rsyslog/runtime/netns_socket.c
Billie Alsup c78c587caf net: Add NetworkNamespace APIs
This moves NetworkNamespace functionality into
the net module. This allows the same code to
be reused across multiple tools and plugins.
The first usage is with omfwd, which is changed
to use the common net implementation.  Note
the net implementation is based on the original
omfwd implementation.  Subsequent PRs will be
opened for integrating this into omuxsock and
imtcp.

The original test case tcp_forwarding_ns_tpl.sh
was broken due to use of single quotes rather
than double quotes, thus preventing the proper
port number argument to be passed to the
listener.  Note this test must be run as
root.

Development has occurred across Fedora41 and
Fedora42, which uses glibc 2.40 and glibc2.41
respectively.  The valgrind suppressions are
updated to handle new glibc issues
accordingly.

Four new functions are callable through the
net module. These allow one to save a handle
to the current namespace, switch to a new
namespace by name, and restore the namespace
from the saved handle.  A fourth higher
level wrapper is used to open a socket in
a named network namespace, and handles the
invocation of the lower level functions.
Ideally this would be the only public
function, however it simplifies integration
into imtcp in the future (which doesn't
directly open sockets).  This may change
in the future as network namespaces are
integrated into more modules and plugins.

Signed-off-by: Billie Alsup <balsup@cisco.com>
2025-09-12 08:48:31 -07:00

167 lines
5.1 KiB
C

/* Implementation for netns_socket API
*
* This file is part of rsyslog.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
* -or-
* see COPYING.ASL20 in the source distribution
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "config.h"
#include <stdio.h>
#include <errno.h>
#include <fcntl.h>
#include <sched.h>
#include <string.h>
#include <sys/socket.h>
#include <sys/types.h>
#include "rsyslog.h"
#include "debug.h"
#include "errmsg.h"
#include "netns_socket.h"
/* Change to the given network namespace.
* This function based on previous implementation
* of tools/omfwd.c function changeToNs.
*/
rsRetVal netns_switch(const char *ns) {
DEFiRet;
#ifdef HAVE_SETNS
int ns_fd = -1;
char *nsPath = NULL;
if (ns && *ns) {
/* Build network namespace path */
if (asprintf(&nsPath, "/var/run/netns/%s", ns) == -1) {
// Some implementations say nsPath would be undefined on failure
nsPath = NULL;
LogError(0, RS_RET_OUT_OF_MEMORY, "%s: asprintf failed", __func__);
ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);
}
/* Open file descriptor of destination network namespace */
ns_fd = open(nsPath, O_RDONLY);
if (ns_fd < 0) {
LogError(errno, RS_RET_IO_ERROR, "%s: could not open namespace '%s'", __func__, ns);
ABORT_FINALIZE(RS_RET_IO_ERROR);
}
/* Change to the destination network namespace */
if (setns(ns_fd, CLONE_NEWNET) != 0) {
LogError(errno, RS_RET_IO_ERROR, "%s: could not change to namespace '%s'", __func__, ns);
ABORT_FINALIZE(RS_RET_IO_ERROR);
}
dbgprintf("%s: changed to network namespace '%s'\n", __func__, ns);
}
finalize_it:
free(nsPath);
if ((ns_fd >= 0) && (close(ns_fd) != 0)) {
LogError(errno, RS_RET_IO_ERROR, "%s: failed to close namespace '%s'", __func__, ns);
}
#else // ndef HAVE_SETNS
if (ns && *ns) {
LogError(ENOSYS, RS_RET_VALUE_NOT_SUPPORTED, "%s: could not change to namespace '%s'", __func__, ns);
ABORT_FINALIZE(RS_RET_VALUE_NOT_SUPPORTED);
}
finalize_it:
#endif // ndef HAVE_SETNS
RETiRet;
}
/* Return to the startup network namespace.
* This function based on code in tools/omfwd.c
*/
rsRetVal ATTR_NONNULL() netns_restore(int *fd) {
DEFiRet;
#ifdef HAVE_SETNS
if (*fd >= 0) {
if (setns(*fd, CLONE_NEWNET) != 0) {
LogError(errno, RS_RET_IO_ERROR, "%s: could not return to startup namespace", __func__);
ABORT_FINALIZE(RS_RET_IO_ERROR);
}
dbgprintf("%s: returned to startup network namespace\n", __func__);
}
finalize_it:
#endif // def HAVE_SETNS
if (*fd >= 0 && close(*fd) != 0) {
LogError(errno, RS_RET_IO_ERROR, "%s: could not close startup namespace fd", __func__);
}
*fd = -1;
RETiRet;
}
/* Save the current network namespace fd
*/
rsRetVal ATTR_NONNULL() netns_save(int *fd) {
DEFiRet;
/*
* The fd must always point to either a valid fd
* or to -1. We expect it to be -1 on entry here.
* To avoid bugs, or possible descriptor leaks,
* check that it is always -1 on entry.
*/
#ifdef HAVE_SETNS
if (*fd != -1) {
LogError(0, RS_RET_CODE_ERR, "%s: called with uninitialized descriptor", __func__);
ABORT_FINALIZE(RS_RET_CODE_ERR);
}
*fd = open("/proc/self/ns/net", O_RDONLY);
if (*fd == -1) {
LogError(errno, RS_RET_IO_ERROR, "%s: could not access startup namespace", __func__);
ABORT_FINALIZE(RS_RET_IO_ERROR);
}
dbgprintf("%s: saved startup network namespace\n", __func__);
finalize_it:
#endif // def HAVE_SETNS
RETiRet;
}
rsRetVal netns_socket(int *fdp, int domain, int type, int protocol, const char *ns) {
DEFiRet;
int fd = -1;
#ifndef HAVE_SETNS
if (ns && *ns) {
LogError(0, RS_RET_VALUE_NOT_SUPPORTED, "Network namespaces are not supported");
ABORT_FINALIZE(RS_RET_VALUE_NOT_SUPPORTED);
}
#else /* def HAVE_SETNS */
rsRetVal iRet_restore;
int ns_fd = -1;
if (ns && *ns) {
CHKiRet(netns_save(&ns_fd));
CHKiRet(netns_switch(ns));
}
#endif /* def HAVE_SETNS */
*fdp = fd = socket(domain, type, protocol);
if (fd == -1) {
LogError(errno, RS_RET_NO_SOCKET, "%s: socket(%d, %d, %d) failed", __func__, domain, type, protocol);
ABORT_FINALIZE(RS_RET_NO_SOCKET);
}
finalize_it:
#ifdef HAVE_SETNS
iRet_restore = netns_restore(&ns_fd);
if (iRet == RS_RET_OK) iRet = iRet_restore;
#endif /* def HAVE_SETNS */
if (iRet != RS_RET_OK && fd != -1) {
(void)close(fd);
*fdp = -1;
}
RETiRet;
}