Replace the pure CAS spin loop with a Test-Test-And-Set (TTAS) pattern
in k_spin_lock() for the non-ticket spinlock path.
The current implementation calls atomic_cas() on every iteration, which
takes exclusive ownership of the cache line even when the compare fails.
Under contention, this generates unnecessary cache coherence traffic as
each spinning CPU invalidates the cache line of all other CPUs.
Use a CAS-first approach: attempt atomic_cas() directly on the first
iteration (no penalty in the uncontended common case), and only fall
back to spinning on atomic_get() reads after the first CAS failure.
The atomic_get() reads from the local cache without bus traffic, and
the expensive atomic_cas() is only re-attempted when the lock appears
free. This significantly reduces cache line bouncing under contention
on multi-core SMP systems.
Fixes: #106616
Signed-off-by: Anand Kumar <anandvtu16158@gmail.com>
W/o MBEDTLS_ECP_NIST_OPTIM enabled DTLS handshakes are considerably
slower, making core Thread functionalities like commissioner/joiner
defunct on slower platforms. Therefore, enable the optimizations if
ECC cryptography is in use.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
This PR enables the DNS packet forwarding without taking into account
the return value of `dns_validate_msg` function.
This is to accomodate scenarios like where ANCOUNT is set to 0, or other
cases in which internal DNS implementation will not validate the
message.
Signed-off-by: Cristian Bulacu <cristian.bulacu@nxp.com>
Add a runtime assertion in z_unpend_all_locked() to verify that
_sched_spinlock is actually held by the caller. This catches misuse
early given the function call depth involved.
Extend the availability of z_spin_is_locked() from CONFIG_SMP &&
CONFIG_TEST to also include CONFIG_ASSERT, so the check can be
used in __ASSERT() outside of test builds.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
When halt_thread() calls k_thread_perms_all_clear() under
_sched_spinlock, the permission cleanup can trigger k_free() on
dynamic objects. k_heap_free() then calls z_unpend_all() which
attempts to take _sched_spinlock again, causing a recursive lock.
Fix this by introducing k_heap_free_sched_locked() and
k_free_sched_locked() variants that use z_unpend_all_locked()
to operate on the wait queue without re-acquiring the scheduler
lock. The existing z_unpend_all() becomes a wrapper that takes
the lock and delegates to z_unpend_all_locked().
unref_check() gains a sched_locked parameter: the abort path
(clear_perms_cb) passes true to use the locked free variant,
while k_thread_perms_clear() passes false for the normal path.
Fixes#106659
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
LPADC_DoOffsetCalibration() triggers offset calibration but returns
immediately without waiting for it to complete. If gain calibration
is requested before offset calibration finishes, the hardware silently
rejects it and GCC[RDY] never asserts, causing
LPADC_FinishAutoCalibration() to spin forever at boot (~50% of the
time on affected boards like FRDM-MCXA266).
Fix by replacing LPADC_DoAutoCalibration() with the split HAL API:
LPADC_PrepareAutoCalibration() + k_busy_wait(1U) +
LPADC_FinishAutoCalibration()
The 1us busy-wait gives offset calibration sufficient time to complete
on all LPADC platforms before the gain calibration request is issued.
Only the HAS_CTRL_CALOFS path is affected (the path that also calls
DoOffsetCalibration); the HAS_CFG_CALOFS path has no preceding offset
calibration and is left unchanged.
Fixes#105652
Signed-off-by: Jjateen Gundesha <jjateen97@gmail.com>
Correct some behaviors that were broken in the original SPI DMA + RTIO
work:
* Properly fallback to interrupt behavior if no DMA channels are assigned
to a given SPI peripheral.
* Properly handle held CS lines across transactions when using hardware CS
control.
Signed-off-by: Pete Johanson <pete.johanson@analog.com>
The VIM IP has a register that informs the maximum
number of interrupts that can be supported.
The ASSERT must check whether the number of IRQs
is LESSER THAN OR EQUAL to above register value,
not a strict equality check.
Signed-off-by: Shreyas Shankar <s-shankar@ti.com>
Disable CONFIG_ETH_DRIVER for RW612 BGA board to align with FRDMRW612
board. The ethernet will be disabled in wifi example by default for
RW612 board. Enable this flag if user wants to use ethernet.
Signed-off-by: Hui Bai <hui.bai@nxp.com>
Previously, the purpose of disabling AVPLL and TDDR is to save power
consumption in low power. But all examples of RW612 use clock_init()
and disable AVPLL and TDDR without condition check is not proper.
Move disable AVPLL and TDDR to condition when they are not in use.
The ethernet uses TDDR clock and enet is enabled by default for RW612.
