Make sure that the calculated ratio values are used just once.
Without this, if the ratio is even insignificantly larger than 1, the
actual clock quickly drifts to really large numbers. This causes the
sync procedure to restart too often.
Signed-off-by: Tomasz Gorochowik <tgorochowik@antmicro.com>
The problem with the previous approach was that the response timestamp
callback which calls net_pkt_unref could be skipped if the callback was
already registered for another packet. The net_pkt_ref function was
always called which led to memory leaks.
This commit simply disallows handling multiple pdelay requests at once.
If the timestamp callback is already registered, the received request
will not be handled.
Signed-off-by: Tomasz Gorochowik <tgorochowik@antmicro.com>
The issue was that the length field of the tlv extension in the sync
follow up packets was wrong. It is supposed to skip the length of the
header of that extension. The easiest fix was separating the header and
the actual contents into separate structs and that's what this commit
does.
Signed-off-by: Tomasz Gorochowik <tgorochowik@antmicro.com>
As the PTP clock should return the correct time, use that
instead of zephyr uptime for time as that has only ms accuracy.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Use priority 3 (critial app) for outgoing event messages (Sync,
Pdelay_Req and Pdelay_Resp). Use priority 6 (Internetwork Control)
for all other outgoing packets.
See IEEE 802.1Q chapter 8.4.4 for more details.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Split out definition of net_app_init() and its parameter flags from
net_app.h header to new net_config.h header. As we do this, rename
the function to net_config_init() and flags to NET_CONFIG_NEED_*.
This is a second step in splitting out network configuration API
out of net_app API, started in the c60df1311 commit.
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
There is no link address for Dummy network technology, so check
that before trying to print link address.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
This allows zephyr to listen LLMNR DNS queries sent by Windows
and respond to them. See RFC 4795 for details.
The feature requires that hostname is set properly to the
zephyr device and LLMNR is configured properly.
Typically following config options are enough for this support:
CONFIG_NET_HOSTNAME_ENABLE=y
CONFIG_NET_HOSTNAME="zephyr-device"
CONFIG_DNS_RESOLVER=y
CONFIG_LLMNR_RESPONDER=y
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
This implements LLMNR client from RFC 4795. This means that caller
is able to resolve DNS resource records using multicast DNS.
The LLMNR is used in Windows networks.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
An ARP entry, if not free, will be either in pending list or in the ARP
table. What differentiate both is the type of data they hold: either a
pending packet or an actual ethernet address.
It is then possible to unite these 2 attributes to save 4 bytes
per-entry.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Until now, each ARP entry had a k_delayed_work to be used as a timer
when the ARP entry is a pending one, waiting for the ARP request to
succeed in order to get to the ARP table.
But k_delayed_work is not a small object (40 bytes). Thus reworking the
ARP request timer by having one central k_delayed_work and a timestamp
in every ARP entry properly handled at every timeout (1+ entry might
have reached the timeout then).
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Instead of looping over the entries array, which will keep the same
order, let's use slist: one for free entries, one for pending ones and
finally one as the actual ARP table.
This permits some optimizations in how to look up and making small
heuristics by changing entries order when it seems relevant.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
This "net gptp [port id]" command will give some extra info about
gPTP status if gPTP is enabled in config file.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
The application does not do much, it just registers to a callback
in order to get information about gPTP.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
The subsys/net/ directory is more logical place for L2 code instead
of ip/ directory. No functionality changes by this commit.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Defines a PTP clock driver that can be implemented in those network
interface drivers that provide gPTP support.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Signed-off-by: Julien Chevrier <julien.chevrier@intel.com>
This is actually the same as #7229 in which we missed this side of
conversion (only PCP to packet priority was implemented).
The conversion is actually the same both ways, thus it uses the map
added earlier.
Signed-off-by: Tomasz Gorochowik <tgorochowik@antmicro.com>
Allow creation of TX timestamp thread which will collect TX timestamp
information from device drivers. If the callback is registered, then
it will pass that timestamp information to the relevant party for
further processing. This support will be used by gPTP code in
subsequent commits.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Signed-off-by: Julien Chevrier <julien.chevrier@intel.com>
Due to the fact that cmake doesn't allow to build empty "library"
(which would be the case when neither CONFIG_NET_APP_SERVER nor
CONFIG_NET_APP_CLIENT is defined).
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
net_app library consisted of 2 disjoint parts: a) library to
setup/configure networking on the application startup; b) library
to setup client and server connections.
