Check if a network interface is managed by a network manager before
falling back to offload API.
Signed-off-by: Chaitanya Tata <Chaitanya.Tata@nordicsemi.no>
This introduces support for Wi-Fi network managers in Zephyr. The
motivation is for the Wi-Fi management layer to work with both
Network managers and offloaded Wi-Fi drivers. The device driver
decides which one to use.
network manager : Apps -> Wi-Fi Mgmt -> Network Manager -> Wi-Fi
interface
offloaded : Apps -> Wi-Fi Mgmt -> Wi-Fi offloaded interface
Support for multiple network managers has been added, each device can
choose its own network manager and there can be mix and match:
wlan0 - Offloaded
wlan1 - Network manager 1
wlan2 - Network manager 2
Signed-off-by: Chaitanya Tata <Chaitanya.Tata@nordicsemi.no>
Decouple interface and Wi-Fi APIs, Wi-Fi APIs are common independent of
Wi-Fi offload or implemented natively (This is preparation for
introducing Native Wi-Fi).
Signed-off-by: Chaitanya Tata <Chaitanya.Tata@nordicsemi.no>
The management thread *barely* fits in 768 bytes of stack; when thread
local storage is in use and TLS variables are also allocated from the same
region, this stack can overflow.
Increase to 800 bytes to leave plenty of room for TLS variables.
Signed-off-by: Keith Packard <keithp@keithp.com>
Removes redundant ACK state from `struct ieee802154_context` and
simplifies the ACK procedure.
Signed-off-by: Florian Grandel <fgrandel@code-for-humans.de>
This change introduces test coverage for association request and
response. Based on this coverage, several closely related issues were
found in the association process which cannot be split into separate
changes without breaking the build.
Most notably did the associate and disassociate net_mgmt commands send
already encoded IEEE 802.15.4 MPDUs to L3 rather than L2. L3 treated
them as payload and made L2 wrap them with another LL header/footer
which produced invalid packets.
The tests also enforce better aligment of the association process with
the IEEE 802.15.4-2020 standard:
* Association requests now ask for ACK as required by the standard. The
fake driver was enhanced to produce ACK packages when requested.
* macPanId and macCoordinator* MAC PIB attributes are set in the right
order for improved filtering of association responses.
* The coordinator may decide not to assign a short address to the end
device even when associated. This is now supported.
* The coordinator may or may not use a short address. Coordinators
choosing not to support short addresses are now supported.
* Updating the association will now remove any previously added short
address from the hardware filter.
* The short address may no longer be changed by the user while
associated to a PAN. Only the coordinator is allowed to allocate short
addresses.
* Validation of outgoing and incoming association request/response
packets is improved.
All changes are documented by pointers into the spec.
Signed-off-by: Florian Grandel <fgrandel@code-for-humans.de>
Fixes an off-by-one bug in the parsing routine of the coordinator
address when associating via shell command.
Signed-off-by: Florian Grandel <fgrandel@code-for-humans.de>
Input validation of some of the IEEE 802.15.4 net_mgmt commands was
incomplete and/or inconsistent. This change introduces a consistent
approach to input validation that is easier to follow.
Signed-off-by: Florian Grandel <fgrandel@code-for-humans.de>
Calculates macResponseWaitTime and applies it to the association
process.
As the timing calculation re-uses symbol period calculations and other
PHY timing constants previously introduced, these are now shared as
utility functions.
Signed-off-by: Florian Grandel <fgrandel@code-for-humans.de>
The IEEE 802.15.4 stack defines radio API helpers that provide
simplified and encapsulated access to radio API features.
These helpers were missing the `_radio_` infix. This infix is introduced
to clearly distinguish between MAC and PHY concerns. While PHY features
may be shared between L2 implementations (including the functions
concerned here), this is not true for MAC features.
Signed-off-by: Florian Grandel <fgrandel@code-for-humans.de>
The IEEE 802.15.4-2020 standard introduces an association type field to
support fast association, see sections 6.4.3 and 7.5.2. We do not yet
implement fast association but we introduce the flag to make this
obvious.
