp); 1251 } 1252} 1253#endif /* DEBUG || DIAGNOSTIC */ 1254 1255#ifdef DEBUG 1256/* 1257 * List all of the locked vnodes in the system. 1258 * Called when debugging the kernel. 1259 */ 1260void 1261printlockedvnodes(void) 1262{ 1263 struct mount *mp; 1264 struct vnode *vp; 1265 1266 printf("Locked vnodes\n"); 1267 1268 TAILQ_FOREACH(mp, &mountlist, mnt_list) { 1269 if (vfs_busy(mp, VB_READ|VB_NOWAIT)) 1270 continue; 1271 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 1272 if (VOP_ISLOCKED(vp)) 1273 vprint(NULL, vp); 1274 } 1275 vfs_unbusy(mp); 1276 } 1277 1278} 1279#endif 1280 1281/* 1282 * Top level filesystem related information gathering. 1283 */ 1284int 1285vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1286 size_t newlen, struct proc *p) 1287{ 1288 struct vfsconf *vfsp, *tmpvfsp; 1289 int ret; 1290 1291 /* all sysctl names at this level are at least name and field */ 1292 if (namelen < 2) 1293 return (ENOTDIR); /* overloaded */ 1294 1295 if (name[0] != VFS_GENERIC) { 1296 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1297 if (vfsp->vfc_typenum == name[0]) 1298 break; 1299 1300 if (vfsp == NULL || vfsp->vfc_vfsops->vfs_sysctl == NULL) 1301 return (EOPNOTSUPP); 1302 1303 return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, 1304 oldp, oldlenp, newp, newlen, p)); 1305 } 1306 1307 switch (name[1]) { 1308 case VFS_MAXTYPENUM: 1309 return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); 1310 1311 case VFS_CONF: 1312 if (namelen < 3) 1313 return (ENOTDIR); /* overloaded */ 1314 1315 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 1316 if (vfsp->vfc_typenum == name[2]) 1317 break; 1318 1319 if (vfsp == NULL) 1320 return (EOPNOTSUPP); 1321 1322 /* Make a copy, clear out kernel pointers */ 1323 tmpvfsp = malloc(sizeof(*tmpvfsp), M_TEMP, M_WAITOK|M_ZERO); 1324 memcpy(tmpvfsp, vfsp, sizeof(*tmpvfsp)); 1325 tmpvfsp->vfc_vfsops = NULL; 1326 tmpvfsp->vfc_next = NULL; 1327 1328 ret = sysctl_rdstruct(oldp, oldlenp, newp, tmpvfsp, 1329 sizeof(struct vfsconf)); 1330 1331 free(tmpvfsp, M_TEMP, sizeof(*tmpvfsp)); 1332 return (ret); 1333 case VFS_BCACHESTAT: /* buffer cache statistics */ 1334 ret = sysctl_rdstruct(oldp, oldlenp, newp, &bcstats, 1335 sizeof(struct bcachestats)); 1336 return(ret); 1337 } 1338 return (EOPNOTSUPP); 1339} 1340 1341/* 1342 * Check to see if a filesystem is mounted on a block device. 1343 */ 1344int 1345vfs_mountedon(struct vnode *vp) 1346{ 1347 struct vnode *vq; 1348 int error = 0; 1349 1350 if (vp->v_specmountpoint != NULL) 1351 return (EBUSY); 1352 if (vp->v_flag & VALIASED) { 1353 for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { 1354 if (vq->v_rdev != vp->v_rdev || 1355 vq->v_type != vp->v_type) 1356 continue; 1357 if (vq->v_specmountpoint != NULL) { 1358 error = EBUSY; 1359 break; 1360 } 1361 } 1362 } 1363 return (error); 1364} 1365 1366#ifdef NFSSERVER 1367/* 1368 * Build hash lists of net addresses and hang them off the mount point. 1369 * Called by vfs_export() to set up the lists of export addresses. 