We do not need promiscuous mode setting in host side when testing
bridging with native-sim, so just ignore errors in this case.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
The legacy bridging code prevented normal IP traffic to the
bridged Ethernet interfaces. This is not intuitive and differs
how bridging setup works in Linux. This commit changes that and
creates a separate virtual interface that is doing the actual
bridging. This enables the bridged Ethernet interfaces to work
normally and provide IP connectivity.
How this works in practice:
* User needs to enable CONFIG_NET_ETHERNET_BRIDGE
* User needs to have a device with more than one Ethernet
interface
* After booting, the net-shell or program API can be used
to add interfaces to the bridge like this.
net bridge addif 1 3 2
where the 1 is the bridge interface index and
2 and 3 are the Ethernet interface indices.
* The bridging is then finally enabled / started when the
bridge interface 1 is taken up
net iface up 1
* If bridged interfaces are removed from the bridge (minimum
of two interfaces are needed there), then the bridging is
disabled automatically. The bridge interface stays up in
this case and can be taken down manually.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
Adds missing checks for get_capabilities method not being NULL.
Fixes crash with netusb and possibly other drivers.
Signed-off-by: Maciej Panek <panekmaciej@outlook.com>
All the network related shell activities should be under
network shell so moving it to "net bridge ..." command.
Add this information to migration guide.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
Utilize a code spell-checking tool to scan for and correct spelling errors
in all files within the `subsys/net/l2` directory.
Signed-off-by: Pisit Sawangvonganan <pisit@ndrsolution.com>
In ethernet_recv(), it will call ethernet_update_length() to check for
ipv4 or ipv6 packet if it needs to get rid of the padding in the
packets. But for EAPOL packets, no need to do this, which may modify
the packet length of EAP type packet wrongly and leads to EAPoL
packets dropped in supplicant.
Signed-off-by: Maochen Wang <maochen.wang@nxp.com>
In case of ACD Probe/Announcement, all we need is to generate ARP
packet, we don't really want any cache entries to be created or searched
for. There was a bug, that a cache entry was created for the
Announcement sent, resulting in skipped ARP packet generation and
malformed packet being sent by the ACD module.
Therefore, simplify all this, by simply returning early in case of
conflict detection packets.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
The autoconf module can now reuse generic address conflict detection,
which was added for all address types.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Add support for IPv4 conflict detection, as specified in RFC 5227.
The new feature is optional and disabled by default.
Address conflict detection was implemented as a part of the IPv4
autoconf feature can be generalized to be available for all address
types.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Namespaced the generated headers with `zephyr` to prevent
potential conflict with other headers.
Introduce a temporary Kconfig `LEGACY_GENERATED_INCLUDE_PATH`
that is enabled by default. This allows the developers to
continue the use of the old include paths for the time being
until it is deprecated and eventually removed. The Kconfig will
generate a build-time warning message, similar to the
`CONFIG_TIMER_RANDOM_GENERATOR`.
Updated the includes path of in-tree sources accordingly.
Most of the changes here are scripted, check the PR for more
info.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
According to RFC3927 and RFC5227, an ARP probe target HW address should
be set to all-zeroes:
"The 'target hardware address' field is ignored and
SHOULD be set to all zeroes."
Hence, we should allow the ARP probes to have all-zeroes target HW
address as well.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
If time aware system has more than 1 gptp port, because there is only
one boolean to handle callback registering, a race condition can occur
for instance in case 2 master ports are waiting for a timestamp following
sending of sync message. More in details, callback may be unregistered
by port 1 while port 2 is waiting for timestamp event.
The issue is solved by declaring as much as boolean and struct
net_if_timestamp_cb as the number of declared ports
(=CONFIG_NET_GPTP_NUM_PORTS).
Signed-off-by: Jean-Nicolas Graux <jean-nicolas.graux@st.com>
Setting a ration of 0 breaks the computation of scaled_rate_offset
in gptp_md_follow_up_prepare().
