|
|
1.1 ! root 1: /* ! 2: * Copyright (c) 1999 Apple Computer, Inc. All rights reserved. ! 3: * ! 4: * @APPLE_LICENSE_HEADER_START@ ! 5: * ! 6: * Portions Copyright (c) 1999 Apple Computer, Inc. All Rights ! 7: * Reserved. This file contains Original Code and/or Modifications of ! 8: * Original Code as defined in and that are subject to the Apple Public ! 9: * Source License Version 1.1 (the "License"). You may not use this file ! 10: * except in compliance with the License. Please obtain a copy of the ! 11: * License at http://www.apple.com/publicsource and read it before using ! 12: * this file. ! 13: * ! 14: * The Original Code and all software distributed under the License are ! 15: * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER ! 16: * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, ! 17: * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, ! 18: * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the ! 19: * License for the specific language governing rights and limitations ! 20: * under the License. ! 21: * ! 22: * @APPLE_LICENSE_HEADER_END@ ! 23: */ ! 24: ! 25: /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */ ! 26: /* ! 27: * Copyright (c) 1989, 1993 ! 28: * The Regents of the University of California. All rights reserved. ! 29: * (c) UNIX System Laboratories, Inc. ! 30: * All or some portions of this file are derived from material licensed ! 31: * to the University of California by American Telephone and Telegraph ! 32: * Co. or Unix System Laboratories, Inc. and are reproduced herein with ! 33: * the permission of UNIX System Laboratories, Inc. ! 34: * ! 35: * Redistribution and use in source and binary forms, with or without ! 36: * modification, are permitted provided that the following conditions ! 37: * are met: ! 38: * 1. Redistributions of source code must retain the above copyright ! 39: * notice, this list of conditions and the following disclaimer. ! 40: * 2. Redistributions in binary form must reproduce the above copyright ! 41: * notice, this list of conditions and the following disclaimer in the ! 42: * documentation and/or other materials provided with the distribution. ! 43: * 3. All advertising materials mentioning features or use of this software ! 44: * must display the following acknowledgement: ! 45: * This product includes software developed by the University of ! 46: * California, Berkeley and its contributors. ! 47: * 4. Neither the name of the University nor the names of its contributors ! 48: * may be used to endorse or promote products derived from this software ! 49: * without specific prior written permission. ! 50: * ! 51: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND ! 52: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE ! 53: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ! 54: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE ! 55: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL ! 56: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS ! 57: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) ! 58: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT ! 59: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY ! 60: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF ! 61: * SUCH DAMAGE. ! 62: * ! 63: * @(#)vfs_subr.c 8.31 (Berkeley) 5/26/95 ! 64: */ ! 65: ! 66: /* ! 67: * External virtual filesystem routines ! 68: */ ! 69: ! 70: #include <mach_nbc.h> ! 71: #include <sys/param.h> ! 72: #include <sys/systm.h> ! 73: #include <sys/proc.h> ! 74: #include <sys/mount.h> ! 75: #include <sys/time.h> ! 76: #include <sys/vnode.h> ! 77: #include <sys/stat.h> ! 78: #include <sys/namei.h> ! 79: #include <sys/ucred.h> ! 80: #include <sys/buf.h> ! 81: #include <sys/errno.h> ! 82: #include <sys/malloc.h> ! 83: #include <sys/domain.h> ! 84: #include <sys/mbuf.h> ! 85: #include <sys/syslog.h> ! 86: ! 87: #include <sys/vm.h> ! 88: #include <sys/sysctl.h> ! 89: ! 90: #include <miscfs/specfs/specdev.h> ! 91: ! 92: #if MACH_NBC ! 93: #include <kern/mapfs.h> ! 94: #endif /* MACH_NBC */ ! 95: ! 96: enum vtype iftovt_tab[16] = { ! 97: VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, ! 98: VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, ! 99: }; ! 100: int vttoif_tab[9] = { ! 101: 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, ! 102: S_IFSOCK, S_IFIFO, S_IFMT, ! 103: }; ! 104: ! 105: /* ! 106: * Insq/Remq for the vnode usage lists. ! 107: */ ! 108: #define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) ! 109: #define bufremvn(bp) { \ ! 110: LIST_REMOVE(bp, b_vnbufs); \ ! 111: (bp)->b_vnbufs.le_next = NOLIST; \ ! 112: } ! 113: TAILQ_HEAD(freelst, vnode) vnode_free_list; /* vnode free list */ ! 114: struct mntlist mountlist; /* mounted filesystem list */ ! 115: ! 116: /* ! 117: * Have to declare first two locks as actual data even if !MACH_SLOCKS, since ! 118: * a pointers to them get passed around. ! 119: */ ! 120: simple_lock_data_t mountlist_slock; ! 121: simple_lock_data_t mntvnode_slock; ! 122: decl_simple_lock_data(,mntid_slock); ! 123: decl_simple_lock_data(,vnode_free_list_slock); ! 124: decl_simple_lock_data(,spechash_slock); ! 125: ! 126: extern struct lock__bsd__ exchangelock; ! 127: ! 128: /* ! 129: * vnodetarget is the amount of vnodes we expect to get back from the ! 130: * VM object cache. As vm_object_cache_steal() is a cpu bound operation ! 131: * for faster processers this number could be higher. ! 132: * Having this number too high introduces longer delays in the execution ! 133: * of getnewvnode(). ! 134: */ ! 135: unsigned long vnodetarget; /* target for vm_object_cache_steal() */ ! 136: #define VNODE_FREE_TARGET 20 /* Default value for vnodetarget */ ! 137: ! 138: /* ! 139: * We need quite a few vnodes on the free list to sustain the ! 140: * rapid stat() the compilation process does, and still benefit from the name ! 141: * cache. Having too few vnodes on the free list causes serious disk ! 142: * thrashing as we cycle through them. ! 143: */ ! 144: #define VNODE_FREE_MIN 100 /* freelist should have at least these many */ ! 145: ! 146: /* ! 147: * We need to get vnodes back from the VM object cache when a certain # ! 148: * of vnodes are reused from the freelist. This is essential for the ! 149: * caching to be effective in the namecache and the buffer cache [for the ! 150: * metadata]. ! 151: */ ! 152: #define VNODE_TOOMANY_REUSED (VNODE_FREE_MIN/4) ! 153: ! 154: /* ! 155: * If we have enough vnodes on the freelist we do not want to reclaim ! 156: * the vnodes from the VM object cache. ! 157: */ ! 158: #define VNODE_FREE_ENOUGH (VNODE_FREE_MIN + (VNODE_FREE_MIN/2)) ! 159: /* ! 160: * Initialize the vnode management data structures. ! 161: */ ! 162: void ! 163: vntblinit() ! 164: { ! 165: simple_lock_init(&mountlist_slock); ! 166: simple_lock_init(&mntvnode_slock); ! 167: simple_lock_init(&mntid_slock); ! 168: simple_lock_init(&spechash_slock); ! 169: TAILQ_INIT(&vnode_free_list); ! 170: simple_lock_init(&vnode_free_list_slock); ! 171: CIRCLEQ_INIT(&mountlist); ! 172: lockinit(&exchangelock, PVFS, "exchange", 0, 0); ! 173: ! 174: if (!vnodetarget) ! 175: vnodetarget = VNODE_FREE_TARGET; ! 176: ! 177: adjust_vm_object_cache(0, desiredvnodes - VNODE_FREE_MIN); ! 178: } ! 179: ! 180: /* Reset the VM Object Cache with the values passed in */ ! 181: kern_return_t ! 182: reset_vmobjectcache(unsigned int val1, unsigned int val2) ! 183: { ! 184: return(adjust_vm_object_cache(val1-VNODE_FREE_MIN, val2 - VNODE_FREE_MIN)); ! 185: } ! 186: ! 187: ! 188: ! 189: /* ! 