vfs_subr.c

来自「基于组件方式开发操作系统的OSKIT源代码」· C语言 代码 · 共 2,733 行 · 第 1/5 页

		    "vgone", 0, &vp->v_interlock);
		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 != NULL) {
			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. The test of the back
	 * pointer and the reference count of zero is because
	 * it will be removed from the free list by getnewvnode,
	 * but will not have its reference count incremented until
	 * after calling vgone. If the reference count were
	 * incremented first, vgone would (incorrectly) try to
	 * close the previous instance of the underlying object.
	 * So, the back pointer is explicitly set to `0xdeadb' in
	 * getnewvnode after removing it from the freelist to ensure
	 * that we do not try to move it here.
	 */
	if (vp->v_usecount == 0) {
		simple_lock(&vnode_free_list_slock);
		if (vp->v_holdcnt > 0)
			panic("vgonel: not clean, vp %p", vp);
		if (vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb &&
		    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);
	}
	vp->v_type = VBAD;
}

/*
 * Lookup a vnode by device number.
 */
int
vfinddev(dev, type, vpp)
	dev_t dev;
	enum vtype type;
	struct vnode **vpp;
{
	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);
}
#endif /*OSKIT*/

#ifndef OSKIT
/*
 * 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;
	struct vnode *vp;
{
	char buf[64];

	if (label != NULL)
		printf("%s: ", label);
	printf("tag %d type %s, usecount %d, writecount %ld, refcount %ld,",
	    vp->v_tag, 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");
#ifndef OSKIT
		VOP_PRINT(vp);
#endif
	}
}

#ifdef DEBUG
/*
 * List all of the locked vnodes in the system.
 * Called when debugging the kernel.
 */
void
printlockedvnodes()
{
	struct mount *mp, *nmp;
	struct vnode *vp;

	printf("Locked vnodes\n");
	simple_lock(&mountlist_slock);
	for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
			nmp = mp->mnt_list.cqe_next;
			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 = mp->mnt_list.cqe_next;
		vfs_unbusy(mp);
	}
	simple_unlock(&mountlist_slock);
}
#endif

extern const char *mountcompatnames[];
extern const int nmountcompatnames;

/*
 * 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;
{
#if defined(COMPAT_09) || defined(COMPAT_43) || defined(COMPAT_44)
	struct vfsconf vfc;
#endif
	struct vfsops *vfsp;

	/* all sysctl names at this level are at least name and field */
	if (namelen < 2)
		return (ENOTDIR);		/* overloaded */

	/* Not generic: goes to file system. */
	if (name[0] != VFS_GENERIC) {
		if (name[0] >= nmountcompatnames || name[0] < 0 ||
		    mountcompatnames[name[0]] == NULL)
			return (EOPNOTSUPP);
		vfsp = vfs_getopsbyname(mountcompatnames[name[0]]);
		if (vfsp == NULL || vfsp->vfs_sysctl == NULL)
			return (EOPNOTSUPP);
		return ((*vfsp->vfs_sysctl)(&name[1], namelen - 1,
		    oldp, oldlenp, newp, newlen, p));
	}

	/* The rest are generic vfs sysctls. */
	switch (name[1]) {
	case VFS_USERMOUNT:
		return sysctl_int(oldp, oldlenp, newp, newlen, &dovfsusermount);
#if defined(COMPAT_09) || defined(COMPAT_43) || defined(COMPAT_44)
	case VFS_MAXTYPENUM:
		/*
		 * Provided for 4.4BSD-Lite2 compatibility.
		 */
		return (sysctl_rdint(oldp, oldlenp, newp, nmountcompatnames));
	case VFS_CONF:
		/*
		 * Special: a node, next is a file system name.
		 * Provided for 4.4BSD-Lite2 compatibility.
		 */
		if (namelen < 3)
			return (ENOTDIR);	/* overloaded */
		if (name[2] >= nmountcompatnames || name[2] < 0 ||
		    mountcompatnames[name[2]] == NULL)
			return (EOPNOTSUPP);
		vfsp = vfs_getopsbyname(mountcompatnames[name[2]]);
		if (vfsp == NULL)
			return (EOPNOTSUPP);
		vfc.vfc_vfsops = vfsp;
		strncpy(vfc.vfc_name, vfsp->vfs_name, MFSNAMELEN);
		vfc.vfc_typenum = name[2];
		vfc.vfc_refcount = vfsp->vfs_refcount;
		vfc.vfc_flags = 0;
		vfc.vfc_mountroot = vfsp->vfs_mountroot;
		vfc.vfc_next = NULL;
		return (sysctl_rdstruct(oldp, oldlenp, newp, &vfc,
		    sizeof(struct vfsconf)));
#endif
	default:
		break;
	}
	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;
{
	struct mount *mp, *nmp;
	struct vnode *nvp, *vp;
	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)) {
			nmp = mp->mnt_list.cqe_next;
			continue;
		}
		savebp = bp;
again:
		simple_lock(&mntvnode_slock);
		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);
			}
			simple_unlock(&mntvnode_slock);
			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);
	}
	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)
	struct vnode *vp;
{
	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.
 */
static int
vfs_hang_addrlist(mp, nep, argp)
	struct mount *mp;
	struct netexport *nep;
	struct export_args *argp;
{
	struct netcred *np, *enp;
	struct radix_node_head *rnh;
	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);
	memset((caddr_t)np, 0, 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 */
		if (rn == 0) {
			enp = (struct netcred *)(*rnh->rnh_lookup)(saddr,
				smask, rnh);
			if (enp == 0) {
				error = EPERM;
				goto out;
			}
		} else 
			enp = (struct netcred *)rn;

		if (enp->netc_exflags != argp->ex_flags ||
		    enp->netc_anon.cr_uid != argp->ex_anon.cr_uid ||
		    enp->netc_anon.cr_gid != argp->ex_anon.cr_gid ||
		    enp->netc_anon.cr_ngroups != argp->ex_anon.cr_ngroups ||
		    memcmp(&enp->netc_anon.cr_groups, &argp->ex_anon.cr_groups,
			enp->netc_anon.cr_ngroups))
				error = EPERM;
		else
			error = 0;
		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 */
static int
vfs_free_netcred(rn, w)
	struct radix_node *rn;
	void *w;
{
	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.
 */
static void
vfs_free_addrlist(nep)
	struct netexport *nep;
{
	int i;
	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;