Experiment with supporting Windows Arm64 natively

So far I haven't found any way to run native Arm64 code on Windows Arm64
without using MSVC. When I build a PE binary from scratch that should be
a valid Windows Arm64 program, the OS refuses to run it. Possibly due to
requiring additional content like XML manifests or relocation or control
flow integrity data that isn't normally required on x64. I've also tried
using VirtualAlloc2() to JIT an Arm64 native function, but VirtualAlloc2
always fails with invalid parameter. I tried using MSVC to create an ARM
DLL that my x64 emulated program can link at runtime, to pass a function
pointer with ARM code, but LoadLibrary() rejects ARM DLLs as invalid exe

The only option left, is likely to write a new program like ape/ape-m1.c
which can be compiled by MSVC to load and run an AARCH64 ELF executable.
The emulated x64 binary would detect emulation using IsWow64Process2 and
then drop the loader executable in a temporary folder, and re-launch the
original executable, using the Arm64 segments of the cosmocc fat binary.

examples/BUILD.mk

@@ -53,6 +53,7 @@ EXAMPLES_DIRECTDEPS =						\
 	LIBC_NEXGEN32E						\
 	LIBC_NT_ADVAPI32					\
 	LIBC_NT_IPHLPAPI					\
+	LIBC_NT_MEMORY						\
 	LIBC_NT_KERNEL32					\
 	LIBC_NT_NTDLL						\
 	LIBC_NT_USER32						\

libc/nt/enum/pageflags.h

@@ -23,4 +23,7 @@
 #define kNtSecLargePages     0x80000000
 #define kNtSecWritecombine   0x40000000
 
+#define kNtPageTargetsInvalid  0x40000000
+#define kNtPageTargetsNoUpdate 0x40000000
+
 #endif /* COSMOPOLITAN_LIBC_NT_ENUM_PAGEFLAGS_H_ */

libc/nt/kernel32/IsWow64Process2.S

@@ -0,0 +1,18 @@
+#include "libc/nt/codegen.h"
+.imp	kernel32,__imp_IsWow64Process2,IsWow64Process2
+
+	.text.windows
+	.ftrace1
+IsWow64Process2:
+	.ftrace2
+#ifdef __x86_64__
+	push	%rbp
+	mov	%rsp,%rbp
+	mov	__imp_IsWow64Process2(%rip),%rax
+	jmp	__sysv2nt
+#elif defined(__aarch64__)
+	mov	x0,#0
+	ret
+#endif
+	.endfn	IsWow64Process2,globl
+	.previous

libc/nt/master.sh

@@ -113,6 +113,7 @@ imp	'GetCurrentProcessId'					GetCurrentProcessId					kernel32	0
 imp	'GetCurrentProcessorNumberEx'				GetCurrentProcessorNumberEx				kernel32	1
 imp	'GetCurrentThread'					GetCurrentThread					kernel32	0
 imp	'GetCurrentThreadId'					GetCurrentThreadId					kernel32	0
+imp	'GetDynamicTimeZoneInformation'				GetDynamicTimeZoneInformation				kernel32	1
 imp	'GetEnvironmentStrings'					GetEnvironmentStringsW					kernel32	1
 imp	'GetEnvironmentVariable'				GetEnvironmentVariableW					kernel32	3
 imp	'GetExitCodeThread'					GetExitCodeThread					kernel32	2
@@ -168,8 +169,6 @@ imp	'GetSystemTimePreciseAsFileTime'			GetSystemTimePreciseAsFileTime				kernel3
 imp	'GetSystemTimes'					GetSystemTimes						kernel32	3
 imp	'GetTempPath'						GetTempPathW						kernel32	2
 imp	'GetTempPathA'						GetTempPathA						kernel32	2
-imp	'GetDynamicTimeZoneInformation'				GetDynamicTimeZoneInformation				kernel32	1
-imp	'GetTimeZoneInformation'				GetTimeZoneInformation					kernel32	1
 imp	'GetThreadContext'					GetThreadContext					kernel32	2
 imp	'GetThreadDescription'					GetThreadDescription					kernel32	2
 imp	'GetThreadIOPendingFlag'				GetThreadIOPendingFlag					kernel32	2
@@ -178,6 +177,7 @@ imp	'GetThreadPriority'					GetThreadPriority					kernel32	1
 imp	'GetThreadPriorityBoost'				GetThreadPriorityBoost					kernel32	2
 imp	'GetThreadTimes'					GetThreadTimes						kernel32	5
 imp	'GetTickCount64'					GetTickCount64						kernel32	0
+imp	'GetTimeZoneInformation'				GetTimeZoneInformation					kernel32	1
 imp	'GetVersionEx'						GetVersionExW						kernel32	1
 imp	'GetVolumeInformationByHandle'				GetVolumeInformationByHandleW				kernel32	8
 imp	'GetVolumePathName'					GetVolumePathNameW					kernel32	3
@@ -197,6 +197,7 @@ imp	'InitializeCriticalSection'				InitializeCriticalSection				kernel32	1
 imp	'InitializeCriticalSectionAndSpinCount'			InitializeCriticalSectionAndSpinCount			kernel32	2
 imp	'InitializeProcThreadAttributeList'			InitializeProcThreadAttributeList			kernel32	4
 imp	'InitializeSRWLock'					InitializeSRWLock					kernel32	1
+imp	'IsWow64Process2'					IsWow64Process2						kernel32	3
 imp	'LeaveCriticalSection'					LeaveCriticalSection					kernel32	1
 imp	'LoadLibrary'						LoadLibraryW						kernel32	1
 imp	'LoadLibraryA'						LoadLibraryA						kernel32	1

