Installing recv callback with net_context_recv() after
net_context_connect() left an opening for a possible race - in case the
server send some data immediately after establishing TCP connection, and
Zephyr did not manage to install the callback on time, the data would be
lost, corrupting the stream.
This can be avoided, by installing the recv callback before the
connection is triggered. As net_context_recv() called w/o timeout only
registers the callback function, it should have no negative impact. The
only change on the TCP side is when the connection is closed - in case
TCP is in connect stage, do not call the recv callback (before this
change it'd be NULL at that point).
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
According to RFC768 UDP packets with zero checksum are allowed with IPv4.
Enable this by default.
For example, some routers use zero checksum in DHCP packets.
Signed-off-by: Nils Ruf <nils.ruf@endress.com>
In certain scenarios, it may be necessary to get values of additional
options from the application layer. With this patch, this can be
accomplished by registering a callback with the DHCP client.
This change has been tested using the posix build in qemu.
Signed-off-by: Matthias Breithaupt <m.breithaupt@vogl-electronic.com>
Until now iterable sections APIs have been part of the toolchain
(common) headers. They are not strictly related to a toolchain, they
just rely on linker providing support for sections. Most files relied on
indirect includes to access the API, now, it is included as needed.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Shifting "timeout <<= conn->zwp_retries" by more then 63 bits results in
undefined behaviour.
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
If a socket has DSCP set then the packets from the socket should also be
marked with appropriate priority in case traffic classes are used in
networking stack.
Signed-off-by: Krishna T <krishna.t@nordicsemi.no>
In the FIN_WAIT_1 any incoming data is dropped, but anknowledged. Add a
check to see if the data is already acknowleged to prevent double
acknowledging of the data and bringing the acknowlegde counter out of
sync. When the acknowledge counter gets out of sync, the connection will
never properly terminate any more.
Signed-off-by: Sjors Hettinga <s.a.hettinga@gmail.com>
Provide an extra struct net_if * iface parameter to
net_if_ipv*_maddr_join/leave functions, so that the corresponding
interface context, the mcast address belong to, can be locked for the
operation.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
As `net_if_get_link_addr()` returns a pointer to the link layer
structure, some extra protection is needed to prevent its contents from
being changed while accessed. Using the mutex lock associated with an
interface should do the trick.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Instead of having a single global mutex, covering, in most cases
unrelated, access to all network interfaces, introduce a separate mutex
for each network interface.
Wherever interface function operates on an individual interface context
data, use local mutex protection instead. The global mutex remains only
to protect access to shared resources (IPv4/6 context pool, timer lists
etc.).
Finally, rework the timer expiry handlers, to prevent interface
functions from being called while the global lock is held (the global
mutex has to be held when iterating over the timer list, as lists are
not thread safe). The reason for this is to avoid possible deadlock
scenario. The global mutex is sometimes being locked, while inside
interface function (protected by a local mutex), so if it was allowed
the other way as well, a deadlock could occur.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Fix possible NULL dereference at net_pkt_cursor_operate() and
resolve 'Dereference after null check' warnings from the Coverity
static code analysis.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
Print new NET_IF_IPV6_NO_ND and NET_IF_IPV6_NO_MLD flags when
the "net iface" command is used.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Add new network interface flag, which allows to disable Multicast
Listener Discovery protocol from being used on the interface. This
allows to interfaces that do not support ND (like OpenThread) to
coexist with other IPv6 interfaces.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Add new network interface flag, which allows to disable Neighbour
Discovery protocol from being used on the interface. This allows to
interfaces that do not support ND (like OpenThread) to coexist with
other IPv6 interfaces.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Fix dhcpv4_create_message() dereferenced NULL return value from
net_pkt_alloc_with_buffer() in case of no memory available.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
Fix net_ipv4_send_fragmented_pkt() dereference of NULL return value
from net_pkt_get_data() possible with non contiguous packet data.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
The arm-clang compiler/linker does not optimize away unused function
symbols and thus will error if symbols that are referenced are not
defined. To fix this add needed ifdef'ry.
