stm32h5: TRACE_ETM support and stm32h563nucleo reference

H5 hangs its debug AP if trace CoreSight is touched unclocked (recover =
power-cycle): the reference's AfterTargetConnect clocks the DBGMCU trace
domain but defers IOEN to firmware, or the mid-boot clock switch desyncs
the decoder. Stock solder bridges make the CN5 path marginal: validated
config is 100 MHz core, width 1, +5 ns (board.h selects the reduced clock
for TRACE_ETM builds); width 4 / 250 MHz retest waits on SB removal.

hw/bsp/stm32h5/boards/stm32h563nucleo/board.h

@@ -88,7 +88,12 @@ static inline void SystemClock_Config(void) {
   RCC_OscInitStruct.PLL.PLLState = RCC_PLL_ON;
   RCC_OscInitStruct.PLL.PLLSource = RCC_PLL1_SOURCE_HSE;
   RCC_OscInitStruct.PLL.PLLM = 4;
+  #ifdef TRACE_ETM
+  RCC_OscInitStruct.PLL.PLLN = 100; // 100 MHz core: the Nucleo trace path (CN5 via solder bridges)
+                                    // corrupts the trace stream at higher TRACECLK (= SYSCLK/2)
+  #else
   RCC_OscInitStruct.PLL.PLLN = 250;
+  #endif
   RCC_OscInitStruct.PLL.PLLP = 2;
   RCC_OscInitStruct.PLL.PLLQ = 2;
   RCC_OscInitStruct.PLL.PLLR = 2;

hw/bsp/stm32h5/boards/stm32h563nucleo/ozone/stm32h563.jdebug

@@ -0,0 +1,113 @@
+/*********************************************************************
+*
+*       OnProjectLoad
+*
+* Function description
+*   Project load routine. Required.
+*
+* Notes
+*   Nucleo-H563ZI trace: PE2-PE6 reach the MIPI-20 connector (CN5) only
+*   with SB8, SB9, SB64, SB68, SB70, SB71, SB78 removed (MB1404 UM).
+*   Firmware must be built with TRACE_ETM=1 (trace pin + DBGMCU init).
+*
+*   The trace path through the solder bridges is signal-marginal: reliable
+*   only at 100 MHz core (TRACE_ETM builds select this automatically in
+*   board.h), port width 1 and +5 ns sample timing (validated empirically;
+*   width 4 or 250 MHz core corrupts the stream within ~100 ms).
+*
+**********************************************************************
+*/
+void OnProjectLoad (void) {
+  Project.SetTraceSource ("Trace Pins");
+  Project.SetTracePortWidth (1);
+  Project.SetTraceTiming (5000, 5000, 5000, 5000);
+  Project.SetSWO (0);
+  Edit.SysVar (VAR_TRACE_CORE_CLOCK, 100000000);
+  Project.AddSvdFile ("$(InstallDir)/Config/CPU/Cortex-M33F.svd");
+
+  Project.SetDevice ("STM32H563ZI");
+  Project.SetHostIF ("USB", "");
+  Project.SetTargetIF ("SWD");
+  Project.SetTIFSpeed ("4 MHz");
+
+  File.Open ("../../../../../../examples/cmake-build-stm32h563nucleo/device/cdc_msc/cdc_msc.elf");
+}
+
+/*********************************************************************
+*
+*       AfterTargetConnect
+*
+* Function description
+*   Enable the trace clock domain (DBGMCU_CR: TRACE_IOEN | TRACE_CLKEN |
+*   TRACE_MODE=4-bit) BEFORE Ozone touches the trace CoreSight components.
+*   On STM32H5 an access to unclocked trace components hangs the debug AP
+*   ("Failed to read target status") until the board is power-cycled.
+*
+**********************************************************************
+*/
+void AfterTargetConnect (void) {
+  unsigned int cr;
+  cr = Target.ReadU32(0x44024004);   // DBGMCU_CR
+  // clock the domain (CLKEN|MODE=4-bit) but keep the pins OFF (clear IOEN):
+  // the pins must only go live via firmware trace_etm_init() AFTER the system
+  // clock switch, or the mid-trace frequency change desyncs the decoder.
+  Target.WriteU32(0x44024004, (cr & 0xFFFFFFEF) | 0xE0);
+}
+
+/*********************************************************************
+*
+*       AfterTargetReset
+*
+* Function description
+*   Event handler routine.
+*    - Sets the PC register to program reset value.
+*    - Sets the SP register to program reset value on Cortex-M.
+*
+**********************************************************************
+*/
+void AfterTargetReset (void) {
+  unsigned int SP;
+  unsigned int PC;
+  unsigned int VectorTableAddr;
+
+  VectorTableAddr = Elf.GetBaseAddr();
+
+  if (VectorTableAddr == 0xFFFFFFFF) {
+    Util.Log("Project file error: failed to get program base");
+  } else {
+    SP = Target.ReadU32(VectorTableAddr);
+    Target.SetReg("SP", SP);
+
+    PC = Target.ReadU32(VectorTableAddr + 4);
+    Target.SetReg("PC", PC);
+  }
+}
+
+/*********************************************************************
+*
+*       AfterTargetDownload
+*
+* Function description
+*   Event handler routine.
+*    - Sets the PC register to program reset value.
+*    - Sets the SP register to program reset value on Cortex-M.
+*
+**********************************************************************
+*/
+void AfterTargetDownload (void) {
+  unsigned int SP;
+  unsigned int PC;
+  unsigned int VectorTableAddr;
+
+  VectorTableAddr = Elf.GetBaseAddr();
+
+  if (VectorTableAddr == 0xFFFFFFFF) {
+    Util.Log("Project file error: failed to get program base");
+  } else {
+    SP = Target.ReadU32(VectorTableAddr);
+    Target.SetReg("SP", SP);
+
+    PC = Target.ReadU32(VectorTableAddr + 4);
+    Target.SetReg("PC", PC);
+  }
+}