1370 */ 1371int 1372vfs_hang_addrlist(struct mount *mp, struct netexport *nep, 1373 struct export_args *argp) 1374{ 1375 struct netcred *np; 1376 struct radix_node_head *rnh; 1377 int nplen, i; 1378 struct radix_node *rn; 1379 struct sockaddr *saddr, *smask = 0; 1380 int error; 1381 1382 if (argp->ex_addrlen == 0) { 1383 if (mp->mnt_flag & MNT_DEFEXPORTED) 1384 return (EPERM); 1385 np = &nep->ne_defexported; 1386 /* fill in the kernel's ucred from userspace's xucred */ 1387 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1388 return (error); 1389 mp->mnt_flag |= MNT_DEFEXPORTED; 1390 goto finish; 1391 } 1392 if (argp->ex_addrlen > MLEN || argp->ex_masklen > MLEN || 1393 argp->ex_addrlen < 0 || argp->ex_masklen < 0) 1394 return (EINVAL); 1395 nplen = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; 1396 np = (struct netcred *)malloc(nplen, M_NETADDR, M_WAITOK|M_ZERO); 1397 saddr = (struct sockaddr *)(np + 1); 1398 error = copyin(argp->ex_addr, saddr, argp->ex_addrlen); 1399 if (error) 1400 goto out; 1401 if (saddr->sa_len > argp->ex_addrlen) 1402 saddr->sa_len = argp->ex_addrlen; 1403 if (argp->ex_masklen) { 1404 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); 1405 error = copyin(argp->ex_mask, smask, argp->ex_masklen); 1406 if (error) 1407 goto out; 1408 if (smask->sa_len > argp->ex_masklen) 1409 smask->sa_len = argp->ex_masklen; 1410 } 1411 /* fill in the kernel's ucred from userspace's xucred */ 1412 if ((error = crfromxucred(&np->netc_anon, &argp->ex_anon))) 1413 goto out; 1414 i = saddr->sa_family; 1415 switch (i) { 1416 case AF_INET: 1417 if ((rnh = nep->ne_rtable_inet) == NULL) { 1418 if (!rn_inithead((void **)&nep->ne_rtable_inet, 1419 offsetof(struct sockaddr_in, sin_addr))) { 1420 error = ENOBUFS; 1421 goto out; 1422 } 1423 rnh = nep->ne_rtable_inet; 1424 } 1425 break; 1426 default: 1427 error = EINVAL; 1428 goto out; 1429 } 1430 rn = rn_addroute(saddr, smask, rnh, np->netc_rnodes, 0); 1431 if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ 1432 error = EPERM; 1433 goto out; 1434 } 1435finish: 1436 np->netc_exflags = argp->ex_flags; 1437 return (0); 1438out: 1439 free(np, M_NETADDR, nplen); 1440 return (error); 1441} 1442 1443int 1444vfs_free_netcred(struct radix_node *rn, void *w, u_int id) 1445{ 1446 struct radix_node_head *rnh = (struct radix_node_head *)w; 1447 1448 rn_delete(rn->rn_key, rn->rn_mask, rnh, NULL); 1449 free(rn, M_NETADDR, 0); 1450 return (0); 1451} 1452 1453/* 1454 * Free the net address hash lists that are hanging off the mount points. 1455 */ 1456void 1457vfs_free_addrlist(struct netexport *nep) 1458{ 1459 struct radix_node_head *rnh; 1460 1461 if ((rnh = nep->ne_rtable_inet) != NULL) { 1462 rn_walktree(rnh, vfs_free_netcred, rnh); 1463 free(rnh, M_RTABLE, 0); 1464 nep->ne_rtable_inet = NULL; 1465 } 1466} 1467#endif /* NFSSERVER */ 1468 1469int 1470vfs_export(struct mount *mp, struct netexport *nep, struct export_args *argp) 1471{ 1472#ifdef NFSSERVER 1473 int error; 1474 1475 if (argp->ex_flags & MNT_DELEXPORT) { 1476 vfs_free_addrlist(nep); 1477 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); 1478 } 1479 if (argp->ex_flags & MNT_EXPORTED) { 1480 if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) 1481 return (error); 1482 mp->mnt_flag |= MNT_EXPORTED; 1483 } 1484 return (0); 1485#else 1486 return (ENOTSUP); 1487#endif /* NFSSERVER */ 1488} 1489 1490struct netcred * 1491vfs_export_lookup(struct mount *mp, struct netexport *nep, struct mbuf *nam) 1492{ 1493#ifdef NFSSERVER 1494 struct netcred *np; 1495 struct radix_node_head *rnh; 1496 struct sockaddr *saddr; 1497 1498 np = NULL; 1499 if (mp->mnt_flag & MNT_EXPORTED) { 1500 /* 1501 * Lookup in the export list first. 