uringmachine 0.1

data/vendor/liburing/test/rename.c

@@ -0,0 +1,132 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Description: run various rename tests
+ *
+ */
+#include <errno.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <string.h>
+#include <fcntl.h>
+#include <sys/stat.h>
+
+#include "liburing.h"
+
+static int test_rename(struct io_uring *ring, const char *old, const char *new)
+{
+	struct io_uring_cqe *cqe;
+	struct io_uring_sqe *sqe;
+	int ret;
+
+	sqe = io_uring_get_sqe(ring);
+	if (!sqe) {
+		fprintf(stderr, "get sqe failed\n");
+		goto err;
+	}
+
+	memset(sqe, 0, sizeof(*sqe));
+
+	io_uring_prep_rename(sqe, old, new);
+	
+	ret = io_uring_submit(ring);
+	if (ret <= 0) {
+		fprintf(stderr, "sqe submit failed: %d\n", ret);
+		goto err;
+	}
+
+	ret = io_uring_wait_cqe(ring, &cqe);
+	if (ret < 0) {
+		fprintf(stderr, "wait completion %d\n", ret);
+		goto err;
+	}
+	ret = cqe->res;
+	io_uring_cqe_seen(ring, cqe);
+	return ret;
+err:
+	return 1;
+}
+
+static int stat_file(const char *buf)
+{
+	struct stat sb;
+
+	if (!stat(buf, &sb))
+		return 0;
+
+	return errno;
+}
+
+int main(int argc, char *argv[])
+{
+	struct io_uring ring;
+	char src[32] = "./XXXXXX";
+	char dst[32] = "./XXXXXX";
+	int ret;
+
+	if (argc > 1)
+		return 0;
+
+	ret = io_uring_queue_init(1, &ring, 0);
+	if (ret) {
+		fprintf(stderr, "ring setup failed: %d\n", ret);
+		return 1;
+	}
+
+	ret = mkstemp(src);
+	if (ret < 0) {
+		perror("mkstemp");
+		return 1;
+	}
+	close(ret);
+
+	ret = mkstemp(dst);
+	if (ret < 0) {
+		perror("mkstemp");
+		return 1;
+	}
+	close(ret);
+
+	if (stat_file(src) != 0) {
+		perror("stat");
+		return 1;
+	}
+	if (stat_file(dst) != 0) {
+		perror("stat");
+		return 1;
+	}
+
+	ret = test_rename(&ring, src, dst);
+	if (ret < 0) {
+		if (ret == -EBADF || ret == -EINVAL) {
+			fprintf(stdout, "Rename not supported, skipping\n");
+			goto out;
+		}
+		fprintf(stderr, "rename: %s\n", strerror(-ret));
+		goto err;
+	} else if (ret)
+		goto err;
+
+	if (stat_file(src) != ENOENT) {
+		fprintf(stderr, "stat got %s\n", strerror(ret));
+		return 1;
+	}
+
+	if (stat_file(dst) != 0) {
+		perror("stat");
+		return 1;
+	}
+
+	ret = test_rename(&ring, "/x/y/1/2", "/2/1/y/x");
+	if (ret != -ENOENT) {
+		fprintf(stderr, "test_rename invalid failed: %d\n", ret);
+		return ret;
+	}
+out:
+	unlink(dst);
+	return 0;
+err:
+	unlink(src);
+	unlink(dst);
+	return 1;
+}

data/vendor/liburing/test/ring-leak.c

@@ -0,0 +1,283 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Based on description from Al Viro - this demonstrates a leak of the
+ * io_uring instance, by sending the io_uring fd over a UNIX socket.
