- New upstream major release with tons of new features and LTS (see: https://www.haproxy.com/blog/announcing-haproxy-2-4/)
- Update haproxy download URL and hash
- Activate promtheus exporter support the new way (using USE_PROMEX=1)
- Cleaned up haproxy-specific CFLAGS
- Changed the halog build to make use of the new Makefile target (admin/halog/halog)
Signed-off-by: Christian Lachner <gladiac@gmail.com>
Rrsync is a perl script that is supplied as an extra with the rsync program.
It must be used in conjunction with openssh-server or openssh-server-pam
as it requires ~/.ssh/authorized_keys which is not supported by dropbear.
Rrsync allows selective access to subdirectories in either read-only, write-only or read-write,
depending on settings in authorized_keys. This allows for safe, restrictive access.
It's particularly useful for automated backup purposes.
An example usage would be this entry:
command="/usr/bin/rrsync -ro /home" <public key here>
This would allow a system connecting with this public key to be able to rsync FROM the
/home directory tree only. It could not write to this directory, nor read from any other directory.
Signed-off-by: Matt Reeve <matt@mreeve.com>
Recreate symbolic link if it's missing after a sysupgrade with a private and public key present in /etc/atlas/
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
* Create working directory when it is not present. Apparently
some recent change made adguardhome fail to start when working
directory is missing.
* Full changelog available at:
* https://github.com/AdguardTeam/AdGuardHome/releases/tag/v0.106.1
Signed-off-by: Dobroslaw Kijowski <dobo90@gmail.com>
* fix pid file processing of the background monitor plus child
processes (bug reported in the forum)
* made the enabled/disabled switch of the background monitor functional
Signed-off-by: Dirk Brenken <dev@brenken.org>
Samplicator receives UDP datagrams on a given port and resends those
datagrams to a specified set of receivers.
Use Cases:
- replicate Flow Samples to multiple receivers
- use with conntrackd to synchronize via unicast to multiple targets
Signed-off-by: Nick Hainke <vincent@systemli.org>
In the procd refactor, support for interfaces with no tracking IPs was
inadvertentiy removed. This commit restores the previous behavior
Signed-off-by: Aaron Goodman <aaronjg@stanford.edu>
Fixes the following security issues:
* CVE-2021-25215 - named crashed when a DNAME record placed in the ANSWER
section during DNAME chasing turned out to be the final
answer to a client query.
* CVE-2021-25214 - Insufficient IXFR checks could result in named serving a
zone without an SOA record at the apex, leading to a
RUNTIME_CHECK assertion failure when the zone was
subsequently refreshed. This has been fixed by adding an
owner name check for all SOA records which are included
in a zone transfer.
Signed-off-by: Noah Meyerhans <frodo@morgul.net>
Using `$(INSTALL_CONF)` will cause the program has no access to
configurations file when someone enabled the selinux support.
Signed-off-by: Tianling Shen <cnsztl@immortalwrt.org>
Xray now is no longer planning to keep compatibility with original
v2ray. Remove PROVIDES before it is totally broken.
Signed-off-by: Tianling Shen <cnsztl@immortalwrt.org>
From mosquitto 2.x, port became optional and deprecated in the config,
and it was recommended that listeners be used instead. Drop the hard
requirement in our config conversion script.
Reported in: https://github.com/openwrt/packages/issues/15506
Signed-off-by: <karlp@etactica.com>
Maintainer: @neheb / @BKPepe / @zhanhb
Compile tested: ipq806x, generic, netgear_r7800, master
Run tested: ipq806x, generic, netgear_r7800, openwrt-19.07
Description:
Squid now only support HTTPS proxy in TCP tunnel mode (e.g. `ssl_bump splice all`):
https_port 3128 ssl-bump tls-cert=/etc/squid/squid.pem generate-host-certificates=on
ssl_bump splice all
In order to operate in SSL Bump mode, we need to compile with `--enable-ssl-crtd` for following configuration:
https_port 3128 ssl-bump tls-cert=/etc/squid/squid.pem generate-host-certificates=on
sslcrtd_program /usr/lib/squid/security_file_certgen -s /car/cache/squid/ssl_db -M 4MB
ssl_bump stare all
ssl_bump bump all
This PR switch the `SQUID_enable-ssl-crtd` into `default y`, therefore default enable SSL Bump mode.