According to IEEE802.1AS 10.2.2.3.9, when sent by clock master
entity, rate_ratio shall be equal to gmRateRatio. But since
in gptp_md_follow_up_prepare(), in case we are the grand master
clock, we take the assumption that clock source is the local clock,
we may set this ratio to 1.0.
Signed-off-by: Jean-Nicolas Graux <jean-nicolas.graux@st.com>
Without this patch, in case time aware system is grand master capable
but is not the grand master clock and behave as a bridge so that it
shall ONLY relay SYNC and FUP messages from slave port to master port(s),
SYNC and FUP are mistakenly generated by ClockMasterSyncSend state
machine.
This patch only call ClockMasterSyncSend state machine in case a Grand
Master clock is present and is this time aware system.
This check is not described by IEEE802.1AS. Instead, according to
10.2.9.3, the SiteSyncSync state machine shall not take into account
information from ClockMasterSyncSend in case this time aware system is
not grand-master capable. Current implementation of ClockMasterSyncSend
state machine sends sync indication to the PortSync entities, instead of
sending it to the SiteSyncSync entity. And the SiteSyncSync state machine
does not make any check.
Signed-off-by: Jean-Nicolas Graux <jean-nicolas.graux@st.com>
Commit 413eceeddb ("net: gptp: fix follow_up message correction_field")
breaks computation of follow up correction field in case time aware system
is not the grand master clock but behave as a transparent clock which relay
sync and follow-up messages from one slave port to one or several master
port(s). In this case, according to IEEE802.AS 11.1.3 and 11.2.14.2.3,
md sync send state machine shall compute the sum of link propagation delay
and residence time, expressed in grand master time base. Then this quantity
shall be added to last received fup correction field to build value of
correction field. This patch checks whether time aware system is the grand
master clock or not, and computes value correction field accordingly.
Signed-off-by: Jean-Nicolas Graux <jean-nicolas.graux@st.com>
Add support for gratuitous ARP transmission by Zephyr network stack.
This allows to prematurely fill the peer ARP table, so there's no need
to send an explicit request when peer needs to send an actual packet.
The gratuitous ARP is send when the network interface is brought up
(joins the network) or a new IP address is added. The gratuitous ARP
request is also sent periodically, with a configurable interval time.
The gratuitous ARP should also be sent whenever MAC address of the
interface is changed, but as Zephyr only allows to do this when
interface is down, this is already covered by the first case (interface
brought up).
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
We were missing a helper function that can be used to check
whether the given function is the virtual VLAN interface.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
The network interface parameter for net_eth_get_vlan_tag() should
be the VLAN interface so use the search loop properly.
Earlier the main interface could be checked.
Add also test cases for this so that we can catch that the func
works properly.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
eth_stats.h is using ethernet api but not including the header,
it must be working by luck in files that include it after
something else that includes ethernet.h, fix by just including it.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
The gPTP is not suppose to be run on top of VLAN and the
earlier support was just for testing purposes. Remove VLAN
support now after the VLAN overhaul.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
Print also network interface index together with the pointer value
in order to get more useful information what is going on in the system.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
Re-implement the VLAN support inside the network stack.
All the user facing APIs stay as is but internally the VLANs
are implemented using the L2 virtual interfaces.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
During stress test with WiFi connect, disconnect, ping and throughput
traffic, ARP table updating failed issue may occur.
In arp_prepare(), if packet allocate failed, should add the arp
entry back to arp_free_entries, to avoid this entry is leak forever.
Signed-off-by: Maochen Wang <maochen.wang@nxp.com>
Convert various networking subsystem files to use ARRAY_FOR_EACH
macro to make the looping more robust.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
The netmask should be tied to the IPv4 address instead of being
global for the network interface.
If there is only one IPv4 address specified to the network interface,
nothing changes from user point of view. But if there are more than
one IPv4 address / network interface, the netmask must be specified
to each address separately.
This means that net_if_ipv4_get_netmask() and net_if_ipv4_set_netmask()
functions should not be used as they only work reliably if there is
only one IPv4 address in the network interface.
The new net_if_ipv4_get_netmask_by_addr() and
net_if_ipv4_set_netmask_by_addr() functions should be used as they make
sure that the netmask is tied to correct IPv4 address in the network
interface.
