Feat/improve linker scripts

.changesets/refactor-linker-script.md

@@ -0,0 +1,36 @@
+release: major
+summary: Refactor linker script, startup code, and memory model with unified copy/zero tables, configurable ITCM, and weak BoardInit. Also fixes LTO.
+
+## Implemented
+
+### Linker Script (`LinkerScript.ld`)
+- **Unified copy table**: all initialized data sections (ISR vector, ITCM code, `.data`, `.dtcm_rodata`, D1/D2/D3 NC and cached data/rodata) are copied from FLASH to RAM by a single generic loop in `Reset_Handler`. No more hand-written copy per section.
+- **Unified zero table**: all BSS sections (DTCM `.bss`, D1/D2/D3 NC and cached BSS) are zero-initialized by the same generic mechanism.
+- **Configurable ITCM**: `__ITCM_SIZE` and `__ITCM_BUILD` macros (passed via C preprocessor `-D` at build time) control ITCM size and whether `.text` is placed in ITCM instead of FLASH.
+- **ISR vector always in ITCM**: the interrupt vector table is unconditionally placed in ITCMRAM and copied at startup.
+- **DTCM constants**: `.dtcm_rodata` section for constants explicitly placed in DTCM (via `DTCM_RODATA` / `DTCM_RODATA_INLINE` macros), initialized from FLASH by the copy table.
+- **Initialized data for all memory domains**: D1, D2, D3 each have non-cached and cached data/rodata sections (`>RAM_Dx AT> FLASH`) copied by the copy table.
+- **Null-pointer guard**: 32-byte MPU region at `0x00000000` (NO_ACCESS) catches runtime null dereferences. Vector table relocated to a 1024-byte aligned location after the 32 security bytes, loaded in VTOR via `.null_guard` NOLOAD section. Region 12 > region 11 → takes priority.
+- **LMA collision guard**: `BYTE(0)` in every `AT> FLASH` section prevents empty sections from clustering at the same PhysAddr, suppressing STM32_Programmer "overlapping segments" warnings.
+
+### Startup Code (`StartupCode.s`)
+- Replaced per-section manual copy with generic copy-table loop (`__copy_table_start` → `__copy_table_end`).
+- Replaced per-section manual BSS zeroing with generic zero-table loop (`__zero_table_start` → `__zero_table_end`).
+- Removed `Reset_Handler` reliance on hardcoded section symbols; everything is table-driven.
+- Calls `weak BoardInit()` between `SystemInit` and global constructors, allowing early hardware initialization from C++.
+
+### System / Board Init (`System.c`)
+- Moved from template-project to ST-LIB so all projects benefit.
+- Provides `weak BoardInit()` and `weak ConfigurationChecker()` — user overrides in firmware.
+
+### MPU (`MPU.hpp`)
+- Fixed `MPUDomain::Instance::as()` and `construct()`: `Request::e` → `Target.e` for valid constexpr member access.
+- Fixed `static_assert` logic: `is_nc && volatile` → `!is_nc || volatile` (cached buffers no longer trigger the assert).
+- Added null-pointer guard region (region 12, `0x00000000`, 32 bytes, NO_ACCESS).
+- Inline specifiers for all `*_INLINE` macros (since inline things are placed in special COMDAT sections that must be separate one from another and everything else).
+
+## Not implemented (left as weak symbols)
+- `ConfigurationChecker`: a function to validate ITCM size and other build-time configuration at runtime. User can define it to add custom startup checks.
+
+## Migration guide (template-project)
+The corresponding template-project PR adapts `BoardInit` to the new weak function contract, and changes the CMAKE as needed. See the template-project for details.