Signed-off-by: Wong Hoi Sing Edison <hswong3i@pantarei-design.com>
Staging certificates have the advantage that their retry limits are loose.
Therefore they can be obtained quickly when automatic retries are used.
Unfortunately they can not be used for deployments because their CA is not
accepted by clients. Production certificates do not have this limitation, but
their retry limits are strict. For production certificates, automatic retries
can only be performed a few times per hour. This makes automatic obtainment of
certificates tenacious.
With use_auto_staging=1, the advantages of the two certificate types are
combined. Uacme will first obtain a staging certificate. When the staging
certificate is successfully obtained, uacme will switch and obtain a production
certificate. Since the staging certificate has already been successfully
obtained, we can ensure that the production certificate is successfully
obtained in the first attempt. This means that "retries" are performed on the
staging certificate and the production certificate is obtained in the first
attempt.
In summary, this feature enables fast obtaining of production certificates when
automatic retries are used.
By default, this feature is set to use_auto_staging=0, which means that
uacme will behave as before by default.
Signed-off-by: Leonardo Mörlein <git@irrelefant.net>
With this commit, issue_cert() can be called multiple times alternating
between staging and production certificates within a script.
Before this commit, the production state dir was stored in $STATE_DIR.
But in the case of $use_staging=1, this variable was overwritten in
issue_cert() with $STAGING_STATE_DIR. This made it impossible to call
issue_cert() with $use_staging=0 afterwards. Now the production state
dir is stored in $PRODUCTION_STATE_DIR. This way it is not overridden
anymore and issue_cert() can be called multiple times alternating with
production and staging.
Signed-off-by: Leonardo Mörlein <git@irrelefant.net>
The get_bool() functionality was already merged to lib/functions.sh, so
it is redundant in the init script. Remove it.
Signed-off-by: Oldřich Jedlička <oldium.pro@gmail.com>
- ignore Content-Length from backend if 101 Switching Protocols
- close HTTP/2 connection after bad password
- skip cert chain build for self-issued certs
- meson zstd fix
- ls-hpack upstream update
- discard some HTTP/2 DATA frames received after response
Signed-off-by: Glenn Strauss <gstrauss@gluelogic.com>
- Exit start if a probe_key is not present
- Add create_key command to generate a private_key based on the provided username in the atlas config.
- Add registration instruction in /etc/atlas
- Rework script to save probe_key on sysupgrade (the key are now adviced to be placed in the /etc/atlas dir and a link is used to make them accessible in the atlas-sw-scripts etc dir)
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
* lots of fixes for many subsystems
* new messenger group chat service
* 'abd' temporarily removed due to upstream issue
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Fixes two related security vulnerabilities (CVE-2020-15078) which
under very specific circumstances allow tricking a server using delayed
authentication (plugin or management) into returning a PUSH_REPLY before
the AUTH_FAILED message, which can possibly be used to gather
information about a VPN setup. In combination with "--auth-gen-token" or
a user-specific token auth solution it can be possible to get access to
a VPN with an otherwise-invalid account.
OpenVPN 2.5.2 also includes other bug fixes and improvements.
Add CI build test script.
Signed-off-by: Magnus Kroken <mkroken@gmail.com>
* add a "whitelist only" mode, this option allows to restrict Internet
access from/to a small number of secure websites/IPs, and block access
from/to the rest of the Internet.
Signed-off-by: Dirk Brenken <dev@brenken.org>
This commit is largely based on the work from Daniel Dickinson in
PR #2096 which was never merged. I tweaked it in a number of ways.
All bugs with this package are mine, not his.
Signed-off-by: Aaron Curley <accwebs@gmail.com>
* support the RPZ trigger 'RPZ-CLIENT-IP' to always allow/block certain
clients based on their IP (currently only supported by bind!)
* avoid promiscuous mode in tcpdump setup for adblock reporting
* speed up dns report preparation
* support dns report mailing (/etc/init.d/adblock report mail)
* fix bind autodetection
* update LuCI-frontend (separate PR)
* update readme
Signed-off-by: Dirk Brenken <dev@brenken.org>
NLS means Native Language Support and when you have it enabled (it is
not default), clamav can not be compiled as it shows following error:
Package clamav is missing dependencies for the following libraries:
libiconv.so.2
Also, it is required that package libiconv-full is compiled first/before
than clamav and then try to compile clamav.
