esp_restart: fix possible race while stalling other CPU, enable WDT early
Previously esp_restart would stall the other CPU before enabling RTC_WDT. If the other CPU was executing an s32c1i instruction, the lock signal from CPU to the arbiter would still be held after CPU was stalled. If the CPU running esp_restart would then try to access the same locked memory pool, it would be stuck, because lock signal would never be released. With this change, esp_restart resets the other CPU before stalling it. Ideally, we would want to reset the CPU and keep it in reset, but the hardware doesn't have such feature for PRO_CPU (it is possible to hold APP_CPU in reset using DPORT register). Given that ROM code will not use s32c1i in the first few hundred cycles, doing reset and then stall seems to be safe. In addition to than, RTC_WDT initialization is moved to the beginning of the function, to prevent possible lock-up if CPU stalling still has any issue.
This commit is contained in:
parent
07543dfec4
commit
e324707cc8
6 changed files with 45 additions and 14 deletions
|
@ -185,3 +185,12 @@ void esp_dport_access_int_deinit(void)
|
|||
#endif
|
||||
portEXIT_CRITICAL_ISR(&g_dport_mux);
|
||||
}
|
||||
|
||||
|
||||
void esp_dport_access_int_abort(void)
|
||||
{
|
||||
dport_core_state[0] = DPORT_CORE_STATE_IDLE;
|
||||
#ifndef CONFIG_FREERTOS_UNICORE
|
||||
dport_core_state[1] = DPORT_CORE_STATE_IDLE;
|
||||
#endif
|
||||
}
|
||||
|
|
|
@ -23,6 +23,7 @@ void esp_dport_access_stall_other_cpu_start(void);
|
|||
void esp_dport_access_stall_other_cpu_end(void);
|
||||
void esp_dport_access_int_init(void);
|
||||
void esp_dport_access_int_deinit(void);
|
||||
void esp_dport_access_int_abort(void);
|
||||
|
||||
#if defined(BOOTLOADER_BUILD) || defined(CONFIG_FREERTOS_UNICORE) || !defined(ESP_PLATFORM)
|
||||
#define DPORT_STALL_OTHER_CPU_START()
|
||||
|
|
|
@ -256,22 +256,31 @@ void IRAM_ATTR esp_restart(void)
|
|||
*/
|
||||
void IRAM_ATTR esp_restart_noos()
|
||||
{
|
||||
const uint32_t core_id = xPortGetCoreID();
|
||||
const uint32_t other_core_id = core_id == 0 ? 1 : 0;
|
||||
esp_cpu_stall(other_core_id);
|
||||
// Disable interrupts
|
||||
xt_ints_off(0xFFFFFFFF);
|
||||
|
||||
// other core is now stalled, can access DPORT registers directly
|
||||
esp_dport_access_int_deinit();
|
||||
|
||||
// We need to disable TG0/TG1 watchdogs
|
||||
// First enable RTC watchdog to be on the safe side
|
||||
// Enable RTC watchdog for 1 second
|
||||
REG_WRITE(RTC_CNTL_WDTWPROTECT_REG, RTC_CNTL_WDT_WKEY_VALUE);
|
||||
REG_WRITE(RTC_CNTL_WDTCONFIG0_REG,
|
||||
RTC_CNTL_WDT_FLASHBOOT_MOD_EN_M |
|
||||
(RTC_WDT_STG_SEL_RESET_SYSTEM << RTC_CNTL_WDT_STG0_S) |
|
||||
(RTC_WDT_STG_SEL_RESET_RTC << RTC_CNTL_WDT_STG1_S) |
|
||||
(1 << RTC_CNTL_WDT_SYS_RESET_LENGTH_S) |
|
||||
(1 << RTC_CNTL_WDT_CPU_RESET_LENGTH_S) );
|
||||
REG_WRITE(RTC_CNTL_WDTCONFIG1_REG, 128000);
|
||||
|
||||
// Reset and stall the other CPU.
|
||||
// CPU must be reset before stalling, in case it was running a s32c1i
|
||||
// instruction. This would cause memory pool to be locked by arbiter
|
||||
// to the stalled CPU, preventing current CPU from accessing this pool.