Signed-off-by: Florian Grandel <fgrandel@code-for-humans.de>
Minor simplification in the definition of IEEE 802.15.4 Kconfig packet
fragmentation configuration.
Signed-off-by: Florian Grandel <fgrandel@code-for-humans.de>
The additional hooks provide infrastructure to construct
rules on another network stack levels. Main benefit of this
approach is packets are pre-parsed and e.g. IP filter is
easier to implement. These hooks are equivalent of prerouting
and local_in in linux's netfilter.
Signed-off-by: Marcin Gasiorek <marcin.gasiorek@nordicsemi.no>
Fixes an issue where a variable was used without being correctly
set by other parts of the code.
Signed-off-by: Jamie McCrae <jamie.mccrae@nordicsemi.no>
This moves copying of parameters after application return value
validation so that the metadata is not stored if rejected.
Signed-off-by: Mariusz Skamra <mariusz.skamra@codecoup.pl>
Merging the two structs cleans up significant amount of code
and makes it easier to expand later.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
Previously the calculation of the PSN was done using
just a timer. This would not work correctly when
attempting to send multiple packets in a single SDU
interval (e.g. to enqueue 2 or more).
Previously we only ever attempted to send 1 packet in
the `sent` callback but scheduling the work item. However
in the case that the `sent` callback was called twice
before the work item had been triggered (possible due
to the priority of the RX thread being higher than the
system workqueue thread). This has been modified so that
if we can enqueue more packets in the lc3_audio_send_data
function, we re-enqueue the work item. We re-enqueue rather
than sending multiple in a single call to avoid blocking
other items on the system workqueue from being scheduled.
This could cause possible missed intervals (if other workqueue
items are processed), but it should be OK from a ISO perspective,
and the new PSN calculator handles this much better.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
Modify lc3_audio_send_data to be triggered on a delayable
k_work so that we can retry it with a delay if it fails for whatever
reason.
This also moves the calculation of the seq_num closer
to when it is used, which makes sense as it is based on
a timed calculation.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
The current API for changing the bondable mode uses the global flag.
With Zephyr support for multiple Bluetooth identities, the API for
changing the bondable mode should be more fine-grained.
The bondable requirements of one identity should not have an impact on
another identity which can have a different set of requirements.
This change introduces function to overlay bondable flag per
connection.
Signed-off-by: Mateusz Kapala <mateusz.kapala@nordicsemi.no>
Current structure of Bluetooth Mesh Kconfig options is quite messy.
This makes impossible to understand which configuration belongs to
which feature. Especially when using interactive Kconfig interface, like
menuconfig or guiconfig.
This commit restructures the options grouping them by protocol layer
they belong to (Network, Transport, Access, etc.), or specific feature
(LPN, Friend, Proxy, Relay, etc.), or implementation (Advertiser, Shell,
Persistent storage). Amount of supported keys, subnets, group addresses
and labels are grouped under Capabilities menu. Generic options that
don't fall to any category are kept at the root menu. For better
visibility, if a specific feature or layer has more than 1 option, they
are hidden under menu.
Signed-off-by: Pavel Vasilyev <pavel.vasilyev@nordicsemi.no>
The sizeof is used in a wrong way which causes incorrect checking a
version of a command. Use NUM_BITS instead.
Signed-off-by: Dawid Niedzwiecki <dawidn@google.com>
The response buffer has to be cleared every command not to pass
unintended content e.g. response from a previous command, or stack
content.
Signed-off-by: Dawid Niedzwiecki <dawidn@google.com>
Update the response buffer size, passed to a command handler, every
command, since a backend could change it in runtime.
Signed-off-by: Dawid Niedzwiecki <dawidn@google.com>
PR adds the statistic module to estimate frame handling.
The module helps to understand the ratio of
the received\relayed\dropped\transmited frames.
That shows the efficiency of the current configuration\implementation.
