Merge branch 'bugfix/tw17012_wifi_interface_stop' into 'release/v2.1'

Bugfix/tw17012 wifi interface stop

See merge request idf/esp-idf!2363
This commit is contained in:
Jiang Jiang Jian 2018-05-09 10:06:15 +08:00
commit c5498468d8
5 changed files with 206 additions and 29 deletions

View File

@ -2163,7 +2163,6 @@ BaseType_t xAlreadyYielded = pdFALSE;
{ {
/* We can schedule the awoken task on this CPU. */ /* We can schedule the awoken task on this CPU. */
xYieldPending[xPortGetCoreID()] = pdTRUE; xYieldPending[xPortGetCoreID()] = pdTRUE;
break;
} }
else else
{ {
@ -3023,6 +3022,8 @@ BaseType_t xTaskRemoveFromEventList( const List_t * const pxEventList )
{ {
TCB_t *pxUnblockedTCB; TCB_t *pxUnblockedTCB;
BaseType_t xReturn; BaseType_t xReturn;
BaseType_t xTaskCanBeReady;
UBaseType_t i, uxTargetCPU;
/* THIS FUNCTION MUST BE CALLED FROM A CRITICAL SECTION. It can also be /* THIS FUNCTION MUST BE CALLED FROM A CRITICAL SECTION. It can also be
called from a critical section within an ISR. */ called from a critical section within an ISR. */
@ -3046,7 +3047,24 @@ BaseType_t xReturn;
return pdFALSE; return pdFALSE;
} }
if( uxSchedulerSuspended[ xPortGetCoreID() ] == ( UBaseType_t ) pdFALSE ) /* Determine if the task can possibly be run on either CPU now, either because the scheduler
the task is pinned to is running or because a scheduler is running on any CPU. */
xTaskCanBeReady = pdFALSE;
if ( pxUnblockedTCB->xCoreID == tskNO_AFFINITY ) {
uxTargetCPU = xPortGetCoreID();
for (i = 0; i < portNUM_PROCESSORS; i++) {
if ( uxSchedulerSuspended[ i ] == ( UBaseType_t ) pdFALSE ) {
xTaskCanBeReady = pdTRUE;
break;
}
}
} else {
uxTargetCPU = pxUnblockedTCB->xCoreID;
xTaskCanBeReady = uxSchedulerSuspended[ uxTargetCPU ] == ( UBaseType_t ) pdFALSE;
}
if( xTaskCanBeReady )
{ {
( void ) uxListRemove( &( pxUnblockedTCB->xGenericListItem ) ); ( void ) uxListRemove( &( pxUnblockedTCB->xGenericListItem ) );
prvAddTaskToReadyList( pxUnblockedTCB ); prvAddTaskToReadyList( pxUnblockedTCB );
@ -3054,8 +3072,8 @@ BaseType_t xReturn;
else else
{ {
/* The delayed and ready lists cannot be accessed, so hold this task /* The delayed and ready lists cannot be accessed, so hold this task
pending until the scheduler is resumed. */ pending until the scheduler is resumed on this CPU. */
vListInsertEnd( &( xPendingReadyList[ xPortGetCoreID() ] ), &( pxUnblockedTCB->xEventListItem ) ); vListInsertEnd( &( xPendingReadyList[ uxTargetCPU ] ), &( pxUnblockedTCB->xEventListItem ) );
} }
if ( tskCAN_RUN_HERE(pxUnblockedTCB->xCoreID) && pxUnblockedTCB->uxPriority >= pxCurrentTCB[ xPortGetCoreID() ]->uxPriority ) if ( tskCAN_RUN_HERE(pxUnblockedTCB->xCoreID) && pxUnblockedTCB->uxPriority >= pxCurrentTCB[ xPortGetCoreID() ]->uxPriority )