Add CONFIG_ETH_DRIVER along with enet and CONFIG_NET_L2_ETHERNET in
clock_init(). If any of the condition is not match, the TDDR clock
will be disabled.
Signed-off-by: Hui Bai <hui.bai@nxp.com>
The root cause of the build failure in this test on kit_psc3m5_evk has
been addressed. The platform_exclude for this board is no longer
necessary.
Signed-off-by: John Batch <john.batch@infineon.com>
Updates the legacy HAL version of the Infineon clock control driver to
be excluded for devices not using the legacy HAL based drivers.
This fixes a failure in the libraries.devicetree.api_ext test, which
on PSC3M5 introduces a "fixed-clock" DTS note that causes
DT_HAS_FIXED_CLOCK_ENABLED and SOC_FAMILY_INFINEON_CAT1 to both be
selected. This results in the HAL based driver being incorrectly
included in the build, causing build errors.
Assisted-by: GitHub Copilot:claude-opus-4.6
Signed-off-by: John Batch <john.batch@infineon.com>
Fix build error when independent reset is enabled on custom host.
Add out-of-band reset kconfig option to guard dependency on host GPIO,
so that in-band independent reset can be supported by default.
Signed-off-by: Fengming Ye <frank.ye@nxp.com>
In the `FLEXCAN_SetTimingConfig` function(fsl_flexcan.c), the following
statement is implemented when the Enhanced Bit Timing Register is enabled:
```
base->ENCBT = CAN_ENCBT_NRJW(pConfig->rJumpwidth) |
CAN_ENCBT_NTSEG1((uint32_t)pConfig->phaseSeg1 +
pConfig->propSeg + 1U) |
CAN_ENCBT_NTSEG2(pConfig->phaseSeg2);
```
`prop_seg + 1` represents the length of the propagation segment.
- `pConfig->propSeg` = `flexcan_config.timingConfig.propSeg`
- Therefore, flexcan_config.timingConfig.propSeg(can_mcux_flexcan.c) should
be equal to `data->timing.prop_seg - 1U`.
Fixes: #106238
Signed-off-by: Jianchao Wang <Jianchao.wang_1@nxp.com>
Fixes the logic in detecting which partition is being executed
from, also fixes a test for this feature
Signed-off-by: Jamie McCrae <jamie.mccrae@nordicsemi.no>
Add a `LITTLE_ENDIAN` dependency to `LORAWAN_FRAG_TRANSPORT` due to the
use of bitfields and packed structures in decoding on-air payloads.
Signed-off-by: Jordan Yates <jordan@embeint.com>
Update the fragmented data transport parsing to always consume complete
control messages, regardless of any validation errors. Simplify the
implementation of this by defining the packets as packed structures.
Signed-off-by: Jordan Yates <jordan@embeint.com>
Append the documentation link to Github annotation. This make it easier for
users to go directly to the relevant page from a pull request.
Signed-off-by: Pieter De Gendt <pieter.degendt@basalte.be>
After commit b1afb494b0, zephyr,flash points to the
soc-nv-flash child node instead of the
nxp,imx-flexspi-nor controller. This shifts the DTS
hierarchy by one level, so
DT_PARENT(DT_CHOSEN(zephyr_flash)) now returns the
flash controller node, not the FlexSPI bus node.
Use DT_CHOSEN(zephyr_flash_controller) to directly
reference the flash controller, and DT_PARENT of it
to reach the FlexSPI bus where needed.
Signed-off-by: Ofir Shemesh <ofirshemesh777@gmail.com>
Previous versions of Mbed TLS and related repos were close to the final
release point, but not exactly that. This commit bumps all the repos to
the official release tag:
- Mbed TLS: 4.1.0
- TF-PSA-Crypto: 1.1.0
- mldsa-native: 5772b4f4a0105694b1203abb582273f78fa951b7 (what is being
pointed to TF-PSA-Crypto at 1.1.0 release tag)
Signed-off-by: Valerio Setti <vsetti@baylibre.com>
Having a differing number of ACL buffers available between the host
and controller causes unnecessary "impedance" when communicating over HCI.
The Zephyr Bluetooth host also warns about this during initialization:
Num of Controller's ACL packets != ACL bt_conn_tx contexts (15 != 3)
Update the default for BT_BUF_ACL_TX_COUNT to match the controller. Also
update BT_BUF_EVT_RX_COUNT and BT_CONN_TX_MAX appropriately. The latter
should generally match TX_COUNT, while the former needs to be at least as
large to deal with Number of Completed Packets events from the controller.
Signed-off-by: Johan Hedberg <johan.hedberg@silabs.com>
do_create() is a classmethod and cannot access self.logger.
Use a module-level _logger instead, consistent with core.py
and __init__.py in the runners package.
