Make uart_aesc_dev_cfg_##no "static const struct" instead of
"static struct" to move this struct into ROM. This reduces the
memory usage for this driver.
Signed-off-by: Daniel Schultz <dnltz@aesc-silicon.de>
Make gpio_aesc_dev_cfg_##no "static const struct" instead of
"static struct" to move this struct into ROM. This reduces the
memory usage for this driver.
Signed-off-by: Daniel Schultz <dnltz@aesc-silicon.de>
Commit f4a2a27071 ("kernel: timeout: make the system clock tick
interface unsigned") changed sys_clock_set_timeout() to take a
uint32_t tick count, and commit c33571b372 ("kernel: timeout: cap
the timeout passed to drivers centrally") removed K_TICKS_FOREVER
from the driver interface. This driver was missed by both, breaking
the build with a conflicting-types error on boards using the RTC JDP
system timer, e.g. frdm_mcxe31b.
Adopt the unsigned prototype, drop the now-meaningless negative-tick
clamp, and replace the K_TICKS_FOREVER test with the new
no-pending-timeout convention: under CONFIG_SYSTEM_CLOCK_SLOPPY_IDLE
the core signals "no deadline" with ticks == SYS_CLOCK_MAX_WAIT, in
which case wait as long as the hardware allows. Other values are
finite and already saturated against cycles_max.
Assisted-by: Claude:claude-fable-5
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
The driver's local RTC_BASE macro collides with the RTC_BASE
peripheral base address macro defined by the NXP MCUX SDK device
headers (e.g. MCXE31B_COMMON.h, pulled in through soc.h), which
breaks the build with a macro-redefinition error on frdm_mcxe31b
when the RTC JDP system timer is enabled.
Rename the driver-local macro to RTC_JDP_BASE so it cannot shadow
the SDK definition. No functional change.
Assisted-by: Claude:claude-fable-5
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Commit f4a2a27071 ("kernel: timeout: make the system clock tick
interface unsigned") changed sys_clock_set_timeout() to take a
uint32_t tick count, and commit c33571b372 ("kernel: timeout: cap
the timeout passed to drivers centrally") removed K_TICKS_FOREVER
from the driver interface. This driver was missed by both, breaking
the build with a conflicting-types error on any board using it, e.g.
sr100_rdk/sr100 via tests/drivers/build_all/timer.
Assisted-by: Claude:claude-fable-5
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
The syscall is declared with a uint32_t level; the verify handler
had uint8_t, which fails to compile once the generated marshalling
include resolves again.
Signed-off-by: Jonathan E. Peace <jep@alphabetiq.com>
CONFIG_LEGACY_GENERATED_INCLUDE_PATH is deprecated and no longer
enabled by default, so generated headers must be included with the
zephyr/ prefix. Fix the legacy <syscalls/haptics.h> include in the
API header and the <syscalls/*_mrsh.c> includes in the syscall
handlers.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
CONFIG_LEGACY_GENERATED_INCLUDE_PATH is deprecated and no longer
enabled by default, so generated headers must be included with the
zephyr/ prefix. Fix the legacy <autoconf.h> include so the driver
builds without the deprecated compatibility option.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Hello, it's great that this driver has finally been released,
and I'm very impressed with the sheer amount of code. I'm reading
it and can't understand everything, but I have a few questions.
- `silabs_efr32_filter`: I see that the Zephyr API requires setting
and unsetting filters, and the Silabs platform also supports unsetting
filters, but why do we return ENOTSUP when the set variable is false?
- `silabs_efr32_cca` isn't implemented and returns ENOTSUP, but
I see a function called `unslotted_csma_ca_channel_access`
in the upper layer that retrieves the return value of
`ieee802154_radio_cca` (this function wraps the cca API).
If it returns an error other than EBUSY, it will be considered
a fatal error and abort the transmission.
I wonder if it's advisable to return 0 (no errror) so it means
that the CCA software doesn't do anything because the CCA
hardware has already performed the task?
- I've added a variable `testing`
if the `CONFIG_IEEE802154_CARRIER_FUNCTIONS`
configuration is enabled.