View File

@ -12,7 +12,7 @@
#include "driver/timer.h" #include "driver/timer.h"
static SemaphoreHandle_t isr_semaphore; static SemaphoreHandle_t isr_semaphore;
static volatile unsigned isr_count, task_count; static volatile unsigned isr_count;
/* Timer ISR increments an ISR counter, and signals a /* Timer ISR increments an ISR counter, and signals a
mutex semaphore to wake up another counter task */ mutex semaphore to wake up another counter task */
@ -29,33 +29,42 @@ static void timer_group0_isr(void *vp_arg)
} }
} }
static void counter_task_fn(void *ignore) typedef struct {
SemaphoreHandle_t trigger_sem;
volatile unsigned counter;
} counter_config_t;
static void counter_task_fn(void *vp_config)
{ {
counter_config_t *config = (counter_config_t *)vp_config;
printf("counter_task running...\n"); printf("counter_task running...\n");
while(1) { while(1) {
xSemaphoreTake(isr_semaphore, portMAX_DELAY); xSemaphoreTake(config->trigger_sem, portMAX_DELAY);
task_count++; config->counter++;
} }
} }
/* This test verifies that an interrupt can wake up a task while the scheduler is disabled. /* This test verifies that an interrupt can wake up a task while the scheduler is disabled.
In the FreeRTOS implementation, this exercises the xPendingReadyList for that core. In the FreeRTOS implementation, this exercises the xPendingReadyList for that core.
*/ */
TEST_CASE("Handle pending context switch while scheduler disabled", "[freertos]") TEST_CASE("Scheduler disabled can handle a pending context switch on resume", "[freertos]")
{ {
task_count = 0;
isr_count = 0; isr_count = 0;
isr_semaphore = xSemaphoreCreateMutex(); isr_semaphore = xSemaphoreCreateMutex();
TaskHandle_t counter_task; TaskHandle_t counter_task;
counter_config_t count_config = {
.trigger_sem = isr_semaphore,
.counter = 0,
};
xTaskCreatePinnedToCore(counter_task_fn, "counter", 2048, xTaskCreatePinnedToCore(counter_task_fn, "counter", 2048,
NULL, UNITY_FREERTOS_PRIORITY + 1, &count_config, UNITY_FREERTOS_PRIORITY + 1,
&counter_task, UNITY_FREERTOS_CPU); &counter_task, UNITY_FREERTOS_CPU);
/* Configure timer ISR */ /* Configure timer ISR */
const timer_config_t config = { const timer_config_t timer_config = {
.alarm_en = 1, .alarm_en = 1,
.auto_reload = 1, .auto_reload = 1,
.counter_dir = TIMER_COUNT_UP, .counter_dir = TIMER_COUNT_UP,
@ -64,7 +73,7 @@ TEST_CASE("Handle pending context switch while scheduler disabled", "[freertos]"
.counter_en = TIMER_PAUSE, .counter_en = TIMER_PAUSE,
}; };
/* Configure timer */ /* Configure timer */
timer_init(TIMER_GROUP_0, TIMER_0, &config); timer_init(TIMER_GROUP_0, TIMER_0, &timer_config);
timer_pause(TIMER_GROUP_0, TIMER_0); timer_pause(TIMER_GROUP_0, TIMER_0);
timer_set_counter_value(TIMER_GROUP_0, TIMER_0, 0); timer_set_counter_value(TIMER_GROUP_0, TIMER_0, 0);
timer_set_alarm_value(TIMER_GROUP_0, TIMER_0, 1000); timer_set_alarm_value(TIMER_GROUP_0, TIMER_0, 1000);