Build various network samples that utilize ethernet but don't have
CONFIG_NET_PROMISCUOUS_MODE will get a link error for:
Error: L6218E: Undefined symbol
net_mgmt_NET_REQUEST_ETHERNET_SET_PROMISC_MODE
(referred from ethernet.o).
Fix by adding ifdef protection around promisc code.
Signed-off-by: Kumar Gala <kumar.gala@intel.com>
Fix tcp_send_zwp() retransmission timeout calculation to avoid sign
extension on bit shift operation: the uint16_t value was implicitly
promoted to 32-bit signed type with sign extension to uint64_t.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
Added a feature of socket connect
being asynchronous. If socket is set
to nonblock with O_NONBLOCK flag,
then connect() is non-blocking aswell.
App can normally poll the socket to
test when the connection is established.
Signed-off-by: Daniel Nejezchleb <dnejezchleb@hwg.cz>
Fixes#56657, incorrect TCP MSS calculation for IPv6.
Previously the supported MSS would incorrectly returned as the MTU if it
was less than the MTU, probably to account for the case where
the network interface MTU check returns 0.
New behaviour is to return the supported MSS as MTU minus header length,
using default MTU for this calculation if network interface MTU check
returns 0.
Signed-off-by: Kenny Johansson <wirehell@gmail.com>
The init infrastructure, found in `init.h`, is currently used by:
- `SYS_INIT`: to call functions before `main`
- `DEVICE_*`: to initialize devices
They are all sorted according to an initialization level + a priority.
`SYS_INIT` calls are really orthogonal to devices, however, the required
function signature requires a `const struct device *dev` as a first
argument. The only reason for that is because the same init machinery is
used by devices, so we have something like:
```c
struct init_entry {
int (*init)(const struct device *dev);
/* only set by DEVICE_*, otherwise NULL */
const struct device *dev;
}
```
As a result, we end up with such weird/ugly pattern:
```c
static int my_init(const struct device *dev)
{
/* always NULL! add ARG_UNUSED to avoid compiler warning */
ARG_UNUSED(dev);
...
}
```
This is really a result of poor internals isolation. This patch proposes
a to make init entries more flexible so that they can accept sytem
initialization calls like this:
```c
static int my_init(void)
{
...
}
```
This is achieved using a union:
```c
union init_function {
/* for SYS_INIT, used when init_entry.dev == NULL */
int (*sys)(void);
/* for DEVICE*, used when init_entry.dev != NULL */
int (*dev)(const struct device *dev);
};
struct init_entry {
/* stores init function (either for SYS_INIT or DEVICE*)
union init_function init_fn;
/* stores device pointer for DEVICE*, NULL for SYS_INIT. Allows
* to know which union entry to call.
*/
const struct device *dev;
}
```
This solution **does not increase ROM usage**, and allows to offer clean
public APIs for both SYS_INIT and DEVICE*. Note that however, init
machinery keeps a coupling with devices.
**NOTE**: This is a breaking change! All `SYS_INIT` functions will need
to be converted to the new signature. See the script offered in the
following commit.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
init: convert SYS_INIT functions to the new signature
Conversion scripted using scripts/utils/migrate_sys_init.py.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
manifest: update projects for SYS_INIT changes
Update modules with updated SYS_INIT calls:
- hal_ti
- lvgl
- sof
- TraceRecorderSource
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
tests: devicetree: devices: adjust test
Adjust test according to the recently introduced SYS_INIT
infrastructure.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
tests: kernel: threads: adjust SYS_INIT call
Adjust to the new signature: int (*init_fn)(void);
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
TCP packet allocation timeout is currently 100ms, but there are cases
where it is not enough and as a side effect, the kernel internals are
printing some errors on the log before retrying again, create a
Kconfig parameter to be able to tune this value.