1502 */ 1503 if (nam != NULL) { 1504 saddr = mtod(nam, struct sockaddr *); 1505 switch(saddr->sa_family) { 1506 case AF_INET: 1507 rnh = nep->ne_rtable_inet; 1508 break; 1509 default: 1510 rnh = NULL; 1511 break; 1512 } 1513 if (rnh != NULL) 1514 np = (struct netcred *)rn_match(saddr, rnh); 1515 } 1516 /* 1517 * If no address match, use the default if it exists. 1518 */ 1519 if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) 1520 np = &nep->ne_defexported; 1521 } 1522 return (np); 1523#else 1524 return (NULL); 1525#endif /* NFSSERVER */ 1526} 1527 1528/* 1529 * Do the usual access checking. 1530 * file_mode, uid and gid are from the vnode in question, 1531 * while acc_mode and cred are from the VOP_ACCESS parameter list 1532 */ 1533int 1534vaccess(enum vtype type, mode_t file_mode, uid_t uid, gid_t gid, 1535 mode_t acc_mode, struct ucred *cred) 1536{ 1537 mode_t mask; 1538 1539 /* User id 0 always gets read/write access. */ 1540 if (cred->cr_uid == 0) { 1541 /* For VEXEC, at least one of the execute bits must be set. */ 1542 if ((acc_mode & VEXEC) && type != VDIR && 1543 (file_mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0) 1544 return EACCES; 1545 return 0; 1546 } 1547 1548 mask = 0; 1549 1550 /* Otherwise, check the owner. */ 1551 if (cred->cr_uid == uid) { 1552 if (acc_mode & VEXEC) 1553 mask |= S_IXUSR; 1554 if (acc_mode & VREAD) 1555 mask |= S_IRUSR; 1556 if (acc_mode & VWRITE) 1557 mask |= S_IWUSR; 1558 return (file_mode & mask) == mask ? 0 : EACCES; 1559 } 1560 1561 /* Otherwise, check the groups. */ 1562 if (groupmember(gid, cred)) { 1563 if (acc_mode & VEXEC) 1564 mask |= S_IXGRP; 1565 if (acc_mode & VREAD) 1566 mask |= S_IRGRP; 1567 if (acc_mode & VWRITE) 1568 mask |= S_IWGRP; 1569 return (file_mode & mask) == mask ? 0 : EACCES; 1570 } 1571 1572 /* Otherwise, check everyone else. */ 1573 if (acc_mode & VEXEC) 1574 mask |= S_IXOTH; 1575 if (acc_mode & VREAD) 1576 mask |= S_IROTH; 1577 if (acc_mode & VWRITE) 1578 mask |= S_IWOTH; 1579 return (file_mode & mask) == mask ? 0 : EACCES; 1580} 1581 1582int 1583vfs_readonly(struct mount *mp, struct proc *p) 1584{ 1585 int error; 1586 1587 error = vfs_busy(mp, VB_WRITE|VB_WAIT); 1588 if (error) { 1589 printf("%s: busy\n", mp->mnt_stat.f_mntonname); 1590 return (error); 1591 } 1592 uvm_vnp_sync(mp); 1593 error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p); 1594 if (error) { 1595 printf("%s: failed to sync\n", mp->mnt_stat.f_mntonname); 1596 vfs_unbusy(mp); 1597 return (error); 1598 } 1599 1600 mp->mnt_flag |= MNT_UPDATE | MNT_RDONLY; 1601 mp->mnt_flag &= ~MNT_SOFTDEP; 1602 error = VFS_MOUNT(mp, mp->mnt_stat.f_mntonname, NULL, NULL, curproc); 1603 if (error) { 1604 printf("%s: failed to remount rdonly, error %d\n", 1605 mp->mnt_stat.f_mntonname, error); 1606 vfs_unbusy(mp); 1607 return (error); 1608 } 1609 if (mp->mnt_syncer != NULL) 1610 vgone(mp->mnt_syncer); 1611 mp->mnt_syncer = NULL; 1612 vfs_unbusy(mp); 1613 return (error); 1614} 1615 1616/* 1617 * Read-only all file systems. 1618 * We traverse the list in reverse order under the assumption that doing so 1619 * will avoid needing to worry about dependencies. 1620 */ 1621void 1622vfs_rofs(struct proc *p) 1623{ 1624 struct mount *mp, *nmp; 1625 1626 TAILQ_FOREACH_REVERSE_SAFE(mp, &mountlist, mntlist, mnt_list, nmp) { 1627 /* XXX Here is a race, the next pointer is not locked. */ 1628 (void) vfs_readonly(mp, p); 1629 } 1630} 1631 1632/* 1633 * Sync and unmount file systems before shutting down. 1634 */ 1635void 1636vfs_shutdown(struct proc *p) 1637{ 1638#ifdef ACCOUNTING 1639 acct_shutdown(); 1640#endif 1641 1642 printf("syncing disks... "); 1643 1644 if (panicstr == 0) { 1645 /* Take all filesystems to read-only */ 1646 sys_sync(p, NULL, NULL); 1647 vfs_rofs(p); 1648 } 1649 1650 if (vfs_syncwait(p, 1)) 1651 printf("giving up\n"); 1652 else 1653 printf("done\n"); 1654 1655#if NSOFTRAID > 0 1656 sr_shutdown(); 1657#endif 1658} 1659 1660/* 1661 * perform sync() operation and wait for buffers to flush. 