Merge branch 'fix/fix_load_store_failure_mepc' into 'master'

fix(system): esp32p4: fix mepc when load/store failure occurred

See merge request espressif/esp-idf!30252

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

@@ -294,6 +294,14 @@ void panic_arch_fill_info(void *frame, panic_info_t *info)
 
     info->description = "Exception was unhandled.";
 
+#if SOC_ASYNCHRONOUS_BUS_ERROR_MODE
+    uintptr_t bus_error_pc = rv_utils_asynchronous_bus_get_error_pc();
+    if (bus_error_pc) {
+        /* Change mepc with the fault pc address */
+        regs->mepc = bus_error_pc;
+    }
+#endif // SOC_ASYNCHRONOUS_BUS_ERROR_MODE
+
     info->addr = (void *) regs->mepc;
 }
 

components/riscv/include/riscv/csr.h

@@ -155,6 +155,20 @@ extern "C" {
 #define TDATA1_HIT_S     (20)
 
 
+/********************************************************
+   Espressif's bus error exceptions registers and fields
+ ********************************************************/
+
+#define MEXSTATUS   0x7E1
+#define MHINT       0x7C5
+
+#define LDPC0       0xBE0
+#define LDPC1       0xBE1
+
+#define STPC0       0xBF0
+#define STPC1       0xBF1
+#define STPC2       0xBF2
+
 /* RISC-V CSR macros
  * Adapted from https://github.com/michaeljclark/riscv-probe/blob/master/libfemto/include/arch/riscv/machine.h
  */

components/riscv/include/riscv/rv_utils.h

@@ -138,12 +138,12 @@ FORCE_INLINE_ATTR void rv_utils_intr_global_disable(void)
  * and `interrupt_intc.h`.
  */
 
-#if SOC_CPU_HAS_FPU
-
 /* ------------------------------------------------- FPU Related ----------------------------------------------------
  *
  * ------------------------------------------------------------------------------------------------------------------ */
 
+#if SOC_CPU_HAS_FPU
+
 FORCE_INLINE_ATTR bool rv_utils_enable_fpu(void)
 {
     /* Set mstatus[14:13] to 0b01 to start the floating-point unit initialization */
@@ -172,6 +172,48 @@ FORCE_INLINE_ATTR void rv_utils_disable_fpu(void)
  *
  * ------------------------------------------------------------------------------------------------------------------ */
 
+#if SOC_ASYNCHRONOUS_BUS_ERROR_MODE
+
+FORCE_INLINE_ATTR uintptr_t rv_utils_asynchronous_bus_get_error_pc(void)
+{
+    uint32_t error_pc;
+    uint32_t mcause, mexstatus;
+
+    mexstatus = RV_READ_CSR(MEXSTATUS);
+    /* MEXSTATUS: Bit 8: Indicates that a load/store access fault (MCAUSE=5/7)
+     * is due to bus-error exception. If this bit is not cleared before exiting
+     * the exception handler, it will trigger a bus error again.
+     * Since we have not mechanisms to recover a normal program execution after
+     * load/store error appears, do nothing. */
+    if ((mexstatus & BIT(8)) == 0) {
+        return 0;
+    }
+    mcause = RV_READ_CSR(mcause) & 0xFF;
+    if (mcause == 5) { /* Load access fault */
+        /* Get the oldest PC at which the load instruction failed */
+        error_pc = RV_READ_CSR(LDPC1);
+        if (error_pc == 0) {
+            error_pc = RV_READ_CSR(LDPC0);
+        }
+    } else if (mcause == 7) { /* Store access fault */
+        /* Get the oldest PC at which the store instruction failed */
+        error_pc = RV_READ_CSR(STPC2);
+        if (error_pc == 0) {
+            error_pc = RV_READ_CSR(STPC1);
+            if (error_pc == 0) {
+                error_pc = RV_READ_CSR(STPC0);
+            }
+        }
+    } else {
+        return 0;
+    }
+    /* Bit 0: Valid bit indicating that this CSR holds the PC (program counter).
+     * Clear this bit */
+    return error_pc & ~(1);
+}
+
+#endif // SOC_ASYNCHRONOUS_BUS_ERROR_MODE
+
 /* ---------------------------------------------------- Debugging ------------------------------------------------------
  *
  * ------------------------------------------------------------------------------------------------------------------ */

components/soc/esp32p4/include/soc/Kconfig.soc_caps.in

@@ -1551,6 +1551,10 @@ config SOC_MEM_NON_CONTIGUOUS_SRAM
     bool
     default y
 
