These are recommended practices by REC-22 and REC-24 of RFC6092:
"Recommended Simple Security Capabilities in Customer Premises Equipment
(CPE) for Providing Residential IPv6 Internet Service"
Fixes FS#640
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Packets which are merely forwarded by the router and which are neither
involved in any DNAT/SNAT nor originate locally, are considered INVALID
from a conntrack point of view, causing them to get dropped in the
zone_*_dest_ACCEPT chains, since those only allow stream with state NEW
or UNTRACKED.
Remove the ctstate restriction on dest accept chains to properly pass-
through unrelated 3rd party traffic.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Update to current HEAD in order to drop automatic generation of per-zone
NOTRACK rules.
The NOTRACK rules used to provide a little performance improvement but the
later introduction of the netfilter conntrack cache made those rules largely
unnecessary. Additionally, those rules caused various issues which broke
stateful firewalling in some scenarios.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Update to latest Git head in order to import several fixes and enhancements.
- Disable drop invalid by default (FS#73, FS#154)
Instead of dropping packets with conntrack state INVALID, only allow streams
with explicit NEW or UNTRACKED conntrack state.
This change gives user defined rules the chance to accept traffic like ICMPv6
multicast which would be filtered away by the very early ctstate INVALID drop
rule otherwise.
The old behaviour can be restored by explicitely setting "drop_invalid" to 1
in the global firewall config section.
- Fix re-initialization of loadable iptables extensions on musl (FS#31)
Since musl does not implement actual dlclose() semantics, it is impossible to
re-run initializers on subsequent dlopen() calls.
The firewall3 executable now intercepts the extension registration calls
instead in order to be able to re-call them when needed.
This also allowed us to switch to libxtables' builtin extension loader as a
positive side-effect.
- Fix masquerade rules for multiple negated IP addresses (FS#248)
When building MASQUERADE rules for zones which specify multiple negated
addresses in masq_src or masq_dest, emit -j RETURN rules which jump out of
the masquerading chain instead of creating multiple rules with inverted "-s"
arguments.
- Tag own rules using comments
Instead of relying on the nonstandard xt_id match, use the xt_comment match
to mark own rules. Existing comments are prefixed with "!fw3: " while
uncommented rules are marked with a sole "!fw3" string.
This allows removing the xt_id match entirely in a later commit.
- Make missing ubus connection nonfatal
Technically, firewall3 is able to operate without ubus just fine as long as
the zones are declared using "option device" or "option subnet" instead of
"option network" so do not abort execution if ubus could not be connected or
of no network namespace is exported in ubus.
This allows running firewall3 on ordinary Linux systems.
- Fix conntrack requirement detection for indirectly connected zones
The current code fails to apply the conntrack requirement flag recursively to
zones, leading to stray NOTRACK rules which break conntrack based traffic
policing.
Change the implementation to iteratively reapply the conntrack fixup logic
until no more zones had been changed in order to ensure that all directly and
indirectly connected zones receive the conntrack requirement flag.
- Add support for iptables 1.6.x
Adds support for the xtables version 11 api in order to allow building
against iptables 1.6.x
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
* Change git packages to xz
* Update mirror checksums in packages where they are used
* Change a few source tarballs to xz if available upstream
* Remove unused lines in packages we're touching, requested by jow- and blogic
* We're relying more on xz-utils so add official mirror as primary source, master site as secondary.
* Add SHA256 checksums to multiple git tarball packages
Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
Update to latest HEAD in order to fix MARK rule generation for local traffic,
also fix a possible race condition during firewall start.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
* Enable drop_invalid by default to catch unnatted packets (#21738)
* Fix processing of inversions for -i, -o, -s, -d and -p flags
* Remove delegate_* chain indirection but rely on xt_id to identify own rules
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 48551
There is no RFC requirement that DHCPv6 servers must reply with a link local
address and some ISP servers in the wild appear to using addresses in the ULA
range to send DHCPv6 offers.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 47048
Our ruleset requires kernel support for conntrack state matching, therfore
depend on the require kmod. Fixes#20542.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 46990
Seems like my second try was again whitespace broken. Sorry for the noise.
Remove src_port from firewall.config to receive dhcpv6 replies. Fixes#20295.
Signed-off-by: Anselm Eberhardt <a.eberhardt@cygnusnetworks.de>
SVN-Revision: 46842
Use shared libipt{,4,6}ext.so libraries instead of statically linking
the userspace matches into the fw3 executable.
As a side effect the match initialization is extremely simplified
compared to the weak function pointer juggling performed before.
This also fixes the initialization of the multiport match.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 45764
The WAN port should at least respond to IGMP and MLD queries as
otherwise a snooping bridge/switch might drop traffic.
RFC4890 recommends to leave IGMP and MLD unfiltered as they are always
link-scoped anyways.
Signed-off-by: Linus Lüssing <linus.luessing@c0d3.blue>
SVN-Revision: 45613
Note, that licensing stuff is a nightmare: many packages does not clearly
state their licenses, and often multiple source files are simply copied
together - each with different licensing information in the file headers.
I tried hard to ensure, that the license information extracted into the OpenWRT's
makefiles fit the "spirit" of the packages, e.g. such small packages which
come without a dedicated source archive "inherites" the OpenWRT's own license
in my opinion.
However, I can not garantee that I always picked the correct information
and/or did not miss license information.
Signed-off-by: Michael Heimpold <mhei@heimpold.de>
SVN-Revision: 43155
somebody started to set a function returncode in the validation
stuff and everybody copies it, e.g.
myfunction()
{
fire_command
return $?
}
a function automatically returns with the last returncode,
so we can safely remove the command 'return $?'. reference:
http://tldp.org/LDP/abs/html/exit-status.html
"The last command executed in the function or script determines the exit status."
Signed-off-by: Bastian Bittorf <bittorf@bluebottle.com>
SVN-Revision: 42278
Utilize the new selective conntrack flushing facility to clear
out active conntrack entries referring to old IP addresses after
a firewall reload.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 42114
Properly parse and pass arbritary netmasks to iptables, this allows
specifying ranges like '::c23f:eff:fe7a:a094/::ffff:ffff:ffff:ffff' to
match the host part of an IPv6 address regardless of the currently active
IPv6 prefix.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 41760
The commit 92281eb747b56e748b7c3d754055919c23befdd4 broke fw3_ubus_addresses() so that
no addresses where returned at all, this caused fw3 to not emit NAT reflection rules
anymore.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 41556
- Do not consider bitmap storage for IPv6 family sets
- Move ipset family parameter before any additional option
- Only emit family parameter for hash sets
- Do not allow IPv6 iprange for IPv4 sets and vice versa
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 39647
The firewall3 implementation as well as the shell implementation predating it
used to process the tcp_ecnoption as boolean while it actually is an integer.
Change the code to parse tcp_ecn as integer.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 39122
- instead of writing one (or more) ACCEPT rules in the filter table
for each redirect install a global ctstate DNAT accept rule per zone
- discard rules and redirects which have invalid options set instead
of silently skipping the invalid values
SVN-Revision: 38849
* Use network.interface dump call instead of individual status calls
to reduce overall netifd lookups and invokes to 1 per fw3 process.
* Allow protocol handlers to assign a firewall zone for an interface
in the data section to allow for dynamic firewall zone assignment.
SVN-Revision: 38504