Signed-off-by: Sylvio Alves <sylvio.alves@espressif.com>
This commit addresses two integer handling issues in the ataes132a
crypto driver identified by coverity scans.
1. In ataes132a_send_command, added a centralized validation check
to ensure the 'nparams' value, when combined with the 5-byte
packet overhead, does not exceed the 8-bit 'count' limit or the
physical 64-byte command buffer. This prevents a potential wrap-
around that would cause the chip to receive an invalid length byte.
2. In the Atmel CRC calculation, added an explicit cast to uint16_t
during the bit-shift operation. This prevents unintended integer
promotion and satisfies static analysis regarding potential
overflows during the 16-bit CRC generation.
Fixes#84683Fixes#84690
Signed-off-by: David J. Leach, Jr. <tasmar@gmail.com>
This commit fixes the bug by ensuring `to_write` is properly bounded.
It is now calculated as the minimum of the remaining requested bytes
(`nbytes - written`) and the remaining space in the current block
(`block_size - block_off`).
Fixes#106276
Signed-off-by: Ashirwad Paswan <ashi06712@gmail.com>
Code example for `pm_device_runtime_enable` referenced a return value
that does not exist anymore.
Signed-off-by: Eden Frosst <eden.frosst@rbr-global.com>
Move signal_pending_ipi() inside the K_SPINLOCK block in
z_get_next_switch_handle(). Calling it after the lock release creates a
window where a CPU can consume its own pending IPI bit via atomic_clear
in signal_pending_ipi(), then silently drop it in
arch_sched_directed_ipi() which skips the calling CPU (i == id).
In configurations where secondary CPUs have a single pinned thread and
take no timer or external interrupts, this can lead to a permanent hang:
the idle CPU can only be woken by IPIs, but no IPIs are pending and no
timeslicing IPIs will be generated since the idle thread is not sliceable.
This was reproduced when running under QEMU with the following sequence
of events observed:
CPU 0 CPU 1
───── ─────
Thread calls k_poll(K_MSEC(1))
z_pend_curr():
mark thread PENDING
z_add_timeout(1ms)
do_swap() to idle thread
WFI
Timer tick fires
sys_clock_announce():
slice_timeout(cpu1):
flag_ipi(BIT(1))
signal_pending_ipi():
MSIP[cpu1] = 1
CPU1 wakes from WFI
z_get_next_switch_handle():
acquire _sched_spinlock
next_up() → idle
(thread still PENDING,
timeout hasn't fired yet)
release _sched_spinlock
Timer tick fires
sys_clock_announce():
z_thread_timeout(thread):
z_unpend_thread(thread)
z_ready_thread(thread):
flag_ipi(BIT(1))
signal_pending_ipi():
atomic_clear(pending_ipi)
returns BIT(1)
arch_sched_directed_ipi(BIT(1))
skips self, IPI silently lost
return to idle thread
WFI
thread still on ready queue
Such an interleaving of events is, of course, likely only reproducible in
practice in virtualized environments where (v)CPUs can be descheduled.
With signal_pending_ipi() inside the lock, next_up() and the IPI
dispatch are atomic. Either the concurrent flag_ipi lands before the
lock is acquired (and next_up sees the thread), or it lands after the
lock is released (and the caller dispatches the IPI). There is no
window where a CPU can consume its own bit for a thread it hasn't seen.
Similar races exist in reschedule() and z_reschedule_irqlock() as well.
Although they won't cause the same permanent hang described above, it
can result in unnecessary rescheduling latency. Fix reschedule(), and
add a TODO to z_reschedule_irqlock(); it doesn't not currently take
the sched spinlock.
Signed-off-by: Andrew Bresticker <abrestic@meta.com>
Replace the hardcoded WAMP RPC type values used when encoding OCPP
messages with their corresponding symbolic constants.
- Replace '2' with OCP_WAMP_RPC_REQ
- replace '3' with OCP_WAMP_RPC_RESP
This removes magic numbers and improves human readability.
Signed-off-by: David J. Leach, Jr. <tasmar@gmail.com>
Static analysis has a difficult time tracing whether a call to
ocpp_send_to_server() needs to initialize sndlock and rspsig because
the message type is set very much earlier in the program logic flow.
Note that coverity is creating a single issue for sndlock and for rspsig
in each function that calls ocpp_send_to_server() without initializing
them.
Adding a NULL test will prevent future static analysis false positives.
Fixes#100026Fixes#100025Fixes#100024Fixes#100017Fixes#100015Fixes#100013Fixes#100011Fixes#100010Fixes#100008Fixes#100006
Signed-off-by: David J. Leach, Jr. <tasmar@gmail.com>