libc/nt/runtime.h

@@ -43,6 +43,8 @@ bool32 SetDefaultDllDirectories(unsigned dirflags);
 bool32 ProcessPrng(void *RandomBuffer, uint32_t RandomBufferLength);
 uint32_t GetModuleFileName(int64_t hModule, char16_t *lpFilename,
                            uint32_t nSize);
+bool32 IsWow64Process2(intptr_t hProcess, uint16_t *out_pProcessMachine,
+                       uint16_t *out_opt_pNativeMachine);
 
 #if ShouldUseMsabiAttribute()
 #include "libc/nt/thunk/runtime.inc"

libc/nt/struct/arm64.h

@@ -0,0 +1,20 @@
+#ifndef COSMOPOLITAN_LIBC_NT_STRUCT_ARM64_H_
+#define COSMOPOLITAN_LIBC_NT_STRUCT_ARM64_H_
+
+struct NtArm64RuntimeFunction {
+  uint32_t BeginAddress;
+  union {
+    uint32_t UnwindData;
+    struct {
+      uint32_t Flag : 2;
+      uint32_t FunctionLength : 11;
+      uint32_t RegF : 3;
+      uint32_t RegI : 4;
+      uint32_t H : 1;
+      uint32_t CR : 2;
+      uint32_t FrameSize : 9;
+    };
+  };
+};
+
+#endif /* COSMOPOLITAN_LIBC_NT_STRUCT_ARM64_H_ */

libc/nt/struct/memextendedparameter.h

@@ -9,26 +9,28 @@
 #define kNtMemExtendedParameterPartitionHandle     3
 #define kNtMemExtendedParameterUserPhysicalHandle  4
 #define kNtMemExtendedParameterAttributeFlags      5
-#define kNtMemExtendedParameterMax                 6
+#define kNtMemExtendedParameterImageMachine        6
+#define kNtMemExtendedParameterMax                 7
 
 #define kNtMemExtendedParameterGraphics          0x00000001
 #define kNtMemExtendedParameterNonpaged          0x00000002
 #define kNtMemExtendedParameterZeroPagesOptional 0x00000004
 #define kNtMemExtendedParameterNonpagedLarge     0x00000008
 #define kNtMemExtendedParameterNonpagedHuge      0x00000010
+#define kNtMemExtendedParameterSoftFaultPages    0x00000020
+#define kNtMemExtendedParameterEcCode            0x00000040
+#define kNtMemExtendedParameterImageNoHpat       0x00000080
 
 struct NtMemExtendedParameter {
-  struct {
+  uint8_t Type;
-    uint64_t Type : kNtMemExtendedParameterTypeBits;
+  uint8_t Reserved[7];
-    uint64_t Reserved : 64 - kNtMemExtendedParameterTypeBits;
-  } DUMMYSTRUCTNAME;
   union {
     uint64_t ULong64;
     void *Pointer;
     size_t Size;
     intptr_t Handle;
     unsigned ULong;
-  } DUMMYUNIONNAME;
+  };
 };
 
 #endif /* COSMOPOLITAN_LIBC_NT_STRUCT_MEMEXTENDEDPARAMETER_H_ */

tool/build/elf2pe.c

@@ -229,6 +229,17 @@ static struct Segment *NewSegment(void) {
   return s;
 }
 
+static int ConvertElfMachineToPe(struct Elf *elf) {
+  switch (elf->ehdr->e_machine) {
+    case EM_NEXGEN32E:
+      return kNtImageFileMachineNexgen32e;
+    case EM_AARCH64:
+      return kNtImageFileMachineArm64;
+    default:
+      Die(elf->path, "unsupported e_machine");
+  }
+}
+
 static Elf64_Addr RelocateVaddrWithinSegment(struct Elf *elf,
                                              Elf64_Addr vaddr_old,
                                              struct Segment *segment) {
@@ -811,7 +822,17 @@ static uint32_t GetPeSectionCharacteristics(struct Segment *s) {
 // originally in the elf image that ld linked. in order for this to work
 // the executable needs to be linked in `ld -q` mode, since it'll retain
 // the .rela sections we'll need later to fixup the binary.
-static struct ImagePointer GeneratePe(struct Elf *elf, char *fp, int64_t vp) {
+static struct ImagePointer GeneratePe(struct Elf *elf, char *fp) {
+
+  int64_t vp = 0;
+  Elf64_Phdr *phdr;
+  for (int i = 0; i < elf->ehdr->e_phnum; ++i) {
+    if ((phdr = GetElfProgramHeaderAddress(elf->ehdr, elf->size, i)) &&
+        phdr->p_type == PT_LOAD) {
+      vp = phdr->p_vaddr;
+      break;
+    }
+  }
 