+config SOC_ASYNCHRONOUS_BUS_ERROR_MODE
+    bool
+    default y
+
 config SOC_EMAC_USE_IO_MUX
     bool
     default y

components/soc/esp32p4/include/soc/soc_caps.h

@@ -629,6 +629,7 @@
 /*-------------------------- Memory CAPS --------------------------*/
 #define SOC_MEM_TCM_SUPPORTED                      (1)
 #define SOC_MEM_NON_CONTIGUOUS_SRAM                (1)
+#define SOC_ASYNCHRONOUS_BUS_ERROR_MODE            (1)
 /*--------------------------- EMAC --------------------------------*/
 #define SOC_EMAC_USE_IO_MUX                        (1) /*!< GPIO matrix is used to select GPIO pads */
 

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

@@ -39,6 +39,8 @@ void test_panic_extram_stack(void);
 void test_task_wdt_cpu1(void);
 #endif
 
+void test_loadprohibited(void);
+
 void test_storeprohibited(void);
 
 void test_cache_error(void);

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

@@ -101,6 +101,7 @@ void app_main(void)
 #if !CONFIG_FREERTOS_UNICORE
     HANDLE_TEST(test_name, test_task_wdt_cpu1);
 #endif
+    HANDLE_TEST(test_name, test_loadprohibited);
     HANDLE_TEST(test_name, test_storeprohibited);
     HANDLE_TEST(test_name, test_cache_error);
     HANDLE_TEST(test_name, test_int_wdt_cache_disabled);

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

@@ -140,7 +140,15 @@ void test_task_wdt_cpu1(void)
 
 void __attribute__((no_sanitize_undefined)) test_storeprohibited(void)
 {
-    *(int*) 0x1 = 0;
+    *(int*) 0x4 = 0;
+    test_task_wdt_cpu0(); /* Trap for unhandled asynchronous bus errors */
+}
+
+void __attribute__((no_sanitize_undefined)) test_loadprohibited(void)
+{
+    static int __attribute__((used)) var;
+    var = *(int*) 0x4;
+    test_task_wdt_cpu0(); /* Trap for unhandled asynchronous bus errors */
 }
 
 void IRAM_ATTR test_cache_error(void)
@@ -222,7 +230,7 @@ void test_ub(void)
  * used for memory protection.
  *
  * However, this test is not valid for S2 and S3, because they have PMS
- * enabled on top of this, giving unpredicatable results.
+ * enabled on top of this, giving unpredictable results.
  */
 #if CONFIG_IDF_TARGET_ESP32
 void test_illegal_access(void)

tools/test_apps/system/panic/pytest_panic.py

@@ -365,14 +365,12 @@ def test_illegal_instruction(
     common_test(dut, config, expected_backtrace=get_default_backtrace(test_func_name))
 
 
-@pytest.mark.parametrize('config', CONFIGS, indirect=True)
+def check_x_prohibited(dut: PanicTestDut, config: str, test_func_name: str, operation: str) -> None:
-@pytest.mark.generic
-def test_storeprohibited(dut: PanicTestDut, config: str, test_func_name: str) -> None:
     dut.run_test_func(test_func_name)
     if dut.is_xtensa:
-        dut.expect_gme('StoreProhibited')
+        dut.expect_gme(f'{operation}Prohibited')
     else:
-        dut.expect_gme('Store access fault')
+        dut.expect_gme(f'{operation} access fault')
     dut.expect_reg_dump(0)
     if dut.is_xtensa:
         dut.expect_backtrace()
@@ -384,6 +382,18 @@ def test_storeprohibited(dut: PanicTestDut, config: str, test_func_name: str) ->
     common_test(dut, config, expected_backtrace=get_default_backtrace(test_func_name))
 
 
+@pytest.mark.parametrize('config', CONFIGS, indirect=True)
+@pytest.mark.generic
+def test_storeprohibited(dut: PanicTestDut, config: str, test_func_name: str) -> None:
+    check_x_prohibited(dut, config, test_func_name, 'Store')
+
+
+@pytest.mark.parametrize('config', CONFIGS, indirect=True)
+@pytest.mark.generic
+def test_loadprohibited(dut: PanicTestDut, config: str, test_func_name: str) -> None:
+    check_x_prohibited(dut, config, test_func_name, 'Load')
+
+
 @pytest.mark.parametrize('config', CONFIGS, indirect=True)
 @pytest.mark.generic
 def test_abort(dut: PanicTestDut, config: str, test_func_name: str) -> None: