Disable unstable tests cases for mps2_an385, which fail sporadically
due to timer stability issues on that platform (see #48608).
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Loosen some timing constraints in order to hack arond spurious
failures in upstream Zephyr's CI while working on an unrelated task.
Signed-off-by: Martí Bolívar <marti.bolivar@nordicsemi.no>
This commit disables the `net.socket.tcp.preempt` test on the
`mps2_an385` platform because it fails very frequently due to the
system timer stability issues.
For more details, refer to the issue zephyrproject-rtos/zephyr#48608.
Signed-off-by: Stephanos Ioannidis <root@stephanos.io>
Add a bunch of missing "zephyr/" prefixes to #include statements in
various test and test framework files.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
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>
Previously the test waited for the worst case time till all sockets
where closed, this causes the test to wait for a long unnecessary time.
Secondly it does not ensure the sockets are actually closed before
starting the next tests.
For tests that are not conducted in user mode, wait until all the contexts
are close for a maximum time. If the contexts are not closed in time raise
an explicit assertion.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Move af_packet_ipproto_raw tests to use new ztest API
TEST=twister -T tests/net/socket/af_packet_ipproto_raw/
Signed-off-by: Hu Zhenyu <zhenyu.hu@intel.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>
Increase threads stack sizes from 2k to 3k, so that it will pass without
stack overflow on qemu_x86 platform when SHA384 support will be reenabled.
Signed-off-by: Marcin Niestroj <m.niestroj@emb.dev>
Datagrams should either be fully sent or not sent at all if networking
buffers or network interface MTU does not allow that.
Verify that by trying to send MTU+1 bytes in case of IPv4 (as IP level
fragmentation is not implemented) and IPv6 (should fail when IPv6
fragmentation support is not enabled and succeed otherwise). In case of
IPv6 try to send "total number of network buffers + 1", so that even with
IPv6 fragmentation enabled requested datagram will not be sent. In all
tested cases when datagram is too big, check that ENOMEM error code is set.
NOTE: Tested behavior is not 100% compliant with Linux, as on Linux
EMSGSIZE error code is set when trying to send datagram bigger than MTU,
when manually disabling IP fragmentation (by setting IP_MTU_DISCOVER to
IP_PMTUDISC_DO). However, it is not trivial to implement such
behavior (EMSGSIZE error due to MTU) now and there is always a risk of
running out of buffers (and getting ENOMEM), so for now implemented tests
just assume the latter case to always happen.
Add 3rd testcase, which enables IPv6 fragmentation support, so that UDP
socket behavior can be tested in that context.
Signed-off-by: Marcin Niestroj <m.niestroj@emb.dev>
A reproducible case to the behavior when during a connection
the link gets broken and all data is lost, the TCP connection
should properly terminate.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Fix the existing DGRAM packet socket test. The test will now specify
LL destination address and protocol type. Also, update the fake ethernet
device send handler, so that it sets the packet network interface
corretly, so that the packet is not rejected at Ethernet L2.
Additionally, add two additional test cases verifying datagram packet
sockets. Fist test will exchange packet between RAW and DGRAM socket,
making sure that net stack forwards/removes L2 header correctly. The
second test verifies that it's possible to recieve a packet on both RAW
and DGRAM socket, making sure the received data has correct format.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Datagrams (for DTLS connection) need to be sent using single API call to
mbedTLS and should not result in sending each fragment in a separate
datagram. Check if that is the case, so allow only single fragment data to
be sent and expect an error when multiple fragments were passed to
sendmsg().
Signed-off-by: Marcin Niestroj <m.niestroj@emb.dev>
When there is no response from the server, a client side close
is obstructed, it should terminate and clean the context. This
tests breaks the connection after the accept and validates
this behavior.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
The test suite is failing on qemu_cotext_a9 due to stack overflow.
Apparently this issue can also be observed on main after enabling
STACK_SENTINEL, however the test only started crashing after TCP
changes.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
In order to bring consistency in-tree, migrate all tests to the new
prefix <zephyr/...>. Note that the conversion has been scripted, refer
to #45388 for more details.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
The close function on a TCP socket can return before the complete
socket has been closed, as there might be packets still in flight.
