On Mon, May 05, 2025 at 11:48:43AM -0700, Kuniyuki Iwashima wrote:

From: Christian Brauner <brauner@kernel.org>
Date: Mon, 05 May 2025 13:13:42 +0200

quoted

@@ -801,6 +846,40 @@ void do_coredump(const kernel_siginfo_t *siginfo)
 		}
 		break;
 	}
+	case COREDUMP_SOCK: {
+		struct file *file __free(fput) = NULL;
+#ifdef CONFIG_UNIX
+		struct socket *socket;
+
+		/*
+		 * It is possible that the userspace process which is
+		 * supposed to handle the coredump and is listening on
+		 * the AF_UNIX socket coredumps. Userspace should just
+		 * mark itself non dumpable.
+		 */
+
+		retval = sock_create_kern(&init_net, AF_UNIX, SOCK_STREAM, 0, &socket);
+		if (retval < 0)
+			goto close_fail;
+
+		file = sock_alloc_file(socket, 0, NULL);
+		if (IS_ERR(file)) {
+			sock_release(socket);
+			retval = PTR_ERR(file);
+			goto close_fail;
+		}
+
+		retval = kernel_connect(socket,
+					(struct sockaddr *)(&coredump_unix_socket),
+					COREDUMP_UNIX_SOCKET_ADDR_SIZE, 0);

This blocks forever if the listener's accept() queue is full.

I think we don't want that and should pass O_NONBLOCK.

To keep the queue clean is userspace responsibility, and we don't
need to care about a weird user.

That seems fine to me. I've changed that.