190: * Mark a mount point as busy. Used to synchronize access and to delay ! 191: * unmounting. Interlock is not released on failure. ! 192: */ ! 193: int ! 194: vfs_busy(mp, flags, interlkp, p) ! 195: struct mount *mp; ! 196: int flags; ! 197: struct slock *interlkp; ! 198: struct proc *p; ! 199: { ! 200: int lkflags; ! 201: ! 202: if (mp->mnt_flag & MNT_UNMOUNT) { ! 203: if (flags & LK_NOWAIT) ! 204: return (ENOENT); ! 205: mp->mnt_flag |= MNT_MWAIT; ! 206: if (interlkp) ! 207: simple_unlock(interlkp); ! 208: /* ! 209: * Since all busy locks are shared except the exclusive ! 210: * lock granted when unmounting, the only place that a ! 211: * wakeup needs to be done is at the release of the ! 212: * exclusive lock at the end of dounmount. ! 213: */ ! 214: sleep((caddr_t)mp, PVFS); ! 215: if (interlkp) ! 216: simple_lock(interlkp); ! 217: return (ENOENT); ! 218: } ! 219: lkflags = LK_SHARED; ! 220: if (interlkp) ! 221: lkflags |= LK_INTERLOCK; ! 222: if (lockmgr(&mp->mnt_lock, lkflags, interlkp, p)) ! 223: panic("vfs_busy: unexpected lock failure"); ! 224: return (0); ! 225: } ! 226: ! 227: /* ! 228: * Free a busy filesystem. ! 229: */ ! 230: void ! 231: vfs_unbusy(mp, p) ! 232: struct mount *mp; ! 233: struct proc *p; ! 234: { ! 235: ! 236: lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, p); ! 237: } ! 238: ! 239: /* ! 240: * Lookup a filesystem type, and if found allocate and initialize ! 241: * a mount structure for it. ! 242: * ! 243: * Devname is usually updated by mount(8) after booting. ! 244: */ ! 245: int ! 246: vfs_rootmountalloc(fstypename, devname, mpp) ! 247: char *fstypename; ! 248: char *devname; ! 249: struct mount **mpp; ! 250: { ! 251: struct proc *p = current_proc(); /* XXX */ ! 252: struct vfsconf *vfsp; ! 253: struct mount *mp; ! 254: ! 255: for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) ! 256: if (!strcmp(vfsp->vfc_name, fstypename)) ! 257: break; ! 258: if (vfsp == NULL) ! 259: return (ENODEV); ! 260: mp = _MALLOC_ZONE((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); ! 261: bzero((char *)mp, (u_long)sizeof(struct mount)); ! 262: lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); ! 263: (void)vfs_busy(mp, LK_NOWAIT, 0, p); ! 264: LIST_INIT(&mp->mnt_vnodelist); ! 265: mp->mnt_vfc = vfsp; ! 266: mp->mnt_op = vfsp->vfc_vfsops; ! 267: mp->mnt_flag = MNT_RDONLY; ! 268: mp->mnt_vnodecovered = NULLVP; ! 269: vfsp->vfc_refcount++; ! 270: mp->mnt_stat.f_type = vfsp->vfc_typenum; ! 271: mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; ! 272: strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); ! 273: mp->mnt_stat.f_mntonname[0] = '/'; ! 274: (void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); ! 275: *mpp = mp; ! 276: return (0); ! 277: } ! 278: ! 279: /* ! 280: * Find an appropriate filesystem to use for the root. If a filesystem ! 281: * has not been preselected, walk through the list of known filesystems ! 282: * trying those that have mountroot routines, and try them until one ! 283: * works or we have tried them all. ! 284: */ ! 285: int ! 286: vfs_mountroot() ! 287: { ! 288: struct vfsconf *vfsp; ! 289: extern int (*mountroot)(void); ! 290: int error; ! 291: ! 292: if (mountroot != NULL) { ! 293: error = (*mountroot)(); ! 294: return (error); ! 295: } ! 296: ! 297: for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) { ! 298: if (vfsp->vfc_mountroot == NULL) ! 299: continue; ! 300: if ((error = (*vfsp->vfc_mountroot)()) == 0) ! 301: return (0); ! 302: printf("%s_mountroot failed: %d\n", vfsp->vfc_name, error); ! 303: } ! 304: return (ENODEV); ! 305: } ! 306: ! 307: /* ! 308: * Lookup a mount point by filesystem identifier. ! 309: */ ! 310: struct mount * ! 311: vfs_getvfs(fsid) ! 312: fsid_t *fsid; ! 313: { ! 314: register struct mount *mp; ! 315: ! 316: simple_lock(&mountlist_slock); ! 317: for (mp = mountlist.cqh_first; mp != (void *)&mountlist; ! 318: mp = mp->mnt_list.cqe_next) { ! 319: if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && ! 320: mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { ! 321: simple_unlock(&mountlist_slock); ! 322: return (mp); ! 323: } ! 324: } ! 325: simple_unlock(&mountlist_slock); ! 326: return ((struct mount *)0); ! 327: } ! 328: ! 329: /* ! 330: * Get a new unique fsid ! 331: */ ! 332: void ! 333: vfs_getnewfsid(mp) ! 334: struct mount *mp; ! 335: { ! 336: static u_short xxxfs_mntid; ! 337: ! 338: fsid_t tfsid; ! 339: int mtype; ! 340: ! 341: simple_lock(&mntid_slock); ! 342: mtype = mp->mnt_vfc->vfc_typenum; ! 343: mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); ! 344: mp->mnt_stat.f_fsid.val[1] = mtype; ! 345: if (xxxfs_mntid == 0) ! 346: ++xxxfs_mntid; ! 347: tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); ! 348: tfsid.val[1] = mtype; ! 349: if (mountlist.cqh_first != (void *)&mountlist) { ! 350: while (vfs_getvfs(&tfsid)) { ! 351: tfsid.val[0]++; ! 352: xxxfs_mntid++; ! 353: } ! 354: } ! 355: mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; ! 356: simple_unlock(&mntid_slock); ! 357: } ! 358: ! 359: /* ! 360: * Set vnode attributes to VNOVAL ! 361: */ ! 362: void ! 363: vattr_null(vap) ! 364: register struct vattr *vap; ! 365: { ! 366: ! 367: vap->va_type = VNON; ! 368: vap->va_size = vap->va_bytes = VNOVAL; ! 369: vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = ! 370: vap->va_fsid = vap->va_fileid = ! 371: vap->va_blocksize = vap->va_rdev = ! 372: vap->va_atime.tv_sec = vap->va_atime.tv_nsec = ! 373: vap->va_mtime.tv_sec = vap->va_mtime.tv_nsec = ! 374: vap->va_ctime.tv_sec = vap->va_ctime.tv_nsec = ! 375: vap->va_flags = vap->va_gen = VNOVAL; ! 376: vap->va_vaflags = 0; ! 377: } ! 378: ! 379: /* ! 380: * Routines having to do with the management of the vnode table. ! 381: */ ! 382: extern int (**dead_vnodeop_p)(); ! 383: static void vclean __P((struct vnode *vp, int flag, struct proc *p)); ! 384: extern void vgonel __P((struct vnode *vp, struct proc *p)); ! 385: long numvnodes, freevnodes; ! 386: ! 387: extern struct vattr va_null; ! 388: ! 389: /* ! 390: * Return the next vnode from the free list. ! 391: */ ! 392: int ! 393: getnewvnode(tag, mp, vops, vpp) ! 394: enum vtagtype tag; ! 395: struct mount *mp; ! 396: int (**vops)(); ! 397: struct vnode **vpp; ! 398: { ! 399: struct proc *p = current_proc(); /* XXX */ ! 400: struct vnode *vp; ! 401: int cnt, didretry = 0; ! 402: static int reused = 0; /* track the reuse rate */ ! 403: ! 404: retry: ! 405: simple_lock(&vnode_free_list_slock); ! 406: /* ! 407: * MALLOC a vnode if the number of vnodes is not reached the desired ! 408: * value. There might be vnodes on the free list, but we do not ! 409: * reuse from the freelist because reusing a vnode implies reusing ! 410: * the name cache entry. ! 411: */ ! 412: if (numvnodes < desiredvnodes) { ! 413: numvnodes++; ! 414: simple_unlock(&vnode_free_list_slock); ! 415: MALLOC_ZONE(vp, struct vnode *, sizeof *vp, M_VNODE, M_WAITOK); ! 416: bzero((char *)vp, sizeof *vp); ! 417: simple_lock_init(&vp->v_interlock); ! 418: } else { ! 419: /* ! 420: * Once the desired number of vnodes are allocated, we start reusing ! 421: * from the freelist. ! 422: */ ! 423: if (freevnodes < VNODE_FREE_MIN) { ! 424: /* ! 425: * if we are low on vnodes on the freelist attempt to get ! 426: * some back from the VM object cache ! 427: */ ! 428: simple_unlock(&vnode_free_list_slock); ! 429: vm_object_cache_steal(vnodetarget); ! 430: simple_lock(&vnode_free_list_slock); ! 431: } ! 432: ! 433: for (cnt = 0, vp = vnode_free_list.tqh_first; ! 434: vp != NULLVP; cnt++, vp = vp->v_freelist.tqe_next) { ! 