* FreeBSD
* GNU/Linux
- * Mac OS X
+ * macOS
* NetBSD
* OpenBSD
* Solaris
+Windows is not supported but you can use
+[WinLLDPService](https://github.com/raspi/WinLLDPService/) as a
+transmit-only agent.
+
Installation
------------
-To compile lldpd, use the following:
+For general instructions [prefer the
+website](http://vincentbernat.github.io/lldpd/installation.html),
+including building from released tarballs.
+
+To compile lldpd from Git, use the following commands:
+ ./autogen.sh
./configure
make
sudo make install
-You need libevent that you can grab from http://libevent.org or
-install from your package system (libevent-dev for Debian/Ubuntu and
-libevent-devel for Redhat/Fedora/CentOS/SuSE).
-
-If your system does not have libevent, ./configure will use the
-shipped copy and compile it statically.
-
-If it complains about a missing agent/struct.h, your installation of
-Net-SNMP is incomplete. The easiest way to fix this is to provide an
-empty struct.h:
-
- touch src/struct.h
-
lldpd uses privilege separation to increase its security. Two
processes, one running as root and doing minimal stuff and the other
running as an unprivileged user into a chroot doing most of the stuff,
are cooperating. You need to create a user called `_lldpd` in a group
`_lldpd` (this can be change with `./configure`). You also need to
-create an empty directory `/var/run/lldpd` (it needs to be owned by
-root, not `_lldpd`!). If you get fuzzy timestamps from syslog, copy
-`/etc/locatime` into the chroot.
+create an empty directory `/usr/local/var/run/lldpd` (it needs to be
+owned by root, not `_lldpd`!). If you get fuzzy timestamps from
+syslog, copy `/etc/locatime` into the chroot.
`lldpcli` lets one query information collected through the command
line. If you don't want to run it as root, just install it setuid or
setgid `_lldpd`.
-Installation (Mac OS X)
+Installation (macOS)
-----------------------
-The same procedure as above applies for Mac OS X. However, there are
+The same procedure as above applies for macOS. However, there are
simpler alternatives:
- 1. Use [Homebrew](http://mxcl.github.io/homebrew/):
+ 1. Use [Homebrew](https://brew.sh):
brew install lldpd
# Or, for the latest version:
brew install https://raw.github.com/vincentbernat/lldpd/master/osx/lldpd.rb
- 2. Build an OSX installer package which should work on the same
- version of OS X (it is important to use a separate build
- directory):
+ 2. Build an macOS installer package which should work on the same
+ version of macOS:
mkdir build && cd build
- ../configure --prefix=/usr --localstatedir=/var --sysconfdir=/private/etc --with-embedded-libevent
- make -C osx pkg ARCHS="i386 x86_64"
+ ../configure --prefix=/usr/local --localstatedir=/var --sysconfdir=/private/etc --with-embedded-libevent \
+ --without-snmp
+ make -C osx pkg
- If you want to compile for an older version of Mac OS X, you need
+ If you want to compile for an older version of macOS, you need
to find the right SDK and issues commands like those:
SDK=/Developer/SDKs/MacOSX10.6.sdk
mkdir build && cd build
- ../configure --prefix=/usr --localstatedir=/var --sysconfdir=/private/etc --with-embedded-libevent \
+ ../configure --prefix=/usr/local --localstatedir=/var --sysconfdir=/private/etc --with-embedded-libevent \
+ --without-snmp \
CFLAGS="-mmacosx-version-min=10.6 -isysroot $SDK" \
LDFLAGS="-mmacosx-version-min=10.6 -isysroot $SDK"
- make -C osx pkg ARCHS="i386 x86_64"
+ make -C osx pkg
+
+ With recent SDK, you don't need to specify an alternate SDK. They
+ are organized in a way that should enable compatibility with older
+ versions of OSX:
+
+ mkdir build && cd build
+ ../configure --prefix=/usr/local --localstatedir=/var --sysconfdir=/private/etc --with-embedded-libevent \
+ --without-snmp \
+ CFLAGS="-mmacosx-version-min=10.9" \
+ LDFLAGS="-mmacosx-version-min=10.9"
+ make -C osx pkg
+
+ You can check with `otool -l` that you got what you expected in
+ term of supported versions.
If you don't follow the above procedures, you will have to create the
user/group `_lldpd`. Have a look at how this is done in
through EDP/CDP/FDP/SONMP are integrated with other informations and
can be queried with `lldpcli` or through SNMP.
+More information:
+ * http://en.wikipedia.org/wiki/Link_Layer_Discovery_Protocol
+ * http://standards.ieee.org/getieee802/download/802.1AB-2005.pdf
+ * http://wiki.wireshark.org/LinkLayerDiscoveryProtocol
+
+Compatibility with older kernels
+--------------------------------
+
+If you have a kernel older than Linux 2.6.39, you need to compile
+lldpd with `--enable-oldies` to enable some compatibility functions:
+otherwise, lldpd will only rely on Netlink to receive bridge, bond and
+VLAN information.
+
For bonding, you need 2.6.24 (in previous version, PACKET_ORIGDEV
affected only non multicast packets). See:
* http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=8032b46489e50ef8f3992159abd0349b5b8e476c
Otherwise, a packet received on a bond will be affected to all
-interfaces of the bond.
