diff --git a/cores/arduino/Cache.cpp b/cores/arduino/Cache.cpp
index 31f3c25..38ab726 100644
--- a/cores/arduino/Cache.cpp
+++ b/cores/arduino/Cache.cpp
@@ -242,15 +242,16 @@ bool Cache::Disable() noexcept
 
 #if SAME70
 
+extern "C" [[noreturn]] void vAssertCalled(uint32_t line, const char *file) noexcept;
+
 void Cache::Flush(const volatile void *start, size_t length) noexcept
 {
 	if ((SCB->CCR & SCB_CCR_DC_Msk) != 0)			// if data cache is enabled
 	{
-		// We assume that the DMA buffer is entirely inside or entirely outside the non-cached RAM area
-		if (start < (void*)&_nocache_ram_start || start >= (void*)&_nocache_ram_end)
+		// The DMA buffer should be entirely inside the non-cached RAM area
+		if ((const char *)start < (const char *)&_nocache_ram_start || (const char *)start + length >= (const char *)&_nocache_ram_end)
 		{
-			const uint32_t startAddr = reinterpret_cast<uint32_t>(start);
-			SCB_CleanDCache_by_Addr(reinterpret_cast<uint32_t*>(startAddr & ~3), length + (startAddr & 3));
+			vAssertCalled(__LINE__, __FILE__);
 		}
 	}
 }
@@ -262,13 +263,10 @@ void Cache::Invalidate(const volatile void *start, size_t length) noexcept
 #if SAME70
 	if ((SCB->CCR & SCB_CCR_DC_Msk) != 0)			// if data cache is enabled
 	{
-		// We assume that the DMA buffer is entirely inside or entirely outside the non-cached RAM area
-		if (start < (void*)&_nocache_ram_start || start >= (void*)&_nocache_ram_end)
+		// The DMA buffer should be entirely inside the non-cached RAM area
+		if ((const char *)start < (const char *)&_nocache_ram_start || (const char *)start + length >= (const char *)&_nocache_ram_end)
 		{
-			// Caution! if any part of the cache line is dirty, the written data will be lost!
-			//TODO make this an abort instead of trying to invalidate the cache
-			const uint32_t startAddr = reinterpret_cast<uint32_t>(start);
-			SCB_InvalidateDCache_by_Addr(reinterpret_cast<uint32_t*>(startAddr & ~3), length + (startAddr & 3));
+			vAssertCalled(__LINE__, __FILE__);
 		}
 	}
 #else
diff --git a/cores/arduino/Flash.cpp b/cores/arduino/Flash.cpp
index aa95a00..4b57445 100644
--- a/cores/arduino/Flash.cpp
+++ b/cores/arduino/Flash.cpp
@@ -156,8 +156,7 @@ extern "C" {
  * \param pus_page The first page accessed.
  * \param pus_offset Byte offset in the first page.
  */
-static void translate_address(Efc **pp_efc, uint32_t ul_addr,
-		uint16_t *pus_page, uint16_t *pus_offset)
+static void translate_address(Efc **pp_efc, uint32_t ul_addr, uint16_t *pus_page, uint16_t *pus_offset) noexcept
 {
 	Efc *p_efc;
 	uint16_t us_page;
@@ -228,7 +227,7 @@ static void translate_address(Efc **pp_efc, uint32_t ul_addr,
  * \param us_offset Byte offset inside page.
  * \param pul_addr Computed address (optional).
  */
-static void compute_address(Efc *p_efc, uint16_t us_page, uint16_t us_offset, uint32_t *pul_addr)
+static void compute_address(Efc *p_efc, uint16_t us_page, uint16_t us_offset, uint32_t *pul_addr) noexcept
 {
 	uint32_t ul_addr;
 
@@ -278,7 +277,7 @@ static void compute_address(Efc *p_efc, uint16_t us_page, uint16_t us_offset, ui
  * \param pul_actual_start Actual start address of lock range.
  * \param pul_actual_end Actual end address of lock range.
  */
-static void compute_lock_range(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end)
+static void compute_lock_range(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end) noexcept
 {
 	uint32_t ul_actual_start, ul_actual_end;
 
@@ -303,7 +302,7 @@ static void compute_lock_range(uint32_t ul_start, uint32_t ul_end, uint32_t *pul
  *
  * \return 0 if successful; otherwise returns an error code.
  */
-uint32_t flash_init(uint32_t ul_mode, uint32_t ul_fws)
+uint32_t flash_init(uint32_t ul_mode, uint32_t ul_fws) noexcept
 {
 	efc_init(EFC, ul_mode, ul_fws);
 