435: if (simple_lock_try(&vp->v_interlock)) ! 436: break; ! 437: } ! 438: /* ! 439: * Unless this is a bad time of the month, at most ! 440: * the first NCPUS items on the free list are ! 441: * locked, so this is close enough to being empty. ! 442: */ ! 443: if (vp == NULLVP) { ! 444: simple_unlock(&vnode_free_list_slock); ! 445: if (!(didretry++) && (vm_object_cache_steal(vnodetarget) > 0)) ! 446: goto retry; ! 447: tablefull("vnode"); ! 448: log(LOG_EMERG, "%d vnodes locked, %d desired, %d numvnodes\n", ! 449: cnt, desiredvnodes, numvnodes); ! 450: *vpp = 0; ! 451: return (ENFILE); ! 452: } ! 453: ! 454: if (vp->v_usecount) ! 455: panic("free vnode isn't: v_type = %d, v_usecount = %d?", ! 456: vp->v_type, vp->v_usecount); ! 457: ! 458: if (reused > VNODE_TOOMANY_REUSED) { ! 459: reused = 0; ! 460: if (freevnodes < VNODE_FREE_ENOUGH) { ! 461: simple_unlock(&vnode_free_list_slock); ! 462: simple_unlock(&vp->v_interlock); ! 463: vm_object_cache_steal(vnodetarget); ! 464: /* ! 465: * The vnode we have right now can potentially have dirty ! 466: * buffers associated with it. So we do not want to reuse it ! 467: * given a choice. The vnodes reclimed from VM object cache are ! 468: * put on the front of the freelist. So retry can potentially ! 469: * avoid IO, which is a good thing. ! 470: */ ! 471: goto retry; ! 472: } ! 473: } ! 474: ! 475: TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); ! 476: /* see comment on why 0xdeadb is set at end of vgone (below) */ ! 477: vp->v_freelist.tqe_prev = (struct vnode **)0xdeadb; ! 478: freevnodes--; ! 479: reused++; ! 480: simple_unlock(&vnode_free_list_slock); ! 481: vp->v_lease = NULL; ! 482: if (vp->v_type != VBAD) ! 483: vgonel(vp, p); ! 484: else ! 485: simple_unlock(&vp->v_interlock); ! 486: #if DIAGNOSTIC ! 487: if (vp->v_data) ! 488: panic("cleaned vnode isn't"); ! 489: { ! 490: int s = splbio(); ! 491: if (vp->v_numoutput) ! 492: panic("Clean vnode has pending I/O's"); ! 493: splx(s); ! 494: } ! 495: #endif ! 496: vp->v_flag = 0; ! 497: vp->v_lastr = 0; ! 498: vp->v_ralen = 0; ! 499: vp->v_maxra = 0; ! 500: vp->v_lastw = 0; ! 501: vp->v_lasta = 0; ! 502: vp->v_cstart = 0; ! 503: vp->v_clen = 0; ! 504: vp->v_socket = 0; ! 505: vp->v_bread = vp->v_consumed = 0; ! 506: } ! 507: vp->v_power = 5; /* 32k speculative reads */ ! 508: vp->v_trigger = 16 * 8; ! 509: vp->v_type = VNON; ! 510: cache_purge(vp); ! 511: vp->v_tag = tag; ! 512: vp->v_op = vops; ! 513: insmntque(vp, mp); ! 514: *vpp = vp; ! 515: vp->v_usecount = 1; ! 516: vp->v_data = 0; ! 517: return (0); ! 518: } ! 519: ! 520: /* ! 521: * Move a vnode from one mount queue to another. ! 522: */ ! 523: void ! 524: insmntque(vp, mp) ! 525: struct vnode *vp; ! 526: struct mount *mp; ! 527: { ! 528: ! 529: simple_lock(&mntvnode_slock); ! 530: /* ! 531: * Delete from old mount point vnode list, if on one. ! 532: */ ! 533: if (vp->v_mount != NULL) ! 534: LIST_REMOVE(vp, v_mntvnodes); ! 535: /* ! 536: * Insert into list of vnodes for the new mount point, if available. ! 537: */ ! 538: if ((vp->v_mount = mp) != NULL) ! 539: LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); ! 540: simple_unlock(&mntvnode_slock); ! 541: } ! 542: ! 543: /* ! 544: * Update outstanding I/O count and do wakeup if requested. ! 545: */ ! 546: void ! 547: vwakeup(bp) ! 548: register struct buf *bp; ! 549: { ! 550: register struct vnode *vp; ! 551: ! 552: bp->b_flags &= ~B_WRITEINPROG; ! 553: if (vp = bp->b_vp) { ! 554: if (--vp->v_numoutput < 0) ! 555: panic("vwakeup: neg numoutput"); ! 556: if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) { ! 557: if (vp->v_numoutput < 0) ! 558: panic("vwakeup: neg numoutput 2"); ! 559: vp->v_flag &= ~VBWAIT; ! 560: wakeup((caddr_t)&vp->v_numoutput); ! 561: } ! 562: } ! 563: } ! 564: ! 565: /* ! 566: * Flush out and invalidate all buffers associated with a vnode. ! 567: * Called with the underlying object locked. ! 568: */ ! 569: int ! 570: vinvalbuf(vp, flags, cred, p, slpflag, slptimeo) ! 571: register struct vnode *vp; ! 572: int flags; ! 573: struct ucred *cred; ! 574: struct proc *p; ! 575: int slpflag, slptimeo; ! 576: { ! 577: register struct buf *bp; ! 578: struct buf *nbp, *blist; ! 579: int s, error = 0; ! 580: ! 581: if (flags & V_SAVE) { ! 582: #if MACH_NBC ! 583: if ((vp->v_type == VREG) && (vp->v_vm_info) && (vp->v_vm_info->mapped)) ! 584: if ((error = mapfs_fsync(vp))) ! 585: return (error); ! 586: #endif /* MACH_NBC */ ! 587: if (error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) ! 588: return (error); ! 589: if (vp->v_dirtyblkhd.lh_first != NULL) ! 590: panic("vinvalbuf: dirty bufs"); ! 591: } ! 592: ! 593: /* ! 594: * make sure we don't have any lingering state ! 595: * associated with cluster writes ! 596: */ ! 597: vp->v_cstart = 0; ! 598: vp->v_clen = 0; ! 599: vp->v_lasta = 0; ! 600: vp->v_lastw = 0; ! 601: ! 602: #if MACH_NBC ! 603: if (vp->v_type == VREG) { ! 604: error = mapfs_invalidate(vp); ! 605: #if DIAGNOSTIC ! 606: if (error) ! 607: kprintf("vinvalbuf: mapfs_invalidate(0x%x) returned %d\n", vp, error); ! 608: #endif ! 609: } ! 610: #endif /* MACH_NBC */ ! 611: ! 612: for (;;) { ! 613: if ((blist = vp->v_cleanblkhd.lh_first) && flags & V_SAVEMETA) ! 614: while (blist && blist->b_lblkno < 0) ! 615: blist = blist->b_vnbufs.le_next; ! 616: if (!blist && (blist = vp->v_dirtyblkhd.lh_first) && ! 617: (flags & V_SAVEMETA)) ! 618: while (blist && blist->b_lblkno < 0) ! 619: blist = blist->b_vnbufs.le_next; ! 620: if (!blist) ! 621: break; ! 622: ! 623: for (bp = blist; bp; bp = nbp) { ! 624: nbp = bp->b_vnbufs.le_next; ! 625: if (flags & V_SAVEMETA && bp->b_lblkno < 0) ! 626: continue; ! 627: s = splbio(); ! 628: if (bp->b_flags & B_BUSY) { ! 629: bp->b_flags |= B_WANTED; ! 630: error = tsleep((caddr_t)bp, ! 631: slpflag | (PRIBIO + 1), "vinvalbuf", ! 632: slptimeo); ! 633: splx(s); ! 634: if (error) ! 635: return (error); ! 636: break; ! 637: } ! 638: bremfree(bp); ! 639: bp->b_flags |= B_BUSY; ! 640: splx(s); ! 641: /* ! 642: * XXX Since there are no node locks for NFS, I believe ! 643: * there is a slight chance that a delayed write will ! 644: * occur while sleeping just above, so check for it. ! 645: */ ! 646: if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { ! 647: (void) VOP_BWRITE(bp); ! 648: break; ! 649: } ! 650: bp->b_flags |= B_INVAL; ! 651: brelse(bp); ! 652: } ! 653: } ! 654: if (!(flags & V_SAVEMETA) && ! 655: (vp->v_dirtyblkhd.lh_first || vp->v_cleanblkhd.lh_first)) ! 656: panic("vinvalbuf: flush failed"); ! 657: return (0); ! 658: } ! 659: ! 660: /* ! 661: * Associate a buffer with a vnode. ! 662: */ ! 663: void ! 664: bgetvp(vp, bp) ! 665: register struct vnode *vp; ! 666: register struct buf *bp; ! 667: { ! 668: ! 669: if (bp->b_vp) ! 670: panic("bgetvp: not free"); ! 671: VHOLD(vp); ! 672: bp->b_vp = vp; ! 673: if (vp->v_type == VBLK || vp->v_type == VCHR) ! 674: bp->b_dev = vp->v_rdev; ! 675: else ! 676: bp->b_dev = NODEV; ! 677: /* ! 678: * Insert onto list for new vnode. ! 679: */ ! 680: bufinsvn(bp, &vp->v_cleanblkhd); ! 681: } ! 682: ! 683: /* ! 684: * Disassociate a buffer from a vnode. ! 685: */ ! 686: void ! 687: brelvp(bp) ! 688: register struct buf *bp; ! 689: { ! 690: struct vnode *vp; ! 691: ! 692: if (bp->b_vp == (struct vnode *) 0) ! 693: panic("brelvp: NULL"); ! 694: /* ! 695: * Delete from old vnode list, if on one. ! 696: */ ! 697: if (bp->b_vnbufs.le_next != NOLIST) ! 698: bufremvn(bp); ! 699: vp = bp->b_vp; ! 700: bp->b_vp = (struct vnode *) 0; ! 701: HOLDRELE(vp); ! 702: } ! 703: ! 704: /* ! 