opnsense-src/tests/sys/netinet/socket_afinet.c

599 lines
16 KiB
C
Raw Permalink Normal View History

/*-
* SPDX-License-Identifier: BSD-2-Clause
*
* Copyright (c) 2019 Bjoern A. Zeeb
inpcb: Remove bogus SO_REUSEPORT(_LB) checks in in_pcbbind() This check for SO_REUSEPORT was added way back in commit 52b65dbe85faf. Per the commit log, this commit restricted this port-stealing check to unicast addresses, and then only if the existing socket does not have SO_REUSEPORT set. In other words, if there exists a socket bound to INADDR_ANY, and we bind a socket to INADDR_ANY with the same port, then the two sockets need not be owned by the same user if the existing socket has SO_REUSEPORT set. This is a surprising semantic; bugzilla PR 7713 gives some additional context. That PR makes a case for the behaviour described above when binding to a multicast address. But, the SO_REUSEPORT check is only applied when binding to a non-multicast address, so it doesn't really make sense. In the PR the committer notes that "unicast applications don't set SO_REUSEPORT", which makes some sense, but also refers to "multicast applications that bind to INADDR_ANY", which sounds a bit suspicious. OpenBSD performs the multicast check, but not the SO_REUSEPORT check. DragonflyBSD removed the SO_REUSEPORT (and INADDR_ANY) checks back in 2014 (commit 0323d5fde12a4). NetBSD explicitly copied our logic and still has it. The plot thickens: 20 years later, SO_REUSEPORT_LB was ported from DragonflyBSD: this option provides similar semantics to SO_REUSEPORT, but for unicast addresses it causes incoming connections/datagrams to be distributed among all sockets in the group. This commit (1a43cff92a20d) inverted the check for SO_REUSEPORT while adding one for SO_REUSEPORT_LB; this appears to have been inadvertent. However: - apparently no one has noticed that the semantics were changed; - sockets belonging to different users can now be bound to the same port so long as they belong to a single lbgroup bound to INADDR_ANY, which is not correct. Simply remove the SO_REUSEPORT(_LB) checks, as their original justification was dubious and their current implementation is wrong; add some tests. Reviewed by: glebius MFC after: 1 month Sponsored by: Klara, Inc. Sponsored by: Stormshield Differential Revision: https://reviews.freebsd.org/D47832 (cherry picked from commit 4f02a7d739b354eef38e19b25866f64842d69414)
2024-12-12 09:06:06 -05:00
* Copyright (c) 2024 Stormshield
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
inpcb: Remove bogus SO_REUSEPORT(_LB) checks in in_pcbbind() This check for SO_REUSEPORT was added way back in commit 52b65dbe85faf. Per the commit log, this commit restricted this port-stealing check to unicast addresses, and then only if the existing socket does not have SO_REUSEPORT set. In other words, if there exists a socket bound to INADDR_ANY, and we bind a socket to INADDR_ANY with the same port, then the two sockets need not be owned by the same user if the existing socket has SO_REUSEPORT set. This is a surprising semantic; bugzilla PR 7713 gives some additional context. That PR makes a case for the behaviour described above when binding to a multicast address. But, the SO_REUSEPORT check is only applied when binding to a non-multicast address, so it doesn't really make sense. In the PR the committer notes that "unicast applications don't set SO_REUSEPORT", which makes some sense, but also refers to "multicast applications that bind to INADDR_ANY", which sounds a bit suspicious. OpenBSD performs the multicast check, but not the SO_REUSEPORT check. DragonflyBSD removed the SO_REUSEPORT (and INADDR_ANY) checks back in 2014 (commit 0323d5fde12a4). NetBSD explicitly copied our logic and still has it. The plot thickens: 20 years later, SO_REUSEPORT_LB was ported from DragonflyBSD: this option provides similar semantics to SO_REUSEPORT, but for unicast addresses it causes incoming connections/datagrams to be distributed among all sockets in the group. This commit (1a43cff92a20d) inverted the check for SO_REUSEPORT while adding one for SO_REUSEPORT_LB; this appears to have been inadvertent. However: - apparently no one has noticed that the semantics were changed; - sockets belonging to different users can now be bound to the same port so long as they belong to a single lbgroup bound to INADDR_ANY, which is not correct. Simply remove the SO_REUSEPORT(_LB) checks, as their original justification was dubious and their current implementation is wrong; add some tests. Reviewed by: glebius MFC after: 1 month Sponsored by: Klara, Inc. Sponsored by: Stormshield Differential Revision: https://reviews.freebsd.org/D47832 (cherry picked from commit 4f02a7d739b354eef38e19b25866f64842d69414)