@@ -322,7 +321,7 @@ uint32_t flash_init(uint32_t ul_mode, uint32_t ul_fws)
  *
  * \return 0 if successful; otherwise returns an error code.
  */
-uint32_t flash_set_wait_state(uint32_t ul_address, uint32_t ul_fws)
+uint32_t flash_set_wait_state(uint32_t ul_address, uint32_t ul_fws) noexcept
 {
 	Efc *p_efc;
 
@@ -341,7 +340,7 @@ uint32_t flash_set_wait_state(uint32_t ul_address, uint32_t ul_fws)
  *
  * \return The actual descriptor length.
  */
-uint32_t flash_get_descriptor(uint32_t ul_address, uint32_t *pul_flash_descriptor, uint32_t ul_size)
+uint32_t flash_get_descriptor(uint32_t ul_address, uint32_t *pul_flash_descriptor, uint32_t ul_size) noexcept
 {
 	Efc *p_efc;
 	uint32_t ul_tmp;
@@ -377,7 +376,7 @@ uint32_t flash_get_descriptor(uint32_t ul_address, uint32_t *pul_flash_descripto
  *
  * \return The flash total page count.
  */
-uint32_t flash_get_page_count(const uint32_t *pul_flash_descriptor)
+uint32_t flash_get_page_count(const uint32_t *pul_flash_descriptor) noexcept
 {
 	return (pul_flash_descriptor[1] / pul_flash_descriptor[2]);
 }
@@ -392,7 +391,7 @@ uint32_t flash_get_page_count(const uint32_t *pul_flash_descriptor)
  *
  * \return The flash page count per region (plane).
  */
-uint32_t flash_get_page_count_per_region(const uint32_t *pul_flash_descriptor)
+uint32_t flash_get_page_count_per_region(const uint32_t *pul_flash_descriptor) noexcept
 {
 	return (pul_flash_descriptor[4] / pul_flash_descriptor[2]);
 }
@@ -407,7 +406,7 @@ uint32_t flash_get_page_count_per_region(const uint32_t *pul_flash_descriptor)
  *
  * \return The flash region (plane) count.
  */
-uint32_t flash_get_region_count(const uint32_t *pul_flash_descriptor)
+uint32_t flash_get_region_count(const uint32_t *pul_flash_descriptor) noexcept
 {
 	return (pul_flash_descriptor[3]);
 }
@@ -423,7 +422,7 @@ uint32_t flash_get_region_count(const uint32_t *pul_flash_descriptor)
  *
  * \return 0 if successful; otherwise returns an error code.
  */
-uint32_t flash_erase_all(uint32_t ul_address)
+uint32_t flash_erase_all(uint32_t ul_address) noexcept
 {
 	Efc *p_efc;
 