705: * Reassign a buffer from one vnode to another. ! 706: * Used to assign file specific control information ! 707: * (indirect blocks) to the vnode to which they belong. ! 708: */ ! 709: void ! 710: reassignbuf(bp, newvp) ! 711: register struct buf *bp; ! 712: register struct vnode *newvp; ! 713: { ! 714: register struct buflists *listheadp; ! 715: ! 716: if (newvp == NULL) { ! 717: printf("reassignbuf: NULL"); ! 718: return; ! 719: } ! 720: /* ! 721: * Delete from old vnode list, if on one. ! 722: */ ! 723: if (bp->b_vnbufs.le_next != NOLIST) ! 724: bufremvn(bp); ! 725: /* ! 726: * If dirty, put on list of dirty buffers; ! 727: * otherwise insert onto list of clean buffers. ! 728: */ ! 729: if (bp->b_flags & B_DELWRI) ! 730: listheadp = &newvp->v_dirtyblkhd; ! 731: else ! 732: listheadp = &newvp->v_cleanblkhd; ! 733: bufinsvn(bp, listheadp); ! 734: } ! 735: ! 736: /* ! 737: * Create a vnode for a block device. ! 738: * Used for root filesystem, argdev, and swap areas. ! 739: * Also used for memory file system special devices. ! 740: */ ! 741: int ! 742: bdevvp(dev, vpp) ! 743: dev_t dev; ! 744: struct vnode **vpp; ! 745: { ! 746: register struct vnode *vp; ! 747: struct vnode *nvp; ! 748: int error; ! 749: ! 750: if (dev == NODEV) { ! 751: *vpp = NULLVP; ! 752: return (ENODEV); ! 753: } ! 754: error = getnewvnode(VT_NON, (struct mount *)0, spec_vnodeop_p, &nvp); ! 755: if (error) { ! 756: *vpp = NULLVP; ! 757: return (error); ! 758: } ! 759: vp = nvp; ! 760: vp->v_type = VBLK; ! 761: if (nvp = checkalias(vp, dev, (struct mount *)0)) { ! 762: vput(vp); ! 763: vp = nvp; ! 764: } ! 765: *vpp = vp; ! 766: return (0); ! 767: } ! 768: ! 769: /* ! 770: * Check to see if the new vnode represents a special device ! 771: * for which we already have a vnode (either because of ! 772: * bdevvp() or because of a different vnode representing ! 773: * the same block device). If such an alias exists, deallocate ! 774: * the existing contents and return the aliased vnode. The ! 775: * caller is responsible for filling it with its new contents. ! 776: */ ! 777: struct vnode * ! 778: checkalias(nvp, nvp_rdev, mp) ! 779: register struct vnode *nvp; ! 780: dev_t nvp_rdev; ! 781: struct mount *mp; ! 782: { ! 783: struct proc *p = current_proc(); /* XXX */ ! 784: struct vnode *vp; ! 785: struct vnode **vpp; ! 786: ! 787: if (nvp->v_type != VBLK && nvp->v_type != VCHR) ! 788: return (NULLVP); ! 789: ! 790: vpp = &speclisth[SPECHASH(nvp_rdev)]; ! 791: loop: ! 792: simple_lock(&spechash_slock); ! 793: for (vp = *vpp; vp; vp = vp->v_specnext) { ! 794: if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) ! 795: continue; ! 796: /* ! 797: * Alias, but not in use, so flush it out. ! 798: */ ! 799: simple_lock(&vp->v_interlock); ! 800: if (vp->v_usecount == 0) { ! 801: simple_unlock(&spechash_slock); ! 802: vgonel(vp, p); ! 803: goto loop; ! 804: } ! 805: if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { ! 806: simple_unlock(&spechash_slock); ! 807: goto loop; ! 808: } ! 809: break; ! 810: } ! 811: if (vp == NULL || vp->v_tag != VT_NON) { ! 812: MALLOC_ZONE(nvp->v_specinfo, struct specinfo *, ! 813: sizeof(struct specinfo), M_VNODE, M_WAITOK); ! 814: bzero(nvp->v_specinfo, sizeof(struct specinfo)); ! 815: nvp->v_rdev = nvp_rdev; ! 816: nvp->v_hashchain = vpp; ! 817: nvp->v_specnext = *vpp; ! 818: nvp->v_specflags = 0; ! 819: simple_unlock(&spechash_slock); ! 820: *vpp = nvp; ! 821: if (vp != NULLVP) { ! 822: nvp->v_flag |= VALIASED; ! 823: vp->v_flag |= VALIASED; ! 824: vput(vp); ! 825: } ! 826: return (NULLVP); ! 827: } ! 828: simple_unlock(&spechash_slock); ! 829: VOP_UNLOCK(vp, 0, p); ! 830: simple_lock(&vp->v_interlock); ! 831: vclean(vp, 0, p); ! 832: vp->v_op = nvp->v_op; ! 833: vp->v_tag = nvp->v_tag; ! 834: nvp->v_type = VNON; ! 835: insmntque(vp, mp); ! 836: return (vp); ! 837: } ! 838: ! 839: /* ! 840: * Grab a particular vnode from the free list, increment its ! 841: * reference count and lock it. The vnode lock bit is set the ! 842: * vnode is being eliminated in vgone. The process is awakened ! 843: * when the transition is completed, and an error returned to ! 844: * indicate that the vnode is no longer usable (possibly having ! 845: * been changed to a new file system type). ! 846: */ ! 847: int ! 848: vget(vp, flags, p) ! 849: struct vnode *vp; ! 850: int flags; ! 851: struct proc *p; ! 852: { ! 853: int error; ! 854: ! 855: /* ! 856: * If the vnode is in the process of being cleaned out for ! 857: * another use, we wait for the cleaning to finish and then ! 858: * return failure. Cleaning is determined by checking that ! 859: * the VXLOCK flag is set. ! 860: */ ! 861: if ((flags & LK_INTERLOCK) == 0) ! 862: simple_lock(&vp->v_interlock); ! 863: if (vp->v_flag & VXLOCK) { ! 864: vp->v_flag |= VXWANT; ! 865: simple_unlock(&vp->v_interlock); ! 866: tsleep((caddr_t)vp, PINOD, "vget", 0); ! 867: return (ENOENT); ! 868: } ! 869: if (vp->v_usecount == 0) { ! 870: simple_lock(&vnode_free_list_slock); ! 871: TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); ! 872: freevnodes--; ! 873: simple_unlock(&vnode_free_list_slock); ! 874: } ! 875: vp->v_usecount++; ! 876: if (flags & LK_TYPE_MASK) { ! 877: if (error = vn_lock(vp, flags | LK_INTERLOCK, p)) ! 878: vrele(vp); ! 879: return (error); ! 880: } ! 881: simple_unlock(&vp->v_interlock); ! 882: return (0); ! 883: } ! 884: ! 885: /* ! 886: * Stubs to use when there is no locking to be done on the underlying object. ! 887: * A minimal shared lock is necessary to ensure that the underlying object ! 888: * is not revoked while an operation is in progress. So, an active shared ! 889: * count is maintained in an auxillary vnode lock structure. ! 890: */ ! 891: int ! 892: vop_nolock(ap) ! 893: struct vop_lock_args /* { ! 894: struct vnode *a_vp; ! 895: int a_flags; ! 896: struct proc *a_p; ! 897: } */ *ap; ! 898: { ! 899: #ifdef notyet ! 900: /* ! 901: * This code cannot be used until all the non-locking filesystems ! 902: * (notably NFS) are converted to properly lock and release nodes. ! 903: * Also, certain vnode operations change the locking state within ! 904: * the operation (create, mknod, remove, link, rename, mkdir, rmdir, ! 905: * and symlink). Ideally these operations should not change the ! 906: * lock state, but should be changed to let the caller of the ! 907: * function unlock them. Otherwise all intermediate vnode layers ! 908: * (such as union, umapfs, etc) must catch these functions to do ! 909: * the necessary locking at their layer. Note that the inactive ! 910: * and lookup operations also change their lock state, but this ! 911: * cannot be avoided, so these two operations will always need ! 912: * to be handled in intermediate layers. ! 913: */ ! 914: struct vnode *vp = ap->a_vp; ! 915: int vnflags, flags = ap->a_flags; ! 916: ! 917: if (vp->v_vnlock == NULL) { ! 918: if ((flags & LK_TYPE_MASK) == LK_DRAIN) ! 919: return (0); ! 920: MALLOC_ZONE(vp->v_vnlock, struct lock__bsd__ *, ! 921: sizeof(struct lock__bsd__), M_VNODE, M_WAITOK); ! 922: lockinit(vp->v_vnlock, PVFS, "vnlock", 0, 0); ! 923: } ! 924: switch (flags & LK_TYPE_MASK) { ! 925: case LK_DRAIN: ! 926: vnflags = LK_DRAIN; ! 927: break; ! 928: case LK_EXCLUSIVE: ! 929: case LK_SHARED: ! 930: vnflags = LK_SHARED; ! 931: break; ! 932: case LK_UPGRADE: ! 933: case LK_EXCLUPGRADE: ! 934: case LK_DOWNGRADE: ! 935: return (0); ! 936: case LK_RELEASE: ! 937: default: ! 938: panic("vop_nolock: bad operation %d", flags & LK_TYPE_MASK); ! 939: } ! 940: if (flags & LK_INTERLOCK) ! 941: vnflags |= LK_INTERLOCK; ! 