+ *
+ * See:
+ *
+ * https://lore.kernel.org/linux-block/20190129192702.3605-1-axboe@kernel.dk/T/#m6c87fc64e4d063786af6ec6fadce3ac1e95d3184
+ *
+ */
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <signal.h>
+#include <inttypes.h>
+#include <sys/types.h>
+#include <sys/syscall.h>
+#include <sys/socket.h>
+#include <sys/wait.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <string.h>
+#include <linux/fs.h>
+
+#include "liburing.h"
+#include "helpers.h"
+#include "../src/syscall.h"
+
+static int __io_uring_register_files(int ring_fd, int fd1, int fd2)
+{
+	__s32 fds[2] = { fd1, fd2 };
+
+	return __sys_io_uring_register(ring_fd, IORING_REGISTER_FILES, fds, 2);
+}
+
+static int get_ring_fd(void)
+{
+	struct io_uring_params p;
+	int fd;
+
+	memset(&p, 0, sizeof(p));
+
+	fd = __sys_io_uring_setup(2, &p);
+	if (fd < 0) {
+		perror("io_uring_setup");
+		return -1;
+	}
+
+	return fd;
+}
+
+static int send_fd(int socket, int fd)
+{
+	char buf[CMSG_SPACE(sizeof(fd))];
+	struct cmsghdr *cmsg;
+	struct msghdr msg;
+
+	memset(buf, 0, sizeof(buf));
+	memset(&msg, 0, sizeof(msg));
+
+	msg.msg_control = buf;
+	msg.msg_controllen = sizeof(buf);
+
+	cmsg = CMSG_FIRSTHDR(&msg);
+	cmsg->cmsg_level = SOL_SOCKET;
+	cmsg->cmsg_type = SCM_RIGHTS;
+	cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
+
+	memmove(CMSG_DATA(cmsg), &fd, sizeof(fd));
+
+	msg.msg_controllen = CMSG_SPACE(sizeof(fd));
+
+	if (sendmsg(socket, &msg, 0) < 0) {
+		if (errno == EINVAL)
+			return T_EXIT_SKIP;
+		perror("sendmsg");
+		return T_EXIT_FAIL;
+	}
+
+	return T_EXIT_PASS;
+}
+
+static int test_iowq_request_cancel(void)
+{
+	char buffer[128];
+	struct io_uring ring;
+	struct io_uring_sqe *sqe;
+	int ret, fds[2];
+
+	ret = io_uring_queue_init(8, &ring, 0);
+	if (ret < 0) {
+		fprintf(stderr, "failed to init io_uring: %s\n", strerror(-ret));
+		return ret;
+	}
+	if (pipe(fds)) {
+		perror("pipe");
+		return -1;
+	}
+	ret = io_uring_register_files(&ring, fds, 2);
+	if (ret) {
+		fprintf(stderr, "file_register: %d\n", ret);
+		return ret;
+	}
+	close(fds[1]);
+
+	sqe = io_uring_get_sqe(&ring);
+	if (!sqe) {
+		fprintf(stderr, "%s: failed to get sqe\n", __FUNCTION__);
+		return 1;
+	}
+	/* potentially sitting in internal polling */
+	io_uring_prep_read(sqe, 0, buffer, 10, 0);
+	sqe->flags |= IOSQE_FIXED_FILE;
+
+	sqe = io_uring_get_sqe(&ring);
+	if (!sqe) {
+		fprintf(stderr, "%s: failed to get sqe\n", __FUNCTION__);
+		return 1;
+	}
+	/* staying in io-wq */
+	io_uring_prep_read(sqe, 0, buffer, 10, 0);
+	sqe->flags |= IOSQE_FIXED_FILE | IOSQE_ASYNC;
+
+	ret = io_uring_submit(&ring);
+	if (ret != 2) {
+		fprintf(stderr, "%s: got %d, wanted 1\n", __FUNCTION__, ret);
+		return 1;
+	}
+
+	/* should unregister files and close the write fd */
+	io_uring_queue_exit(&ring);
+
+	/*
+	 * We're trying to wait for the ring to "really" exit, that will be
+	 * done async. For that rely on the registered write end to be closed
+	 * after ring quiesce, so failing read from the other pipe end.