1662 */ 1663int 1664vfs_syncwait(struct proc *p, int verbose) 1665{ 1666 struct buf *bp; 1667 int iter, nbusy, dcount, s; 1668#ifdef MULTIPROCESSOR 1669 int hold_count; 1670#endif 1671 1672 sys_sync(p, NULL, NULL); 1673 1674 /* Wait for sync to finish. */ 1675 dcount = 10000; 1676 for (iter = 0; iter < 20; iter++) { 1677 nbusy = 0; 1678 LIST_FOREACH(bp, &bufhead, b_list) { 1679 if ((bp->b_flags & (B_BUSY|B_INVAL|B_READ)) == B_BUSY) 1680 nbusy++; 1681 /* 1682 * With soft updates, some buffers that are 1683 * written will be remarked as dirty until other 1684 * buffers are written. 1685 */ 1686 if (bp->b_flags & B_DELWRI) { 1687 s = splbio(); 1688 bremfree(bp); 1689 buf_acquire(bp); 1690 splx(s); 1691 nbusy++; 1692 bawrite(bp); 1693 if (dcount-- <= 0) { 1694 if (verbose) 1695 printf("softdep "); 1696 return 1; 1697 } 1698 } 1699 } 1700 if (nbusy == 0) 1701 break; 1702 if (verbose) 1703 printf("%d ", nbusy); 1704#ifdef MULTIPROCESSOR 1705 if (_kernel_lock_held()) 1706 hold_count = __mp_release_all(&kernel_lock); 1707 else 1708 hold_count = 0; 1709#endif 1710 DELAY(40000 * iter); 1711#ifdef MULTIPROCESSOR 1712 if (hold_count) 1713 __mp_acquire_count(&kernel_lock, hold_count); 1714#endif 1715 } 1716 1717 return nbusy; 1718} 1719 1720/* 1721 * posix file system related system variables. 1722 */ 1723int 1724fs_posix_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, 1725 void *newp, size_t newlen, struct proc *p) 1726{ 1727 /* all sysctl names at this level are terminal */ 1728 if (namelen != 1) 1729 return (ENOTDIR); 1730 1731 switch (name[0]) { 1732 case FS_POSIX_SETUID: 1733 if (newp && securelevel > 0) 1734 return (EPERM); 1735 return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); 1736 default: 1737 return (EOPNOTSUPP); 1738 } 1739 /* NOTREACHED */ 1740} 1741 1742/* 1743 * file system related system variables. 1744 */ 1745int 1746fs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1747 size_t newlen, struct proc *p) 1748{ 1749 sysctlfn *fn; 1750 1751 switch (name[0]) { 1752 case FS_POSIX: 1753 fn = fs_posix_sysctl; 1754 break; 1755 default: 1756 return (EOPNOTSUPP); 1757 } 1758 return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); 1759} 1760 1761 1762/* 1763 * Routines dealing with vnodes and buffers 1764 */ 1765 1766/* 1767 * Wait for all outstanding I/Os to complete 1768 * 1769 * Manipulates v_numoutput. Must be called at splbio() 1770 */ 1771int 1772vwaitforio(struct vnode *vp, int slpflag, char *wmesg, int timeo) 1773{ 1774 int error = 0; 1775 1776 splassert(IPL_BIO); 1777 1778 while (vp->v_numoutput) { 1779 vp->v_bioflag |= VBIOWAIT; 1780 error = tsleep(&vp->v_numoutput, 1781 slpflag | (PRIBIO + 1), wmesg, timeo); 1782 if (error) 1783 break; 1784 } 1785 1786 return (error); 1787} 1788 1789/* 1790 * Update outstanding I/O count and do wakeup if requested. 