test/ring-leak.c

/* 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 "../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 void 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)
		perror("sendmsg");
}

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 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;
	}
	send_fd(sp[0], ring.ring_fd);

	/* 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);

	/*
	 * 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) {
			fprintf(stderr, "test_scm_cycles() failed %i\n",
				update);
			return 1;
		}
		break;
	}

	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)
		send_fd(sp[0], ring_fd);

	close(ring_fd);
	close(sp[0]);
	close(sp[1]);
	return 0;
}