]> git.ipfire.org Git - thirdparty/lldpd.git/log
thirdparty/lldpd.git
4 years agolldp: when receiving a shutdown LLDPU, don't clear chassis information fix/chassis-shutdown 357/head
Vincent Bernat [Tue, 8 Oct 2019 17:35:41 +0000 (19:35 +0200)] 
lldp: when receiving a shutdown LLDPU, don't clear chassis information

The chassis may be shared with another port. When the MSAP is known
and we receive a shutdown LLDPDU, just leave the original chassis as
is instead of copying information from the new chassis to the old
chassis.

Fix #348.

4 years agolldp: validate a bit more received LLDP frames
Vincent Bernat [Tue, 1 Oct 2019 19:42:42 +0000 (21:42 +0200)] 
lldp: validate a bit more received LLDP frames

Notably, we ensure the order and unicity of Chassis ID, Port ID and
TTL TLV. For Chassis ID and Port ID, we also ensure the maximum size
does not exceed 256.

Fix #351.

4 years agointerfaces: only register protocol handler for LLDP when only LLDP enabled
Vincent Bernat [Tue, 1 Oct 2019 04:18:52 +0000 (06:18 +0200)] 
interfaces: only register protocol handler for LLDP when only LLDP enabled

On Linux, the drop counter is increased on unhandled packets. We are
using a raw socket with ETH_P_ALL, so we get a copy of the packet. The
original packet is ultimately dropped later and this increases the
drop counter associated to the interface on Linux.

When listening only to LLDP, use ETH_P_LLDP instead of ETH_P_ALL to
avoid this.

Fix #296.

4 years agolldpctl: don't reopen control socket each time we want to lock
Vincent Bernat [Sun, 29 Sep 2019 12:09:54 +0000 (14:09 +0200)] 
lldpctl: don't reopen control socket each time we want to lock

4 years agolldpctl: put a lock around some commands to avoid race conditions
Vincent Bernat [Tue, 10 Sep 2019 07:55:54 +0000 (09:55 +0200)] 
lldpctl: put a lock around some commands to avoid race conditions

Fix #343

4 years agolog: don't use black for coloring DBG
Vincent Bernat [Sun, 29 Sep 2019 11:58:23 +0000 (13:58 +0200)] 
log: don't use black for coloring DBG

As foreground, it may be gray, but it could be too dark. Use dark cyan
instead.

4 years agonetlink: ignore IFLA_LINK if we previously received IFLA_LINK_NETNSID
Vincent Bernat [Mon, 23 Sep 2019 19:35:14 +0000 (21:35 +0200)] 
netlink: ignore IFLA_LINK if we previously received IFLA_LINK_NETNSID

Starting from 5.2 (commit feadc4b6cf42), IFLA_LINK appear after
IFLA_LINK_NETNSID. Handle this case correctly.

4 years agointerfaces: don't log errno when detecting loops
Vincent Bernat [Sun, 15 Sep 2019 18:53:21 +0000 (20:53 +0200)] 
interfaces: don't log errno when detecting loops

4 years agotests: fix skip instruction
Vincent Bernat [Sun, 15 Sep 2019 17:40:03 +0000 (19:40 +0200)] 
tests: fix skip instruction

4 years agotests: don't use pytest.config
Vincent Bernat [Sun, 15 Sep 2019 17:19:06 +0000 (19:19 +0200)] 
tests: don't use pytest.config

It's deprecated in 5.0.

4 years agointerfaces: enable matching on interface name for management address feature/management-device
Vincent Bernat [Sun, 15 Sep 2019 15:45:52 +0000 (17:45 +0200)] 
interfaces: enable matching on interface name for management address

We allow the user to match a management address using the interface
name by specifying the interface name as a pattern. The same rules as
for specifying IP patterns apply but there is no notion of exact match
for an interface.

Fix #349.

4 years agolldpd: fix bitfield
Wataru Ashihara [Wed, 4 Sep 2019 06:52:55 +0000 (15:52 +0900)] 
lldpd: fix bitfield

4 years agoclient: fix parameter order
Wataru Ashihara [Wed, 4 Sep 2019 02:59:33 +0000 (11:59 +0900)] 
client: fix parameter order

4 years agolib: fix memory leak
Wataru Ashihara [Wed, 4 Sep 2019 02:59:03 +0000 (11:59 +0900)] 
lib: fix memory leak

4 years agofix typo
Wataru Ashihara [Wed, 4 Sep 2019 02:52:56 +0000 (11:52 +0900)] 
fix typo

4 years agolib: fix enumeration
Wataru Ashihara [Tue, 6 Feb 2018 12:09:40 +0000 (21:09 +0900)] 
lib: fix enumeration

4 years agobuild: disable warnings on cast alignments
Vincent Bernat [Sat, 27 Jul 2019 17:23:12 +0000 (19:23 +0200)] 
build: disable warnings on cast alignments

clang is often wront about it (it increases alignment requirement, but
the surrounding structure ensure the alignment is correct). Dunno if
gcc is smarter or just ignore most of these problems.

4 years agotests: be verbose when running make on CI when an error happens
Vincent Bernat [Sat, 27 Jul 2019 16:53:22 +0000 (18:53 +0200)] 
tests: be verbose when running make on CI when an error happens

Otherwise, we don't really know why we get errors.

4 years agoclient: don't display management interface index if unknown
Vincent Bernat [Sat, 27 Jul 2019 15:36:35 +0000 (17:36 +0200)] 
client: don't display management interface index if unknown

4 years agointerfaces: compute interface index for fixed management address
Vincent Bernat [Sat, 27 Jul 2019 15:28:01 +0000 (17:28 +0200)] 
interfaces: compute interface index for fixed management address

When management address is provided without a pattern, fetch the
appropriate interface index if the interface is known. Thanks to
@kefins for the actual patch.

Fix #338.

4 years agolib: expose management interface index
Vincent Bernat [Sat, 27 Jul 2019 15:11:40 +0000 (17:11 +0200)] 
lib: expose management interface index

4 years agolib: fix missing words in lldpctl.h
Vincent Bernat [Sat, 27 Jul 2019 15:00:25 +0000 (17:00 +0200)] 
lib: fix missing words in lldpctl.h

4 years agoinclude: update headers to 4.19
Vincent Bernat [Sun, 7 Jul 2019 17:24:21 +0000 (19:24 +0200)] 
include: update headers to 4.19

It makes things compile correctly with Jessie

4 years agoJune 15th 2019 was a Saturday, not a Sunday
eb3095 [Wed, 17 Jul 2019 21:56:46 +0000 (17:56 -0400)] 
June 15th 2019 was a Saturday, not a Sunday

This addresses the following issue,

RPM build errors:
error: Bad exit status from /var/tmp/rpm-tmp.tUWJAJ (%prep)
    bogus date in %changelog: Sun Jun 15 2019 Vincent Bernat <bernat@luffy.cx> - 1.0.4-1

4 years agoosx: update recipe for HomeBrew
Vincent Bernat [Sat, 15 Jun 2019 09:25:40 +0000 (11:25 +0200)] 
osx: update recipe for HomeBrew

4 years agorelease: 1.0.4 1.0.4
Vincent Bernat [Sat, 15 Jun 2019 09:09:44 +0000 (11:09 +0200)] 
release: 1.0.4

4 years agolldpd: document expectation for LLDP-MED device configuration
Vincent Bernat [Wed, 12 Jun 2019 20:10:56 +0000 (22:10 +0200)] 
lldpd: document expectation for LLDP-MED device configuration

When using class II and class III endpoints, some additional TLVs are
mandatory. This is not enforced by lldpd.

See #335.

