feat(esp_system): Print backtrace for both CPUs when cache error does not determine CPU

components/esp_system/port/arch/xtensa/panic_arch.c

@@ -340,6 +340,7 @@ static inline void print_cache_err_details(const void *f)
             break;
         case EXTMEM_DCACHE_WRITE_FLASH_ST:
             panic_print_str("Write back error occurred while dcache tries to write back to flash\r\n");
+            panic_print_str("The following backtrace may not indicate the code that caused Cache invalid access\r\n");
             break;
         case EXTMEM_MMU_ENTRY_FAULT_ST:
             vaddr = REG_READ(EXTMEM_CACHE_MMU_FAULT_VADDR_REG);

components/esp_system/port/panic_handler.c

@@ -151,9 +151,20 @@ static void panic_handler(void *frame, bool pseudo_excause)
             busy_wait();
         } else if (panic_get_cause(frame) == PANIC_RSN_INTWDT_CPU1 && core_id == 0) {
             busy_wait();
-        } else if (panic_get_cause(frame) == PANIC_RSN_CACHEERR && core_id != esp_cache_err_get_cpuid()) {
-            g_exc_frames[core_id] = NULL; // Only print the backtrace for the offending core
-            busy_wait();
+        } else if (panic_get_cause(frame) == PANIC_RSN_CACHEERR) {
+            // The invalid cache access interrupt calls to the panic handler.
+            // When the cache interrupt happens, we can not determine the CPU where the
+            // invalid cache access has occurred.
+            if (esp_cache_err_get_cpuid() == -1) {
+                // We can not determine the CPU where the invalid cache access has occurred.
+                // Print backtraces for both CPUs.
+                if (core_id != 0) {
+                    busy_wait();
+                }
+            } else if (core_id != esp_cache_err_get_cpuid()) {
+                g_exc_frames[core_id] = NULL; // Only print the backtrace for the offending core
+                busy_wait();
+            }
         }
 #if CONFIG_ESP_SYSTEM_HW_STACK_GUARD
         else if (panic_get_cause(frame) == ETS_ASSIST_DEBUG_INUM &&
@@ -183,7 +194,7 @@ static void panic_handler(void *frame, bool pseudo_excause)
 #if __XTENSA__
         if (!(esp_ptr_executable(esp_cpu_pc_to_addr(panic_get_address(frame))) && (panic_get_address(frame) & 0xC0000000U))) {
             /* Xtensa ABI sets the 2 MSBs of the PC according to the windowed call size
-             * Incase the PC is invalid, GDB will fail to translate addresses to function names
+             * In case the PC is invalid, GDB will fail to translate addresses to function names
              * Hence replacing the PC to a placeholder address in case of invalid PC
              */
             panic_set_address(frame, (uint32_t)&_invalid_pc_placeholder);

tools/test_apps/system/panic/main/include/test_panic.h

@@ -59,6 +59,10 @@ void test_assert(void);
 
 void test_assert_cache_disabled(void);
 
+void test_assert_cache_write_back_error_can_print_backtrace(void);
+
+void test_assert_cache_write_back_error_can_print_backtrace2(void);
+
 void test_illegal_access(void);
 
 void test_capture_dram(void);

tools/test_apps/system/panic/main/test_app_main.c

@@ -114,6 +114,8 @@ void app_main(void)
     HANDLE_TEST(test_name, test_ub);
     HANDLE_TEST(test_name, test_assert);
     HANDLE_TEST(test_name, test_assert_cache_disabled);
+    HANDLE_TEST(test_name, test_assert_cache_write_back_error_can_print_backtrace);
+    HANDLE_TEST(test_name, test_assert_cache_write_back_error_can_print_backtrace2);
 #if CONFIG_IDF_TARGET_ESP32
     HANDLE_TEST(test_name, test_illegal_access);
 #endif

tools/test_apps/system/panic/main/test_panic.c

@@ -7,10 +7,12 @@
 #include <unistd.h>
 #include <assert.h>
 #include <string.h>
+#include <inttypes.h>
 
 #include "esp_partition.h"
 #include "esp_flash.h"
 #include "esp_system.h"
+#include "spi_flash_mmap.h"
 
 #include "esp_private/cache_utils.h"
 #include "esp_memory_utils.h"
@@ -20,6 +22,7 @@
 #include "freertos/task.h"
 