2024-12-12 09:06:06 -05:00
#include <sys/param.h>
#include <sys/socket.h>
inpcb: Remove bogus SO_REUSEPORT(_LB) checks in in_pcbbind() This check for SO_REUSEPORT was added way back in commit 52b65dbe85faf. Per the commit log, this commit restricted this port-stealing check to unicast addresses, and then only if the existing socket does not have SO_REUSEPORT set. In other words, if there exists a socket bound to INADDR_ANY, and we bind a socket to INADDR_ANY with the same port, then the two sockets need not be owned by the same user if the existing socket has SO_REUSEPORT set. This is a surprising semantic; bugzilla PR 7713 gives some additional context. That PR makes a case for the behaviour described above when binding to a multicast address. But, the SO_REUSEPORT check is only applied when binding to a non-multicast address, so it doesn't really make sense. In the PR the committer notes that "unicast applications don't set SO_REUSEPORT", which makes some sense, but also refers to "multicast applications that bind to INADDR_ANY", which sounds a bit suspicious. OpenBSD performs the multicast check, but not the SO_REUSEPORT check. DragonflyBSD removed the SO_REUSEPORT (and INADDR_ANY) checks back in 2014 (commit 0323d5fde12a4). NetBSD explicitly copied our logic and still has it. The plot thickens: 20 years later, SO_REUSEPORT_LB was ported from DragonflyBSD: this option provides similar semantics to SO_REUSEPORT, but for unicast addresses it causes incoming connections/datagrams to be distributed among all sockets in the group. This commit (1a43cff92a20d) inverted the check for SO_REUSEPORT while adding one for SO_REUSEPORT_LB; this appears to have been inadvertent. However: - apparently no one has noticed that the semantics were changed; - sockets belonging to different users can now be bound to the same port so long as they belong to a single lbgroup bound to INADDR_ANY, which is not correct. Simply remove the SO_REUSEPORT(_LB) checks, as their original justification was dubious and their current implementation is wrong; add some tests. Reviewed by: glebius MFC after: 1 month Sponsored by: Klara, Inc. Sponsored by: Stormshield Differential Revision: https://reviews.freebsd.org/D47832 (cherry picked from commit 4f02a7d739b354eef38e19b25866f64842d69414)
2024-12-12 09:06:06 -05:00
#include <sys/wait.h>
#include <netinet/in.h>
inpcb: Remove bogus SO_REUSEPORT(_LB) checks in in_pcbbind() This check for SO_REUSEPORT was added way back in commit 52b65dbe85faf. Per the commit log, this commit restricted this port-stealing check to unicast addresses, and then only if the existing socket does not have SO_REUSEPORT set. In other words, if there exists a socket bound to INADDR_ANY, and we bind a socket to INADDR_ANY with the same port, then the two sockets need not be owned by the same user if the existing socket has SO_REUSEPORT set. This is a surprising semantic; bugzilla PR 7713 gives some additional context. That PR makes a case for the behaviour described above when binding to a multicast address. But, the SO_REUSEPORT check is only applied when binding to a non-multicast address, so it doesn't really make sense. In the PR the committer notes that "unicast applications don't set SO_REUSEPORT", which makes some sense, but also refers to "multicast applications that bind to INADDR_ANY", which sounds a bit suspicious. OpenBSD performs the multicast check, but not the SO_REUSEPORT check. DragonflyBSD removed the SO_REUSEPORT (and INADDR_ANY) checks back in 2014 (commit 0323d5fde12a4). NetBSD explicitly copied our logic and still has it. The plot thickens: 20 years later, SO_REUSEPORT_LB was ported from DragonflyBSD: this option provides similar semantics to SO_REUSEPORT, but for unicast addresses it causes incoming connections/datagrams to be distributed among all sockets in the group. This commit (1a43cff92a20d) inverted the check for SO_REUSEPORT while adding one for SO_REUSEPORT_LB; this appears to have been inadvertent. However: - apparently no one has noticed that the semantics were changed; - sockets belonging to different users can now be bound to the same port so long as they belong to a single lbgroup bound to INADDR_ANY, which is not correct. Simply remove the SO_REUSEPORT(_LB) checks, as their original justification was dubious and their current implementation is wrong; add some tests. Reviewed by: glebius MFC after: 1 month Sponsored by: Klara, Inc. Sponsored by: Stormshield Differential Revision: https://reviews.freebsd.org/D47832 (cherry picked from commit 4f02a7d739b354eef38e19b25866f64842d69414)
2024-12-12 09:06:06 -05:00
#include <errno.h>
#include <poll.h>