4 years agonetlink: make the netlink socket totally non blocking
Vincent Bernat [Mon, 10 Jun 2019 09:17:52 +0000 (11:17 +0200)] 
netlink: make the netlink socket totally non blocking

See #333.

4 years agonetlink: handle blocking read from netlink socket
Vincent Bernat [Sun, 9 Jun 2019 06:13:06 +0000 (08:13 +0200)] 
netlink: handle blocking read from netlink socket

It seems it is possible to run into a condition where the netlink
socket is not available for read. Set the MSG_DONTWAIT flag and fetch
an error if there is any.

Fix #333

4 years agosnmp: implement lldpRemOrgDefInfoTable for remote custom TLVs
Vincent Bernat [Wed, 29 May 2019 16:58:09 +0000 (18:58 +0200)] 
snmp: implement lldpRemOrgDefInfoTable for remote custom TLVs

As a simplification, lldpRemOrgDefInfoIndex is 1 for the first custom
TLV of a given port and is increased by 1 for each new TLV. This is
not what is encouraged in the MIB:

> An agent is encouraged to assign monotonically increasing index
> values to new entries, starting with one, after each reboot. It is
> considered unlikely that the lldpRemOrgDefInfoIndex will wrap
> between reboots.

However, it is simpler to implement it this way as we don't need to
record the index inside the `lldpd_custom` structure. Also, the index
will increase even for a different OUI or subtype as we do not want to
sort the custom TLVs.

Fix #330.

4 years agodebian: depends on lsb-base
Vincent Bernat [Sun, 2 Jun 2019 07:02:42 +0000 (09:02 +0200)] 
debian: depends on lsb-base

This is needed for SysV init script

4 years agodebian: no need to BD on lsb-release anymore
Vincent Bernat [Sun, 2 Jun 2019 07:00:47 +0000 (09:00 +0200)] 
debian: no need to BD on lsb-release anymore

4 years agodoc: update Android instructions for ARM64
Vincent Bernat [Wed, 15 May 2019 04:20:27 +0000 (06:20 +0200)] 
doc: update Android instructions for ARM64

Nowadays, this is a more common platform than ARM.

4 years agoclient: use bold instead of a color for command completions
Vincent Bernat [Sat, 4 May 2019 06:58:38 +0000 (08:58 +0200)] 
client: use bold instead of a color for command completions

The color needs to be readable on both light and dark backgrounds and
should be readable on most themes. 1;35m would be a fit, but let's
stay safe by just using bold.

Fix #326

5 years agoagent: do not increase statsAgeoutsTotal when receiving a shutdown frame
Vincent Bernat [Tue, 9 Apr 2019 05:07:14 +0000 (07:07 +0200)] 
agent: do not increase statsAgeoutsTotal when receiving a shutdown frame

Fix #325

5 years agolldp: increase statsTLVsUnrecognizedTotal on unknown TLV
Vincent Bernat [Sat, 6 Apr 2019 19:17:25 +0000 (21:17 +0200)] 
lldp: increase statsTLVsUnrecognizedTotal on unknown TLV

This was done for organization TLVs, but not for other TLVs.

Fix #323.

5 years agotests: update pytest
Vincent Bernat [Sat, 6 Apr 2019 12:29:13 +0000 (14:29 +0200)] 
tests: update pytest

And also, don't specify too many dependencies, just the essential
ones.

5 years agoclient: rename max-neighbor command
Vincent Bernat [Sat, 6 Apr 2019 11:20:51 +0000 (13:20 +0200)] 
client: rename max-neighbor command

5 years agotests: add test for global configuration settings
Vincent Bernat [Sat, 6 Apr 2019 11:15:38 +0000 (13:15 +0200)] 
tests: add test for global configuration settings

5 years agoclient: ability to set maximum of neighbors from lldpcli
Vincent Bernat [Sat, 6 Apr 2019 10:13:24 +0000 (12:13 +0200)] 
client: ability to set maximum of neighbors from lldpcli