Signed-off-by: Phuc Hoang <donp172748@gmail.com>
When CONFIG_NXP_WIFI_IPV6 is not enabled, the
WIFI_EVENT_UAP_NET_ADDR_CONFIG message is never sent to wlcmgr_task
because it relies solely on the NET_EVENT_IPV6_DAD_SUCCEED callback.
This causes WLAN_REASON_UAP_SUCCESS to never be triggered, leaving
applications waiting indefinitely for WIFI_STATUS_AP_SUCCESS.
Send WIFI_EVENT_UAP_NET_ADDR_CONFIG directly in
net_configure_address() when IPv6 is disabled, so the uAP state
machine can proceed to CM_UAP_IP_UP and emit the success event
regardless of IPv6 configuration.
Assisted-by: WChat:Claude
Signed-off-by: Maochen Wang <maochen.wang@nxp.com>
The RTC (rtc_jdp) is a single hardware instance that can back either the
counter API or the system timer, but not both at once. When the
zephyr,system-timer chosen property points at an nxp,rtc-jdp node, skip
that instance in the counter driver so the counter and system timer
drivers never bind the same hardware. This mirrors the arbitration in
counter_mcux_lptmr.c.
The driver body is also guarded so that when the only nxp,rtc-jdp
instance is the chosen system timer, no unused counter device or static
functions remain.
Build-tested on frdm_mcxe31b: tests/drivers/counter/counter_basic_api
still instantiates the RTC counter in the default configuration, while
tests/kernel/timer/timer_api builds with the RTC as the system timer and
the counter instance compiled out.
Signed-off-by: Holt Sun <holt.sun@nxp.com>
Add a system timer driver for the NXP RTC (rtc_jdp) IP on the MCXE31x
family. It implements the standard system clock interfaces and is
selected through the zephyr,system-timer chosen property pointing at an
nxp,rtc-jdp node.
The RTC is a free-running 32-bit up-counter with a single absolute
compare register (RTCVAL); the driver runs tickless against it. Because
the RTC keeps counting in all low-power states, it can serve as the
kernel tick source without a companion timer.
RTCVAL is an equality compare that synchronizes into the slow RTC clock
domain and sets INV_RTC while in progress, during which further RTCVAL
writes are ignored. The driver waits for INV_RTC to clear before a write
and, after it, re-reads the counter and retries with the compare bumped
forward if the count has already reached it, so a match is never lost to
the sync latency (a miss would only recur after a full 32-bit wrap). The
scheduling horizon is capped at half the counter range so a compare
cannot wrap past the counter and look like the past, mirroring
mcux_stm_timer.c.
The counter rate (clock-frequency / prescaler) is published at runtime
via TIMER_READS_ITS_FREQUENCY_AT_RUNTIME, making the devicetree
prescaler the single source of truth for the tick rate. The MCXE31x SoC
defconfig disables SysTick and defaults the tick rate when this timer is
selected.
Validated on frdm_mcxe31b: tests/kernel/timer/timer_api passes 15/15.
Signed-off-by: Holt Sun <holt.sun@nxp.com>
On STM32H74x/H75x, prepare_regulator_voltage_scale() selected voltage
scale 0 (VOS0) on every board. VOS0 there is entered via the overdrive
bit SYSCFG_PWRCR.ODEN and is only valid when Vcore is generated by the
LDO (H7 Data Sheets; stm32h7xx_hal_pwr.h). On a board whose pwr node
selects an SMPS-direct/bypass supply (LDO off), setting ODEN drives the
regulator out of spec: ~1.5 V Vcore measured on an STM32H755, above the
1.40 V absolute maximum, risking permanent damage.
This was latent until the VOS0/ODEN fix in v4.4.0 (PR #104749): before
that the ODEN write was silently dropped and the part stayed at VOS1.
On H74x/H75x, select VOS0 only for the supply modes where the LDO
generates Vcore (ldo, smps-ldo, smps-ext-ldo) and use VOS1 otherwise;
enable the SYSCFG clock first so the ODEN write takes effect (on a cold
boot it is otherwise gated until later in SoC init). Other H7 lines do
not use ODEN for VOS0, so their behavior is unchanged.