hw/bsp/stm32h5/family.c

@@ -99,6 +99,24 @@ static UART_HandleTypeDef UartHandle = {
 };
 #endif
 
+#ifdef TRACE_ETM
+static void trace_etm_init(void) {
+  // H5 trace pins are PE2 to PE6 (Nucleo-144: requires trace solder-bridge config, see board docs)
+  GPIO_InitTypeDef gpio_init;
+  gpio_init.Pin       = GPIO_PIN_2 | GPIO_PIN_3 | GPIO_PIN_4 | GPIO_PIN_5 | GPIO_PIN_6;
+  gpio_init.Mode      = GPIO_MODE_AF_PP;
+  gpio_init.Pull      = GPIO_PULLUP;
+  gpio_init.Speed     = GPIO_SPEED_FREQ_VERY_HIGH;
+  gpio_init.Alternate = GPIO_AF0_TRACE;
+  HAL_GPIO_Init(GPIOE, &gpio_init);
+
+  // Enable trace port + clock, synchronous 4-bit mode
+  DBGMCU->CR |= DBGMCU_CR_TRACE_IOEN | DBGMCU_CR_TRACE_CLKEN | DBGMCU_CR_TRACE_MODE;
+}
+#else
+#define trace_etm_init()
+#endif
+
 void board_init(void) {
   // Cache UID before ICACHE is enabled (STM32H5 errata: reading UID_BASE with ICACHE causes hard fault)
   volatile uint32_t* stm32_uuid = (volatile uint32_t*) UID_BASE;
@@ -127,6 +145,8 @@ void board_init(void) {
   __HAL_RCC_GPIOI_CLK_ENABLE();
   #endif
 
+  trace_etm_init();
+
   #if CFG_TUSB_OS == OPT_OS_NONE
   // 1ms tick timer
   SysTick_Config(SystemCoreClock / 1000);