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>
Fix the dead lock between rx_q and tx_q thread.
When tx_q thread prepares to send ARP packet, it might get
the net_if_tx_lock in net_if_tx(), then in net_arp_prepare(),
it will try to get the arp_mutex.
At the same time, if the rx_q thread receives an ARP reply
packet, in arp_update(), it will get the arp_mutex first,
and flush the packets in arp pending_queue and try to get
the net_if_tx_lock. Then the dead lock occurs, two threads
stuck and all the packcets can't be freed.
In arp_update(), taking the net_if_tx_lock first then taking
the arp_mutex can fix this issue.
Signed-off-by: Maochen Wang <maochen.wang@nxp.com>
As defined in IEEE802.1AS-2020 ch. 10.2.12.2.1, the port identity
of an MDSyncSend structure sent from a port shall be set to the
port identity of the sending port according to and ch. 8.5.2.
This commit replaces the port identity before forwarding a sync.
Fixes#68385
Signed-off-by: Manuel Schappacher <manuel.schappacher@hs-offenburg.de>
This set of changes brings in raw packet tx injection mode
feature to ethernet and removes it from being a Wi-Fi only feature.
It was earlier envisaged as Wi-Fi net management feature only.
Signed-off-by: Vivekananda Uppunda <vivekananda.uppunda@nordicsemi.no>
So that it's possible to register ARP entries manually. Needed for DHCP
server implementation, which in unicast mode needs to reply to an IP
address that is not registered on the peer interface yet (hence no ARP
reply will be sent). It's needed to add an ARP entry manually in that
case, as hardware address is already known at that point.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
A problem occurred while running PTP on a multi-port target
(RENESAS RZT2M) with two ports enabled. Announce messages on the
switched devices master port always contained local clock information
instead information from received on the slave port from the better
GM clock. Depending on the BMCA config this turned into having more
than one GM in the system.
Sending always the locally stored GM information helped to overcome
this issue.
Signed-off-by: Manuel Schappacher <manuel.schappacher@hs-offenburg.de>
This reverts commit 6b644dff67.
Reason: breaks Peer-to-Peer gPTP connection. A better solution should be
found to handle the optional TLV on the announce message (chapter 10.5.1
IEEE 802.1AS-2011)
Signed-off-by: Mario Paja <mario.paja@zal.aero>
Commit 55802e5e86 fixed error handling of
TX errors, in case ARP request was generated. There are however also
other places where post-ARP cleanup should be done on the TX path (like
running out of buffers for Ethernet L2 header allocation).
This commit fixes those cases in ethernet_send(), where function would
exit early and report error after ARP prepare stage.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
This fix addresses wrong announce message length warning message.
TLV is a variable length (4+8N) based on the 802.1AS-2011 (table 10-8). In
Zephyr TLV is fixed to 12 bytes. TLV type and length are already taken
into account in the announcement message length.
Signed-off-by: Mario Paja <mario.paja@zal.aero>
The Zephyr's core ethernet code had to be adjusted to support setting T1S
PLCA parameters from user Zephyr programs.
Such approach allows more flexibility, as T1S network configuration;
especially PLCA node numbers, can be assigned not only via device tree
at compile time. For example user can read them from EEPROM and then
configure the network accordingly.
For now - the union in struct ethernet_t1s_param only consists of plca
structure. This can change in the future, when other T1S OA parameters -
like Receive/Transmit Cut-Through Enable (bits RXCTE/TXCTE in OA_CONFIG0
register) are made adjustable from user program.
Signed-off-by: Lukasz Majewski <lukma@denx.de>
A recent iface lock removal in ed17320c3d
exposed issues with concurrent access on TX to drivers that are not
re-entrant.
Reverting that commit does not really solve the problem, as it would
still exist if multiple Traffic Class queues are in use.
Therefore, introduce a separate mutex for TX data path, protecting the
L2/driver from concurrent transfers from several threads.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Move the syscall_handler.h header, used internally only to a dedicated
internal folder that should not be used outside of Zephyr.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
