Fix for Windows arm64

CMakeLists.txt

@@ -15,6 +15,9 @@ if(NOT FUNCHOOK_CPU)
   if (FUNCHOOK_CPU MATCHES "aarch64")
       set(FUNCHOOK_CPU arm64)
   endif ()
+  if (FUNCHOOK_CPU MATCHES "ARM64")
+      set(FUNCHOOK_CPU arm64)
+  endif ()
 endif ()
 
 if (FUNCHOOK_CPU MATCHES "arm64")

src/funchook.c

@@ -282,7 +282,7 @@ static int funchook_prepare_internal(funchook_t *funchook, void **target_func, v
     for (i = 0; i < LITERAL_POOL_NUM; i++) {
         size_t *addr = (size_t*)(entry->trampoline + LITERAL_POOL_OFFSET + i * 2);
         if (*addr != 0) {
-            funchook_log(funchook, "    %016lx : 0x%lx\n", (size_t)addr, *addr);
+            funchook_log(funchook, "    "ADDR_FMT" : 0x%"PRIxPTR"\n", (size_t)addr, *addr);
         }
     }
 #endif

src/funchook_arm64.c

@@ -102,8 +102,8 @@ static int to_regno(funchook_t *funchook, uint32_t avail_regs, uint32_t *regno)
 static int funchook_write_jump32(funchook_t *funchook, const uint32_t *src, const uint32_t *dst, uint32_t *out)
 {
     intptr_t imm = ROUND_DOWN((size_t)dst, PAGE_SIZE) - ROUND_DOWN((size_t)src, PAGE_SIZE);
-    size_t immlo = (imm >> 12) & 0x03;
-    size_t immhi = (imm >> 14) & 0x7FFFFul;
+    uint32_t immlo = (uint32_t)(imm >> 12) & 0x03;
+    uint32_t immhi = (uint32_t)(imm >> 14) & 0x7FFFFul;
 
     /* adrp x9, dst */
     out[0] = 0x90000009 | (immlo << 29) | (immhi << 5);
@@ -170,7 +170,7 @@ int funchook_make_trampoline(funchook_t *funchook, ip_displacement_t *disp, cons
     size_t *literal_pool = (size_t*)(trampoline + LITERAL_POOL_OFFSET);
 
 #define LDR_ADDR(regno, addr) do { \
-    int imm19__ = ((size_t)literal_pool - (size_t)ctx.dst) >> 2; \
+    int imm19__ = (int)((size_t)literal_pool - (size_t)ctx.dst) >> 2; \
     *(literal_pool++) = (addr); \
     *(ctx.dst++) = 0x58000000 | TO_IMM19(imm19__) | (regno); \
 } while (0)
@@ -319,6 +319,13 @@ int funchook_make_trampoline(funchook_t *funchook, ip_displacement_t *disp, cons
             break;
         }
         ctx.src++;
+
+        // special case
+        if ((func[0] & 0xFC000000) == 0x14000000 && (func[1] & 0xFFFF0000) == 0) {
+            // The first instruction is B (unconditional jump).
+            // The second is UDF (permanently undefined).
+            ctx.src = func + 2;
+        }
         if (ctx.src - func >= JUMP32_SIZE) {
             rv = to_regno(funchook, avail_regs, &regno);
             if (rv != 0) {

src/funchook_internal.h

@@ -66,7 +66,7 @@
 #define ADDR_FMT "%08" PRIxPTR
 #endif
 
-#if defined __aarch64__
+#if defined _M_ARM64 || defined __aarch64__
 #define CPU_ARM64
 #define CPU_64BIT
 #endif

src/funchook_windows.c

@@ -30,6 +30,7 @@
  */
 #define PSAPI_VERSION 1
 #include <stdint.h>
+#include <stdio.h>
 #include <windows.h>
 #include <psapi.h>
 #include "funchook_internal.h"
@@ -87,7 +88,7 @@ static int alloc_page_info(funchook_t *funchook, page_list_t **pl_out, void *hin
 {
     void *addr;
     page_list_t *pl;
-#ifdef CPU_X86_64
+#ifdef CPU_64BIT
     void *old_hint = hint;
     while (1) {
         MEMORY_BASIC_INFORMATION mbi;
@@ -339,7 +340,7 @@ void *funchook_resolve_func(funchook_t *funchook, void *func)
     HMODULE hMod;
     BOOL ok;
     IMAGE_IMPORT_DESCRIPTOR *desc_head, *desc;
-    uint8_t *fn = (uint8_t*)func;
+    insn_t *fn = (insn_t*)func;
     size_t pos = 0;
     DWORD cnt;
 