Signed-off-by: Thomas Peter <thomas.peter@volpi-group.com>
Replace wifi_get_country_code() with wlan_get_country_code() in the
shim driver. The shim driver should call WLAN layer APIs (wlan.h)
rather than internal wifi driver APIs (wifi.h) directly.
Assisted-by: WChat:Claude
Signed-off-by: Maochen Wang <maochen.wang@nxp.com>
Add support for CONFIG_WIFI_MGMT_RAW_SCAN_RESULTS in the NXP WiFi
driver. When enabled, the driver emits NET_EVENT_WIFI_RAW_SCAN_RESULT
events containing reconstructed 802.11 beacon frames with raw IEs
from each scanned BSS.
This allows applications to access information not available in the
standard wifi_scan_result struct, such as country code, RSN IE, and
mobility domain.
When CONFIG_WIFI_MGMT_RAW_SCAN_RESULTS_ONLY is set, normal parsed
scan results are suppressed.
The raw frame data includes a reconstructed 802.11 MAC header (24B)
and fixed fields (12B) prepended to the IEs, making it compatible
with the existing wifi shell raw scan result handler.
Assisted-by: WChat:Claude
Signed-off-by: Maochen Wang <maochen.wang@nxp.com>
Add support for scanning PCIe buses beyond bus 0 using a devicetree
property. Some platforms have PCIe devices on
non-zero buses that are not reachable through bridge
enumeration from bus 0.
Signed-off-by: Bindu S <bindu.s@intel.com>
This adds a driver for the ARM CMSDK APB timer as a system clock.
There exists a similar driver for the ARM CMSDK APB timer as a counter.
Signed-off-by: Mike J. Chen <mjchen@google.com>
Co-authored-by: Wojciech Sipak <wsipak@antmicro.com>
Signed-off-by: Wojciech Sipak <wsipak@antmicro.com>
The changes are required to allow using a system timer driver
for the same device with the same compatible string.
The counter driver is used for each instance that is not used as
the chosen system timer.
Signed-off-by: Wojciech Sipak <wsipak@antmicro.com>
MXCRYPTOLITE mfd driver intended to supported a crypto driver and
a trng (entropy) driver.
Assisted-by: Claude:claude-opus-4.8
Signed-off-by: McAtee Maxwell <maxwell.mcatee@infineon.com>
FUEL_GAUGE_CYCLE_COUNT was documented and treated inconsistently:
Drivers had inconsistently been reporting full charge/discharge cycles or
"1/100ths" of a cycle in the ``FUEL_GAUGE_CYCLE_COUNT`` property.
The property now consistently reports full cycles, and drivers that
previously reported fractions of a cycle (i.e. ADP5360 and BQ27Z746) have
been updated to report full cycles instead.
Applications that relied on the old behavior should be updated.
Signed-off-by: Florian Moesch <florian.moesch@draeger.com>
Replace nested MIN/MAX macro calls with the min3/max3 APIs which are
safer (evaluate arguments only once) and cleaner.
Assisted-by: Gemini:gemini-3.1-pro
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
With sys_clock_set_timeout() and sys_clock_announce() now taking an
unsigned tick count, the per-driver clamps and casts that assumed a
signed argument are no longer right:
- The (int32_t) cast on the CLAMP upper bound (MAX_TICKS and friends)
is unnecessary now that ticks is unsigned; drop it.
- CLAMP(ticks - 1, 0, MAX) underflows when ticks == 0: the unsigned
ticks - 1 wraps to UINT32_MAX and clamps to MAX_TICKS, turning an
'announce ASAP' request into a maximum sleep. Use
CLAMP(ticks, 1, MAX) - 1, which floors at 1 before subtracting so
ticks 0 and 1 both map to 0 with no underflow. stm32_lptim keeps its
minimum of one scheduled tick via CLAMP(ticks, 2, base) - 1. The
renesas_rx and renesas_ra_ulpt MAX_TICKS can be 0 (a short counter
relative to CYCLES_PER_TICK), which would invert that clamp range, so
they keep the original guarded subtract-then-cap, to be revisited.