5 years agopriv: include limits.h for PATH_MAX
Sören Tempel [Mon, 25 Mar 2019 17:43:03 +0000 (18:43 +0100)] 
priv: include limits.h for PATH_MAX

5 years agolib: use an unique variable as iterator in foreach macro
Vincent Bernat [Fri, 15 Mar 2019 07:25:09 +0000 (08:25 +0100)] 
lib: use an unique variable as iterator in foreach macro

This lessen the chance of the `iter` variable to shadow a user-defined
variable. This is also a tentative to help #312, even if the scope of
the `iter` variable should ensure we can nest two loops without any
issue.

5 years agobuild: warn on implicit fallthrough for case statements
Vincent Bernat [Wed, 13 Mar 2019 07:54:22 +0000 (08:54 +0100)] 
build: warn on implicit fallthrough for case statements

5 years agobuild: enable increased reliability of stack overflow detection
Vincent Bernat [Thu, 10 Jan 2019 21:36:55 +0000 (22:36 +0100)] 
build: enable increased reliability of stack overflow detection

See https://developers.redhat.com/blog/2018/03/21/compiler-and-linker-flags-gcc/

5 years agointerfaces/linux: make dsa special
Brandon Streiff [Mon, 16 Jul 2018 21:53:56 +0000 (16:53 -0500)] 
interfaces/linux: make dsa special

There used to be specific exemptions carved out for "veth" and "dsa",
which were removed in b8db52bd7c7d ("interfaces/linux: blacklist some
drivers instead of whitelisting"). "veth" was restored in 2958b9d48940
("interfaces/linux: make veth special"). This commit restores the
whitelist for dsa devices as well.

5 years agobuild: fix compilation with `--enable-fdp --disable-cdp`
Vincent Bernat [Tue, 8 Jan 2019 22:19:15 +0000 (17:19 -0500)] 
build: fix compilation with `--enable-fdp --disable-cdp`

Fix #315

5 years agodaemon: allow lseek for seccomp
Vincent Bernat [Mon, 10 Dec 2018 16:21:41 +0000 (17:21 +0100)] 
daemon: allow lseek for seccomp

5 years agoosx: update version in HomeBrew formula
Vincent Bernat [Mon, 10 Dec 2018 16:20:50 +0000 (17:20 +0100)] 
osx: update version in HomeBrew formula

5 years agorelease: 1.0.3 1.0.3
Vincent Bernat [Mon, 10 Dec 2018 14:02:58 +0000 (15:02 +0100)] 
release: 1.0.3

5 years agodaemon: fix creation of chroot directory
Vincent Bernat [Mon, 10 Dec 2018 13:54:22 +0000 (14:54 +0100)] 
daemon: fix creation of chroot directory

Because the current path was uninitialized, we were creating
semi-random directory at the root of the filesystem.

5 years agoosx: update version in HomeBrew formula
Vincent Bernat [Tue, 4 Dec 2018 07:20:06 +0000 (08:20 +0100)] 
osx: update version in HomeBrew formula

5 years agoosx: sync HomeBrew formula
Vincent Bernat [Tue, 4 Dec 2018 07:19:26 +0000 (08:19 +0100)] 
osx: sync HomeBrew formula

5 years agodaemon: check if started through systemd first
Vincent Bernat [Mon, 3 Dec 2018 08:33:18 +0000 (09:33 +0100)] 
daemon: check if started through systemd first

When running with `-d`, we may still need to send a notification to
systemd, even if we didn't plan to daemonize.

Fix #309.

5 years agodaemon: remove unused agent_default_agentx_socket() function
Vincent Bernat [Sat, 1 Dec 2018 09:47:37 +0000 (10:47 +0100)] 
daemon: remove unused agent_default_agentx_socket() function

It's not needed anymore since 83577a2b3b88 and on older versions of
Net-SNMP, it uses an unexposed directive.