1791 * 1792 * Manipulates v_numoutput. Must be called at splbio() 1793 */ 1794void 1795vwakeup(struct vnode *vp) 1796{ 1797 splassert(IPL_BIO); 1798 1799 if (vp != NULL) { 1800 if (vp->v_numoutput-- == 0) 1801 panic("vwakeup: neg numoutput"); 1802 if ((vp->v_bioflag & VBIOWAIT) && vp->v_numoutput == 0) { 1803 vp->v_bioflag &= ~VBIOWAIT; 1804 wakeup(&vp->v_numoutput); 1805 } 1806 } 1807} 1808 1809/* 1810 * Flush out and invalidate all buffers associated with a vnode. 1811 * Called with the underlying object locked. 1812 */ 1813int 1814vinvalbuf(struct vnode *vp, int flags, struct ucred *cred, struct proc *p, 1815 int slpflag, int slptimeo) 1816{ 1817 struct buf *bp; 1818 struct buf *nbp, *blist; 1819 int s, error; 1820 1821#ifdef VFSLCKDEBUG 1822 if ((vp->v_flag & VLOCKSWORK) && !VOP_ISLOCKED(vp)) 1823 panic("vinvalbuf(): vp isn't locked"); 1824#endif 1825 1826 if (flags & V_SAVE) { 1827 s = splbio(); 1828 vwaitforio(vp, 0, "vinvalbuf", 0); 1829 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1830 splx(s); 1831 if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) 1832 return (error); 1833 s = splbio(); 1834 if (vp->v_numoutput > 0 || 1835 !LIST_EMPTY(&vp->v_dirtyblkhd)) 1836 panic("vinvalbuf: dirty bufs"); 1837 } 1838 splx(s); 1839 } 1840loop: 1841 s = splbio(); 1842 for (;;) { 1843 if ((blist = LIST_FIRST(&vp->v_cleanblkhd)) && 1844 (flags & V_SAVEMETA)) 1845 while (blist && blist->b_lblkno < 0) 1846 blist = LIST_NEXT(blist, b_vnbufs); 1847 if (blist == NULL && 1848 (blist = LIST_FIRST(&vp->v_dirtyblkhd)) && 1849 (flags & V_SAVEMETA)) 1850 while (blist && blist->b_lblkno < 0) 1851 blist = LIST_NEXT(blist, b_vnbufs); 1852 if (!blist) 1853 break; 1854 1855 for (bp = blist; bp; bp = nbp) { 1856 nbp = LIST_NEXT(bp, b_vnbufs); 1857 if (flags & V_SAVEMETA && bp->b_lblkno < 0) 1858 continue; 1859 if (bp->b_flags & B_BUSY) { 1860 bp->b_flags |= B_WANTED; 1861 error = tsleep(bp, slpflag | (PRIBIO + 1), 1862 "vinvalbuf", slptimeo); 1863 if (error) { 1864 splx(s); 1865 return (error); 1866 } 1867 break; 1868 } 1869 bremfree(bp); 1870 /* 1871 * XXX Since there are no node locks for NFS, I believe 1872 * there is a slight chance that a delayed write will 1873 * occur while sleeping just above, so check for it. 1874 */ 1875 if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { 1876 buf_acquire(bp); 1877 splx(s); 1878 (void) VOP_BWRITE(bp); 1879 goto loop; 1880 } 1881 buf_acquire_nomap(bp); 1882 bp->b_flags |= B_INVAL; 1883 brelse(bp); 1884 } 1885 } 1886 if (!(flags & V_SAVEMETA) && 1887 (!LIST_EMPTY(&vp->v_dirtyblkhd) || !LIST_EMPTY(&vp->v_cleanblkhd))) 1888 panic("vinvalbuf: flush failed"); 1889 splx(s); 1890 return (0); 1891} 1892 1893void 1894vflushbuf(struct vnode *vp, int sync) 1895{ 1896 struct buf *bp, *nbp; 1897 int s; 1898 1899loop: 1900 s = splbio(); 1901 LIST_FOREACH_SAFE(bp, &vp->v_dirtyblkhd, b_vnbufs, nbp) { 1902 if ((bp->b_flags & B_BUSY)) 1903 continue; 1904 if ((bp->b_flags & B_DELWRI) == 0) 1905 panic("vflushbuf: not dirty"); 1906 bremfree(bp); 1907 buf_acquire(bp); 1908 splx(s); 1909 /* 1910 * Wait for I/O associated with indirect blocks to complete, 1911 * since there is no way to quickly wait for them below. 1912 */ 1913 if (bp->b_vp == vp || sync == 0) 1914 (void) bawrite(bp); 1915 else 1916 (void) bwrite(bp); 1917 goto loop; 1918 } 1919 if (sync == 0) { 1920 splx(s); 1921 return; 1922 } 1923 vwaitforio(vp, 0, "vflushbuf", 0); 1924 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 1925 splx(s); 1926#ifdef DIAGNOSTIC 1927 vprint("vflushbuf: dirty", vp); 1928#endif 1929 goto loop; 1930 } 1931 splx(s); 1932} 1933 1934/* 1935 * Associate a buffer with a vnode. 