FPU context switching is always performed on demand through the FPU
access exception handler. Actual task switching only grants or denies
FPU access depending on the current FPU owner.
Because RISC-V doesn't have a dedicated FPU access exception, we must
catch the Illegal Instruction exception and look for actual FP opcodes.
There is no longer a need to allocate FPU storage on the stack for every
exception making esf smaller and stack overflows less likely.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Instead of saving/restoring FPU content on every exception and task
switch, this replaces FPU sharing support with a "lazy" (on-demand)
context switching algorithm similar to the one used on ARM64.
Every thread starts with FPU access disabled. On the first access the
FPU trap is invoked to:
- flush the FPU content to the previous thread's memory storage;
- restore the current thread's FPU content from memory.
When a thread loads its data in the FPU, it becomes the FPU owner.
FPU content is preserved across task switching, however FPU access is
either allowed if the new thread is the FPU owner, or denied otherwise.
A thread may claim FPU ownership only through the FPU trap. This way,
threads that don't use the FPU won't force an FPU context switch.
If only one running thread uses the FPU, there will be no FPU context
switching to do at all.
It is possible to do FP accesses in ISRs and syscalls. This is not the
norm though, so the same principle is applied here, although exception
contexts may not own the FPU. When they access the FPU, the FPU content
is flushed and the exception context is granted FPU access for the
duration of the exception. Nested IRQs are disallowed in that case to
dispense with the need to save and restore exception's FPU context data.
This is the core implementation only to ease reviewing. It is not yet
hooked into the build.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Right now this is hardcoded to z_sched_ipi(). Make it so that other IPI
services can be added in the future.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
This reverts commit 5824ac90ecd8dd06d9ea117d345dbc6b069e5840.
The resulting revert takes into account changes to the predicate that
guards the use of `resp_addr`.
In the resulting revert, the predicate guarding the use of `resp_addr`
is differs from the naive revert because the intention of that predicate
is to know when `conn.c:bt_conn_le_create` decided to use the Host
resolver. And, the logic in `conn.c:bt_conn_le_create` has changed. The
changes to the logic in `conn.c` are in the following commits:
89780d715d5a09325ba9
Signed-off-by: Aleksander Wasaznik <aleksander.wasaznik@nordicsemi.no>
Fix filter accept list and privacy feature Kconfig default
based on whether host has them enabled.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Change the http timeout mechanism to use poll instead of shutdown.
This should fix a problem where the shutdown will be called in a
different thread context which can lead to deadlocks on certain
driver implementations like offloaded modem drivers.
Fixes#53967
Signed-off-by: Wouter Cappelle <wouter.cappelle@crodeon.com>
Change the name of the custom macro defined for the stm32 devices
to fit the VND_PWM_xxx model
Keeping old deprecated macro, though.
Signed-off-by: Francois Ramu <francois.ramu@st.com>
When removing a CIG, the instance and associated CIS instances were not
always released. This causes repeated Create_CIG/Remove_CIG operations
to fail assertion with resource error.
Enables CONFIG_BT_CTLR_CENTRAL_ISO in bsim_test_audio.
Fixes issue #53301.
Signed-off-by: Morten Priess <mtpr@oticon.com>
The sample application shows how to configure NXP S32 Network Controller
(NETC) for different use-cases.
Signed-off-by: Manuel Arguelles <manuel.arguelles@nxp.com>
Introduce DT nodes for NETC complex and enable its usage for
s32z270dc2_r52 boards. Using PSI0 as default networking interface and
Switch Port0 as it's the only port available on this board.
Signed-off-by: Manuel Arguelles <manuel.arguelles@nxp.com>
Introduce Ethernet low-level driver for NXP S32 Network Controller
(NETC). Current driver allows to manage from Zephyr a Physical Station
Interface (SI) and/or a Virtual SI. The NETC has an integrated Ethernet
Switch. Currently the Switch is initialized from this driver with a
default configuration, and all ports are enabled and transparent for
the user. A separate Switch driver should be addressed in future patches.
Signed-off-by: Manuel Arguelles <manuel.arguelles@nxp.com>
Introduce NXP S32 NETC External MDIO controller driver. Driver supports
a single instance, as current support is based on NXP S32Z/E SoCs.
Signed-off-by: Manuel Arguelles <manuel.arguelles@nxp.com>
Advertise Gigabit Ethernet if the PHY supports it. As with the
other speeds, it is assumed the PHY supports both duplex modes.
Signed-off-by: Manuel Arguelles <manuel.arguelles@nxp.com>
Introduce NXP S32 Message Receive Unit (MRU) driver based
of Mbox API. The MRU couples with a processor and allows to
receive messages from senders, which are other modules or
processors.
Signed-off-by: Manuel Arguelles <manuel.arguelles@nxp.com>
Similar to the ad_init that automatically adds advertising data
from other modules, pa_ad_init has been added to do the same
for periodic advertising.
The only use case so far is the BAP broadcast source
data, if enabled.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
The call to audio_ad_data_add will now either call
connectable_ad_data_add for connectable advertising, or
unconnectable_ad_data_add for unconnectbale advertising.
Most service UUIDs etc. does not make sense to advertise if
not connectable.
This also adds the BAP broadcast advertising data to
unconnectable.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
If the advertising is connectable, and we have privacy,
then according to the HAP spec, we should not advertise
the HAS uuid.
This was done by simply truncating the advertising data, but
it was only truncated by 1, instead of by the size of the
UUID.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
As we are now compiling in parallel all apps, build errors
are not anymore next to the compile line, so let's tell
developers for app we are printing the build output
to ease debugging.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Not all toolchains support newlib so tests that require newlib need
to have a filter to we don't try and build those tests on those
testcases. Some newer tests are missing:
filter: TOOLCHAIN_HAS_NEWLIB == 1
so add that to testcases that needed.