5 years agorelease: 1.0.2 1.0.2
Vincent Bernat [Sat, 1 Dec 2018 09:21:01 +0000 (10:21 +0100)] 
release: 1.0.2

5 years agodaemon: don't enable ProtectSystem by default
Vincent Bernat [Fri, 30 Nov 2018 21:48:36 +0000 (22:48 +0100)] 
daemon: don't enable ProtectSystem by default

If the chroot is in `/usr` (like `/usr/local/var/run/lldpd` which is
the default), neither systemd nor lldpd will be able to create and
write to it. This may be solved with `ReadWritePaths` (unsure if it
would create the directory), but this doesn't exist in older versions
of systemd.

Just comment the directive to let people know it exists and should
work in most cases.

Fix #308

5 years agointerfaces: remove specific handling for bonds except with --enable-oldies
Vincent Bernat [Wed, 28 Nov 2018 13:56:47 +0000 (14:56 +0100)] 
interfaces: remove specific handling for bonds except with --enable-oldies

Starting from Linux 4.19, LLDP packets are transmitted back to the
bond devices and it seems the original interface is lost in the
process. Therefore, packets are duplicated to both members. Upstream
commit is:

    commit 6a9e461f6fe4434e6172304b69774daff9a3ac4c
    Author: Mahesh Bandewar <maheshb@google.com>
    Date:   Mon Sep 24 14:39:42 2018 -0700

    bonding: pass link-local packets to bonding master also.

    Commit b89f04c61efe ("bonding: deliver link-local packets with
    skb->dev set to link that packets arrived on") changed the behavior
    of how link-local-multicast packets are processed. The change in
    the behavior broke some legacy use cases where these packets are
    expected to arrive on bonding master device also.

    This patch passes the packet to the stack with the link it arrived
    on as well as passes to the bonding-master device to preserve the
    legacy use case.

Fixes: b89f04c61efe ("bonding: deliver link-local packets with skb->dev set to link that packets arrived on")
Reported-by: Michal Soltys <soltys@ziu.info>
Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code to handle bond devices is not needed since Linux 2.6.27.
Therefore, move it to the `--enable-oldies` option.

Fix #306.

5 years agodoc: update documentation for Android
Vincent Bernat [Wed, 7 Nov 2018 12:25:35 +0000 (13:25 +0100)] 
doc: update documentation for Android

See #305 for more details.

5 years agoAllow linux kernel to autoprobe correct socket_nl.nl_pid address
Pavel Shirshov [Sat, 20 Oct 2018 02:15:44 +0000 (19:15 -0700)] 
Allow linux kernel to autoprobe correct socket_nl.nl_pid address

5 years agodaemon: do not explicitely inline functions
Vincent Bernat [Tue, 2 Oct 2018 18:36:37 +0000 (20:36 +0200)] 
daemon: do not explicitely inline functions

As we are using `-Winline`, if it fails, we get a warning. Let the
compiler decide if something has to be inlined. As we use only static
functions, it should be easy to inline if possible.

5 years agobuild: ability to disable libbsd with --without-libbsd
Vincent Bernat [Tue, 25 Sep 2018 08:57:49 +0000 (10:57 +0200)] 
build: ability to disable libbsd with --without-libbsd

This is useful when user wants to ensure reproducibility of the build
whatever libbsd is present or not.

5 years agomarshal: avoid NULL pointer arithmetic
Vincent Bernat [Mon, 24 Sep 2018 13:20:43 +0000 (15:20 +0200)] 
marshal: avoid NULL pointer arithmetic

This is UB and clang warns about this. Ensure we use an int of the
same size as a pointer and convert it directly.