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
In case native IPv6 is disabled, Kconfig entries related to native IPv6
stack should not be enabled. Otherwise, circular dependencies can be
created if native stack is disabled, as in case of recent changes in
NET_IPV6_RA_RDNSS option (where a dependency for a native stack module
was enabled).
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
SO_RCVBUF option processing at the TCP level was broken. The option
value was only checked once, when the TCP context was allocated. This
made little sense, as at this point the option would not even get a
chance to have custom value. If the user modified the option after the
socket (net_context) was created, it had no effect on the TCP operation.
This commit fixes this, by checking the option value whenever new packet
is processed, so that the configured window size is updated at the TCP
level before we report it in the TCP ACK. In order to achieve this,
introduce a new helper function, to refresh the configured window sizes,
to avoid bloating `tcp_in()` even further.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Adds dummy link layer for offloaded ifaces, allowing
ifaces to directly receive l2_enable calls
Signed-off-by: Georges Oates_Larsen <georges.larsen@nordicsemi.no>
Packet handlers registered with net_conn API should not use NET_CONTINUE
to report packet processing status. As they register for a specific
local port, it cannot be assumed someone else will handle the packet,
and for net_conn this status is not really handled differently from
NET_OK. Therefore, reporting NET_CONTINUE, and not freeing the packet,
can result in a packet leak.
Too short packets should not really be handled differently from other
malformed packets in DHCPv4 module, therefore report NET_DROP instead of
NET_CONTINUE.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
TCP implementation abused the packet processing result reporting
mechanism, by reporting NET_DROP for every packet other than data
packets (which were passed to the application). This simplified the TCP
implementation a bit, as it didn't need to deal with packet releasing.
While technically it worked, it gave incorrect results when it comes to
packet statistics, as a lot of valid TCP packets were counted as
dropped.
This commit fixes this. Each packet that was properly processed by the
TCP stack is not either released at the TCP level, or passed to the
application. In both cases, NET_OK is returned to the network core,
indicating the packet has been properly processed and consumed.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
Handle RA RDNSS and use the first DNS server fetched. This is needed
when building IPv6 only without static DNS server IP addresses.
This implementation does not handle the lifetime, because the current
resolve logic does not have support for a DNS server lifetime.
Signed-off-by: Stig Bjørlykke <stig.bjorlykke@nordicsemi.no>
There is no reason why the default size should be 2,
because an event is only a combination of
{uint32_t + void*} + info
Using 2 as default causes a high risk of loosing events,
setting the default to a more sensible value of 5 still
restricts memory usage but keeps it more safe.
Signed-off-by: Jeroen van Dooren <jeroen.van.dooren@nobleo.nl>
The current implementation doesn't handle
re-entrant calls correctly, causing events
to get lost.
Under specific conditions, re-entrancy happens
from event callback done by net_mgmt.c back into
'net_mgmt_event_notify_with_info' causing circular
buffer administration corruption.
These conditions are:
- llmnr_iface_event_handler
- mdns_iface_event_handler
- wifi_mgmt_raise_connect_result_event
- dhcpv4_immediate_timeout uses a workaround
Validated the fix on a Nucleo_H743ZI, using an
sample app.
Signed-off-by: Jeroen van Dooren <jeroen.van.dooren@nobleo.nl>
According to RFC3927, hosts with only IPv4 LL address should still be
able to send packets to destination w/o IPv4 LL address:
A host with an IPv4 Link-Local address may send to a destination
which does not have an IPv4 Link-Local address. If the host is not
multi-homed, the procedure is simple and unambiguous: Using ARP and
forwarding directly to on-link destinations is the default route
This behaviour however was not possible with Zephyr, which only allowed
to use IPv4 LL source address for IPv4 LL destinations.
Fix this, by introducing a final fallback (only if IPv4 autoconf is
enabled), to select IPv4 LL address as a source address if no other
address is available.
Additionally, modify the ARP routine a bit if IPv4 LL address is in use.
There's no really point to forward the packet to gateway if IPv4 LL
address is used, as such addresses are not routable. Instead, try to
find the peer in local network in such case.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
We need to make sure the user_data of the network buffer has enough room to
store the L2CAP metadata.