inpcb: Remove bogus SO_REUSEPORT(_LB) checks in in_pcbbind() This check for SO_REUSEPORT was added way back in commit 52b65dbe85faf. Per the commit log, this commit restricted this port-stealing check to unicast addresses, and then only if the existing socket does not have SO_REUSEPORT set. In other words, if there exists a socket bound to INADDR_ANY, and we bind a socket to INADDR_ANY with the same port, then the two sockets need not be owned by the same user if the existing socket has SO_REUSEPORT set. This is a surprising semantic; bugzilla PR 7713 gives some additional context. That PR makes a case for the behaviour described above when binding to a multicast address. But, the SO_REUSEPORT check is only applied when binding to a non-multicast address, so it doesn't really make sense. In the PR the committer notes that "unicast applications don't set SO_REUSEPORT", which makes some sense, but also refers to "multicast applications that bind to INADDR_ANY", which sounds a bit suspicious. OpenBSD performs the multicast check, but not the SO_REUSEPORT check. DragonflyBSD removed the SO_REUSEPORT (and INADDR_ANY) checks back in 2014 (commit 0323d5fde12a4). NetBSD explicitly copied our logic and still has it. The plot thickens: 20 years later, SO_REUSEPORT_LB was ported from DragonflyBSD: this option provides similar semantics to SO_REUSEPORT, but for unicast addresses it causes incoming connections/datagrams to be distributed among all sockets in the group. This commit (1a43cff92a20d) inverted the check for SO_REUSEPORT while adding one for SO_REUSEPORT_LB; this appears to have been inadvertent. However: - apparently no one has noticed that the semantics were changed; - sockets belonging to different users can now be bound to the same port so long as they belong to a single lbgroup bound to INADDR_ANY, which is not correct. Simply remove the SO_REUSEPORT(_LB) checks, as their original justification was dubious and their current implementation is wrong; add some tests. Reviewed by: glebius MFC after: 1 month Sponsored by: Klara, Inc. Sponsored by: Stormshield Differential Revision: https://reviews.freebsd.org/D47832 (cherry picked from commit 4f02a7d739b354eef38e19b25866f64842d69414)
2024-12-12 09:06:06 -05:00
#include <pwd.h>
#include <stdio.h>
#include <unistd.h>
#include <atf-c.h>
ATF_TC_WITHOUT_HEAD(socket_afinet);
ATF_TC_BODY(socket_afinet, tc)
{
int sd;
sd = socket(PF_INET, SOCK_DGRAM, 0);
ATF_CHECK(sd >= 0);
close(sd);
}
ATF_TC_WITHOUT_HEAD(socket_afinet_bind_zero);
ATF_TC_BODY(socket_afinet_bind_zero, tc)
{
int sd, rc;
struct sockaddr_in sin;
if (atf_tc_get_config_var_as_bool_wd(tc, "ci", false))
atf_tc_skip("doesn't work when mac_portacl(4) loaded (https://bugs.freebsd.org/238781)");
sd = socket(PF_INET, SOCK_DGRAM, 0);
ATF_CHECK(sd >= 0);
bzero(&sin, sizeof(sin));
/*
* For AF_INET we do not check the family in in_pcbbind_setup(9),
* sa_len gets set from the syscall argument in getsockaddr(9),
* so we bind to 0:0.
*/
rc = bind(sd, (struct sockaddr *)&sin, sizeof(sin));
ATF_CHECK_EQ(0, rc);
close(sd);
}
ATF_TC_WITHOUT_HEAD(socket_afinet_bind_ok);
ATF_TC_BODY(socket_afinet_bind_ok, tc)
{
int sd, rc;
struct sockaddr_in sin;
sd = socket(PF_INET, SOCK_DGRAM, 0);
ATF_CHECK(sd >= 0);
bzero(&sin, sizeof(sin));
sin.sin_family = AF_INET;
sin.sin_len = sizeof(sin);
sin.sin_port = htons(0);
sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
rc = bind(sd, (struct sockaddr *)&sin, sizeof(sin));
ATF_CHECK_EQ(0, rc);
close(sd);
}
ATF_TC_WITHOUT_HEAD(socket_afinet_poll_no_rdhup);
ATF_TC_BODY(socket_afinet_poll_no_rdhup, tc)
{
int ss, ss2, cs, rc;
struct sockaddr_in sin;
socklen_t slen;
struct pollfd pfd;
int one = 1;
/* Verify that we don't expose POLLRDHUP if not requested. */
/* Server setup. */
ss = socket(PF_INET, SOCK_STREAM, 0);
ATF_CHECK(ss >= 0);
rc = setsockopt(ss, SOL_SOCKET, SO_REUSEPORT, &one, sizeof(one));
ATF_CHECK_EQ(0, rc);
bzero(&sin, sizeof(sin));
sin.sin_family = AF_INET;
sin.sin_len = sizeof(sin);
sin.sin_port = htons(0);
sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
rc = bind(ss, (struct sockaddr *)&sin, sizeof(sin));
ATF_CHECK_EQ(0, rc);
rc = listen(ss, 1);
ATF_CHECK_EQ(0, rc);
slen = sizeof(sin);
rc = getsockname(ss, (struct sockaddr *)&sin, &slen);
ATF_CHECK_EQ(0, rc);
/* Client connects, server accepts. */
cs = socket(PF_INET, SOCK_STREAM, 0);
ATF_CHECK(cs >= 0);
rc = connect(cs, (struct sockaddr *)&sin, sizeof(sin));
ATF_CHECK_EQ(0, rc);
ss2 = accept(ss, NULL, NULL);
ATF_CHECK(ss2 >= 0);
/* Server can write, sees only POLLOUT. */
pfd.fd = ss2;
pfd.events = POLLIN | POLLOUT;
rc = poll(&pfd, 1, 0);
ATF_CHECK_EQ(1, rc);
ATF_CHECK_EQ(POLLOUT, pfd.revents);
/* Client closes socket! */
rc = close(cs);
ATF_CHECK_EQ(0, rc);
/*
* Server now sees POLLIN, but not POLLRDHUP because we didn't ask.
