ardupilot/libraries/AP_HAL_ChibiOS/Scheduler.cpp

831 lines
24 KiB
C++

/*
* This file is free software: you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This file is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
* See the GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License along
* with this program. If not, see <http://www.gnu.org/licenses/>.
*
* Code by Andrew Tridgell and Siddharth Bharat Purohit
*/
#include <AP_HAL/AP_HAL_Boards.h>
#ifndef HAL_SCHEDULER_ENABLED
#define HAL_SCHEDULER_ENABLED 1
#endif
#if HAL_SCHEDULER_ENABLED
#include <AP_HAL/AP_HAL.h>
#include <hal.h>
#include "AP_HAL_ChibiOS.h"
#include "Scheduler.h"
#include "Util.h"
#include "GPIO.h"
#include <AP_HAL_ChibiOS/UARTDriver.h>
#include <AP_HAL_ChibiOS/AnalogIn.h>
#include <AP_HAL_ChibiOS/Storage.h>
#include <AP_HAL_ChibiOS/RCOutput.h>
#include <AP_HAL_ChibiOS/RCInput.h>
#include <AP_HAL_ChibiOS/CANIface.h>
#include <AP_InternalError/AP_InternalError.h>
#if CH_CFG_USE_DYNAMIC == TRUE
#include <AP_Logger/AP_Logger.h>
#include <AP_Scheduler/AP_Scheduler.h>
#include <AP_BoardConfig/AP_BoardConfig.h>
#include "hwdef/common/stm32_util.h"
#include "hwdef/common/flash.h"
#include "hwdef/common/watchdog.h"
#include <AP_Filesystem/AP_Filesystem.h>
#include "shared_dma.h"
#include <AP_Common/ExpandingString.h>
#if HAL_WITH_IO_MCU
#include <AP_IOMCU/AP_IOMCU.h>
extern AP_IOMCU iomcu;
#endif
using namespace ChibiOS;
extern const AP_HAL::HAL& hal;
#ifndef HAL_NO_TIMER_THREAD
THD_WORKING_AREA(_timer_thread_wa, TIMER_THD_WA_SIZE);
#endif
#ifndef HAL_NO_RCOUT_THREAD
THD_WORKING_AREA(_rcout_thread_wa, RCOUT_THD_WA_SIZE);
#endif
#ifndef HAL_NO_RCIN_THREAD
THD_WORKING_AREA(_rcin_thread_wa, RCIN_THD_WA_SIZE);
#endif
#ifndef HAL_USE_EMPTY_IO
THD_WORKING_AREA(_io_thread_wa, IO_THD_WA_SIZE);
#endif
#ifndef HAL_USE_EMPTY_STORAGE
THD_WORKING_AREA(_storage_thread_wa, STORAGE_THD_WA_SIZE);
#endif
#ifndef HAL_NO_MONITOR_THREAD
THD_WORKING_AREA(_monitor_thread_wa, MONITOR_THD_WA_SIZE);
#endif
Scheduler::Scheduler()
{
}
void Scheduler::init()
{
chBSemObjectInit(&_timer_semaphore, false);
chBSemObjectInit(&_io_semaphore, false);
#ifndef HAL_NO_MONITOR_THREAD
// setup the monitor thread - this is used to detect software lockups
_monitor_thread_ctx = chThdCreateStatic(_monitor_thread_wa,
sizeof(_monitor_thread_wa),
APM_MONITOR_PRIORITY, /* Initial priority. */
_monitor_thread, /* Thread function. */
this); /* Thread parameter. */
#endif
#ifndef HAL_NO_TIMER_THREAD
// setup the timer thread - this will call tasks at 1kHz
_timer_thread_ctx = chThdCreateStatic(_timer_thread_wa,
sizeof(_timer_thread_wa),
APM_TIMER_PRIORITY, /* Initial priority. */
_timer_thread, /* Thread function. */
this); /* Thread parameter. */
#endif
#ifndef HAL_NO_RCOUT_THREAD
// setup the RCOUT thread - this will call tasks at 1kHz
_rcout_thread_ctx = chThdCreateStatic(_rcout_thread_wa,
sizeof(_rcout_thread_wa),
APM_RCOUT_PRIORITY, /* Initial priority. */
_rcout_thread, /* Thread function. */
this); /* Thread parameter. */
#endif
#ifndef HAL_NO_RCIN_THREAD
// setup the RCIN thread - this will call tasks at 1kHz
_rcin_thread_ctx = chThdCreateStatic(_rcin_thread_wa,
sizeof(_rcin_thread_wa),
APM_RCIN_PRIORITY, /* Initial priority. */
_rcin_thread, /* Thread function. */