Signed-off-by: Jonathan Rico <jonathan.rico@nordicsemi.no>
Bypass the IP stack and go directly to L2 with
SOCK_RAW/IPPROTO_RAW sockets in net_if_send_data().
Signed-off-by: Jani Hirsimäki <jani.hirsimaki@nordicsemi.no>
IEEE802154 drivers expect, that a complete 802.15.4 frame fits into a
single net_buf provided in a net_pkt.
There is a corner case with header compression when this could not be
the case. If the IPv6 packet before IPHC exceeded 802.15.4 MTU, it'd
consist of several net_buf's. This was not an issue, if the IPv6 packet
after header compression still required fragmentation, as the
fragmentation module took care of formatting individual net_buf's
properly.
For short range of packet sizes however this was no the case. The IPv6
packet, after header compression, might not require fragmentation any
more. As the IPHC logic only modified the IPv6 header part, by trimming
the buffer in front of the net_buf, such a packet would still consist of
two net_bufs, even though it should fit into a single 15.4 frame. Such
packet was passed to the driver, causing the driver to send only part of
the packet, from the first net_buf.
Fix this, by using net_pkt functions to manipulate the packet, instead
of operating on net_buf directly. net_pkt_pull() will not trim the
buffer in front, but rather move the entire packet content to the front.
On the other hand, net_pkt_compact() will assure that there's no gaps in
the net_bufs. In result, packets that do not require fragmentation,
should be placed into a single net_buf, as expected by drivers.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
It should be possible to configure/initialize the DHCPv4 for an
interface when the interface is down, the DHCPv4 however should not
proceed with the procedure in such case.
Add a simple check to prevent DHCPv4 from sending packets when interface
is down. As the module registers to the NET_IF_UP event, the procedure
will be restarted when the interface goes up.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
During the requesting state, ignore NAKs from any
non-requesting server. This gives the requesting
server a chance to ACK the request.
Signed-off-by: Ryan Erickson <ryan.erickson@lairdconnect.com>
When IPv4 autoconf is enabled and IPv4 LL address is configured on an
interface, the subnet mask should be updated accordingly, otherwise it
can lead to unexpected behaviour (like for example not identifying peer
LL address as a part of the same subnet).
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
For socket families that are niether INET/INET6 e.g., PACKET, we cannot
estimate header length, so, if the payload is zero as well, networking
stack drop the packet.
Instead, allow for zero header + payload packets and let L2 take care of
handling them, for some L2's this is still a valid frame.
E..g, In case of Wi-Fi this is sent as 802.11 Header + LLC Header
(no payload).
Add unittests for both AF_INET + Zero payload and AF_PACKET + Zero
payload.
Signed-off-by: Krishna T <krishna.t@nordicsemi.no>
This mirrors the behaviour of MLD packets, the only other ICMPv6 subtype
that is tracked separately.
Inbound ND packets are already counted as received ICMP packets by way of
going through the ICMP receive path.
Signed-off-by: Armin Brauns <armin.brauns@embedded-solutions.at>
This caused outgoing MAC address lookups to count as one received and two
sent Neighbour Discovery packets, when it's really one sent Neighbour
Solicitation and one received Neighbour Advertisement.
Signed-off-by: Armin Brauns <armin.brauns@embedded-solutions.at>
The net_pkt pointer provided to net pkt commands was not validated in
any way. Therefore it was fairly easy to crash an application by
providing invalid address.
This commit adds the pointer validation. It's checked whether the
pointer provided belongs to any net_pkt pools known to the net stack,
and if the pointer offset within the slab actually points to the
beginning of the net_pkt structure.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
There is a pending TODO regarding unifying net_tcp_unref() and
net_tcp_put(). Given that net_tcp_unref() is no longer used by the upper
layer (it should only use get/put APIs), the function can be removed
from external TCP API, as referencing/dereferencing is now only used
internally by the TCP stack.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