Add a wait at the end of the test so we are sure all sockets have
been closed before the next test starts.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
The close function on a TCP socket can return before the complete
socket has been closed, as there might be packets still in flight.
Modify the test to wait briefly (less then the retransmit time), before
counting the number of still open net_contexts. This makes the test
outcome not dependent on the scheduling order of the different tasks.
Secondly the test actually checks the number of open contexts to be zero
so there is no need to wait for any open contexts to still close.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
When there is no response from the server, a client side connect
should return a ETIMEOUT. This tests breaks the connection and validates
this behavior.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Add unit tests for socket dispatcher module. The tests verify that:
* a socket is automatically dispatched to a regular socket on any
first socket call, if the socket was not bound already before
(according to the default priority rules)
* a socket is correctly dispatched to an offloaded/native socket
implementation after a SO_BINDTODEVICE call
* a socket is correctly dispatched to a native TLS socket, with an
underlying native/offloaded socket, if TLS_NATIVE and SO_BINDOTODEVICE
socket options are used
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
A reproducible case to see what happens when the TCP stack runs out
of buffers. It transfers a block, bigger then the number of buffers
available.
Also test by introducing packet loss in the loopback driver.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Some names of the test cases are duplicated within the project.
This commit contains the proposed names of the test scenarios.
Signed-off-by: Katarzyna Giadla <katarzyna.giadla@nordicsemi.no>
Since the loopback driver makes use of the RX packet pool now, adjust
the packet count for the test.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
A common pattern here was to take the work item as the subfield of a
containing object. But the contained field is not a k_work, it's a
k_work_delayable.
Things were working only because the work field was first, so the
pointers had the same value. Do things right and fix things to
produce correct code if/when that field ever moves within delayable.
Signed-off-by: Yong Cong Sin <yongcong.sin@gmail.com>
The print specifier for `atomic_t` should be updated
to `%ld`, `%lu`, or `%lx` to account for the type
change of `atomic_t` to `long`.
Signed-off-by: Christopher Friedt <chrisfriedt@gmail.com>
Previous timeouts values were done when local TCP replies wourd occur
within sender's thread. This was a known behavior: no context switch,
shorter timings for sending packets (locally only) thus shorter
timeouts.
Such behavior changed when local replies had to go through TCP's work
queue. See commit 798588e709Fixes#40129
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Client thread might run before the server gets to put itself on accept.
Leading to the server waiting forever.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Due to the previous change on when to send TCP packet on local IP, pkt
may be held in a queue which is to run on a k_work. This changes the
scheduling, and due to that one of the test is failing to allocate a
RX net_pkt at the time it wants to. (previous TCP connection is not yet
fully closed and still own PKT that new connection cannot get then).
Of course all those waiting paquets require buffers so raising them.
It was verified that there is no leak, adding net_pkt_print() at
tcp_conn_unref() shows that when all tcp connection are finally unrefed:
all net_pkt get freed as well.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
When creating a socket, all of the registered socket implementation are
processed in a sequence, allowing to find appropriate socket
implementation for specified family/type/protocol. So far however,
the order of processing was not clearly defined, leaving ambiguity if
multiple implmentations supported the same set of parameters.
Fix this, by registering socket priority along with implementation. This
makes the processing order of particular socket implementations
explicit, giving more flexibility to the user, for example when it's
neeed to prioritze one implementation over another if they support the
same set of parameters.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Fixes: #38403
Adding NET_DRIVERS menuconfig so that network drivers are grouped
together in its own menu entry under drivers, similar to most other
drivers.
This further has the advantages that `CONFIG_NET_DRIVERS` can be used
for testing to determine if network drivers has been selected.
This changed revealed a dependency loop where both `select` (for SLIP)
and `depends` (for PPP) which both depends on NET_DRIVERS` where in use
in the dependency tree for Qemu networking, especially NET_SLIP_TAP.
This is handled by defaulting `NET_DRIVERS` to `y` when building for a
Qemu target.
`SLIP` had a dependency to `!QEMU_TARGET || NET_QEMU_SLIP`. This is
changed so that SLIP prompt depends on `!QEMU_TARGET` which provides
full user control in hardware but makes the symbol promptless on Qemu
targets.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