@@ -444,7 +443,7 @@ uint32_t flash_erase_all(uint32_t ul_address)
  *
  * \return 0 if successful; otherwise returns an error code.
  */
-uint32_t flash_erase_page(uint32_t ul_address, uint8_t uc_page_num)
+uint32_t flash_erase_page(uint32_t ul_address, uint8_t uc_page_num) noexcept
 {
 	Efc *p_efc;
 	uint16_t us_page;
@@ -477,7 +476,7 @@ uint32_t flash_erase_page(uint32_t ul_address, uint8_t uc_page_num)
  *
  * \return 0 if successful; otherwise returns an error code.
  */
-uint32_t flash_erase_sector(uint32_t ul_address)
+uint32_t flash_erase_sector(uint32_t ul_address) noexcept
 {
 	Efc *p_efc;
 	uint16_t us_page;
@@ -508,7 +507,7 @@ uint32_t flash_erase_sector(uint32_t ul_address)
  *
  * \return 0 if successful, otherwise returns an error code.
  */
-uint32_t flash_write(uint32_t ul_address, const void *p_buffer, uint32_t ul_size, uint32_t ul_erase_flag)
+uint32_t flash_write(uint32_t ul_address, const void *p_buffer, uint32_t ul_size, uint32_t ul_erase_flag) noexcept
 {
 	Efc *p_efc;
 	uint32_t ul_fws_temp;
@@ -603,7 +602,7 @@ uint32_t flash_write(uint32_t ul_address, const void *p_buffer, uint32_t ul_size
  *
  * \return 0 if successful, otherwise returns an error code.
  */
-uint32_t flash_lock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end)
+uint32_t flash_lock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end) noexcept
 {
 	Efc *p_efc;
 	uint32_t ul_actual_start, ul_actual_end;
@@ -651,7 +650,7 @@ uint32_t flash_lock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_sta
  *
  * \return 0 if successful, otherwise returns an error code.
  */
-uint32_t flash_unlock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end)
+uint32_t flash_unlock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end) noexcept
 {
 	Efc *p_efc;
 	uint32_t ul_actual_start, ul_actual_end;
@@ -694,7 +693,7 @@ uint32_t flash_unlock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_s
  *
  * \return The number of locked regions inside the given address range.
  */
-uint32_t flash_is_locked(uint32_t ul_start, uint32_t ul_end)
+uint32_t flash_is_locked(uint32_t ul_start, uint32_t ul_end) noexcept
 {
 	Efc *p_efc;
 	uint16_t us_start_page, us_end_page;
@@ -771,7 +770,7 @@ uint32_t flash_is_locked(uint32_t ul_start, uint32_t ul_end)
  *
  * \return 0 if successful; otherwise returns an error code.
  */
-uint32_t flash_set_gpnvm(uint32_t ul_gpnvm)
+uint32_t flash_set_gpnvm(uint32_t ul_gpnvm) noexcept
 {
 	if (ul_gpnvm >= GPNVM_NUM_MAX) {
 		return FLASH_RC_INVALID;
@@ -795,7 +794,7 @@ uint32_t flash_set_gpnvm(uint32_t ul_gpnvm)
  *
  * \return 0 if successful; otherwise returns an error code.
  */
-uint32_t flash_clear_gpnvm(uint32_t ul_gpnvm)
+uint32_t flash_clear_gpnvm(uint32_t ul_gpnvm) noexcept
 {
 	if (ul_gpnvm >= GPNVM_NUM_MAX) {
 		return FLASH_RC_INVALID;
@@ -832,7 +831,7 @@ uint32_t flash_read_gpnvm_bits() noexcept
  * \retval 0 If the given GPNVM bit is currently cleared.
  * otherwise returns an error code.
  */
-uint32_t flash_is_gpnvm_set(uint32_t ul_gpnvm)
+uint32_t flash_is_gpnvm_set(uint32_t ul_gpnvm) noexcept
 {
 	uint32_t ul_gpnvm_bits;
 
diff --git a/cores/arduino/Flash.h b/cores/arduino/Flash.h
index 3e221cf..c7d1d66 100644
--- a/cores/arduino/Flash.h
+++ b/cores/arduino/Flash.h
@@ -94,26 +94,26 @@ typedef enum flash_farg_page_num {
 #define FLASH_ACCESS_MODE_64     EFC_ACCESS_MODE_64
 //! @}
 
-uint32_t flash_init(uint32_t ul_mode, uint32_t ul_fws);
-uint32_t flash_set_wait_state(uint32_t ul_address, uint32_t ul_fws);
-uint32_t flash_get_descriptor(uint32_t ul_address, uint32_t *pul_flash_descriptor,	uint32_t ul_size);
-uint32_t flash_get_page_count(const uint32_t *pul_flash_descriptor);
-uint32_t flash_get_page_count_per_region(const uint32_t *pul_flash_descriptor);
-uint32_t flash_get_region_count(const uint32_t *pul_flash_descriptor);
-uint32_t flash_erase_all(uint32_t ul_address);
+uint32_t flash_init(uint32_t ul_mode, uint32_t ul_fws) noexcept;
+uint32_t flash_set_wait_state(uint32_t ul_address, uint32_t ul_fws) noexcept;
+uint32_t flash_get_descriptor(uint32_t ul_address, uint32_t *pul_flash_descriptor,	uint32_t ul_size) noexcept;
+uint32_t flash_get_page_count(const uint32_t *pul_flash_descriptor) noexcept;
+uint32_t flash_get_page_count_per_region(const uint32_t *pul_flash_descriptor) noexcept;
+uint32_t flash_get_region_count(const uint32_t *pul_flash_descriptor) noexcept;
+uint32_t flash_erase_all(uint32_t ul_address) noexcept;
 
 #if (SAM4S || SAM4E || SAM4N || SAM4C || SAMG || SAM4CP || SAM4CM || SAMV71 || SAMV70 || SAMS70 || SAME70)
-uint32_t flash_erase_page(uint32_t ul_address, uint8_t uc_page_num);
-uint32_t flash_erase_sector(uint32_t ul_address);
+uint32_t flash_erase_page(uint32_t ul_address, uint8_t uc_page_num) noexcept;
+uint32_t flash_erase_sector(uint32_t ul_address) noexcept;
 #endif
 