CMakeLists.txt

@@ -414,6 +414,7 @@ add_library(${STLIB_LIBRARY} OBJECT
   $<$<AND:$<BOOL:${CMAKE_CROSSCOMPILING}>,$<BOOL:${USE_ETHERNET}>>:${HALAL_C_ETH_PHY}>
 
 
+  $<$<BOOL:${CMAKE_CROSSCOMPILING}>:${CMAKE_CURRENT_LIST_DIR}/System.c>
   $<$<BOOL:${CMAKE_CROSSCOMPILING}>:${CPP_UTILITIES_C}>
   $<$<BOOL:${CMAKE_CROSSCOMPILING}>:${CPP_UTILITIES_CPP}>
 
@@ -480,6 +481,8 @@ target_compile_options(${STLIB_LIBRARY} PRIVATE
   $<$<BOOL:${CMAKE_CROSSCOMPILING}>:-fdata-sections>
   $<$<BOOL:${CMAKE_CROSSCOMPILING}>:-fno-exceptions>
 
+  $<$<AND:$<BOOL:${ENABLE_LTO}>,$<NOT:$<CONFIG:DEBUG>>>:-flto=auto>
+
   -Wno-psabi
 
   $<$<COMPILE_LANGUAGE:C>:-w>
@@ -530,10 +533,6 @@ target_include_directories(${STLIB_LIBRARY} PUBLIC
   ${CMAKE_CURRENT_LIST_DIR}/Inc/ST-LIB_HIGH
 )
 
-if(ENABLE_LTO)
-  set_property(TARGET ${STLIB_LIBRARY} PROPERTY INTERPROCEDURAL_OPTIMIZATION_RELEASE ON)
-  set_property(TARGET ${STLIB_LIBRARY} PROPERTY INTERPROCEDURAL_OPTIMIZATION_RELWITHDEBINFO ON)
-endif()
 
 if(STLIB_USE_PCH)
   if(COMMAND target_precompile_headers)

Inc/HALAL/Models/SPI/SPI2.hpp

@@ -1326,7 +1326,7 @@ struct SPIDomain {
     }
 
     template <std::size_t N> struct Init {
-        static inline std::array<Instance, N> instances{};
+        static inline std::array<Instance, N> instances;
 
         static void init(
             std::span<const Config, N> cfgs,

Inc/HALAL/Models/TimerDomain/TimerDomain.hpp

@@ -230,8 +230,6 @@ struct InputCaptureInfo {
     uint32_t frequency;
 };
 
-static ST_LIB::InputCaptureInfo input_capture_info_dummy;
-
 constexpr std::array<uint8_t, 25> create_timer_idxmap() {
     std::array<uint8_t, 25> result{};
 
@@ -284,6 +282,7 @@ struct TimerDomain {
     /* 2x as big as necessary but this makes indexing easier & faster */
     static InputCaptureInfo* input_capture_info[max_instances][input_capture_channels];
     static InputCaptureInfo input_capture_info_backing[max_instances][input_capture_channels];
+    static InputCaptureInfo input_capture_info_dummy;
 
     struct Entry {
         std::array<char, 8> name; /* max length = 7 */

Inc/HALAL/Services/ADC/ADC.hpp

@@ -13,10 +13,10 @@
 
 #include "HALAL/Models/Pin.hpp"
 #ifdef SIM_ON
-#define STLIB_ADC_DMA_BUFFER_ATTR
+#define STLIB_ADC_DMA_BUFFER_ATTR(x) inline
 #else
 #include "HALAL/Models/MPU.hpp"
-#define STLIB_ADC_DMA_BUFFER_ATTR D1_NC
+#define STLIB_ADC_DMA_BUFFER_ATTR(x) D2_NC_BSS_INLINE(x)
 #endif
 
 using std::array;
@@ -780,8 +780,8 @@ struct ADCDomain {
             "ADC DMA buffer size exceeds max_instances"
         );
 
