Skip IP header checksum calculation when the network interface reports
support for this feature.
Signed-off-by: Daniel DeGrasse <daniel.degrasse@nxp.com>
According to RFC 793, ch 3.9 Event Processing, receving SYN flag after
the connection has been established is an error codition:
If the SYN is in the window it is an error, send a reset, any
outstanding RECEIVEs and SEND should receive "reset" responses,
all segment queues should be flushed, the user should also
receive an unsolicited general "connection reset" signal, enter
the CLOSED state, delete the TCB, and return."
Currently TCP stack ignored such event, causing interoperability test
failures. Fix this, by verifying if the SYN flag is set in a packet in
any state other than TCP_LISTEN and TCP_SYN_SENT.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
While a previous change had already decoupled the IEEE 802.15.4 L2 from
IP upper layers, this only worked when at least one other interface of
the device supported IP.
This change removes this requirement and thereby fixes a build error
that occurred when disabling IP support while maintaining IEEE 802.15.4
L2 support.
Fixes: #48718
Signed-off-by: Florian Grandel <jerico.dev@gmail.com>
net_if_up() unconditionally calls iface_ipv6_start() and
net_ipv4_autoconf_start(). This is not correct for interfaces that do
not support IPv4/v6.
This patch therefore introduces a check on the interface's flags before
proceeding.
Signed-off-by: Florian Grandel <jerico.dev@gmail.com>
This change decouples the IEEE 802.15.4 (L2) layer from all IPv6 (L3)
concerns.
Applications may now choose to set CONFIG_NET_6LO=n and
CONFIG_NET_L2_IEEE802154=y at the same time.
Setting CONFIG_NET_6LO=n will build a vanilla IEEE 802.15.4-2006 specs
compliant L2 layer without any reference to 6LoWPAN or IPv6. This allows
application developers to design custom non-IP protocols on top of
IEEE 802.15.4-2006 and thereby makes the L2 layer much more re-usable.
Fixes#48585.
Signed-off-by: Florian Grandel <jerico.dev@gmail.com>
This change contains some merely editorial changes to inline comments
plus updates references from the IEEE 802.15.4-2003 spec to
IEEE 802.15.4-2006 which corresponds to the implementation level of
the module.
Signed-off-by: Florian Grandel <jerico.dev@gmail.com>
Network connection was not unregistered properly for packet socket,
which resulted in dangling connection when net_context was dereferenced
(i. e. when packet socket was closed). This could lead to a crash on a
consecutive incoming packet.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
The L2 protocol type information is not carried to the upper layers.
This is problematic for packet sockets, as the address structure in
recvfrom() is supposed to provide this information.
Fix this by adding ll_proto_type field in the net_pkt structure.
Set the protocol type in the Ethernet L2 when packet is processed.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
It can happen that two similar stacks enter a retransmission cycle
due to a packet collision. If the transmission timeout is the same
both stacks will retry at the same moment resulting in another
collision.
By introducing a randomized retry timeout, the chance of
a second collision is reduced and it reduces furter the more
retransmissions occur.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Apply an exponentially increasing wait time between tcp retries. This
is a good balance between reacting fast to single lost packets and reduce
the network load when an outage takes a little longer. It also allows the
connection to survive a longer interruption with less retransmissions.
Update the test to accommodate for the increased socket closure timeout
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Fix compilation warning in GCC 12 when it's not obvious that we want to
compare the adresses of the first elements of two arrays
Signed-off-by: Hristo Mitrev <hr.mitrev@gmail.com>
If we're disabling certain features and instead providing function stubs,
it doesn't make sense to use macros for those routines that return values
that the user may check unless we want to provide them with vague errors
like:
src/coap-server.c:101:55: error: expected expression before ';' token
101 | ret = net_ipv6_mld_join(iface, &mcast_addr.sin6_addr);
When a function we're mocking up returns a value, use a static inline
stub.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@huawei.com>
If the call to net_send_data() fails, for example if the forwading
interface is down, then the pkt will leak. The reference taken by
net_pkt_shallow_clone() will never be released. Fix the problem
by dropping the rerefence count in the error path.
Signed-off-by: Andriy Gelman <andriy.gelman@gmail.com>
Logging v1 has been removed and log_strdup wrapper function is no
longer needed. Removing the function and its use in the tree.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
So far, TCP cloned a packet with data on an RX path for the application,
leaving the original packet intact. This isn't really needed, as the
original packet is unconditionally freed later anyway, so the TCP can as
well simply queue the original packet for the application, while
informing the network processing core, that the packet was consumed by
the TCP layer.
