vfsunlink.c

/*
 * Copyright (c) 1982, 1986, 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.
 * Copyright 2015-2026 Rivoreo
 *
 * 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
 *    in this position and unchanged.
 * 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.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``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 AUTHOR 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.
 */

#include <sys/cdefs.h>
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/module.h>
#include <sys/sysctl.h>
#include <sys/filio.h>
#include <sys/namei.h>
#include <sys/mount.h>
#include <sys/vnode.h>
#include <sys/buf.h>
#include <sys/priv.h>
#include <sys/conf.h>
#if __FreeBSD_version < 1100066
#include <sys/capability.h>
#else
#include <sys/capsicum.h>
#endif

#include <security/mac/mac_framework.h>
#include <security/audit/audit.h>
#include <sys/jail.h>

#ifndef SDT_PROBE2
#define SDT_PROBE2(...)
#endif

#ifndef SDT_PROBE3
#define SDT_PROBE3(...)
#endif

#if __FreeBSD_version < 1100507
#define vrefact VREF
#endif

#include <sys/stat.h>

#include <ufs/ufs/quota.h>
#include <ufs/ufs/inode.h>
#include <ufs/ufs/extattr.h>
#include <ufs/ufs/ufsmount.h>
#include <ufs/ufs/ufs_extern.h>
#define UFS_DIP_SET DIP_SET
#define UFS_VTOI VTOI
#define ufs_inode inode

#if 0
/*
 * For dotdot lookups in capability mode, only allow the component
 * lookup to succeed if the resulting directory was already traversed
 * during the operation.  Also fail dotdot lookups for non-local
 * filesystems, where external agents might assist local lookups to
 * escape the compartment.
 */
static int
nameicap_check_dotdot(struct nameidata *ndp, struct vnode *dp, int lookup_cap_dotdot_nonlocal)
{
#if 0
	struct nameicap_tracker *nt;
	struct mount *mp;

	if ((ndp->ni_lcf & NI_LCF_CAP_DOTDOT) == 0 || dp == NULL ||
	    dp->v_type != VDIR)
		return (0);
	mp = dp->v_mount;
	if (lookup_cap_dotdot_nonlocal == 0 && mp != NULL &&
	    (mp->mnt_flag & MNT_LOCAL) == 0)
		return (ENOTCAPABLE);
	TAILQ_FOREACH_REVERSE(nt, &ndp->ni_cap_tracker, nameicap_tracker_head,
	    nm_link) {
		if (dp == nt->dp)
			return (0);
	}
#endif
	return (ENOTCAPABLE);
}
#endif

static int compute_cn_lkflags(struct mount *mp, int lkflags, int cnflags) {
	if (mp == NULL || ((lkflags & LK_SHARED) &&
	    (!(mp->mnt_kern_flag & MNTK_LOOKUP_SHARED) ||
	    ((cnflags & ISDOTDOT) &&
	    (mp->mnt_kern_flag & MNTK_LOOKUP_EXCL_DOTDOT))))) {
		lkflags &= ~LK_SHARED;
		lkflags |= LK_EXCLUSIVE;
	}
	lkflags |= LK_NODDLKTREAT;
	return (lkflags);
}

static int needs_exclusive_leaf(struct mount *mp, int flags) {
	/*
	 * Intermediate nodes can use shared locks, we only need to
	 * force an exclusive lock for leaf nodes.
	 */
	if ((flags & (ISLASTCN | LOCKLEAF)) != (ISLASTCN | LOCKLEAF))
		return (0);

	/* Always use exclusive locks if LOCKSHARED isn't set. */
	if (!(flags & LOCKSHARED))
		return (1);

	/*
	 * For lookups during open(), if the mount point supports
	 * extended shared operations, then use a shared lock for the
	 * leaf node, otherwise use an exclusive lock.
	 */
	if ((flags & ISOPEN) != 0)
		return (!MNT_EXTENDED_SHARED(mp));

	/*
	 * Lookup requests outside of open() that specify LOCKSHARED
	 * only need a shared lock on the leaf vnode.
	 */
	return (0);
}

#if 0
struct nameicap_tracker {
	struct vnode *dp;
	TAILQ_ENTRY(nameicap_tracker) nm_link;
};

static void nameicap_cleanup(struct nameidata *ndp)
{
	struct nameicap_tracker *nt, *nt1;