Signed-off-by: Aleksandr Khromykh <aleksandr.khromykh@nordicsemi.no>
The broadcast_id and the pa_sync fields should be set before calling
broadcast_sink_add_src as broadcast_sink_add_src will those values to
compare against other receive states.
The function was also missing a call to broadcast_sink_cleanup
for the newly allocated sink in case there was an error.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
The CIG state variable is set to CIG_STATE_CONFIGURABLE, but then
cleared by memcpy from configuration cache. Set state after memcpy.
Fixes EBQ test /HCI/CIS/BI-10-C.
Signed-off-by: Morten Priess <mtpr@oticon.com>
In the case where memq_deinit in LLL flushing ends up with the free
link being the link provided by the CIS instance, and pointed to by
lll.link_tx_free, the free counter data in the link element is
overwritten during central CIS creation.
This has the effect that when starting the next CIG, there will suddenly
be 0 links available, and controller fails assertion.
By saving- and restoring the lll.link_tx before and after CIS
configuration cache copying, the free counter is intact.
Signed-off-by: Morten Priess <mtpr@oticon.com>
An advertising report must not be generated unless the AUX_ADV_IND
has been received (for ADV_EXT_IND with an auxptr)
Signed-off-by: Troels Nilsson <trnn@demant.com>
Use CONFIG_BT_CTLR_PHY_CODED to flag out coded phy code
Don't schedule a scan for coded phy when it is not supported
Signed-off-by: Troels Nilsson <trnn@demant.com>
When a stream leaves the streaming state the `stopped` callback
will now be called, similar to how the `started` callback works.
This ensures that `stopped` is always called and not just when
the CIS disconnects.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
So they depend or select on the right NATIVE_BUILD
instead of NATIVE_APPLICATION.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
parse_number() helper function accepts long * pointer or the output,
however in several places, an address of a variable of incompatible
type was passed to the function (for example an address of a bool
variable was cast to (long *) and provided to the function). This
could cause memory overwrites or other unexpected behaviour.
Fix this, by defining a helper variable of type long, and use it with
the parse_number() function. Only after successful parsing, the value is
then assigned to the proper destination.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
When PMAX value is changed, it should update all events.
I believe there is a bug that caused the code only to update
events that are ongoing (to be send).
Now if PMAX changes, next event timestamp is recalculated.
Fixes#59397
Signed-off-by: Seppo Takalo <seppo.takalo@nordicsemi.no>
The enum used for connection types gets named bt_conn_type to guard
against accidental usage of generic integers with relation to it.
The added default case in several switch statements avoids warnings
against unhandled enum values.
Signed-off-by: Arkadiusz Kozdra <akozdra@antmicro.com>
Fail gracefully if an HCI event of one type arrives for a handle of a
different connection type. The requested types are currently based on
what fields are used, not on the usage context, in order to keep every
correct use so far still working.
A warning is logged if the connection identified by the handle does not
match the requested connection type.
Signed-off-by: Arkadiusz Kozdra <akozdra@antmicro.com>
The "virtio_" prefix is used by the open-amp library API.
Rename local functions using "ipc_virtio_" prefix to avoid
function redefinition.
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@foss.st.com>
Instead of waiting for ase_buf to be available, the read of an
ASE in IDLE state can be handled without using ase_buf.
Signed-off-by: Mariusz Skamra <mariusz.skamra@codecoup.pl>
This fixes missing control point status and ASE status change, when
failed to disconnect the CIS. As the server may, but does not have to
disconnect the CIS, the operation should not be considered as failed
when it happens.
Signed-off-by: Mariusz Skamra <mariusz.skamra@codecoup.pl>
PR instantiates cdb and core keys to prevent scenario
when key is changed over cdb API and gets new ID but
core still operates with old ID(potentially invalid).
Known issue the keys will be desynchronized
during\after key refresh procedure.
Signed-off-by: Aleksandr Khromykh <aleksandr.khromykh@nordicsemi.no>
This commits eliminates the old way of
pacs notifications, replacing k_work with
simple bt_gatt_notify.
Signed-off-by: Szymon Czapracki <szymon.czapracki@codecoup.pl>