1936 * 1937 * Manipulates buffer vnode queues. Must be called at splbio(). 1938 */ 1939void 1940bgetvp(struct vnode *vp, struct buf *bp) 1941{ 1942 splassert(IPL_BIO); 1943 1944 1945 if (bp->b_vp) 1946 panic("bgetvp: not free"); 1947 vhold(vp); 1948 bp->b_vp = vp; 1949 if (vp->v_type == VBLK || vp->v_type == VCHR) 1950 bp->b_dev = vp->v_rdev; 1951 else 1952 bp->b_dev = NODEV; 1953 /* 1954 * Insert onto list for new vnode. 1955 */ 1956 bufinsvn(bp, &vp->v_cleanblkhd); 1957} 1958 1959/* 1960 * Disassociate a buffer from a vnode. 1961 * 1962 * Manipulates vnode buffer queues. Must be called at splbio(). 1963 */ 1964void 1965brelvp(struct buf *bp) 1966{ 1967 struct vnode *vp; 1968 1969 splassert(IPL_BIO); 1970 1971 if ((vp = bp->b_vp) == (struct vnode *) 0) 1972 panic("brelvp: NULL"); 1973 /* 1974 * Delete from old vnode list, if on one. 1975 */ 1976 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 1977 bufremvn(bp); 1978 if ((vp->v_bioflag & VBIOONSYNCLIST) && 1979 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 1980 vp->v_bioflag &= ~VBIOONSYNCLIST; 1981 LIST_REMOVE(vp, v_synclist); 1982 } 1983 bp->b_vp = NULL; 1984 1985 vdrop(vp); 1986} 1987 1988/* 1989 * Replaces the current vnode associated with the buffer, if any, 1990 * with a new vnode. 1991 * 1992 * If an output I/O is pending on the buffer, the old vnode 1993 * I/O count is adjusted. 1994 * 1995 * Ignores vnode buffer queues. Must be called at splbio(). 1996 */ 1997void 1998buf_replacevnode(struct buf *bp, struct vnode *newvp) 1999{ 2000 struct vnode *oldvp = bp->b_vp; 2001 2002 splassert(IPL_BIO); 2003 2004 if (oldvp) 2005 brelvp(bp); 2006 2007 if ((bp->b_flags & (B_READ | B_DONE)) == 0) { 2008 newvp->v_numoutput++; /* put it on swapdev */ 2009 vwakeup(oldvp); 2010 } 2011 2012 bgetvp(newvp, bp); 2013 bufremvn(bp); 2014} 2015 2016/* 2017 * Used to assign buffers to the appropriate clean or dirty list on 2018 * the vnode and to add newly dirty vnodes to the appropriate 2019 * filesystem syncer list. 2020 * 2021 * Manipulates vnode buffer queues. Must be called at splbio(). 2022 */ 2023void 2024reassignbuf(struct buf *bp) 2025{ 2026 struct buflists *listheadp; 2027 int delay; 2028 struct vnode *vp = bp->b_vp; 2029 2030 splassert(IPL_BIO); 2031 2032 /* 2033 * Delete from old vnode list, if on one. 2034 */ 2035 if (LIST_NEXT(bp, b_vnbufs) != NOLIST) 2036 bufremvn(bp); 2037 2038 /* 2039 * If dirty, put on list of dirty buffers; 2040 * otherwise insert onto list of clean buffers. 2041 */ 2042 if ((bp->b_flags & B_DELWRI) == 0) { 2043 listheadp = &vp->v_cleanblkhd; 2044 if ((vp->v_bioflag & VBIOONSYNCLIST) && 2045 LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { 2046 vp->v_bioflag &= ~VBIOONSYNCLIST; 2047 LIST_REMOVE(vp, v_synclist); 2048 } 2049 } else { 2050 listheadp = &vp->v_dirtyblkhd; 2051 if ((vp->v_bioflag & VBIOONSYNCLIST) == 0) { 2052 switch (vp->v_type) { 2053 case VDIR: 2054 delay = syncdelay / 2; 2055 break; 2056 case VBLK: 2057 if (vp->v_specmountpoint != NULL) { 2058 delay = syncdelay / 3; 2059 break; 2060 } 2061 /* FALLTHROUGH */ 2062 default: 2063 delay = syncdelay; 2064 } 2065 vn_syncer_add_to_worklist(vp, delay); 2066 } 2067 } 2068 bufinsvn(bp, listheadp); 2069} 2070 2071int 2072vfs_register(struct vfsconf *vfs) 2073{ 2074 struct vfsconf *vfsp; 2075 struct vfsconf **vfspp; 2076 2077#ifdef DIAGNOSTIC 2078 /* Paranoia? */ 2079 if (vfs->vfc_refcount != 0) 2080 printf("vfs_register called