this); /* Thread parameter. */
#endif
#ifndef HAL_USE_EMPTY_IO
// the IO thread runs at lower priority
_io_thread_ctx = chThdCreateStatic(_io_thread_wa,
sizeof(_io_thread_wa),
APM_IO_PRIORITY, /* Initial priority. */
_io_thread, /* Thread function. */
this); /* Thread parameter. */
#endif
#ifndef HAL_USE_EMPTY_STORAGE
// the storage thread runs at just above IO priority
_storage_thread_ctx = chThdCreateStatic(_storage_thread_wa,
sizeof(_storage_thread_wa),
APM_STORAGE_PRIORITY, /* Initial priority. */
_storage_thread, /* Thread function. */
this); /* Thread parameter. */
#endif
}
void Scheduler::delay_microseconds(uint16_t usec)
{
if (usec == 0) { //chibios faults with 0us sleep
return;
}
uint32_t ticks;
ticks = chTimeUS2I(usec);
if (ticks == 0) {
// calling with ticks == 0 causes a hard fault on ChibiOS
ticks = 1;
}
ticks = MIN(TIME_MAX_INTERVAL, ticks);
chThdSleep(MAX(ticks,CH_CFG_ST_TIMEDELTA)); //Suspends Thread for desired microseconds
}
/*
wrapper around sem_post that boosts main thread priority
*/
static void set_high_priority()
{
#if APM_MAIN_PRIORITY_BOOST != APM_MAIN_PRIORITY
hal_chibios_set_priority(APM_MAIN_PRIORITY_BOOST);
#endif
}
/*
return the main thread to normal priority
*/
void Scheduler::boost_end(void)
{
#if APM_MAIN_PRIORITY_BOOST != APM_MAIN_PRIORITY
if (in_main_thread() && _priority_boosted) {
_priority_boosted = false;
hal_chibios_set_priority(APM_MAIN_PRIORITY);
}
#endif
}
/*
a variant of delay_microseconds that boosts priority to
APM_MAIN_PRIORITY_BOOST for APM_MAIN_PRIORITY_BOOST_USEC
microseconds when the time completes. This significantly improves
the regularity of timing of the main loop
*/
void Scheduler::delay_microseconds_boost(uint16_t usec)
{
if (!_priority_boosted && in_main_thread()) {
set_high_priority();
_priority_boosted = true;
_called_boost = true;
}
delay_microseconds(usec); //Suspends Thread for desired microseconds
}
/*
return true if delay_microseconds_boost() has been called since last check
*/
bool Scheduler::check_called_boost(void)
{
if (!_called_boost) {
return false;
}
_called_boost = false;
return true;
}
void Scheduler::delay(uint16_t ms)
{
uint64_t start = AP_HAL::micros64();
while ((AP_HAL::micros64() - start)/1000 < ms) {
delay_microseconds(1000);
if (_min_delay_cb_ms <= ms) {
if (in_main_thread()) {
call_delay_cb();
}
}
}
}
void Scheduler::register_timer_process(AP_HAL::MemberProc proc)
{
chBSemWait(&_timer_semaphore);
for (uint8_t i = 0; i < _num_timer_procs; i++) {
if (_timer_proc[i] == proc) {
chBSemSignal(&_timer_semaphore);
return;
}
}
if (_num_timer_procs < CHIBIOS_SCHEDULER_MAX_TIMER_PROCS) {
_timer_proc[_num_timer_procs] = proc;
_num_timer_procs++;
} else {
DEV_PRINTF("Out of timer processes\n");
}
chBSemSignal(&_timer_semaphore);
}
void Scheduler::register_io_process(AP_HAL::MemberProc proc)
{
chBSemWait(&_io_semaphore);
for (uint8_t i = 0; i < _num_io_procs; i++) {
if (_io_proc[i] == proc) {
chBSemSignal(&_io_semaphore);
return;
}
}
if (_num_io_procs < CHIBIOS_SCHEDULER_MAX_TIMER_PROCS) {
_io_proc[_num_io_procs] = proc;
_num_io_procs++;
} else {
DEV_PRINTF("Out of IO processes\n");
}
chBSemSignal(&_io_semaphore);
}
void Scheduler::register_timer_failsafe(AP_HAL::Proc failsafe, uint32_t period_us)
{
_failsafe = failsafe;
}
void Scheduler::reboot(bool hold_in_bootloader)