Signed-off-by: Josef Schlehofer <pepe.schlehofer@gmail.com>
/etc/profile.d/50-openvpn-easy-rsa.sh was not listed as configfile
and changes were lost during upgrades.
Signed-off-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
libseccomp can't be built on ARC, so we must disable the option here as
well. A different fix was first proposed by @zxlhhyccc in #15377.
Fixes: #15313
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
Add patch fixing compilation without deprecated OpenSSL APIs.
Fix installation. This never worked as the section was misnamed.
Updated tool names.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
chacha20policy1305 is also an AEAD cipher, and hence does not
permit a hash algorithm.
Fixes issue #15397.
Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
This patch prevents multiple cron jobs from being created to run the
safe-search-maintenance script.
To reproduce this bug, perform the following:
- Install safe-search
- Perform an OpenWRT firmware upgrade (choose to preserve user settings)
- Install safe-search again
Signed-off-by: Gregory L. Dietsche <gregory.dietsche@cuw.edu>
The strongswan-libnttfft package should not select the strongswan
package, but should depend on it instead. Otherwise a circular
dependency is created.
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
Rework the bonding.sh protocol handler to accept slave interface names
encoded in uci list notation. Also replace ifconfig up/down with ip
link calls while we're at it.
Fixes: #11455
Fixes: https://github.com/openwrt/luci/issues/4473
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
MacOS ignores Bonjour services for which TXT records are not returned. This changes forces umdns service to return a TXT record (`daemon=ksmbd`) for the ksmbd service. The exact content is unimportant and to the best of my knowledge nothing reads the `daemon` tag.
Symptoms of the problem (which are also debugging steps):
* Finder refuses to open the OpenWRT "computer" in the Network list.
* Discovery.app (Bonjour Browser) lists the _ssh._tcp service, but the submenu for it doesn't unfold and no address is shown.
* `dns-sd -L OpenWrt _smb._tcp` doesn't return any address.
Signed-off-by: Kirill Nikolaev <cyril7@gmail.com>
This is a security fix, affecting 2.0.0 through to 2.0.9. Mosquitto instances
could be remotely DoS'd by authenticated clients.
Release notes at: https://github.com/eclipse/mosquitto/blob/v2.0.10/ChangeLog.txt
CVE number has not yet been assigned.
Signed-off-by: Karl Palsson <karlp@etactica.com>
Prior to this commit, the acme service attempted to obtain certificates
once and then terminated, regardless of whether the certificate could be
obtained or not. This commit introduces a new uci option "retries" to
the "certificate" section. If this option is set to N, the acme service
will attempt to obtain the certificate up to N times before terminating.
There is a waiting pause between the retries to comply with the rate
limits of Let'sEncrypt.
The waiting pause is:
- 2 minutes for staging certificates
- 24 minutes for production certificates
The current "Failed Validation" rate limits of Let'sEncrypt are:
- staging: 60 per hour -> 1 failure every 1 minute in avg.
- production: 5 per hour -> 1 failure every 12 minutes in avg.
This means that we are within rate limits by a factor of two.
By default the option "retries" is set to "1", which means that acme
behaves as before by default. If the variable is set to "0", infinite
retries are performed.
This feature is helpful, when you already want to initiate the
certificate request, but you are still waiting for your dns server to be
configured, your network to appear or other conditions.
Signed-off-by: Leonardo Mörlein <git@irrelefant.net>
Before this commit, issue_cert always returned 1 no matter if uacme
returned 1, 2, 3, ... With this commit, the return code of the uacme
binary is propagated. Therefore the caller of issue_cert can
differentiate between "no renew necessary" and "an error occurred".
Signed-off-by: Leonardo Mörlein <me@irrelefant.net>
With this commit, the run-acme script can be included into other scripts
by setting INLCUDE_ONLY=1.
Signed-off-by: Leonardo Mörlein <me@irrelefant.net>
Derived from the ipsec initd script, with the following changes:
(1) various code improvements, corrections (get rid of left/right
updown scripts, since there's only one), etc;
(2) add reauth and fragmentation parameters;
(3) add x.509 certificate-based authentication;
and other minor changes.
Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
netifyd supports a '-F' filter option in 'bpf' notation to filter
packets from its consideration.