with vfc_refcount > 0\n"); 2081#endif 2082 2083 /* Check if filesystem already known */ 2084 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2085 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) 2086 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2087 return (EEXIST); 2088 2089 if (vfs->vfc_typenum > maxvfsconf) 2090 maxvfsconf = vfs->vfc_typenum; 2091 2092 vfs->vfc_next = NULL; 2093 2094 /* Add to the end of the list */ 2095 *vfspp = vfs; 2096 2097 /* Call vfs_init() */ 2098 if (vfs->vfc_vfsops->vfs_init) 2099 (*(vfs->vfc_vfsops->vfs_init))(vfs); 2100 2101 return 0; 2102} 2103 2104int 2105vfs_unregister(struct vfsconf *vfs) 2106{ 2107 struct vfsconf *vfsp; 2108 struct vfsconf **vfspp; 2109 int maxtypenum; 2110 2111 /* Find our vfsconf struct */ 2112 for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; 2113 vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { 2114 if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) 2115 break; 2116 } 2117 2118 if (!vfsp) /* Not found */ 2119 return (ENOENT); 2120 2121 if (vfsp->vfc_refcount) /* In use */ 2122 return (EBUSY); 2123 2124 /* Remove from list and free */ 2125 *vfspp = vfsp->vfc_next; 2126 2127 maxtypenum = 0; 2128 2129 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 2130 if (vfsp->vfc_typenum > maxtypenum) 2131 maxtypenum = vfsp->vfc_typenum; 2132 2133 maxvfsconf = maxtypenum; 2134 return 0; 2135} 2136 2137/* 2138 * Check if vnode represents a disk device 2139 */ 2140int 2141vn_isdisk(struct vnode *vp, int *errp) 2142{ 2143 if (vp->v_type != VBLK && vp->v_type != VCHR) 2144 return (0); 2145 2146 return (1); 2147} 2148 2149#ifdef DDB 2150#include <machine/db_machdep.h> 2151#include <ddb/db_interface.h> 2152 2153void 2154vfs_buf_print(void *b, int full, 2155 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2156{ 2157 struct buf *bp = b; 2158 2159 (*pr)(" vp %p lblkno 0x%llx blkno 0x%llx dev 0x%x\n" 2160 " proc %p error %d flags %lb\n", 2161 bp->b_vp, (int64_t)bp->b_lblkno, (int64_t)bp->b_blkno, bp->b_dev, 2162 bp->b_proc, bp->b_error, bp->b_flags, B_BITS); 2163 2164 (*pr)(" bufsize 0x%lx bcount 0x%lx resid 0x%lx\n" 2165 " data %p saveaddr %p dep %p iodone %p\n", 2166 bp->b_bufsize, bp->b_bcount, (long)bp->b_resid, 2167 bp->b_data, bp->b_saveaddr, 2168 LIST_FIRST(&bp->b_dep), bp->b_iodone); 2169 2170 (*pr)(" dirty {off 0x%x end 0x%x} valid {off 0x%x end 0x%x}\n", 2171 bp->b_dirtyoff, bp->b_dirtyend, bp->b_validoff, bp->b_validend); 2172 2173#ifdef FFS_SOFTUPDATES 2174 if (full) 2175 softdep_print(bp, full, pr); 2176#endif 2177} 2178 2179const char *vtypes[] = { VTYPE_NAMES }; 2180const char *vtags[] = { VTAG_NAMES }; 2181 2182void 2183vfs_vnode_print(void *v, int full, 2184 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2185{ 2186 struct vnode *vp = v; 2187 2188 (*pr)("tag %s(%d) type %s(%d) mount %p typedata %p\n", 2189 (u_int)vp->v_tag >= nitems(vtags)? "<unk>":vtags[vp->v_tag], 2190 vp->v_tag, 2191 (u_int)vp->v_type >= nitems(vtypes)? "<unk>":vtypes[vp->v_type], 2192 vp->v_type, vp->v_mount, vp->v_mountedhere); 2193 2194 (*pr)("data %p usecount %d writecount %d holdcnt %d numoutput %d\n", 2195 vp->v_data, vp->v_usecount, vp->v_writecount, 2196 vp->v_holdcnt, vp->v_numoutput); 2197 2198 /* uvm_object_printit(&vp->v_uobj, full, pr); */ 2199 2200 if (full) { 2201 struct buf *bp; 2202 2203 (*pr)("clean bufs:\n"); 2204 LIST_FOREACH(bp, &vp->v_cleanblkhd, b_vnbufs) { 2205 (*pr)(" bp %p\n", bp); 