{
// disarm motors to ensure they are off during a bootloader upload
hal.rcout->force_safety_on();
#if HAL_WITH_IO_MCU
if (AP_BoardConfig::io_enabled()) {
iomcu.shutdown();
}
#endif
#if HAL_LOGGING_ENABLED
//stop logging
if (AP_Logger::get_singleton()) {
AP::logger().StopLogging();
}
// unmount filesystem, if active
AP::FS().unmount();
#endif
#if AP_FASTBOOT_ENABLED
// setup RTC for fast reboot
set_fast_reboot(hold_in_bootloader?RTC_BOOT_HOLD:RTC_BOOT_FAST);
#endif
// disable all interrupt sources
port_disable();
// reboot
NVIC_SystemReset();
}
void Scheduler::_run_timers()
{
if (_in_timer_proc) {
return;
}
_in_timer_proc = true;
int num_procs = 0;
chBSemWait(&_timer_semaphore);
num_procs = _num_timer_procs;
chBSemSignal(&_timer_semaphore);
// now call the timer based drivers
for (int i = 0; i < num_procs; i++) {
if (_timer_proc[i]) {
_timer_proc[i]();
}
}
// and the failsafe, if one is setup
if (_failsafe != nullptr) {
_failsafe();
}
#if HAL_USE_ADC == TRUE && !defined(HAL_DISABLE_ADC_DRIVER)
// process analog input
((AnalogIn *)hal.analogin)->_timer_tick();
#endif
_in_timer_proc = false;
}
void Scheduler::_timer_thread(void *arg)
{
Scheduler *sched = (Scheduler *)arg;
chRegSetThreadName("timer");
while (!sched->_hal_initialized) {
sched->delay_microseconds(1000);
}
while (true) {
sched->delay_microseconds(1000);
// run registered timers
sched->_run_timers();
if (sched->in_expected_delay()) {
sched->watchdog_pat();
}
}
}
void Scheduler::_rcout_thread(void *arg)
{
#ifndef HAL_NO_RCOUT_THREAD
Scheduler *sched = (Scheduler *)arg;
chRegSetThreadName("rcout");
while (!sched->_hal_initialized) {
sched->delay_microseconds(1000);
}
#if HAL_USE_PWM == TRUE
// trampoline into the rcout thread
((RCOutput*)hal.rcout)->rcout_thread();
#endif
#endif
}
/*
return true if we are in a period of expected delay. This can be
used to suppress error messages
*/
bool Scheduler::in_expected_delay(void) const
{
if (!_initialized) {
// until setup() is complete we expect delays
return true;
}
if (expect_delay_start != 0) {
uint32_t now = AP_HAL::millis();
if (now - expect_delay_start <= expect_delay_length) {
return true;
}
}
#if !defined(HAL_NO_FLASH_SUPPORT) && !defined(HAL_BOOTLOADER_BUILD)
if (stm32_flash_recent_erase()) {
return true;
}
#endif
return false;
}
#ifndef HAL_NO_MONITOR_THREAD
void Scheduler::_monitor_thread(void *arg)
{
Scheduler *sched = (Scheduler *)arg;
chRegSetThreadName("monitor");
while (!sched->_initialized) {
sched->delay(100);
}
bool using_watchdog = AP_BoardConfig::watchdog_enabled();
#if HAL_LOGGING_ENABLED
uint8_t log_wd_counter = 0;
#endif
while (true) {
sched->delay(100);
if (using_watchdog) {
stm32_watchdog_save((uint32_t *)&hal.util->persistent_data, (sizeof(hal.util->persistent_data)+3)/4);
}
// if running memory guard then check all allocations
malloc_check(nullptr);
uint32_t now = AP_HAL::millis();
uint32_t loop_delay = now - sched->last_watchdog_pat_ms;
if (loop_delay >= 200) {
// the main loop has been stuck for at least
// 200ms. Starting logging the main loop state
#if HAL_LOGGING_ENABLED
const AP_HAL::Util::PersistentData &pd = hal.util->persistent_data;
if (AP_Logger::get_singleton()) {
const struct log_MON mon{
LOG_PACKET_HEADER_INIT(LOG_MON_MSG),
time_us : AP_HAL::micros64(),