	KASSERT(TAILQ_EMPTY(&ndp->ni_cap_tracker) ||
	    (ndp->ni_lcf & NI_LCF_CAP_DOTDOT) != 0, ("not strictrelative"));
	TAILQ_FOREACH_SAFE(nt, &ndp->ni_cap_tracker, nm_link, nt1) {
		TAILQ_REMOVE(&ndp->ni_cap_tracker, nt, nm_link);
		vdrop(nt->dp);
		uma_zfree(nt_zone, nt);
	}
}
#endif

static void namei_cleanup_cnp(struct componentname *cnp) {
	uma_zfree(namei_zone, cnp->cn_pnbuf);
}

static int namei_handle_root(struct nameidata *ndp, struct vnode **dpp) {
	struct componentname *cnp = &ndp->ni_cnd;
#ifdef NI_LCF_STRICTRELATIVE
	if ((ndp->ni_lcf & NI_LCF_STRICTRELATIVE) != 0) {
#ifdef KTRACE
		if (KTRPOINT(curthread, KTR_CAPFAIL))
			ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
#endif
		printf("vfsroot debug: namei_handle_root: ENOTCAPABLE\n");
		return (ENOTCAPABLE);
	}
#endif
	while (*(cnp->cn_nameptr) == '/') {
		cnp->cn_nameptr++;
		ndp->ni_pathlen--;
	}
	*dpp = ndp->ni_rootdir;
	vrefact(*dpp);
	return (0);
}

static int relaxed_lookup(struct nameidata *ndp, int lookup_shared) {
	char *cp;			/* pointer into pathname argument */
	char *prev_ni_next;		/* saved ndp->ni_next */
	struct vnode *dp = NULL;	/* the directory we are searching */
	struct vnode *tdp;		/* saved dp */
	struct mount *mp;		/* mount table entry */
	struct prison *pr;
	size_t prev_ni_pathlen;		/* saved ndp->ni_pathlen */
	int docache;			/* == 0 do not cache last component */
	int wantparent;			/* 1 => wantparent or lockparent flag */
	int rdonly;			/* lookup read-only flag bit */
	int error = 0;
	int dpunlocked = 0;		/* dp has already been unlocked */
	int relookup = 0;		/* do not consume the path component */
	struct componentname *cnp = &ndp->ni_cnd;
	int lkflags_save;
	int ni_dvp_unlocked;
	int lookup_cap_dotdot_nonlocal = 1;	// TODO: get from sysctl
	/*
	 * Setup: break out flag bits into variables.
	 */
	ni_dvp_unlocked = 0;
	wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
	KASSERT(cnp->cn_nameiop == LOOKUP || wantparent,
	    ("CREATE, DELETE, RENAME require LOCKPARENT or WANTPARENT."));
	docache = (cnp->cn_flags & NOCACHE) ^ NOCACHE;
	if (cnp->cn_nameiop == DELETE ||
	    (wantparent && cnp->cn_nameiop != CREATE &&
	     cnp->cn_nameiop != LOOKUP)) {
		docache = 0;
	}
	rdonly = cnp->cn_flags & RDONLY;
	cnp->cn_flags &= ~ISSYMLINK;
	ndp->ni_dvp = NULL;
	/*
	 * We use shared locks until we hit the parent of the last cn then
	 * we adjust based on the requesting flags.
	 */
	cnp->cn_lkflags = lookup_shared ? LK_SHARED : LK_EXCLUSIVE;
	dp = ndp->ni_startdir;
	ndp->ni_startdir = NULLVP;
	vn_lock(dp,
	    compute_cn_lkflags(dp->v_mount, cnp->cn_lkflags | LK_RETRY,
	    cnp->cn_flags));

dirloop:
	/*
	 * Search a new directory.
	 *
	 * The last component of the filename is left accessible via
	 * cnp->cn_nameptr for callers that need the name. Callers needing
	 * the name set the SAVENAME flag. When done, they assume
	 * responsibility for freeing the pathname buffer.
	 */
	for (cp = cnp->cn_nameptr; *cp != 0 && *cp != '/'; cp++)
		continue;
	cnp->cn_namelen = cp - cnp->cn_nameptr;
	if (cnp->cn_namelen > NAME_MAX) {
		error = ENAMETOOLONG;
		goto bad;
	}
	prev_ni_pathlen = ndp->ni_pathlen;
	ndp->ni_pathlen -= cnp->cn_namelen;
	KASSERT(ndp->ni_pathlen <= PATH_MAX,
	    ("%s: ni_pathlen underflow to %zd\n", __func__, ndp->ni_pathlen));
	prev_ni_next = ndp->ni_next;
	ndp->ni_next = cp;

	/*
	 * Replace multiple slashes by a single slash and trailing slashes
	 * by a null.  This must be done before VOP_LOOKUP() because some
	 * fs's don't know about trailing slashes.  Remember if there were
	 * trailing slashes to handle symlinks, existing non-directories
	 * and non-existing files that won't be directories specially later.
	 */
	while (*cp == '/' && (cp[1] == '/' || cp[1] == '\0')) {
		cp++;
		ndp->ni_pathlen--;
		if (*cp == '\0') {
			*ndp->ni_next = '\0';
			cnp->cn_flags |= TRAILINGSLASH;
		}
	}
	ndp->ni_next = cp;

	cnp->cn_flags |= MAKEENTRY;
	if (*cp == '\0' && docache == 0)
		cnp->cn_flags &= ~MAKEENTRY;
	if (cnp->cn_namelen == 2 &&
	    cnp->cn_nameptr[1] == '.' && cnp->cn_nameptr[0] == '.') {
		cnp->cn_flags |= ISDOTDOT;
	} else {
		cnp->cn_flags &= ~ISDOTDOT;
	}
	if (*ndp->ni_next == 0) {
		cnp->cn_flags |= ISLASTCN;
	} else {
		cnp->cn_flags &= ~ISLASTCN;
	}