   Elf64_Sym *entry;
   if (!(entry = FindGlobal(elf, "__win32_start")) &&
@@ -855,7 +876,7 @@ static struct ImagePointer GeneratePe(struct Elf *elf, char *fp, int64_t vp) {
   struct NtImageFileHeader *filehdr;
   filehdr = (struct NtImageFileHeader *)fp;
   fp += sizeof(struct NtImageFileHeader);
-  filehdr->Machine = kNtImageFileMachineNexgen32e;
+  filehdr->Machine = ConvertElfMachineToPe(elf);
   filehdr->TimeDateStamp = 1690072024;
   filehdr->Characteristics =
       kNtPeFileExecutableImage | kNtImageFileLargeAddressAware |
@@ -873,7 +894,9 @@ static struct ImagePointer GeneratePe(struct Elf *elf, char *fp, int64_t vp) {
   opthdr->FileAlignment = 512;
   opthdr->SectionAlignment = MAX(4096, elf->align);
   opthdr->MajorOperatingSystemVersion = 6;
+  opthdr->MinorOperatingSystemVersion = 2;
   opthdr->MajorSubsystemVersion = 6;
+  opthdr->MinorSubsystemVersion = 2;
   opthdr->Subsystem = kNtImageSubsystemWindowsCui;
   opthdr->DllCharacteristics = kNtImageDllcharacteristicsNxCompat |
                                kNtImageDllcharacteristicsHighEntropyVa;
@@ -1116,7 +1139,7 @@ int main(int argc, char *argv[]) {
   // translate executable
   struct Elf *elf = OpenElf(argv[optind]);
   char *buf = Memalign(MAX_ALIGN, 134217728);
-  struct ImagePointer ip = GeneratePe(elf, buf, 0x00400000);
+  struct ImagePointer ip = GeneratePe(elf, buf);
   if (creat(outpath, 0755) == -1)
     DieSys(elf->path);
   Pwrite(3, buf, ip.fp - buf, 0);

tool/build/elf2pe.h

@@ -1,8 +1,8 @@
 #ifndef COSMOPOLITAN_TOOL_BUILD_ELF2PE_H_
 #define COSMOPOLITAN_TOOL_BUILD_ELF2PE_H_
 
-#define __dll_import(DLL, RET, FUNC, ARGS)                      \
+#define __dll_import(DLL, RET, FUNC, ARGS)                  \
-  extern RET(*const __attribute__((__ms_abi__, __weak__)) FUNC) \
+  extern RET(*const __msabi __attribute__((__weak__)) FUNC) \
       ARGS __asm__("\"dll$" DLL "$" #FUNC "\"")
 
 #endif /* COSMOPOLITAN_TOOL_BUILD_ELF2PE_H_ */

tool/hello/BUILD.mk

@@ -79,7 +79,7 @@ o/$(MODE)/tool/hello/hello-pe.ape:			\
 # elf2pe can generate binaries that don't have dll imports
 o/$(MODE)/tool/hello/life-pe.dbg:			\
 		o/$(MODE)/tool/hello/life-pe.o
-	@$(COMPILE) -ALINK.elf $(LINK) $(LINKARGS) $(OUTPUT_OPTION) -q -e WinMain
+	@$(COMPILE) -ALINK.elf $(LINK) $(LINKARGS) $(OUTPUT_OPTION) -q -e WinMain #-Ttext-segment=0x140000000
 o/$(MODE)/tool/hello/life-pe.ape:			\
 		o/$(MODE)/tool/hello/life-pe.dbg	\
 		o/$(MODE)/tool/build/elf2pe

tool/hello/hello-pe.c

@@ -7,6 +7,7 @@
 │   • http://creativecommons.org/publicdomain/zero/1.0/            │
 ╚─────────────────────────────────────────────────────────────────*/
 #endif
+#include "libc/nt/thunk/msabi.h"
 #include "tool/build/elf2pe.h"
 
 #define STD_OUTPUT_HANDLE -11u
@@ -15,7 +16,7 @@ __dll_import("kernel32.dll", long, GetStdHandle, (unsigned));
 __dll_import("kernel32.dll", int, WriteFile,
              (long, const void *, unsigned, unsigned *, void *));
 
-__attribute__((__ms_abi__)) long WinMain(void) {
+__msabi long WinMain(void) {
   WriteFile(GetStdHandle(STD_OUTPUT_HANDLE), "hello world\n", 12, 0, 0);
   return 0;
 }

tool/hello/life-pe.c

@@ -7,7 +7,8 @@
 │   • http://creativecommons.org/publicdomain/zero/1.0/            │
 ╚─────────────────────────────────────────────────────────────────*/
 #endif
+#include "libc/nt/thunk/msabi.h"
 
-__attribute__((__ms_abi__)) long WinMain(void) {
+__msabi long WinMain(void) {
   return 42 << 8;
 }