switch: arm rec EntryPoints need rx. fix arm & dsp cache flush

core/hw/aica/dsp_arm64.cpp

@@ -328,7 +328,7 @@ public:
 		FinalizeCode();
 
 		vmem_platform_flush_cache(
-			GetBuffer()->GetStartAddress<void*>(), GetBuffer()->GetEndAddress<void*>(),
+			GetBuffer()->GetStartAddress<char*>() + rx_offset, GetBuffer()->GetEndAddress<char*>() + rx_offset,
 			GetBuffer()->GetStartAddress<void*>(), GetBuffer()->GetEndAddress<void*>());
 	}
 
@@ -455,7 +455,7 @@ void dsp_recompile()
 void dsp_rec_init()
 {
 #ifdef FEAT_NO_RWX_PAGES
-	verify(vmem_platform_prepare_jit_block(dsp.DynCode, sizeof(dsp.DynCode), (void**)&pCodeBuffer, (uintptr_t *)&rx_offset));
+	verify(vmem_platform_prepare_jit_block(dsp.DynCode, sizeof(dsp.DynCode), (void**)&pCodeBuffer, &rx_offset));
 #else
 	verify(vmem_platform_prepare_jit_block(dsp.DynCode, sizeof(dsp.DynCode), (void**)&pCodeBuffer));
 #endif

core/hw/arm7/arm7_rec.cpp

@@ -562,7 +562,7 @@ void compile()
 	// also the size of the EntryPoints table. This way the dynarec
 	// main loop doesn't have to worry about the actual aica
 	// ram size. The aica ram always wraps to 8 MB anyway.
-	EntryPoints[(pc & (ARAM_SIZE_MAX - 1)) / 4] = (void (*)())rv;
+	EntryPoints[(pc & (ARAM_SIZE_MAX - 1)) / 4] = (void (*)())writeToExec(rv);
 
 	block_ops.clear();
 
@@ -657,7 +657,7 @@ void flush()
 void init()
 {
 #ifdef FEAT_NO_RWX_PAGES
-	verify(vmem_platform_prepare_jit_block(ARM7_TCB, ICacheSize, (void**)&ICache, (uintptr_t *)&rx_offset));
+	verify(vmem_platform_prepare_jit_block(ARM7_TCB, ICacheSize, (void**)&ICache, &rx_offset));
 #else
 	verify(vmem_platform_prepare_jit_block(ARM7_TCB, ICacheSize, (void**)&ICache));
 #endif

core/hw/arm7/arm7_rec_arm64.cpp

@@ -602,7 +602,7 @@ public:
 		FinalizeCode();
 		verify(GetBuffer()->GetCursorOffset() <= GetBuffer()->GetCapacity());
 		vmem_platform_flush_cache(
-				GetBuffer()->GetStartAddress<void*>(), GetBuffer()->GetEndAddress<void*>(),
+				recompiler::writeToExec(GetBuffer()->GetStartAddress<void*>()), recompiler::writeToExec(GetBuffer()->GetEndAddress<void*>()),
 				GetBuffer()->GetStartAddress<void*>(), GetBuffer()->GetEndAddress<void*>());
 		recompiler::advance(GetBuffer()->GetSizeInBytes());
 
@@ -683,7 +683,7 @@ public:
 
 		FinalizeCode();
 		vmem_platform_flush_cache(
-				GetBuffer()->GetStartAddress<void*>(), GetBuffer()->GetEndAddress<void*>(),
+				recompiler::writeToExec(GetBuffer()->GetStartAddress<void*>()), recompiler::writeToExec(GetBuffer()->GetEndAddress<void*>()),
 				GetBuffer()->GetStartAddress<void*>(), GetBuffer()->GetEndAddress<void*>());
 		recompiler::advance(GetBuffer()->GetSizeInBytes());
 	}

core/hw/mem/_vmem.h

@@ -29,7 +29,7 @@ void vmem_platform_destroy();
 bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code_area_rwx);
 // Same as above but uses two address spaces one with RX and RW protections.
 // Note: this function doesnt have to be implemented, it's a fallback for the above one.
-bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code_area_rw, uintptr_t *rx_offset);
+bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code_area_rw, ptrdiff_t *rx_offset);
 // This might not need an implementation (ie x86/64 cpus).
 void vmem_platform_flush_cache(void *icache_start, void *icache_end, void *dcache_start, void *dcache_end);
 