5 years agolib: use constants from lldp-const.h for Dot3 MAU
Vincent Bernat [Mon, 24 Sep 2018 13:12:21 +0000 (15:12 +0200)] 
lib: use constants from lldp-const.h for Dot3 MAU

5 years agodot3: fix 100BASE-T4 MAU
Vincent Bernat [Mon, 24 Sep 2018 13:00:42 +0000 (15:00 +0200)] 
dot3: fix 100BASE-T4 MAU

Fix #297

5 years agointerfaces: provide clearer messages about ethtool problems
Vincent Bernat [Sat, 11 Aug 2018 17:43:20 +0000 (19:43 +0200)] 
interfaces: provide clearer messages about ethtool problems

Logs can confuse people on how important a problem is. Tell them
exactly what this would affect.

5 years agodaemon: implement mkdir -p directly in lldpd
Vincent Bernat [Wed, 8 Aug 2018 21:06:39 +0000 (23:06 +0200)] 
daemon: implement mkdir -p directly in lldpd

It's difficult to know the path to mkdir. If we use the one from
autoconf (@mkdir_p@), we get the path from the host, not the target.
If we hardcode `/bin/mkdir`, we may not work on platforms like NixOS.

See https://github.com/NixOS/nixpkgs/issues/44507.

5 years agodoc: add latest commit to NEWS
Vincent Bernat [Fri, 20 Jul 2018 22:01:34 +0000 (00:01 +0200)] 
doc: add latest commit to NEWS

5 years agoMerge pull request #289 from GustavWi/master
Vincent Bernat [Sun, 15 Jul 2018 17:23:15 +0000 (19:23 +0200)] 
Merge pull request #289 from GustavWi/master

Add support for CDP PD PoE negotiation.

5 years agoAdd support for PD PoE negotiation. 289/head
Gustav Wiklander [Thu, 21 Jun 2018 08:49:37 +0000 (10:49 +0200)] 
Add support for PD PoE negotiation.

Power requests refer to the power at the PSE.
Thus the loss offset caused by the cable has to be added
to the power request. Also the power received from the PSE
must subtract the cable loss to be compatible with lldp.
There are three TLVs for CDPv2 PoE negotiation.
Power Consumption: Current maximum power consumption of PD.
Power Request: Wanted maximum power consumption of PD.
Power Available: Power output from PSE.

Only used if lldp PoE is not supported by switch.

A cisco switch which does support both lldp and cdp will
use the protocol which is first to transmit a package.

Change-Id: Ib45ae582799c50a8d839120c28a5080e582c66e9

5 years agoredhat: install lldpcli as setuid _lldpd
Vincent Bernat [Wed, 4 Jul 2018 09:45:51 +0000 (11:45 +0200)] 
redhat: install lldpcli as setuid _lldpd

5 years agoosx: update brew formula
Vincent Bernat [Wed, 20 Jun 2018 16:56:52 +0000 (18:56 +0200)] 
osx: update brew formula

5 years agoMerge pull request #285 from vincentbernat/feature/linux-caps
Vincent Bernat [Sat, 16 Jun 2018 17:43:56 +0000 (19:43 +0200)] 
Merge pull request #285 from vincentbernat/feature/linux-caps

priv: drop most privileges in monitor, only keep CAP_NET_RAW/ADMIN

5 years agotests: request CAP_DAC_OVERRIDE feature/linux-caps 285/head
Vincent Bernat [Sat, 16 Jun 2018 15:59:32 +0000 (17:59 +0200)] 
tests: request CAP_DAC_OVERRIDE

CAP_FOWNER is for being able to use chown/chmod. The permission we
need to ignore permissions is CAP_DAC_OVERRIDE. It is quite a large
permission, unfortunately.

5 years agotests: add SNMP-related tests
Vincent Bernat [Sat, 16 Jun 2018 15:30:06 +0000 (17:30 +0200)] 
tests: add SNMP-related tests

5 years agopriv: always request CAP_FOWNER
Vincent Bernat [Sat, 16 Jun 2018 15:53:33 +0000 (17:53 +0200)] 
priv: always request CAP_FOWNER

While setting ifalias has some additional checks to ensure we can do
that with CAP_NET_ADMIN, we also need CAP_FOWNER to pass the sysfs
owner check. And we have to have both as the other test still needs to
pass.

