| Age | Commit message (Collapse) | Author |
|---|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Deregister netlink event socket in case of error different
from ENOBUFS or failure to set netlink rx buffer size
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Start with a netlink receive buffersize of 130k for the netlink event socket;
double the netlink receive buffer size in case an ENOBUFS is reported.
Also align function naming and its parameters.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Rework the IPv6 relay logic to make it more robust by making
use of libnltiny to process netlink messages.
At the same time reimplement the IPv6 relay finite state machine
to make the relay logic less error prone
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Interface defaults are whiped when clean interface is executed
resulting into loss of default behavior for the ndproxy_routing
and ra_management uci parameters
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Make adding and removing of routes via apply_lease dependant
on the assignment bound state as it makes no sense to delete
routes for an assignment which is not bound.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Use libnl-tiny to construct and process netlink messages when
manipulating IPv6 routes and fetching IPv6 addresses.
This fixes lingering netlink error messages on the netlink socket
in case route deletion failed causing fetching of IPv6 addresses
to be aborted and odhcpd faultly assuming no IPv6 addresses being
present on the interface.
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Rework the IPv6 address dump logic to make it more robust
and generate syslog traces in case of error situations
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Add extra syslog tracing for trouble shooting
and make some syslog traces more self explaining
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
This reverts commit 70fbf936839e02d526929be06ecba3059f78be3c.
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Rework prio of syslog messages allowing to filter out the relevant messages
by setting the correct log level
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
ia_reconf is only set but nowhere used; so let's remove it ...
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
In case a netlink route event is received for a default route
netlink data attributes were faulty processed as neigbor netlink
data. Fix this by not processing any netlink attribute for
any route event.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Function append_reply filters out addresses having invalid prefix length
when constructing prefixes being delegated to a client.
Now also filter out addresses with an invalid prefix length when writing
the delegated prefixes in the statefile and during logging of the delegated
prefixes.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Fix assignment dereference by dhcpv6_log in case
the assignment is freed when nothing has been
assigned.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
DHCPv6 client
Fix displaying IPv6 addresses which were not assigned to a DHCPv6 client
Clean up the code by using macro ADDR_ENTRY_VALID_IA_ADDR on different
places to align the checks.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
The odhcpd documentation currently implies that "option ignore 1" in a section
of type "dhcp" will disable any services on the referenced interface while the
code actually ignores the setting when initializing referenced interfaces.
This commonly leads to situations where users think that "option disabled"
will prevent any server activity on a given interface while in reality odhcpd
effectively acts as rogue DHCPv6 server in its default configuration.
To fully inhibit any activity, one has to specify both "option ignore 1" as
well as "option ra disabled", "option ndp disabled" and
"option dhcpv6 disabled" which is highly counter intuitive at best.
A previous commit (5f425ed Respect interface "ignore" settings as documented.)
attempted to address the problem by observing the value of i->ignore instead
of unconditionally enabling all protocols but broke the ability to use DHCPv6
relay mode while having DHCPv4 disabled.
This patch changes ...
- the enabling of services on an interface to use the value of the ignore
option as default choice instead of hardcoding "true"
- the options "ra", "ndp", "dhcpv4" and "dhcpv6" to take precedence over the
default "ignore" state to allow selectively enabling services on ignored
interfaces
- the README to clarify the fact that "ra", "ndp", "dhcpv4" and "dhcpv6" can
now override "option ignore"
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Acked-by: Karl Palsson <karlp@etactica.com>
Tested-by: Stijn Tintel <stijn@linux-ipv6.be>
Tested-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Fix static assignment check failure which was
introduced in commit 561890e
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Update valid_until lifetime only for non static DHCPv6 leases otherwise
the lifetime of a static lease is not infinite anymore
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Statefull IPv6 addresses are assigned to DHCPv6 clients from a
local prefix based on criteria which need to be fulfilled.
One of those criteria requires the IPv6 address still having
preferred lifetime which was not checked when constructing
the contents of the statefile.
Another criteria requires the prefix length not being smaller
than or equal to 64 when ra_management equals 0 or 1 which was
neither correctly checked when constructing the statefile.
This can lead to invalid hostname/IPv6 entries in the statefile
and thus possible connectivity issues when using DNS resolving.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Function free_dhcpv6_assignment already removes the assignment in
the list so there's no need to call list_del in reconf_timer
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Create statefile dir if the specified dir does not yet exist
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Further align coding style
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Fix hostname not being reported as "-" in the DHCP statefile in case
no hostname is reported by the DHCPv4 client
(https://:wqgithub.com/openwrt/odhcpd/issues/94)
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
In case infinite leasetime is assigned to a non static DHCPv4/v6 lease
override the infinite lifetime of the lease when either a DHCPv4/v6
decline or release is received.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
This reverts commit 5f425ed1e5f92397e27ec4f44820b6ef677b8134.
Both dnsmasq and odhcpd use the ignore option. This is fine for normal
operation, but it breaks DHCPv6 relay when dnsmasq is disabled on the
wan interface.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
|
|
It does not make sense to scan the entire routing table if the decision
will be overridden afterwards.
This makes it possible to use odhcpd on a router with full IPv6 BGP
table, by setting ra_default to 2. Without this change, odhcpd is
constantly busy reading /proc/net/ipv6_route, and doesn't respond to
router solicitations in time.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
|
|
Set bound flag for DHCPv4 and DHCPv6 assignments when the IPv6/IPv4 address
is leased to a client.
This will prevent the printing of leases and hostname/IPv4/IPv6 address
combinations in the state file for for which the IPv4/6 address has not
been assigned.
Also this will fix the printing of assignments which have been declined
by the clients
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
In case multiple dhcp uci sections are configured and one section contains
RA config while the other section does not; the RA interface config of the
first section will be overwritten by the second section.
Fix this by setting the default values for ra_management and ndp_proxy_routing
uci parameters when the interface is created and by removing the overwrite check
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
The valid_until parameter which is set to 0 for static DHCPv6
assignments was overwritten depending on the received DHCPv6 message
which turned the assignment into having a finite lifetime and thus not
static anymore.
Fix this by checking if the valid_until parameter holds infinite
lifetime before updating it.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Fix assignment of static DHCPv6 leases which was broken
by commit 6c3bc0141dacc88992583d6f2cdc8a360df7422d.
At the same time fix lifetime check of existing assignments
having a finite lifetime
Fixes FS#309
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
|
|
Add a CMake FIND_PATH and INCLUDE_DIRECTORIES searching for
libubox/uloop.h. Some external toolchains which do not include standard
locations would fail to find the header otherwise.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
|