+	 */
+	ret = read(fds[0], buffer, 10);
+	if (ret < 0)
+		perror("read");
+	close(fds[0]);
+	return 0;
+}
+
+static void trigger_unix_gc(void)
+{
+	int fd;
+
+	fd = socket(AF_UNIX, SOCK_DGRAM, 0);
+	if (fd < 0)
+		perror("socket dgram");
+	else
+		close(fd);
+}
+
+static int test_scm_cycles(bool update)
+{
+	char buffer[128];
+	struct io_uring ring;
+	int i, ret;
+	int sp[2], fds[2], reg_fds[4];
+
+	if (socketpair(AF_UNIX, SOCK_DGRAM, 0, sp) != 0) {
+		perror("Failed to create Unix-domain socket pair\n");
+		return 1;
+	}
+	ret = io_uring_queue_init(8, &ring, 0);
+	if (ret < 0) {
+		fprintf(stderr, "failed to init io_uring: %s\n", strerror(-ret));
+		return ret;
+	}
+	if (pipe(fds)) {
+		perror("pipe");
+		return -1;
+	}
+	ret = send_fd(sp[0], ring.ring_fd);
+	if (ret != T_EXIT_PASS)
+		return ret;
+
+	/* register an empty set for updates */
+	if (update) {
+		for (i = 0; i < 4; i++)
+			reg_fds[i] = -1;
+		ret = io_uring_register_files(&ring, reg_fds, 4);
+		if (ret) {
+			fprintf(stderr, "file_register: %d\n", ret);
+			return ret;
+		}
+	}
+
+	reg_fds[0] = fds[0];
+	reg_fds[1] = fds[1];
+	reg_fds[2] = sp[0];
+	reg_fds[3] = sp[1];
+	if (update) {
+		ret = io_uring_register_files_update(&ring, 0, reg_fds, 4);
+		if (ret != 4) {
+			fprintf(stderr, "file_register: %d\n", ret);
+			return ret;
+		}
+	} else {
+		ret = io_uring_register_files(&ring, reg_fds, 4);
+		if (ret) {
+			fprintf(stderr, "file_register: %d\n", ret);
+			return ret;
+		}
+	}
+
+	close(fds[1]);
+	close(sp[0]);
+	close(sp[1]);
+
+	/* should unregister files and close the write fd */
+	io_uring_queue_exit(&ring);
+
+	trigger_unix_gc();
+
+	/*
+	 * We're trying to wait for the ring to "really" exit, that will be
+	 * done async. For that rely on the registered write end to be closed
+	 * after ring quiesce, so failing read from the other pipe end.
+	 */
+	ret = read(fds[0], buffer, 10);
+	if (ret < 0)
+		perror("read");
+	close(fds[0]);
+	return 0;
+}
+
+int main(int argc, char *argv[])
+{
+	int sp[2], pid, ring_fd, ret;
+	int i;
+
+	if (argc > 1)
+		return 0;
+
+	ret = test_iowq_request_cancel();
+	if (ret) {
+		fprintf(stderr, "test_iowq_request_cancel() failed\n");
+		return 1;
+	}
+
+	for (i = 0; i < 2; i++) {
+		bool update = !!(i & 1);
+
+		ret = test_scm_cycles(update);
+		if (ret == T_EXIT_SKIP)
+			return T_EXIT_SKIP;
+		if (ret) {
+			fprintf(stderr, "test_scm_cycles() failed %i\n",
+				update);
+			return 1;
+		}
+	}
+
+	if (socketpair(AF_UNIX, SOCK_DGRAM, 0, sp) != 0) {
+		perror("Failed to create Unix-domain socket pair\n");
+		return 1;
+	}
+
+	ring_fd = get_ring_fd();
+	if (ring_fd < 0)
+		return 1;
+
+	ret = __io_uring_register_files(ring_fd, sp[0], sp[1]);
+	if (ret < 0) {
+		perror("register files");
+		return 1;
+	}
+
+	pid = fork();
+	if (pid) {
+		ret = send_fd(sp[0], ring_fd);
+		if (ret != T_EXIT_PASS)
+			return ret;
+	}
+
+	close(ring_fd);
+	close(sp[0]);
+	close(sp[1]);
+	return 0;
+}

data/vendor/liburing/test/ring-leak2.c

@@ -0,0 +1,249 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Description: Test two ring deadlock. A buggy kernel will end up
+ * 		having io_wq_* workers pending, as the circular reference
+ * 		will prevent full exit.