+interfaces of the bond. In this case, lldpd will affect a received
+randomly to one of the interface (so a neighbor may be affected to the
+wrong interface).
-On 2.6.27, we are able to receive packets on real interface for bonded
+On 2.6.27, we are able to receive packets on real interface for enslaved
devices. This allows one to get neighbor information on active/backup
bonds. Without the 2.6.27, lldpd won't receive any information on
inactive slaves. Here are the patchs (thanks to Joe Eykholt):
* http://www.freebsd.org/cgi/query-pr.cgi?pr=138620
-Some devices (notably Cisco IOS) send frames on the native VLAN while
-they should send them untagged. If your network card does not support
-accelerated VLAN, you will receive those frames as well. However, if
-your network card handles VLAN encapsulation/decapsulation (check with
+Some devices (notably Cisco IOS) send frames tagged with the native
+VLAN while they should send them untagged. If your network card does
+not support accelerated VLAN, you will receive those frames as long as
+the corresponding interface exists (see below). However, if your
+network card handles VLAN encapsulation/decapsulation (check with
`ethtool -k`), you need a recent kernel to be able to receive those
frames without listening on all available VLAN. Starting from Linux
2.6.27, lldpd is able to capture VLAN frames when VLAN acceleration is
supported by the network card. Here is the patch:
- http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=bc1d0411b804ad190cdadabac48a10067f17b9e6
-On some versions, frames are sent on VLAN 1. If your network card
-support accelerated VLAN, you need to subscribe to this VLAN as
-well. The Linux kernel does not provide any interface for this. The
-easiest way is to create the VLAN for each port:
+ * http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=bc1d0411b804ad190cdadabac48a10067f17b9e6
+
+On some other versions, frames are sent on VLAN 1. If this is not the
+native VLAN and if your network card support accelerated VLAN, you
+need to subscribe to this VLAN as well. The Linux kernel does not
+provide any interface for this. The easiest way is to create the VLAN
+for each port:
ip link add link eth0 name eth0.1 type vlan id 1
ip link set up dev eth0.1
10:54:06.431154 f0:29:29:1d:7c:01 > 01:80:c2:00:00:0e, ethertype 802.1Q (0x8100), length 363: vlan 1, p 7, ethertype LLDP, LLDP, name SW-APP-D07.VTY, length 345
-In this case, just create VLAN 1 will fix the situation.
+In this case, just create VLAN 1 will fix the situation. There are
+other solutions:
-More information:
- * http://en.wikipedia.org/wiki/Link_Layer_Discovery_Protocol
- * http://standards.ieee.org/getieee802/download/802.1AB-2005.pdf
- * http://wiki.wireshark.org/LinkLayerDiscoveryProtocol
+ 1. Disable VLAN acceleration on the receive side (`ethtool -K eth0
+ rxvlan off`) but this may or may not work. Check if there are
+ similar properties that could apply with `ethtool -k eth0`.
+ 2. Put the interface in promiscuous mode with `ip link set
+ promisc on dev eth0`.
+
+The last solution can be done directly by `lldpd` (on Linux only) by
+using the option `configure system interface promiscuous`.
+
+On modern networks, the performance impact should be nonexistent.
Development
-----------
During development, you may want to execute lldpd at its current
-location inside of doing `make install`. The correct way to do this is
+location instead of doing `make install`. The correct way to do this is
to issue the following command:
sudo libtool execute src/daemon/lldpd -L $PWD/src/client/lldpcli -d
`lldpcli` it will start in an unconfigured mode and won't send or
accept LLDP frames.
+You can use [afl](http://lcamtuf.coredump.cx/afl/) to test some
+aspects of lldpd. To test frame decoding, you can do something like
+that:
+
+ export AFL_USE_ASAN=1 # only on 32bit arch
+ ./configure CC=afl-gcc
+ make clean check
+ cd tests
+ mkdir inputs
+ mv *.pcap inputs
+ afl-fuzz -i inputs -o outputs ./decode @@
+
+There is a general test suite with `make check`. It's also possible to
+run integration tests. They need [py.test](http://pytest.org/latest/)
+and rely on Linux containers to be executed.
+
+To enable code coverage, use:
+
+ ../configure --prefix=/usr --sysconfdir=/etc --localstatedir=/var \
+ --enable-sanitizers --enable-gcov --with-snmp \
+ CFLAGS="-O0 -g"
+ make
+ make check
+ # maybe, run integration tests
+ lcov --base-directory $PWD/src/lib \
+ --directory src --capture --output-file gcov.info
+ genhtml gcov.info --output-directory coverage
+
Embedding
---------
tcpdump -s0 -vv -pni eth0 ether dst 01:80:c2:00:00:0e
+Intel X710 cards may handle LLDP themselves, intercepting any incoming
+packets. If you don't see anything through `tcpdump`, check if you
+have such a card (with `lspci`) and stop the embedded LLDP daemon:
+
+ for f in /sys/kernel/debug/i40e/*/command; do
+ echo lldp stop > $f
+ done
+
License
-------
projects / thirdparty / lldpd.git / blobdiff