#if 0
	if ((cnp->cn_flags & ISLASTCN) != 0 &&
	    cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.' &&
	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
		error = EINVAL;
		goto bad;
	}
#endif

	//nameicap_tracker_add(ndp, dp);

	/*
	 * Check for degenerate name (e.g. / or "")
	 * which is a way of talking about a directory,
	 * e.g. like "/." or ".".
	 */
	if (cnp->cn_nameptr[0] == '\0') {
		if (dp->v_type != VDIR) {
			error = ENOTDIR;
			goto bad;
		}
		if (cnp->cn_nameiop != LOOKUP) {
			error = EISDIR;
			goto bad;
		}
		if (wantparent) {
			ndp->ni_dvp = dp;
			VREF(dp);
		}
		ndp->ni_vp = dp;

		if (cnp->cn_flags & AUDITVNODE1)
			AUDIT_ARG_VNODE1(dp);
		else if (cnp->cn_flags & AUDITVNODE2)
			AUDIT_ARG_VNODE2(dp);

		if (!(cnp->cn_flags & (LOCKPARENT | LOCKLEAF)))
			VOP_UNLOCK(dp, 0);
		/* XXX This should probably move to the top of function. */
		if (cnp->cn_flags & SAVESTART)
			panic("lookup: SAVESTART");
		goto success;
	}

	/*
	 * Handle "..": five special cases.
	 * 0. If doing a capability lookup and lookup_cap_dotdot is
	 *    disabled, return ENOTCAPABLE.
	 * 1. Return an error if this is the last component of
	 *    the name and the operation is DELETE or RENAME.
	 * 2. If at root directory (e.g. after chroot)
	 *    or at absolute root directory
	 *    then ignore it so can't get out.
	 * 3. If this vnode is the root of a mounted
	 *    filesystem, then replace it with the
	 *    vnode which was mounted on so we take the
	 *    .. in the other filesystem.
	 * 4. If the vnode is the top directory of
	 *    the jail or chroot, don't let them out.
	 * 5. If doing a capability lookup and lookup_cap_dotdot is
	 *    enabled, return ENOTCAPABLE if the lookup would escape
	 *    from the initial file descriptor directory.  Checks are
	 *    done by ensuring that namei() already traversed the
	 *    result of dotdot lookup.
	 */
	if (cnp->cn_flags & ISDOTDOT) {
#ifdef NI_LCF_STRICTRELATIVE
		if ((ndp->ni_lcf & (NI_LCF_STRICTRELATIVE | NI_LCF_CAP_DOTDOT))
		    == NI_LCF_STRICTRELATIVE) {
#ifdef KTRACE
			if (KTRPOINT(curthread, KTR_CAPFAIL))
				ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
#endif
			error = ENOTCAPABLE;
			printf("vfsroot debug: lookup: ENOTCAPABLE\n");
			goto bad;
		}
#endif
#if 0
		if ((cnp->cn_flags & ISLASTCN) != 0 &&
		    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
			error = EINVAL;
			goto bad;
		}
#endif
		for (;;) {
			for (pr = cnp->cn_cred->cr_prison; pr != NULL;
			     pr = pr->pr_parent) {
				if (dp == pr->pr_root) break;
			}
			if (dp == ndp->ni_rootdir || 
			    dp == ndp->ni_topdir || 
			    dp == rootvnode ||
			    pr != NULL ||
			    ((dp->v_vflag & VV_ROOT) != 0 &&
			     (cnp->cn_flags & NOCROSSMOUNT) != 0)) {
				ndp->ni_dvp = dp;
				ndp->ni_vp = dp;
				VREF(dp);
				goto nextname;
			}
			if ((dp->v_vflag & VV_ROOT) == 0)
				break;
			if (dp->v_iflag & VI_DOOMED) {	/* forced unmount */
				error = ENOENT;
				goto bad;
			}
			tdp = dp;
			dp = dp->v_mount->mnt_vnodecovered;
			VREF(dp);
			vput(tdp);
			vn_lock(dp,
			    compute_cn_lkflags(dp->v_mount, cnp->cn_lkflags |
			    LK_RETRY, ISDOTDOT));
#if 0
			error = nameicap_check_dotdot(ndp, dp, lookup_cap_dotdot_nonlocal);
			if (error != 0) {
				printf("vfsunlink debug: %s:%u: error = %u\n", __FILE__, __LINE__, error);
#ifdef KTRACE
				if (KTRPOINT(curthread, KTR_CAPFAIL))
					ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
#endif
				goto bad;
			}
#endif
		}
	}

	/*
	 * We now have a segment name to search for, and a directory to search.
	 */
unionlookup:
#ifdef MAC
	if ((cnp->cn_flags & NOMACCHECK) == 0) {
		error = mac_vnode_check_lookup(cnp->cn_thread->td_ucred, dp,
		    cnp);
		if (error)
			goto bad;
	}
#endif
	ndp->ni_dvp = dp;
	ndp->ni_vp = NULL;
	ASSERT_VOP_LOCKED(dp, "lookup");
	/*
	 * If we have a shared lock we may need to upgrade the lock for the
	 * last operation.
	 */
	if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN) &&
	    VOP_ISLOCKED(dp) == LK_SHARED) {
		vn_lock(dp, LK_UPGRADE|LK_RETRY);
	}
	if ((dp->v_iflag & VI_DOOMED) != 0) {
		error = ENOENT;
		goto bad;
	}
	/*
	 * If we're looking up the last component and we need an exclusive
	 * lock, adjust our lkflags.
	 */
	if (needs_exclusive_leaf(dp->v_mount, cnp->cn_flags)) {
		cnp->cn_lkflags = LK_EXCLUSIVE;
	}
#ifdef NAMEI_DIAGNOSTIC
	vn_printf(dp, "lookup in ");
#endif
	lkflags_save = cnp->cn_lkflags;
	cnp->cn_lkflags = compute_cn_lkflags(dp->v_mount, cnp->cn_lkflags,
	    cnp->cn_flags);
	error = VOP_LOOKUP(dp, &ndp->ni_vp, cnp);
	cnp->cn_lkflags = lkflags_save;
	if (error != 0) {
		KASSERT(ndp->ni_vp == NULL, ("leaf should be empty"));
#ifdef NAMEI_DIAGNOSTIC
		printf("not found\n");
#endif
		if ((error == ENOENT) &&
		    (dp->v_vflag & VV_ROOT) && (dp->v_mount != NULL) &&
		    (dp->v_mount->mnt_flag & MNT_UNION)) {
			tdp = dp;
			dp = dp->v_mount->mnt_vnodecovered;
			VREF(dp);
			vput(tdp);
			vn_lock(dp,
			    compute_cn_lkflags(dp->v_mount, cnp->cn_lkflags |
			    LK_RETRY, cnp->cn_flags));
			//nameicap_tracker_add(ndp, dp);
			goto unionlookup;
		}

#ifdef ERELOOKUP
		if (error == ERELOOKUP) {
			vref(dp);
			ndp->ni_vp = dp;
			error = 0;
			relookup = 1;
			goto good;
		}
#endif

		if (error != EJUSTRETURN)
			goto bad;
		/*
		 * At this point, we know we're at the end of the
		 * pathname.  If creating / renaming, we can consider
		 * allowing the file or directory to be created / renamed,
		 * provided we're not on a read-only filesystem.
		 */
		if (rdonly) {
			error = EROFS;
			goto bad;
		}
		/* trailing slash only allowed for directories */
		if ((cnp->cn_flags & TRAILINGSLASH) &&
		    !(cnp->cn_flags & WILLBEDIR)) {
			error = ENOENT;
			goto bad;
		}
		if ((cnp->cn_flags & LOCKPARENT) == 0)
			VOP_UNLOCK(dp, 0);
		/*
		 * We return with ni_vp NULL to indicate that the entry
		 * doesn't currently exist, leaving a pointer to the
		 * (possibly locked) directory vnode in ndp->ni_dvp.
		 */
		if (cnp->cn_flags & SAVESTART) {
			ndp->ni_startdir = ndp->ni_dvp;
			VREF(ndp->ni_startdir);
		}
		goto success;
	}

good:
#ifdef NAMEI_DIAGNOSTIC
	printf("found\n");
#endif
	dp = ndp->ni_vp;

#if 0
	/*
	 * Check to see if the vnode has been mounted on;
	 * if so find the root of the mounted filesystem.
	 */
	while (dp->v_type == VDIR && (mp = dp->v_mountedhere) &&
	       (cnp->cn_flags & NOCROSSMOUNT) == 0) {
		if (vfs_busy(mp, 0))
			continue;
		vput(dp);
		if (dp != ndp->ni_dvp)
			vput(ndp->ni_dvp);
		else
			vrele(ndp->ni_dvp);
		vrefact(vp_crossmp);
		ndp->ni_dvp = vp_crossmp;
		error = VFS_ROOT(mp, compute_cn_lkflags(mp, cnp->cn_lkflags,
		    cnp->cn_flags), &tdp);
		vfs_unbusy(mp);
		if (vn_lock(vp_crossmp, LK_SHARED | LK_NOWAIT))
			panic("vp_crossmp exclusively locked or reclaimed");
		if (error) {
			dpunlocked = 1;
			goto bad2;
		}
		ndp->ni_vp = dp = tdp;
	}
#endif