+ *
+ * Based on a test case from Josef <josef.grieb@gmail.com>
+ *
+ */
+#include <errno.h>
+#include <fcntl.h>
+#include <netinet/in.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <strings.h>
+#include <poll.h>
+#include <sys/socket.h>
+#include <unistd.h>
+#include <sys/eventfd.h>
+#include <pthread.h>
+
+#include "liburing.h"
+#include "../src/syscall.h"
+
+enum {
+	ACCEPT,
+	READ,
+	WRITE,
+	POLLING_IN,
+	POLLING_RDHUP,
+	CLOSE,
+	EVENTFD_READ,
+};
+
+typedef struct conn_info {
+	__u32 fd;
+	__u16 type;
+	__u16 bid;
+} conn_info;
+
+static char read_eventfd_buffer[8];
+
+static pthread_mutex_t lock;
+static struct io_uring *client_ring;
+
+static int client_eventfd = -1;
+
+static int setup_io_uring(struct io_uring *ring)
+{
+	struct io_uring_params p = { };
+	int ret;
+
+	ret = io_uring_queue_init_params(8, ring, &p);
+	if (ret) {
+		fprintf(stderr, "Unable to setup io_uring: %s\n",
+			strerror(-ret));
+		return 1;
+	}
+	return 0;
+}
+
+static void add_socket_eventfd_read(struct io_uring *ring, int fd)
+{
+	struct io_uring_sqe *sqe;
+	conn_info conn_i = {
+		.fd = fd,
+		.type = EVENTFD_READ,
+	};
+
+	sqe = io_uring_get_sqe(ring);
+	io_uring_prep_read(sqe, fd, &read_eventfd_buffer, 8, 0);
+	io_uring_sqe_set_flags(sqe, IOSQE_ASYNC);
+
+	memcpy(&sqe->user_data, &conn_i, sizeof(conn_i));
+}
+
+static void add_socket_pollin(struct io_uring *ring, int fd)
+{
+	struct io_uring_sqe *sqe;
+	conn_info conn_i = {
+		.fd = fd,
+		.type = POLLING_IN,
+	};
+
+	sqe = io_uring_get_sqe(ring);
+	io_uring_prep_poll_add(sqe, fd, POLL_IN);
+
+	memcpy(&sqe->user_data, &conn_i, sizeof(conn_i));
+}
+
+static void *server_thread(void *arg)
+{
+	struct sockaddr_in serv_addr;
+	int port = 0;
+	int sock_listen_fd, evfd;
+	const int val = 1;
+	struct io_uring ring;
+       
+	sock_listen_fd = socket(AF_INET, SOCK_STREAM | SOCK_NONBLOCK, 0);
+	setsockopt(sock_listen_fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
+
+	memset(&serv_addr, 0, sizeof(serv_addr));
+	serv_addr.sin_family = AF_INET;
+	serv_addr.sin_port = htons(port);
+	serv_addr.sin_addr.s_addr = INADDR_ANY;
+
+	evfd = eventfd(0, EFD_CLOEXEC);
+
+	// bind and listen
+	if (bind(sock_listen_fd, (struct sockaddr *)&serv_addr, sizeof(serv_addr)) < 0) {
+		perror("Error binding socket...\n");
+		exit(1);
+	}
+	if (listen(sock_listen_fd, 1) < 0) {
+		perror("Error listening on socket...\n");
+		exit(1);
+	}
+
+	setup_io_uring(&ring);
+	add_socket_eventfd_read(&ring, evfd);
+	add_socket_pollin(&ring, sock_listen_fd);
+
+	while (1) {
+		struct io_uring_cqe *cqe;
+		unsigned head;
+		unsigned count = 0;
+
+		io_uring_submit_and_wait(&ring, 1);