* Need non-zero timeout to wait for the FIN to arrive and trigger the
* socket to become readable.
*/
pfd.fd = ss2;
pfd.events = POLLIN;
rc = poll(&pfd, 1, 60000);
ATF_CHECK_EQ(1, rc);
ATF_CHECK_EQ(POLLIN, pfd.revents);
close(ss2);
close(ss);
}
ATF_TC_WITHOUT_HEAD(socket_afinet_poll_rdhup);
ATF_TC_BODY(socket_afinet_poll_rdhup, tc)
{
int ss, ss2, cs, rc;
struct sockaddr_in sin;
socklen_t slen;
struct pollfd pfd;
char buffer;
int one = 1;
/* Verify that server sees POLLRDHUP if it asks for it. */
/* Server setup. */
ss = socket(PF_INET, SOCK_STREAM, 0);
ATF_CHECK(ss >= 0);
rc = setsockopt(ss, SOL_SOCKET, SO_REUSEPORT, &one, sizeof(one));
ATF_CHECK_EQ(0, rc);
bzero(&sin, sizeof(sin));
sin.sin_family = AF_INET;
sin.sin_len = sizeof(sin);
sin.sin_port = htons(0);
sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
rc = bind(ss, (struct sockaddr *)&sin, sizeof(sin));
ATF_CHECK_EQ(0, rc);
rc = listen(ss, 1);
ATF_CHECK_EQ(0, rc);
slen = sizeof(sin);
rc = getsockname(ss, (struct sockaddr *)&sin, &slen);
ATF_CHECK_EQ(0, rc);
/* Client connects, server accepts. */
cs = socket(PF_INET, SOCK_STREAM, 0);
ATF_CHECK(cs >= 0);
rc = connect(cs, (struct sockaddr *)&sin, sizeof(sin));
ATF_CHECK_EQ(0, rc);
ss2 = accept(ss, NULL, NULL);
ATF_CHECK(ss2 >= 0);
/* Server can write, so sees POLLOUT. */
pfd.fd = ss2;
pfd.events = POLLIN | POLLOUT | POLLRDHUP;
rc = poll(&pfd, 1, 0);
ATF_CHECK_EQ(1, rc);
ATF_CHECK_EQ(POLLOUT, pfd.revents);
/* Client writes two bytes, server reads only one of them. */
rc = write(cs, "xx", 2);
ATF_CHECK_EQ(2, rc);
rc = read(ss2, &buffer, 1);
ATF_CHECK_EQ(1, rc);
/* Server can read, so sees POLLIN. */
pfd.fd = ss2;
pfd.events = POLLIN | POLLOUT | POLLRDHUP;
rc = poll(&pfd, 1, 0);
ATF_CHECK_EQ(1, rc);
ATF_CHECK_EQ(POLLIN | POLLOUT, pfd.revents);
/* Client closes socket! */
rc = close(cs);
ATF_CHECK_EQ(0, rc);
/*
* Server sees Linux-style POLLRDHUP. Note that this is the case even
* though one byte of data remains unread.
*
* This races against the delivery of FIN caused by the close() above.
* Sometimes (more likely when run under truss or if another system
* call is added in between) it hits the path where sopoll_generic()
* immediately sees SBS_CANTRCVMORE, and sometimes it sleeps with flag
* SB_SEL so that it's woken up almost immediately and runs again,
* which is why we need a non-zero timeout here.
*/
pfd.fd = ss2;
pfd.events = POLLRDHUP;
rc = poll(&pfd, 1, 60000);
ATF_CHECK_EQ(1, rc);
ATF_CHECK_EQ(POLLRDHUP, pfd.revents);
close(ss2);
close(ss);
}
ATF_TC_WITHOUT_HEAD(socket_afinet_stream_reconnect);
ATF_TC_BODY(socket_afinet_stream_reconnect, tc)
{
struct sockaddr_in sin;
socklen_t slen;
int ss, cs, rc;
/*
* Make sure that an attempt to connect(2) a connected or disconnected
* stream socket fails with EISCONN.