5 years agopriv: request CAP_FOWNER only if SNMP is enabled on command-line
Vincent Bernat [Sat, 16 Jun 2018 15:47:14 +0000 (17:47 +0200)] 
priv: request CAP_FOWNER only if SNMP is enabled on command-line

5 years agopriv: don't use %m in format string
Vincent Bernat [Sat, 16 Jun 2018 15:36:16 +0000 (17:36 +0200)] 
priv: don't use %m in format string

This is not portable, use %s, even if we are on Linux.

5 years agodebian/redhat: build-depends on libcap-dev/libcap-devel
Vincent Bernat [Fri, 15 Jun 2018 08:57:05 +0000 (10:57 +0200)] 
debian/redhat: build-depends on libcap-dev/libcap-devel

5 years agotests: don't start lldpd twice in a namespace
Vincent Bernat [Fri, 15 Jun 2018 06:11:25 +0000 (08:11 +0200)] 
tests: don't start lldpd twice in a namespace

5 years agotests: add a test for interface alias handling
Vincent Bernat [Fri, 15 Jun 2018 05:27:38 +0000 (07:27 +0200)] 
tests: add a test for interface alias handling

5 years agobuild: also test directly for -lcap when no pkg-config
Vincent Bernat [Wed, 13 Jun 2018 06:06:45 +0000 (08:06 +0200)] 
build: also test directly for -lcap when no pkg-config

5 years agopriv: when SNMP AgentX socket is filesystem-based, acquire CAP_FOWNER
Vincent Bernat [Wed, 13 Jun 2018 05:56:48 +0000 (07:56 +0200)] 
priv: when SNMP AgentX socket is filesystem-based, acquire CAP_FOWNER

5 years agopriv: drop most privileges in monitor, only keep CAP_NET_RAW/ADMIN
Vincent Bernat [Tue, 12 Jun 2018 21:17:21 +0000 (23:17 +0200)] 
priv: drop most privileges in monitor, only keep CAP_NET_RAW/ADMIN

On Linux, we mostly rely on CAP_NET_RAW. Only keep that one. However,
we also write to ifalias, which needs CAP_NET_ADMIN. We could let user
choose at runtime if they want to grant this capability or not.
Currently, a user can turn it on/off at any time.

Access to SNMP socket may also be problematic. We need some solid
solution about that before merging.

Is it safe to use the same UID for the monitored and the unprivileged
process? Signals are mostly harmless. As for ptrace, since the
monitored process as more capabilities, this will not be allowed by
Linux.

5 years agoMerge pull request #286 from GustavWi/master
Vincent Bernat [Wed, 13 Jun 2018 11:27:29 +0000 (13:27 +0200)] 
Merge pull request #286 from GustavWi/master

Read all notifications in lldpctl_recv.

5 years agoRead all notifications in lldpctl_recv. 286/head
Gustav Wiklander [Wed, 13 Jun 2018 09:35:15 +0000 (11:35 +0200)] 
Read all notifications in lldpctl_recv.

Can otherwise lead to unbounded growth in input_buffer if
lldp devices send notifications simultaneously thus
a socket callback contains multiple notifications
and only the first one is cleared. This leads to continous
growth of the input buffer and will crash the system.

Change-Id: I3331de2d8201f574a91b8e8f2d51a98010f7a7ab

5 years agolib: document the inability to do anything else when processing notifications
Vincent Bernat [Wed, 13 Jun 2018 06:29:46 +0000 (08:29 +0200)] 
lib: document the inability to do anything else when processing notifications

The code doesn't handle notifications with regular operations on the
same connection. Tell the user to open a another connection.

5 years agobuild: use lowercase for PKG_CHECK_MODULES
Vincent Bernat [Wed, 13 Jun 2018 05:41:00 +0000 (07:41 +0200)] 
build: use lowercase for PKG_CHECK_MODULES

Otherwise, this is a bit ugly.