- CLAMP(ticks, 0, MAX) for unsigned ticks has a pointless lower bound;
use MIN(ticks, MAX).
- Drop the now-redundant (int32_t) casts on the sys_clock_announce()
arguments, and make stm32wb0's dticks unsigned to match.
The recurring 'ticks - 1' is most likely the sign of a suboptimal
computation, but that will be dealt with elsewhere.
No functional change beyond the ticks == 0 underflow fix.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
The HPET main counter and comparator are handled as 64-bit, so the only
thing HPET_MAX_TICKS (== INT32_MAX) and the /2 headroom ever bounded was
the old signed sys_clock_announce() range plus IRQ-latency room. The kernel
now caps the requested timeout (SYS_CLOCK_MAX_WAIT) and keeps the announced
tick count in range, providing both, so the driver clamp is redundant;
remove it.
Convert the deep-idle stop (clearing GCONF_ENABLE) to the
IS_ENABLED(CONFIG_SYSTEM_CLOCK_SLOPPY_IDLE) && ticks == SYS_CLOCK_MAX_WAIT
sentinel, matching the other timer drivers, and drop the dead
K_TICKS_FOREVER mapping. No driver now references K_TICKS_FOREVER.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
These drivers clamped the programmed deadline against the smaller of two
limits: the sys_clock_announce() range (INT32_MAX ticks) and the
cycle_diff_t width used for the cycle-to-tick division. The kernel now caps
the requested timeout in ticks (SYS_CLOCK_MAX_WAIT) and keeps the announced
tick count in range, so the announce-range limit is obsolete.
Drop it (CYCLES_MAX_1 and the MIN), leaving only the cycle-width limit. The
1/4 headroom is kept: it covers the *cycle*-domain late report (a delayed
ISR must still see a cycle delta that fits in cycle_diff_t before the
cast), which the kernel's tick-domain slack does not address. With the
announce range gone, CYCLES_MAX no longer depends on CYC_PER_TICK and
becomes a compile-time constant, so the runtime precompute is removed too.
Also drop the now-dead K_TICKS_FOREVER branches (the kernel never passes
it).
No functional change for the timeouts the kernel actually requests.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
These drivers stop or disable the timer when the kernel has no pending
timeout, which is only correct when CONFIG_SYSTEM_CLOCK_SLOPPY_IDLE lets
the uptime skew during idle. That used to be implicit: the kernel only
passed K_TICKS_FOREVER under sloppy idle. The core now always passes a
finite value and signals "no pending deadline" with ticks ==
SYS_CLOCK_MAX_WAIT regardless of the option, so the gate must be made
explicit or a non-sloppy build would stop the timer and lose time.
Test IS_ENABLED(CONFIG_SYSTEM_CLOCK_SLOPPY_IDLE) && ticks ==
SYS_CLOCK_MAX_WAIT before stopping, matching the ite/stm32 drivers already
converted, and drop the now-redundant idle argument from the condition. The
normal path re-enables the timer on the next sys_clock_set_timeout(), so
the stop is safe. nrf_rtc only re-keys its "no pending timeout" far-compare
and sys_busy flag to the sentinel; it does not stop the timer.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
The kernel no longer passes K_TICKS_FOREVER to sys_clock_set_timeout(), so
the guards that skipped reprogramming (or whose inverted form gated the
normal path) on it are dead.
Unwrap the "program only when not forever" guards (cc23x0_rtc,
cc23x0_systim) and the "idle and not forever" power-management hooks
(esp32, xtensa) since the condition is now always satisfied. Drop the
idle/forever skip-returns (max32_rv32, max32_wut, openrisc). renesas_rx_cmt
and renesas_ra_ulpt tested the forever sentinel alongside INT32_MAX (==
SYS_CLOCK_MAX_WAIT); keep the single SYS_CLOCK_MAX_WAIT "no near deadline"
skip and drop the dead disjunct.
No functional change.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
The kernel core no longer passes K_TICKS_FOREVER to
sys_clock_set_timeout(): next_timeout() always returns a finite value
capped at SYS_CLOCK_MAX_WAIT. The per-driver "map K_TICKS_FOREVER to the
hardware maximum" handling is therefore dead: the requested tick count is
already bounded by the kernel and each of these drivers re-clamps to its
own hardware limit anyway.