loop_delay : loop_delay,
current_task : pd.scheduler_task,
internal_error_mask : pd.internal_errors,
internal_error_count : pd.internal_error_count,
internal_error_line : pd.internal_error_last_line,
mavmsg : pd.last_mavlink_msgid,
mavcmd : pd.last_mavlink_cmd,
semline : pd.semaphore_line,
spicnt : pd.spi_count,
i2ccnt : pd.i2c_count
};
AP::logger().WriteCriticalBlock(&mon, sizeof(mon));
}
#endif
}
if (loop_delay >= 500 && !sched->in_expected_delay()) {
// at 500ms we declare an internal error
INTERNAL_ERROR(AP_InternalError::error_t::main_loop_stuck);
}
#if AP_CRASHDUMP_ENABLED
if (loop_delay >= 1800 && using_watchdog) {
// we are about to watchdog, better to trigger a hardfault
// now and get a crash dump file
void *ptr = (void*)0xE000FFFF;
typedef void (*fptr)();
fptr gptr = (fptr) (void *)ptr;
gptr();
}
#endif
#if HAL_LOGGING_ENABLED
if (log_wd_counter++ == 10 && hal.util->was_watchdog_reset()) {
log_wd_counter = 0;
// log watchdog message once a second
const AP_HAL::Util::PersistentData &pd = hal.util->last_persistent_data;
struct log_WDOG wdog{
LOG_PACKET_HEADER_INIT(LOG_WDOG_MSG),
time_us : AP_HAL::micros64(),
scheduler_task : pd.scheduler_task,
internal_errors : pd.internal_errors,
internal_error_count : pd.internal_error_count,
internal_error_last_line : pd.internal_error_last_line,
last_mavlink_msgid : pd.last_mavlink_msgid,
last_mavlink_cmd : pd.last_mavlink_cmd,
semaphore_line : pd.semaphore_line,
fault_line : pd.fault_line,
fault_type : pd.fault_type,
fault_addr : pd.fault_addr,
fault_thd_prio : pd.fault_thd_prio,
fault_icsr : pd.fault_icsr,
fault_lr : pd.fault_lr
};
memcpy(wdog.thread_name4, pd.thread_name4, ARRAY_SIZE(wdog.thread_name4));
AP::logger().WriteCriticalBlock(&wdog, sizeof(wdog));
}
#endif // HAL_LOGGING_ENABLED
#ifndef IOMCU_FW
// setup GPIO interrupt quotas
hal.gpio->timer_tick();
#endif
}
}
#endif // HAL_NO_MONITOR_THREAD
void Scheduler::_rcin_thread(void *arg)
{
Scheduler *sched = (Scheduler *)arg;
chRegSetThreadName("rcin");
while (!sched->_hal_initialized) {
sched->delay_microseconds(20000);
}
while (true) {
sched->delay_microseconds(1000);
((RCInput *)hal.rcin)->_timer_tick();
}
}
void Scheduler::_run_io(void)
{
if (_in_io_proc) {
return;
}
_in_io_proc = true;
int num_procs = 0;
chBSemWait(&_io_semaphore);
num_procs = _num_io_procs;
chBSemSignal(&_io_semaphore);
// now call the IO based drivers
for (int i = 0; i < num_procs; i++) {
if (_io_proc[i]) {
_io_proc[i]();
}
}
_in_io_proc = false;
}
void Scheduler::_io_thread(void* arg)
{
Scheduler *sched = (Scheduler *)arg;
chRegSetThreadName("io");
while (!sched->_hal_initialized) {
sched->delay_microseconds(1000);
}
#if HAL_LOGGING_ENABLED
uint32_t last_sd_start_ms = AP_HAL::millis();
#endif
#if CH_DBG_ENABLE_STACK_CHECK == TRUE
uint32_t last_stack_check_ms = 0;
#endif
while (true) {
sched->delay_microseconds(1000);
// run registered IO processes
sched->_run_io();
#if HAL_LOGGING_ENABLED || CH_DBG_ENABLE_STACK_CHECK == TRUE
uint32_t now = AP_HAL::millis();
#endif
#if HAL_LOGGING_ENABLED
if (!hal.util->get_soft_armed()) {
// if sdcard hasn't mounted then retry it every 3s in the IO
// thread when disarmed
if (now - last_sd_start_ms > 3000) {
last_sd_start_ms = now;
AP::FS().retry_mount();
}
}
#endif
