diff --git a/lib/libutil/mntopts.c b/lib/libutil/mntopts.c
index 4a064a086fd5..0412007ebcdf 100644
--- a/lib/libutil/mntopts.c
+++ b/lib/libutil/mntopts.c
@@ -138,7 +138,7 @@ checkpath_allow_file(const char *path, char *resolved)
 
 	if (realpath(path, resolved) == NULL || stat(resolved, &sb) != 0)
 		return (1);
-	if (!S_ISDIR(sb.st_mode) && !S_ISREG(sb.st_mode)) {
+	if (!S_ISDIR(sb.st_mode) && !S_ISREG(sb.st_mode) && !S_ISSOCK(sb.st_mode)) {
 		errno = ENOTDIR;
 		return (1);
 	}
diff --git a/sbin/mount_nullfs/mount_nullfs.8 b/sbin/mount_nullfs/mount_nullfs.8
index b3cf57fd9dea..2ea14dca2259 100644
--- a/sbin/mount_nullfs/mount_nullfs.8
+++ b/sbin/mount_nullfs/mount_nullfs.8
@@ -30,7 +30,7 @@
 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 .\" SUCH DAMAGE.
 .\"
-.Dd March 24, 2024
+.Dd August 24, 2026
 .Dt MOUNT_NULLFS 8
 .Os
 .Sh NAME
@@ -62,7 +62,9 @@ but in other respects it is indistinguishable from the original.
 .Pp
 The
 .Nm
-utility supports mounting both directories and single files.
+utility supports mounting directories, regular files and
+.Xr unix 4
+sockets.
 Both
 .Ar target
 and
diff --git a/sbin/mount_nullfs/mount_nullfs.c b/sbin/mount_nullfs/mount_nullfs.c
index fc04961e6247..d1096bd1bad3 100644
--- a/sbin/mount_nullfs/mount_nullfs.c
+++ b/sbin/mount_nullfs/mount_nullfs.c
@@ -98,13 +98,13 @@ main(int argc, char *argv[])
 		err(EX_USAGE, "%s", target);
 	if (stat_realpath(argv[1], mountpoint, &mountpoint_stat) != 0)
 		err(EX_USAGE, "%s", mountpoint);
-	if (!S_ISDIR(target_stat.st_mode) && !S_ISREG(target_stat.st_mode))
-		errx(EX_USAGE, "%s: must be either a file or directory",
+	if (!S_ISDIR(target_stat.st_mode) && !S_ISREG(target_stat.st_mode) && !S_ISSOCK(target_stat.st_mode))
+		errx(EX_USAGE, "%s: must be either a file, a socket or a directory",
 		    target);
 	if ((target_stat.st_mode & S_IFMT) !=
 	    (mountpoint_stat.st_mode & S_IFMT))
 		errx(EX_USAGE,
-		    "%s: must be same type as %s (file or directory)",
+		    "%s: must be same type as %s (file, socket or directory)",
 		    mountpoint, target);
 
 	build_iovec(&iov, &iovlen, "fstype", nullfs, (size_t)-1);
diff --git a/sys/fs/nullfs/null_vfsops.c b/sys/fs/nullfs/null_vfsops.c
index 0ec4f9c87297..b3fd266c7fe0 100644
--- a/sys/fs/nullfs/null_vfsops.c
+++ b/sys/fs/nullfs/null_vfsops.c
@@ -175,7 +175,7 @@ nullfs_mount(struct mount *mp)
 	 * Lower vnode must be the same type as the covered vnode - we
 	 * don't allow mounting directories to files or vice versa.
 	 */
-	if ((lowerrootvp->v_type != VDIR && lowerrootvp->v_type != VREG) ||
+	if ((lowerrootvp->v_type != VDIR && lowerrootvp->v_type != VREG && lowerrootvp->v_type != VSOCK) ||
 	    lowerrootvp->v_type != mp->mnt_vnodecovered->v_type) {
 		NULLFSDEBUG("nullfs_mount: target must be same type as fspath");
 		vput(lowerrootvp);
diff --git a/sys/kern/vfs_cache.c b/sys/kern/vfs_cache.c
index 9e1c9cb757b5..6692ae6facf9 100644
--- a/sys/kern/vfs_cache.c
+++ b/sys/kern/vfs_cache.c
@@ -3292,7 +3292,7 @@ kern___realpathat(struct thread *td, int fd, const char *path, char *buf,
 	if ((error = namei(&nd)) != 0)
 		return (error);
 