942: return(lockmgr(vp->v_vnlock, vnflags, &vp->v_interlock, ap->a_p)); ! 943: #else /* for now */ ! 944: /* ! 945: * Since we are not using the lock manager, we must clear ! 946: * the interlock here. ! 947: */ ! 948: if (ap->a_flags & LK_INTERLOCK) ! 949: simple_unlock(&ap->a_vp->v_interlock); ! 950: return (0); ! 951: #endif ! 952: } ! 953: ! 954: /* ! 955: * Decrement the active use count. ! 956: */ ! 957: int ! 958: vop_nounlock(ap) ! 959: struct vop_unlock_args /* { ! 960: struct vnode *a_vp; ! 961: int a_flags; ! 962: struct proc *a_p; ! 963: } */ *ap; ! 964: { ! 965: struct vnode *vp = ap->a_vp; ! 966: ! 967: if (vp->v_vnlock == NULL) ! 968: return (0); ! 969: return (lockmgr(vp->v_vnlock, LK_RELEASE, NULL, ap->a_p)); ! 970: } ! 971: ! 972: /* ! 973: * Return whether or not the node is in use. ! 974: */ ! 975: int ! 976: vop_noislocked(ap) ! 977: struct vop_islocked_args /* { ! 978: struct vnode *a_vp; ! 979: } */ *ap; ! 980: { ! 981: struct vnode *vp = ap->a_vp; ! 982: ! 983: if (vp->v_vnlock == NULL) ! 984: return (0); ! 985: return (lockstatus(vp->v_vnlock)); ! 986: } ! 987: ! 988: /* ! 989: * Vnode reference. ! 990: */ ! 991: void ! 992: vref(vp) ! 993: struct vnode *vp; ! 994: { ! 995: ! 996: simple_lock(&vp->v_interlock); ! 997: if (vp->v_usecount <= 0) ! 998: panic("vref used where vget required"); ! 999: vp->v_usecount++; ! 1000: simple_unlock(&vp->v_interlock); ! 1001: } ! 1002: ! 1003: /* ! 1004: * vput(), just unlock and vrele() ! 1005: */ ! 1006: void ! 1007: vput(vp) ! 1008: struct vnode *vp; ! 1009: { ! 1010: struct proc *p = current_proc(); /* XXX */ ! 1011: ! 1012: #if DIGANOSTIC ! 1013: if (vp == NULL) ! 1014: panic("vput: null vp"); ! 1015: #endif ! 1016: simple_lock(&vp->v_interlock); ! 1017: vp->v_usecount--; ! 1018: if (vp->v_usecount > 0) { ! 1019: simple_unlock(&vp->v_interlock); ! 1020: VOP_UNLOCK(vp, 0, p); ! 1021: return; ! 1022: } ! 1023: #if DIAGNOSTIC ! 1024: if (vp->v_usecount < 0 || vp->v_writecount != 0) { ! 1025: vprint("vput: bad ref count", vp); ! 1026: panic("vput: ref cnt"); ! 1027: } ! 1028: #endif ! 1029: /* ! 1030: * insert at tail of LRU list ! 1031: */ ! 1032: simple_lock(&vnode_free_list_slock); ! 1033: TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); ! 1034: freevnodes++; ! 1035: simple_unlock(&vnode_free_list_slock); ! 1036: simple_unlock(&vp->v_interlock); ! 1037: VOP_INACTIVE(vp, p); ! 1038: } ! 1039: ! 1040: /* ! 1041: * Vnode release. ! 1042: * If count drops to zero, call inactive routine and return to freelist. ! 1043: */ ! 1044: void ! 1045: vrele(vp) ! 1046: struct vnode *vp; ! 1047: { ! 1048: struct proc *p = current_proc(); /* XXX */ ! 1049: ! 1050: #if DIAGNOSTIC ! 1051: if (vp == NULL) ! 1052: panic("vrele: null vp"); ! 1053: #endif ! 1054: simple_lock(&vp->v_interlock); ! 1055: vp->v_usecount--; ! 1056: if (vp->v_usecount > 0) { ! 1057: simple_unlock(&vp->v_interlock); ! 1058: return; ! 1059: } ! 1060: #if DIAGNOSTIC ! 1061: if (vp->v_usecount < 0 || vp->v_writecount != 0) { ! 1062: vprint("vrele: bad ref count", vp); ! 1063: panic("vrele: ref cnt"); ! 1064: } ! 1065: #endif ! 1066: /* ! 1067: * insert at tail of LRU list ! 1068: */ ! 1069: simple_lock(&vnode_free_list_slock); ! 1070: if (vp->v_flag & VAGE) { ! 1071: TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); ! 1072: vp->v_flag &= ~VAGE; ! 1073: } else ! 1074: TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); ! 1075: freevnodes++; ! 1076: simple_unlock(&vnode_free_list_slock); ! 1077: if ((vp->v_flag & VXLOCK) == 0) { ! 1078: if (vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK, p) == 0) ! 1079: VOP_INACTIVE(vp, p); ! 1080: #if DIAGNOSTIC ! 1081: else ! 1082: kprintf("vrele: vn_lock() failed for vp = 0x%08x\n", vp); ! 1083: } else { ! 1084: kprintf("vrele: attempted deadlock!\n"); ! 1085: #endif ! 1086: simple_unlock(&vp->v_interlock); ! 1087: } ! 1088: ! 1089: } ! 1090: ! 1091: /* ! 1092: * Page or buffer structure gets a reference. ! 1093: */ ! 1094: void ! 1095: vhold(vp) ! 1096: register struct vnode *vp; ! 1097: { ! 1098: ! 1099: simple_lock(&vp->v_interlock); ! 1100: vp->v_holdcnt++; ! 1101: simple_unlock(&vp->v_interlock); ! 1102: } ! 1103: ! 1104: /* ! 1105: * Page or buffer structure frees a reference. ! 1106: */ ! 1107: void ! 1108: holdrele(vp) ! 1109: register struct vnode *vp; ! 1110: { ! 1111: ! 1112: simple_lock(&vp->v_interlock); ! 1113: if (vp->v_holdcnt <= 0) ! 1114: panic("holdrele: holdcnt"); ! 1115: vp->v_holdcnt--; ! 1116: simple_unlock(&vp->v_interlock); ! 1117: } ! 1118: ! 1119: /* ! 1120: * Remove any vnodes in the vnode table belonging to mount point mp. ! 1121: * ! 1122: * If MNT_NOFORCE is specified, there should not be any active ones, ! 1123: * return error if any are found (nb: this is a user error, not a ! 1124: * system error). If MNT_FORCE is specified, detach any active vnodes ! 1125: * that are found. ! 1126: */ ! 1127: #if DIAGNOSTIC ! 1128: int busyprt = 0; /* print out busy vnodes */ ! 1129: struct ctldebug debug1 = { "busyprt", &busyprt }; ! 1130: #endif ! 1131: ! 1132: int ! 1133: vflush(mp, skipvp, flags) ! 1134: struct mount *mp; ! 1135: struct vnode *skipvp; ! 1136: int flags; ! 1137: { ! 1138: struct proc *p = current_proc(); /* XXX */ ! 1139: struct vnode *vp, *nvp; ! 1140: int busy = 0; ! 1141: ! 1142: simple_lock(&mntvnode_slock); ! 1143: loop: ! 1144: for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { ! 1145: if (vp->v_mount != mp) ! 1146: goto loop; ! 1147: nvp = vp->v_mntvnodes.le_next; ! 1148: /* ! 1149: * Skip over a selected vnode. ! 1150: */ ! 1151: if (vp == skipvp) ! 1152: continue; ! 1153: ! 1154: simple_lock(&vp->v_interlock); ! 1155: /* ! 1156: * Skip over a vnodes marked VSYSTEM. ! 1157: */ ! 1158: if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { ! 1159: simple_unlock(&vp->v_interlock); ! 1160: continue; ! 1161: } ! 1162: /* ! 1163: * If WRITECLOSE is set, only flush out regular file ! 1164: * vnodes open for writing. ! 1165: */ ! 1166: if ((flags & WRITECLOSE) && ! 1167: (vp->v_writecount == 0 || vp->v_type != VREG)) { ! 1168: simple_unlock(&vp->v_interlock); ! 1169: continue; ! 1170: } ! 1171: /* ! 1172: * With v_usecount == 0, all we need to do is clear ! 1173: * out the vnode data structures and we are done. ! 1174: */ ! 1175: if (vp->v_usecount == 0) { ! 1176: simple_unlock(&mntvnode_slock); ! 1177: vgonel(vp, p); ! 1178: simple_lock(&mntvnode_slock); ! 1179: continue; ! 1180: } ! 1181: /* ! 1182: * If FORCECLOSE is set, forcibly close the vnode. ! 1183: * For block or character devices, revert to an ! 1184: * anonymous device. For all other files, just kill them. ! 1185: */ ! 1186: if (flags & FORCECLOSE) { ! 1187: simple_unlock(&mntvnode_slock); ! 1188: if (vp->v_type != VBLK && vp->v_type != VCHR) { ! 1189: vgonel(vp, p); ! 1190: } else { ! 1191: vclean(vp, 0, p); ! 1192: vp->v_op = spec_vnodeop_p; ! 1193: insmntque(vp, (struct mount *)0); ! 1194: } ! 1195: simple_lock(&mntvnode_slock); ! 1196: continue; ! 1197: } ! 1198: #if DIAGNOSTIC ! 1199: if (busyprt) ! 1200: vprint("vflush: busy vnode", vp); ! 1201: #endif ! 1202: simple_unlock(&vp->v_interlock); ! 1203: busy++; ! 1204: } ! 1205: simple_unlock(&mntvnode_slock); ! 1206: if (busy) ! 1207: return (EBUSY); ! 1208: return (0); ! 1209: } ! 1210: ! 1211: /* ! 1212: * Disassociate the underlying file system from a vnode. ! 1213: * The vnode interlock is held on entry. ! 1214: */ ! 1215: static void ! 1216: vclean(vp, flags, p) ! 1217: struct vnode *vp; ! 1218: int flags; ! 1219: struct proc *p; ! 1220: { ! 1221: int active; ! 1222: ! 