-uint32_t flash_write(uint32_t ul_address, const void *p_buffer, uint32_t ul_size, uint32_t ul_erase_flag);
-uint32_t flash_lock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end);
-uint32_t flash_unlock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end);
-uint32_t flash_is_locked(uint32_t ul_start, uint32_t ul_end);
-uint32_t flash_set_gpnvm(uint32_t ul_gpnvm);
-uint32_t flash_clear_gpnvm(uint32_t ul_gpnvm);
-uint32_t flash_is_gpnvm_set(uint32_t ul_gpnvm);
+uint32_t flash_write(uint32_t ul_address, const void *p_buffer, uint32_t ul_size, uint32_t ul_erase_flag) noexcept;
+uint32_t flash_lock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end) noexcept;
+uint32_t flash_unlock(uint32_t ul_start, uint32_t ul_end, uint32_t *pul_actual_start, uint32_t *pul_actual_end) noexcept;
+uint32_t flash_is_locked(uint32_t ul_start, uint32_t ul_end) noexcept;
+uint32_t flash_set_gpnvm(uint32_t ul_gpnvm) noexcept;
+uint32_t flash_clear_gpnvm(uint32_t ul_gpnvm) noexcept;
+uint32_t flash_is_gpnvm_set(uint32_t ul_gpnvm) noexcept;
 uint32_t flash_read_gpnvm_bits() noexcept;
 uint32_t flash_read_unique_id(uint32_t *pul_data) noexcept;		// read 4 dwords of unique ID
 
diff --git a/variants/same70/startup_same70.c b/variants/same70/startup_same70.c
index 149ae03..21399d3 100644
--- a/variants/same70/startup_same70.c
+++ b/variants/same70/startup_same70.c
@@ -319,7 +319,7 @@ const DeviceVectors exception_table = {
         .pfnRSWDT_Handler  = (void*) RSWDT_Handler   /* 63 Reinforced Secure Watchdog Timer */
 };
 
-// This must be marked noinline so that R0 is loaded with the required value for SP
+// This must be marked noinline so that R0 is loaded with the required value for SP. Return is via LR so it's OK to return with a different SP.
 __attribute((noinline)) void SetStackPointer(uint32_t *topOfStack)
 {
 	__asm volatile("msr msp, r0");
@@ -329,7 +329,7 @@ __attribute((noinline)) void SetStackPointer(uint32_t *topOfStack)
  * \brief This is the code that gets called on processor reset.
  * To initialize the device, and call the main() routine.
  */
-__attribute__((noreturn)) void Reset_Handler(void)
+__attribute__((noreturn, naked)) void Reset_Handler(void)
 {
 	// If TCM is allocated then SP may point beyond the end of RAM, so move it
 	SetStackPointer(&_ezero_nocache);
@@ -348,6 +348,12 @@ __attribute__((noreturn)) void Reset_Handler(void)
 		}
 	}
 
+	/* Clear the zero segment */
+	for (uint32_t *pDest = &_szero; pDest < &_ezero;)
+	{
+		*pDest++ = 0;
+	}
+
 	// Check that no TCM is allocated before we relocate the stack to the top of memory. This uses a RAMFUNC, so we must initialise the relocate segment before here.
 	// The temporary stack we are on is in the non-cached memory segment, so don't clear that segment until after we have relocated the stack.
 	if (flash_read_gpnvm_bits() & ((1ul << 7) | (1ul << 8)))
@@ -372,13 +378,7 @@ __attribute__((noreturn)) void Reset_Handler(void)
 		*pDest++ = 0;
 	}
 
-	/* Clear the zero segment */
-	for (uint32_t *pDest = &_szero; pDest < &_ezero;)
-	{
-		*pDest++ = 0;
-	}
-
-	/* Set the vector table base address */
+	// Set the vector table base address
 	const uint32_t * const pSrc = (uint32_t *) & _sfixed;
 	SCB->VTOR = ((uint32_t) pSrc & SCB_VTOR_TBLOFF_Msk);
 
@@ -401,6 +401,7 @@ __attribute__((noreturn)) void Reset_Handler(void)
  */
 void Dummy_Handler(void)
 {
-	while (1) {
-	}
+	while (1) { }
 }
+
+// End