/* $OpenBSD: vfs_subr.c,v 1.45 2000/04/25 22:41:57 csapuntz Exp $ */ /* $NetBSD: vfs_subr.c,v 1.53 1996/04/22 01:39:13 christos Exp $ */ /* * Copyright (c) 1989, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)vfs_subr.c 8.13 (Berkeley) 4/18/94 */ /* * External virtual filesystem routines */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if defined(UVM) #include #endif enum vtype iftovt_tab[16] = { VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, }; int vttoif_tab[9] = { 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, S_IFSOCK, S_IFIFO, S_IFMT, }; int doforce = 1; /* 1 => permit forcible unmounting */ int prtactive = 0; /* 1 => print out reclaim of active vnodes */ int suid_clear = 1; /* 1 => clear SUID / SGID on owner change */ /* * Insq/Remq for the vnode usage lists. */ #define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) #define bufremvn(bp) { \ LIST_REMOVE(bp, b_vnbufs); \ (bp)->b_vnbufs.le_next = NOLIST; \ } struct freelst vnode_hold_list; /* list of vnodes referencing buffers */ struct freelst vnode_free_list; /* vnode free list */ struct mntlist mountlist; /* mounted filesystem list */ struct simplelock mountlist_slock; static struct simplelock mntid_slock; struct simplelock mntvnode_slock; struct simplelock vnode_free_list_slock; struct simplelock spechash_slock; void insmntque __P((struct vnode *, struct mount *)); int getdevvp __P((dev_t, struct vnode **, enum vtype)); int vfs_hang_addrlist __P((struct mount *, struct netexport *, struct export_args *)); int vfs_free_netcred __P((struct radix_node *, void *)); void vfs_free_addrlist __P((struct netexport *)); static __inline__ void vputonfreelist __P((struct vnode *)); #ifdef DEBUG void printlockedvnodes __P((void)); #endif /* * Initialize the vnode management data structures. */ void vntblinit() { simple_lock_init(&mntvnode_slock); simple_lock_init(&mntid_slock); simple_lock_init(&spechash_slock); TAILQ_INIT(&vnode_hold_list); TAILQ_INIT(&vnode_free_list); simple_lock_init(&vnode_free_list_slock); CIRCLEQ_INIT(&mountlist); /* * Initialize the filesystem syncer. */ vn_initialize_syncerd(); } /* * Mark a mount point as busy. Used to synchronize access and to delay * unmounting. Interlock is not released on failure. */ int vfs_busy(mp, flags, interlkp, p) struct mount *mp; int flags; struct simplelock *interlkp; struct proc *p; { int lkflags; if (mp->mnt_flag & MNT_UNMOUNT) { if (flags & LK_NOWAIT) return (ENOENT); mp->mnt_flag |= MNT_MWAIT; if (interlkp) simple_unlock(interlkp); /* * Since all busy locks are shared except the exclusive * lock granted when unmounting, the only place that a * wakeup needs to be done is at the release of the * exclusive lock at the end of dounmount. */ sleep((caddr_t)mp, PVFS); if (interlkp) simple_lock(interlkp); return (ENOENT); } lkflags = LK_SHARED; if (interlkp) lkflags |= LK_INTERLOCK; if (lockmgr(&mp->mnt_lock, lkflags, interlkp, p)) panic("vfs_busy: unexpected lock failure"); return (0); } /* * Free a busy file system */ void vfs_unbusy(mp, p) struct mount *mp; struct proc *p; { lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, p); } /* * Lookup a filesystem type, and if found allocate and initialize * a mount structure for it. * * Devname is usually updated by mount(8) after booting. */ int vfs_rootmountalloc(fstypename, devname, mpp) char *fstypename; char *devname; struct mount **mpp; { struct proc *p = curproc; /* XXX */ struct vfsconf *vfsp; struct mount *mp; for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) if (!strcmp(vfsp->vfc_name, fstypename)) break; if (vfsp == NULL) return (ENODEV); mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); bzero((char *)mp, (u_long)sizeof(struct mount)); lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); (void)vfs_busy(mp, LK_NOWAIT, 0, p); LIST_INIT(&mp->mnt_vnodelist); mp->mnt_vfc = vfsp; mp->mnt_op = vfsp->vfc_vfsops; mp->mnt_flag = MNT_RDONLY; mp->mnt_vnodecovered = NULLVP; vfsp->vfc_refcount++; mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); mp->mnt_stat.f_mntonname[0] = '/'; (void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0); *mpp = mp; return (0); } /* * Find an appropriate filesystem to use for the root. If a filesystem * has not been preselected, walk through the list of known filesystems * trying those that have mountroot routines, and try them until one * works or we have tried them all. */ int vfs_mountroot() { struct vfsconf *vfsp; extern int (*mountroot)(void); int error; if (mountroot != NULL) return ((*mountroot)()); for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) { if (vfsp->vfc_mountroot == NULL) continue; if ((error = (*vfsp->vfc_mountroot)()) == 0) return (0); printf("%s_mountroot failed: %d\n", vfsp->vfc_name, error); } return (ENODEV); } /* * Lookup a mount point by filesystem identifier. */ struct mount * vfs_getvfs(fsid) fsid_t *fsid; { register struct mount *mp; simple_lock(&mountlist_slock); CIRCLEQ_FOREACH(mp, &mountlist, mnt_list) { if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { simple_unlock(&mountlist_slock); return (mp); } } simple_unlock(&mountlist_slock); return ((struct mount *)0); } /* * Get a new unique fsid */ void vfs_getnewfsid(mp) struct mount *mp; { static u_short xxxfs_mntid; fsid_t tfsid; int mtype; simple_lock(&mntid_slock); mtype = mp->mnt_vfc->vfc_typenum; mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); mp->mnt_stat.f_fsid.val[1] = mtype; if (xxxfs_mntid == 0) ++xxxfs_mntid; tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); tfsid.val[1] = mtype; if (mountlist.cqh_first != (void *)&mountlist) { while (vfs_getvfs(&tfsid)) { tfsid.val[0]++; xxxfs_mntid++; } } mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; simple_unlock(&mntid_slock); } /* * Make a 'unique' number from a mount type name. * Note that this is no longer used for ffs which * now has an on-disk filesystem id. */ long makefstype(type) char *type; { long rv; for (rv = 0; *type; type++) { rv <<= 2; rv ^= *type; } return rv; } /* * Set vnode attributes to VNOVAL */ void vattr_null(vap) register struct vattr *vap; { vap->va_type = VNON; /* XXX These next two used to be one line, but for a GCC bug. */ vap->va_size = VNOVAL; vap->va_bytes = VNOVAL; vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = vap->va_fsid = vap->va_fileid = vap->va_blocksize = vap->va_rdev = vap->va_atime.tv_sec = vap->va_atime.tv_nsec = vap->va_mtime.tv_sec = vap->va_mtime.tv_nsec = vap->va_ctime.tv_sec = vap->va_ctime.tv_nsec = vap->va_flags = vap->va_gen = VNOVAL; vap->va_vaflags = 0; } /* * Routines having to do with the management of the vnode table. */ extern int (**dead_vnodeop_p) __P((void *)); long numvnodes; /* * Return the next vnode from the free list. */ int getnewvnode(tag, mp, vops, vpp) enum vtagtype tag; struct mount *mp; int (**vops) __P((void *)); struct vnode **vpp; { struct proc *p = curproc; /* XXX */ struct freelst *listhd; static int toggle; struct vnode *vp; #ifdef DIAGNOSTIC int s; #endif /* * We must choose whether to allocate a new vnode or recycle an * existing one. The criterion for allocating a new one is that * the total number of vnodes is less than the number desired or * there are no vnodes on either free list. Generally we only * want to recycle vnodes that have no buffers associated with * them, so we look first on the vnode_free_list. If it is empty, * we next consider vnodes with referencing buffers on the * vnode_hold_list. The toggle ensures that half the time we * will use a buffer from the vnode_hold_list, and half the time * we will allocate a new one unless the list has grown to twice * the desired size. We are reticent to recycle vnodes from the * vnode_hold_list because we will lose the identity of all its * referencing buffers. */ toggle ^= 1; if (numvnodes > 2 * desiredvnodes) toggle = 0; simple_lock(&vnode_free_list_slock); if ((numvnodes < desiredvnodes) || ((TAILQ_FIRST(listhd = &vnode_free_list) == NULL) && ((TAILQ_FIRST(listhd = &vnode_hold_list) == NULL) || toggle))) { simple_unlock(&vnode_free_list_slock); vp = (struct vnode *)malloc((u_long)sizeof *vp, M_VNODE, M_WAITOK); bzero((char *)vp, sizeof *vp); numvnodes++; } else { for (vp = TAILQ_FIRST(listhd); vp != NULLVP; vp = TAILQ_NEXT(vp, v_freelist)) { if (simple_lock_try(&vp->v_interlock)) break; } /* * Unless this is a bad time of the month, at most * the first NCPUS items on the free list are * locked, so this is close enough to being empty. */ if (vp == NULLVP) { simple_unlock(&vnode_free_list_slock); tablefull("vnode"); *vpp = 0; return (ENFILE); } if (vp->v_usecount) { vprint("free vnode", vp); panic("free vnode isn't"); } TAILQ_REMOVE(listhd, vp, v_freelist); vp->v_flag &= ~VONFREELIST; simple_unlock(&vnode_free_list_slock); vp->v_lease = NULL; if (vp->v_type != VBAD) vgonel(vp, p); else simple_unlock(&vp->v_interlock); #ifdef DIAGNOSTIC if (vp->v_data) { vprint("cleaned vnode", vp); panic("cleaned vnode isn't"); } s = splbio(); if (vp->v_numoutput) panic("Clean vnode has pending I/O's"); splx(s); #endif vp->v_flag = 0; vp->v_lastr = 0; vp->v_ralen = 0; vp->v_maxra = 0; vp->v_lastw = 0; vp->v_lasta = 0; vp->v_cstart = 0; vp->v_clen = 0; vp->v_socket = 0; } vp->v_type = VNON; cache_purge(vp); vp->v_tag = tag; vp->v_op = vops; insmntque(vp, mp); *vpp = vp; vp->v_usecount = 1; vp->v_data = 0; #ifdef UVM simple_lock_init(&vp->v_uvm.u_obj.vmobjlock); #endif return (0); } /* * Move a vnode from one mount queue to another. */ void insmntque(vp, mp) register struct vnode *vp; register struct mount *mp; { simple_lock(&mntvnode_slock); /* * Delete from old mount point vnode list, if on one. */ if (vp->v_mount != NULL) LIST_REMOVE(vp, v_mntvnodes); /* * Insert into list of vnodes for the new mount point, if available. */ if ((vp->v_mount = mp) != NULL) LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); simple_unlock(&mntvnode_slock); } /* * Create a vnode for a block device. * Used for root filesystem, argdev, and swap areas. * Also used for memory file system special devices. */ int bdevvp(dev, vpp) dev_t dev; struct vnode **vpp; { return (getdevvp(dev, vpp, VBLK)); } /* * Create a vnode for a character device. * Used for kernfs and some console handling. */ int cdevvp(dev, vpp) dev_t dev; struct vnode **vpp; { return (getdevvp(dev, vpp, VCHR)); } /* * Create a vnode for a device. * Used by bdevvp (block device) for root file system etc., * and by cdevvp (character device) for console and kernfs. */ int getdevvp(dev, vpp, type) dev_t dev; struct vnode **vpp; enum vtype type; { register struct vnode *vp; struct vnode *nvp; int error; if (dev == NODEV) { *vpp = NULLVP; return (0); } error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp); if (error) { *vpp = NULLVP; return (error); } vp = nvp; vp->v_type = type; if ((nvp = checkalias(vp, dev, NULL)) != 0) { vput(vp); vp = nvp; } *vpp = vp; return (0); } /* * Check to see if the new vnode represents a special device * for which we already have a vnode (either because of * bdevvp() or because of a different vnode representing * the same block device). If such an alias exists, deallocate * the existing contents and return the aliased vnode. The * caller is responsible for filling it with its new contents. */ struct vnode * checkalias(nvp, nvp_rdev, mp) register struct vnode *nvp; dev_t nvp_rdev; struct mount *mp; { struct proc *p = curproc; register struct vnode *vp; struct vnode **vpp; if (nvp->v_type != VBLK && nvp->v_type != VCHR) return (NULLVP); vpp = &speclisth[SPECHASH(nvp_rdev)]; loop: simple_lock(&spechash_slock); for (vp = *vpp; vp; vp = vp->v_specnext) { simple_lock(&vp->v_interlock); if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) { simple_unlock(&vp->v_interlock); continue; } /* * Alias, but not in use, so flush it out. */ if (vp->v_usecount == 0) { simple_unlock(&spechash_slock); vgonel(vp, p); goto loop; } if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { simple_unlock(&spechash_slock); goto loop; } break; } /* * Common case is actually in the if statement */ if (vp == NULL || !(vp->v_tag == VT_NON && vp->v_type == VBLK)) { MALLOC(nvp->v_specinfo, struct specinfo *, sizeof(struct specinfo), M_VNODE, M_WAITOK); nvp->v_rdev = nvp_rdev; nvp->v_hashchain = vpp; nvp->v_specnext = *vpp; nvp->v_specmountpoint = NULL; nvp->v_speclockf = NULL; simple_unlock(&spechash_slock); *vpp = nvp; if (vp != NULLVP) { nvp->v_flag |= VALIASED; vp->v_flag |= VALIASED; vput(vp); } return (NULLVP); } /* * This code is the uncommon case. It is called in case * we found an alias that was VT_NON && vtype of VBLK * This means we found a block device that was created * using bdevvp. * An example of such a vnode is the root partition device vnode * craeted in ffs_mountroot. * * The vnodes created by bdevvp should not be aliased (why?). */ simple_unlock(&spechash_slock); VOP_UNLOCK(vp, 0, p); simple_lock(&vp->v_interlock); vclean(vp, 0, p); vp->v_op = nvp->v_op; vp->v_tag = nvp->v_tag; nvp->v_type = VNON; insmntque(vp, mp); return (vp); } /* * Grab a particular vnode from the free list, increment its * reference count and lock it. The vnode lock bit is set the * vnode is being eliminated in vgone. The process is awakened * when the transition is completed, and an error returned to * indicate that the vnode is no longer usable (possibly having * been changed to a new file system type). */ int vget(vp, flags, p) struct vnode *vp; int flags; struct proc *p; { int error; /* * If the vnode is in the process of being cleaned out for * another use, we wait for the cleaning to finish and then * return failure. Cleaning is determined by checking that * the VXLOCK flag is set. */ if ((flags & LK_INTERLOCK) == 0) simple_lock(&vp->v_interlock); if (vp->v_flag & VXLOCK) { vp->v_flag |= VXWANT; simple_unlock(&vp->v_interlock); tsleep((caddr_t)vp, PINOD, "vget", 0); return (ENOENT); } if ((vp->v_flag & VONFREELIST) && (vp->v_usecount == 0)) { simple_lock(&vnode_free_list_slock); if (vp->v_holdcnt > 0) TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); else TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); simple_unlock(&vnode_free_list_slock); vp->v_flag &= ~VONFREELIST; } vp->v_usecount++; if (flags & LK_TYPE_MASK) { if ((error = vn_lock(vp, flags | LK_INTERLOCK, p)) != 0) { vp->v_usecount--; if (vp->v_usecount == 0) vputonfreelist(vp); simple_unlock(&vp->v_interlock); } return (error); } simple_unlock(&vp->v_interlock); return (0); } #ifdef DIAGNOSTIC /* * Vnode reference. */ void vref(vp) struct vnode *vp; { simple_lock(&vp->v_interlock); if (vp->v_usecount <= 0) panic("vref used where vget required"); vp->v_usecount++; simple_unlock(&vp->v_interlock); } #endif /* DIAGNOSTIC */ static __inline__ void vputonfreelist(vp) struct vnode *vp; { struct freelst *lst; /* * insert at tail of LRU list */ #ifdef DIAGNOSTIC if (vp->v_usecount != 0) { panic("Use count is not zero!"); } if (vp->v_flag & VONFREELIST) { vprint ("vnode already on free list: ", vp); panic ("vnode already on free list"); return; } #endif vp->v_flag |= VONFREELIST; simple_lock(&vnode_free_list_slock); if (vp->v_holdcnt > 0) lst = &vnode_hold_list; else lst = &vnode_free_list; if (vp->v_type == VBAD) TAILQ_INSERT_HEAD(lst, vp, v_freelist); else TAILQ_INSERT_TAIL(lst, vp, v_freelist); simple_unlock(&vnode_free_list_slock); } /* * vput(), just unlock and vrele() */ void vput(vp) register struct vnode *vp; { struct proc *p = curproc; /* XXX */ #ifdef DIAGNOSTIC if (vp == NULL) panic("vput: null vp"); #endif simple_lock(&vp->v_interlock); vp->v_usecount--; if (vp->v_usecount > 0) { simple_unlock(&vp->v_interlock); VOP_UNLOCK(vp, 0, p); return; } #ifdef DIAGNOSTIC if (vp->v_usecount < 0 || vp->v_writecount != 0) { vprint("vput: bad ref count", vp); panic("vput: ref cnt"); } #endif vputonfreelist(vp); simple_unlock(&vp->v_interlock); VOP_INACTIVE(vp, p); } /* * Vnode release - use for active VNODES. * If count drops to zero, call inactive routine and return to freelist. */ void vrele(vp) register struct vnode *vp; { struct proc *p = curproc; /* XXX */ #ifdef DIAGNOSTIC if (vp == NULL) panic("vrele: null vp"); #endif simple_lock(&vp->v_interlock); vp->v_usecount--; if (vp->v_usecount > 0) { simple_unlock(&vp->v_interlock); return; } #ifdef DIAGNOSTIC if (vp->v_usecount < 0 || vp->v_writecount != 0) { vprint("vrele: bad ref count", vp); panic("vrele: ref cnt"); } #endif vputonfreelist(vp); if (vn_lock(vp, LK_EXCLUSIVE|LK_INTERLOCK, p) == 0) VOP_INACTIVE(vp, p); } #ifdef DIAGNOSTIC /* * Page or buffer structure gets a reference. */ void vhold(vp) register struct vnode *vp; { /* * If it is on the freelist and the hold count is currently * zero, move it to the hold list. */ simple_lock(&vp->v_interlock); if ((vp->v_flag & VONFREELIST) && vp->v_holdcnt == 0 && vp->v_usecount == 0) { simple_lock(&vnode_free_list_slock); TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); TAILQ_INSERT_TAIL(&vnode_hold_list, vp, v_freelist); simple_unlock(&vnode_free_list_slock); } vp->v_holdcnt++; simple_unlock(&vp->v_interlock); } /* * Page or buffer structure frees a reference. */ void holdrele(vp) register struct vnode *vp; { simple_lock(&vp->v_interlock); if (vp->v_holdcnt <= 0) panic("holdrele: holdcnt"); vp->v_holdcnt--; /* * If it is on the holdlist and the hold count drops to * zero, move it to the free list. */ if ((vp->v_flag & VONFREELIST) && vp->v_holdcnt == 0 && vp->v_usecount == 0) { simple_lock(&vnode_free_list_slock); TAILQ_REMOVE(&vnode_hold_list, vp, v_freelist); TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); simple_unlock(&vnode_free_list_slock); } simple_unlock(&vp->v_interlock); } #endif /* DIAGNOSTIC */ /* * Remove any vnodes in the vnode table belonging to mount point mp. * * If MNT_NOFORCE is specified, there should not be any active ones, * return error if any are found (nb: this is a user error, not a * system error). If MNT_FORCE is specified, detach any active vnodes * that are found. */ #ifdef DEBUG int busyprt = 0; /* print out busy vnodes */ struct ctldebug debug1 = { "busyprt", &busyprt }; #endif int vflush(mp, skipvp, flags) struct mount *mp; struct vnode *skipvp; int flags; { struct proc *p = curproc; register struct vnode *vp, *nvp; int busy = 0; simple_lock(&mntvnode_slock); loop: for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { if (vp->v_mount != mp) goto loop; nvp = vp->v_mntvnodes.le_next; /* * Skip over a selected vnode. */ if (vp == skipvp) continue; simple_lock(&vp->v_interlock); /* * Skip over a vnodes marked VSYSTEM. */ if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) { simple_unlock(&vp->v_interlock); continue; } /* * If WRITECLOSE is set, only flush out regular file * vnodes open for writing. */ if ((flags & WRITECLOSE) && (vp->v_writecount == 0 || vp->v_type != VREG)) { simple_unlock(&vp->v_interlock); continue; } /* * With v_usecount == 0, all we need to do is clear * out the vnode data structures and we are done. */ if (vp->v_usecount == 0) { simple_unlock(&mntvnode_slock); vgonel(vp, p); simple_lock(&mntvnode_slock); continue; } /* * If FORCECLOSE is set, forcibly close the vnode. * For block or character devices, revert to an * anonymous device. For all other files, just kill them. */ if (flags & FORCECLOSE) { simple_unlock(&mntvnode_slock); if (vp->v_type != VBLK && vp->v_type != VCHR) { vgonel(vp, p); } else { vclean(vp, 0, p); vp->v_op = spec_vnodeop_p; insmntque(vp, (struct mount *)0); } simple_lock(&mntvnode_slock); continue; } #ifdef DEBUG if (busyprt) vprint("vflush: busy vnode", vp); #endif simple_unlock(&vp->v_interlock); busy++; } simple_unlock(&mntvnode_slock); if (busy) return (EBUSY); return (0); } /* * Disassociate the underlying file system from a vnode. * The vnode interlock is held on entry. */ void vclean(vp, flags, p) register struct vnode *vp; int flags; struct proc *p; { int active; /* * Check to see if the vnode is in use. * If so we have to reference it before we clean it out * so that its count cannot fall to zero and generate a * race against ourselves to recycle it. */ if ((active = vp->v_usecount) != 0) vp->v_usecount++; /* * Prevent the vnode from being recycled or * brought into use while we clean it out. */ if (vp->v_flag & VXLOCK) panic("vclean: deadlock"); vp->v_flag |= VXLOCK; #ifdef UVM /* * clean out any VM data associated with the vnode. */ uvm_vnp_terminate(vp); #endif /* * Even if the count is zero, the VOP_INACTIVE routine may still * have the object locked while it cleans it out. The VOP_LOCK * ensures that the VOP_INACTIVE routine is done with its work. * For active vnodes, it ensures that no other activity can * occur while the underlying object is being cleaned out. */ VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK, p); /* * Clean out any buffers associated with the vnode. */ if (flags & DOCLOSE) vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0); /* * If purging an active vnode, it must be closed and * deactivated before being reclaimed. Note that the * VOP_INACTIVE will unlock the vnode */ if (active) { if (flags & DOCLOSE) VOP_CLOSE(vp, FNONBLOCK, NOCRED, p); VOP_INACTIVE(vp, p); } else { /* * Any other processes trying to obtain this lock must first * wait for VXLOCK to clear, then call the new lock operation. */ VOP_UNLOCK(vp, 0, p); } /* * Reclaim the vnode. */ if (VOP_RECLAIM(vp, p)) panic("vclean: cannot reclaim"); if (active) { simple_lock(&vp->v_interlock); vp->v_usecount--; if (vp->v_usecount == 0) { if (vp->v_holdcnt > 0) panic("vclean: not clean"); vputonfreelist(vp); } simple_unlock(&vp->v_interlock); } cache_purge(vp); if (vp->v_vnlock) { if ((vp->v_vnlock->lk_flags & LK_DRAINED) == 0) vprint("vclean: lock not drained", vp); FREE(vp->v_vnlock, M_VNODE); vp->v_vnlock = NULL; } /* * Done with purge, notify sleepers of the grim news. */ vp->v_op = dead_vnodeop_p; vp->v_tag = VT_NON; vp->v_flag &= ~VXLOCK; #ifdef DIAGNOSTIC vp->v_flag &= ~VLOCKSWORK; #endif if (vp->v_flag & VXWANT) { vp->v_flag &= ~VXWANT; wakeup((caddr_t)vp); } } /* * Recycle an unused vnode to the front of the free list. * Release the passed interlock if the vnode will be recycled. */ int vrecycle(vp, inter_lkp, p) struct vnode *vp; struct simplelock *inter_lkp; struct proc *p; { simple_lock(&vp->v_interlock); if (vp->v_usecount == 0) { if (inter_lkp) simple_unlock(inter_lkp); vgonel(vp, p); return (1); } simple_unlock(&vp->v_interlock); return (0); } /* * Eliminate all activity associated with a vnode * in preparation for reuse. */ void vgone(vp) register struct vnode *vp; { struct proc *p = curproc; simple_lock (&vp->v_interlock); vgonel(vp, p); } /* * vgone, with the vp interlock held. */ void vgonel(vp, p) struct vnode *vp; struct proc *p; { register struct vnode *vq; struct vnode *vx; /* * If a vgone (or vclean) is already in progress, * wait until it is done and return. */ if (vp->v_flag & VXLOCK) { vp->v_flag |= VXWANT; simple_unlock(&vp->v_interlock); tsleep((caddr_t)vp, PINOD, "vgone", 0); return; } /* * Clean out the filesystem specific data. */ vclean(vp, DOCLOSE, p); /* * Delete from old mount point vnode list, if on one. */ if (vp->v_mount != NULL) insmntque(vp, (struct mount *)0); /* * If special device, remove it from special device alias list * if it is on one. */ if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) { simple_lock(&spechash_slock); if (*vp->v_hashchain == vp) { *vp->v_hashchain = vp->v_specnext; } else { for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { if (vq->v_specnext != vp) continue; vq->v_specnext = vp->v_specnext; break; } if (vq == NULL) panic("missing bdev"); } if (vp->v_flag & VALIASED) { vx = NULL; for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) continue; if (vx) break; vx = vq; } if (vx == NULL) panic("missing alias"); if (vq == NULL) vx->v_flag &= ~VALIASED; vp->v_flag &= ~VALIASED; } simple_unlock(&spechash_slock); FREE(vp->v_specinfo, M_VNODE); vp->v_specinfo = NULL; } /* * If it is on the freelist and not already at the head, * move it to the head of the list. */ vp->v_type = VBAD; if ((vp->v_flag & VONFREELIST) && vp->v_usecount == 0) { simple_lock(&vnode_free_list_slock); if (vp->v_holdcnt > 0) panic("vgonel: not clean"); if (TAILQ_FIRST(&vnode_free_list) != vp) { TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); } simple_unlock(&vnode_free_list_slock); } } /* * Lookup a vnode by device number. */ int vfinddev(dev, type, vpp) dev_t dev; enum vtype type; struct vnode **vpp; { register struct vnode *vp; int rc =0; simple_lock(&spechash_slock); for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { if (dev != vp->v_rdev || type != vp->v_type) continue; *vpp = vp; rc = 1; break; } simple_unlock(&spechash_slock); return (rc); } /* * Revoke all the vnodes corresponding to the specified minor number * range (endpoints inclusive) of the specified major. */ void vdevgone(maj, minl, minh, type) int maj, minl, minh; enum vtype type; { struct vnode *vp; int mn; for (mn = minl; mn <= minh; mn++) if (vfinddev(makedev(maj, mn), type, &vp)) VOP_REVOKE(vp, REVOKEALL); } /* * Calculate the total number of references to a special device. */ int vcount(vp) struct vnode *vp; { struct vnode *vq, *vnext; int count; loop: if ((vp->v_flag & VALIASED) == 0) return (vp->v_usecount); simple_lock(&spechash_slock); for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) { vnext = vq->v_specnext; if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) continue; /* * Alias, but not in use, so flush it out. */ if (vq->v_usecount == 0 && vq != vp) { simple_unlock(&spechash_slock); vgone(vq); goto loop; } count += vq->v_usecount; } simple_unlock(&spechash_slock); return (count); } /* * Print out a description of a vnode. */ static char *typename[] = { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; void vprint(label, vp) char *label; register struct vnode *vp; { char buf[64]; if (label != NULL) printf("%s: ", label); printf("type %s, usecount %d, writecount %d, holdcount %ld,", typename[vp->v_type], vp->v_usecount, vp->v_writecount, vp->v_holdcnt); buf[0] = '\0'; if (vp->v_flag & VROOT) strcat(buf, "|VROOT"); if (vp->v_flag & VTEXT) strcat(buf, "|VTEXT"); if (vp->v_flag & VSYSTEM) strcat(buf, "|VSYSTEM"); if (vp->v_flag & VXLOCK) strcat(buf, "|VXLOCK"); if (vp->v_flag & VXWANT) strcat(buf, "|VXWANT"); if (vp->v_flag & VBWAIT) strcat(buf, "|VBWAIT"); if (vp->v_flag & VALIASED) strcat(buf, "|VALIASED"); if (buf[0] != '\0') printf(" flags (%s)", &buf[1]); if (vp->v_data == NULL) { printf("\n"); } else { printf("\n\t"); VOP_PRINT(vp); } } #ifdef DEBUG /* * List all of the locked vnodes in the system. * Called when debugging the kernel. */ void printlockedvnodes() { struct proc *p = curproc; register struct mount *mp, *nmp; register struct vnode *vp; printf("Locked vnodes\n"); simple_lock(&mountlist_slock); for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; mp = nmp) { if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { nmp = CIRCLEQ_NEXT(mp, mnt_list); continue; } for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = vp->v_mntvnodes.le_next) { if (VOP_ISLOCKED(vp)) vprint((char *)0, vp); } simple_lock(&mountlist_slock); nmp = CIRCLEQ_NEXT(mp, mnt_list); vfs_unbusy(mp, p); } simple_unlock(&mountlist_slock); } #endif /* * Top level filesystem related information gathering. */ int vfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) int *name; u_int namelen; void *oldp; size_t *oldlenp; void *newp; size_t newlen; struct proc *p; { struct vfsconf *vfsp; /* all sysctl names at this level are at least name and field */ if (namelen < 2) return (ENOTDIR); /* overloaded */ if (name[0] != VFS_GENERIC) { for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) if (vfsp->vfc_typenum == name[0]) break; if (vfsp == NULL) return (EOPNOTSUPP); return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1, oldp, oldlenp, newp, newlen, p)); } switch (name[1]) { case VFS_MAXTYPENUM: return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf)); case VFS_CONF: if (namelen < 3) return (ENOTDIR); /* overloaded */ for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) if (vfsp->vfc_typenum == name[2]) break; if (vfsp == NULL) return (EOPNOTSUPP); return (sysctl_rdstruct(oldp, oldlenp, newp, vfsp, sizeof(struct vfsconf))); } return (EOPNOTSUPP); } int kinfo_vdebug = 1; int kinfo_vgetfailed; #define KINFO_VNODESLOP 10 /* * Dump vnode list (via sysctl). * Copyout address of vnode followed by vnode. */ /* ARGSUSED */ int sysctl_vnode(where, sizep, p) char *where; size_t *sizep; struct proc *p; { register struct mount *mp, *nmp; struct vnode *vp, *nvp; register char *bp = where, *savebp; char *ewhere; int error; #define VPTRSZ sizeof (struct vnode *) #define VNODESZ sizeof (struct vnode) if (where == NULL) { *sizep = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ); return (0); } ewhere = where + *sizep; simple_lock(&mountlist_slock); for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) { if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) { nmp = mp->mnt_list.cqe_next; continue; } savebp = bp; again: for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { /* * Check that the vp is still associated with * this filesystem. RACE: could have been * recycled onto the same filesystem. */ if (vp->v_mount != mp) { simple_unlock(&mntvnode_slock); if (kinfo_vdebug) printf("kinfo: vp changed\n"); bp = savebp; goto again; } nvp = vp->v_mntvnodes.le_next; if (bp + VPTRSZ + VNODESZ > ewhere) { simple_unlock(&mntvnode_slock); *sizep = bp - where; return (ENOMEM); } if ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) || (error = copyout((caddr_t)vp, bp + VPTRSZ, VNODESZ))) return (error); bp += VPTRSZ + VNODESZ; simple_lock(&mntvnode_slock); } simple_unlock(&mntvnode_slock); simple_lock(&mountlist_slock); nmp = mp->mnt_list.cqe_next; vfs_unbusy(mp, p); } simple_unlock(&mountlist_slock); *sizep = bp - where; return (0); } /* * Check to see if a filesystem is mounted on a block device. */ int vfs_mountedon(vp) register struct vnode *vp; { register struct vnode *vq; int error = 0; if (vp->v_specmountpoint != NULL) return (EBUSY); if (vp->v_flag & VALIASED) { simple_lock(&spechash_slock); for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) continue; if (vq->v_specmountpoint != NULL) { error = EBUSY; break; } } simple_unlock(&spechash_slock); } return (error); } /* * Build hash lists of net addresses and hang them off the mount point. * Called by ufs_mount() to set up the lists of export addresses. */ int vfs_hang_addrlist(mp, nep, argp) struct mount *mp; struct netexport *nep; struct export_args *argp; { register struct netcred *np; register struct radix_node_head *rnh; register int i; struct radix_node *rn; struct sockaddr *saddr, *smask = 0; struct domain *dom; int error; if (argp->ex_addrlen == 0) { if (mp->mnt_flag & MNT_DEFEXPORTED) return (EPERM); np = &nep->ne_defexported; np->netc_exflags = argp->ex_flags; np->netc_anon = argp->ex_anon; np->netc_anon.cr_ref = 1; mp->mnt_flag |= MNT_DEFEXPORTED; return (0); } i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK); bzero((caddr_t)np, i); saddr = (struct sockaddr *)(np + 1); error = copyin(argp->ex_addr, (caddr_t)saddr, argp->ex_addrlen); if (error) goto out; if (saddr->sa_len > argp->ex_addrlen) saddr->sa_len = argp->ex_addrlen; if (argp->ex_masklen) { smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); error = copyin(argp->ex_mask, (caddr_t)smask, argp->ex_masklen); if (error) goto out; if (smask->sa_len > argp->ex_masklen) smask->sa_len = argp->ex_masklen; } i = saddr->sa_family; if ((rnh = nep->ne_rtable[i]) == 0) { /* * Seems silly to initialize every AF when most are not * used, do so on demand here */ for (dom = domains; dom; dom = dom->dom_next) if (dom->dom_family == i && dom->dom_rtattach) { dom->dom_rtattach((void **)&nep->ne_rtable[i], dom->dom_rtoffset); break; } if ((rnh = nep->ne_rtable[i]) == 0) { error = ENOBUFS; goto out; } } rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, np->netc_rnodes); if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ error = EPERM; goto out; } np->netc_exflags = argp->ex_flags; np->netc_anon = argp->ex_anon; np->netc_anon.cr_ref = 1; return (0); out: free(np, M_NETADDR); return (error); } /* ARGSUSED */ int vfs_free_netcred(rn, w) struct radix_node *rn; void *w; { register struct radix_node_head *rnh = (struct radix_node_head *)w; (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh); free((caddr_t)rn, M_NETADDR); return (0); } /* * Free the net address hash lists that are hanging off the mount points. */ void vfs_free_addrlist(nep) struct netexport *nep; { register int i; register struct radix_node_head *rnh; for (i = 0; i <= AF_MAX; i++) if ((rnh = nep->ne_rtable[i]) != NULL) { (*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh); free((caddr_t)rnh, M_RTABLE); nep->ne_rtable[i] = 0; } } int vfs_export(mp, nep, argp) struct mount *mp; struct netexport *nep; struct export_args *argp; { int error; if (argp->ex_flags & MNT_DELEXPORT) { vfs_free_addrlist(nep); mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); } if (argp->ex_flags & MNT_EXPORTED) { if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0) return (error); mp->mnt_flag |= MNT_EXPORTED; } return (0); } struct netcred * vfs_export_lookup(mp, nep, nam) register struct mount *mp; struct netexport *nep; struct mbuf *nam; { register struct netcred *np; register struct radix_node_head *rnh; struct sockaddr *saddr; np = NULL; if (mp->mnt_flag & MNT_EXPORTED) { /* * Lookup in the export list first. */ if (nam != NULL) { saddr = mtod(nam, struct sockaddr *); rnh = nep->ne_rtable[saddr->sa_family]; if (rnh != NULL) { np = (struct netcred *) (*rnh->rnh_matchaddr)((caddr_t)saddr, rnh); if (np && np->netc_rnodes->rn_flags & RNF_ROOT) np = NULL; } } /* * If no address match, use the default if it exists. */ if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) np = &nep->ne_defexported; } return (np); } /* * Do the usual access checking. * file_mode, uid and gid are from the vnode in question, * while acc_mode and cred are from the VOP_ACCESS parameter list */ int vaccess(file_mode, uid, gid, acc_mode, cred) mode_t file_mode; uid_t uid; gid_t gid; mode_t acc_mode; struct ucred *cred; { mode_t mask; /* User id 0 always gets access. */ if (cred->cr_uid == 0) return 0; mask = 0; /* Otherwise, check the owner. */ if (cred->cr_uid == uid) { if (acc_mode & VEXEC) mask |= S_IXUSR; if (acc_mode & VREAD) mask |= S_IRUSR; if (acc_mode & VWRITE) mask |= S_IWUSR; return (file_mode & mask) == mask ? 0 : EACCES; } /* Otherwise, check the groups. */ if (cred->cr_gid == gid || groupmember(gid, cred)) { if (acc_mode & VEXEC) mask |= S_IXGRP; if (acc_mode & VREAD) mask |= S_IRGRP; if (acc_mode & VWRITE) mask |= S_IWGRP; return (file_mode & mask) == mask ? 0 : EACCES; } /* Otherwise, check everyone else. */ if (acc_mode & VEXEC) mask |= S_IXOTH; if (acc_mode & VREAD) mask |= S_IROTH; if (acc_mode & VWRITE) mask |= S_IWOTH; return (file_mode & mask) == mask ? 0 : EACCES; } /* * Unmount all file systems. * We traverse the list in reverse order under the assumption that doing so * will avoid needing to worry about dependencies. */ void vfs_unmountall() { register struct mount *mp, *nmp; int allerror, error, again = 1; retry: for (allerror = 0, mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { nmp = mp->mnt_list.cqe_prev; if ((error = dounmount(mp, MNT_FORCE, curproc)) != 0) { printf("unmount of %s failed with error %d\n", mp->mnt_stat.f_mntonname, error); allerror = 1; } } if (allerror) { printf("WARNING: some file systems would not unmount\n"); if (again) { printf("retrying\n"); again = 0; goto retry; } } } /* * Sync and unmount file systems before shutting down. */ void vfs_shutdown() { /* XXX Should suspend scheduling. */ (void) spl0(); printf("syncing disks... "); if (panicstr == 0) { /* Release inodes held by texts before update. */ #if !defined(UVM) vnode_pager_umount(NULL); #endif #ifdef notdef vnshutdown(); #endif /* Sync before unmount, in case we hang on something. */ sys_sync(&proc0, (void *)0, (register_t *)0); /* Unmount file systems. */ vfs_unmountall(); } if (vfs_syncwait(1)) printf("giving up\n"); else printf("done\n"); } /* * perform sync() operation and wait for buffers to flush. * assumtions: called w/ scheduler disabled and physical io enabled * for now called at spl0() XXX */ int vfs_syncwait(verbose) int verbose; { register struct buf *bp; int iter, nbusy, dcount, s; struct proc *p; p = curproc? curproc : &proc0; sys_sync(p, (void *)0, (register_t *)0); /* Wait for sync to finish. */ dcount = 10000; for (iter = 0; iter < 20; iter++) { nbusy = 0; for (bp = &buf[nbuf]; --bp >= buf; ) { if ((bp->b_flags & (B_BUSY|B_INVAL)) == B_BUSY) nbusy++; /* * With soft updates, some buffers that are * written will be remarked as dirty until other * buffers are written. */ if (bp->b_flags & B_DELWRI) { s = splbio(); bremfree(bp); bp->b_flags |= B_BUSY; splx(s); nbusy++; bawrite(bp); if (dcount-- <= 0) { printf("softdep "); return 1; } } } if (nbusy == 0) break; if (verbose) printf("%d ", nbusy); DELAY(40000 * iter); } return nbusy; } /* * posix file system related system variables. */ int fs_posix_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) int *name; u_int namelen; void *oldp; size_t *oldlenp; void *newp; size_t newlen; struct proc *p; { /* all sysctl names at this level are terminal */ if (namelen != 1) return (ENOTDIR); switch (name[0]) { case FS_POSIX_SETUID: if (newp && securelevel > 0) return (EPERM); return(sysctl_int(oldp, oldlenp, newp, newlen, &suid_clear)); default: return (EOPNOTSUPP); } /* NOTREACHED */ } /* * file system related system variables. */ int fs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p) int *name; u_int namelen; void *oldp; size_t *oldlenp; void *newp; size_t newlen; struct proc *p; { sysctlfn *fn; switch (name[0]) { case FS_POSIX: fn = fs_posix_sysctl; break; default: return (EOPNOTSUPP); } return (*fn)(name + 1, namelen - 1, oldp, oldlenp, newp, newlen, p); } /* * Routines dealing with vnodes and buffers */ /* * Update outstanding I/O count and do wakeup if requested. * * Manipulates v_numoutput. Must be called at splbio() */ void vwakeup(bp) register struct buf *bp; { register struct vnode *vp; bp->b_flags &= ~B_WRITEINPROG; if ((vp = bp->b_vp) != NULL) { if (--vp->v_numoutput < 0) panic("vwakeup: neg numoutput"); if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) { vp->v_flag &= ~VBWAIT; wakeup((caddr_t)&vp->v_numoutput); } } } /* * Flush out and invalidate all buffers associated with a vnode. * Called with the underlying object locked. */ int vinvalbuf(vp, flags, cred, p, slpflag, slptimeo) register struct vnode *vp; int flags; struct ucred *cred; struct proc *p; int slpflag, slptimeo; { register struct buf *bp; struct buf *nbp, *blist; int s, error; if (flags & V_SAVE) { s = splbio(); while (vp->v_numoutput) { vp->v_flag |= VBWAIT; sleep((caddr_t)&vp->v_numoutput, PRIBIO + 1); } if (vp->v_dirtyblkhd.lh_first != NULL) { splx(s); if ((error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) != 0) return (error); s = splbio(); if (vp->v_numoutput > 0 || vp->v_dirtyblkhd.lh_first != NULL) panic("vinvalbuf: dirty bufs"); } splx(s); } loop: s = splbio(); for (;;) { if ((blist = vp->v_cleanblkhd.lh_first) && (flags & V_SAVEMETA)) while (blist && blist->b_lblkno < 0) blist = blist->b_vnbufs.le_next; if (!blist && (blist = vp->v_dirtyblkhd.lh_first) && (flags & V_SAVEMETA)) while (blist && blist->b_lblkno < 0) blist = blist->b_vnbufs.le_next; if (!blist) break; for (bp = blist; bp; bp = nbp) { nbp = bp->b_vnbufs.le_next; if (flags & V_SAVEMETA && bp->b_lblkno < 0) continue; if (bp->b_flags & B_BUSY) { bp->b_flags |= B_WANTED; error = tsleep((caddr_t)bp, slpflag | (PRIBIO + 1), "vinvalbuf", slptimeo); if (error) { splx(s); return (error); } break; } bp->b_flags |= B_BUSY | B_VFLUSH; /* * XXX Since there are no node locks for NFS, I believe * there is a slight chance that a delayed write will * occur while sleeping just above, so check for it. */ if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { splx(s); (void) VOP_BWRITE(bp); goto loop; } bp->b_flags |= B_INVAL; brelse(bp); } } if (!(flags & V_SAVEMETA) && (vp->v_dirtyblkhd.lh_first || vp->v_cleanblkhd.lh_first)) panic("vinvalbuf: flush failed"); splx(s); return (0); } void vflushbuf(vp, sync) register struct vnode *vp; int sync; { register struct buf *bp, *nbp; int s; loop: s = splbio(); for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { nbp = bp->b_vnbufs.le_next; if ((bp->b_flags & B_BUSY)) continue; if ((bp->b_flags & B_DELWRI) == 0) panic("vflushbuf: not dirty"); bp->b_flags |= B_BUSY | B_VFLUSH; splx(s); /* * Wait for I/O associated with indirect blocks to complete, * since there is no way to quickly wait for them below. */ if (bp->b_vp == vp || sync == 0) (void) bawrite(bp); else (void) bwrite(bp); goto loop; } if (sync == 0) { splx(s); return; } while (vp->v_numoutput) { vp->v_flag |= VBWAIT; tsleep((caddr_t)&vp->v_numoutput, PRIBIO + 1, "vflushbuf", 0); } splx(s); if (vp->v_dirtyblkhd.lh_first != NULL) { vprint("vflushbuf: dirty", vp); goto loop; } } /* * Associate a buffer with a vnode. * * Manipulates buffer vnode queues. Must be called at splbio(). */ void bgetvp(vp, bp) register struct vnode *vp; register struct buf *bp; { if (bp->b_vp) panic("bgetvp: not free"); VHOLD(vp); bp->b_vp = vp; if (vp->v_type == VBLK || vp->v_type == VCHR) bp->b_dev = vp->v_rdev; else bp->b_dev = NODEV; /* * Insert onto list for new vnode. */ bufinsvn(bp, &vp->v_cleanblkhd); } /* * Disassociate a buffer from a vnode. * * Manipulates vnode buffer queues. Must be called at splbio(). */ void brelvp(bp) register struct buf *bp; { struct vnode *vp; if ((vp = bp->b_vp) == (struct vnode *) 0) panic("brelvp: NULL"); /* * Delete from old vnode list, if on one. */ if (bp->b_vnbufs.le_next != NOLIST) bufremvn(bp); if ((vp->v_flag & VONSYNCLIST) && LIST_FIRST(&vp->v_dirtyblkhd) == NULL) { vp->v_flag &= ~VONSYNCLIST; LIST_REMOVE(vp, v_synclist); } bp->b_vp = (struct vnode *) 0; HOLDRELE(vp); } /* * Reassign a buffer from one vnode to another. Used to assign buffers * to the appropriate clean or dirty list and to add newly dirty vnodes * to the appropriate filesystem syncer list. * * Manipulates vnode buffer queues. Must be called at splbio(). */ void reassignbuf(bp, newvp) register struct buf *bp; register struct vnode *newvp; { struct buflists *listheadp; int delay; if (newvp == NULL) { printf("reassignbuf: NULL"); return; } /* * Delete from old vnode list, if on one. */ if (bp->b_vnbufs.le_next != NOLIST) bufremvn(bp); /* * If dirty, put on list of dirty buffers; * otherwise insert onto list of clean buffers. */ if ((bp->b_flags & B_DELWRI) == 0) { listheadp = &newvp->v_cleanblkhd; if ((newvp->v_flag & VONSYNCLIST) && LIST_FIRST(&newvp->v_dirtyblkhd) == NULL) { newvp->v_flag &= ~VONSYNCLIST; LIST_REMOVE(newvp, v_synclist); } } else { listheadp = &newvp->v_dirtyblkhd; if ((newvp->v_flag & VONSYNCLIST) == 0) { switch (newvp->v_type) { case VDIR: delay = syncdelay / 2; break; case VBLK: if (newvp->v_specmountpoint != NULL) { delay = syncdelay / 3; break; } /* fall through */ default: delay = syncdelay; } vn_syncer_add_to_worklist(newvp, delay); } } bufinsvn(bp, listheadp); } int vfs_register(vfs) struct vfsconf *vfs; { struct vfsconf *vfsp; struct vfsconf **vfspp; #ifdef DIAGNOSTIC /* Paranoia? */ if (vfs->vfc_refcount != 0) printf("vfs_register called with vfc_refcount > 0\n"); #endif /* Check if filesystem already known */ for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) return (EEXIST); if (vfs->vfc_typenum > maxvfsconf) maxvfsconf = vfs->vfc_typenum; vfs->vfc_next = NULL; /* Add to the end of the list */ *vfspp = vfs; /* Call vfs_init() */ if (vfs->vfc_vfsops->vfs_init) (*(vfs->vfc_vfsops->vfs_init))(vfs); return 0; } int vfs_unregister(vfs) struct vfsconf *vfs; { struct vfsconf *vfsp; struct vfsconf **vfspp; int maxtypenum; /* Find our vfsconf struct */ for (vfspp = &vfsconf, vfsp = vfsconf; vfsp; vfspp = &vfsp->vfc_next, vfsp = vfsp->vfc_next) { if (strcmp(vfsp->vfc_name, vfs->vfc_name) == 0) break; } if (!vfsp) /* Not found */ return (ENOENT); if (vfsp->vfc_refcount) /* In use */ return (EBUSY); /* Remove from list and free */ *vfspp = vfsp->vfc_next; maxtypenum = 0; for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) if (vfsp->vfc_typenum > maxtypenum) maxtypenum = vfsp->vfc_typenum; maxvfsconf = maxtypenum; return 0; }