Remove the dead K_TICKS_FOREVER mapping from the drivers whose value is
still bounded by a following hardware clamp. mchp_sam_pit64b and
xlnx_psttc, where the ternary was also the only guard before a uint32 cycle
multiply, keep an explicit clamp to their hardware maximum. native_sim,
which has no hardware limit, simply drops the special case.
No functional change.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
sys_clock_announce() reports ticks elapsed since the previous announce. The
quantity that must stay in range is therefore the distance of the next
timeout from that announce, not its distance from "now" as next_timeout()
previously bounded. When an early timeout is repeatedly replaced by a later
one, announces are deferred and that distance can exceed the range even
though each set_timeout() delay was in range, which every driver has had to
defend against on its own.
Cap it in the core instead: SYS_CLOCK_MAX_WAIT becomes UINT32_MAX/2 (the
clamp, with the upper half of the range left as slack for a late announce)
and next_timeout() clamps the inter-announce distance to it. Drivers no
longer need to clamp for the announce range and only have to honour their
own cycle-count limits.
The core no longer emits K_TICKS_FOREVER; the reported delay is always
finite, saturating at SYS_CLOCK_MAX_WAIT. Drivers that stop the timer
entirely when idle now key off ticks == SYS_CLOCK_MAX_WAIT under
CONFIG_SYSTEM_CLOCK_SLOPPY_IDLE.
z_get_next_timeout_expiry() still reports to the idle and PM paths as a
signed int32_t. It keeps its K_TICKS_FOREVER default and adopts the
next_timeout() value only when that fits, so the unsigned cap can never
surface there as a negative number.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
The sys_clock_set_timeout(), sys_clock_announce() and
sys_clock_announce_locked() interfaces carry a number of ticks to be
scheduled or announced. Those ticks have no negative meaning, so a signed
argument is both wasteful and error prone: it invites incorrect handling in
drivers and it halves the range that sys_clock_announce() can represent.
Switch the tick argument of these interfaces, along with the internal
announce_remaining accounting, from int32_t to uint32_t. This is a
mechanical change: driver bodies perform plain arithmetic on the value and
are unaffected by the signedness, and the kernel never passes a negative
tick count. The freed sign bit doubles the representable announce range,
which a later change will use to move the announce range limit out of the
drivers and into the core.
One config is not strictly neutral: under CONFIG_TIMEOUT_64BIT,
K_TICKS_FOREVER is a 64-bit value that an unsigned 32-bit argument can no
longer compare equal to, so the K_TICKS_FOREVER tests in drivers stop
matching there. K_TICKS_FOREVER is a k_ticks_t concept and has no business
in this tick count interface; a follow-up removes it from the driver side
entirely, which closes this gap.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Rename HL78XX network operator fields to avoid using operator as a
member name.
Use operator_info for the cached network operator structure and
operator_name for the operator string. This makes the field names clearer
and avoids ambiguity with the C++ operator keyword when the public modem
API header is included from C++ code.
Update COPS parsing and HL78XX API helpers to use the renamed fields.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Reset the receive staging ring when clearing receive transaction state.
Use the captured receive length as the actual socket read length instead
of assuming the requested modem length was fully staged. This allows
short reads to complete without treating them as fatal data mismatches,
while still rejecting invalid, empty, or over-captured transactions.
Return the actual copied length to the socket layer and update the modem
packet queue by the same amount. Also reset receive transaction state
when invalid lengths or ring buffer mismatches are detected so stale
staged data cannot affect the next read.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Remove obsolete helpers, duplicate definitions, and now-unused code
left after the HL78xx structure and API refactoring.
Also apply small formatting fixes where declarations or constants were
touched by the refactor.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Reorganize HL78xx socket runtime data into logical sub-structures.
Group socket control, receive, parser, network, synchronization,
connection, and device state instead of keeping all socket fields at
the top level. Update socket handling code to use the new access paths.