#if CH_DBG_ENABLE_STACK_CHECK == TRUE
if (now - last_stack_check_ms > 1000) {
last_stack_check_ms = now;
sched->check_stack_free();
}
#endif
}
}
#if defined(STM32H7)
/*
the H7 has 64k of ITCM memory at address zero. We reserve 1k of it
to prevent nullptr being valid. This function checks that memory is
always zero
*/
void Scheduler::check_low_memory_is_zero()
{
const uint32_t *lowmem = nullptr;
// we start at address 0x1 as reading address zero causes a fault
for (uint16_t i=1; i<256; i++) {
if (lowmem[i] != 0) {
// re-use memory guard internal error
AP_memory_guard_error(1023);
break;
}
}
// we can't do address 0, but can check next 3 bytes
const uint8_t *addr0 = (const uint8_t *)0;
for (uint8_t i=1; i<4; i++) {
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Warray-bounds"
if (addr0[i] != 0) {
AP_memory_guard_error(1023);
break;
}
#pragma GCC diagnostic pop
}
}
#endif // STM32H7
void Scheduler::_storage_thread(void* arg)
{
Scheduler *sched = (Scheduler *)arg;
chRegSetThreadName("storage");
while (!sched->_hal_initialized) {
sched->delay_microseconds(10000);
}
#if defined STM32H7
uint16_t memcheck_counter=0;
#endif
while (true) {
sched->delay_microseconds(1000);
// process any pending storage writes
hal.storage->_timer_tick();
#if defined STM32H7
if (memcheck_counter++ % 500 == 0) {
// run check at 2Hz
sched->check_low_memory_is_zero();
}
#endif
}
}
void Scheduler::set_system_initialized()
{
if (_initialized) {
AP_HAL::panic("PANIC: scheduler::set_system_initialized called"
"more than once");
}
_initialized = true;
}
/*
disable interrupts and return a context that can be used to
restore the interrupt state. This can be used to protect
critical regions
*/
void *Scheduler::disable_interrupts_save(void)
{
return (void *)(uintptr_t)chSysGetStatusAndLockX();
}
/*
restore interrupt state from disable_interrupts_save()
*/
void Scheduler::restore_interrupts(void *state)
{
chSysRestoreStatusX((syssts_t)(uintptr_t)state);
}
/*
trampoline for thread create
*/
void Scheduler::thread_create_trampoline(void *ctx)
{
AP_HAL::MemberProc *t = (AP_HAL::MemberProc *)ctx;
(*t)();
free(t);
}
// calculates an integer to be used as the priority for a newly-created thread
uint8_t Scheduler::calculate_thread_priority(priority_base base, int8_t priority) const
{
uint8_t thread_priority = APM_IO_PRIORITY;
static const struct {
priority_base base;
uint8_t p;
} priority_map[] = {
{ PRIORITY_BOOST, APM_MAIN_PRIORITY_BOOST},
{ PRIORITY_MAIN, APM_MAIN_PRIORITY},
{ PRIORITY_SPI, APM_SPI_PRIORITY},
{ PRIORITY_I2C, APM_I2C_PRIORITY},
{ PRIORITY_CAN, APM_CAN_PRIORITY},
{ PRIORITY_TIMER, APM_TIMER_PRIORITY},
{ PRIORITY_RCOUT, APM_RCOUT_PRIORITY},
{ PRIORITY_LED, APM_LED_PRIORITY},
{ PRIORITY_RCIN, APM_RCIN_PRIORITY},
{ PRIORITY_IO, APM_IO_PRIORITY},
{ PRIORITY_UART, APM_UART_PRIORITY},
{ PRIORITY_STORAGE, APM_STORAGE_PRIORITY},
{ PRIORITY_SCRIPTING, APM_SCRIPTING_PRIORITY},
};
for (uint8_t i=0; i<ARRAY_SIZE(priority_map); i++) {
if (priority_map[i].base == base) {
thread_priority = constrain_int16(priority_map[i].p + priority, LOWPRIO, HIGHPRIO);
break;
}
}
return thread_priority;
}
/*
create a new thread
*/
bool Scheduler::thread_create(AP_HAL::MemberProc proc, const char *name, uint32_t stack_size, priority_base base, int8_t priority)
{