As the configuration library is universally useful and is a generic
Zephyr networking component, split it out to a seperate top-level
networking library under net/lib/config/.
Fixes: #7658
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
This patch changes the backoff factor to be random between zero and
2^be-1, as defined by the standard.
The previous implementation generated either a power-of-two number or
zero for bo_n, because it only overlaps a single bit with the random
number. The overlap is very rare, as in random, and results in the
backoff time being zero most of the time.
Signed-off-by: Franco Saworski <franco.saworski@blik.io>
When the priority of the LwM2M engine was lowered, it causes an
occasional registration update to fall outside of the registration
lifetime. This shows up as the following error:
Failed with code 4.4. Retrying registration
Let's try and retry a bit earlier to account for the priority
change.
Signed-off-by: Michael Scott <michael@opensourcefoundries.com>
The ARRAY_SIZE() utility macro will actually test the parameter types,
and ensure that it is only called with arrays, and not arrays decayed
to pointers.
Changes were performed with a simple Coccinelle script.
Signed-off-by: Leandro Pereira <leandro.pereira@intel.com>
'default false' should have been 'default n', though they happen to have
the same effect here, due to undefined Kconfig symbols ('false')
evaluating to 'n' in a boolean sense.
Kconfig bool symbols implicitly default to 'n', so remove the default
rather than fixing it.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
Invalid configurations should be detected during configuration instead
of during compilation whenever possible.
This patch replaces a BUILD_ASSERT on CONFIG_NET_BUF_USER_DATA_SIZE
with what is intended to be an equivalent Kconfig restriction.
Signed-off-by: Sebastian Bøe <sebastian.boe@nordicsemi.no>
Currently, we add TCP options only to SYN+ACK reply to peer's SYN
(i.e. passive open). For consistency, add them also when we send
SYN ourselves (active open). In both cases, we add just MSS option
currently.
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
Ethernet header is always filled in the first fragment of a packet,
so passing it as a separate function paramter is supefluous.
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
There's an apparent typo in testing net_buf headroom. Also, after
adding VLAN header support, its size should be used too.
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
There was no proper support to timeout an ARP requests which meant
that trying to resolve non-existent IP address left network packet
pending on ARP cache.
Fixes#8019
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
For debugging purposes it is useful to know which interface
the dropped packet was received.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
The 'source' of subsys/net/ip/Kconfig in subsys/net/Kconfig is already
within an 'if NETWORKING' block, so the NETWORKING dependency in
subsys/net/ip/Kconfig is redundant.
Remove the redundant dependency.
This gets rid of a bunch of 'NETWORKING && NETWORKING' dependencies in
the auto-generated Kconfig docs.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
Because per-interface statistics rely on interface pointer stored in a
net_pkt, it should not be unreferenced before stats are updated.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Functions for per-interface statistics collection used a pointer to a
packet that could've been deallocated in the net_conn callback function.
In result, application could crash when interface related to the packet
was referenced. To fix that, packet interface is stored earlier, so it
can be used instead for statistics collection.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
The IP stack drops any TCP segment which doesn't fit into our
receive window. However, we still must accept Zero Window Probe
segments, which are segments, usually with data length of 1, which
a peer sends to us after we stayed with zero window for some time.
In this case, we need to repeat an ACK with the old ack number.
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
Catch interface up/down events so that we can renew the
address if interface goes down and is then restored.
Fixes#7553
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Add net_eth_carrier_on() and net_eth_carrier_off() functions that
can be called by ethernet device driver when it detects that carrier
is lost or found.
Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
It's rather confusing to not see current TCP state in any way (it
makes distinguishing different TCP contexts very hard). And nobody
can know/remember that it's printed with CONFIG_NET_DEBUG_TCP
defined. So, just make it be printed always (initially I thought
about printing just numeric value if CONFIG_NET_DEBUG_TCP isn't
defined, but why, if we can print symbolic name easily).
Also, add a hint that defining CONFIG_NET_DEBUG_TCP will still
print even more info (like unacked pkt list) - similarly to
similar helpful hints we have in other parts of net shell.
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
Let's set it by default when allocating net_pkt. A macro will avoid
ifdefs as well
CONFIG_NET_TX_DEFAULT_PRIORITY is always defined.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
When disabling an ethernet interface, only its cache entries need to be
cleared up and not the whole cache. This is meaninful in case there is
2+ ethernet interface instances.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