	/*
	 * Check for symbolic link
	 */
	if ((dp->v_type == VLNK) &&
	    ((cnp->cn_flags & FOLLOW) || (cnp->cn_flags & TRAILINGSLASH) ||
	     *ndp->ni_next == '/')) {
		cnp->cn_flags |= ISSYMLINK;
		if (dp->v_iflag & VI_DOOMED) {
			/*
			 * We can't know whether the directory was mounted with
			 * NOSYMFOLLOW, so we can't follow safely.
			 */
			error = ENOENT;
			goto bad2;
		}
		if (dp->v_mount->mnt_flag & MNT_NOSYMFOLLOW) {
			error = EACCES;
			goto bad2;
		}
		/*
		 * Symlink code always expects an unlocked dvp.
		 */
		if (ndp->ni_dvp != ndp->ni_vp) {
			VOP_UNLOCK(ndp->ni_dvp, 0);
			ni_dvp_unlocked = 1;
		}
		goto success;
	}

nextname:
	/*
	 * Not a symbolic link that we will follow.  Continue with the
	 * next component if there is any; otherwise, we're done.
	 */
	KASSERT((cnp->cn_flags & ISLASTCN) || *ndp->ni_next == '/',
	    ("lookup: invalid path state."));
	if (relookup) {
		relookup = 0;
		ndp->ni_pathlen = prev_ni_pathlen;
		ndp->ni_next = prev_ni_next;
		if (ndp->ni_dvp != dp)
			vput(ndp->ni_dvp);
		else
			vrele(ndp->ni_dvp);
		goto dirloop;
	}
#if 0
	if (cnp->cn_flags & ISDOTDOT) {
		error = nameicap_check_dotdot(ndp, ndp->ni_vp, lookup_cap_dotdot_nonlocal);
		if (error != 0) {
			printf("vfsunlink debug: %s:%u: error = %u\n", __FILE__, __LINE__, error);
#ifdef KTRACE
			if (KTRPOINT(curthread, KTR_CAPFAIL))
				ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
#endif
			goto bad2;
		}
	}
#endif
	if (*ndp->ni_next == '/') {
		cnp->cn_nameptr = ndp->ni_next;
		while (*cnp->cn_nameptr == '/') {
			cnp->cn_nameptr++;
			ndp->ni_pathlen--;
		}
		if (ndp->ni_dvp != dp)
			vput(ndp->ni_dvp);
		else
			vrele(ndp->ni_dvp);
		goto dirloop;
	}
	/*
	 * If we're processing a path with a trailing slash,
	 * check that the end result is a directory.
	 */
	if ((cnp->cn_flags & TRAILINGSLASH) && dp->v_type != VDIR) {
		error = ENOTDIR;
		goto bad2;
	}
	/*
	 * Disallow directory write attempts on read-only filesystems.
	 */
	if (rdonly &&
	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
		error = EROFS;
		goto bad2;
	}
	if (cnp->cn_flags & SAVESTART) {
		ndp->ni_startdir = ndp->ni_dvp;
		VREF(ndp->ni_startdir);
	}
	if (!wantparent) {
		ni_dvp_unlocked = 2;
		if (ndp->ni_dvp != dp)
			vput(ndp->ni_dvp);
		else
			vrele(ndp->ni_dvp);
	} else if ((cnp->cn_flags & LOCKPARENT) == 0 && ndp->ni_dvp != dp) {
		VOP_UNLOCK(ndp->ni_dvp, 0);
		ni_dvp_unlocked = 1;
	}

	if (cnp->cn_flags & AUDITVNODE1)
		AUDIT_ARG_VNODE1(dp);
	else if (cnp->cn_flags & AUDITVNODE2)
		AUDIT_ARG_VNODE2(dp);

	if ((cnp->cn_flags & LOCKLEAF) == 0)
		VOP_UNLOCK(dp, 0);
success:
	/*
	 * Because of lookup_shared we may have the vnode shared locked, but
	 * the caller may want it to be exclusively locked.
	 */
	if (needs_exclusive_leaf(dp->v_mount, cnp->cn_flags) &&
	    VOP_ISLOCKED(dp) != LK_EXCLUSIVE) {
		vn_lock(dp, LK_UPGRADE | LK_RETRY);
		if (dp->v_iflag & VI_DOOMED) {
			error = ENOENT;
			goto bad2;
		}
	}
	return (0);

bad2:
	if (ni_dvp_unlocked != 2) {
		if (dp != ndp->ni_dvp && !ni_dvp_unlocked) {
			vput(ndp->ni_dvp);
		} else {
			vrele(ndp->ni_dvp);
		}
	}
bad:
	if (!dpunlocked)
		vput(dp);
	ndp->ni_vp = NULL;
	return (error);
}