This reduces the size of the flat socket data structure and makes
ownership of parser, receive, connection, and device state clearer.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Update HL78xx GNSS integration to use the grouped device and status
state layout.
Replace direct GNSS device pointer accesses with data->devices.gnss
and update low power status references to the new grouped state. This
keeps GNSS handling consistent with the rest of the driver data layout.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Replace hardcoded modem response strings in chat match definitions with
named HL78xx response constants.
Use shared constants for OK, ERROR, CME ERROR, CMS ERROR, CONNECT, and
related modem responses. This avoids duplicate string literals and
keeps chat handling aligned with the rest of the driver.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Split the network information query implementation into smaller helper
functions.
The previous implementation handled all network information types in a
single large switch body. Move the individual APN, RAT, operator, TAC,
MCC, MNC, cell ID, IP address, DNS, active band, and SINR handling into
dedicated helpers.
This reduces function complexity and makes each network information
path easier to review and maintain.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Simplify the HL78xx AT command API by replacing the variadic printf
and scanf style helpers with functions that accept pre-formatted AT
command strings.
The caller is now responsible for formatting commands before passing
them to the driver. This removes local formatting logic, drops the
stdarg dependency from the public header, and reduces the complexity of
the command execution path.
Expose the new public helpers and update the internal implementation to
use an explicit command context. Also move common modem response
strings into named constants so they can be shared by chat, socket, and
AT command handling.
No functional behavior is changed for already formatted commands.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Move PSM, eDRX, power-down, and low power restore state under the
status.lpm sub-structure.
This keeps low power related state together and makes the relationship
between low power events, timers, and state transitions clearer. Update
all users to the new access paths.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Move related modem status fields into dedicated sub-structures.
Group signal, band, restart, and KCELLMEAS state instead of keeping
these fields directly in struct modem_status. This reduces the size of
the flat status namespace and makes future status updates easier to
review.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Move delayable work items and semaphores into dedicated sub-structures
in the HL78xx driver data.
This reduces the number of top-level fields in struct hl78xx_data and
makes the purpose of each synchronization or work item easier to
understand from its access path.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Move HL78xx related device pointers under the devices sub-structure.
Replace direct uses of the modem, GNSS, and offload device pointers
with the corresponding data->devices members. This keeps device
references grouped with the rest of the driver runtime state and
prepares the code for further structure cleanup.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Introduce grouped state structures for the HL78xx driver to reduce
the number of flat fields in the main driver data and status objects.
Add dedicated containers for signal, band, restart, KCELLMEAS, low
power, work, semaphore, device, and runtime band state. This prepares
the driver for follow-up mechanical moves and makes ownership of each
piece of state clearer.
No functional behavior is changed.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Fix CESQ_RSRP and CESQ_RSRQ conversion macros in hl78xx_gnss_parsers.c:
the previous formulas were shifted by one RSRP/RSRQ step, producing
off-by-one signal values.
Add HL78XX_CESQ_RSRP_BASE, HL78XX_CESQ_RSRP_STEP, and equivalent
RSRQ constants to hl78xx_apis.h so callers can apply the same
conversion without duplicating magic numbers.
Add the AT error-code encoding constants HL78XX_MODEM_AT_ERROR,
HL78XX_MODEM_AT_CME_ERROR, and HL78XX_MODEM_AT_CMS_ERROR alongside
the hl78xx_modem_at_printf(), hl78xx_modem_at_scanf(), and
hl78xx_modem_at_cmd() public inline wrappers in hl78xx_apis.h.
Add hl78xx_sim_slot, hl78xx_modem_restart_mode, and
hl78xx_power_down_response enums to hl78xx_apis.h. Add
HL78XX_VGPIO_LOW, HL78XX_VGPIO_HIGH, HL78XX_GPIO6_LOW, and
HL78XX_GPIO6_HIGH event types. Add public declarations for
hl78xx_restart_modem(), hl78xx_set_wake_pin_low(),
hl78xx_set_wake_pin_high(), hl78xx_set_active_sim(), and the
hl78xx_power_down_trigger/cancel/respond/confirm() API.