*/
/* Server setup. */
ss = socket(PF_INET, SOCK_STREAM, 0);
ATF_CHECK(ss >= 0);
bzero(&sin, sizeof(sin));
sin.sin_family = AF_INET;
sin.sin_len = sizeof(sin);
sin.sin_port = htons(0);
sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
rc = bind(ss, (struct sockaddr *)&sin, sizeof(sin));
ATF_CHECK_EQ(0, rc);
rc = listen(ss, 1);
ATF_CHECK_EQ(0, rc);
slen = sizeof(sin);
rc = getsockname(ss, (struct sockaddr *)&sin, &slen);
ATF_CHECK_EQ(0, rc);
/* Client connects, shuts down. */
cs = socket(PF_INET, SOCK_STREAM, 0);
ATF_CHECK(cs >= 0);
rc = connect(cs, (struct sockaddr *)&sin, sizeof(sin));
ATF_CHECK_EQ(0, rc);
rc = shutdown(cs, SHUT_RDWR);
ATF_CHECK_EQ(0, rc);
/* A subsequent connect(2) fails with EISCONN. */
rc = connect(cs, (struct sockaddr *)&sin, sizeof(sin));
ATF_CHECK_EQ(-1, rc);
ATF_CHECK_EQ(errno, EISCONN);
rc = close(cs);
ATF_CHECK_EQ(0, rc);
rc = close(ss);
ATF_CHECK_EQ(0, rc);
}
inpcb: Remove bogus SO_REUSEPORT(_LB) checks in in_pcbbind() This check for SO_REUSEPORT was added way back in commit 52b65dbe85faf. Per the commit log, this commit restricted this port-stealing check to unicast addresses, and then only if the existing socket does not have SO_REUSEPORT set. In other words, if there exists a socket bound to INADDR_ANY, and we bind a socket to INADDR_ANY with the same port, then the two sockets need not be owned by the same user if the existing socket has SO_REUSEPORT set. This is a surprising semantic; bugzilla PR 7713 gives some additional context. That PR makes a case for the behaviour described above when binding to a multicast address. But, the SO_REUSEPORT check is only applied when binding to a non-multicast address, so it doesn't really make sense. In the PR the committer notes that "unicast applications don't set SO_REUSEPORT", which makes some sense, but also refers to "multicast applications that bind to INADDR_ANY", which sounds a bit suspicious. OpenBSD performs the multicast check, but not the SO_REUSEPORT check. DragonflyBSD removed the SO_REUSEPORT (and INADDR_ANY) checks back in 2014 (commit 0323d5fde12a4). NetBSD explicitly copied our logic and still has it. The plot thickens: 20 years later, SO_REUSEPORT_LB was ported from DragonflyBSD: this option provides similar semantics to SO_REUSEPORT, but for unicast addresses it causes incoming connections/datagrams to be distributed among all sockets in the group. This commit (1a43cff92a20d) inverted the check for SO_REUSEPORT while adding one for SO_REUSEPORT_LB; this appears to have been inadvertent. However: - apparently no one has noticed that the semantics were changed; - sockets belonging to different users can now be bound to the same port so long as they belong to a single lbgroup bound to INADDR_ANY, which is not correct. Simply remove the SO_REUSEPORT(_LB) checks, as their original justification was dubious and their current implementation is wrong; add some tests. Reviewed by: glebius MFC after: 1 month Sponsored by: Klara, Inc. Sponsored by: Stormshield Differential Revision: https://reviews.freebsd.org/D47832 (cherry picked from commit 4f02a7d739b354eef38e19b25866f64842d69414)
2024-12-12 09:06:06 -05:00
/*
* Make sure that unprivileged users can't set the IP_BINDANY or IPV6_BINDANY
* socket options.
*/
ATF_TC(socket_afinet_bindany);
ATF_TC_HEAD(socket_afinet_bindany, tc)
{
atf_tc_set_md_var(tc, "require.user", "unprivileged");
}
ATF_TC_BODY(socket_afinet_bindany, tc)
{
int s;
s = socket(AF_INET, SOCK_STREAM, 0);
ATF_REQUIRE(s >= 0);
ATF_REQUIRE_ERRNO(EPERM,
setsockopt(s, IPPROTO_IP, IP_BINDANY, &(int){1}, sizeof(int)) ==
-1);
ATF_REQUIRE(close(s) == 0);
s = socket(AF_INET, SOCK_DGRAM, 0);
ATF_REQUIRE(s >= 0);
ATF_REQUIRE_ERRNO(EPERM,
setsockopt(s, IPPROTO_IP, IP_BINDANY, &(int){1}, sizeof(int)) ==
-1);
ATF_REQUIRE(close(s) == 0);
s = socket(AF_INET6, SOCK_STREAM, 0);
ATF_REQUIRE(s >= 0);
ATF_REQUIRE_ERRNO(EPERM,
setsockopt(s, IPPROTO_IPV6, IPV6_BINDANY, &(int){1}, sizeof(int)) ==
-1);
ATF_REQUIRE(close(s) == 0);
s = socket(AF_INET6, SOCK_DGRAM, 0);
ATF_REQUIRE(s >= 0);
ATF_REQUIRE_ERRNO(EPERM,
setsockopt(s, IPPROTO_IPV6, IPV6_BINDANY, &(int){1}, sizeof(int)) ==
-1);
ATF_REQUIRE(close(s) == 0);
}
/*
* Bind a socket to the specified address, optionally dropping privileges and
* setting one of the SO_REUSE* options first.
*
* Returns true if the bind succeeded, and false if it failed with EADDRINUSE.