+
+		io_uring_for_each_cqe(&ring, head, cqe) {
+			struct conn_info conn_i;
+
+			count++;
+			memcpy(&conn_i, &cqe->user_data, sizeof(conn_i));
+
+			if (conn_i.type == ACCEPT) {
+				int sock_conn_fd = cqe->res;
+				// only read when there is no error, >= 0
+				if (sock_conn_fd > 0) {
+					add_socket_pollin(&ring, sock_listen_fd);
+
+					pthread_mutex_lock(&lock);
+					io_uring_submit(client_ring);
+					pthread_mutex_unlock(&lock);
+
+				}
+			} else if (conn_i.type == POLLING_IN) {
+				break;
+			}
+		}
+		io_uring_cq_advance(&ring, count);
+	}
+}
+
+static void *client_thread(void *arg)
+{
+	struct io_uring ring;
+	int ret;
+
+	setup_io_uring(&ring);
+	client_ring = &ring;
+
+	client_eventfd = eventfd(0, EFD_CLOEXEC);
+	pthread_mutex_lock(&lock);
+	add_socket_eventfd_read(&ring, client_eventfd);
+	pthread_mutex_unlock(&lock);
+
+	while (1) {
+		struct io_uring_cqe *cqe;
+		unsigned head;
+		unsigned count = 0;
+
+		pthread_mutex_lock(&lock);
+		io_uring_submit(&ring);
+		pthread_mutex_unlock(&lock);
+
+		ret = __sys_io_uring_enter(ring.ring_fd, 0, 1, IORING_ENTER_GETEVENTS, NULL);
+		if (ret < 0) {
+			perror("Error io_uring_enter...\n");
+			exit(1);
+		}
+
+		// go through all CQEs
+		io_uring_for_each_cqe(&ring, head, cqe) {
+			struct conn_info conn_i;
+			int type;
+
+			count++;
+			memcpy(&conn_i, &cqe->user_data, sizeof(conn_i));
+
+			type = conn_i.type;
+			if (type == READ) {
+				pthread_mutex_lock(&lock);
+
+				if (cqe->res <= 0) {
+					// connection closed or error
+					shutdown(conn_i.fd, SHUT_RDWR);
+				} else {
+					pthread_mutex_unlock(&lock);
+					break;
+				}
+				add_socket_pollin(&ring, conn_i.fd);
+				pthread_mutex_unlock(&lock);
+			} else if (type == WRITE) {
+			} else if (type == POLLING_IN) {
+				break;
+			} else if (type == POLLING_RDHUP) {
+				break;
+			} else if (type == CLOSE) {
+			} else if (type == EVENTFD_READ) {
+				add_socket_eventfd_read(&ring, client_eventfd);
+			}
+		}
+
+		io_uring_cq_advance(&ring, count);
+	}
+}
+
+static void sig_alrm(int sig)
+{
+	exit(0);
+}
+
+int main(int argc, char *argv[])
+{
+	pthread_t server_thread_t, client_thread_t;
+	struct sigaction act;
+
+	if (argc > 1)
+		return 0;
+
+	if (pthread_mutex_init(&lock, NULL) != 0) {
+		printf("\n mutex init failed\n");
+		return 1;
+	}
+
+	pthread_create(&server_thread_t, NULL, &server_thread, NULL);
+	pthread_create(&client_thread_t, NULL, &client_thread, NULL);
+
+	memset(&act, 0, sizeof(act));
+	act.sa_handler = sig_alrm;
+	act.sa_flags = SA_RESTART;
+	sigaction(SIGALRM, &act, NULL);
+	alarm(1);
+
+	pthread_join(server_thread_t, NULL);
+	return 0;
+}