 #include "hal/mpu_hal.h"
+#include "rom/cache.h"
 
 /* Test utility function */
 
@@ -177,6 +180,46 @@ void IRAM_ATTR test_assert_cache_disabled(void)
     assert(0);
 }
 
+const char TEST_STR[] = "my_tag";
+void test_assert_cache_write_back_error_can_print_backtrace(void)
+{
+    printf("1) %p\n", TEST_STR);
+    *(uint32_t*)TEST_STR = 3; // We changed the rodata string.
+    // All chips except ESP32S3 stop execution here and raise a LoadStore error on the line above.
+#if CONFIG_IDF_TARGET_ESP32S3
+    // On the ESP32S3, the error occurs later when the cache writeback is triggered
+    // (in this test, a direct call to Cache_WriteBack_All).
+    Cache_WriteBack_All(); // Cache writeback triggers the invalid cache access interrupt.
+#endif
+    // We are testing that the backtrace is printed instead of TG1WDT.
+    printf("2) %p\n", TEST_STR); // never get to this place.
+}
+
+void test_assert_cache_write_back_error_can_print_backtrace2(void)
+{
+    printf("1) %p\n", TEST_STR);
+    *(uint32_t*)TEST_STR = 3; // We changed the rodata string.
+    // All chips except ESP32S3 stop execution here and raise a LoadStore error on the line above.
+    // On the ESP32S3, the error occurs later when the cache writeback is triggered
+    // (in this test, a large range of DRAM is mapped and read, causing an error).
+    uint8_t temp = 0;
+    size_t map_size = SPI_FLASH_SEC_SIZE * 512;
+    const void *map;
+    spi_flash_mmap_handle_t out_handle;
+    esp_err_t err = spi_flash_mmap(0, map_size, SPI_FLASH_MMAP_DATA, &map, &out_handle);
+    if (err != ESP_OK) {
+        printf("spi_flash_mmap failed %x\n", err);
+        return;
+    }
+    const uint8_t *rodata = map;
+    for (size_t i = 0; i < map_size; i++) {
+        temp = rodata[i];
+    }
+    // Cache writeback triggers the invalid cache access interrupt.
+    // We are testing that the backtrace is printed instead of TG1WDT.
+    printf("2) %p 0x%" PRIx8 " \n", TEST_STR, temp); // never get to this place.
+}
+
 /**
  * This function overwrites the stack beginning from the valid area continuously towards and beyond
  * the end of the stack (stack base) of the current task.

tools/test_apps/system/panic/pytest_panic.py

@@ -535,6 +535,42 @@ def test_assert_cache_disabled(
     )
 
 
+def cache_error_log_check(dut: PanicTestDut) -> None:
+    if dut.is_xtensa:
+        if dut.target == 'esp32s3':
+            dut.expect_exact("Guru Meditation Error: Core  / panic'ed (Cache disabled but cached memory region accessed)")
+            dut.expect_exact('Write back error occurred while dcache tries to write back to flash')
+            dut.expect_exact('The following backtrace may not indicate the code that caused Cache invalid access')
+        else:
+            dut.expect_exact("Guru Meditation Error: Core  0 panic'ed (LoadStoreError)")
+    else:
+        dut.expect_exact("Guru Meditation Error: Core  0 panic'ed (Store access fault)")
+    dut.expect_reg_dump(0)
+    if dut.target == 'esp32s3':
+        dut.expect_reg_dump(1)
+    dut.expect_cpu_reset()
+
+
+@pytest.mark.generic
+@pytest.mark.supported_targets
+@pytest.mark.parametrize('config', ['panic'], indirect=True)
+def test_assert_cache_write_back_error_can_print_backtrace(
+    dut: PanicTestDut, config: str, test_func_name: str
+) -> None:
+    dut.run_test_func(test_func_name)
+    cache_error_log_check(dut)
+
+
+@pytest.mark.generic
+@pytest.mark.supported_targets
+@pytest.mark.parametrize('config', ['panic'], indirect=True)
+def test_assert_cache_write_back_error_can_print_backtrace2(
+    dut: PanicTestDut, config: str, test_func_name: str
+) -> None:
+    dut.run_test_func(test_func_name)
+    cache_error_log_check(dut)
+
+
 @pytest.mark.esp32
 @pytest.mark.generic
 @pytest.mark.parametrize('config', ['panic_delay'], indirect=True)