1223: /* ! 1224: * Check to see if the vnode is in use. ! 1225: * If so we have to reference it before we clean it out ! 1226: * so that its count cannot fall to zero and generate a ! 1227: * race against ourselves to recycle it. ! 1228: */ ! 1229: if (active = vp->v_usecount) ! 1230: vp->v_usecount++; ! 1231: /* ! 1232: * Prevent the vnode from being recycled or ! 1233: * brought into use while we clean it out. ! 1234: */ ! 1235: if (vp->v_flag & VXLOCK) ! 1236: panic("vclean: deadlock"); ! 1237: vp->v_flag |= VXLOCK; ! 1238: /* ! 1239: * Even if the count is zero, the VOP_INACTIVE routine may still ! 1240: * have the object locked while it cleans it out. The VOP_LOCK ! 1241: * ensures that the VOP_INACTIVE routine is done with its work. ! 1242: * For active vnodes, it ensures that no other activity can ! 1243: * occur while the underlying object is being cleaned out. ! 1244: */ ! 1245: VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK, p); ! 1246: /* ! 1247: * Clean out any buffers associated with the vnode. ! 1248: */ ! 1249: if (flags & DOCLOSE) ! 1250: vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); ! 1251: ! 1252: if ((vp->v_type == VREG) && (vp->v_vm_info != NULL)) ! 1253: { ! 1254: vm_info_free(vp); ! 1255: vp->v_vm_info = NULL; ! 1256: } ! 1257: /* ! 1258: * If purging an active vnode, it must be closed and ! 1259: * deactivated before being reclaimed. Note that the ! 1260: * VOP_INACTIVE will unlock the vnode. ! 1261: */ ! 1262: if (active) { ! 1263: if (flags & DOCLOSE) ! 1264: VOP_CLOSE(vp, IO_NDELAY, NOCRED, p); ! 1265: VOP_INACTIVE(vp, p); ! 1266: } else { ! 1267: /* ! 1268: * Any other processes trying to obtain this lock must first ! 1269: * wait for VXLOCK to clear, then call the new lock operation. ! 1270: */ ! 1271: VOP_UNLOCK(vp, 0, p); ! 1272: } ! 1273: /* ! 1274: * Reclaim the vnode. ! 1275: */ ! 1276: if (VOP_RECLAIM(vp, p)) ! 1277: panic("vclean: cannot reclaim"); ! 1278: if (active) ! 1279: vrele(vp); ! 1280: cache_purge(vp); ! 1281: if (vp->v_vnlock) { ! 1282: if ((vp->v_vnlock->lk_flags & LK_DRAINED) == 0) ! 1283: vprint("vclean: lock not drained", vp); ! 1284: FREE_ZONE(vp->v_vnlock, sizeof (struct lock__bsd__), M_VNODE); ! 1285: vp->v_vnlock = NULL; ! 1286: } ! 1287: ! 1288: ! 1289: /* ! 1290: * Done with purge, notify sleepers of the grim news. ! 1291: */ ! 1292: vp->v_op = dead_vnodeop_p; ! 1293: vp->v_tag = VT_NON; ! 1294: vp->v_flag &= ~VXLOCK; ! 1295: if (vp->v_flag & VXWANT) { ! 1296: vp->v_flag &= ~VXWANT; ! 1297: wakeup((caddr_t)vp); ! 1298: } ! 1299: } ! 1300: ! 1301: /* ! 1302: * Eliminate all activity associated with the requested vnode ! 1303: * and with all vnodes aliased to the requested vnode. ! 1304: */ ! 1305: int ! 1306: vop_revoke(ap) ! 1307: struct vop_revoke_args /* { ! 1308: struct vnode *a_vp; ! 1309: int a_flags; ! 1310: } */ *ap; ! 1311: { ! 1312: struct vnode *vp, *vq; ! 1313: struct proc *p = current_proc(); /* XXX */ ! 1314: ! 1315: #if DIAGNOSTIC ! 1316: if ((ap->a_flags & REVOKEALL) == 0) ! 1317: panic("vop_revoke"); ! 1318: #endif ! 1319: ! 1320: vp = ap->a_vp; ! 1321: simple_lock(&vp->v_interlock); ! 1322: ! 1323: if (vp->v_flag & VALIASED) { ! 1324: /* ! 1325: * If a vgone (or vclean) is already in progress, ! 1326: * wait until it is done and return. ! 1327: */ ! 1328: if (vp->v_flag & VXLOCK) { ! 1329: vp->v_flag |= VXWANT; ! 1330: simple_unlock(&vp->v_interlock); ! 1331: tsleep((caddr_t)vp, PINOD, "vop_revokeall", 0); ! 1332: return (0); ! 1333: } ! 1334: /* ! 1335: * Ensure that vp will not be vgone'd while we ! 1336: * are eliminating its aliases. ! 1337: */ ! 1338: vp->v_flag |= VXLOCK; ! 1339: simple_unlock(&vp->v_interlock); ! 1340: while (vp->v_flag & VALIASED) { ! 1341: simple_lock(&spechash_slock); ! 1342: for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { ! 1343: if (vq->v_rdev != vp->v_rdev || ! 1344: vq->v_type != vp->v_type || vp == vq) ! 1345: continue; ! 1346: simple_unlock(&spechash_slock); ! 1347: vgone(vq); ! 1348: break; ! 1349: } ! 1350: if (vq == NULLVP) ! 1351: simple_unlock(&spechash_slock); ! 1352: } ! 1353: /* ! 1354: * Remove the lock so that vgone below will ! 1355: * really eliminate the vnode after which time ! 1356: * vgone will awaken any sleepers. ! 1357: */ ! 1358: simple_lock(&vp->v_interlock); ! 1359: vp->v_flag &= ~VXLOCK; ! 1360: } ! 1361: vgonel(vp, p); ! 1362: return (0); ! 1363: } ! 1364: ! 1365: /* ! 1366: * Recycle an unused vnode to the front of the free list. ! 1367: * Release the passed interlock if the vnode will be recycled. ! 1368: */ ! 1369: int ! 1370: vrecycle(vp, inter_lkp, p) ! 1371: struct vnode *vp; ! 1372: struct slock *inter_lkp; ! 1373: struct proc *p; ! 1374: { ! 1375: ! 1376: simple_lock(&vp->v_interlock); ! 1377: if (vp->v_usecount == 0) { ! 1378: if (inter_lkp) ! 1379: simple_unlock(inter_lkp); ! 1380: vgonel(vp, p); ! 1381: return (1); ! 1382: } ! 1383: simple_unlock(&vp->v_interlock); ! 1384: return (0); ! 1385: } ! 1386: ! 1387: /* ! 1388: * Eliminate all activity associated with a vnode ! 1389: * in preparation for reuse. ! 1390: */ ! 1391: void ! 1392: vgone(vp) ! 1393: struct vnode *vp; ! 1394: { ! 1395: struct proc *p = current_proc(); /* XXX */ ! 1396: ! 1397: simple_lock(&vp->v_interlock); ! 1398: vgonel(vp, p); ! 1399: } ! 1400: ! 1401: /* ! 1402: * vgone, with the vp interlock held. ! 1403: */ ! 1404: void ! 1405: vgonel(vp, p) ! 1406: struct vnode *vp; ! 1407: struct proc *p; ! 1408: { ! 1409: struct vnode *vq; ! 1410: struct vnode *vx; ! 1411: ! 1412: /* ! 1413: * If a vgone (or vclean) is already in progress, ! 1414: * wait until it is done and return. ! 1415: */ ! 1416: if (vp->v_flag & VXLOCK) { ! 1417: vp->v_flag |= VXWANT; ! 1418: simple_unlock(&vp->v_interlock); ! 1419: tsleep((caddr_t)vp, PINOD, "vgone", 0); ! 1420: return; ! 1421: } ! 1422: /* ! 1423: * Clean out the filesystem specific data. ! 1424: */ ! 1425: vclean(vp, DOCLOSE, p); ! 1426: /* ! 1427: * Delete from old mount point vnode list, if on one. ! 1428: */ ! 1429: if (vp->v_mount != NULL) ! 1430: insmntque(vp, (struct mount *)0); ! 1431: /* ! 1432: * If special device, remove it from special device alias list ! 1433: * if it is on one. ! 1434: */ ! 1435: if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { ! 1436: simple_lock(&spechash_slock); ! 1437: if (*vp->v_hashchain == vp) { ! 1438: *vp->v_hashchain = vp->v_specnext; ! 1439: } else { ! 1440: for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { ! 1441: if (vq->v_specnext != vp) ! 1442: continue; ! 1443: vq->v_specnext = vp->v_specnext; ! 1444: break; ! 1445: } ! 1446: if (vq == NULL) ! 1447: panic("missing bdev"); ! 1448: } ! 1449: if (vp->v_flag & VALIASED) { ! 1450: vx = NULL; ! 1451: for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { ! 1452: if (vq->v_rdev != vp->v_rdev || ! 1453: vq->v_type != vp->v_type) ! 1454: continue; ! 1455: if (vx) ! 1456: break; ! 1457: vx = vq; ! 1458: } ! 1459: if (vx == NULL) ! 1460: panic("missing alias"); ! 1461: if (vq == NULL) ! 1462: vx->v_flag &= ~VALIASED; ! 1463: vp->v_flag &= ~VALIASED; ! 1464: } ! 1465: simple_unlock(&spechash_slock); ! 1466: FREE_ZONE(vp->v_specinfo, sizeof (struct specinfo), M_VNODE); ! 1467: vp->v_specinfo = NULL; ! 1468: } ! 1469: /* ! 1470: * If it is on the freelist and not already at the head, ! 1471: * move it to the head of the list. The test of the back ! 1472: * pointer and the reference count of zero is because ! 1473: * it will be removed from the free list by getnewvnode, ! 1474: * but will not have its reference count incremented until ! 1475: * after calling vgone. If the reference count were ! 1476: * incremented first, vgone would (incorrectly) try to ! 1477: * close the previous instance of the underlying object. ! 1478: * So, the back pointer is explicitly set to `0xdeadb' in ! 1479: * getnewvnode after removing it from the freelist to ensure ! 1480: * that we do not try to move it here. ! 1481: */ ! 1482: if (vp->v_usecount == 0) { ! 1483: simple_lock(&vnode_free_list_slock); ! 1484: if ((vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb) && ! 1485: vnode_free_list.tqh_first != vp) { ! 1486: TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); ! 1487: TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); ! 1488: } ! 1489: simple_unlock(&vnode_free_list_slock); ! 1490: } ! 1491: vp->v_type = VBAD; ! 1492: } ! 1493: ! 1494: /* ! 1495: * Lookup a vnode by device number. ! 1496: */ ! 1497: int ! 1498: vfinddev(dev, type, vpp) ! 1499: dev_t dev; ! 1500: enum vtype type; ! 1501: struct vnode **vpp; ! 1502: { ! 1503: struct vnode *vp; ! 1504: int rc = 0; ! 1505: ! 1506: simple_lock(&spechash_slock); ! 1507: for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { ! 1508: if (dev != vp->v_rdev || type != vp->v_type) ! 1509: continue; ! 1510: *vpp = vp; ! 1511: rc = 1; ! 1512: break; ! 1513: } ! 1514: simple_unlock(&spechash_slock); ! 1515: return (rc); ! 1516: } ! 1517: ! 1518: /* ! 1519: * Calculate the total number of references to a special device. ! 1520: */ ! 1521: int ! 1522: vcount(vp) ! 1523: struct vnode *vp; ! 1524: { ! 1525: struct vnode *vq, *vnext; ! 1526: int count; ! 1527: ! 1528: loop: ! 1529: if ((vp->v_flag & VALIASED) == 0) ! 1530: return (vp->v_usecount); ! 1531: simple_lock(&spechash_slock); ! 1532: for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { ! 1533: vnext = vq->v_specnext; ! 1534: if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) ! 1535: continue; ! 1536: /* ! 1537: * Alias, but not in use, so flush it out. ! 1538: */ ! 1539: if (vq->v_usecount == 0 && vq != vp) { ! 1540: simple_unlock(&spechash_slock); ! 1541: vgone(vq); ! 1542: goto loop; ! 1543: } ! 1544: count += vq->v_usecount; ! 1545: } ! 1546: simple_unlock(&spechash_slock); ! 1547: return (count); ! 1548: } ! 1549: ! 1550: /* ! 1551: * Print out a description of a vnode. ! 1552: */ ! 1553: static char *typename[] = ! 1554: { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; ! 1555: ! 1556: void ! 1557: vprint(label, vp) ! 1558: char *label; ! 1559: register struct vnode *vp; ! 1560: { ! 1561: char buf[64]; ! 1562: ! 1563: if (label != NULL) ! 1564: printf("%s: ", label); ! 1565: printf("type %s, usecount %d, writecount %d, refcount %d,", ! 1566: typename[vp->v_type], vp->v_usecount, vp->v_writecount, ! 1567: vp->v_holdcnt); ! 1568: buf[0] = '\0'; ! 1569: if (vp->v_flag & VROOT) ! 1570: strcat(buf, "|VROOT"); ! 1571: if (vp->v_flag & VTEXT) ! 1572: strcat(buf, "|VTEXT"); ! 1573: if (vp->v_flag & VSYSTEM) ! 1574: strcat(buf, "|VSYSTEM"); ! 1575: if (vp->v_flag & VXLOCK) ! 1576: strcat(buf, "|VXLOCK"); ! 1577: if (vp->v_flag & VXWANT) ! 1578: strcat(buf, "|VXWANT"); ! 1579: if (vp->v_flag & VBWAIT) ! 1580: strcat(buf, "|VBWAIT"); ! 1581: if (vp->v_flag & VALIASED) ! 1582: strcat(buf, "|VALIASED"); ! 1583: if (buf[0] != '\0') ! 1584: printf(" flags (%s)", &buf[1]); ! 1585: if (vp->v_data == NULL) { ! 1586: printf("\n"); ! 1587: } else { ! 1588: printf("\n\t"); ! 1589: VOP_PRINT(vp); ! 1590: } ! 1591: } ! 1592: ! 1593: #ifdef DEBUG ! 1594: /* ! 1595: * List all of the locked vnodes in the system. ! 1596: * Called when debugging the kernel. ! 1597: */ ! 1598: void ! 1599: printlockedvnodes() ! 1600: { ! 1601: struct proc *p = current_proc(); /* XXX */ ! 1602: struct mount *mp, *nmp; ! 1603: struct vnode *vp; ! 1604: ! 1605: printf("Locked vnodes\n"); ! 1606: simple_lock(&mountlist_slock); ! 1607: for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) { ! 1608: if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { ! 1609: nmp = mp->mnt_list.cqe_next; ! 1610: continue; ! 1611: } ! 1612: for (vp = mp->mnt_vnodelist.lh_first; ! 1613: vp != NULL; ! 1614: vp = vp->v_mntvnodes.le_next) { ! 1615: if (VOP_ISLOCKED(vp)) ! 1616: vprint((char *)0, vp); ! 1617: } ! 1618: simple_lock(&mountlist_slock); ! 1619: nmp = mp->mnt_list.cqe_next; ! 1620: vfs_unbusy(mp, p); ! 1621: } ! 1622: simple_unlock(&mountlist_slock); ! 1623: } ! 1624: #endif ! 1625: ! 1626: /* ! 1627: * Top level filesystem related information gathering. ! 1628: */ ! 1629: int ! 1630: vfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) ! 1631: int *name; ! 1632: u_int namelen; ! 1633: void *oldp; ! 1634: size_t *oldlenp; ! 1635: void *newp; ! 1636: size_t newlen; ! 1637: struct proc *p; ! 1638: { ! 1639: struct ctldebug *cdp; ! 1640: struct vfsconf *vfsp; ! 1641: ! 1642: #ifdef NeXT ! 1643: if (name[0] == VFS_NUMMNTOPS) { ! 1644: extern unsigned int vfs_nummntops; ! 1645: return (sysctl_rdint(oldp, oldlenp, newp, vfs_nummntops)); ! 1646: } ! 1647: #endif ! 1648: /* all sysctl names at this level are at least name and field */ ! 1649: if (namelen < 2) ! 1650: return (ENOTDIR); /* overloaded */ ! 1651: if (name[0] != VFS_GENERIC) { ! 1652: for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) ! 1653: if (vfsp->vfc_typenum == name[0]) ! 1654: break; ! 1655: if (vfsp == NULL) ! 1656: return (EOPNOTSUPP); ! 1657: return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, ! 1658: oldp, oldlenp, newp, newlen, p)); ! 1659: } ! 1660: switch (name[1]) { ! 1661: case VFS_MAXTYPENUM: ! 1662: return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); ! 1663: case VFS_CONF: ! 1664: if (namelen < 3) ! 1665: return (ENOTDIR); /* overloaded */ ! 1666: for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) ! 1667: if (vfsp->vfc_typenum == name[2]) ! 1668: break; ! 1669: if (vfsp == NULL) ! 1670: return (EOPNOTSUPP); ! 1671: return (sysctl_rdstruct(oldp, oldlenp, newp, vfsp, ! 1672: sizeof(struct vfsconf))); ! 1673: } ! 1674: return (EOPNOTSUPP); ! 1675: } ! 1676: ! 1677: int kinfo_vdebug = 1; ! 1678: int kinfo_vgetfailed; ! 1679: #define KINFO_VNODESLOP 10 ! 1680: /* ! 1681: * Dump vnode list (via sysctl). ! 1682: * Copyout address of vnode followed by vnode. ! 1683: */ ! 1684: /* ARGSUSED */ ! 1685: int ! 1686: sysctl_vnode(where, sizep, p) ! 1687: char *where; ! 1688: size_t *sizep; ! 1689: struct proc *p; ! 1690: { ! 1691: struct mount *mp, *nmp; ! 1692: struct vnode *nvp, *vp; ! 1693: char *bp = where, *savebp; ! 1694: char *ewhere; ! 1695: int error; ! 1696: ! 1697: #define VPTRSZ sizeof (struct vnode *) ! 1698: #define VNODESZ sizeof (struct vnode) ! 1699: if (where == NULL) { ! 1700: *sizep = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ); ! 1701: return (0); ! 1702: } ! 1703: ewhere = where + *sizep; ! 1704: ! 1705: simple_lock(&mountlist_slock); ! 1706: for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) { ! 1707: if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { ! 1708: nmp = mp->mnt_list.cqe_next; ! 1709: continue; ! 1710: } ! 1711: savebp = bp; ! 1712: again: ! 1713: simple_lock(&mntvnode_slock); ! 1714: for (vp = mp->mnt_vnodelist.lh_first; ! 1715: vp != NULL; ! 1716: vp = nvp) { ! 1717: /* ! 1718: * Check that the vp is still associated with ! 1719: * this filesystem. RACE: could have been ! 1720: * recycled onto the same filesystem. ! 1721: */ ! 1722: if (vp->v_mount != mp) { ! 1723: simple_unlock(&mntvnode_slock); ! 1724: if (kinfo_vdebug) ! 1725: printf("kinfo: vp changed\n"); ! 1726: bp = savebp; ! 1727: goto again; ! 1728: } ! 