static int relaxed_namei(struct nameidata *ndp) {
	struct filedesc *fdp;	/* pointer to file descriptor state */
	char *cp;		/* pointer into pathname argument */
	struct vnode *dp;	/* the directory we are searching */
	struct iovec aiov;		/* uio for reading symbolic links */
	struct componentname *cnp;
	struct thread *td;
	struct proc *p;
	cap_rights_t rights;
	struct uio auio;
	int error, linklen, startdir_used;

	int lookup_shared = 1;		// TODO: get from sysctl
	int lookup_cap_dotdot = 1;	// TODO: get from sysctl

	cnp = &ndp->ni_cnd;
	td = cnp->cn_thread;
	p = td->td_proc;
	ndp->ni_cnd.cn_cred = ndp->ni_cnd.cn_thread->td_ucred;
	KASSERT(cnp->cn_cred && p, ("namei: bad cred/proc"));
	KASSERT((cnp->cn_nameiop & (~OPMASK)) == 0,
	    ("namei: nameiop contaminated with flags"));
	KASSERT((cnp->cn_flags & OPMASK) == 0,
	    ("namei: flags contaminated with nameiops"));
	MPASS(ndp->ni_startdir == NULL || ndp->ni_startdir->v_type == VDIR ||
	    ndp->ni_startdir->v_type == VBAD);
	if (!lookup_shared)
		cnp->cn_flags &= ~LOCKSHARED;
	fdp = p->p_fd;
#ifdef NI_LCF_STRICTRELATIVE
	TAILQ_INIT(&ndp->ni_cap_tracker);
	ndp->ni_lcf = 0;
#endif

	/* We will set this ourselves if we need it. */
	cnp->cn_flags &= ~TRAILINGSLASH;

	/*
	 * Get a buffer for the name to be translated, and copy the
	 * name into the buffer.
	 */
	if ((cnp->cn_flags & HASBUF) == 0) {
		cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
	}
	if (ndp->ni_segflg == UIO_SYSSPACE) {
		error = copystr(ndp->ni_dirp, cnp->cn_pnbuf, MAXPATHLEN,
		    &ndp->ni_pathlen);
	} else {
		error = copyinstr(ndp->ni_dirp, cnp->cn_pnbuf, MAXPATHLEN,
		    &ndp->ni_pathlen);
	}

	/*
	 * Don't allow empty pathnames.
	 */
	if (error == 0 && *cnp->cn_pnbuf == '\0')
		error = ENOENT;

#ifdef CAPABILITY_MODE
	/*
	 * In capability mode, lookups must be restricted to happen in
	 * the subtree with the root specified by the file descriptor:
	 * - The root must be real file descriptor, not the pseudo-descriptor
	 *   AT_FDCWD.
	 * - The passed path must be relative and not absolute.
	 * - If lookup_cap_dotdot is disabled, path must not contain the
	 *   '..' components.
	 * - If lookup_cap_dotdot is enabled, we verify that all '..'
	 *   components lookups result in the directories which were
	 *   previously walked by us, which prevents an escape from
	 *   the relative root.
	 */
	if (error == 0 && IN_CAPABILITY_MODE(td) &&
	    (cnp->cn_flags & NOCAPCHECK) == 0) {
#ifdef NI_LCF_STRICTRELATIVE
		ndp->ni_lcf |= NI_LCF_STRICTRELATIVE;
#endif
		if (ndp->ni_dirfd == AT_FDCWD) {
#ifdef KTRACE
			if (KTRPOINT(td, KTR_CAPFAIL)) {
				ktrcapfail(CAPFAIL_LOOKUP, NULL, NULL);
			}
#endif
			error = ECAPMODE;
		}
	}
#endif
	if (error != 0) {
		namei_cleanup_cnp(cnp);
		ndp->ni_vp = NULL;
		return (error);
	}
	ndp->ni_loopcnt = 0;
#ifdef KTRACE
	if (KTRPOINT(td, KTR_NAMEI)) {
		KASSERT(cnp->cn_thread == curthread,
		    ("namei not using curthread"));
		ktrnamei(cnp->cn_pnbuf);
	}
#endif
	/*
	 * Get starting point for the translation.
	 */
	FILEDESC_SLOCK(fdp);
	ndp->ni_rootdir = fdp->fd_rdir;
	vrefact(ndp->ni_rootdir);
	ndp->ni_topdir = fdp->fd_jdir;

