need to track the delta changes in the stack height when performing t…

…he resumes

make

@@ -26,6 +26,7 @@ C_FLAGS = (
 )
 CXX_FLAGS = (
     '-std=c++14 '
+    '-fno-exceptions '
 )
 CXX_FLAGS_UNIT = (
     '-I ./deps/catch/ '
@@ -54,7 +55,7 @@ def release():
     C_FLAGS = C_FLAGS.replace('-O0', '-O2')
     C_FLAGS = C_FLAGS.replace('-ggdb', '')
     C_FLAGS += ' -DNDEBUG -DRED_RELEASE -fdata-sections -ffunction-sections -flto '
-    LD_FLAGS += '-flto ' #-Wl,--gc-sections -Wl,--print-gc-sections '
+    LD_FLAGS += '-flto -O2 ' #-Wl,--gc-sections -Wl,--print-gc-sections '
     clean()
     build()
     Run('mkdir -p release')
@@ -172,12 +173,13 @@ def deps():
     if not os.path.isdir('build'):
         Shell('mkdir -p build')
     if not os.path.isfile('deps/udis86/libudis86/.libs/libudis86.so') or not os.path.isfile('deps/udis86/libudis86/itab.h'):
-          Shell('cd deps/udis86 && ./autogen.sh && PYTHON=`which python2` ./configure && '
-                #"sed -i '/^CFLAGS\ =/ s/$/\ \-mgeneral\-regs\-only/' Makefile &&"
-                'make V=1', shell=True)
+          Shell('cd deps/udis86 && ./autogen.sh && PYTHON=`which python2` ./configure && ' +
+                #("sed -i '/^CFLAGS\ =/ s/$/\ \-flto/' Makefile &&" if RELEASE else '') +
+                'make V=1 CFLAGS=' + ('"-Wall -O2 -flto"' if RELEASE else '"-Wall -ggdb"')
+                , shell=True)
     if not os.path.isfile('build/asmjit/libasmjit.so'):
         Shell('mkdir -p build/asmjit')
-        asm_flags = ''  # -DASMJIT_ALLOC=test123
+        asm_flags = '\-fno-exceptions\ '  # -DASMJIT_ALLOC=test123
         cm_args = '-DASMJIT_DISABLE_COMPILER=1 -DASMJIT_CFLAGS=\'==REPLACE_ME==\' -DCMAKE_CXX_COMPILER=g++ -DCMAKE_C_COMPILER=gcc'
         if RELEASE:
             Shell('cd build/asmjit && cmake ../../deps/asmjit {} -DASMJIT_RELEASE=1'.format(cm_args), shell=True)

src/config.h

@@ -7,7 +7,6 @@
 #endif
 
 #ifdef CONF_RELEASE_BUILD
-# define
 # define CONF_BUILD_TOGGLE(debug, release) release
 #else
 // configure the system to perform more traces to attempt to debug

src/cpp_allocator.h

@@ -2,6 +2,7 @@
 #define REDMAGIC_CPP_ALLOCATOR_H_
 
 #include <memory>
+#include <cstdlib>
 
 extern "C" void *__real_malloc(size_t size);
 extern "C" void __real_free(void *ptr);
@@ -32,8 +33,13 @@ namespace redmagic {
 
     pointer allocate(size_type n, const_pointer hint = 0) {
       void* p = __real_malloc(n * sizeof(T));
-      if (!p)
+      if (!p) {
+#ifdef __EXCEPTIONS
         throw std::bad_alloc();
+#else
+        std::abort();
+#endif
+      }
       return static_cast<pointer>(p);
     }
 

src/jit_internal.h

@@ -136,6 +136,14 @@ namespace redmagic {
     //   >
     RealMallocMap<void*, branch_info> branches;
 
+    struct merge_location_info {
+      RealMallocSet<mem_loc_t> rips;
+#ifdef CONF_MERGE_BACK_ON_RET
+      bool is_method_return = false;
+#endif
+
+    };
+
 #ifdef CONF_CHECK_MERGE_RIP
     // std::unordered_map<
     //   mem_loc_t,
@@ -205,6 +213,10 @@ namespace redmagic {
   struct tracer_merge_block_stack_s {
     mem_loc_t merge_head = 0; // head of linked list for this merge point
 
+#ifdef CONF_MERGE_BACK_ON_RET
+    bool method_merge = false;
+#endif
+
     tracer_merge_block_stack_s() {}
   };
 

src/manager.cc

@@ -995,7 +995,6 @@ namespace {
 }
 
 bool Manager::should_trace_method(void *id) {
-
   if(no_trace_methods.find(id) != no_trace_methods.end())
     return false;
 

src/simple_compiler.cc

@@ -361,8 +361,9 @@ void SimpleCompiler::ResumeBlockJump(mem_loc_t resume_pc) {
   // this will allow for it to easily write in a direct jump, as being designed now, we will have to redirect the jump through this indirection
   // so first conditional jump followed by direct jump
   // also, this will not work with concurrent threads
-  lea(x86::r9, x86::ptr(label));
-  mov(x86::r8, imm_u(0xfbfbfbfbfbfbfbfb));
+  lea(x86::r9, x86::ptr(label)); // patch address
+  mov(x86::r8, imm_u(0xfbfbfbfbfbfbfbfb)); // merge point
+
 
   jmp(imm_ptr(&red_asm_restart_trace));
 

src/tracer.cc

@@ -753,8 +753,9 @@ void Tracer::TempEnableTrace(void *resume_pc) {
   set_pc((mem_loc_t)resume_pc);
   SimpleCompiler compiler(buffer);
   // the "normal" return address will be set to ris when this returns from the temp disabled region
-  compiler.TestRegister((mem_loc_t)&red_asm_jump_rsi, RSI, (register_t)resume_pc, &merge_block_stack.back());
+  auto wb = compiler.TestRegister((mem_loc_t)&red_asm_jump_rsi, RSI, (register_t)resume_pc, &merge_block_stack.back());
   auto written = compiler.finalize();
+  wb.replace_stump<uint64_t>(0xfafafafafafafafa, written.getRawBuffer());
   write_interrupt_block();
 }
 
@@ -1766,6 +1767,7 @@ void Tracer::evaluate_instruction() {
 #ifdef CONF_MERGE_BACK_ON_RET
       merge_block_stack.push_back(tracer_merge_block_stack_s());
       method_stack.back().corresponding_merge_block = merge_block_stack.size();
+      merge_block_stack.back().method_merge = true;
 #endif
     }