@ -75,20 +84,20 @@ TEST_CASE("Handle pending context switch while scheduler disabled", "[freertos]"
vTaskDelay(5); vTaskDelay(5);
// Check some counts have been triggered via the ISR // Check some counts have been triggered via the ISR
TEST_ASSERT(task_count > 10); TEST_ASSERT(count_config.counter > 10);
TEST_ASSERT(isr_count > 10); TEST_ASSERT(isr_count > 10);
for (int i = 0; i < 20; i++) { for (int i = 0; i < 20; i++) {
vTaskSuspendAll(); vTaskSuspendAll();
esp_intr_noniram_disable(); esp_intr_noniram_disable();
unsigned no_sched_task = task_count; unsigned no_sched_task = count_config.counter;
// scheduler off on this CPU... // scheduler off on this CPU...
ets_delay_us(20 * 1000); ets_delay_us(20 * 1000);
//TEST_ASSERT_NOT_EQUAL(no_sched_isr, isr_count); //TEST_ASSERT_NOT_EQUAL(no_sched_isr, isr_count);
TEST_ASSERT_EQUAL(task_count, no_sched_task); TEST_ASSERT_EQUAL(count_config.counter, no_sched_task);
// disable timer interrupts // disable timer interrupts
timer_disable_intr(TIMER_GROUP_0, TIMER_0); timer_disable_intr(TIMER_GROUP_0, TIMER_0);
@ -98,9 +107,139 @@ TEST_CASE("Handle pending context switch while scheduler disabled", "[freertos]"
esp_intr_noniram_enable(); esp_intr_noniram_enable();
xTaskResumeAll(); xTaskResumeAll();
TEST_ASSERT_NOT_EQUAL(task_count, no_sched_task); TEST_ASSERT_NOT_EQUAL(count_config.counter, no_sched_task);
} }
vTaskDelete(counter_task); vTaskDelete(counter_task);
vSemaphoreDelete(isr_semaphore); vSemaphoreDelete(isr_semaphore);
} }
/* Multiple tasks on different cores can be added to the pending ready list
while scheduler is suspended, and should be started once the scheduler
resumes.
*/
TEST_CASE("Scheduler disabled can wake multiple tasks on resume", "[freertos]")
{
#define TASKS_PER_PROC 4
TaskHandle_t tasks[portNUM_PROCESSORS][TASKS_PER_PROC] = { 0 };
counter_config_t counters[portNUM_PROCESSORS][TASKS_PER_PROC] = { 0 };
/* Start all the tasks, they will block on isr_semaphore */
for (int p = 0; p < portNUM_PROCESSORS; p++) {
for (int t = 0; t < TASKS_PER_PROC; t++) {
counters[p][t].trigger_sem = xSemaphoreCreateMutex();
TEST_ASSERT_NOT_NULL( counters[p][t].trigger_sem );
TEST_ASSERT( xSemaphoreTake(counters[p][t].trigger_sem, 0) );
xTaskCreatePinnedToCore(counter_task_fn, "counter", 2048,
&counters[p][t], UNITY_FREERTOS_PRIORITY + 1,
&tasks[p][t], p);
TEST_ASSERT_NOT_NULL( tasks[p][t] );
}
}
/* takes a while to initialize tasks on both cores, sometimes... */
vTaskDelay(TASKS_PER_PROC * portNUM_PROCESSORS * 3);
/* Check nothing is counting, each counter should be blocked on its trigger_sem */
for (int p = 0; p < portNUM_PROCESSORS; p++) {
for (int t = 0; t < TASKS_PER_PROC; t++) {
TEST_ASSERT_EQUAL(0, counters[p][t].counter);
}
}
/* Suspend scheduler on this CPU */
vTaskSuspendAll();
/* Give all the semaphores once. This will wake tasks immediately on the other
CPU, but they are deferred here until the scheduler resumes.
*/
for (int p = 0; p < portNUM_PROCESSORS; p++) {
for (int t = 0; t < TASKS_PER_PROC; t++) {
xSemaphoreGive(counters[p][t].trigger_sem);
}
}
ets_delay_us(200); /* Let the other CPU do some things */
for (int p = 0; p < portNUM_PROCESSORS; p++) {
for (int t = 0; t < TASKS_PER_PROC; t++) {
int expected = (p == UNITY_FREERTOS_CPU) ? 0 : 1; // Has run if it was on the other CPU
ets_printf("Checking CPU %d task %d (expected %d actual %d)\n", p, t, expected, counters[p][t].counter);
TEST_ASSERT_EQUAL(expected, counters[p][t].counter);
}
}
/* Resume scheduler */
xTaskResumeAll();
/* Now the tasks on both CPUs should have been woken once and counted once. */
for (int p = 0; p < portNUM_PROCESSORS; p++) {
for (int t = 0; t < TASKS_PER_PROC; t++) {
ets_printf("Checking CPU %d task %d (expected 1 actual %d)\n", p, t, counters[p][t].counter);
TEST_ASSERT_EQUAL(1, counters[p][t].counter);
}
}
/* Clean up */
for (int p = 0; p < portNUM_PROCESSORS; p++) {
for (int t = 0; t < TASKS_PER_PROC; t++) {
vTaskDelete(tasks[p][t]);
vSemaphoreDelete(counters[p][t].trigger_sem);
}
}
}
static volatile bool sched_suspended;
static void suspend_scheduler_5ms_task_fn(void *ignore)
{
vTaskSuspendAll();
sched_suspended = true;
for (int i = 0; i <5; i++) {
ets_delay_us(1000);
}
xTaskResumeAll();
sched_suspended = false;
vTaskDelete(NULL);
}
#ifndef CONFIG_FREERTOS_UNICORE
/* If the scheduler is disabled on one CPU (A) with a task blocked on something, and a task
on B (where scheduler is running) wakes it, then the task on A should be woken on resume.
*/
TEST_CASE("Scheduler disabled on CPU B, tasks on A can wake", "[freertos]")
{
TaskHandle_t counter_task;
SemaphoreHandle_t wake_sem = xSemaphoreCreateMutex();
xSemaphoreTake(wake_sem, 0);
counter_config_t count_config = {
.trigger_sem = wake_sem,
.counter = 0,
};
xTaskCreatePinnedToCore(counter_task_fn, "counter", 2048,
&count_config, UNITY_FREERTOS_PRIORITY + 1,
&counter_task, !UNITY_FREERTOS_CPU);
xTaskCreatePinnedToCore(suspend_scheduler_5ms_task_fn, "suspender", 2048,
NULL, UNITY_FREERTOS_PRIORITY - 1,
NULL, !UNITY_FREERTOS_CPU);
/* counter task is now blocked on other CPU, waiting for wake_sem, and we expect
that this CPU's scheduler will be suspended for 5ms shortly... */
while(!sched_suspended) { }
xSemaphoreGive(wake_sem);
ets_delay_us(1000);
// Bit of a race here if the other CPU resumes its scheduler, but 5ms is a long time... */
TEST_ASSERT(sched_suspended);
TEST_ASSERT_EQUAL(0, count_config.counter); // the other task hasn't woken yet, because scheduler is off
TEST_ASSERT(sched_suspended);
/* wait for the rest of the 5ms... */
while(sched_suspended) { }
ets_delay_us(100);
TEST_ASSERT_EQUAL(1, count_config.counter); // when scheduler resumes, counter task should immediately count
vTaskDelete(counter_task);
}
#endif

View File

@ -59,6 +59,12 @@ void spi_flash_op_unlock()
{ {
xSemaphoreGive(s_flash_op_mutex); xSemaphoreGive(s_flash_op_mutex);
} }
/*
If you're going to modify this, keep in mind that while the flash caches of the pro and app
cpu are separate, the psram cache is *not*. If one of the CPUs returns from a flash routine
with its cache enabled but the other CPUs cache is not enabled yet, you will have problems
when accessing psram from the former CPU.
*/
void IRAM_ATTR spi_flash_op_block_func(void* arg) void IRAM_ATTR spi_flash_op_block_func(void* arg)
{ {
@ -67,8 +73,6 @@ void IRAM_ATTR spi_flash_op_block_func(void* arg)
// Restore interrupts that aren't located in IRAM // Restore interrupts that aren't located in IRAM
esp_intr_noniram_disable(); esp_intr_noniram_disable();
uint32_t cpuid = (uint32_t) arg; uint32_t cpuid = (uint32_t) arg;
// Disable cache so that flash operation can start
spi_flash_disable_cache(cpuid, &s_flash_op_cache_state[cpuid]);
// s_flash_op_complete flag is cleared on *this* CPU, otherwise the other // s_flash_op_complete flag is cleared on *this* CPU, otherwise the other
// CPU may reset the flag back to false before IPC task has a chance to check it // CPU may reset the flag back to false before IPC task has a chance to check it
// (if it is preempted by an ISR taking non-trivial amount of time) // (if it is preempted by an ISR taking non-trivial amount of time)
@ -122,8 +126,12 @@ void IRAM_ATTR spi_flash_disable_interrupts_caches_and_other_cpu()
} }
// Kill interrupts that aren't located in IRAM // Kill interrupts that aren't located in IRAM
esp_intr_noniram_disable(); esp_intr_noniram_disable();
// Disable cache on this CPU as well // This CPU executes this routine, with non-IRAM interrupts and the scheduler
// disabled. The other CPU is spinning in the spi_flash_op_block_func task, also
// with non-iram interrupts and the scheduler disabled. None of these CPUs will
// touch external RAM or flash this way, so we can safely disable caches.
spi_flash_disable_cache(cpuid, &s_flash_op_cache_state[cpuid]); spi_flash_disable_cache(cpuid, &s_flash_op_cache_state[cpuid]);
spi_flash_disable_cache(other_cpuid, &s_flash_op_cache_state[other_cpuid]);
} }
void IRAM_ATTR spi_flash_enable_interrupts_caches_and_other_cpu() void IRAM_ATTR spi_flash_enable_interrupts_caches_and_other_cpu()
@ -133,22 +141,20 @@ void IRAM_ATTR spi_flash_enable_interrupts_caches_and_other_cpu()
#ifndef NDEBUG #ifndef NDEBUG
// Sanity check: flash operation ends on the same CPU as it has started // Sanity check: flash operation ends on the same CPU as it has started
assert(cpuid == s_flash_op_cpu); assert(cpuid == s_flash_op_cpu);
// More sanity check: if scheduler isn't started, only CPU0 can call this.
assert(!(xTaskGetSchedulerState() == taskSCHEDULER_NOT_STARTED && cpuid != 0));
s_flash_op_cpu = -1; s_flash_op_cpu = -1;
#endif #endif
// Re-enable cache on this CPU // Re-enable cache on both CPUs. After this, cache (flash and external RAM) should work again.
spi_flash_restore_cache(cpuid, s_flash_op_cache_state[cpuid]); spi_flash_restore_cache(cpuid, s_flash_op_cache_state[cpuid]);
if (xTaskGetSchedulerState() == taskSCHEDULER_NOT_STARTED) {
// Scheduler is not running yet — this means we are running on PRO CPU.
// other_cpuid is APP CPU, and it is either in reset or is spinning in
// user_start_cpu1, which is in IRAM. So we can simply reenable cache.
assert(other_cpuid == 1);
spi_flash_restore_cache(other_cpuid, s_flash_op_cache_state[other_cpuid]); spi_flash_restore_cache(other_cpuid, s_flash_op_cache_state[other_cpuid]);
} else {
if (xTaskGetSchedulerState() != taskSCHEDULER_NOT_STARTED) {
// Signal to spi_flash_op_block_task that flash operation is complete // Signal to spi_flash_op_block_task that flash operation is complete
s_flash_op_complete = true; s_flash_op_complete = true;
} }
// Re-enable non-iram interrupts // Re-enable non-iram interrupts
esp_intr_noniram_enable(); esp_intr_noniram_enable();

View File

@ -95,6 +95,11 @@ void IRAM_ATTR spi_flash_guard_set(const spi_flash_guard_funcs_t *funcs)
s_flash_guard_ops = funcs; s_flash_guard_ops = funcs;
} }
const spi_flash_guard_funcs_t *IRAM_ATTR spi_flash_guard_get()
{
return s_flash_guard_ops;
}
size_t IRAM_ATTR spi_flash_get_chip_size() size_t IRAM_ATTR spi_flash_get_chip_size()
{ {
return g_rom_flashchip.chip_size; return g_rom_flashchip.chip_size;

View File

@ -297,6 +297,15 @@ typedef struct {
*/ */
void spi_flash_guard_set(const spi_flash_guard_funcs_t* funcs); void spi_flash_guard_set(const spi_flash_guard_funcs_t* funcs);
/**
* @brief Get the guard functions used for flash access
*
* @return The guard functions that were set via spi_flash_guard_set(). These functions
* can be called if implementing custom low-level SPI flash operations.
*/
const spi_flash_guard_funcs_t *spi_flash_guard_get();
/** /**
* @brief Default OS-aware flash access guard functions * @brief Default OS-aware flash access guard functions
*/ */