@@ -349,6 +350,7 @@ void *funchook_resolve_func(funchook_t *funchook, void *func)
             funchook_log(funchook, "  func %p is in %.*s\n", func, (int)len, path);
         }
     }
+#if defined CPU_X86 || defined CPU_X86_64
     if (fn[0] == 0xe9) {
         fn = (fn + 5) + *(int*)(fn + 1);
         funchook_log(funchook, "  relative jump to %p\n", fn);
@@ -361,6 +363,31 @@ void *funchook_resolve_func(funchook_t *funchook, void *func)
 #endif
         funchook_log(funchook, "  indirect jump to addresss at %p\n", (void*)pos);
     }
+#endif
+#if defined CPU_ARM64
+#define ADRP_XIP0       0x90000010
+#define ADRP_XIP0_MASK  0x9F00001F
+#define ADRP_XIP0_IMMLO 0x60000000
+#define ADRP_XIP0_IMMHI 0x00FFFFE0
+#define LDR_XIP0        0xF9400210
+#define LDR_XIP0_MASK   0xFFC003FF
+#define LDR_XIP0_IMM12  0x003FFC00
+#define BR_XIP0         0xD61F0200
+    if ((fn[0] & ADRP_XIP0_MASK) == ADRP_XIP0 &&
+        (fn[1] & LDR_XIP0_MASK) == LDR_XIP0 &&
+        fn[2] == BR_XIP0) {
+        // fn[0]: addrp xip0, immhi&immlo
+        // fn[1]: ldr   xip0, [xip0,imm12]
+        // fn[2]: br    xip0
+        size_t addr = (size_t)fn & ~((1 << 12) - 1);
+        size_t immhi = ((size_t)(fn[0] & ADRP_XIP0_IMMHI) >> 5) << (12 + 2);
+        size_t immlo = ((size_t)(fn[0] & ADRP_XIP0_IMMLO) >> 29) << 12;
+        size_t imm12 = ((size_t)(fn[1] & LDR_XIP0_IMM12) >> 10) << 3;
+        pos = addr + immhi + immlo + imm12;
+        // fprintf(stderr, "%016I64x: %08x %08x %08x : %I64x %I64x %I64x %I64x\n", (size_t)fn, fn[0], fn[1], fn[2], addr, immhi, immlo, imm12);
+        funchook_log(funchook, "  indirect jump to addresss at %p\n", (void*)pos);
+    }
+#endif
     if (pos == 0) {
         return func;
     }

test/CMakeLists.txt

@@ -23,11 +23,14 @@ if (FUNCHOOK_CPU STREQUAL x86)
 endif ()
 
 
-if (MSVC)
+if (NOT MSVC)
+  set(FUNCHOOK_ASM_SUFFIX _gas.S)
+elseif (FUNCHOOK_CPU MATCHES "^x86")
   enable_language(ASM_MASM)
   set(FUNCHOOK_ASM_SUFFIX _masm.asm)
 else ()
-  set(FUNCHOOK_ASM_SUFFIX _gas.S)
+  set(FUNCHOOK_CPU noasm)
+  set(FUNCHOOK_ASM_SUFFIX .c)
 endif ()
 
 add_library(funchook_test SHARED libfunchook_test.c libfunchook_test_${FUNCHOOK_CPU}${FUNCHOOK_ASM_SUFFIX})

test/libfunchook_test_noasm.c

@@ -0,0 +1,19 @@
+#ifdef WIN32
+#define DLLEXPORT __declspec(dllexport)
+#else
+#define DLLEXPORT
+#endif
+
+extern int get_val_in_dll(void);
+
+DLLEXPORT
+int call_get_val_in_dll(void)
+{
+  return get_val_in_dll();
+}
+
+DLLEXPORT
+int jump_get_val_in_dll(void)
+{
+  return get_val_in_dll();
+}