	/*
	 * If we are auditing the kernel pathname, save the user pathname.
	 */
	if (cnp->cn_flags & AUDITVNODE1) {
		AUDIT_ARG_UPATH1(td, ndp->ni_dirfd, cnp->cn_pnbuf);
	}
	if (cnp->cn_flags & AUDITVNODE2) {
		AUDIT_ARG_UPATH2(td, ndp->ni_dirfd, cnp->cn_pnbuf);
	}

	startdir_used = 0;
	dp = NULL;
	cnp->cn_nameptr = cnp->cn_pnbuf;
	if (cnp->cn_pnbuf[0] == '/') {
		error = namei_handle_root(ndp, &dp);
	} else {
		if (ndp->ni_startdir != NULL) {
			dp = ndp->ni_startdir;
			startdir_used = 1;
		} else if (ndp->ni_dirfd == AT_FDCWD) {
			dp = fdp->fd_cdir;
			vrefact(dp);
		} else {
			rights = ndp->ni_rightsneeded;
			cap_rights_set(&rights, CAP_LOOKUP);

			if (cnp->cn_flags & AUDITVNODE1)
				AUDIT_ARG_ATFD1(ndp->ni_dirfd);
			if (cnp->cn_flags & AUDITVNODE2)
				AUDIT_ARG_ATFD2(ndp->ni_dirfd);
			error = fgetvp_rights(td, ndp->ni_dirfd,
			    &rights, &ndp->ni_filecaps, &dp);
			if (error == EINVAL)
				error = ENOTDIR;
#if defined NI_LCF_STRICTRELATIVE && defined CAPABILITIES
			/*
			 * If file descriptor doesn't have all rights,
			 * all lookups relative to it must also be
			 * strictly relative.
			 */
			CAP_ALL(&rights);
			if (!cap_rights_contains(&ndp->ni_filecaps.fc_rights,
			    &rights) ||
			    ndp->ni_filecaps.fc_fcntls != CAP_FCNTL_ALL ||
			    ndp->ni_filecaps.fc_nioctls != -1) {
				ndp->ni_lcf |= NI_LCF_STRICTRELATIVE;
			}
#endif
		}
		if (error == 0 && dp->v_type != VDIR)
			error = ENOTDIR;
	}
	FILEDESC_SUNLOCK(fdp);
	if (ndp->ni_startdir != NULL && !startdir_used) {
		vrele(ndp->ni_startdir);
	}
	if (error != 0) {
		if (dp != NULL)
			vrele(dp);
		goto out;
	}
#ifdef NI_LCF_STRICTRELATIVE
	if ((ndp->ni_lcf & NI_LCF_STRICTRELATIVE) != 0 &&
	    lookup_cap_dotdot != 0) {
		ndp->ni_lcf |= NI_LCF_CAP_DOTDOT;
	}
#endif
	SDT_PROBE3(vfs, namei, lookup, entry, dp, cnp->cn_pnbuf,
	    cnp->cn_flags);
	for (;;) {
		ndp->ni_startdir = dp;
		error = relaxed_lookup(ndp, lookup_shared);
		if (error != 0)
			goto out;
		/*
		 * If not a symbolic link, we're done.
		 */
		if ((cnp->cn_flags & ISSYMLINK) == 0) {
			vrele(ndp->ni_rootdir);
			if ((cnp->cn_flags & (SAVENAME | SAVESTART)) == 0) {
				namei_cleanup_cnp(cnp);
			} else {
				cnp->cn_flags |= HASBUF;
			}
			//nameicap_cleanup(ndp);
			SDT_PROBE2(vfs, namei, lookup, return, 0, ndp->ni_vp);
			return (0);
		}
		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
			error = ELOOP;
			break;
		}
#ifdef MAC
		if ((cnp->cn_flags & NOMACCHECK) == 0) {
			error = mac_vnode_check_readlink(td->td_ucred,
			    ndp->ni_vp);
			if (error != 0)
				break;
		}
#endif
		if (ndp->ni_pathlen > 1) {
			cp = uma_zalloc(namei_zone, M_WAITOK);
		} else {
			cp = cnp->cn_pnbuf;
		}
		aiov.iov_base = cp;
		aiov.iov_len = MAXPATHLEN;
		auio.uio_iov = &aiov;
		auio.uio_iovcnt = 1;
		auio.uio_offset = 0;
		auio.uio_rw = UIO_READ;
		auio.uio_segflg = UIO_SYSSPACE;
		auio.uio_td = td;
		auio.uio_resid = MAXPATHLEN;
		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
		if (error != 0) {
			if (ndp->ni_pathlen > 1) {
				uma_zfree(namei_zone, cp);
			}
			break;
		}
		linklen = MAXPATHLEN - auio.uio_resid;
		if (linklen == 0) {
			if (ndp->ni_pathlen > 1) {
				uma_zfree(namei_zone, cp);
			}
			error = ENOENT;
			break;
		}
		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
			if (ndp->ni_pathlen > 1)
				uma_zfree(namei_zone, cp);
			error = ENAMETOOLONG;
			break;
		}
		if (ndp->ni_pathlen > 1) {
			bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
			uma_zfree(namei_zone, cnp->cn_pnbuf);
			cnp->cn_pnbuf = cp;
		} else {
			cnp->cn_pnbuf[linklen] = '\0';
		}
		ndp->ni_pathlen += linklen;
		vput(ndp->ni_vp);
		dp = ndp->ni_dvp;
		/*
		 * Check if root directory should replace current directory.
		 */
		cnp->cn_nameptr = cnp->cn_pnbuf;
		if (*(cnp->cn_nameptr) == '/') {
			vrele(dp);
			error = namei_handle_root(ndp, &dp);
			if (error != 0)
				goto out;
		}
	}
	vput(ndp->ni_vp);
	ndp->ni_vp = NULL;
	vrele(ndp->ni_dvp);
out:
	vrele(ndp->ni_rootdir);
	namei_cleanup_cnp(cnp);
	//nameicap_cleanup(ndp);
	SDT_PROBE2(vfs, namei, lookup, return, error, NULL);
	return (error);
}