Add HL78XX_RAT_MODE_NONE initializer constant to hl78xx_apis.h and
use it in the hello_hl78xx sample to make the initial RAT mode
explicit rather than relying on zero-initialisation.
Add HL78XX_LOG_HEXDUMP_DBG macro to hl78xx.h for optional verbose
hex dump logging.
Change the post-restart AT command timeout in hl78xx_chat.h from a
short constant to HL78XX_CMD_TIMEOUT_LONG so commands sent
immediately after modem restart have adequate time to complete.
Guard the KEDRXCFG command in the PMC init script with
CONFIG_MODEM_HL78XX_RAT_NBNTN so the NB-NTN eDRX parameter is only
applied on builds that include NTN support.
Add MODEM_HL78XX_AT_SHELL depends on SHELL to Kconfig.hl78xx to
prevent a Kconfig warning when the shell subsystem is disabled.
Port hl78xx_hl7800.c to the updated signal-quality API: replace
direct field access with the new CESQ helper macros.
Port hl78xx_hl7800.c signal-quality path and hl78xx_gnss_parsers.c
CESQ parsing to the corrected RSRP/RSRQ formulas.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Add hl78xx_at_cmd_capture_ctx to track the terminal result (OK,
ERROR, CME ERROR, CMS ERROR) and any captured response text for a
dynamic command in flight. Allocate hl78xx_at_cmd_terminal_result
and hl78xx_at_cmd_error_type enums in hl78xx.h alongside the context
struct. Add hl78xx_at_cmd_capture_ctx at_cmd_capture and
hl78xx_at_cmd_capture_ctx at_cmd_capture fields to hl78xx_data.
Add hl78xx_dynamic_cmd_prepare_matches() to normalise a caller-
supplied match array into data->buffers.dynamic_matches before each
command, injecting default ok/error/cme/cms callbacks where the
caller left callback = NULL. Wire it into
hl78xx_dynamic_cmd_prepare_script() and switch
modem_chat_script_chat_set_response_matches() to use the normalised
pointers. Introduce hl78xx_dynamic_cmd_reset_terminal_result() and
call it at the top of hl78xx_dynamic_cmd_send_impl().
Change hl78xx_dynamic_cmd_send_impl() tx_lock acquisition to use a
calculated k_timeout_t (response_timeout or HL78XX_CMD_TIMEOUT_SHORT
for synchronous calls; K_NO_WAIT for async). Log a warning with the
command text when the mutex times out. Propagate the exact errno
through all early-exit paths instead of returning -1. After a
successful script run, check terminal_result and return -EIO if the
modem replied with an error response.
Add hl78xx_at_cmd_execute() as the single internal executor used by
both hl78xx_send_cmd() (no capture) and the new public AT helpers.
Add hl78xx_modem_at_exec(), hl78xx_modem_at_printf_va(),
hl78xx_modem_at_cmd_va(), and hl78xx_modem_at_scanf_va() in
hl78xx_apis.c as the implementations backing the inlines in
hl78xx_apis.h. Add hl78xx_modem_at_encode_error() to translate a
capture error type + code into the packed HL78XX_MODEM_AT_* encoding.
Add struct kband_syntax active_band to modem_status and
hl78xx_on_kbnd() to populate it from AT+KBND? URC responses.
Add hl78xx_runtime_band_provider_t, runtime_band_provider, and
runtime_band_provider_user_data to hl78xx_data. Add
hl78xx_get_runtime_band_override(), hl78xx_band_bitmap_from_single_
band(), and hl78xx_get_expected_band_config_for_rat() to
hl78xx_cfg.c. Switch hl78xx_band_cfg() to call
hl78xx_get_expected_band_config_for_rat() instead of
hl78xx_get_band_default_config_for_rat() so a runtime provider can
override the configured band at runtime.
