diff --git a/cr-restore.c b/cr-restore.c
index f7887fd82..c8a616d33 100644
--- a/cr-restore.c
+++ b/cr-restore.c
@@ -1552,6 +1552,10 @@ static void sigreturn_restore(pid_t pstree_pid, pid_t pid)
 
 	rst_mutex_init(&task_args->rst_lock);
 
+	strncpy(task_args->ns_last_pid_path,
+		LAST_PID_PATH,
+		sizeof(task_args->ns_last_pid_path));
+
 	if (pstree_entry.nr_threads) {
 		int i;
 
diff --git a/include/crtools.h b/include/crtools.h
index 8af27bd4a..e4d9e4a99 100644
--- a/include/crtools.h
+++ b/include/crtools.h
@@ -55,6 +55,7 @@ struct cr_fd_desc_tmpl {
 #define FMT_FNAME_SIGACTS	"sigacts-%d.img"
 
 #define LAST_PID_PATH		"/proc/sys/kernel/ns_last_pid"
+#define LAST_PID_PERM		0666
 
 /* file descriptors */
 struct cr_fd_desc {
diff --git a/include/restorer.h b/include/restorer.h
index 4a2dec8ee..aa1e38ece 100644
--- a/include/restorer.h
+++ b/include/restorer.h
@@ -66,7 +66,11 @@ struct task_restore_core_args {
 	int				pid;			/* task pid */
 	int				fd_core;		/* opened core file */
 	int				fd_self_vmas;		/* opened file with running VMAs to unmap */
-	char				self_vmas_path[64];	/* path to it, to unlink it once we're done */
+	union {
+		char			self_vmas_path[64];	/* path to it, to unlink it once we're done */
+		char			last_pid_buf[64];	/* internal buffer to save stack space  */
+	};
+	char				ns_last_pid_path[64];
 	bool				restore_threads;	/* if to restore threads */
 	rst_mutex_t			rst_lock;
 
@@ -231,6 +235,37 @@ static void always_inline write_num_n(long num)
 	sys_write(1, &c, 1);
 }
 
+static long always_inline vprint_num(char *buf, long num)
+{
+	unsigned long d = 1000000000000000000;
+	unsigned int started = 0;
+	unsigned int minus = 0;
+	unsigned int i = 0;
+	unsigned int c;
+
+	if (num < 0) {
+		num = -num;
+		buf[i++] = '-';
+	}
+
+	while (d) {
+		c = num / d;
+		num -= d * c;
+		d /= 10;
+		if (!c && !started)
+			continue;
+		if (!started)
+			started = 1;
+		add_ord(c);
+		buf[i++] = c;
+
+	}
+
+	buf[i++] = 0;
+
+	return i;
+}
+
 static void always_inline write_hex_n(unsigned long num)
 {
 	unsigned char *s = (unsigned char *)&num;
diff --git a/restorer.c b/restorer.c
index 5154ee2fc..62217be03 100644
--- a/restorer.c
+++ b/restorer.c
@@ -31,6 +31,14 @@ long restore_thread(long cmd, struct thread_restore_args *args)
 		struct core_entry *core_entry;
 		struct rt_sigframe *rt_sigframe;
 		unsigned long new_sp, fsgs_base;
+		int my_pid = sys_gettid();
+
+		if (my_pid != args->pid) {
+			write_num_n(__LINE__);
+			write_num_n(my_pid);
+			write_num_n(args->pid);
+			goto core_restore_end;
+		}
 
 		core_entry = (struct core_entry *)&args->mem_zone.heap;
 
@@ -413,10 +421,24 @@ self_len_end:
 		if (args->nr_threads) {
 			struct thread_restore_args *thread_args = args->thread_args;
 			long clone_flags = CLONE_VM | CLONE_FILES | CLONE_SIGHAND	|
-					   CLONE_THREAD | CLONE_SYSVSEM			|
-					   CLONE_CHILD_USEPID;
+					   CLONE_THREAD | CLONE_SYSVSEM;
+			long last_pid_len;
 			long parent_tid;
-			int i;
+			int i, fd;
+
+			fd = sys_open(args->ns_last_pid_path, O_RDWR, LAST_PID_PERM);
+			if (fd < 0) {
+				write_num_n(__LINE__);
+				write_num_n(fd);
+				goto core_restore_end;
+			}
+
+			ret = sys_flock(fd, LOCK_EX);
+			if (ret) {
+				write_num_n(__LINE__);
+				write_num_n(ret);
+				goto core_restore_end;
+			}
 
 			for (i = 0; i < args->nr_threads; i++) {
 
@@ -430,6 +452,9 @@ self_len_end:
 					RESTORE_ALIGN_STACK((long)thread_args[i].mem_zone.stack,
 							    sizeof(thread_args[i].mem_zone.stack));
 
+				last_pid_len = vprint_num(args->last_pid_buf, thread_args[i].pid - 1);
+				sys_write(fd, args->last_pid_buf, last_pid_len);
+
 				/*
 				 * To achieve functionality like libc's clone()
 				 * we need a pure assembly here, because clone()'ed
@@ -476,6 +501,13 @@ self_len_end:
 						"g"(&thread_args[i])
 					: "rax", "rdi", "rsi", "rdx", "r10", "memory");
 			}
+
+			ret = sys_flock(fd, LOCK_UN);
+			if (ret) {
+				write_num_n(__LINE__);
+				write_num_n(ret);
+				goto core_restore_end;
+			}
 		}
 
 		/*