core/hw/sh4/dyna/driver.cpp

@@ -33,7 +33,7 @@ u8 SH4_TCB[CODE_SIZE + TEMP_CODE_SIZE + 4096]
 
 u8* CodeCache;
 u8* TempCodeCache;
-uintptr_t cc_rx_offset;
+ptrdiff_t cc_rx_offset;
 
 u32 LastAddr;
 u32 LastAddr_min;

core/hw/sh4/dyna/ngen.h

@@ -54,7 +54,7 @@
 // (ie. exceptions) are RX pointers. These two macros convert between them by
 // sub/add the pointer offset. CodeCache will point to the RW pointer for simplicity.
 #ifdef FEAT_NO_RWX_PAGES
-	extern uintptr_t cc_rx_offset;
+	extern ptrdiff_t cc_rx_offset;
 	#define CC_RW2RX(ptr) (void*)(((uintptr_t)(ptr)) + cc_rx_offset)
 	#define CC_RX2RW(ptr) (void*)(((uintptr_t)(ptr)) - cc_rx_offset)
 #else

core/linux/libnx_vmem.cpp

@@ -206,13 +206,9 @@ bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code
 }
 
 // Use two addr spaces: need to remap something twice, therefore use allocate_shared_filemem()
-bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code_area_rw, uintptr_t *rx_offset)
+bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code_area_rw, ptrdiff_t *rx_offset)
 {
 	const unsigned size_aligned = ((size + PAGE_SIZE) & (~(PAGE_SIZE-1)));
-	void *ptr_rx = code_area;
-
-	if (ptr_rx != code_area)
-		return false;
 
 	virtmemLock();
 	void* ptr_rw = virtmemFindAslr(size_aligned, 0);
@@ -226,8 +222,8 @@ bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code
 	}
 
 	*code_area_rw = ptr_rw;
-	*rx_offset = (char*)ptr_rx - (char*)ptr_rw;
-	INFO_LOG(DYNAREC, "Info: Using NO_RWX mode, rx ptr: %p, rw ptr: %p, offset: %lu\n", ptr_rx, ptr_rw, (unsigned long)*rx_offset);
+	*rx_offset = (char*)code_area - (char*)ptr_rw;
+	INFO_LOG(DYNAREC, "Info: Using NO_RWX mode, rx ptr: %p, rw ptr: %p, offset: %ld\n", code_area, ptr_rw, (long)*rx_offset);
 
 	return true;
 }

core/linux/posix_vmem.cpp

@@ -258,7 +258,7 @@ bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code
 }
 
 // Use two addr spaces: need to remap something twice, therefore use allocate_shared_filemem()
-bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code_area_rw, uintptr_t *rx_offset) {
+bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code_area_rw, ptrdiff_t *rx_offset) {
 	shmem_fd2 = allocate_shared_filemem(size);
 	if (shmem_fd2 < 0)
 		return false;
@@ -278,7 +278,7 @@ bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code
 
 	*code_area_rw = ptr_rw;
 	*rx_offset = (char*)ptr_rx - (char*)ptr_rw;
-	INFO_LOG(DYNAREC, "Info: Using NO_RWX mode, rx ptr: %p, rw ptr: %p, offset: %lu", ptr_rx, ptr_rw, (unsigned long)*rx_offset);
+	INFO_LOG(DYNAREC, "Info: Using NO_RWX mode, rx ptr: %p, rw ptr: %p, offset: %ld", ptr_rx, ptr_rw, (long)*rx_offset);
 
 	return (ptr_rw != MAP_FAILED);
 }

core/windows/win_vmem.cpp

@@ -206,7 +206,7 @@ static void* mem_file_map(void *addr, unsigned size) {
 }
 
 // Use two addr spaces: need to remap something twice, therefore use CreateFileMapping()
-bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code_area_rw, uintptr_t *rx_offset) {
+bool vmem_platform_prepare_jit_block(void *code_area, unsigned size, void **code_area_rw, ptrdiff_t *rx_offset) {
 	mem_handle2 = CreateFileMapping(INVALID_HANDLE_VALUE, 0, PAGE_EXECUTE_READWRITE, 0, size, 0);
 
 	// Get the RX page close to the code_area