This allows to improve the download throughput even further, since the
CPU don't waste time on needles packet copying.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
The current default of 250ms appears to be too low in case connection
teardown takes place in lossy networks - in case of FIN packet
retransmission, the connection on the Zephyr side could have already
been dismissed due to low TIME_WAIT state delay, resulting in ICMP
Destination Unreachable replies.
Increase the default value to 1500ms - this is still pretty low, but at
least gives the peer some time to retransmit the FIN packet.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
In case TCP stack enters TIMED_WAIT state (after receiving FIN/ACK reply
from peer), it should stil be ready to reply with ACK for any
consecutive FIN attempts. Othewise, in case the final ACK from Zephyr
side is lost, the connection is not properly closed on the other end,
and peer keeps retransmitting the final FIN packet.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Implement a mechanism, according to RFC 813, which allows to prevent so
called "Silly Window Syndrome" - a scenario where the TCP receiver keeps
reporting small window sizes in the acknowledgments, effectively
limiting the connection throughput. This allows to improve performance
in low-buffer configurations, where the maximum window size is small,
and the issue was hitting quite often.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
As the stack can safely keep allocating data during retransmission mode
there is no need to take the tx_sem during retransmission any more.
Data stored in the send_data buffer will be transmitted upon the ack of
the data for which an ack is pending. This the application being fully
stalled when the TCP connection enters retransmission mode.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
After the window_full function has been fixed by looking at the
send_data_total instead of the unacked_len. There is no risk
in sending data in transmission mode.
This reverts commit 0088aaefa0.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
This commit extends the `struct net_pkt` structure with
`ieee802154_txpwr` field that contains signed value of the desired
transmission power of a IEEE 802.15.4 frame in dBm.
Signed-off-by: Jedrzej Ciupis <jedrzej.ciupis@nordicsemi.no>
To improve the performance with small chunks send, implement Nagle's
algorithm. Provide the option TCP_NODELAY to disable the algorithm.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
In the function net_tcp_queue_data. When an -ENOBUFS by
tcp_send_queued_data is returned, it throws away the whole block of size
len from the send_data. If the len is > MSS, it could happen that the
first section is transmitted, but at the second an -ENOBUFS occurs.
In that case the data is transmitted, but later on removed from the
send_data.
To circumvent this problem, check if the len + unacked_len is smaller
then the send_data_total. If so, the data can safely be removed from
send_data. Otherwise, just pretend the transmission went OK. The
acknowledgment and retransmit path will eventually take care of it.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
The window full computation was corrected to use the send_total instead
of the unacked_len. This conflicted with the new polling implementation
due to the moment when these values are changed.
Move taking the tx_sem outside of tcp_send_queued_data to handle the
-ENOBUF situation properly in case called from net_tcp_queue_data.
net_tcp_queue_data removes data from the send_data in case the
transmission failed with -ENOBUF. This cause the buffer to be not full
any more.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Log an error to explicitly log a failed buffer allocation in TCP
retransmission. This avoids silently failing retransmissions due to
repeating buffer allocation failures.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
When there was no room to transmit the a next packet to be transmitted,
the -ENOBUFS could cause the retransmission to fail.
Secondly the conn->unacked_len can be set to 0 in the retransmission
process, causing the subscribtion to the transmit timer to fail. Use the
variable send_data_total instead.
Make sure that is the send_data buffer becomes empty the send_timer is
cancelled, but make sure any pending data still keeps on being transmitted.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
The function tcp_data_get tries to update the TCP receive window using
net_context_update_recv_wnd. This function graps the context lock while
the tcp_data_get is called from a situation where it already has tcp lock
is already. Transmission actions do first grab the the context lock and
try to grab the tcp lock afterwards. The combination of both can cause a
deadlock.
By taking the shortcut to directly update the tcp receive window without
going through the net context, the context lock is not required avoiding a
possible deadlock situation.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
In the existing the value received from the other side by the TCP options
is used as MSS for transmission. Since the MSS options are an
announcement rather then a negotionation, it is likely the receiver will
have a different and possibly bigger MSS than allowed by our side.
This allow potentially for different a MSS in the receive and transmit
path.
Directly using the received MSS could cause problems when our MSS is only
allowed to be small. At transmission, for that reason take the minimum of
the received MSS and our desired MSS to find a value compatible to both
sides of the link.
Rename the function to net_tcp_get_recv_mss to net_tcp_get_supported_mss
to better reflect its function in the new situation.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Currently a shallow clone of a packet will bump the reference count on
all the fragments. The net_pkt_unref() function, however, only drops the
reference count on the head fragment. Fix this by only bumping the ref
count on the head buf during shallow clone.
Only bumping the ref count of head is more in line with the idea that
head buf is not responsible for the fragments of its child.