Add support for a uci 'filter' option. eg. filter to exclude SSDP
multicasts from a particularly noisy device:
option filter 'not (udp and dst 239.255.255.250 and dst port 1900 and src 192.168.1.5)'
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Even it's only cosmetic and should not affect the function of regular system,
fix the name of the IPKG_INSTROOT variable.
Typo was added long ago with 8400c9a6ec.
Signed-off-by: Sven Roederer <devel-sven@geroedel.de>
Since v1.4.1, Xray has introduced a new feature to transfer data via
browsers, which can disguise itself as a normal browser to cheat
network censorship.
For more details, see https://github.com/XTLS/Xray-core/pull/421.
Signed-off-by: Tianling Shen <cnsztl@immortalwrt.org>
If you shutdown ipsec service, and it doesn't clean up
/var/ipsec/ipsec.conf, then when you start swanctl service it
might see an incompatible file on startup. Remedy is to
remove unneeded files when shutting down the service. They
can always be regenerated when the service starts again.
Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
This commit adds a number of fixes to the OpenVPN up/down hotplug command
wrapper which currently fails to actually invoke user defined up and down
commands for uci configurations not using external native configurations.
- Use the `--setenv` to pass the user configured `up` and `down` commands
as `user_up` and `user_down` environment variables respectively
- Instead of attempting to scrape the `up` and `down` settings from the
(possibly generated) native OpenVPN configuration in
`/etc/hotplug.d/openvpn/01-user`, read them from the respective
environment variables instead
- Fix parsing of native configuration values in `get_openvpn_option()`;
first try to parse a given setting as single quoted value, then as
double quoted and finally as non-quoted, potentially white-space
escaped one. This ensures that `up '/bin/foo'` is interpreted as
`/bin/foo` and not `'/bin/foo'`
Ref: https://forum.openwrt.org/t/openvpn-up-down-configuration-ignored/91126
Supersedes: #15121, #15284
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
The `tmate` tool is a fork of `tmux` which allows remote access to a
device without setting up any port forwarding. This commits adds the
backend server which handles connections.
Signed-off-by: Paul Spooren <mail@aparcar.org>
These config files are only used by the ipsec interface to charon,
and shouldn't be part of the base package.
Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
* rework the central iptables function to significantly
reduce the code complexity and the overall number of iptables calls
* check early and only once in the chain for ctstate NEW and
return otherwise (thanks @ldir-EDB0)
* made the whitelist ordering within the chain more flexible
Signed-off-by: Dirk Brenken <dev@brenken.org>
faster to compile.
A small selection of packages was tested going from:
Executed in 696.30 secs fish external
usr time 82.98 mins 395.00 micros 82.98 mins
sys time 9.02 mins 0.00 micros 9.02 mins
to:
Executed in 592.20 secs fish external
usr time 84.84 mins 361.00 micros 84.84 mins
sys time 8.85 mins 57.00 micros 8.85 mins
Tested by running make -j 12 and wiping staging/build_dir/target_x
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Having scripts diddle user written config files seems potentially
dangerous. Plus there's really no downside to including some
empty files. Best to just make the includes be permanent.
Additional feature suggested by Luiz: if a -opkg version of the
config file was created unnecessarily, remove it as part of the
upgrade process since changes won't be happening to that file
as an artifact of the service starting. The include lines are
now permanent, which means that (1) additional configuration
synthesized by UCI won't be anywhere that opkg (or sysupgrade,
for that matter) cares about since it won't be persistent, and
(2) if changes are being made, then they're being done by a
person with an editor and they really should be distinguished.
Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
It seems the command name output from netstat can be truncated in weird
ways, so let's get the binary name from /proc instead and use that for
matching which listener we have.
Fixes#15071.
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
* fix another IPv4/IPv6 related iptables chain creation problem
* fix counter during ipset creation
* fix regex for debug counters
* fix ipset housekeeping for local sources
Signed-off-by: Dirk Brenken <dev@brenken.org>
Reorganize Makefile for consistency between packages.
Switch to AUTORELEASE for simplicity.
Switch to building with Ninja for faster compilation.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
* add a restrictive "jail mode only" variant, just point your
jail directory to your primary dns directory
* update readme
Signed-off-by: Dirk Brenken <dev@brenken.org>
