Return vfs_mount to stage1

It is needed for all the probing Windows does to issue
the stage2 mount request. Without zfsvfs it would just
reply with errors, and mount would not happen.

Instead, add vfs_changeowner to move vnodes over from
dcb to vcb.

module/os/windows/spl/spl-vnode.c

@@ -2084,3 +2084,18 @@ vnode_pager_setsize(void *fo, vnode_t *vp, uint64_t size, boolean_t delay)
 	}
 
 }
+
+void
+vfs_changeowner(mount_t *from, mount_t *to)
+{
+	struct vnode *rvp;
+	mutex_enter(&vnode_all_list_lock);
+	for (rvp = list_head(&vnode_all_list);
+	    rvp;
+	    rvp = list_next(&vnode_all_list, rvp)) {
+		if (rvp->v_mount == from)
+			rvp->v_mount = to;
+
+	}
+	mutex_exit(&vnode_all_list_lock);
+}

module/os/windows/zfs/zfs_vnops_windows_mount.c

@@ -814,6 +814,13 @@ zfs_windows_mount(zfs_cmd_t *zc)
 
 	// Mount will temporarily be pointing to "dcb" until the
 	// zfs_vnop_mount() below corrects it to "vcb".
+	status = zfs_vfs_mount(zmo_dcb, NULL, (user_addr_t)&mnt_args, NULL);
+	dprintf("%s: zfs_vfs_mount() returns %d\n", __func__, status);
+
+	if (status) {
+		IoDeleteDevice(diskDeviceObject);
+		return (status);
+	}
 
 	// Check if we are to mount with driveletter, or path
 	// We already check that path is "\\??\\" above, and
@@ -1199,36 +1206,21 @@ zfs_vnop_mount(PDEVICE_OBJECT DiskDevice, PIRP Irp, PIO_STACK_LOCATION IrpSp)
 
 	volDeviceObject->Flags |= DO_BUS_ENUMERATED_DEVICE;
 
-
-	struct zfs_mount_args mnt_args;
-	char *dsname[ZFS_MAX_DATASET_NAME_LEN];
-	ULONG length = 0;
-
-	RtlUnicodeToUTF8N(dsname, ZFS_MAX_DATASET_NAME_LEN, &length,
-	    dcb->name.Buffer, dcb->name.Length);
-	dsname[length] = 0;
-
-	mnt_args.struct_size = sizeof (struct zfs_mount_args);
-	mnt_args.optlen = 0;
-	mnt_args.mflag = 0; // Set flags
-	mnt_args.fspec = dsname;
-
-	status = zfs_vfs_mount(vcb, NULL, (user_addr_t)&mnt_args, NULL);
-	dprintf("%s: zfs_vfs_mount() returns %d\n", __func__, status);
-
-	if (status) {
-		IoDeleteDevice(volDeviceObject);
-		return (status);
-	}
-	vfs_setfsprivate(dcb, vfs_fsprivate(vcb)); // HACK
-
-	zfsvfs_t *zfsvfs = vfs_fsprivate(vcb);
-	if (zfsvfs == NULL) {
+	zfsvfs_t *zfsvfs = vfs_fsprivate(dcb);
+	int giveup = 0;
+	while (zfsvfs == NULL) {
+		delay(hz / 10);
 		dprintf("zfsvfs not resolved yet\n");
-		status = STATUS_MOUNT_POINT_NOT_RESOLVED;
-		goto out;
+		zfsvfs = vfs_fsprivate(dcb);
+		if (giveup++ > 50)
+			return (STATUS_UNRECOGNIZED_VOLUME);
 	}
 	zfsvfs->z_vfs = vcb;
+	vfs_setfsprivate(vcb, zfsvfs);
+	// a bit hacky this bit, but we created some vnodes under
+	// dcb during this mount hand over, make them be owned by
+	// vcb
+	vfs_changeowner(dcb, vcb);
 
 	// Remember the parent device, so during unmount we can free both.
 	vcb->parent_device = dcb;