*/
static bool
child_bind(const atf_tc_t *tc, int type, struct sockaddr *sa, int opt,
bool unpriv)
inpcb: Remove bogus SO_REUSEPORT(_LB) checks in in_pcbbind() This check for SO_REUSEPORT was added way back in commit 52b65dbe85faf. Per the commit log, this commit restricted this port-stealing check to unicast addresses, and then only if the existing socket does not have SO_REUSEPORT set. In other words, if there exists a socket bound to INADDR_ANY, and we bind a socket to INADDR_ANY with the same port, then the two sockets need not be owned by the same user if the existing socket has SO_REUSEPORT set. This is a surprising semantic; bugzilla PR 7713 gives some additional context. That PR makes a case for the behaviour described above when binding to a multicast address. But, the SO_REUSEPORT check is only applied when binding to a non-multicast address, so it doesn't really make sense. In the PR the committer notes that "unicast applications don't set SO_REUSEPORT", which makes some sense, but also refers to "multicast applications that bind to INADDR_ANY", which sounds a bit suspicious. OpenBSD performs the multicast check, but not the SO_REUSEPORT check. DragonflyBSD removed the SO_REUSEPORT (and INADDR_ANY) checks back in 2014 (commit 0323d5fde12a4). NetBSD explicitly copied our logic and still has it. The plot thickens: 20 years later, SO_REUSEPORT_LB was ported from DragonflyBSD: this option provides similar semantics to SO_REUSEPORT, but for unicast addresses it causes incoming connections/datagrams to be distributed among all sockets in the group. This commit (1a43cff92a20d) inverted the check for SO_REUSEPORT while adding one for SO_REUSEPORT_LB; this appears to have been inadvertent. However: - apparently no one has noticed that the semantics were changed; - sockets belonging to different users can now be bound to the same port so long as they belong to a single lbgroup bound to INADDR_ANY, which is not correct. Simply remove the SO_REUSEPORT(_LB) checks, as their original justification was dubious and their current implementation is wrong; add some tests. Reviewed by: glebius MFC after: 1 month Sponsored by: Klara, Inc. Sponsored by: Stormshield Differential Revision: https://reviews.freebsd.org/D47832 (cherry picked from commit 4f02a7d739b354eef38e19b25866f64842d69414)
2024-12-12 09:06:06 -05:00
{
const char *user;
pid_t child;
if (unpriv) {
if (!atf_tc_has_config_var(tc, "unprivileged_user"))
atf_tc_skip("unprivileged_user not set");
user = atf_tc_get_config_var(tc, "unprivileged_user");
} else {
user = NULL;
}
child = fork();
ATF_REQUIRE(child != -1);
if (child == 0) {
int s;
if (user != NULL) {
struct passwd *passwd;
passwd = getpwnam(user);
if (seteuid(passwd->pw_uid) != 0)
_exit(1);
}
s = socket(sa->sa_family, type, 0);
if (s < 0)
_exit(2);
if (bind(s, sa, sa->sa_len) == 0)
_exit(3);
if (errno != EADDRINUSE)
_exit(4);
if (opt != 0) {
if (setsockopt(s, SOL_SOCKET, opt, &(int){1},
sizeof(int)) != 0)
_exit(5);
}
if (bind(s, sa, sa->sa_len) == 0)
_exit(6);
if (errno != EADDRINUSE)
_exit(7);
_exit(0);
} else {
int status;
ATF_REQUIRE_EQ(waitpid(child, &status, 0), child);
ATF_REQUIRE(WIFEXITED(status));
status = WEXITSTATUS(status);
ATF_REQUIRE_MSG(status == 0 || status == 6,
"child exited with %d", status);
return (status == 6);
}
}
static bool
child_bind_priv(const atf_tc_t *tc, int type, struct sockaddr *sa, int opt)
{
return (child_bind(tc, type, sa, opt, false));
}
static bool
child_bind_unpriv(const atf_tc_t *tc, int type, struct sockaddr *sa, int opt)
{
return (child_bind(tc, type, sa, opt, true));
}
static int
bind_socket(int domain, int type, int opt, bool unspec, struct sockaddr *sa)
{
socklen_t slen;
int s;
s = socket(domain, type, 0);
ATF_REQUIRE(s >= 0);
if (domain == AF_INET) {
struct sockaddr_in sin;
bzero(&sin, sizeof(sin));
sin.sin_family = AF_INET;
sin.sin_len = sizeof(sin);
sin.sin_addr.s_addr = htonl(unspec ?
INADDR_ANY : INADDR_LOOPBACK);
sin.sin_port = htons(0);
ATF_REQUIRE(bind(s, (struct sockaddr *)&sin, sizeof(sin)) == 0);
slen = sizeof(sin);
} else /* if (domain == AF_INET6) */ {
struct sockaddr_in6 sin6;
bzero(&sin6, sizeof(sin6));
sin6.sin6_family = AF_INET6;
sin6.sin6_len = sizeof(sin6);
sin6.sin6_addr = unspec ? in6addr_any : in6addr_loopback;
sin6.sin6_port = htons(0);
ATF_REQUIRE(bind(s, (struct sockaddr *)&sin6, sizeof(sin6)) == 0);
slen = sizeof(sin6);
}
if (opt != 0) {
ATF_REQUIRE(setsockopt(s, SOL_SOCKET, opt, &(int){1},
sizeof(int)) == 0);
}
ATF_REQUIRE(getsockname(s, sa, &slen) == 0);
return (s);
}
static void
multibind_test(const atf_tc_t *tc, int domain, int type)
{
struct sockaddr_storage ss;
int opts[4] = { 0, SO_REUSEADDR, SO_REUSEPORT, SO_REUSEPORT_LB };
int s;
bool flags[2] = { false, true };
bool res;
for (size_t flagi = 0; flagi < nitems(flags); flagi++) {
for (size_t opti = 0; opti < nitems(opts); opti++) {
s = bind_socket(domain, type, opts[opti], flags[flagi],
(struct sockaddr *)&ss);
for (size_t optj = 0; optj < nitems(opts); optj++) {
int opt;
opt = opts[optj];
res = child_bind_priv(tc, type,
(struct sockaddr *)&ss, opt);
/*
* Multi-binding is only allowed when both
* sockets have SO_REUSEPORT or SO_REUSEPORT_LB
* set.