Signed-off-by: Andriy Gelman <andriy.gelman@gmail.com>
Datagrams should either be fully sent or not sent at all if networking
buffers or network interface MTU does not allow that. So far the behavior
was to truncate outgoing packets, even for datagram sockets.
When there is not enough available payload buffer to fit all requested
data, fail if that happens for datagram socket.
Signed-off-by: Marcin Niestroj <m.niestroj@emb.dev>
To be able to accept DHCP discover/request incoming packets the ip stack
neeeds to accept unspecified src addr if the destination is broadcast
address.
Signed-off-by: Olof Winge <olof@skyshaper.net>
It turns out that all of the items listed under TODO section for packet
socket implmeentation has been adressed over time, threfore remove the
obsolete TODO.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Datagram AF_PACKET sockets were not processed properly by the net stack.
Instead of receving a packet already processed L2, and thus with L2
header trimmed, it was receiving a raw, unprocessed packet.
Fix this by calling net_packet_socket_input() for the second time, after
L2 has processed the packet. An updated connection handler module will
forward the packet correctly based on the corresponding socket type and
packet L2 processing status.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Currently, the packet socket implementation in net_context required that
netowrk interface to transmit the packet to was set on every sendto()
call. This spoils the whole idea of binding a socket. Fix this, by
checking first if the net_conext is already bound to a particular
interface, and if so, do not throw an error in case sll_ifindex field is
not a valid interface.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
In the stack both unacked_len and send_data_total track the amount
of data for retransmission. send_data_total actually accounts the
total bytes in the buffer, where unacked_len is used to control the
retransmission progress.
Using unacked_len is sometimes reset to 0, this can lead to more data
being allowd in the send_data buffer. In worse case this can cause
depletion of the net buffers, causing a stall and crash of the connection.
The value send_data_total actually accounts the total amount of data in
the send_data buffer, so it is the proper value to used in the
tcp_window_full function.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
In the function tcp_send_data, the variable conn->unacked_len in copied
into a local variable pos. This value is only used in one location and
used mixed with the original conn->unacked_len.
This fix removes pos and switches to use conn->unacked_len everywhere
to reduce the chance of confusion. This does not functionally change the
code.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Instead of sending ZWP from send context, when it is detected that
window is full due to zero-window, implement a proper persistent timer,
that is scheduled once zero-window is detected. The timer is responsible
for sending ZWP to the peer and is canceled once non-zero-window is
notified by the peer.
Additionally, in case peer reported zero-window, do not trigger
retransmission from net_tcp_queue_data(), as it won't be transmitted
anyway by the stack.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
The semaphore is reset when TCP layer would normally reject transfer
request (either due to TX window being full or entering retransmission
mode). Once data is acnowledged, or the reatransmission is done, the
semaphore is set again.
Upper layers can monitor the semaphore with `k_poll()` instead of
waiting blindly before attempting to transmit again.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Make use of the status argument in the recv_cb() callback function -
instead of blindly reporting ECONNRESET whenever TCP context is
dereferenced, indicate whether an actual error condition happened (by
setting respective errno value) or a graceful shutdown took place (by
setting status to 0).
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Remove NET_TCP_BACKLOG_SIZE from KConfig and from test,
because it's not present anymore in current version of TCP stack.
Signed-off-by: Andrey Dodonov <Andrey.Dodonov@endress.com>
After introducing SO_SNDBUF socket option, a possible deadlock situation
slipped into the TCP implementation. The scenario for the deadlock:
* application thread tries to send some data, it enters
net_context_send() which locks the context mutex,
* internal context_sendto() blocks on a TX packet allocation, if the
TX pool is empty rescheduling takes place,
* now, if at the same time some incoming packet has arrived (ACK for
example), TCP stack enters tcp_in() function from a different
thread. The function locks the TCP connection mutex, and tries to
obtain the SNDBUF option value. net_context_get_option() tries to
lock the context mutex, but it is already held by the transmitting
thread, so the receiver thread blocks
* when TX packet is available again, the transmitting thread unblocks
and tries to pass the packet down to TCP stack. net_tcp_queue_data()
is called which attempts to lock the TCP connection mutex, but it is
already held by the receiving thread. Both threads are in a deadlock
now with no chance to recover.
Fix this, by obtaining the SNDBUF option value in tcp_in() before
locking the TCP connection mutex.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
In order to bring consistency in-tree, migrate all subsystems code to
the new prefix <zephyr/...>. Note that the conversion has been scripted,
refer to zephyrproject-rtos#45388 for more details.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
This complements the Kconfig possibility, and allows setting an
interface as default on runtime. Changing the default interface also
works around limitations when trying to use an offloaded interface
together with a native one.
Signed-off-by: Ole Morten Haaland <omh@icsys.no>