-        alignas(32) STLIB_ADC_DMA_BUFFER_ATTR
-            static inline uint16_t dma_buffer_pool[total_dma_slots > 0 ? total_dma_slots : 1]{};
+        alignas(32) static uint16_t STLIB_ADC_DMA_BUFFER_ATTR(dma_buffer_pool
+        ) dma_buffer_pool[total_dma_slots > 0 ? total_dma_slots : 1]{};
 
         static constexpr bool is_resolved_config(const Config& cfg) {
             return cfg.peripheral != Peripheral::AUTO && cfg.channel != Channel::AUTO;

Inc/HALAL/Services/DFSDM/DFSDM.hpp

@@ -6,7 +6,7 @@
 #include "HALAL/Models/DMA/DMA2.hpp"
 #include "HALAL/Models/MPU.hpp"
 
-#define STLIB_DFSDM_DMA_BUFFER_ATTR D1_NC
+#define STLIB_DFSDM_DMA_BUFFER_ATTR(x) D2_NC_BSS_INLINE(x)
 
 #define Oversampling_MAX 1024
 #define Oversampling_MAX_Filter_4 215
@@ -992,14 +992,14 @@ struct DFSDM_CHANNEL_DOMAIN {
             sizes.filter0 + sizes.filter1 + sizes.filter2 + sizes.filter3;
 
         // Filter Buffers
-        alignas(32) STLIB_DFSDM_DMA_BUFFER_ATTR
-            static inline int32_t Buffer_Filter0[sizes.filter0 > 0 ? sizes.filter0 : 1]{};
-        alignas(32) STLIB_DFSDM_DMA_BUFFER_ATTR
-            static inline int32_t Buffer_Filter1[sizes.filter1 > 0 ? sizes.filter1 : 1]{};
-        alignas(32) STLIB_DFSDM_DMA_BUFFER_ATTR
-            static inline int32_t Buffer_Filter2[sizes.filter2 > 0 ? sizes.filter2 : 1]{};
-        alignas(32) STLIB_DFSDM_DMA_BUFFER_ATTR
-            static inline int32_t Buffer_Filter3[sizes.filter3 > 0 ? sizes.filter3 : 1]{};
+        alignas(32) STLIB_DFSDM_DMA_BUFFER_ATTR(Buffer_Filter0
+        ) static int32_t Buffer_Filter0[sizes.filter0 > 0 ? sizes.filter0 : 1]{};
+        alignas(32) STLIB_DFSDM_DMA_BUFFER_ATTR(Buffer_Filter1
+        ) static int32_t Buffer_Filter1[sizes.filter1 > 0 ? sizes.filter1 : 1]{};
+        alignas(32) STLIB_DFSDM_DMA_BUFFER_ATTR(Buffer_Filter2
+        ) static int32_t Buffer_Filter2[sizes.filter2 > 0 ? sizes.filter2 : 1]{};
+        alignas(32) STLIB_DFSDM_DMA_BUFFER_ATTR(Buffer_Filter3
+        ) static int32_t Buffer_Filter3[sizes.filter3 > 0 ? sizes.filter3 : 1]{};
 
         static inline std::array<Instance, N> instances{};
 

Inc/MockedDrivers/stm32h7xx_hal_mock.h

@@ -1242,6 +1242,7 @@ enum {
 #define MPU_REGION_NUMBER8 (0x08U)
 #define MPU_REGION_NUMBER10 (0x0AU)
 #define MPU_REGION_NUMBER11 (0x0BU)
+#define MPU_REGION_NUMBER12 (0x0CU)
 
 #define MPU_REGION_ENABLE (0x01U)
 #define MPU_REGION_SIZE_4GB (0x1FU)

Inc/ST-LIB.hpp

@@ -385,3 +385,10 @@ template <BoardFaultPolicy FaultPolicyT, auto&... devs> struct Board {
 };
 
 } // namespace ST_LIB
+
+/**
+ * @brief   This is a function that gets called early in the startup process,
+ *          before the global constructors and main() are called.
+ *          It is responsible for initializing the hardware and peripherals
+ */
+extern "C" void BoardInit(void);