*/
if (opts[opti] != 0 &&
opts[opti] != SO_REUSEADDR && opti == optj)
ATF_REQUIRE(res);
else
ATF_REQUIRE(!res);
res = child_bind_unpriv(tc, type,
(struct sockaddr *)&ss, opt);
/*
* Multi-binding is only allowed when both
* sockets have the same owner.
*/
ATF_REQUIRE(!res);
inpcb: Remove bogus SO_REUSEPORT(_LB) checks in in_pcbbind() This check for SO_REUSEPORT was added way back in commit 52b65dbe85faf. Per the commit log, this commit restricted this port-stealing check to unicast addresses, and then only if the existing socket does not have SO_REUSEPORT set. In other words, if there exists a socket bound to INADDR_ANY, and we bind a socket to INADDR_ANY with the same port, then the two sockets need not be owned by the same user if the existing socket has SO_REUSEPORT set. This is a surprising semantic; bugzilla PR 7713 gives some additional context. That PR makes a case for the behaviour described above when binding to a multicast address. But, the SO_REUSEPORT check is only applied when binding to a non-multicast address, so it doesn't really make sense. In the PR the committer notes that "unicast applications don't set SO_REUSEPORT", which makes some sense, but also refers to "multicast applications that bind to INADDR_ANY", which sounds a bit suspicious. OpenBSD performs the multicast check, but not the SO_REUSEPORT check. DragonflyBSD removed the SO_REUSEPORT (and INADDR_ANY) checks back in 2014 (commit 0323d5fde12a4). NetBSD explicitly copied our logic and still has it. The plot thickens: 20 years later, SO_REUSEPORT_LB was ported from DragonflyBSD: this option provides similar semantics to SO_REUSEPORT, but for unicast addresses it causes incoming connections/datagrams to be distributed among all sockets in the group. This commit (1a43cff92a20d) inverted the check for SO_REUSEPORT while adding one for SO_REUSEPORT_LB; this appears to have been inadvertent. However: - apparently no one has noticed that the semantics were changed; - sockets belonging to different users can now be bound to the same port so long as they belong to a single lbgroup bound to INADDR_ANY, which is not correct. Simply remove the SO_REUSEPORT(_LB) checks, as their original justification was dubious and their current implementation is wrong; add some tests. Reviewed by: glebius MFC after: 1 month Sponsored by: Klara, Inc. Sponsored by: Stormshield Differential Revision: https://reviews.freebsd.org/D47832 (cherry picked from commit 4f02a7d739b354eef38e19b25866f64842d69414)
2024-12-12 09:06:06 -05:00
}
ATF_REQUIRE(close(s) == 0);
}
}
}
/*
* Try to bind two sockets to the same address/port tuple. Under some
* conditions this is permitted.
*/
ATF_TC(socket_afinet_multibind);
ATF_TC_HEAD(socket_afinet_multibind, tc)
{
atf_tc_set_md_var(tc, "require.user", "root");
atf_tc_set_md_var(tc, "require.config", "unprivileged_user");
}
ATF_TC_BODY(socket_afinet_multibind, tc)
{
multibind_test(tc, AF_INET, SOCK_STREAM);
multibind_test(tc, AF_INET, SOCK_DGRAM);
multibind_test(tc, AF_INET6, SOCK_STREAM);
multibind_test(tc, AF_INET6, SOCK_DGRAM);
}
static void
bind_connected_port_test(const atf_tc_t *tc, int domain)
{
struct sockaddr_in sin;
struct sockaddr_in6 sin6;
struct sockaddr *sinp;
int error, sd[3], tmp;
bool res;
/*
* Create a connected socket pair.