|
||||
const uint32_t core_id = xPortGetCoreID();
|
||||
const uint32_t other_core_id = (core_id == 0) ? 1 : 0;
|
||||
esp_cpu_reset(other_core_id);
|
||||
esp_cpu_stall(other_core_id);
|
||||
|
||||
// Other core is now stalled, can access DPORT registers directly
|
||||
esp_dport_access_int_abort();
|
||||
|
||||
// Disable TG0/TG1 watchdogs
|
||||
TIMERG0.wdt_wprotect=TIMG_WDT_WKEY_VALUE;
|
||||
TIMERG0.wdt_config0.en = 0;
|
||||
|
@ -280,8 +289,6 @@ void IRAM_ATTR esp_restart_noos()
|
|||
TIMERG1.wdt_config0.en = 0;
|
||||
TIMERG1.wdt_wprotect=0;
|
||||
|
||||
// Disable all interrupts
|
||||
xt_ints_off(0xFFFFFFFF);
|
||||
|
||||
// Disable cache
|
||||
Cache_Read_Disable(0);
|
||||
|
@ -322,14 +329,14 @@ void IRAM_ATTR esp_restart_noos()
|
|||
// Reset CPUs
|
||||
if (core_id == 0) {
|
||||
// Running on PRO CPU: APP CPU is stalled. Can reset both CPUs.
|
||||
SET_PERI_REG_MASK(RTC_CNTL_OPTIONS0_REG,
|
||||
RTC_CNTL_SW_PROCPU_RST_M | RTC_CNTL_SW_APPCPU_RST_M);
|
||||
esp_cpu_reset(1);
|
||||
esp_cpu_reset(0);
|
||||
} else {
|
||||
// Running on APP CPU: need to reset PRO CPU and unstall it,
|
||||
// then reset APP CPU
|
||||
SET_PERI_REG_MASK(RTC_CNTL_OPTIONS0_REG, RTC_CNTL_SW_PROCPU_RST_M);
|
||||
esp_cpu_reset(0);
|
||||
esp_cpu_unstall(0);
|
||||
SET_PERI_REG_MASK(RTC_CNTL_OPTIONS0_REG, RTC_CNTL_SW_APPCPU_RST_M);
|
||||
esp_cpu_reset(1);
|
||||
}
|
||||
while(true) {
|
||||
;
|
||||
|
|
|
@ -44,6 +44,12 @@ void IRAM_ATTR esp_cpu_unstall(int cpu_id)
|
|||
}
|
||||
}
|
||||
|
||||
void IRAM_ATTR esp_cpu_reset(int cpu_id)
|
||||
{
|
||||
SET_PERI_REG_MASK(RTC_CNTL_OPTIONS0_REG,
|
||||
cpu_id == 0 ? RTC_CNTL_SW_PROCPU_RST_M : RTC_CNTL_SW_APPCPU_RST_M);
|
||||
}
|
||||
|
||||
bool IRAM_ATTR esp_cpu_in_ocd_debug_mode()
|
||||
{
|
||||
#if CONFIG_ESP32_DEBUG_OCDAWARE
|
||||
|
|
|
@ -85,6 +85,13 @@ void esp_cpu_stall(int cpu_id);
|
|||
*/
|
||||
void esp_cpu_unstall(int cpu_id);
|
||||
|
||||
/**
|
||||
* @brief Reset CPU using RTC controller
|
||||
* @param cpu_id ID of the CPU to reset (0 = PRO, 1 = APP)
|
||||
*/
|
||||
void esp_cpu_reset(int cpu_id);
|
||||
|
||||
|
||||
/**
|
||||
* @brief Returns true if a JTAG debugger is attached to CPU
|
||||
* OCD (on chip debug) port.
|
||||
|
|
|
@ -1718,6 +1718,7 @@
|
|||
#define RTC_WDT_STG_SEL_INT 1
|
||||
#define RTC_WDT_STG_SEL_RESET_CPU 2
|
||||
#define RTC_WDT_STG_SEL_RESET_SYSTEM 3
|
||||
#define RTC_WDT_STG_SEL_RESET_RTC 4
|
||||
|
||||
#define RTC_CNTL_WDTCONFIG1_REG (DR_REG_RTCCNTL_BASE + 0x90)
|
||||
/* RTC_CNTL_WDT_STG0_HOLD : R/W ;bitpos:[31:0] ;default: 32'd128000 ; */
|
||||
|
|
Loading…
Reference in a new issue