2206 vfs_buf_print(bp, full, pr); 2207 } 2208 2209 (*pr)("dirty bufs:\n"); 2210 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) { 2211 (*pr)(" bp %p\n", bp); 2212 vfs_buf_print(bp, full, pr); 2213 } 2214 } 2215} 2216 2217void 2218vfs_mount_print(struct mount *mp, int full, 2219 int (*pr)(const char *, ...) __attribute__((__format__(__kprintf__,1,2)))) 2220{ 2221 struct vfsconf *vfc = mp->mnt_vfc; 2222 struct vnode *vp; 2223 int cnt; 2224 2225 (*pr)("flags %b\nvnodecovered %p syncer %p data %p\n", 2226 mp->mnt_flag, MNT_BITS, 2227 mp->mnt_vnodecovered, mp->mnt_syncer, mp->mnt_data); 2228 2229 (*pr)("vfsconf: ops %p name \"%s\" num %d ref %d flags 0x%x\n", 2230 vfc->vfc_vfsops, vfc->vfc_name, vfc->vfc_typenum, 2231 vfc->vfc_refcount, vfc->vfc_flags); 2232 2233 (*pr)("statvfs cache: bsize %x iosize %x\nblocks %llu free %llu avail %lld\n", 2234 mp->mnt_stat.f_bsize, mp->mnt_stat.f_iosize, mp->mnt_stat.f_blocks, 2235 mp->mnt_stat.f_bfree, mp->mnt_stat.f_bavail); 2236 2237 (*pr)(" files %llu ffiles %llu favail %lld\n", mp->mnt_stat.f_files, 2238 mp->mnt_stat.f_ffree, mp->mnt_stat.f_favail); 2239 2240 (*pr)(" f_fsidx {0x%x, 0x%x} owner %u ctime 0x%llx\n", 2241 mp->mnt_stat.f_fsid.val[0], mp->mnt_stat.f_fsid.val[1], 2242 mp->mnt_stat.f_owner, mp->mnt_stat.f_ctime); 2243 2244 (*pr)(" syncwrites %llu asyncwrites = %llu\n", 2245 mp->mnt_stat.f_syncwrites, mp->mnt_stat.f_asyncwrites); 2246 2247 (*pr)(" syncreads %llu asyncreads = %llu\n", 2248 mp->mnt_stat.f_syncreads, mp->mnt_stat.f_asyncreads); 2249 2250 (*pr)(" fstype \"%s\" mnton \"%s\" mntfrom \"%s\" mntspec \"%s\"\n", 2251 mp->mnt_stat.f_fstypename, mp->mnt_stat.f_mntonname, 2252 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntfromspec); 2253 2254 (*pr)("locked vnodes:"); 2255 /* XXX would take mountlist lock, except ddb has no context */ 2256 cnt = 0; 2257 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2258 if (VOP_ISLOCKED(vp)) { 2259 if (cnt == 0) 2260 (*pr)("\n %p", vp); 2261 else if ((cnt % (72 / (sizeof(void *) * 2 + 4))) == 0) 2262 (*pr)(",\n %p", vp); 2263 else 2264 (*pr)(", %p", vp); 2265 cnt++; 2266 } 2267 } 2268 (*pr)("\n"); 2269 2270 if (full) { 2271 (*pr)("all vnodes:"); 2272 /* XXX would take mountlist lock, except ddb has no context */ 2273 cnt = 0; 2274 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2275 if (cnt == 0) 2276 (*pr)("\n %p", vp); 2277 else if ((cnt % (72 / (sizeof(void *) * 2 + 4))) == 0) 2278 (*pr)(",\n %p", vp); 2279 else 2280 (*pr)(", %p", vp); 2281 cnt++; 2282 } 2283 (*pr)("\n"); 2284 } 2285} 2286#endif /* DDB */ 2287 2288void 2289copy_statfs_info(struct statfs *sbp, const struct mount *mp) 2290{ 2291 const struct statfs *mbp; 2292 2293 strncpy(sbp->f_fstypename, mp->mnt_vfc->vfc_name, MFSNAMELEN); 2294 2295 if (sbp == (mbp = &mp->mnt_stat)) 2296 return; 2297 2298 sbp->f_fsid = mbp->f_fsid; 2299 sbp->f_owner = mbp->f_owner; 2300 sbp->f_flags = mbp->f_flags; 2301 sbp->f_syncwrites = mbp->f_syncwrites; 2302 sbp->f_asyncwrites = mbp->f_asyncwrites; 2303 sbp->f_syncreads = mbp->f_syncreads; 2304 sbp->f_asyncreads = mbp->f_asyncreads; 2305 sbp->f_namemax = mbp->f_namemax; 2306 memcpy(sbp->f_mntonname, mp->mnt_stat.f_mntonname, MNAMELEN); 2307 memcpy(sbp->f_mntfromname, mp->mnt_stat.f_mntfromname, MNAMELEN); 2308 memcpy(sbp->f_mntfromspec, mp->mnt_stat.f_mntfromspec, MNAMELEN); 2309 memcpy(&sbp->mount_info, &mp->mnt_stat.mount_info, 2310 sizeof(union mount_info)); 2311} 2312