Signed-off-by: Kumar Gala <kumar.gala@intel.com>
New LLCP will process the CIS establishement in the TX path, meaning it
will only process once every ACL connection interval.
If ACL connection interval is significantly longer than the ISO
interval, CIS packets corresponding to the ACL connection interval will
be lost. This may cause latency in audio rendition.
By processing the CIS establishment in the RX path, notification will be
sent immediately, and audio path is enabled fast to reduce latency.
When a CIS fails to establish, the controller shall complete the
procedure by sending a CIS_ESTABLISHED event containing the error.
Signed-off-by: Morten Priess <mtpr@oticon.com>
Disconnect CIS immediately if event done extra status indicates MIC
failure. Return reason BT_HCI_ERR_TERM_DUE_TO_MIC_FAIL.
Fixes EBQ test /LL/CIS/PER/BV-27-C.
Signed-off-by: Morten Priess <mtpr@oticon.com>
When the 'instant' of a CIS is postponed by increasing the conn_event of
the CIS_RSP compared to CIS_REQ, the CIS offset shall be calculated as
an equivalent window offset by an integer number of ISO intervals.
This is obtained by increasing the CIS_Offset_Min and CIS_Offset_Max by
the modulus of the connection interval delay time.
Signed-off-by: Morten Priess <mtpr@oticon.com>
To be able to timely start the first CIS/CIG, if the requested instant
is the next connection event, call ull_peripheral_iso_start directly
from rp_cc_state_wait_rx_cis_ind instead of waiting for next time
through the state machine.
To enable ULL/LLL to prevent adding latency when LLCP is waiting for
instant, local- and remote procedures with instant now expose a function
for checking this state.
Fixed event_counter function to prevent one-off in RX path. Unified to
use single function ull_conn_event_counter.
Fixed LLCP unit tests and added new mock function.
Signed-off-by: Morten Priess <mtpr@oticon.com>
Set the packing field when creating the broadcast source. This
is currently missing, and in uninitialized.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
Add test where privacy is enabled and where
the two devices are bonded and have exchanged IRK
before doing the periodic advertising sync.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
Expand the periodic advertising tests with connections,
to verify that two devices can simultaneously be connected
and keep a periodic advertising sync.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
Adds a test suite exclusively for testing
periodic advertising sync.
The first test added is a simple test that does a periodic
advertising sync, and passes afterwards.
The test suite will be expanded with testing PA syncs
with PAST (both from remote and local transfer) as well
as privacy handling.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
Add the overlays usage for bsim mesh configurations instead of
multiple configurations that duplicate each other a lot.
Signed-off-by: Aleksandr Khromykh <aleksandr.khromykh@nordicsemi.no>
This change enables the ISR Hard Reset sent bits, so that
an interrupt is generated when a Hard Reset is sent or
the Hard Reset failed.
Signed-off-by: Sam Hurst <sbh1187@gmail.com>
The Hard Reset sent signal was tested twice in the same
"if else" structure but only handled in the last test. This
change removes the first detection so that Hard Reset can
be correctly detected.
Signed-off-by: Sam Hurst <sbh1187@gmail.com>
USB stack did leak memory on every SET ADDRESS request. UDC control
allocator could cope with up to 13 leaked allocations. Therefore,
issuing bus reset 13 times in a row (in addition to automatic reset
after connecting the USB cable) was enough to exhaust memory and
thus drive USB stack inoperable.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
On board stm32h747i_disco_m7, USB clock is updated to enable USB1ULPI
clock.
Doing so, it should not delete 48MHz domain clock setting.
Signed-off-by: Erwan Gouriou <erwan.gouriou@linaro.org>
On STM32L4 that probvide HSI48 clock, use it as 48MHz domain clock.
This impacts following devices:
-SDMMC
-RNG
-USB
Otherwise, when HSI48 is not available MSI is used.
Signed-off-by: Erwan Gouriou <erwan.gouriou@linaro.org>
Validate the powered state and transitions for an arbitrary device
hanging off a power domain.
Signed-off-by: Jordan Yates <jordan.yates@data61.csiro.au>
Only run the `PM_DEVICE_ACTION_TURN_ON` and `PM_DEVICE_ACTION_TURN_OFF`
actions for child devices that have refered to the domain via the
`power-domain` property.
This prevents multiple actions being run for devices that refer to
several power domains, e.g.
```
test_dev: test_dev {
compatible = "test-device-pm";
status = "okay";
power-domain = <&test_reg_1>;
alternate-domain = <&test_reg_chained>;
};
```
Signed-off-by: Jordan Yates <jordan.yates@data61.csiro.au>
The struct sensor_value type does NOT require val2 to be positive. The
removed code is in fact a rather serious bug, probably put in place
because it makes printing sensor_values easier.
Signed-off-by: Benjamin Lindqvist <benjamin@eub.se>
The sink ASE does not have a disabling state, but since the
stream callbacks does not necessarily match the ASE states,
we need to do our own check to see if a sink ASE has been
disabled.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
Before this commit we simply always notified/indicated
all presets and active index to any bonded device, even
if they never subscribed to anything, or even if we never
registered HAS.
This commit modifies that, so that only the proper
connected clients will get the notifications/indications,
and that we only send them if something has changed.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
The security_changed callback may be called before
bt_has_register has been called, thus making the active_preset_work
handler uninitailized, causing an ASSERT when the work is submitted
in the security changed callback.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
Calling bt_has_register multiple times would cause attempts
to register a HAS multiple times. Fixed by adding a check
and return with EALREADY if it is already registered.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