// take a copy of the MemberProc, it is freed after thread exits
AP_HAL::MemberProc *tproc = (AP_HAL::MemberProc *)malloc(sizeof(proc));
if (!tproc) {
return false;
}
*tproc = proc;
const uint8_t thread_priority = calculate_thread_priority(base, priority);
thread_t *thread_ctx = thread_create_alloc(THD_WORKING_AREA_SIZE(stack_size),
name,
thread_priority,
thread_create_trampoline,
tproc);
if (thread_ctx == nullptr) {
free(tproc);
return false;
}
return true;
}
/*
inform the scheduler that we are calling an operation from the
main thread that may take an extended amount of time. This can
be used to prevent watchdog reset during expected long delays
A value of zero cancels the previous expected delay
*/
void Scheduler::_expect_delay_ms(uint32_t ms)
{
if (!in_main_thread()) {
// only for main thread
return;
}
// pat once immediately
watchdog_pat();
WITH_SEMAPHORE(expect_delay_sem);
if (ms == 0) {
if (expect_delay_nesting > 0) {
expect_delay_nesting--;
}
if (expect_delay_nesting == 0) {
expect_delay_start = 0;
}
} else {
uint32_t now = AP_HAL::millis();
if (expect_delay_start != 0) {
// we already have a delay running, possibly extend it
uint32_t done = now - expect_delay_start;
if (expect_delay_length > done) {
ms = MAX(ms, expect_delay_length - done);
}
}
expect_delay_start = now;
expect_delay_length = ms;
expect_delay_nesting++;
// also put our priority below timer thread if we are boosted
boost_end();
}
}
/*
this is _expect_delay_ms() with check that we are in the main thread
*/
void Scheduler::expect_delay_ms(uint32_t ms)
{
if (!in_main_thread()) {
// only for main thread
return;
}
_expect_delay_ms(ms);
}
// pat the watchdog
void Scheduler::watchdog_pat(void)
{
stm32_watchdog_pat();
last_watchdog_pat_ms = AP_HAL::millis();
#if defined(HAL_GPIO_PIN_EXT_WDOG)
ext_watchdog_pat(last_watchdog_pat_ms);
#endif
}
#if defined(HAL_GPIO_PIN_EXT_WDOG)
// toggle the external watchdog gpio pin
void Scheduler::ext_watchdog_pat(uint32_t now_ms)
{
// toggle watchdog GPIO every WDI_OUT_INTERVAL_TIME_MS
if ((now_ms - last_ext_watchdog_ms) >= EXT_WDOG_INTERVAL_MS) {
palToggleLine(HAL_GPIO_PIN_EXT_WDOG);
last_ext_watchdog_ms = now_ms;
}
}
#endif
#if CH_DBG_ENABLE_STACK_CHECK == TRUE
/*
check we have enough stack free on all threads and the ISR stack
*/
void Scheduler::check_stack_free(void)
{
// we raise an internal error stack_overflow when the available
// stack on any thread or the ISR stack drops below this
// threshold. This means we get an overflow error when we haven't
// yet completely run out of stack. This gives us a good
// pre-warning when we are getting too close
#if defined(STM32F1)
const uint32_t min_stack = 32;
#else
const uint32_t min_stack = 64;
#endif
if (stack_free(&__main_stack_base__) < min_stack) {
// use "line number" of 0xFFFF for ISR stack low
AP::internalerror().error(AP_InternalError::error_t::stack_overflow, 0xFFFF);
}
for (thread_t *tp = chRegFirstThread(); tp; tp = chRegNextThread(tp)) {
if (stack_free(tp->wabase) < min_stack) {
// use task priority for line number. This allows us to
// identify the task fairly reliably
AP::internalerror().error(AP_InternalError::error_t::stack_overflow, tp->realprio);
}
}
}
#endif // CH_DBG_ENABLE_STACK_CHECK == TRUE
#endif // CH_CFG_USE_DYNAMIC
#endif // HAL_SCHEDULER_ENABLED