-	if (nd.ni_vp->v_type == VREG && nd.ni_dvp->v_type != VDIR &&
+	if ((nd.ni_vp->v_type == VREG || nd.ni_vp->v_type == VSOCK) && nd.ni_dvp->v_type != VDIR &&
 	    (nd.ni_vp->v_vflag & VV_ROOT) != 0) {
 		struct vnode *covered_vp;
 
@@ -5729,7 +5729,7 @@ cache_fplookup_climb_mount(struct cache_fpl *fpl)
 	vp = fpl->tvp;
 	vp_seqc = fpl->tvp_seqc;
 
-	VNPASS(vp->v_type == VDIR || vp->v_type == VREG || vp->v_type == VBAD, vp);
+	VNPASS(vp->v_type == VDIR || vp->v_type == VREG || vp->v_type == VSOCK || vp->v_type == VBAD, vp);
 	mp = atomic_load_ptr(&vp->v_mountedhere);
 	if (__predict_false(mp == NULL)) {
 		return (0);
@@ -5786,7 +5786,7 @@ cache_fplookup_cross_mount(struct cache_fpl *fpl)
 	vp = fpl->tvp;
 	vp_seqc = fpl->tvp_seqc;
 
-	VNPASS(vp->v_type == VDIR || vp->v_type == VREG || vp->v_type == VBAD, vp);
+	VNPASS(vp->v_type == VDIR || vp->v_type == VREG || vp->v_type == VSOCK || vp->v_type == VBAD, vp);
 	mp = atomic_load_ptr(&vp->v_mountedhere);
 	if (__predict_false(mp == NULL)) {
 		return (0);
diff --git a/sys/kern/vfs_mount.c b/sys/kern/vfs_mount.c
index 94720615bc3f..265e95ee0a75 100644
--- a/sys/kern/vfs_mount.c
+++ b/sys/kern/vfs_mount.c
@@ -1171,12 +1171,12 @@ vfs_domount_first(
 	if (error == 0)
 		error = vinvalbuf(vp, V_SAVE, 0, 0);
 	if (vfsp->vfc_flags & VFCF_FILEMOUNT) {
-		if (error == 0 && vp->v_type != VDIR && vp->v_type != VREG)
+		if (error == 0 && vp->v_type != VDIR && vp->v_type != VREG && vp->v_type != VSOCK)
 			error = EINVAL;
 		/*
 		 * For file mounts, ensure that there is only one hardlink to the file.
 		 */
-		if (error == 0 && vp->v_type == VREG && va.va_nlink != 1)
+		if (error == 0 && (vp->v_type == VREG || vp->v_type == VSOCK) && va.va_nlink != 1)
 			error = EINVAL;
 	} else {
 		if (error == 0 && vp->v_type != VDIR)
@@ -1700,7 +1700,7 @@ vfs_domount(
 	 * Don't allow stacking file mounts to work around problems with the way
 	 * that namei sets nd.ni_dvp to vp_crossmp for these.
 	 */
-	if (vp->v_type == VREG)
+	if (vp->v_type == VREG || vp->v_type == VSOCK)
 		fsflags |= MNT_NOCOVER;
 	if ((fsflags & MNT_UPDATE) == 0) {
 		if ((vp->v_vflag & VV_ROOT) != 0 &&