*/
sd[0] = socket(domain, SOCK_STREAM, 0);
ATF_REQUIRE_MSG(sd[0] >= 0, "socket failed: %s", strerror(errno));
sd[1] = socket(domain, SOCK_STREAM, 0);
ATF_REQUIRE_MSG(sd[1] >= 0, "socket failed: %s", strerror(errno));
if (domain == PF_INET) {
memset(&sin, 0, sizeof(sin));
sin.sin_family = AF_INET;
sin.sin_len = sizeof(sin);
sin.sin_addr.s_addr = htonl(INADDR_ANY);
sin.sin_port = htons(0);
sinp = (struct sockaddr *)&sin;
} else {
ATF_REQUIRE(domain == PF_INET6);
memset(&sin6, 0, sizeof(sin6));
sin6.sin6_family = AF_INET6;
sin6.sin6_len = sizeof(sin6);
sin6.sin6_addr = in6addr_any;
sin6.sin6_port = htons(0);
sinp = (struct sockaddr *)&sin6;
}
error = bind(sd[0], sinp, sinp->sa_len);
ATF_REQUIRE_MSG(error == 0, "bind failed: %s", strerror(errno));
error = listen(sd[0], 1);
ATF_REQUIRE_MSG(error == 0, "listen failed: %s", strerror(errno));
error = getsockname(sd[0], sinp, &(socklen_t){ sinp->sa_len });
ATF_REQUIRE_MSG(error == 0, "getsockname failed: %s", strerror(errno));
error = connect(sd[1], sinp, sinp->sa_len);
ATF_REQUIRE_MSG(error == 0, "connect failed: %s", strerror(errno));
tmp = accept(sd[0], NULL, NULL);
ATF_REQUIRE_MSG(tmp >= 0, "accept failed: %s", strerror(errno));
ATF_REQUIRE(close(sd[0]) == 0);
sd[0] = tmp;
/* bind() should succeed even from an unprivileged user. */
res = child_bind(tc, SOCK_STREAM, sinp, 0, false);
ATF_REQUIRE(!res);
res = child_bind(tc, SOCK_STREAM, sinp, 0, true);
ATF_REQUIRE(!res);
}
/*
* Normally bind() prevents port stealing by a different user, even when
* SO_REUSE* are specified. However, if the port is bound by a connected
* socket, then it's fair game.
*/
ATF_TC(socket_afinet_bind_connected_port);
ATF_TC_HEAD(socket_afinet_bind_connected_port, tc)
{
atf_tc_set_md_var(tc, "require.user", "root");
atf_tc_set_md_var(tc, "require.config", "unprivileged_user");
}
ATF_TC_BODY(socket_afinet_bind_connected_port, tc)
{
bind_connected_port_test(tc, AF_INET);
bind_connected_port_test(tc, AF_INET6);
}
ATF_TP_ADD_TCS(tp)
{
ATF_TP_ADD_TC(tp, socket_afinet);
ATF_TP_ADD_TC(tp, socket_afinet_bind_zero);
ATF_TP_ADD_TC(tp, socket_afinet_bind_ok);
ATF_TP_ADD_TC(tp, socket_afinet_poll_no_rdhup);
ATF_TP_ADD_TC(tp, socket_afinet_poll_rdhup);
ATF_TP_ADD_TC(tp, socket_afinet_stream_reconnect);
inpcb: Remove bogus SO_REUSEPORT(_LB) checks in in_pcbbind() This check for SO_REUSEPORT was added way back in commit 52b65dbe85faf. Per the commit log, this commit restricted this port-stealing check to unicast addresses, and then only if the existing socket does not have SO_REUSEPORT set. In other words, if there exists a socket bound to INADDR_ANY, and we bind a socket to INADDR_ANY with the same port, then the two sockets need not be owned by the same user if the existing socket has SO_REUSEPORT set. This is a surprising semantic; bugzilla PR 7713 gives some additional context. That PR makes a case for the behaviour described above when binding to a multicast address. But, the SO_REUSEPORT check is only applied when binding to a non-multicast address, so it doesn't really make sense. In the PR the committer notes that "unicast applications don't set SO_REUSEPORT", which makes some sense, but also refers to "multicast applications that bind to INADDR_ANY", which sounds a bit suspicious. OpenBSD performs the multicast check, but not the SO_REUSEPORT check. DragonflyBSD removed the SO_REUSEPORT (and INADDR_ANY) checks back in 2014 (commit 0323d5fde12a4). NetBSD explicitly copied our logic and still has it. The plot thickens: 20 years later, SO_REUSEPORT_LB was ported from DragonflyBSD: this option provides similar semantics to SO_REUSEPORT, but for unicast addresses it causes incoming connections/datagrams to be distributed among all sockets in the group. This commit (1a43cff92a20d) inverted the check for SO_REUSEPORT while adding one for SO_REUSEPORT_LB; this appears to have been inadvertent. However: - apparently no one has noticed that the semantics were changed; - sockets belonging to different users can now be bound to the same port so long as they belong to a single lbgroup bound to INADDR_ANY, which is not correct. Simply remove the SO_REUSEPORT(_LB) checks, as their original justification was dubious and their current implementation is wrong; add some tests. Reviewed by: glebius MFC after: 1 month Sponsored by: Klara, Inc. Sponsored by: Stormshield Differential Revision: https://reviews.freebsd.org/D47832 (cherry picked from commit 4f02a7d739b354eef38e19b25866f64842d69414)
2024-12-12 09:06:06 -05:00
ATF_TP_ADD_TC(tp, socket_afinet_bindany);
ATF_TP_ADD_TC(tp, socket_afinet_multibind);
ATF_TP_ADD_TC(tp, socket_afinet_bind_connected_port);
return atf_no_error();
}