Fix hl78xx_api_func_modem_dynamic_cmd_send() null-pointer check and
hl78xx_api_func_get_modem_info_standard() argument validation to
guard dev and dev->data before dereferencing.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
Add a two-stage power-down flow. hl78xx_power_down_work_handler()
notifies the application via HL78XX_POWER_DOWN_UPDATE and arms a
confirmation timer (CONFIG_MODEM_HL78XX_POWER_DOWN_CONFIRM_TIMEOUT_S).
hl78xx_power_down_shutdown_fn() fires when the timer expires and enters
MODEM_HL78XX_STATE_INIT_POWER_OFF. Add hl78xx_power_down_trigger(),
hl78xx_power_down_cancel(), hl78xx_power_down_respond(), and
hl78xx_power_down_confirm() as the public API surface. Guard
hl78xx_power_down_feed_timer() against calls when feeding is
suppressed or no timer is pending.
Add hl78xx_suspend_uart() and hl78xx_resume_uart() helpers. Use them
in the sleep BUS_CLOSED handler, the sleep DEVICE_AWAKE handler, the
idle BUS_CLOSED handler, and the idle SUSPEND handler. Move
k_sem_give() from hl78xx_on_idle_state_enter() to
hl78xx_idle_event_handler(BUS_CLOSED) so the semaphore is released
only after the bus is closed and the UART is suspended. Add
k_sem_take(K_NO_WAIT) before delegating SUSPEND in
hl78xx_driver_pm_action() to drain any stale count.
Add MODEM_HL78XX_STATE_SOFT_RESET with enter/leave/event handlers.
hl78xx_on_soft_reset_state_enter() sends AT+CFUN=4,1 and waits for
MDM_RESTART; on timeout it falls back via
hl78xx_enter_restart_fallback_state(). Add
MODEM_HL78XX_EVENT_RESTART_REQUESTED and
hl78xx_handle_restart_requested_event() as a pre-dispatch global
handler that routes to SOFT_RESET or RESET_PULSE based on restart
mode. Add hl78xx_api_func_restart() / hl78xx_restart_modem() as the
public restart API. Reset at_cmd_ready_sent in the restart path so
AT_CMD_READY fires once per boot cycle.
Add VGPIO debounce: hl78xx_dispatch_vgpio_state(),
hl78xx_vgpio_debounce_work_handler(), and k_work_init_delayable() in
hl78xx_init(). ISR now schedules the debounce work when
vgpio_debounce_ms > 0; direct dispatch otherwise. Set
HL7812_VGPIO_DEBOUNCE_MS = 300 and assign vgpio_debounce_ms in
hl78xx_variant_ops_hl7812.
Add HL78XX_VGPIO_LOW / HL78XX_VGPIO_HIGH event types.
Add unconditional HL78XX_GPIO6_LOW and HL78XX_GPIO6_HIGH dispatches
in hl78xx_hl7812_gpio6_handle_low/high(). Guard gpio6_handle_low()
against transitions during a config restart. Simplify power-down
state tracking: drop the conditional POWER_DOWN_UPDATE dispatch from
GPIO6 ISR; the new two-stage flow takes that responsibility.
Call hl78xx_invalidate_socket_contexts() from
hl78xx_hl7812_on_ksup_lpm() on unexpected restart.
Guard hl78xx_on_run_init_script_state_enter() and
hl78xx_on_soft_reset_state_enter() to assert the WAKE pin high before
running scripts. Add hl78xx_on_await_power_on_state_enter() call to
hl78xx_power_down_allow_feeding() and resume_uart on TIMEOUT.
Manage config_restart_pending across rat_cfg and pmc_cfg states so
the TIMEOUT handler waits for MDM_RESTART rather than triggering a
second restart. Add MDM_RESTART handling in pmc_cfg to enter
RUN_INIT_SCRIPT directly.
Add hl78xx_set_wake_pin_low(), hl78xx_set_wake_pin_high(),
hl78xx_set_active_sim(), and hl78xx_power_down_trigger/cancel/
respond/confirm() to the public API.
Fix mdm_gpio_sim_switch DT macro: use mdm_sim_switch_gpios instead
of mdm_sim_select_gpios. Fix sim_switch GPIO direction: GPIO_OUTPUT
not GPIO_INPUT.
Add hl78xx_sim_slot, hl78xx_modem_restart_mode, and
hl78xx_power_down_response enums plus VGPIO/GPIO6 event types to
hl78xx_apis.h.
Signed-off-by: Zafer SEN <zafersn93@gmail.com>