1729: nvp = vp->v_mntvnodes.le_next; ! 1730: if (bp + VPTRSZ + VNODESZ > ewhere) { ! 1731: simple_unlock(&mntvnode_slock); ! 1732: *sizep = bp - where; ! 1733: return (ENOMEM); ! 1734: } ! 1735: simple_unlock(&mntvnode_slock); ! 1736: if ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) || ! 1737: (error = copyout((caddr_t)vp, bp + VPTRSZ, VNODESZ))) ! 1738: return (error); ! 1739: bp += VPTRSZ + VNODESZ; ! 1740: simple_lock(&mntvnode_slock); ! 1741: } ! 1742: simple_unlock(&mntvnode_slock); ! 1743: simple_lock(&mountlist_slock); ! 1744: nmp = mp->mnt_list.cqe_next; ! 1745: vfs_unbusy(mp, p); ! 1746: } ! 1747: simple_unlock(&mountlist_slock); ! 1748: ! 1749: *sizep = bp - where; ! 1750: return (0); ! 1751: } ! 1752: ! 1753: /* ! 1754: * Check to see if a filesystem is mounted on a block device. ! 1755: */ ! 1756: int ! 1757: vfs_mountedon(vp) ! 1758: struct vnode *vp; ! 1759: { ! 1760: struct vnode *vq; ! 1761: int error = 0; ! 1762: ! 1763: if (vp->v_specflags & SI_MOUNTEDON) ! 1764: return (EBUSY); ! 1765: if (vp->v_flag & VALIASED) { ! 1766: simple_lock(&spechash_slock); ! 1767: for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { ! 1768: if (vq->v_rdev != vp->v_rdev || ! 1769: vq->v_type != vp->v_type) ! 1770: continue; ! 1771: if (vq->v_specflags & SI_MOUNTEDON) { ! 1772: error = EBUSY; ! 1773: break; ! 1774: } ! 1775: } ! 1776: simple_unlock(&spechash_slock); ! 1777: } ! 1778: return (error); ! 1779: } ! 1780: ! 1781: /* ! 1782: * Unmount all filesystems. The list is traversed in reverse order ! 1783: * of mounting to avoid dependencies. ! 1784: */ ! 1785: void ! 1786: vfs_unmountall() ! 1787: { ! 1788: struct mount *mp, *nmp; ! 1789: struct proc *p = current_proc(); /* XXX */ ! 1790: ! 1791: /* ! 1792: * Since this only runs when rebooting, it is not interlocked. ! 1793: */ ! 1794: for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { ! 1795: nmp = mp->mnt_list.cqe_prev; ! 1796: (void) dounmount(mp, MNT_FORCE, p); ! 1797: } ! 1798: } ! 1799: ! 1800: /* ! 1801: * Build hash lists of net addresses and hang them off the mount point. ! 1802: * Called by ufs_mount() to set up the lists of export addresses. ! 1803: */ ! 1804: static int ! 1805: vfs_hang_addrlist(mp, nep, argp) ! 1806: struct mount *mp; ! 1807: struct netexport *nep; ! 1808: struct export_args *argp; ! 1809: { ! 1810: register struct netcred *np; ! 1811: register struct radix_node_head *rnh; ! 1812: register int i; ! 1813: struct radix_node *rn; ! 1814: struct sockaddr *saddr, *smask = 0; ! 1815: struct domain *dom; ! 1816: int error; ! 1817: ! 1818: if (argp->ex_addrlen == 0) { ! 1819: if (mp->mnt_flag & MNT_DEFEXPORTED) ! 1820: return (EPERM); ! 1821: np = &nep->ne_defexported; ! 1822: np->netc_exflags = argp->ex_flags; ! 1823: np->netc_anon = argp->ex_anon; ! 1824: np->netc_anon.cr_ref = 1; ! 1825: mp->mnt_flag |= MNT_DEFEXPORTED; ! 1826: return (0); ! 1827: } ! 1828: i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; ! 1829: MALLOC(np, struct netcred *, i, M_NETADDR, M_WAITOK); ! 1830: bzero((caddr_t)np, i); ! 1831: saddr = (struct sockaddr *)(np + 1); ! 1832: if (error = copyin(argp->ex_addr, (caddr_t)saddr, argp->ex_addrlen)) ! 1833: goto out; ! 1834: if (saddr->sa_len > argp->ex_addrlen) ! 1835: saddr->sa_len = argp->ex_addrlen; ! 1836: if (argp->ex_masklen) { ! 1837: smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); ! 1838: error = copyin(argp->ex_addr, (caddr_t)smask, argp->ex_masklen); ! 1839: if (error) ! 1840: goto out; ! 1841: if (smask->sa_len > argp->ex_masklen) ! 1842: smask->sa_len = argp->ex_masklen; ! 1843: } ! 1844: i = saddr->sa_family; ! 1845: if ((rnh = nep->ne_rtable[i]) == 0) { ! 1846: /* ! 1847: * Seems silly to initialize every AF when most are not ! 1848: * used, do so on demand here ! 1849: */ ! 1850: for (dom = domains; dom; dom = dom->dom_next) ! 1851: if (dom->dom_family == i && dom->dom_rtattach) { ! 1852: dom->dom_rtattach((void **)&nep->ne_rtable[i], ! 1853: dom->dom_rtoffset); ! 1854: break; ! 1855: } ! 1856: if ((rnh = nep->ne_rtable[i]) == 0) { ! 1857: error = ENOBUFS; ! 1858: goto out; ! 1859: } ! 1860: } ! 1861: rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, ! 1862: np->netc_rnodes); ! 1863: if (rn == 0) { ! 1864: /* ! 1865: * One of the reasons that rnh_addaddr may fail is that ! 1866: * the entry already exists. To check for this case, we ! 1867: * look up the entry to see if it is there. If so, we ! 1868: * do not need to make a new entry but do return success. ! 1869: */ ! 1870: _FREE(np, M_NETADDR); ! 1871: rn = (*rnh->rnh_matchaddr)((caddr_t)saddr, rnh); ! 1872: if (rn != 0 && (rn->rn_flags & RNF_ROOT) == 0 && ! 1873: ((struct netcred *)rn)->netc_exflags == argp->ex_flags && ! 1874: !bcmp((caddr_t)&((struct netcred *)rn)->netc_anon, ! 1875: (caddr_t)&argp->ex_anon, sizeof(struct ucred))) ! 1876: return (0); ! 1877: return (EPERM); ! 1878: } ! 1879: np->netc_exflags = argp->ex_flags; ! 1880: np->netc_anon = argp->ex_anon; ! 1881: np->netc_anon.cr_ref = 1; ! 1882: return (0); ! 1883: out: ! 1884: _FREE(np, M_NETADDR); ! 1885: return (error); ! 1886: } ! 1887: ! 1888: /* ARGSUSED */ ! 1889: static int ! 1890: vfs_free_netcred(rn, w) ! 1891: struct radix_node *rn; ! 1892: caddr_t w; ! 1893: { ! 1894: register struct radix_node_head *rnh = (struct radix_node_head *)w; ! 1895: ! 1896: (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh); ! 1897: _FREE((caddr_t)rn, M_NETADDR); ! 1898: return (0); ! 1899: } ! 1900: ! 1901: /* ! 1902: * Free the net address hash lists that are hanging off the mount points. ! 1903: */ ! 1904: static void ! 1905: vfs_free_addrlist(nep) ! 1906: struct netexport *nep; ! 1907: { ! 1908: register int i; ! 1909: register struct radix_node_head *rnh; ! 1910: ! 1911: for (i = 0; i <= AF_MAX; i++) ! 1912: if (rnh = nep->ne_rtable[i]) { ! 1913: (*rnh->rnh_walktree)(rnh, vfs_free_netcred, ! 1914: (caddr_t)rnh); ! 1915: _FREE((caddr_t)rnh, M_RTABLE); ! 1916: nep->ne_rtable[i] = 0; ! 1917: } ! 1918: } ! 1919: ! 1920: int ! 1921: vfs_export(mp, nep, argp) ! 1922: struct mount *mp; ! 1923: struct netexport *nep; ! 1924: struct export_args *argp; ! 1925: { ! 1926: int error; ! 1927: ! 1928: if (argp->ex_flags & MNT_DELEXPORT) { ! 1929: vfs_free_addrlist(nep); ! 1930: mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); ! 1931: } ! 1932: if (argp->ex_flags & MNT_EXPORTED) { ! 1933: if (error = vfs_hang_addrlist(mp, nep, argp)) ! 1934: return (error); ! 1935: mp->mnt_flag |= MNT_EXPORTED; ! 1936: } ! 1937: return (0); ! 1938: } ! 1939: ! 1940: struct netcred * ! 1941: vfs_export_lookup(mp, nep, nam) ! 1942: register struct mount *mp; ! 1943: struct netexport *nep; ! 1944: struct mbuf *nam; ! 1945: { ! 1946: register struct netcred *np; ! 1947: register struct radix_node_head *rnh; ! 1948: struct sockaddr *saddr; ! 1949: ! 1950: np = NULL; ! 1951: if (mp->mnt_flag & MNT_EXPORTED) { ! 1952: /* ! 1953: * Lookup in the export list first. ! 1954: */ ! 1955: if (nam != NULL) { ! 1956: saddr = mtod(nam, struct sockaddr *); ! 1957: rnh = nep->ne_rtable[saddr->sa_family]; ! 1958: if (rnh != NULL) { ! 1959: np = (struct netcred *) ! 1960: (*rnh->rnh_matchaddr)((caddr_t)saddr, ! 1961: rnh); ! 1962: if (np && np->netc_rnodes->rn_flags & RNF_ROOT) ! 1963: np = NULL; ! 1964: } ! 1965: } ! 1966: /* ! 1967: * If no address match, use the default if it exists. ! 1968: */ ! 1969: if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) ! 1970: np = &nep->ne_defexported; ! 1971: } ! 1972: return (np); ! 1973: } ! 1974:
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.