static void print_ufs_dir_info(struct vnode *dv) {
	const struct ufs_inode *di = UFS_VTOI(dv);
	printf("i_number=%u, i_flag=0x%x, i_effnlink=%d, i_count=%d, i_endoff=%d, i_diroff=%d, i_offset=%d, i_nextclustercg=%d, i_size=%llu, i_gen=%llu\n",
		(unsigned int)di->i_number, (unsigned int)di->i_flag, di->i_effnlink,
		(int)di->i_count, (int)di->i_endoff, (int)di->i_diroff, (int)di->i_offset,
		di->i_nextclustercg, (unsigned long long int)di->i_size, (unsigned long long int)di->i_gen);
}

static int handle_unlink(SYSCTL_HANDLER_ARGS) {
	if(req->oldptr) return ENOTSUP;
	if(!req->newptr) return 0;
	arg2 = req->newlen - req->newidx;
	//uprintf("arg2 = %ld\n", (long int)arg2);
	char path[arg2+1];
	int e = SYSCTL_IN(req, path, arg2);
	path[arg2] = 0;
	if(e) return e;
	//uprintf("path = \"%s\"\n", path);

	struct thread *td = curthread;
	struct mount *mp;
	struct nameidata nd;
	cap_rights_t rights;
restart:
	bwillwrite();
#if 1
	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1, UIO_SYSSPACE,
		path, AT_FDCWD, cap_rights_init(&rights, CAP_UNLINKAT), td);
#else
	NDINIT_ATRIGHTS(&nd, 0, LOCKPARENT | LOCKLEAF | AUDITVNODE1, UIO_SYSSPACE,
		path, AT_FDCWD, cap_rights_init(&rights, CAP_UNLINKAT), td);
#endif
	e = relaxed_namei(&nd);
	//uprintf("e = %d\n", e);
	if(e) return e == EINVAL ? EPERM : e;
	struct vnode *vp = nd.ni_vp;
	if(vp->v_type == VDIR) e = priv_check(td, PRIV_VFS_ADMIN);
	if(!e) {
		e = vn_start_write(nd.ni_dvp, &mp, V_NOWAIT);
		if(e) {
			NDFREE(&nd, NDF_ONLY_PNBUF);
			vput(nd.ni_dvp);
			if(vp == nd.ni_dvp) vrele(vp);
			else vput(vp);
			e = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
			if(e) return e;
			goto restart;
		}
#if 1
		if(strncmp(mp->mnt_stat.f_fstypename, "ufs", MFSNAMELEN) == 0) {
			print_ufs_dir_info(nd.ni_dvp);
			struct vnode *v;
			ino_t ino;
			e = ufs_lookup_ino(nd.ni_dvp, &v, &nd.ni_cnd, &ino);
#if 1
			if(e) printf("ufs_lookup_ino failed: error %d\n", e);
			else printf("ino = %u\n", (unsigned int)ino);
#else
			if(e) {
				printf("ufs_lookup_ino failed: error %d\n", e);
				goto cleanup;
			}
#endif
			print_ufs_dir_info(nd.ni_dvp);
		}
#endif
#ifdef MAC
		e = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp, &nd.ni_cnd);
		if(e) goto skip;
#endif
		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
		e = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
#ifdef MAC
skip:
#endif
		vn_finished_write(mp);
	}
cleanup:
	NDFREE(&nd, NDF_ONLY_PNBUF);
	vput(nd.ni_dvp);
	if(vp == nd.ni_dvp) vrele(vp);
	else vput(vp);
	return e;
}

SYSCTL_OID(_vfs, OID_AUTO, unlink, CTLTYPE_STRING | CTLFLAG_WR, NULL, 0, handle_unlink, "A", "Unlink by path");

static int vfsunlink_module_event(module_t mod, int type, void *data) {
	switch(type) {
		case MOD_LOAD:
			break;
		case MOD_UNLOAD:
			break;
		case MOD_SHUTDOWN:
			break;
		default:
			return EOPNOTSUPP;
	}
	return 0;
}

static moduledata_t module_data = { "vfsunlink", vfsunlink_module_event, NULL };
DECLARE_MODULE(vfsunlink, module_data, SI_SUB_VFS, SI_ORDER_ANY);
MODULE_VERSION(vfsunlink, 1);