6 years agodoc: add more instructions for Android
Vincent Bernat [Wed, 11 Apr 2018 06:45:24 +0000 (08:45 +0200)] 
doc: add more instructions for Android

6 years agobuild: no fatal errors when compiling with embedded libevent on macOS 1.0.1
Vincent Bernat [Mon, 9 Apr 2018 12:45:03 +0000 (14:45 +0200)] 
build: no fatal errors when compiling with embedded libevent on macOS

6 years agobuild: don't override CFLAGS when building libevent
Vincent Bernat [Mon, 9 Apr 2018 12:19:03 +0000 (14:19 +0200)] 
build: don't override CFLAGS when building libevent

We still want for example the minimal macOS version.

6 years agorelease: 1.0.1
Vincent Bernat [Mon, 9 Apr 2018 11:44:08 +0000 (13:44 +0200)] 
release: 1.0.1

6 years agodaemon: use "mkdir -p" instead of "mkdir"
Vincent Bernat [Mon, 9 Apr 2018 11:39:46 +0000 (13:39 +0200)] 
daemon: use "mkdir -p" instead of "mkdir"

We don't use `RuntimeDirectory=` because it needs to be relative to
`/run`. As PRIVSEP_CHROOT can be any value, this is not possible.

6 years agoRevert "daemon: use RuntimeDirectory= instead of mkdir to create chroot"
Vincent Bernat [Mon, 9 Apr 2018 11:39:18 +0000 (13:39 +0200)] 
Revert "daemon: use RuntimeDirectory= instead of mkdir to create chroot"

This reverts commit 478ec07ddc343e68d1691d551dba00919fa79d70.

6 years agodaemon: use RuntimeDirectory= instead of mkdir to create chroot
Vincent Bernat [Mon, 9 Apr 2018 10:53:33 +0000 (12:53 +0200)] 
daemon: use RuntimeDirectory= instead of mkdir to create chroot

This fix a bug introduced in b0ca585f627aa where "mkdir -p" was
replaced by "mkdir" which would fail on restart due to the directory
already existing.

6 years agobuild: try another way to circumvent macOS warning in libevent
Vincent Bernat [Mon, 9 Apr 2018 05:26:18 +0000 (07:26 +0200)] 
build: try another way to circumvent macOS warning in libevent

6 years agobuild: also build with embedded libevent in OSX
Vincent Bernat [Sun, 8 Apr 2018 18:34:35 +0000 (20:34 +0200)] 
build: also build with embedded libevent in OSX

6 years agobuild: don't be picky about deprecated stuff in libevent
Vincent Bernat [Sun, 8 Apr 2018 17:26:43 +0000 (19:26 +0200)] 
build: don't be picky about deprecated stuff in libevent

This should fix:

    evutil_rand.c:177:2: error: 'arc4random_addrandom' is deprecated: first deprecated in macOS 10.12 - use arc4random_stir [-Werror,-Wdeprecated-declarations]

6 years agorelease: prepare 1.0.0 1.0.0
Vincent Bernat [Sun, 8 Apr 2018 16:30:57 +0000 (18:30 +0200)] 
release: prepare 1.0.0

6 years agoMerge pull request #274 from vincentbernat/feature/keep-some-ports
Vincent Bernat [Sun, 18 Mar 2018 11:08:13 +0000 (12:08 +0100)] 
Merge pull request #274 from vincentbernat/feature/keep-some-ports

lldpd: add an option to keep some specified ports

6 years agoMerge pull request #272 from westermo/fix-snmp-remTablesLastChange
Vincent Bernat [Sat, 17 Mar 2018 16:06:38 +0000 (17:06 +0100)] 
Merge pull request #272 from westermo/fix-snmp-remTablesLastChange

handle lldpStatsRemTablesLastChangeTime correctly when items are removed